- ctdb/ib minor bugfixes (error case)

- make ctdb capable of alternative connection (like ib) again, solved the fork problem
- do_debug memory overwrite bugfix (occured using ibwrapper_test with wrong address given)

(This used to be ctdb commit da0b84cda26d544f63841dfd770ed7ebad401944)

ctdb/common/ctdb.c

@@ -31,22 +31,45 @@
 */
 int ctdb_set_transport(struct ctdb_context *ctdb, const char *transport)
 {
+	ctdb->transport = talloc_strdup(ctdb, transport);
+	return 0;
+}
+
+int ctdb_init_transport(struct ctdb_context *ctdb)
+{
+	int	i;
 	int ctdb_tcp_init(struct ctdb_context *ctdb);
+	int	transport_found = 0;
 #ifdef USE_INFINIBAND
 	int ctdb_ibw_init(struct ctdb_context *ctdb);
 #endif /* USE_INFINIBAND */
 
-	if (strcmp(transport, "tcp") == 0) {
-		return ctdb_tcp_init(ctdb);
+	if (strcmp(ctdb->transport, "tcp") == 0) {
+		transport_found = 1;
+		if (ctdb_tcp_init(ctdb))
+			return -1;
 	}
 #ifdef USE_INFINIBAND
-	if (strcmp(transport, "ib") == 0) {
-		return ctdb_ibw_init(ctdb);
+	else if (strcmp(ctdb->transport, "ib") == 0) {
+		transport_found = 1;
+		if (ctdb_ibw_init(ctdb))
+			return -1;
 	}
 #endif /* USE_INFINIBAND */
 
-	ctdb_set_error(ctdb, "Unknown transport '%s'\n", transport);
-	return -1;
+	if (!transport_found) {
+		ctdb_set_error(ctdb, "Unknown transport '%s'\n", ctdb->transport);
+		return -1;
+	}
+
+	for(i=0; i<ctdb->num_nodes; i++) {
+		if (ctdb->methods->add_node(ctdb->nodes[i]) != 0) {
+			DEBUG(0, ("methods->add_node failed at %d\n", i));
+			return -1;
+		}
+	}
+
+	return 0;
 }
 
 /*
@@ -116,11 +139,6 @@ static int ctdb_add_node(struct ctdb_context *ctdb, char *nstr)
 	   will change! */
 	node->vnn = ctdb->num_nodes;
 
-	if (ctdb->methods->add_node(node) != 0) {
-		talloc_free(node);
-		return -1;
-	}
-
 	if (ctdb_same_address(&ctdb->address, &node->address)) {
 		ctdb->vnn = node->vnn;
 	}

ctdb/common/ctdb_client.c

@@ -373,7 +373,10 @@ int ctdb_set_message_handler(struct ctdb_context *ctdb, uint32_t srvid,
 
 	/* if the domain socket is not yet open, open it */
 	if (ctdb->daemon.sd==-1) {
-		ux_socket_connect(ctdb);
+		if (ux_socket_connect(ctdb)) {
+			DEBUG(0, ("ux_socket_connect failed\n"));
+			return -1;
+		}
 	}
 
 	ZERO_STRUCT(c);

ctdb/common/ctdb_daemon.c

@@ -44,6 +44,11 @@ static void daemon_incoming_packet(void *, uint8_t *, uint32_t );
 
 static void ctdb_main_loop(struct ctdb_context *ctdb)
 {
+	/* we are the dispatcher process now, so start the protocol going */
+	if (ctdb_init_transport(ctdb)) {
+		exit(1);
+	}
+
 	ctdb->methods->start(ctdb);
 
