diff options
Diffstat (limited to 'fs/bcachefs/journal_reclaim.c')
| -rw-r--r-- | fs/bcachefs/journal_reclaim.c | 29 |
1 files changed, 28 insertions, 1 deletions
diff --git a/fs/bcachefs/journal_reclaim.c b/fs/bcachefs/journal_reclaim.c index be50455c7f13..f23e5ee9ad75 100644 --- a/fs/bcachefs/journal_reclaim.c +++ b/fs/bcachefs/journal_reclaim.c @@ -874,7 +874,34 @@ static int journal_flush_done(struct journal *j, u64 seq_to_flush, --type) if (journal_flush_pins_or_still_flushing(j, seq_to_flush, BIT(type))) { *did_work = true; - return ret; + + /* + * Question from Dan Carpenter, on the early return: + * + * If journal_flush_pins_or_still_flushing() returns + * true, then the flush hasn't complete and we must + * return 0; we want the outer closure_wait_event() in + * journal_flush_pins() to continue. + * + * The early return is there because we don't want to + * call journal_entry_close() until we've finished + * flushing all outstanding journal pins - otherwise + * seq_to_flush can be U64_MAX, and we'll close a bunch + * of journal entries and write tiny ones completely + * unnecessarily. + * + * Having the early return be in the loop where we loop + * over types is important, because flushing one journal + * pin can cause new journal pins to be added (even of + * the same type, btree node writes may generate more + * btree node writes, when updating the parent pointer + * has a full node and has to trigger a split/compact). + * + * This is part of our shutdown sequence, where order of + * flushing is important in order to make sure that it + * terminates... + */ + return 0; } if (seq_to_flush > journal_cur_seq(j)) |