1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
|
/* SPDX-License-Identifier: GPL-2.0 */
#ifndef _BCACHEFS_JOURNAL_TYPES_H
#define _BCACHEFS_JOURNAL_TYPES_H
#include <linux/cache.h>
#include <linux/workqueue.h>
#include "alloc_types.h"
#include "super_types.h"
#include "fifo.h"
struct journal_res;
/*
* We put two of these in struct journal; we used them for writes to the
* journal that are being staged or in flight.
*/
struct journal_buf {
struct jset *data;
BKEY_PADDED(key);
struct closure_waitlist wait;
unsigned size;
unsigned disk_sectors;
/* bloom filter: */
unsigned long has_inode[1024 / sizeof(unsigned long)];
};
/*
* Something that makes a journal entry dirty - i.e. a btree node that has to be
* flushed:
*/
struct journal_entry_pin_list {
struct list_head list;
struct list_head flushed;
atomic_t count;
struct bch_devs_list devs;
};
struct journal;
struct journal_entry_pin;
typedef void (*journal_pin_flush_fn)(struct journal *j,
struct journal_entry_pin *, u64);
struct journal_entry_pin {
struct list_head list;
journal_pin_flush_fn flush;
u64 seq;
};
/* corresponds to a btree node with a blacklisted bset: */
struct blacklisted_node {
__le64 seq;
enum btree_id btree_id;
struct bpos pos;
};
struct journal_seq_blacklist {
struct list_head list;
u64 start;
u64 end;
struct journal_entry_pin pin;
struct blacklisted_node *entries;
size_t nr_entries;
};
struct journal_res {
bool ref;
u8 idx;
u16 u64s;
u32 offset;
u64 seq;
};
union journal_res_state {
struct {
atomic64_t counter;
};
struct {
u64 v;
};
struct {
u64 cur_entry_offset:20,
idx:1,
prev_buf_unwritten:1,
buf0_count:21,
buf1_count:21;
};
};
/* bytes: */
#define JOURNAL_ENTRY_SIZE_MIN (64U << 10) /* 64k */
#define JOURNAL_ENTRY_SIZE_MAX (4U << 20) /* 4M */
/*
* We stash some journal state as sentinal values in cur_entry_offset:
* note - cur_entry_offset is in units of u64s
*/
#define JOURNAL_ENTRY_OFFSET_MAX ((1U << 20) - 1)
#define JOURNAL_ENTRY_CLOSED_VAL (JOURNAL_ENTRY_OFFSET_MAX - 1)
#define JOURNAL_ENTRY_ERROR_VAL (JOURNAL_ENTRY_OFFSET_MAX)
/*
* JOURNAL_NEED_WRITE - current (pending) journal entry should be written ASAP,
* either because something's waiting on the write to complete or because it's
* been dirty too long and the timer's expired.
*/
enum {
JOURNAL_REPLAY_DONE,
JOURNAL_STARTED,
JOURNAL_NEED_WRITE,
JOURNAL_NOT_EMPTY,
};
/* Embedded in struct bch_fs */
struct journal {
/* Fastpath stuff up front: */
unsigned long flags;
union journal_res_state reservations;
unsigned cur_entry_u64s;
unsigned prev_buf_sectors;
unsigned cur_buf_sectors;
unsigned buf_size_want;
/*
* Two journal entries -- one is currently open for new entries, the
* other is possibly being written out.
*/
struct journal_buf buf[2];
spinlock_t lock;
/* Used when waiting because the journal was full */
wait_queue_head_t wait;
struct closure_waitlist async_wait;
struct closure io;
struct delayed_work write_work;
/* Sequence number of most recent journal entry (last entry in @pin) */
atomic64_t seq;
/* seq, last_seq from the most recent journal entry successfully written */
u64 seq_ondisk;
u64 last_seq_ondisk;
/*
* FIFO of journal entries whose btree updates have not yet been
* written out.
*
* Each entry is a reference count. The position in the FIFO is the
* entry's sequence number relative to @seq.
*
* The journal entry itself holds a reference count, put when the
* journal entry is written out. Each btree node modified by the journal
* entry also holds a reference count, put when the btree node is
* written.
*
* When a reference count reaches zero, the journal entry is no longer
* needed. When all journal entries in the oldest journal bucket are no
* longer needed, the bucket can be discarded and reused.
*/
struct {
u64 front, back, size, mask;
struct journal_entry_pin_list *data;
} pin;
struct journal_entry_pin *flush_in_progress;
wait_queue_head_t pin_flush_wait;
u64 replay_journal_seq;
struct mutex blacklist_lock;
struct list_head seq_blacklist;
struct journal_seq_blacklist *new_blacklist;
BKEY_PADDED(key);
struct write_point wp;
spinlock_t err_lock;
struct delayed_work reclaim_work;
unsigned long last_flushed;
/* protects advancing ja->last_idx: */
struct mutex reclaim_lock;
unsigned write_delay_ms;
unsigned reclaim_delay_ms;
u64 res_get_blocked_start;
u64 need_write_time;
u64 write_start_time;
struct bch2_time_stats *write_time;
struct bch2_time_stats *delay_time;
struct bch2_time_stats *blocked_time;
struct bch2_time_stats *flush_seq_time;
#ifdef CONFIG_DEBUG_LOCK_ALLOC
struct lockdep_map res_map;
#endif
};
/*
* Embedded in struct bch_dev. First three fields refer to the array of journal
* buckets, in bch_sb.
*/
struct journal_device {
/*
* For each journal bucket, contains the max sequence number of the
* journal writes it contains - so we know when a bucket can be reused.
*/
u64 *bucket_seq;
unsigned sectors_free;
/* Journal bucket we're currently writing to */
unsigned cur_idx;
/* Last journal bucket that still contains an open journal entry */
/*
* j->lock and j->reclaim_lock must both be held to modify, j->lock
* sufficient to read:
*/
unsigned last_idx;
unsigned nr;
u64 *buckets;
/* Bio for journal reads/writes to this device */
struct bio *bio;
/* for bch_journal_read_device */
struct closure read;
};
#endif /* _BCACHEFS_JOURNAL_TYPES_H */
|