bcachefs: Allow for unknown btree IDs

We need to allow filesystems with metadata from newer versions to be
mountable and usable by older versions.

This patch enables us to roll out new btrees without a new major version
number; we can now handle btree roots for unknown btree types.

The unknown btree roots will be retained, and fsck (including
backpointers) will check them, the same as other btree types.

We add a dynamic array for the extra, unknown btree roots, in addition
to the fixed size btree root array, and add new helpers for looking up
btree roots.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/backpointers.c

@@ -272,6 +272,7 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
 					 unsigned iter_flags)
 {
 	struct bch_fs *c = trans->c;
+	struct btree_root *r = bch2_btree_id_root(c, bp.btree_id);
 	struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
 	struct bkey_s_c k;
 
@@ -279,7 +280,7 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
 				  bp.btree_id,
 				  bp.pos,
 				  0,
-				  min(bp.level, c->btree_roots[bp.btree_id].level),
+				  min(bp.level, r->level),
 				  iter_flags);
 	k = bch2_btree_iter_peek_slot(iter);
 	if (bkey_err(k)) {
@@ -287,8 +288,8 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
 		return k;
 	}
 
-	if (bp.level == c->btree_roots[bp.btree_id].level + 1)
-		k = bkey_i_to_s_c(&c->btree_roots[bp.btree_id].key);
+	if (bp.level == r->level + 1)
+		k = bkey_i_to_s_c(&r->key);
 
 	if (k.k && extent_matches_bp(c, bp.btree_id, bp.level, k, bucket, bp))
 		return k;
@@ -531,6 +532,7 @@ static int check_btree_root_to_backpointers(struct btree_trans *trans,
 					    struct bpos_level *last_flushed)
 {
 	struct bch_fs *c = trans->c;
+	struct btree_root *r = bch2_btree_id_root(c, btree_id);
 	struct btree_iter iter;
 	struct btree *b;
 	struct bkey_s_c k;
@@ -539,8 +541,7 @@ static int check_btree_root_to_backpointers(struct btree_trans *trans,
 	const union bch_extent_entry *entry;
 	int ret;
 
-	bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0,
-				  c->btree_roots[btree_id].level, 0);
+	bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0, r->level, 0);
 	b = bch2_btree_iter_peek_node(&iter);
 	ret = PTR_ERR_OR_ZERO(b);
 	if (ret)
@@ -640,12 +641,13 @@ static int bch2_check_extents_to_backpointers_pass(struct btree_trans *trans,
 						   struct bpos bucket_start,
 						   struct bpos bucket_end)
 {
+	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	enum btree_id btree_id;
 	struct bpos_level last_flushed = { UINT_MAX };
 	int ret = 0;
 
-	for (btree_id = 0; btree_id < BTREE_ID_NR; btree_id++) {
+	for (btree_id = 0; btree_id < btree_id_nr_alive(c); btree_id++) {
 		unsigned depth = btree_type_has_ptrs(btree_id) ? 0 : 1;
 
 		bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0,

fs/bcachefs/bcachefs.h

@@ -749,7 +749,8 @@ struct bch_fs {
 	struct bio_set		btree_bio;
 	struct workqueue_struct	*io_complete_wq;
 
-	struct btree_root	btree_roots[BTREE_ID_NR];
+	struct btree_root	btree_roots_known[BTREE_ID_NR];
+	DARRAY(struct btree_root) btree_roots_extra;
 	struct mutex		btree_root_lock;
 
 	struct btree_cache	btree_cache;

fs/bcachefs/btree_cache.c

@@ -25,13 +25,15 @@ void bch2_recalc_btree_reserve(struct bch_fs *c)
 {
 	unsigned i, reserve = 16;
 
-	if (!c->btree_roots[0].b)
+	if (!c->btree_roots_known[0].b)
 		reserve += 8;
 
-	for (i = 0; i < BTREE_ID_NR; i++)
-		if (c->btree_roots[i].b)
-			reserve += min_t(unsigned, 1,
-					 c->btree_roots[i].b->c.level) * 8;
+	for (i = 0; i < btree_id_nr_alive(c); i++) {
+		struct btree_root *r = bch2_btree_id_root(c, i);
+
+		if (r->b)
+			reserve += min_t(unsigned, 1, r->b->c.level) * 8;
+	}
 
 	c->btree_cache.reserve = reserve;
 }
@@ -409,9 +411,12 @@ void bch2_fs_btree_cache_exit(struct bch_fs *c)
 
 	kvpfree(c->verify_ondisk, btree_bytes(c));
 
-	for (i = 0; i < BTREE_ID_NR; i++)
-		if (c->btree_roots[i].b)
-			list_add(&c->btree_roots[i].b->list, &bc->live);
+	for (i = 0; i < btree_id_nr_alive(c); i++) {
+		struct btree_root *r = bch2_btree_id_root(c, i);
+
+		if (r->b)
+			list_add(&r->b->list, &bc->live);
+	}
 
 	list_splice(&bc->freeable, &bc->live);
 

fs/bcachefs/btree_cache.h

@@ -97,7 +97,27 @@ static inline unsigned btree_blocks(struct bch_fs *c)
 	(BTREE_FOREGROUND_MERGE_THRESHOLD(c) +			\
 	 (BTREE_FOREGROUND_MERGE_THRESHOLD(c) >> 2))
 
