-#ifndef _BCACHE_JOURNAL_H
-#define _BCACHE_JOURNAL_H
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BCACHEFS_JOURNAL_H
+#define _BCACHEFS_JOURNAL_H
/*
* THE JOURNAL:
*
* Synchronous updates are specified by passing a closure (@flush_cl) to
* bch2_btree_insert() or bch_btree_insert_node(), which then pass that parameter
- * down to the journalling code. That closure will will wait on the journal
- * write to complete (via closure_wait()).
+ * down to the journalling code. That closure will wait on the journal write to
+ * complete (via closure_wait()).
*
* If the index update wasn't synchronous, the journal entry will be
* written out after 10 ms have elapsed, by default (the delay_ms field
#include "journal_types.h"
-/*
- * Only used for holding the journal entries we read in btree_journal_read()
- * during cache_registration
- */
-struct journal_replay {
- struct list_head list;
- struct jset j;
-};
+struct bch_fs;
-static inline struct jset_entry *__jset_entry_type_next(struct jset *jset,
- struct jset_entry *entry, unsigned type)
+static inline void journal_wake(struct journal *j)
{
- while (entry < vstruct_last(jset)) {
- if (JOURNAL_ENTRY_TYPE(entry) == type)
- return entry;
-
- entry = vstruct_next(entry);
- }
-
- return NULL;
+ wake_up(&j->wait);
+ closure_wake_up(&j->async_wait);
}
-#define for_each_jset_entry_type(entry, jset, type) \
- for (entry = (jset)->start; \
- (entry = __jset_entry_type_next(jset, entry, type)); \
- entry = vstruct_next(entry))
-
-#define for_each_jset_key(k, _n, entry, jset) \
- for_each_jset_entry_type(entry, jset, JOURNAL_ENTRY_BTREE_KEYS) \
- vstruct_for_each_safe(entry, k, _n)
+static inline struct journal_buf *journal_cur_buf(struct journal *j)
+{
+ return j->buf + j->reservations.idx;
+}
-#define JOURNAL_PIN (32 * 1024)
+/* Sequence number of oldest dirty journal entry */
-static inline bool journal_pin_active(struct journal_entry_pin *pin)
+static inline u64 journal_last_seq(struct journal *j)
{
- return pin->pin_list != NULL;
+ return j->pin.front;
}
-static inline struct journal_entry_pin_list *
-journal_seq_pin(struct journal *j, u64 seq)
+static inline u64 journal_cur_seq(struct journal *j)
{
- return &j->pin.data[(size_t) seq & j->pin.mask];
+ return atomic64_read(&j->seq);
}
-void bch2_journal_pin_add(struct journal *, struct journal_res *,
- struct journal_entry_pin *, journal_pin_flush_fn);
-void bch2_journal_pin_drop(struct journal *, struct journal_entry_pin *);
-void bch2_journal_pin_add_if_older(struct journal *,
- struct journal_entry_pin *,
- struct journal_entry_pin *,
- journal_pin_flush_fn);
-void bch2_journal_flush_pins(struct journal *, u64);
-
-struct closure;
-struct bch_fs;
-struct keylist;
-
-struct bkey_i *bch2_journal_find_btree_root(struct bch_fs *, struct jset *,
- enum btree_id, unsigned *);
-
-int bch2_journal_seq_should_ignore(struct bch_fs *, u64, struct btree *);
-
-u64 bch2_inode_journal_seq(struct journal *, u64);
+static inline u64 journal_last_unwritten_seq(struct journal *j)
+{
+ return j->seq_ondisk + 1;
+}
static inline int journal_state_count(union journal_res_state s, int idx)
{
- return idx == 0 ? s.buf0_count : s.buf1_count;
+ switch (idx) {
+ case 0: return s.buf0_count;
+ case 1: return s.buf1_count;
+ case 2: return s.buf2_count;
+ case 3: return s.buf3_count;
+ }
+ BUG();
}
