1 files changed, 118 insertions, 308 deletions

diff --git a/libbcachefs/btree_gc.c b/libbcachefs/btree_gc.c
index ecbd9598..e1e899d5 100644
--- a/libbcachefs/btree_gc.c
+++ b/libbcachefs/btree_gc.c
@@ -52,12 +52,6 @@ static struct bkey_s unsafe_bkey_s_c_to_s(struct bkey_s_c k)
 	}}};
 }
 
-static bool should_restart_for_topology_repair(struct bch_fs *c)
-{
-	return c->opts.fix_errors != FSCK_FIX_no &&
-		!(c->recovery_passes_complete & BIT_ULL(BCH_RECOVERY_PASS_check_topology));
-}
-
 static inline void __gc_pos_set(struct bch_fs *c, struct gc_pos new_pos)
 {
 	preempt_disable();
@@ -546,9 +540,9 @@ reconstruct_root:
 			if (!bch2_btree_has_scanned_nodes(c, i)) {
 				mustfix_fsck_err(c, btree_root_unreadable_and_scan_found_nothing,
 						 "no nodes found for btree %s, continue?", bch2_btree_id_str(i));
-				bch2_btree_root_alloc_fake(c, i, 0);
+				bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			} else {
-				bch2_btree_root_alloc_fake(c, i, 1);
+				bch2_btree_root_alloc_fake_trans(trans, i, 1);
 				bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX);
 				ret = bch2_get_scanned_nodes(c, i, 0, POS_MIN, SPOS_MAX);
 				if (ret)
@@ -576,7 +570,7 @@ reconstruct_root:
 				goto reconstruct_root;
 
 			bch_err(c, "empty btree root %s", bch2_btree_id_str(i));
-			bch2_btree_root_alloc_fake(c, i, 0);
+			bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			r->alive = false;
 			ret = 0;
 		}
@@ -603,7 +597,7 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id
 	 * use check_bucket_ref here
 	 */
 	bkey_for_each_ptr_decode(k->k, ptrs_c, p, entry_c) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, p.ptr.dev);
 		struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
 		enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, p, entry_c);
 
@@ -736,7 +730,7 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id
 			 */
 			struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
 			bkey_for_each_ptr(ptrs, ptr) {
-				struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+				struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 				struct bucket *g = PTR_GC_BUCKET(ca, ptr);
 
 				ptr->gen = g->gen;
@@ -747,7 +741,7 @@ static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id
 restart_drop_ptrs:
 			ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
 			bkey_for_each_ptr_decode(bkey_i_to_s(new).k, ptrs, p, entry) {
-				struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
+				struct bch_dev *ca = bch2_dev_bkey_exists(c, p.ptr.dev);
 				struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
 				enum bch_data_type data_type = bch2_bkey_ptr_data_type(bkey_i_to_s_c(new), p, entry);
 
@@ -862,7 +856,7 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
 
 	ret = commit_do(trans, NULL, NULL, 0,
 			bch2_key_trigger(trans, btree_id, level, old,
-					 unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC));
+					 unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_gc));
 fsck_err:
 err:
 	printbuf_exit(&buf);
@@ -872,8 +866,7 @@ err:
 
 static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool initial)
 {
-	struct btree_node_iter iter;
-	struct bkey unpacked;
+	struct btree_and_journal_iter iter;
 	struct bkey_s_c k;
 	int ret = 0;
 
@@ -881,36 +874,33 @@ static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool i
 	if (ret)
 		return ret;
 
-	if (!btree_node_type_needs_gc(btree_node_type(b)))
-		return 0;
-
-	bch2_btree_node_iter_init_from_start(&iter, b);
+	bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b);
 
-	while ((k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked)).k) {
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false,
-				       &k, initial);
+	while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
+		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false, &k, initial);
 		if (ret)
-			return ret;
+			break;
 
-		bch2_btree_node_iter_advance(&iter, b);
+		bch2_btree_and_journal_iter_advance(&iter);
 	}
 
-	return 0;
+	bch2_btree_and_journal_iter_exit(&iter);
+	return ret;
 }
 
 static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id,
-			 bool initial, bool metadata_only)
+			 bool initial)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	struct btree *b;
-	unsigned depth = metadata_only ? 1 : 0;
+	unsigned target_depth = btree_type_has_ptrs(btree_id) ? 0 : 1;
 	int ret = 0;
 
 	gc_pos_set(c, gc_pos_btree(btree_id, POS_MIN, 0));
 
