bcachefs: bch2_trigger_alloc() handles state changes better

bch2_trigger_alloc() kicks off certain tasks on bucket state changes;
e.g. triggering the bucket discard worker and the invalidate worker.

We've observed the discard worker running too often - most runs it
doesn't do any work, according to the tracepoint - so clearly, we're
kicking it off too often.

This adds an explicit statechange() macro to make these checks more
precise.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This commit is contained in:
Kent Overstreet 2024-02-15 21:42:10 -05:00
parent b63570f747
commit 6e9d0558b1

View File

@ -860,23 +860,28 @@ int bch2_trigger_alloc(struct btree_trans *trans,
*bucket_gen(ca, new.k->p.offset) = new_a->gen;
bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, false);
percpu_up_read(&c->mark_lock);
if (new_a->data_type == BCH_DATA_free &&
(!new_a->journal_seq || new_a->journal_seq < c->journal.flushed_seq_ondisk))
#define eval_state(_a, expr) ({ const struct bch_alloc_v4 *a = _a; expr; })
#define statechange(expr) !eval_state(old_a, expr) && eval_state(new_a, expr)
#define bucket_flushed(a) (!a->journal_seq || a->journal_seq <= c->journal.flushed_seq_ondisk)
if (statechange(a->data_type == BCH_DATA_free &&
bucket_flushed(a)))
closure_wake_up(&c->freelist_wait);
if (new_a->data_type == BCH_DATA_need_discard &&
(!bucket_journal_seq || bucket_journal_seq < c->journal.flushed_seq_ondisk))
if (statechange(a->data_type == BCH_DATA_need_discard &&
bucket_flushed(a)) &&
!bch2_bucket_is_open(c, new.k->p.inode, new.k->p.offset))
bch2_do_discards(c);
if (old_a->data_type != BCH_DATA_cached &&
new_a->data_type == BCH_DATA_cached &&
if (statechange(a->data_type == BCH_DATA_cached) &&
!bch2_bucket_is_open(c, new.k->p.inode, new.k->p.offset) &&
should_invalidate_buckets(ca, bch2_dev_usage_read(ca)))
bch2_do_invalidates(c);
if (new_a->data_type == BCH_DATA_need_gc_gens)
if (statechange(a->data_type == BCH_DATA_need_gc_gens))
bch2_do_gc_gens(c);
percpu_up_read(&c->mark_lock);
}
if ((flags & BTREE_TRIGGER_GC) &&