static inline void journal_state_inc(union journal_res_state *s)
{
s->buf0_count += s->idx == 0;
s->buf1_count += s->idx == 1;
-}
-
-static inline void bch2_journal_set_has_inode(struct journal_buf *buf, u64 inum)
-{
- set_bit(hash_64(inum, ilog2(sizeof(buf->has_inode) * 8)), buf->has_inode);
+ s->buf2_count += s->idx == 2;
+ s->buf3_count += s->idx == 3;
}
/*
return u64s + sizeof(struct jset_entry) / sizeof(u64);
}
-static inline void bch2_journal_add_entry_at(struct journal_buf *buf,
- const void *data, size_t u64s,
- unsigned type, enum btree_id id,
- unsigned level, unsigned offset)
+static inline int journal_entry_overhead(struct journal *j)
{
- struct jset_entry *entry = vstruct_idx(buf->data, offset);
+ return sizeof(struct jset) / sizeof(u64) + j->entry_u64s_reserved;
+}
+static inline struct jset_entry *
+bch2_journal_add_entry_noreservation(struct journal_buf *buf, size_t u64s)
+{
+ struct jset *jset = buf->data;
+ struct jset_entry *entry = vstruct_idx(jset, le32_to_cpu(jset->u64s));
+
+ memset(entry, 0, sizeof(*entry));
entry->u64s = cpu_to_le16(u64s);
- entry->btree_id = id;
- entry->level = level;
- entry->flags = 0;
- SET_JOURNAL_ENTRY_TYPE(entry, type);
- memcpy_u64s(entry->_data, data, u64s);
+ le32_add_cpu(&jset->u64s, jset_u64s(u64s));
+
+ return entry;
}
-static inline void bch2_journal_add_keys(struct journal *j, struct journal_res *res,
- enum btree_id id, const struct bkey_i *k)
+static inline struct jset_entry *
+journal_res_entry(struct journal *j, struct journal_res *res)
{
- struct journal_buf *buf = &j->buf[res->idx];
- unsigned actual = jset_u64s(k->k.u64s);
+ return vstruct_idx(j->buf[res->idx].data, res->offset);
+}
- EBUG_ON(!res->ref);
- BUG_ON(actual > res->u64s);
+static inline unsigned journal_entry_init(struct jset_entry *entry, unsigned type,
+ enum btree_id id, unsigned level,
+ unsigned u64s)
+{
+ entry->u64s = cpu_to_le16(u64s);
+ entry->btree_id = id;
+ entry->level = level;
+ entry->type = type;
+ entry->pad[0] = 0;
+ entry->pad[1] = 0;
+ entry->pad[2] = 0;
+ return jset_u64s(u64s);
+}
+
+static inline unsigned journal_entry_set(struct jset_entry *entry, unsigned type,
+ enum btree_id id, unsigned level,
+ const void *data, unsigned u64s)
+{
+ unsigned ret = journal_entry_init(entry, type, id, level, u64s);
+
+ memcpy_u64s_small(entry->_data, data, u64s);
+ return ret;
+}
- bch2_journal_set_has_inode(buf, k->k.p.inode);
+static inline struct jset_entry *
+bch2_journal_add_entry(struct journal *j, struct journal_res *res,
+ unsigned type, enum btree_id id,
+ unsigned level, unsigned u64s)
+{
+ struct jset_entry *entry = journal_res_entry(j, res);
+ unsigned actual = journal_entry_init(entry, type, id, level, u64s);
- bch2_journal_add_entry_at(buf, k, k->k.u64s,
- JOURNAL_ENTRY_BTREE_KEYS, id,
- 0, res->offset);
+ EBUG_ON(!res->ref);
+ EBUG_ON(actual > res->u64s);
res->offset += actual;
res->u64s -= actual;
+ return entry;
}
-void bch2_journal_buf_put_slowpath(struct journal *, bool);
+static inline bool journal_entry_empty(struct jset *j)
+{
+ if (j->seq != j->last_seq)
+ return false;
+
+ vstruct_for_each(j, i)
+ if (i->type == BCH_JSET_ENTRY_btree_keys && i->u64s)
+ return false;
+ return true;
+}
-static inline void bch2_journal_buf_put(struct journal *j, unsigned idx,
- bool need_write_just_set)
+/*
+ * Drop reference on a buffer index and return true if the count has hit zero.
