1 #ifndef _BCACHEFS_JOURNAL_TYPES_H
2 #define _BCACHEFS_JOURNAL_TYPES_H
4 #include <linux/cache.h>
5 #include <linux/workqueue.h>
7 #include "alloc_types.h"
8 #include "super_types.h"
14 * We put two of these in struct journal; we used them for writes to the
15 * journal that are being staged or in flight.
22 struct closure_waitlist wait;
24 unsigned buf_size; /* size in bytes of @data */
25 unsigned sectors; /* maximum size for current entry */
26 unsigned disk_sectors; /* maximum size entry could have been, if
27 buf_size was bigger */
28 unsigned u64s_reserved;
30 unsigned long has_inode[1024 / sizeof(unsigned long)];
34 * Something that makes a journal entry dirty - i.e. a btree node that has to be
38 struct journal_entry_pin_list {
39 struct list_head list;
40 struct list_head flushed;
42 struct bch_devs_list devs;
46 struct journal_entry_pin;
47 typedef void (*journal_pin_flush_fn)(struct journal *j,
48 struct journal_entry_pin *, u64);
50 struct journal_entry_pin {
51 struct list_head list;
52 journal_pin_flush_fn flush;
65 * For reserving space in the journal prior to getting a reservation on a
66 * particular journal entry:
68 struct journal_preres {
72 union journal_res_state {
82 u64 cur_entry_offset:20,
90 union journal_preres_state {
106 #define JOURNAL_ENTRY_SIZE_MIN (64U << 10) /* 64k */
107 #define JOURNAL_ENTRY_SIZE_MAX (4U << 20) /* 4M */
110 * We stash some journal state as sentinal values in cur_entry_offset:
111 * note - cur_entry_offset is in units of u64s
113 #define JOURNAL_ENTRY_OFFSET_MAX ((1U << 20) - 1)
115 #define JOURNAL_ENTRY_CLOSED_VAL (JOURNAL_ENTRY_OFFSET_MAX - 1)
116 #define JOURNAL_ENTRY_ERROR_VAL (JOURNAL_ENTRY_OFFSET_MAX)
119 * JOURNAL_NEED_WRITE - current (pending) journal entry should be written ASAP,
120 * either because something's waiting on the write to complete or because it's
121 * been dirty too long and the timer's expired.
129 JOURNAL_MAY_GET_UNRESERVED,
132 /* Embedded in struct bch_fs */
134 /* Fastpath stuff up front: */
138 union journal_res_state reservations;
140 /* Max size of current journal entry */
141 unsigned cur_entry_u64s;
142 unsigned cur_entry_sectors;
145 * 0, or -ENOSPC if waiting on journal reclaim, or -EROFS if
146 * insufficient devices:
150 union journal_preres_state prereserved;
152 /* Reserved space in journal entry to be used just prior to write */
153 unsigned entry_u64s_reserved;
155 unsigned buf_size_want;
158 * Two journal entries -- one is currently open for new entries, the
159 * other is possibly being written out.
161 struct journal_buf buf[2];
165 /* if nonzero, we may not open a new journal entry: */
168 /* Used when waiting because the journal was full */
169 wait_queue_head_t wait;
170 struct closure_waitlist async_wait;
171 struct closure_waitlist preres_wait;
174 struct delayed_work write_work;
176 /* Sequence number of most recent journal entry (last entry in @pin) */
179 /* seq, last_seq from the most recent journal entry successfully written */
184 * FIFO of journal entries whose btree updates have not yet been
187 * Each entry is a reference count. The position in the FIFO is the
188 * entry's sequence number relative to @seq.
190 * The journal entry itself holds a reference count, put when the
191 * journal entry is written out. Each btree node modified by the journal
192 * entry also holds a reference count, put when the btree node is
195 * When a reference count reaches zero, the journal entry is no longer
196 * needed. When all journal entries in the oldest journal bucket are no
197 * longer needed, the bucket can be discarded and reused.
200 u64 front, back, size, mask;
201 struct journal_entry_pin_list *data;
204 u64 replay_journal_seq;
205 u64 replay_journal_seq_end;
207 struct write_point wp;
210 struct delayed_work reclaim_work;
211 struct mutex reclaim_lock;
212 unsigned long last_flushed;
213 struct journal_entry_pin *flush_in_progress;
214 wait_queue_head_t pin_flush_wait;
216 /* protects advancing ja->discard_idx: */
217 struct mutex discard_lock;
220 unsigned write_delay_ms;
221 unsigned reclaim_delay_ms;
223 u64 res_get_blocked_start;
225 u64 write_start_time;
227 struct time_stats *write_time;
228 struct time_stats *delay_time;
229 struct time_stats *blocked_time;
230 struct time_stats *flush_seq_time;
232 #ifdef CONFIG_DEBUG_LOCK_ALLOC
233 struct lockdep_map res_map;
238 * Embedded in struct bch_dev. First three fields refer to the array of journal
239 * buckets, in bch_sb.
241 struct journal_device {
243 * For each journal bucket, contains the max sequence number of the
244 * journal writes it contains - so we know when a bucket can be reused.
248 unsigned sectors_free;
251 * discard_idx <= dirty_idx_ondisk <= dirty_idx <= cur_idx:
253 unsigned discard_idx; /* Next bucket to discard */
254 unsigned dirty_idx_ondisk;
256 unsigned cur_idx; /* Journal bucket we're currently writing to */
261 /* Bio for journal reads/writes to this device */
264 /* for bch_journal_read_device */
269 * journal_entry_res - reserve space in every journal entry:
271 struct journal_entry_res {
275 #endif /* _BCACHEFS_JOURNAL_TYPES_H */