-#define btree_node_root(_c, _b)	((_c)->btree_roots[(_b)->c.btree_id].b)
+static inline unsigned btree_id_nr_alive(struct bch_fs *c)
+{
+	return BTREE_ID_NR + c->btree_roots_extra.nr;
+}
+
+static inline struct btree_root *bch2_btree_id_root(struct bch_fs *c, unsigned id)
+{
+	if (likely(id < BTREE_ID_NR)) {
+		return &c->btree_roots_known[id];
+	} else {
+		unsigned idx = id - BTREE_ID_NR;
+
+		EBUG_ON(idx >= c->btree_roots_extra.nr);
+		return &c->btree_roots_extra.data[idx];
+	}
+}
+
+static inline struct btree *btree_node_root(struct bch_fs *c, struct btree *b)
+{
+	return bch2_btree_id_root(c, b->c.btree_id)->b;
+}
 
 void bch2_btree_node_to_text(struct printbuf *, struct bch_fs *,
 			     const struct btree *);

fs/bcachefs/btree_gc.c

@@ -529,8 +529,13 @@ static int bch2_repair_topology(struct bch_fs *c)
 
 	bch2_trans_init(&trans, c, 0, 0);
 
-	for (i = 0; i < BTREE_ID_NR && !ret; i++) {
-		b = c->btree_roots[i].b;
+	for (i = 0; i < btree_id_nr_alive(c)&& !ret; i++) {
+		struct btree_root *r = bch2_btree_id_root(c, i);
+
+		if (!r->alive)
+			continue;
+
+		b = r->b;
 		if (btree_node_fake(b))
 			continue;
 
@@ -883,7 +888,7 @@ static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id,
 		return ret;
 
 	mutex_lock(&c->btree_root_lock);
-	b = c->btree_roots[btree_id].b;
+	b = bch2_btree_id_root(c, btree_id)->b;
 	if (!btree_node_fake(b)) {
 		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
 
@@ -1006,7 +1011,7 @@ static int bch2_gc_btree_init(struct btree_trans *trans,
 	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
-	b = c->btree_roots[btree_id].b;
+	b = bch2_btree_id_root(c, btree_id)->b;
 
 	if (btree_node_fake(b))
 		return 0;
@@ -1072,6 +1077,15 @@ static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only)
 			? bch2_gc_btree_init(&trans, ids[i], metadata_only)
 			: bch2_gc_btree(&trans, ids[i], initial, metadata_only);
 
+	for (i = BTREE_ID_NR; i < btree_id_nr_alive(c) && !ret; i++) {
+		if (!bch2_btree_id_root(c, i)->alive)
+			continue;
+
+		ret = initial
+			? bch2_gc_btree_init(&trans, i, metadata_only)
+			: bch2_gc_btree(&trans, i, initial, metadata_only);
+	}
+
 	if (ret < 0)
 		bch_err_fn(c, ret);
 

fs/bcachefs/btree_io.c

@@ -517,7 +517,7 @@ static void btree_pos_to_text(struct printbuf *out, struct bch_fs *c,
 	prt_printf(out, "%s level %u/%u\n  ",
 	       bch2_btree_ids[b->c.btree_id],
 	       b->c.level,
-	       c->btree_roots[b->c.btree_id].level);
+	       bch2_btree_id_root(c, b->c.btree_id)->level);
 	bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&b->key));
 }
 

fs/bcachefs/btree_iter.c

@@ -238,7 +238,7 @@ static void bch2_btree_path_verify(struct btree_trans *trans,
 	for (i = 0; i < (!path->cached ? BTREE_MAX_DEPTH : 1); i++) {
 		if (!path->l[i].b) {
 			BUG_ON(!path->cached &&
-			       c->btree_roots[path->btree_id].b->c.level > i);
+			       bch2_btree_id_root(c, path->btree_id)->b->c.level > i);
 			break;
 		}
 
@@ -732,7 +732,7 @@ static inline int btree_path_lock_root(struct btree_trans *trans,
 				       unsigned long trace_ip)
 {
 	struct bch_fs *c = trans->c;
-	struct btree *b, **rootp = &c->btree_roots[path->btree_id].b;
+	struct btree *b, **rootp = &bch2_btree_id_root(c, path->btree_id)->b;
 	enum six_lock_type lock_type;
 	unsigned i;
 	int ret;

fs/bcachefs/btree_update_interior.c

@@ -1199,7 +1199,7 @@ static void bch2_btree_set_root_inmem(struct bch_fs *c, struct btree *b)
 	       (b->c.level < btree_node_root(c, b)->c.level ||
 		!btree_node_dying(btree_node_root(c, b))));
 