+ */
+static inline union journal_res_state journal_state_buf_put(struct journal *j, unsigned idx)
{
union journal_res_state s;
s.v = atomic64_sub_return(((union journal_res_state) {
.buf0_count = idx == 0,
.buf1_count = idx == 1,
+ .buf2_count = idx == 2,
+ .buf3_count = idx == 3,
}).v, &j->reservations.counter);
+ return s;
+}
+
+bool bch2_journal_entry_close(struct journal *);
+void bch2_journal_do_writes(struct journal *);
+void bch2_journal_buf_put_final(struct journal *, u64);
+
+static inline void __bch2_journal_buf_put(struct journal *j, unsigned idx, u64 seq)
+{
+ union journal_res_state s;
+
+ s = journal_state_buf_put(j, idx);
+ if (!journal_state_count(s, idx))
+ bch2_journal_buf_put_final(j, seq);
+}
- EBUG_ON(s.idx != idx && !s.prev_buf_unwritten);
+static inline void bch2_journal_buf_put(struct journal *j, unsigned idx, u64 seq)
+{
+ union journal_res_state s;
- /*
- * Do not initiate a journal write if the journal is in an error state
- * (previous journal entry write may have failed)
- */
- if (s.idx != idx &&
- !journal_state_count(s, idx) &&
- s.cur_entry_offset != JOURNAL_ENTRY_ERROR_VAL)
- bch2_journal_buf_put_slowpath(j, need_write_just_set);
+ s = journal_state_buf_put(j, idx);
+ if (!journal_state_count(s, idx)) {
+ spin_lock(&j->lock);
+ bch2_journal_buf_put_final(j, seq);
+ spin_unlock(&j->lock);
+ }
}
/*
if (!res->ref)
return;
- lock_release(&j->res_map, 0, _RET_IP_);
+ lock_release(&j->res_map, _THIS_IP_);
- while (res->u64s) {
- bch2_journal_add_entry_at(&j->buf[res->idx], NULL, 0,
- JOURNAL_ENTRY_BTREE_KEYS,
- 0, 0, res->offset);
- res->offset += jset_u64s(0);
- res->u64s -= jset_u64s(0);
- }
+ while (res->u64s)
+ bch2_journal_add_entry(j, res,
+ BCH_JSET_ENTRY_btree_keys,
+ 0, 0, 0);
- bch2_journal_buf_put(j, res->idx, false);
+ bch2_journal_buf_put(j, res->idx, res->seq);
res->ref = 0;
}
int bch2_journal_res_get_slowpath(struct journal *, struct journal_res *,
- unsigned, unsigned);
+ unsigned);
+
+/* First bits for BCH_WATERMARK: */
+enum journal_res_flags {
+ __JOURNAL_RES_GET_NONBLOCK = BCH_WATERMARK_BITS,
+ __JOURNAL_RES_GET_CHECK,
+};
+
+#define JOURNAL_RES_GET_NONBLOCK (1 << __JOURNAL_RES_GET_NONBLOCK)
+#define JOURNAL_RES_GET_CHECK (1 << __JOURNAL_RES_GET_CHECK)
static inline int journal_res_get_fast(struct journal *j,
struct journal_res *res,
- unsigned u64s_min,
- unsigned u64s_max)
+ unsigned flags)
{
union journal_res_state old, new;
u64 v = atomic64_read(&j->reservations.counter);
* Check if there is still room in the current journal
* entry:
*/
- if (old.cur_entry_offset + u64s_min > j->cur_entry_u64s)
+ if (new.cur_entry_offset + res->u64s > j->cur_entry_u64s)
return 0;
- res->offset = old.cur_entry_offset;
- res->u64s = min(u64s_max, j->cur_entry_u64s -
- old.cur_entry_offset);
+ EBUG_ON(!journal_state_count(new, new.idx));
+
+ if ((flags & BCH_WATERMARK_MASK) < j->watermark)
+ return 0;
- journal_state_inc(&new);
new.cur_entry_offset += res->u64s;
+ journal_state_inc(&new);
+
+ /*
+ * If the refcount would overflow, we have to wait:
+ * XXX - tracepoint this:
+ */
+ if (!journal_state_count(new, new.idx))
+ return 0;
+
+ if (flags & JOURNAL_RES_GET_CHECK)
+ return 1;
} while ((v = atomic64_cmpxchg(&j->reservations.counter,
old.v, new.v)) != old.v);
- res->ref = true;
- res->idx = new.idx;
- res->seq = le64_to_cpu(j->buf[res->idx].data->seq);
+ res->ref = true;
+ res->idx = old.idx;
+ res->offset = old.cur_entry_offset;
+ res->seq = le64_to_cpu(j->buf[old.idx].data->seq);
return 1;