 	__for_each_btree_node(trans, iter, btree_id, POS_MIN,
-			      0, depth, BTREE_ITER_PREFETCH, b, ret) {
+			      0, target_depth, BTREE_ITER_prefetch, b, ret) {
 		bch2_verify_btree_nr_keys(b);
 
 		gc_pos_set(c, gc_pos_btree_node(b));
@@ -941,126 +931,61 @@ static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id,
 static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b,
 				      unsigned target_depth)
 {
-	struct bch_fs *c = trans->c;
-	struct btree_and_journal_iter iter;
-	struct bkey_s_c k;
-	struct bkey_buf cur;
-	struct printbuf buf = PRINTBUF;
-	int ret = 0;
-
-	ret = bch2_btree_node_check_topology(trans, b);
+	int ret = btree_gc_mark_node(trans, b, true);
 	if (ret)
 		return ret;
 
-	bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b);
-	bch2_bkey_buf_init(&cur);
-
-	while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
-		BUG_ON(bpos_lt(k.k->p, b->data->min_key));
-		BUG_ON(bpos_gt(k.k->p, b->data->max_key));
-
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level,
-				       false, &k, true);
-		if (ret)
-			goto fsck_err;
-
-		bch2_btree_and_journal_iter_advance(&iter);
-	}
-
 	if (b->c.level > target_depth) {
-		bch2_btree_and_journal_iter_exit(&iter);
+		struct bch_fs *c = trans->c;
+		struct btree_and_journal_iter iter;
+		struct bkey_s_c k;
+		struct bkey_buf cur;
+
+		bch2_bkey_buf_init(&cur);
 		bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b);
 		iter.prefetch = true;
 
 		while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
-			struct btree *child;
-
 			bch2_bkey_buf_reassemble(&cur, c, k);
 			bch2_btree_and_journal_iter_advance(&iter);
 
-			child = bch2_btree_node_get_noiter(trans, cur.k,
+			struct btree *child =
+				bch2_btree_node_get_noiter(trans, cur.k,
 						b->c.btree_id, b->c.level - 1,
 						false);
 			ret = PTR_ERR_OR_ZERO(child);
-
-			if (bch2_err_matches(ret, EIO)) {
-				bch2_topology_error(c);
-
-				if (__fsck_err(c,
-					  FSCK_CAN_FIX|
-					  FSCK_CAN_IGNORE|
-					  FSCK_NO_RATELIMIT,
-					  btree_node_read_error,
-					  "Unreadable btree node at btree %s level %u:\n"
-					  "  %s",
-					  bch2_btree_id_str(b->c.btree_id),
-					  b->c.level - 1,
-					  (printbuf_reset(&buf),
-					   bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur.k)), buf.buf)) &&
-				    should_restart_for_topology_repair(c)) {
-					bch_info(c, "Halting mark and sweep to start topology repair pass");
-					ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology);
-					goto fsck_err;
-				} else {
-					/* Continue marking when opted to not
-					 * fix the error: */
-					ret = 0;
-					set_bit(BCH_FS_initial_gc_unfixed, &c->flags);
-					continue;
-				}
-			} else if (ret) {
-				bch_err_msg(c, ret, "getting btree node");
+			bch_err_msg(c, ret, "getting btree node");
+			if (ret)
 				break;
-			}
 
-			ret = bch2_gc_btree_init_recurse(trans, child,
-							 target_depth);
+			ret = bch2_gc_btree_init_recurse(trans, child, target_depth);
 			six_unlock_read(&child->c.lock);
 
 			if (ret)
 				break;
 		}
+
+		bch2_bkey_buf_exit(&cur, c);
+		bch2_btree_and_journal_iter_exit(&iter);
 	}
-fsck_err:
-	bch2_bkey_buf_exit(&cur, c);
-	bch2_btree_and_journal_iter_exit(&iter);
-	printbuf_exit(&buf);
+
 	return ret;
 }
 
 static int bch2_gc_btree_init(struct btree_trans *trans,
-			      enum btree_id btree_id,
-			      bool metadata_only)
+			      enum btree_id btree_id)
 {
 	struct bch_fs *c = trans->c;
-	struct btree *b;
-	unsigned target_depth = metadata_only ? 1 : 0;
-	struct printbuf buf = PRINTBUF;
+	/*
+	 * We need to make sure every leaf node is readable before going RW
+	unsigned target_depth = btree_node_type_needs_gc(__btree_node_type(0, btree_id)) ? 0 : 1;
+	*/
+	unsigned target_depth = 0;
 	int ret = 0;
 