-	btree_node_root(c, b) = b;
+	bch2_btree_id_root(c, b->c.btree_id)->b = b;
 	mutex_unlock(&c->btree_root_lock);
 
 	bch2_recalc_btree_reserve(c);
@@ -2402,7 +2402,7 @@ bool bch2_btree_interior_updates_flush(struct bch_fs *c)
 
 void bch2_journal_entry_to_btree_root(struct bch_fs *c, struct jset_entry *entry)
 {
-	struct btree_root *r = &c->btree_roots[entry->btree_id];
+	struct btree_root *r = bch2_btree_id_root(c, entry->btree_id);
 
 	mutex_lock(&c->btree_root_lock);
 
@@ -2428,15 +2428,15 @@ bch2_btree_roots_to_journal_entries(struct bch_fs *c,
 
 	mutex_lock(&c->btree_root_lock);
 
-	for (i = 0; i < BTREE_ID_NR; i++)
-		if (c->btree_roots[i].alive && !test_bit(i, &have)) {
-			journal_entry_set(end,
-					  BCH_JSET_ENTRY_btree_root,
-					  i, c->btree_roots[i].level,
-					  &c->btree_roots[i].key,
-					  c->btree_roots[i].key.k.u64s);
+	for (i = 0; i < btree_id_nr_alive(c); i++) {
+		struct btree_root *r = bch2_btree_id_root(c, i);
+
+		if (r->alive && !test_bit(i, &have)) {
+			journal_entry_set(end, BCH_JSET_ENTRY_btree_root,
+					  i, r->level, &r->key, r->key.k.u64s);
 			end = vstruct_next(end);
 		}
+	}
 
 	mutex_unlock(&c->btree_root_lock);
 

fs/bcachefs/move.c

@@ -632,7 +632,7 @@ int bch2_move_data(struct bch_fs *c,
 	bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
 
 	for (id = start_btree_id;
-	     id <= min_t(unsigned, end_btree_id, BTREE_ID_NR - 1);
+	     id <= min_t(unsigned, end_btree_id, btree_id_nr_alive(c) - 1);
 	     id++) {
 		stats->btree_id = id;
 
@@ -640,6 +640,9 @@ int bch2_move_data(struct bch_fs *c,
 		    id != BTREE_ID_reflink)
 			continue;
 
+		if (!bch2_btree_id_root(c, id)->b)
+			continue;
+
 		ret = __bch2_move_data(&ctxt,
 				       id == start_btree_id ? start_pos : POS_MIN,
 				       id == end_btree_id   ? end_pos   : POS_MAX,
@@ -861,10 +864,13 @@ static int bch2_move_btree(struct bch_fs *c,
 	stats->data_type = BCH_DATA_btree;
 
 	for (id = start_btree_id;
-	     id <= min_t(unsigned, end_btree_id, BTREE_ID_NR - 1);
+	     id <= min_t(unsigned, end_btree_id, btree_id_nr_alive(c) - 1);
 	     id++) {
 		stats->btree_id = id;
 
+		if (!bch2_btree_id_root(c, id)->b)
+			continue;
+
 		bch2_trans_node_iter_init(&trans, &iter, id, POS_MIN, 0, 0,
 					  BTREE_ITER_PREFETCH);
 retry:

fs/bcachefs/recovery.c

@@ -702,13 +702,13 @@ static int journal_replay_entry_early(struct bch_fs *c,
 	case BCH_JSET_ENTRY_btree_root: {
 		struct btree_root *r;
 
-		if (entry->btree_id >= BTREE_ID_NR) {
-			bch_err(c, "filesystem has unknown btree type %u",
-				entry->btree_id);
-			return -EINVAL;
+		while (entry->btree_id >= c->btree_roots_extra.nr + BTREE_ID_NR) {
+			ret = darray_push(&c->btree_roots_extra, (struct btree_root) { NULL });
+			if (ret)
+				return ret;
 		}
 
-		r = &c->btree_roots[entry->btree_id];
+		r = bch2_btree_id_root(c, entry->btree_id);
 
 		if (entry->u64s) {
 			r->level = entry->level;
@@ -980,8 +980,8 @@ static int read_btree_roots(struct bch_fs *c)
 	unsigned i;
 	int ret = 0;
 
-	for (i = 0; i < BTREE_ID_NR; i++) {
-		struct btree_root *r = &c->btree_roots[i];
+	for (i = 0; i < btree_id_nr_alive(c); i++) {
+		struct btree_root *r = bch2_btree_id_root(c, i);
 
 		if (!r->alive)
 			continue;
@@ -1014,7 +1014,7 @@ static int read_btree_roots(struct bch_fs *c)
 	}
 
 	for (i = 0; i < BTREE_ID_NR; i++) {
-		struct btree_root *r = &c->btree_roots[i];
+		struct btree_root *r = bch2_btree_id_root(c, i);
 
 		if (!r->b) {
 			r->alive = false;

fs/bcachefs/super.c

@@ -485,6 +485,7 @@ static void __bch2_fs_free(struct bch_fs *c)
 		for_each_possible_cpu(cpu)
 			kfree(per_cpu_ptr(c->btree_paths_bufs, cpu)->path);
 
+	darray_exit(&c->btree_roots_extra);
 	free_percpu(c->btree_paths_bufs);
 	free_percpu(c->pcpu);
 	mempool_exit(&c->large_bkey_pool);