Fix: free ust channel object after grace period in consumer
[lttng-tools.git] / src / common / ust-consumer / ust-consumer.c
index bdace6586881c3eb76a0505fd9ea553b36bbd012..152455ca64fa7e938834c01d697bb3b97e6198a8 100644 (file)
@@ -79,6 +79,7 @@ static void destroy_channel(struct lttng_consumer_channel *channel)
         */
        if (channel->uchan) {
                lttng_ustconsumer_del_channel(channel);
+               lttng_ustconsumer_free_channel(channel);
        }
        free(channel);
 }
@@ -327,8 +328,12 @@ static int create_ust_streams(struct lttng_consumer_channel *channel,
                /* Keep stream reference when creating metadata. */
                if (channel->type == CONSUMER_CHANNEL_TYPE_METADATA) {
                        channel->metadata_stream = stream;
-                       stream->ust_metadata_poll_pipe[0] = ust_metadata_pipe[0];
-                       stream->ust_metadata_poll_pipe[1] = ust_metadata_pipe[1];
+                       if (channel->monitor) {
+                               /* Set metadata poll pipe if we created one */
+                               memcpy(stream->ust_metadata_poll_pipe,
+                                               ust_metadata_pipe,
+                                               sizeof(ust_metadata_pipe));
+                       }
                }
        }
 
@@ -859,7 +864,7 @@ error:
  * Returns 0 on success, < 0 on error
  */
 static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id,
-               uint64_t max_stream_size, struct lttng_consumer_local_data *ctx)
+               uint64_t nb_packets_per_stream, struct lttng_consumer_local_data *ctx)
 {
        int ret;
        unsigned use_relayd = 0;
@@ -941,12 +946,13 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id,
 
                /*
                 * The original value is sent back if max stream size is larger than
-                * the possible size of the snapshot. Also, we asume that the session
+                * the possible size of the snapshot. Also, we assume that the session
                 * daemon should never send a maximum stream size that is lower than
                 * subbuffer size.
                 */
-               consumed_pos = consumer_get_consumed_maxsize(consumed_pos,
-                               produced_pos, max_stream_size);
+               consumed_pos = consumer_get_consume_start_pos(consumed_pos,
+                               produced_pos, nb_packets_per_stream,
+                               stream->max_sb_size);
 
                while (consumed_pos < produced_pos) {
                        ssize_t read_len;
@@ -1023,7 +1029,12 @@ error:
 }
 
 /*
- * Receive the metadata updates from the sessiond.
+ * Receive the metadata updates from the sessiond. Supports receiving
+ * overlapping metadata, but is needs to always belong to a contiguous
+ * range starting from 0.
+ * Be careful about the locks held when calling this function: it needs
+ * the metadata cache flush to concurrently progress in order to
+ * complete.
  */
 int lttng_ustconsumer_recv_metadata(int sock, uint64_t key, uint64_t offset,
                uint64_t len, struct lttng_consumer_channel *channel,
@@ -1435,6 +1446,15 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                health_code_update();
 
+               if (!len) {
+                       /*
+                        * There is nothing to receive. We have simply
+                        * checked whether the channel can be found.
+                        */
+                       ret_code = LTTCOMM_CONSUMERD_SUCCESS;
+                       goto end_msg_sessiond;
+               }
+
                /* Tell session daemon we are ready to receive the metadata. */
                ret = consumer_send_status_msg(sock, LTTCOMM_CONSUMERD_SUCCESS);
                if (ret < 0) {
@@ -1489,7 +1509,7 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                        ret = snapshot_channel(msg.u.snapshot_channel.key,
                                        msg.u.snapshot_channel.pathname,
                                        msg.u.snapshot_channel.relayd_id,
-                                       msg.u.snapshot_channel.max_stream_size,
+                                       msg.u.snapshot_channel.nb_packets_per_stream,
                                        ctx);
                        if (ret < 0) {
                                ERR("Snapshot channel failed");
@@ -1668,6 +1688,13 @@ void lttng_ustconsumer_del_channel(struct lttng_consumer_channel *chan)
        if (chan->switch_timer_enabled == 1) {
                consumer_timer_switch_stop(chan);
        }
+}
+
+void lttng_ustconsumer_free_channel(struct lttng_consumer_channel *chan)
+{
+       assert(chan);
+       assert(chan->uchan);
+
        consumer_metadata_cache_destroy(chan);
        ustctl_destroy_channel(chan->uchan);
 }
@@ -1768,7 +1795,7 @@ int commit_one_metadata_packet(struct lttng_consumer_stream *stream)
        int ret;
 