-	b = bch2_btree_id_root(c, btree_id)->b;
+	struct btree *b = bch2_btree_id_root(c, btree_id)->b;
 
 	six_lock_read(&b->c.lock, NULL, NULL);
-	printbuf_reset(&buf);
-	bch2_bpos_to_text(&buf, b->data->min_key);
-	if (mustfix_fsck_err_on(!bpos_eq(b->data->min_key, POS_MIN), c,
-				btree_root_bad_min_key,
-			"btree root with incorrect min_key: %s", buf.buf)) {
-		bch_err(c, "repair unimplemented");
-		ret = -BCH_ERR_fsck_repair_unimplemented;
-		goto fsck_err;
-	}
-
-	printbuf_reset(&buf);
-	bch2_bpos_to_text(&buf, b->data->max_key);
-	if (mustfix_fsck_err_on(!bpos_eq(b->data->max_key, SPOS_MAX), c,
-				btree_root_bad_max_key,
-			"btree root with incorrect max_key: %s", buf.buf)) {
-		bch_err(c, "repair unimplemented");
-		ret = -BCH_ERR_fsck_repair_unimplemented;
-		goto fsck_err;
-	}
-
 	if (b->c.level >= target_depth)
 		ret = bch2_gc_btree_init_recurse(trans, b, target_depth);
 
@@ -1070,11 +995,9 @@ static int bch2_gc_btree_init(struct btree_trans *trans,
 		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, true,
 				       &k, true);
 	}
-fsck_err:
 	six_unlock_read(&b->c.lock);
 
 	bch_err_fn(c, ret);
-	printbuf_exit(&buf);
 	return ret;
 }
 
@@ -1084,7 +1007,7 @@ static inline int btree_id_gc_phase_cmp(enum btree_id l, enum btree_id r)
 		(int) btree_id_to_gc_phase(r);
 }
 
-static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only)
+static int bch2_gc_btrees(struct bch_fs *c, bool initial)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
 	enum btree_id ids[BTREE_ID_NR];
@@ -1095,98 +1018,38 @@ static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only)
 		ids[i] = i;
 	bubble_sort(ids, BTREE_ID_NR, btree_id_gc_phase_cmp);
 
-	for (i = 0; i < BTREE_ID_NR && !ret; i++)
-		ret = initial
-			? bch2_gc_btree_init(trans, ids[i], metadata_only)
-			: bch2_gc_btree(trans, ids[i], initial, metadata_only);
+	for (i = 0; i < btree_id_nr_alive(c) && !ret; i++) {
+		unsigned btree = i < BTREE_ID_NR ? ids[i] : i;
 
-	for (i = BTREE_ID_NR; i < btree_id_nr_alive(c) && !ret; i++) {
-		if (!bch2_btree_id_root(c, i)->alive)
+		if (IS_ERR_OR_NULL(bch2_btree_id_root(c, btree)->b))
 			continue;
 
 		ret = initial
-			? bch2_gc_btree_init(trans, i, metadata_only)
-			: bch2_gc_btree(trans, i, initial, metadata_only);
+			? bch2_gc_btree_init(trans, btree)
+			: bch2_gc_btree(trans, btree, initial);
+
+		if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO),
+					c, btree_node_read_error,
+			       "btree node read error for %s",
+			       bch2_btree_id_str(btree)))
+			ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology);
 	}
-
+fsck_err:
 	bch2_trans_put(trans);
 	bch_err_fn(c, ret);
 	return ret;
 }
 
