bcachefs: Fix a trans path overflow in bch2_btree_delete_range_trans()

bch2_btree_delete_range_trans() was using btree_trans_too_many_iters()
to avoid path overflow, but this was buggy here (and also
btree_trans_too_many_iters() is suspect in general).

btree_trans_too_many_iters() only returns true when we're close to the
maximum number of paths - within 8 - but extent insert/delete assumes
that it can use more paths than that.

Instead, we need to call bch2_trans_begin() on every loop iteration.
Since we don't want to call bch2_trans_begin() (restarting the outer
transaction) if the call was a no-op - if we had no work to do - we have
to structure things a bit oddly.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/btree_update_leaf.c

@@ -1651,15 +1651,18 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
 	int ret = 0;
 
 	bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_INTENT);
-retry:
-	while ((k = bch2_btree_iter_peek(&iter)).k &&
-	       !(ret = bkey_err(k) ?:
-		 btree_trans_too_many_iters(trans)) &&
-	       bkey_cmp(iter.pos, end) < 0) {
+	while ((k = bch2_btree_iter_peek(&iter)).k) {
 		struct disk_reservation disk_res =
 			bch2_disk_reservation_init(trans->c, 0);
 		struct bkey_i delete;
 
+		ret = bkey_err(k);
+		if (ret)
+			goto err;
+
+		if (bkey_cmp(iter.pos, end) >= 0)
+			break;
+
 		bkey_init(&delete.k);
 
 		/*
@@ -1688,23 +1691,27 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
 
 			ret = bch2_extent_trim_atomic(trans, &iter, &delete);
 			if (ret)
-				break;
+				goto err;
 		}
 
 		ret   = bch2_trans_update(trans, &iter, &delete, update_flags) ?:
 			bch2_trans_commit(trans, &disk_res, journal_seq,
 					  BTREE_INSERT_NOFAIL);
 		bch2_disk_reservation_put(trans->c, &disk_res);
-		if (ret)
-			break;
-	}
-
-	if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
+err:
+		/*
+		 * the bch2_trans_begin() call is in a weird place because we
+		 * need to call it after every transaction commit, to avoid path
+		 * overflow, but don't want to call it if the delete operation
+		 * is a no-op and we have no work to do:
+		 */
 		bch2_trans_begin(trans);
+
+		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 			ret = 0;
-		goto retry;
+		if (ret)
+			break;
 	}
-
 	bch2_trans_iter_exit(trans, &iter);
 
 	if (!ret && trans_was_restarted(trans, restart_count))