        pthread_mutex_lock(&stream->chan->metadata_cache->lock);
-       if (stream->chan->metadata_cache->contiguous
+       if (stream->chan->metadata_cache->max_offset
                        == stream->ust_metadata_pushed) {
                ret = 0;
                goto end;
@@ -1776,7 +1803,7 @@ int commit_one_metadata_packet(struct lttng_consumer_stream *stream)
 
        write_len = ustctl_write_one_packet_to_channel(stream->chan->uchan,
                        &stream->chan->metadata_cache->data[stream->ust_metadata_pushed],
-                       stream->chan->metadata_cache->contiguous
+                       stream->chan->metadata_cache->max_offset
                        - stream->ust_metadata_pushed);
        assert(write_len != 0);
        if (write_len < 0) {
@@ -1786,7 +1813,7 @@ int commit_one_metadata_packet(struct lttng_consumer_stream *stream)
        }
        stream->ust_metadata_pushed += write_len;
 
-       assert(stream->chan->metadata_cache->contiguous >=
+       assert(stream->chan->metadata_cache->max_offset >=
                        stream->ust_metadata_pushed);
        ret = write_len;
 
@@ -1800,7 +1827,9 @@ end:
  * Sync metadata meaning request them to the session daemon and snapshot to the
  * metadata thread can consumer them.
  *
- * Metadata stream lock MUST be acquired.
+ * Metadata stream lock is held here, but we need to release it when
+ * interacting with sessiond, else we cause a deadlock with live
+ * awaiting on metadata to be pushed out.
  *
  * Return 0 if new metadatda is available, EAGAIN if the metadata stream
  * is empty or a negative value on error.
@@ -1814,6 +1843,7 @@ int lttng_ustconsumer_sync_metadata(struct lttng_consumer_local_data *ctx,
        assert(ctx);
        assert(metadata);
 
+       pthread_mutex_unlock(&metadata->lock);
        /*
         * Request metadata from the sessiond, but don't wait for the flush
         * because we locked the metadata thread.
@@ -1822,6 +1852,7 @@ int lttng_ustconsumer_sync_metadata(struct lttng_consumer_local_data *ctx,
        if (ret < 0) {
                goto end;
        }
+       pthread_mutex_lock(&metadata->lock);
 
        ret = commit_one_metadata_packet(metadata);
        if (ret <= 0) {
@@ -2048,7 +2079,23 @@ retry:
                /*
                 * In live, block until all the metadata is sent.
                 */
+               pthread_mutex_lock(&stream->metadata_timer_lock);
+               assert(!stream->missed_metadata_flush);
+               stream->waiting_on_metadata = true;
+               pthread_mutex_unlock(&stream->metadata_timer_lock);
+
                err = consumer_stream_sync_metadata(ctx, stream->session_id);
+
+               pthread_mutex_lock(&stream->metadata_timer_lock);
+               stream->waiting_on_metadata = false;
+               if (stream->missed_metadata_flush) {
+                       stream->missed_metadata_flush = false;
+                       pthread_mutex_unlock(&stream->metadata_timer_lock);
+                       (void) consumer_flush_ust_index(stream);
+               } else {
+                       pthread_mutex_unlock(&stream->metadata_timer_lock);
+               }
+
                if (err < 0) {
                        goto end;
                }
@@ -2129,7 +2176,7 @@ int lttng_ustconsumer_data_pending(struct lttng_consumer_stream *stream)
                uint64_t contiguous, pushed;
 
                /* Ease our life a bit. */
-               contiguous = stream->chan->metadata_cache->contiguous;
+               contiguous = stream->chan->metadata_cache->max_offset;
                pushed = stream->ust_metadata_pushed;
 
                /*
@@ -2258,6 +2305,10 @@ void lttng_ustconsumer_close_stream_wakeup(struct lttng_consumer_stream *stream)
  * function or any of its callees. Timers have a very strict locking
  * semantic with respect to teardown. Failure to respect this semantic
  * introduces deadlocks.
+ *
+ * DON'T hold the metadata lock when calling this function, else this
+ * can cause deadlock involving consumer awaiting for metadata to be
+ * pushed out due to concurrent interaction with the session daemon.
  */
 int lttng_ustconsumer_request_metadata(struct lttng_consumer_local_data *ctx,
                struct lttng_consumer_channel *channel, int timer, int wait)
This page took 0.02701 seconds and 4 git commands to generate.