bcachefs: Disk space accounting rewrite

Main part of the disk accounting rewrite.

This is a wholesale rewrite of the existing disk space accounting, which
relies on percepu counters that are sharded by journal buffer, and
rolled up and added to each journal write.

With the new scheme, every set of counters is a distinct key in the
accounting btree; this fixes scaling limitations of the old scheme,
where counters took up space in each journal entry and required multiple
percpu counters.

Now, in memory accounting requires a single set of percpu counters - not
multiple for each in flight journal buffer - and in the future we'll
probably also have counters that don't use in memory percpu counters,
they're not strictly required.

An accounting update is now a normal btree update, using the btree write
buffer path. At transaction commit time, we apply accounting updates to
the in memory counters, which are percpu counters indexed in an
eytzinger tree by the accounting key.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

1 files changed, 12 insertions, 5 deletions

diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 5c6bfa9e69d5..2d03b4ff31dd 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -139,8 +139,6 @@ static void replay_now_at(struct journal *j, u64 seq)
 static int bch2_journal_replay_accounting_key(struct btree_trans *trans,
 					      struct journal_key *k)
 {
-	struct journal_keys *keys = &trans->c->journal_keys;
-
 	struct btree_iter iter;
 	bch2_trans_node_iter_init(trans, &iter, k->btree_id, k->k->k.p,
 				  BTREE_MAX_DEPTH, k->level,
@@ -282,6 +280,7 @@ int bch2_journal_replay(struct bch_fs *c)
 		ret = commit_do(trans, NULL, NULL,
 				BCH_TRANS_COMMIT_no_enospc|
 				BCH_TRANS_COMMIT_journal_reclaim|
+				BCH_TRANS_COMMIT_skip_accounting_apply|
 				BCH_TRANS_COMMIT_no_journal_res,
 			     bch2_journal_replay_accounting_key(trans, k));
 		if (bch2_fs_fatal_err_on(ret, c, "error replaying accounting; %s", bch2_err_str(ret)))
@@ -312,6 +311,7 @@ int bch2_journal_replay(struct bch_fs *c)
 			commit_do(trans, NULL, NULL,
 				  BCH_TRANS_COMMIT_no_enospc|
 				  BCH_TRANS_COMMIT_journal_reclaim|
+				  BCH_TRANS_COMMIT_skip_accounting_apply|
 				  (!k->allocated ? BCH_TRANS_COMMIT_no_journal_res : 0),
 			     bch2_journal_replay_key(trans, k));
 		BUG_ON(!ret && !k->overwritten && k->k->k.type != KEY_TYPE_accounting);
@@ -342,6 +342,7 @@ int bch2_journal_replay(struct bch_fs *c)
 
 		ret = commit_do(trans, NULL, NULL,
 				BCH_TRANS_COMMIT_no_enospc|
+				BCH_TRANS_COMMIT_skip_accounting_apply|
 				(!k->allocated
 				 ? BCH_TRANS_COMMIT_no_journal_res|BCH_WATERMARK_reclaim
 				 : 0),
@@ -1050,9 +1051,6 @@ int bch2_fs_initialize(struct bch_fs *c)
 	for (unsigned i = 0; i < BTREE_ID_NR; i++)
 		bch2_btree_root_alloc_fake(c, i, 0);
 
-	for_each_member_device(c, ca)
-		bch2_dev_usage_init(ca);
-
 	ret = bch2_fs_journal_alloc(c);
 	if (ret)
 		goto err;
@@ -1069,6 +1067,15 @@ int bch2_fs_initialize(struct bch_fs *c)
 	if (ret)
 		goto err;
 
+	for_each_member_device(c, ca) {
+		ret = bch2_dev_usage_init(ca);
+		bch_err_msg(c, ret, "initializing device usage");
+		if (ret) {
+			bch2_dev_put(ca);
+			goto err;
+		}
+	}
+
 	/*
 	 * Write out the superblock and journal buckets, now that we can do
 	 * btree updates