-static void mark_metadata_sectors(struct bch_fs *c, struct bch_dev *ca,
-				  u64 start, u64 end,
-				  enum bch_data_type type,
-				  unsigned flags)
-{
-	u64 b = sector_to_bucket(ca, start);
-
-	do {
-		unsigned sectors =
-			min_t(u64, bucket_to_sector(ca, b + 1), end) - start;
-
-		bch2_mark_metadata_bucket(c, ca, b, type, sectors,
-					  gc_phase(GC_PHASE_SB), flags);
-		b++;
-		start += sectors;
-	} while (start < end);
-}
-
-static void bch2_mark_dev_superblock(struct bch_fs *c, struct bch_dev *ca,
-				     unsigned flags)
-{
-	struct bch_sb_layout *layout = &ca->disk_sb.sb->layout;
-	unsigned i;
-	u64 b;
-
-	for (i = 0; i < layout->nr_superblocks; i++) {
-		u64 offset = le64_to_cpu(layout->sb_offset[i]);
-
-		if (offset == BCH_SB_SECTOR)
-			mark_metadata_sectors(c, ca, 0, BCH_SB_SECTOR,
-					      BCH_DATA_sb, flags);
-
-		mark_metadata_sectors(c, ca, offset,
-				      offset + (1 << layout->sb_max_size_bits),
-				      BCH_DATA_sb, flags);
-	}
-
-	for (i = 0; i < ca->journal.nr; i++) {
-		b = ca->journal.buckets[i];
-		bch2_mark_metadata_bucket(c, ca, b, BCH_DATA_journal,
-					  ca->mi.bucket_size,
-					  gc_phase(GC_PHASE_SB), flags);
-	}
-}
-
-static void bch2_mark_superblocks(struct bch_fs *c)
+static int bch2_mark_superblocks(struct bch_fs *c)
 {
 	mutex_lock(&c->sb_lock);
 	gc_pos_set(c, gc_phase(GC_PHASE_SB));
 
-	for_each_online_member(c, ca)
-		bch2_mark_dev_superblock(c, ca, BTREE_TRIGGER_GC);
+	int ret = bch2_trans_mark_dev_sbs_flags(c, BTREE_TRIGGER_gc);
 	mutex_unlock(&c->sb_lock);
+	return ret;
 }
 
-#if 0
-/* Also see bch2_pending_btree_node_free_insert_done() */
-static void bch2_mark_pending_btree_node_frees(struct bch_fs *c)
-{
-	struct btree_update *as;
-	struct pending_btree_node_free *d;
-
-	mutex_lock(&c->btree_interior_update_lock);
-	gc_pos_set(c, gc_phase(GC_PHASE_PENDING_DELETE));
-
-	for_each_pending_btree_node_free(c, as, d)
-		if (d->index_update_done)
-			bch2_mark_key(c, bkey_i_to_s_c(&d->key), BTREE_TRIGGER_GC);
-
-	mutex_unlock(&c->btree_interior_update_lock);
-}
-#endif
-
 static void bch2_gc_free(struct bch_fs *c)
 {
 	genradix_free(&c->reflink_gc_table);
@@ -1204,28 +1067,23 @@ static void bch2_gc_free(struct bch_fs *c)
 	c->usage_gc = NULL;
 }
 
-static int bch2_gc_done(struct bch_fs *c,
-			bool initial, bool metadata_only)
+static int bch2_gc_done(struct bch_fs *c)
 {
 	struct bch_dev *ca = NULL;
 	struct printbuf buf = PRINTBUF;
-	bool verify = !metadata_only &&
-		!c->opts.reconstruct_alloc &&
-		(!initial || (c->sb.compat & (1ULL << BCH_COMPAT_alloc_info)));
 	unsigned i;
 	int ret = 0;
 
 	percpu_down_write(&c->mark_lock);
 
-#define copy_field(_err, _f, _msg, ...)					\
-	if (dst->_f != src->_f &&					\
-	    (!verify ||							\
-	     fsck_err(c, _err, _msg ": got %llu, should be %llu"	\
-		      , ##__VA_ARGS__, dst->_f, src->_f)))		\
+#define copy_field(_err, _f, _msg, ...)						\
+	if (fsck_err_on(dst->_f != src->_f, c, _err,				\
+			_msg ": got %llu, should be %llu" , ##__VA_ARGS__,	\
+			dst->_f, src->_f))					\
 		dst->_f = src->_f
-#define copy_dev_field(_err, _f, _msg, ...)				\
+#define copy_dev_field(_err, _f, _msg, ...)					\
 	copy_field(_err, _f, "dev %u has wrong " _msg, ca->dev_idx, ##__VA_ARGS__)
-#define copy_fs_field(_err, _f, _msg, ...)				\
+#define copy_fs_field(_err, _f, _msg, ...)					\
 	copy_field(_err, _f, "fs has wrong " _msg, ##__VA_ARGS__)
 
 	for (i = 0; i < ARRAY_SIZE(c->usage); i++)
@@ -1258,31 +1116,24 @@ static int bch2_gc_done(struct bch_fs *c,
 		copy_fs_field(fs_usage_btree_wrong,
 			      b.btree,		"btree");
 
