Fix: consumerd errors on exit

- The LTTNG_CONSUMER_STOP command is not used by sessiond anymore, so
  deprecate it,
- Ensure that the normal teardown (with signal handler) does not trigger
  a health error in consumer daemon by distinguishing between errors and
  "should exit" with different lttng_consumer_poll_socket() return values.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

diff --git a/src/common/consumer.c b/src/common/consumer.c
index c842de9358965c79e31f489b789906f3d1deddec..ede214c8025b264fcdf5a538a565b2ab5b710eed 100644 (file)
--- a/src/common/consumer.c
+++ b/src/common/consumer.c
@@ -1095,8 +1095,8 @@ static int update_poll_array(struct lttng_consumer_local_data *ctx,
 }
 
 /*
- * Poll on the should_quit pipe and the command socket return -1 on error and
- * should exit, 0 if data is available on the command socket
+ * Poll on the should_quit pipe and the command socket return -1 on
+ * error, 1 if should exit, 0 if data is available on the command socket
  */
 int lttng_consumer_poll_socket(struct pollfd *consumer_sockpoll)
 {
@@ -1112,16 +1112,13 @@ restart:
                        goto restart;
                }
                PERROR("Poll error");
-               goto exit;
+               return -1;
        }
        if (consumer_sockpoll[0].revents & (POLLIN | POLLPRI)) {
                DBG("consumer_should_quit wake up");
-               goto exit;
+               return 1;
        }
        return 0;
-
-exit:
-       return -1;
 }
 
 /*
@@ -2944,8 +2941,8 @@ static int set_metadata_socket(struct lttng_consumer_local_data *ctx,
        assert(ctx);
        assert(sockpoll);
 
-       if (lttng_consumer_poll_socket(sockpoll) < 0) {
-               ret = -1;
+       ret = lttng_consumer_poll_socket(sockpoll);
+       if (ret) {
                goto error;
        }
        DBG("Metadata connection on client_socket");
@@ -3014,7 +3011,12 @@ void *consumer_thread_sessiond_poll(void *data)
        consumer_sockpoll[1].fd = client_socket;
        consumer_sockpoll[1].events = POLLIN | POLLPRI;
 
-       if (lttng_consumer_poll_socket(consumer_sockpoll) < 0) {
+       ret = lttng_consumer_poll_socket(consumer_sockpoll);
+       if (ret) {
+               if (ret > 0) {
+                       /* should exit */
+                       err = 0;
+               }
                goto end;
        }
        DBG("Connection on client_socket");
@@ -3031,7 +3033,11 @@ void *consumer_thread_sessiond_poll(void *data)
         * command unix socket.
         */
        ret = set_metadata_socket(ctx, consumer_sockpoll, client_socket);
-       if (ret < 0) {
+       if (ret) {
+               if (ret > 0) {
+                       /* should exit */
+                       err = 0;
+               }
                goto end;
        }
 
@@ -3052,15 +3058,15 @@ void *consumer_thread_sessiond_poll(void *data)
                health_poll_entry();
                ret = lttng_consumer_poll_socket(consumer_sockpoll);
                health_poll_exit();
-               if (ret < 0) {
+               if (ret) {
+                       if (ret > 0) {
+                               /* should exit */
+                               err = 0;
+                       }
                        goto end;
                }
                DBG("Incoming command on sock");
                ret = lttng_consumer_recv_cmd(ctx, sock, consumer_sockpoll);
-               if (ret == -ENOENT) {
-                       DBG("Received STOP command");
-                       goto end;
-               }
                if (ret <= 0) {
                        /*
                         * This could simply be a session daemon quitting. Don't output
@@ -3278,7 +3284,9 @@ int consumer_add_relayd_socket(uint64_t net_seq_idx, int sock_type,
        }
 
        /* Poll on consumer socket. */
-       if (lttng_consumer_poll_socket(consumer_sockpoll) < 0) {
+       ret = lttng_consumer_poll_socket(consumer_sockpoll);
+       if (ret) {
+               /* Needing to exit in the middle of a command: error. */
                lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_POLL_ERROR);
                ret = -EINTR;
                goto error_nosignal;

diff --git a/src/common/consumer.h b/src/common/consumer.h
index 84ef2719d47e55eb7abc8626cdf37b3a11da525c..7485e65b259d4b855c301405425225a462b5e0bf 100644 (file)
--- a/src/common/consumer.h
+++ b/src/common/consumer.h
@@ -41,7 +41,7 @@ enum lttng_consumer_command {
        /* pause, delete, active depending on fd state */
        LTTNG_CONSUMER_UPDATE_STREAM,
        /* inform the consumer to quit when all fd has hang up */
-       LTTNG_CONSUMER_STOP,
+       LTTNG_CONSUMER_STOP,    /* deprecated */
        LTTNG_CONSUMER_ADD_RELAYD_SOCKET,
        /* Inform the consumer to kill a specific relayd connection */
        LTTNG_CONSUMER_DESTROY_RELAYD,

diff --git a/src/common/kernel-consumer/kernel-consumer.c b/src/common/kernel-consumer/kernel-consumer.c
index 061123d61160d32016249aa9856b8f0e4a43b5c6..ab4f604d7a5c525447cdf23453633ef0baa4318a 100644 (file)
--- a/src/common/kernel-consumer/kernel-consumer.c
+++ b/src/common/kernel-consumer/kernel-consumer.c
@@ -441,17 +441,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
        health_code_update();
 
-       if (msg.cmd_type == LTTNG_CONSUMER_STOP) {
-               /*
-                * Notify the session daemon that the command is completed.
-                *
-                * On transport layer error, the function call will print an error
-                * message so handling the returned code is a bit useless since we
-                * return an error code anyway.
-                */
-               (void) consumer_send_status_msg(sock, ret_code);
-               return -ENOENT;
-       }
+       /* Deprecated command */
+       assert(msg.cmd_type != LTTNG_CONSUMER_STOP);
 
        health_code_update();
 
@@ -593,9 +584,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                health_poll_entry();
                ret = lttng_consumer_poll_socket(consumer_sockpoll);
                health_poll_exit();
-               if (ret < 0) {
-                       rcu_read_unlock();
-                       return -EINTR;
+               if (ret) {
+                       goto error_fatal;
                }
 
                health_code_update();

diff --git a/src/common/ust-consumer/ust-consumer.c b/src/common/ust-consumer/ust-consumer.c
index 07073260f3262bb3cbd2207a846e3404fcf0713c..0955e66fa4e384fdde2f70703a96e626b8ebfc8c 100644 (file)
--- a/src/common/ust-consumer/ust-consumer.c
+++ b/src/common/ust-consumer/ust-consumer.c
@@ -1116,17 +1116,8 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
        health_code_update();
 
-       if (msg.cmd_type == LTTNG_CONSUMER_STOP) {
-               /*
-                * Notify the session daemon that the command is completed.
-                *
-                * On transport layer error, the function call will print an error
-                * message so handling the returned code is a bit useless since we
-                * return an error code anyway.
-                */
-               (void) consumer_send_status_msg(sock, ret_code);
-               return -ENOENT;
-       }
+       /* deprecated */
+       assert(msg.cmd_type != LTTNG_CONSUMER_STOP);
 
        health_code_update();
 
@@ -1456,7 +1447,7 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                health_poll_entry();
                ret = lttng_consumer_poll_socket(consumer_sockpoll);
                health_poll_exit();
-               if (ret < 0) {
+               if (ret) {
                        goto error_fatal;
                }