}
static inline int bch2_journal_res_get(struct journal *j, struct journal_res *res,
- unsigned u64s_min, unsigned u64s_max)
+ unsigned u64s, unsigned flags)
{
int ret;
EBUG_ON(res->ref);
- EBUG_ON(u64s_max < u64s_min);
+ EBUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
+
+ res->u64s = u64s;
- if (journal_res_get_fast(j, res, u64s_min, u64s_max))
+ if (journal_res_get_fast(j, res, flags))
goto out;
- ret = bch2_journal_res_get_slowpath(j, res, u64s_min, u64s_max);
+ ret = bch2_journal_res_get_slowpath(j, res, flags);
if (ret)
return ret;
out:
- lock_acquire_shared(&j->res_map, 0, 0, NULL, _THIS_IP_);
- EBUG_ON(!res->ref);
+ if (!(flags & JOURNAL_RES_GET_CHECK)) {
+ lock_acquire_shared(&j->res_map, 0,
+ (flags & JOURNAL_RES_GET_NONBLOCK) != 0,
+ NULL, _THIS_IP_);
+ EBUG_ON(!res->ref);
+ }
return 0;
}
-void bch2_journal_wait_on_seq(struct journal *, u64, struct closure *);
-void bch2_journal_flush_seq_async(struct journal *, u64, struct closure *);
+/* journal_entry_res: */
+
+void bch2_journal_entry_res_resize(struct journal *,
+ struct journal_entry_res *,
+ unsigned);
+
+int bch2_journal_flush_seq_async(struct journal *, u64, struct closure *);
void bch2_journal_flush_async(struct journal *, struct closure *);
-void bch2_journal_meta_async(struct journal *, struct closure *);
int bch2_journal_flush_seq(struct journal *, u64);
int bch2_journal_flush(struct journal *);
+bool bch2_journal_noflush_seq(struct journal *, u64);
int bch2_journal_meta(struct journal *);
void bch2_journal_halt(struct journal *);
? -EIO : 0;
}
-static inline bool journal_flushes_device(struct bch_dev *ca)
-{
- return true;
-}
-
-void bch2_journal_start(struct bch_fs *);
-int bch2_journal_mark(struct bch_fs *, struct list_head *);
-void bch2_journal_entries_free(struct list_head *);
-int bch2_journal_read(struct bch_fs *, struct list_head *);
-int bch2_journal_replay(struct bch_fs *, struct list_head *);
+struct bch_dev;
static inline void bch2_journal_set_replay_done(struct journal *j)
{
set_bit(JOURNAL_REPLAY_DONE, &j->flags);
}
-ssize_t bch2_journal_print_debug(struct journal *, char *);
-ssize_t bch2_journal_print_pins(struct journal *, char *);
+void bch2_journal_unblock(struct journal *);
+void bch2_journal_block(struct journal *);
+struct journal_buf *bch2_next_write_buffer_flush_journal_buf(struct journal *j, u64 max_seq);
-int bch2_dev_journal_alloc(struct bch_dev *);
+void __bch2_journal_debug_to_text(struct printbuf *, struct journal *);
+void bch2_journal_debug_to_text(struct printbuf *, struct journal *);
+void bch2_journal_pins_to_text(struct printbuf *, struct journal *);
+bool bch2_journal_seq_pins_to_text(struct printbuf *, struct journal *, u64 *);
-static inline unsigned bch2_nr_journal_buckets(struct bch_sb_field_journal *j)
-{
- return j
- ? (__le64 *) vstruct_end(&j->field) - j->buckets
- : 0;
-}
+int bch2_set_nr_journal_buckets(struct bch_fs *, struct bch_dev *,
+ unsigned nr);
+int bch2_dev_journal_alloc(struct bch_dev *);
+int bch2_fs_journal_alloc(struct bch_fs *);
-int bch2_journal_move(struct bch_dev *);
+void bch2_dev_journal_stop(struct journal *, struct bch_dev *);
void bch2_fs_journal_stop(struct journal *);
+int bch2_fs_journal_start(struct journal *, u64);
+
void bch2_dev_journal_exit(struct bch_dev *);
int bch2_dev_journal_init(struct bch_dev *, struct bch_sb *);
void bch2_fs_journal_exit(struct journal *);
int bch2_fs_journal_init(struct journal *);
-#endif /* _BCACHE_JOURNAL_H */
+#endif /* _BCACHEFS_JOURNAL_H */