 	/* go into a wait loop to allow other nodes to complete */
@@ -594,6 +599,12 @@ int ctdb_start(struct ctdb_context *ctdb)
 		close(fd[0]);
 		close(ctdb->daemon.sd);
 		ctdb->daemon.sd = -1;
+
+		/* Added because of ctdb->methods->allocate_pkt calls */
+		/* TODO: clean */
+		int ctdb_tcp_init(struct ctdb_context *ctdb);
+		ctdb_tcp_init(ctdb);
+
 		return 0;
 	}
 

ctdb/ib/ibwrapper.c

@@ -557,17 +557,17 @@ static void ibw_event_handler_cm(struct event_context *ev,
 	return;
 error:
 	if (event!=NULL && (rc=rdma_ack_cm_event(event))) {
-		sprintf(ibw_lasterr, "rdma_ack_cm_event failed with %d\n", rc);
-		goto error;
+		DEBUG(0, ("rdma_ack_cm_event failed with %d\n", rc));
 	}
 
 	DEBUG(0, ("cm event handler: %s", ibw_lasterr));
 
 	if (cma_id!=pctx->cm_id) {
 		conn = talloc_get_type(cma_id->context, struct ibw_conn);
-		if (conn)
+		if (conn) {
 			conn->state = IBWC_ERROR;
-		pctx->connstate_func(NULL, conn);
+			pctx->connstate_func(NULL, conn);
+		}
 	} else {
 		ctx->state = IBWS_ERROR;
 		pctx->connstate_func(ctx, NULL);

ctdb/include/ctdb.h

@@ -70,6 +70,12 @@ struct ctdb_context *ctdb_init(struct event_context *ev);
 */
 int ctdb_set_transport(struct ctdb_context *ctdb, const char *transport);
 
+/*
+  initialize the chosen transport
+  do this in the process where it is going to be used
+ */
+int ctdb_init_transport(struct ctdb_context *ctdb);
+
 /*
   set the directory for the local databases
 */

ctdb/include/ctdb_private.h

@@ -120,6 +120,7 @@ struct ctdb_context {
 	struct ctdb_address address;
 	const char *name;
 	const char *db_directory;
+	char *transport;
 	uint32_t vnn; /* our own vnn */
 	uint32_t num_nodes;
 	uint32_t num_connected;
@@ -362,6 +363,11 @@ int ctdbd_start(struct ctdb_context *ctdb);
 struct ctdb_call_state *ctdbd_call_send(struct ctdb_db_context *ctdb_db, struct ctdb_call *call);
 int ctdbd_call_recv(struct ctdb_call_state *state, struct ctdb_call *call);
 
+/*
+  allocate a packet for sending via queue
+*/
+void *ctdb_queue_allocate_pkt(TALLOC_CTX *mem_ctx, size_t size);
+
 /*
   queue a packet for sending
 */

ctdb/lib/util/debug.c

@@ -22,6 +22,22 @@
 #include "system/time.h"
 #include <unistd.h>
 
+static int vasprintf2(char **ptr, const char *format, va_list ap)
+{
+    int ret;
+    va_list tmp_ap;
+
+    va_copy(tmp_ap, ap);
+    ret = vsnprintf(NULL, 0, format, tmp_ap);
+    if (ret <= 0) return ret;
+
+    (*ptr) = (char *)malloc(ret+1);
+    if (!*ptr) return -1;
+    ret = vsnprintf(*ptr, ret+1, format, ap);
+
+    return ret;
+}
+
 void do_debug(const char *format, ...)
 {
 	struct timeval tm;
@@ -29,7 +45,7 @@ void do_debug(const char *format, ...)
 	char *s = NULL;
 
 	va_start(ap, format);
-	vasprintf(&s, format, ap);
+	vasprintf2(&s, format, ap);
 	va_end(ap);
 
 	gettimeofday(&tm, NULL);

ctdb/tests/ctdb_bench.c

@@ -204,7 +204,8 @@ int main(int argc, const char *argv[])
 	/* start the protocol running */
 	ret = ctdb_start(ctdb);
 
-	ctdb_set_message_handler(ctdb, 0, ring_message_handler,&msg_count);
+	if (ctdb_set_message_handler(ctdb, 0, ring_message_handler,&msg_count))
+		goto error;
 
 	/* wait until all nodes are connected (should not be needed
 	   outside of test code) */
@@ -212,6 +213,7 @@ int main(int argc, const char *argv[])
 
 	bench_ring(ctdb, ev);
        
+error:
 	/* shut it down */
 	ctdb_shutdown(ctdb);