bcachefs: bch2_trans_unlock() must always be followed by relock() or begin()

We're about to add new asserts for btree_trans locking consistency, and
part of that requires that aren't using the btree_trans while it's
unlocked.

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

10 files changed, 29 insertions, 4 deletions

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 2f91ff67453f..f07373b78174 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -2172,6 +2172,9 @@ int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
 	u64 now;
 	int ret = 0;
 
+	if (bch2_trans_relock(trans))
+		bch2_trans_begin(trans);
+
 	a = bch2_trans_start_alloc_update(trans, &iter,  POS(dev, bucket_nr));
 	ret = PTR_ERR_OR_ZERO(a);
 	if (ret)
diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c
index df4439a38df5..fb8825c4e7ad 100644
--- a/fs/bcachefs/alloc_foreground.c
+++ b/fs/bcachefs/alloc_foreground.c
@@ -1342,6 +1342,10 @@ retry:
 
 	*wp_ret = wp = writepoint_find(trans, write_point.v);
 
+	ret = bch2_trans_relock(trans);
+	if (ret)
+		goto err;
+
 	/* metadata may not allocate on cache devices: */
 	if (wp->data_type != BCH_DATA_user)
 		have_cache = true;
diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 3bef5b1af16d..0d44d613e4ac 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -729,6 +729,8 @@ transaction_restart:							\
 #define for_each_btree_key_upto(_trans, _iter, _btree_id,		\
 				_start, _end, _flags, _k, _do)		\
 ({									\
+	bch2_trans_begin(trans);					\
+									\
 	struct btree_iter _iter;					\
 	bch2_trans_iter_init((_trans), &(_iter), (_btree_id),		\
 			     (_start), (_flags));			\
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 4aa62a74a59b..35f80c97b973 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -737,9 +737,6 @@ err:
 	 */
 	b = READ_ONCE(as->b);
 	if (b) {
-		btree_path_idx_t path_idx = bch2_path_get_unlocked_mut(trans,
-						as->btree_id, b->c.level, b->key.k.p);
-		struct btree_path *path = trans->paths + path_idx;
 		/*
 		 * @b is the node we did the final insert into:
 		 *
@@ -763,6 +760,10 @@ err:
 		 * have here:
 		 */
 		bch2_trans_unlock(trans);
+		bch2_trans_begin(trans);
+		btree_path_idx_t path_idx = bch2_path_get_unlocked_mut(trans,
+						as->btree_id, b->c.level, b->key.k.p);
+		struct btree_path *path = trans->paths + path_idx;
 		btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
 		mark_btree_node_locked(trans, path, b->c.level, BTREE_NODE_INTENT_LOCKED);
 		path->l[b->c.level].lock_seq = six_lock_seq(&b->c.lock);
diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c
index cae4e4eb5329..e8c7c5cec03f 100644
--- a/fs/bcachefs/data_update.c
+++ b/fs/bcachefs/data_update.c
@@ -386,6 +386,8 @@ static void bch2_update_unwritten_extent(struct btree_trans *trans,
 	while (bio_sectors(bio)) {
 		unsigned sectors = bio_sectors(bio);
 
+		bch2_trans_begin(trans);
+
 		bch2_trans_iter_init(trans, &iter, update->btree_id, update->op.pos,
 				     BTREE_ITER_slots);
 		ret = lockrestart_do(trans, ({
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 81cfc74828fd..841bb92e53df 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -1036,6 +1036,10 @@ retry:
 
 		bch2_btree_iter_set_pos(&iter,
 			POS(iter.pos.inode, iter.pos.offset + sectors));
+
+		ret = bch2_trans_relock(trans);
+		if (ret)
+			break;
 	}
 	start = iter.pos.offset;
 	bch2_trans_iter_exit(trans, &iter);
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index 217cb98ed707..73e25250de75 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -1248,6 +1248,10 @@ retry:
 
 		buckets.nr = 0;
 
+		ret = bch2_trans_relock(trans);
+		if (ret)
+			break;
+
 		k = bch2_btree_iter_peek_slot(&iter);
 		ret = bkey_err(k);
 		if (ret)
diff --git a/fs/bcachefs/movinggc.c b/fs/bcachefs/movinggc.c
index d0afd8bc0193..10bfb31c151b 100644
--- a/fs/bcachefs/movinggc.c
+++ b/fs/bcachefs/movinggc.c
@@ -158,6 +158,8 @@ static int bch2_copygc_get_buckets(struct moving_context *ctxt,
 	if (bch2_fs_fatal_err_on(ret, c, "%s: from bch2_btree_write_buffer_tryflush()", bch2_err_str(ret)))
 		return ret;
 
+	bch2_trans_begin(trans);
+
 	ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
 				  lru_pos(BCH_LRU_FRAGMENTATION_START, 0, 0),
 				  lru_pos(BCH_LRU_FRAGMENTATION_START, U64_MAX, LRU_TIME_MAX),
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index c640c0f13976..cf81e5128c3a 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -323,6 +323,8 @@ static int do_rebalance(struct moving_context *ctxt)
 	struct bkey_s_c k;
 	int ret = 0;
 
+	bch2_trans_begin(trans);
+
 	bch2_move_stats_init(&r->work_stats, "rebalance_work");
 	bch2_move_stats_init(&r->scan_stats, "rebalance_scan");
 
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index cc603276f917..27ee27b285bd 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -202,7 +202,7 @@ int bch2_journal_replay(struct bch_fs *c)
 	struct journal *j = &c->journal;
 	u64 start_seq	= c->journal_replay_seq_start;
 	u64 end_seq	= c->journal_replay_seq_start;
-	struct btree_trans *trans = bch2_trans_get(c);
+	struct btree_trans *trans = NULL;
 	bool immediate_flush = false;
 	int ret = 0;
 
@@ -216,6 +216,7 @@ int bch2_journal_replay(struct bch_fs *c)
 	BUG_ON(!atomic_read(&keys->ref));
 
 	move_gap(keys, keys->nr);
+	trans = bch2_trans_get(c);
 
 	/*
 	 * First, attempt to replay keys in sorted order. This is more