-		if (!metadata_only) {
-			copy_fs_field(fs_usage_data_wrong,
-				      b.data,	"data");
-			copy_fs_field(fs_usage_cached_wrong,
-				      b.cached,	"cached");
-			copy_fs_field(fs_usage_reserved_wrong,
-				      b.reserved,	"reserved");
-			copy_fs_field(fs_usage_nr_inodes_wrong,
-				      b.nr_inodes,"nr_inodes");
-
-			for (i = 0; i < BCH_REPLICAS_MAX; i++)
-				copy_fs_field(fs_usage_persistent_reserved_wrong,
-					      persistent_reserved[i],
-					      "persistent_reserved[%i]", i);
-		}
+		copy_fs_field(fs_usage_data_wrong,
+			      b.data,	"data");
+		copy_fs_field(fs_usage_cached_wrong,
+			      b.cached,	"cached");
+		copy_fs_field(fs_usage_reserved_wrong,
+			      b.reserved,	"reserved");
+		copy_fs_field(fs_usage_nr_inodes_wrong,
+			      b.nr_inodes,"nr_inodes");
+
+		for (i = 0; i < BCH_REPLICAS_MAX; i++)
+			copy_fs_field(fs_usage_persistent_reserved_wrong,
+				      persistent_reserved[i],
+				      "persistent_reserved[%i]", i);
 
 		for (i = 0; i < c->replicas.nr; i++) {
 			struct bch_replicas_entry_v1 *e =
 				cpu_replicas_entry(&c->replicas, i);
 
-			if (metadata_only &&
-			    (e->data_type == BCH_DATA_user ||
-			     e->data_type == BCH_DATA_cached))
-				continue;
-
 			printbuf_reset(&buf);
 			bch2_replicas_entry_to_text(&buf, e);
 
@@ -1361,11 +1212,10 @@ static inline bool bch2_alloc_v4_cmp(struct bch_alloc_v4 l,
 
 static int bch2_alloc_write_key(struct btree_trans *trans,
 				struct btree_iter *iter,
-				struct bkey_s_c k,
-				bool metadata_only)
+				struct bkey_s_c k)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, iter->pos.inode);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, iter->pos.inode);
 	struct bucket old_gc, gc, *b;
 	struct bkey_i_alloc_v4 *a;
 	struct bch_alloc_v4 old_convert, new;
@@ -1402,12 +1252,6 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 		bch2_dev_usage_update_m(c, ca, &old_gc, &gc);
 	percpu_up_read(&c->mark_lock);
 
-	if (metadata_only &&
-	    gc.data_type != BCH_DATA_sb &&
-	    gc.data_type != BCH_DATA_journal &&
-	    gc.data_type != BCH_DATA_btree)
-		return 0;
-
 	if (gen_after(old->gen, gc.gen))
 		return 0;
 
@@ -1460,12 +1304,12 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ])
 		a->v.io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
 
-	ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_NORUN);
+	ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_norun);
 fsck_err:
 	return ret;
 }
 
-static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_alloc_done(struct bch_fs *c)
 {
 	int ret = 0;
 
@@ -1474,9 +1318,9 @@ static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
 			for_each_btree_key_upto_commit(trans, iter, BTREE_ID_alloc,
 					POS(ca->dev_idx, ca->mi.first_bucket),
 					POS(ca->dev_idx, ca->mi.nbuckets - 1),
-					BTREE_ITER_SLOTS|BTREE_ITER_PREFETCH, k,
+					BTREE_ITER_slots|BTREE_ITER_prefetch, k,
 					NULL, NULL, BCH_TRANS_COMMIT_lazy_rw,
-				bch2_alloc_write_key(trans, &iter, k, metadata_only)));
+				bch2_alloc_write_key(trans, &iter, k)));
 		if (ret) {
 			percpu_ref_put(&ca->ref);
 			break;
@@ -1487,7 +1331,7 @@ static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
 	return ret;
 }
 
-static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_alloc_start(struct bch_fs *c)
 {
 	for_each_member_device(c, ca) {
 		struct bucket_array *buckets = kvmalloc(sizeof(struct bucket_array) +
@@ -1506,8 +1350,8 @@ static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only)
 
