Fix: relayd: live: erroneous message timestamp observed from live viewer

[lttng-tools.git] / src / bin / lttng-relayd / live.c

diff --git a/src/bin/lttng-relayd/live.c b/src/bin/lttng-relayd/live.c
index f0151331ae2fd707bd354c85b88d5480aff040f4..11e2f799dda14b2e364fffcf701a42a519a4f920 100644 (file)
--- a/src/bin/lttng-relayd/live.c
+++ b/src/bin/lttng-relayd/live.c
@@ -1165,6 +1165,16 @@ int viewer_get_new_streams(struct relay_connection *conn)
         * the viewer's point of view.
         */
        pthread_mutex_lock(&session->lock);
+       /*
+        * If a session rotation is ongoing, do not attempt to open any
+        * stream, because the chunk can be in an intermediate state
+        * due to directory renaming.
+        */
+       if (session->ongoing_rotation) {
+               DBG("Relay session %" PRIu64 " rotation ongoing", session_id);
+               response.status = htobe32(LTTNG_VIEWER_NEW_STREAMS_NO_NEW);
+               goto send_reply_unlock;
+       }
        ret = make_viewer_streams(session,
                        conn->viewer_session,
                        LTTNG_VIEWER_SEEK_BEGINNING, &nb_total, &nb_unsent,
@@ -1305,6 +1315,17 @@ int viewer_attach_session(struct relay_connection *conn)
                goto send_reply;
        }
 
+       /*
+        * If a session rotation is ongoing, do not attempt to open any
+        * stream, because the chunk can be in an intermediate state
+        * due to directory renaming.
+        */
+       if (session->ongoing_rotation) {
+               DBG("Relay session %" PRIu64 " rotation ongoing", session_id);
+               send_streams = 0;
+               goto send_reply;
+       }
+
        ret = make_viewer_streams(session,
                        conn->viewer_session, seek_type,
                        &nb_streams, NULL, NULL, &closed);
@@ -1609,6 +1630,13 @@ int viewer_get_next_index(struct relay_connection *conn)
        metadata_viewer_stream =
                        ctf_trace_get_viewer_metadata_stream(ctf_trace);
 
+       /*
+        * Hold the session lock to protect against concurrent changes
+        * to the chunk files (e.g. rename done by clear), which are
+        * protected by the session ongoing rotation state. Those are
+        * synchronized with the session lock.
+        */
+       pthread_mutex_lock(&rstream->trace->session->lock);
        pthread_mutex_lock(&rstream->lock);
 
        /*
@@ -1775,6 +1803,7 @@ int viewer_get_next_index(struct relay_connection *conn)
 send_reply:
        if (rstream) {
                pthread_mutex_unlock(&rstream->lock);
+               pthread_mutex_unlock(&rstream->trace->session->lock);
        }
 
        if (metadata_viewer_stream) {
@@ -1816,6 +1845,7 @@ end:
 
 error_put:
        pthread_mutex_unlock(&rstream->lock);
+       pthread_mutex_unlock(&rstream->trace->session->lock);
        if (metadata_viewer_stream) {
                viewer_stream_put(metadata_viewer_stream);
        }