X-Git-Url: http://git.lttng.org./?a=blobdiff_plain;f=src%2Fbin%2Flttng-relayd%2Fmain.c;h=238d20431618677943f8197434f65343fca60a03;hb=4d3350dd504d64cb10770c683848afd04a5817ea;hp=dc19a69c77a548df81fffc2aa7e44affd12279b4;hpb=e0547b835d76e0a3a2c3082d7c10e6902a1f3e04;p=lttng-tools.git

diff --git a/src/bin/lttng-relayd/main.c b/src/bin/lttng-relayd/main.c
index dc19a69c7..238d20431 100644
--- a/src/bin/lttng-relayd/main.c
+++ b/src/bin/lttng-relayd/main.c
@@ -70,6 +70,7 @@
 #include "stream.h"
 #include "connection.h"
 #include "tracefile-array.h"
+#include "tcp_keep_alive.h"
 
 /* command line options */
 char *opt_output_path;
@@ -891,6 +892,15 @@ restart:
 					lttcomm_destroy_sock(newsock);
 					goto error;
 				}
+
+				ret = socket_apply_keep_alive_config(newsock->fd);
+				if (ret < 0) {
+					ERR("Failed to apply TCP keep-alive configuration on socket (%i)",
+							newsock->fd);
+					lttcomm_destroy_sock(newsock);
+					goto error;
+				}
+
 				new_conn = connection_create(newsock, type);
 				if (!new_conn) {
 					lttcomm_destroy_sock(newsock);
@@ -969,12 +979,16 @@ static void *relay_thread_dispatcher(void *data)
 
 	health_code_update();
 
-	while (!CMM_LOAD_SHARED(dispatch_thread_exit)) {
+	for (;;) {
 		health_code_update();
 
 		/* Atomically prepare the queue futex */
 		futex_nto1_prepare(&relay_conn_queue.futex);
 
+		if (CMM_LOAD_SHARED(dispatch_thread_exit)) {
+			break;
+		}
+
 		do {
 			health_code_update();
 
@@ -1579,6 +1593,7 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr,
 {
 	int ret;
 	struct lttcomm_relayd_version reply, msg;
+	bool compatible = true;
 
 	conn->version_check_done = 1;
 
@@ -1603,9 +1618,7 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr,
 	if (reply.major != be32toh(msg.major)) {
 		DBG("Incompatible major versions (%u vs %u), deleting session",
 				reply.major, be32toh(msg.major));
-		connection_put(conn);
-		ret = 0;
-		goto end;
+		compatible = false;
 	}
 
 	conn->major = reply.major;
@@ -1624,6 +1637,11 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr,
 		ERR("Relay sending version");
 	}
 
+	if (!compatible) {
+		ret = -1;
+		goto end;
+	}
+
 	DBG("Version check done using protocol %u.%u", conn->major,
 			conn->minor);
 
@@ -2027,8 +2045,13 @@ static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr,
 		/* no flush. */
 		ret = 0;
 	} else {
+		/*
+		 * ret < 0
+		 *
+		 * relay_index_try_flush is responsible for the self-reference
+		 * put of the index object on error.
+		 */
 		ERR("relay_index_try_flush error %d", ret);
-		relay_index_put(index);
 		ret = -1;
 	}
 
@@ -2228,9 +2251,13 @@ static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num,
 		/* No flush. */
 		ret = 0;
 	} else {
-		/* Put self-ref for this index due to error. */
-		relay_index_put(index);
-		index = NULL;
+		/*
+		 * ret < 0
+		 *
+		 * relay_index_try_flush is responsible for the self-reference
+		 * put of the index object on error.
+		 */
+		ERR("relay_index_try_flush error %d", ret);
 		ret = -1;
 	}
 end:
@@ -2680,6 +2707,11 @@ error:
 			destroy_conn,
 			sock_n.node) {
 		health_code_update();
+
+		if (session_abort(destroy_conn->session)) {
+			assert(0);
+		}
+
 		/*
 		 * No need to grab another ref, because we own
 		 * destroy_conn.
@@ -2932,6 +2964,12 @@ exit_init_data:
 	health_app_destroy(health_relayd);
 exit_health_app_create:
 exit_options:
+	/*
+	 * Wait for all pending call_rcu work to complete before tearing
+	 * down data structures. call_rcu worker may be trying to
+	 * perform lookups in those structures.
+	 */
+	rcu_barrier();
 	relayd_cleanup();
 
 	/* Ensure all prior call_rcu are done. */