ctdb: add/implement CTDB_CONTROL_TCP_CLIENT_DISCONNECTED

With multichannel a ctdb connection from smbd may hold multiple
tcp connections, which can be disconnected before the smbd
process terminates the whole ctdb connection, so we a
way to remove undo 'CTDB_CONTROL_TCP_CLIENT' again.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=15523

Signed-off-by: Stefan Metzmacher <[email protected]>
Reviewed-by: Martin Schwenke <[email protected]>

ctdb/include/ctdb_private.h

@@ -892,6 +892,9 @@ int ctdb_set_public_addresses(struct ctdb_context *ctdb, bool check_addresses);
 
 int32_t ctdb_control_tcp_client(struct ctdb_context *ctdb, uint32_t client_id,
 				TDB_DATA indata);
+int32_t ctdb_control_tcp_client_disconnected(struct ctdb_context *ctdb,
+					     uint32_t client_id,
+					     TDB_DATA indata);
 int32_t ctdb_control_tcp_add(struct ctdb_context *ctdb, TDB_DATA indata,
 			     bool tcp_update_needed);
 int32_t ctdb_control_tcp_remove(struct ctdb_context *ctdb, TDB_DATA indata);

ctdb/protocol/protocol.h

@@ -381,6 +381,7 @@ enum ctdb_controls {CTDB_CONTROL_PROCESS_EXISTS          = 0,
 		    CTDB_CONTROL_ECHO_DATA               = 156,
 		    CTDB_CONTROL_DISABLE_NODE            = 157,
 		    CTDB_CONTROL_ENABLE_NODE             = 158,
+		    CTDB_CONTROL_TCP_CLIENT_DISCONNECTED = 159,
 };
 
 #define MAX_COUNT_BUCKETS 16

ctdb/protocol/protocol_control.c

@@ -410,6 +410,10 @@ static size_t ctdb_req_control_data_len(struct ctdb_req_control_data *cd)
 
 	case CTDB_CONTROL_ENABLE_NODE:
 		break;
+
+	case CTDB_CONTROL_TCP_CLIENT_DISCONNECTED:
+		len = ctdb_connection_len(cd->data.conn);
+		break;
 	}
 
 	return len;
@@ -1016,6 +1020,14 @@ static int ctdb_req_control_data_pull(uint8_t *buf, size_t buflen,
 					  &cd->data.echo_data,
 					  &np);
 		break;
+
+	case CTDB_CONTROL_TCP_CLIENT_DISCONNECTED:
+		ret = ctdb_connection_pull(buf,
+					   buflen,
+					   mem_ctx,
+					   &cd->data.conn,
+					   &np);
+		break;
 	}
 
 	if (ret != 0) {
@@ -1376,6 +1388,9 @@ static size_t ctdb_reply_control_data_len(struct ctdb_reply_control_data *cd)
 
 	case CTDB_CONTROL_ENABLE_NODE:
 		break;
+
+	case CTDB_CONTROL_TCP_CLIENT_DISCONNECTED:
+		break;
 	}
 
 	return len;

ctdb/protocol/protocol_debug.c

@@ -245,6 +245,7 @@ static void ctdb_opcode_print(uint32_t opcode, FILE *fp)
 		{ CTDB_CONTROL_ECHO_DATA, "ECHO_DATA" },
 		{ CTDB_CONTROL_DISABLE_NODE, "DISABLE_NODE" },
 		{ CTDB_CONTROL_ENABLE_NODE, "ENABLE_NODE" },
+		{ CTDB_CONTROL_TCP_CLIENT_DISCONNECTED, "TCP_CLIENT_DISCONNECTED" },
 		{ MAP_END, "" },
 	};
 

ctdb/server/ctdb_control.c

@@ -868,6 +868,10 @@ static int32_t ctdb_control_dispatch(struct ctdb_context *ctdb,
 		CHECK_CONTROL_DATA_SIZE(0);
 		return ctdb_control_enable_node(ctdb);
 
+	case CTDB_CONTROL_TCP_CLIENT_DISCONNECTED:
+		CHECK_CONTROL_DATA_SIZE(sizeof(struct ctdb_connection));
+		return ctdb_control_tcp_client_disconnected(ctdb, client_id, indata);
+
 	default:
 		DEBUG(DEBUG_CRIT,(__location__ " Unknown CTDB control opcode %u\n", opcode));
 		return -1;

ctdb/server/ctdb_takeover.c

@@ -1377,6 +1377,92 @@ int32_t ctdb_control_tcp_client(struct ctdb_context *ctdb, uint32_t client_id,
 	return 0;
 }
 
+static bool ctdb_client_remove_tcp(struct ctdb_client *client,
+				   const struct ctdb_connection *conn)
+{
+	struct ctdb_tcp_list *tcp = NULL;
+	struct ctdb_tcp_list *tcp_next = NULL;
+	bool found = false;
+
+	for (tcp = client->tcp_list; tcp != NULL; tcp = tcp_next) {
+		bool same;
+
+		tcp_next = tcp->next;
+
+		same = ctdb_connection_same(conn, &tcp->connection);
+		if (!same) {
+			continue;
+		}
+
+		TALLOC_FREE(tcp);
+		found = true;
+	}
+
+	return found;
+}
+
+/*
+  called by a client to inform us of a TCP connection that was disconnected
+ */
+int32_t ctdb_control_tcp_client_disconnected(struct ctdb_context *ctdb,
+					     uint32_t client_id,
+					     TDB_DATA indata)
+{
+	struct ctdb_client *client = reqid_find(ctdb->idr, client_id, struct ctdb_client);
+	struct ctdb_connection *tcp_sock = NULL;
+	int ret;
+	TDB_DATA data;
+	char conn_str[132] = { 0, };
+	bool found = false;
+
+	tcp_sock = (struct ctdb_connection *)indata.dptr;
+
+	ctdb_canonicalize_ip_inplace(&tcp_sock->src);
+	ctdb_canonicalize_ip_inplace(&tcp_sock->dst);
+
+	ret = ctdb_connection_to_buf(conn_str,
+				     sizeof(conn_str),
+				     tcp_sock,
+				     false,
+				     " -> ");
+	if (ret != 0) {
+		strlcpy(conn_str, "UNKNOWN", sizeof(conn_str));
+	}
+
+	found = ctdb_client_remove_tcp(client, tcp_sock);
+	if (!found) {
+		DBG_DEBUG("TCP connection %s not found "
+			  "(client_id %u pid %u).\n",
+			  conn_str, client_id, client->pid);
+		return 0;
+	}
+
+	D_INFO("deregistered TCP connection %s "
+	       "(client_id %u pid %u)\n",
+	       conn_str, client_id, client->pid);
+
+	data.dptr = (uint8_t *)tcp_sock;
+	data.dsize = sizeof(*tcp_sock);
+
+	/* tell all nodes about this tcp connection is gone */
+	ret = ctdb_daemon_send_control(ctdb,
+				       CTDB_BROADCAST_CONNECTED,
+				       0,
+				       CTDB_CONTROL_TCP_REMOVE,
+				       0,
+				       CTDB_CTRL_FLAG_NOREPLY,
+				       data,
+				       NULL,
+				       NULL);
+	if (ret != 0) {
+		DBG_ERR("Failed to send CTDB_CONTROL_TCP_REMOVE: %s\n",
+			conn_str);
+		return -1;
+	}
+
+	return 0;
+}
+
 /*
   find a tcp address on a list
  */