#include "bcachefs.h"
#include "btree_key_cache.h"
+#include "btree_update.h"
+#include "buckets.h"
#include "errcode.h"
#include "error.h"
#include "journal.h"
#include "journal_io.h"
#include "journal_reclaim.h"
#include "replicas.h"
-#include "super.h"
+#include "sb-members.h"
+#include "trace.h"
#include <linux/kthread.h>
#include <linux/sched/mm.h>
-#include <trace/events/bcachefs.h>
/* Free space calculations: */
clean = j->space[journal_space_clean].total;
total = j->space[journal_space_total].total;
- if (!clean_ondisk &&
- journal_cur_seq(j) == j->seq_ondisk) {
- struct printbuf buf = PRINTBUF;
-
- __bch2_journal_debug_to_text(&buf, j);
- bch_err(c, "journal stuck\n%s", buf.buf);
- printbuf_exit(&buf);
-
- /*
- * Hack: bch2_fatal_error() calls bch2_journal_halt() which
- * takes journal lock:
- */
- spin_unlock(&j->lock);
- bch2_fatal_error(c);
- spin_lock(&j->lock);
-
- ret = JOURNAL_ERR_journal_stuck;
- } else if (!j->space[journal_space_discarded].next_entry)
+ if (!j->space[journal_space_discarded].next_entry)
ret = JOURNAL_ERR_journal_full;
if ((j->space[journal_space_clean_ondisk].next_entry <
j->space[journal_space_clean_ondisk].total) &&
(clean - clean_ondisk <= total / 8) &&
- (clean_ondisk * 2 > clean ))
+ (clean_ondisk * 2 > clean))
set_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
else
clear_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
blkdev_issue_discard(ca->disk_sb.bdev,
bucket_to_sector(ca,
ja->buckets[ja->discard_idx]),
- ca->mi.bucket_size, GFP_NOIO);
+ ca->mi.bucket_size, GFP_NOFS);
spin_lock(&j->lock);
ja->discard_idx = (ja->discard_idx + 1) % ja->nr;
*/
while (!fifo_empty(&j->pin) &&
!atomic_read(&fifo_peek_front(&j->pin).count)) {
- BUG_ON(!list_empty(&fifo_peek_front(&j->pin).list));
- BUG_ON(!list_empty(&fifo_peek_front(&j->pin).flushed));
- BUG_ON(!fifo_pop(&j->pin, temp));
+ fifo_pop(&j->pin, temp);
popped = true;
}
}
}
-static inline void __journal_pin_drop(struct journal *j,
+static inline bool __journal_pin_drop(struct journal *j,
struct journal_entry_pin *pin)
{
struct journal_entry_pin_list *pin_list;
if (!journal_pin_active(pin))
- return;
+ return false;
if (j->flush_in_progress == pin)
j->flush_in_progress_dropped = true;
list_del_init(&pin->list);
/*
- * Unpinning a journal entry make make journal_next_bucket() succeed, if
+ * Unpinning a journal entry may make journal_next_bucket() succeed, if
* writing a new last_seq will now make another bucket available:
*/
- if (atomic_dec_and_test(&pin_list->count) &&
- pin_list == &fifo_peek_front(&j->pin))
- bch2_journal_reclaim_fast(j);
+ return atomic_dec_and_test(&pin_list->count) &&
+ pin_list == &fifo_peek_front(&j->pin);
}
void bch2_journal_pin_drop(struct journal *j,
struct journal_entry_pin *pin)
{
spin_lock(&j->lock);
- __journal_pin_drop(j, pin);
+ if (__journal_pin_drop(j, pin))
+ bch2_journal_reclaim_fast(j);
spin_unlock(&j->lock);
}
+static enum journal_pin_type journal_pin_type(journal_pin_flush_fn fn)
+{
+ if (fn == bch2_btree_node_flush0 ||
+ fn == bch2_btree_node_flush1)
+ return JOURNAL_PIN_btree;
+ else if (fn == bch2_btree_key_cache_journal_flush)
+ return JOURNAL_PIN_key_cache;
+ else
+ return JOURNAL_PIN_other;
+}
+
void bch2_journal_pin_set(struct journal *j, u64 seq,
struct journal_entry_pin *pin,
journal_pin_flush_fn flush_fn)
{
struct journal_entry_pin_list *pin_list;
+ bool reclaim;
spin_lock(&j->lock);
pin_list = journal_seq_pin(j, seq);
- __journal_pin_drop(j, pin);
+ reclaim = __journal_pin_drop(j, pin);
atomic_inc(&pin_list->count);
pin->seq = seq;
pin->flush = flush_fn;
- if (flush_fn == bch2_btree_key_cache_journal_flush)
- list_add(&pin->list, &pin_list->key_cache_list);
- else if (flush_fn)
- list_add(&pin->list, &pin_list->list);
+ if (flush_fn)
+ list_add(&pin->list, &pin_list->list[journal_pin_type(flush_fn)]);
else
list_add(&pin->list, &pin_list->flushed);
+
+ if (reclaim)
+ bch2_journal_reclaim_fast(j);
spin_unlock(&j->lock);
/*
static struct journal_entry_pin *
journal_get_next_pin(struct journal *j,
- bool get_any,