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-					 BTREE_ITER_PREFETCH, k, ({
-			struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
+					 BTREE_ITER_prefetch, k, ({
+			struct bch_dev *ca = bch2_dev_bkey_exists(c, k.k->p.inode);
 			struct bucket *g = gc_bucket(ca, k.k->p.offset);
 
 			struct bch_alloc_v4 a_convert;
@@ -1515,36 +1359,19 @@ static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only)
 
 			g->gen_valid	= 1;
 			g->gen		= a->gen;
-
-			if (metadata_only &&
-			    (a->data_type == BCH_DATA_user ||
-			     a->data_type == BCH_DATA_cached ||
-			     a->data_type == BCH_DATA_parity)) {
-				g->data_type		= a->data_type;
-				g->dirty_sectors	= a->dirty_sectors;
-				g->cached_sectors	= a->cached_sectors;
-				g->stripe		= a->stripe;
-				g->stripe_redundancy	= a->stripe_redundancy;
-			}
-
 			0;
 		})));
 	bch_err_fn(c, ret);
 	return ret;
 }
 
-static void bch2_gc_alloc_reset(struct bch_fs *c, bool metadata_only)
+static void bch2_gc_alloc_reset(struct bch_fs *c)
 {
 	for_each_member_device(c, ca) {
 		struct bucket_array *buckets = gc_bucket_array(ca);
 		struct bucket *g;
 
 		for_each_bucket(g, buckets) {
-			if (metadata_only &&
-			    (g->data_type == BCH_DATA_user ||
-			     g->data_type == BCH_DATA_cached ||
-			     g->data_type == BCH_DATA_parity))
-				continue;
 			g->data_type = 0;
 			g->dirty_sectors = 0;
 			g->cached_sectors = 0;
@@ -1600,35 +1427,27 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_gc_reflink_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_reflink_done(struct bch_fs *c)
 {
 	size_t idx = 0;
 
-	if (metadata_only)
-		return 0;
-
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 				BTREE_ID_reflink, POS_MIN,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_gc_write_reflink_key(trans, &iter, k, &idx)));
 	c->reflink_gc_nr = 0;
 	return ret;
 }
 
-static int bch2_gc_reflink_start(struct bch_fs *c,
-				 bool metadata_only)
+static int bch2_gc_reflink_start(struct bch_fs *c)
 {
-
-	if (metadata_only)
-		return 0;
-
 	c->reflink_gc_nr = 0;
 
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_reflink, POS_MIN,
-				   BTREE_ITER_PREFETCH, k, ({
+				   BTREE_ITER_prefetch, k, ({
 			const __le64 *refcount = bkey_refcount_c(k);
 
 			if (!refcount)
@@ -1651,7 +1470,7 @@ static int bch2_gc_reflink_start(struct bch_fs *c,
 	return ret;
 }
 
-static void bch2_gc_reflink_reset(struct bch_fs *c, bool metadata_only)
+static void bch2_gc_reflink_reset(struct bch_fs *c)
 {
 	struct genradix_iter iter;
 	struct reflink_gc *r;
@@ -1713,20 +1532,17 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_gc_stripes_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_stripes_done(struct bch_fs *c)
 {
-	if (metadata_only)
-		return 0;
-
 	return bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 				BTREE_ID_stripes, POS_MIN,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_gc_write_stripes_key(trans, &iter, k)));
 }
 
-static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only)
+static void bch2_gc_stripes_reset(struct bch_fs *c)
 {
 	genradix_free(&c->gc_stripes);
 }
@@ -1736,7 +1552,6 @@ static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only)
  *
  * @c:			filesystem object
  * @initial:		are we in recovery?
- * @metadata_only:	are we just checking metadata references, or everything?
  *
  * Returns: 0 on success, or standard errcode on failure
  *
@@ -1755,7 +1570,7 @@ static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only)
  *    move around - if references move backwards in the ordering GC
  *    uses, GC could skip past them
  */
-int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only)
+static int bch2_gc(struct bch_fs *c, bool initial)
 {
 	unsigned iter = 0;
 	int ret;
@@ -1767,23 +1582,20 @@ int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only)
 	bch2_btree_interior_updates_flush(c);
 
 	ret   = bch2_gc_start(c) ?:
-		bch2_gc_alloc_start(c, metadata_only) ?:
-		bch2_gc_reflink_start(c, metadata_only);
+		bch2_gc_alloc_start(c) ?:
+		bch2_gc_reflink_start(c);
 	if (ret)
 		goto out;
 again:
 	gc_pos_set(c, gc_phase(GC_PHASE_START));
 
-	bch2_mark_superblocks(c);
-
-	ret = bch2_gc_btrees(c, initial, metadata_only);
+	ret = bch2_mark_superblocks(c);
+	BUG_ON(ret);
 
+	ret = bch2_gc_btrees(c, initial);
 	if (ret)
 		goto out;
 
-#if 0
-	bch2_mark_pending_btree_node_frees(c);
-#endif
 	c->gc_count++;
 
 	if (test_bit(BCH_FS_need_another_gc, &c->flags) ||
@@ -1801,9 +1613,9 @@ again:
 		clear_bit(BCH_FS_need_another_gc, &c->flags);
 		__gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING));
 
-		bch2_gc_stripes_reset(c, metadata_only);
-		bch2_gc_alloc_reset(c, metadata_only);
-		bch2_gc_reflink_reset(c, metadata_only);
+		bch2_gc_stripes_reset(c);
+		bch2_gc_alloc_reset(c);
+		bch2_gc_reflink_reset(c);
 		ret = bch2_gc_reset(c);
 		if (ret)
 			goto out;
@@ -1816,10 +1628,10 @@ out:
 	if (!ret) {
 		bch2_journal_block(&c->journal);
 
-		ret   = bch2_gc_alloc_done(c, metadata_only) ?:
-			bch2_gc_done(c, initial, metadata_only) ?:
-			bch2_gc_stripes_done(c, metadata_only) ?:
-			bch2_gc_reflink_done(c, metadata_only);
+		ret   = bch2_gc_alloc_done(c) ?:
+			bch2_gc_done(c) ?:
+			bch2_gc_stripes_done(c) ?:
+			bch2_gc_reflink_done(c);
 
 		bch2_journal_unblock(&c->journal);
 	}
@@ -1842,6 +1654,11 @@ out:
 	return ret;
 }
 
+int bch2_check_allocations(struct bch_fs *c)
+{
+	return bch2_gc(c, true);
+}
+
 static int gc_btree_gens_key(struct btree_trans *trans,
 			     struct btree_iter *iter,
 			     struct bkey_s_c k)
@@ -1853,7 +1670,7 @@ static int gc_btree_gens_key(struct btree_trans *trans,
 
 	percpu_down_read(&c->mark_lock);
 	bkey_for_each_ptr(ptrs, ptr) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 
 		if (ptr_stale(ca, ptr) > 16) {
 			percpu_up_read(&c->mark_lock);
@@ -1862,7 +1679,7 @@ static int gc_btree_gens_key(struct btree_trans *trans,
 	}
 
 	bkey_for_each_ptr(ptrs, ptr) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 		u8 *gen = &ca->oldest_gen[PTR_BUCKET_NR(ca, ptr)];
 
 		if (gen_after(*gen, ptr->gen))
@@ -1883,7 +1700,7 @@ update:
 static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct btree_iter *iter,
 				       struct bkey_s_c k)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(trans->c, iter->pos.inode);
+	struct bch_dev *ca = bch2_dev_bkey_exists(trans->c, iter->pos.inode);
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
 	struct bkey_i_alloc_v4 *a_mut;
@@ -1944,7 +1761,7 @@ int bch2_gc_gens(struct bch_fs *c)
 			ret = bch2_trans_run(c,
 				for_each_btree_key_commit(trans, iter, i,
 						POS_MIN,
-						BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+						BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
 						k,
 						NULL, NULL,
 						BCH_TRANS_COMMIT_no_enospc,
@@ -1956,7 +1773,7 @@ int bch2_gc_gens(struct bch_fs *c)
 	ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
 				POS_MIN,
-				BTREE_ITER_PREFETCH,
+				BTREE_ITER_prefetch,
 				k,
 				NULL, NULL,
 				BCH_TRANS_COMMIT_no_enospc,
@@ -2023,14 +1840,7 @@ static int bch2_gc_thread(void *arg)
 		last = atomic64_read(&clock->now);
 		last_kick = atomic_read(&c->kick_gc);
 
-		/*
-		 * Full gc is currently incompatible with btree key cache:
-		 */
-#if 0
-		ret = bch2_gc(c, false, false);
-#else
 		bch2_gc_gens(c);
-#endif
 		debug_check_no_locks_held();
 	}