- bool get_key_cache,
- u64 max_seq, u64 *seq)
+ u64 seq_to_flush,
+ unsigned allowed_below_seq,
+ unsigned allowed_above_seq,
+ u64 *seq)
{
struct journal_entry_pin_list *pin_list;
struct journal_entry_pin *ret = NULL;
+ unsigned i;
fifo_for_each_entry_ptr(pin_list, &j->pin, *seq) {
- if (*seq > max_seq && !get_any && !get_key_cache)
+ if (*seq > seq_to_flush && !allowed_above_seq)
break;
- if (*seq <= max_seq || get_any) {
- ret = list_first_entry_or_null(&pin_list->list,
- struct journal_entry_pin, list);
- if (ret)
- return ret;
- }
-
- if (*seq <= max_seq || get_any || get_key_cache) {
- ret = list_first_entry_or_null(&pin_list->key_cache_list,
- struct journal_entry_pin, list);
- if (ret)
- return ret;
- }
+ for (i = 0; i < JOURNAL_PIN_NR; i++)
+ if ((((1U << i) & allowed_below_seq) && *seq <= seq_to_flush) ||
+ ((1U << i) & allowed_above_seq)) {
+ ret = list_first_entry_or_null(&pin_list->list[i],
+ struct journal_entry_pin, list);
+ if (ret)
+ return ret;
+ }
}
return NULL;
}
/* returns true if we did work */
-static size_t journal_flush_pins(struct journal *j, u64 seq_to_flush,
+static size_t journal_flush_pins(struct journal *j,
+ u64 seq_to_flush,
+ unsigned allowed_below_seq,
+ unsigned allowed_above_seq,
unsigned min_any,
unsigned min_key_cache)
{
lockdep_assert_held(&j->reclaim_lock);
while (1) {
+ unsigned allowed_above = allowed_above_seq;
+ unsigned allowed_below = allowed_below_seq;
+
+ if (min_any) {
+ allowed_above |= ~0;
+ allowed_below |= ~0;
+ }
+
+ if (min_key_cache) {
+ allowed_above |= 1U << JOURNAL_PIN_key_cache;
+ allowed_below |= 1U << JOURNAL_PIN_key_cache;
+ }
+
cond_resched();
j->last_flushed = jiffies;
spin_lock(&j->lock);
- pin = journal_get_next_pin(j,
- min_any != 0,
- min_key_cache != 0,
- seq_to_flush, &seq);
+ pin = journal_get_next_pin(j, seq_to_flush, allowed_below, allowed_above, &seq);
if (pin) {
BUG_ON(j->flush_in_progress);
j->flush_in_progress = pin;
min_key_cache = min(bch2_nr_btree_keys_need_flush(c), (size_t) 128);
- trace_journal_reclaim_start(c, direct, kicked,
+ trace_and_count(c, journal_reclaim_start, c,
+ direct, kicked,
min_nr, min_key_cache,
j->prereserved.reserved,
j->prereserved.remaining,
atomic_long_read(&c->btree_key_cache.nr_keys));
nr_flushed = journal_flush_pins(j, seq_to_flush,
+ ~0, 0,
min_nr, min_key_cache);
if (direct)
j->nr_direct_reclaim += nr_flushed;
else
j->nr_background_reclaim += nr_flushed;
- trace_journal_reclaim_finish(c, nr_flushed);
+ trace_and_count(c, journal_reclaim_finish, c, nr_flushed);
if (nr_flushed)
wake_up(&j->reclaim_wait);
j->next_reclaim = now + delay;
while (1) {
- set_current_state(TASK_INTERRUPTIBLE);
+ set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
if (kthread_should_stop())
break;
if (j->reclaim_kicked)
spin_unlock(&j->lock);
if (journal_empty)
- freezable_schedule();
+ schedule();
else if (time_after(j->next_reclaim, jiffies))
- freezable_schedule_timeout(j->next_reclaim - jiffies);
+ schedule_timeout(j->next_reclaim - jiffies);
else
break;
}
mutex_lock(&j->reclaim_lock);
- if (journal_flush_pins(j, seq_to_flush, 0, 0))
+ if (journal_flush_pins(j, seq_to_flush,
+ (1U << JOURNAL_PIN_key_cache)|
+ (1U << JOURNAL_PIN_other), 0, 0, 0) ||
+ journal_flush_pins(j, seq_to_flush,
+ (1U << JOURNAL_PIN_btree), 0, 0, 0))
*did_work = true;
spin_lock(&j->lock);
mutex_lock(&c->replicas_gc_lock);
bch2_replicas_gc_start(c, 1 << BCH_DATA_journal);
- seq = 0;
+ /*
+ * Now that we've populated replicas_gc, write to the journal to mark
+ * active journal devices. This handles the case where the journal might
+ * be empty. Otherwise we could clear all journal replicas and
+ * temporarily put the fs into an unrecoverable state. Journal recovery
+ * expects to find devices marked for journal data on unclean mount.
+ */
+ ret = bch2_journal_meta(&c->journal);
+ if (ret)
+ goto err;
+ seq = 0;
spin_lock(&j->lock);
while (!ret) {
struct bch_replicas_padded replicas;
spin_lock(&j->lock);
}
spin_unlock(&j->lock);
-
+err:
ret = bch2_replicas_gc_end(c, ret);
mutex_unlock(&c->replicas_gc_lock);