Fix: perform local data pending before checking data pending with relayd

[lttng-tools.git] / src / common / consumer / consumer.c

diff --git a/src/common/consumer/consumer.c b/src/common/consumer/consumer.c
index 63a2ab288c865e338582485199a7a268064bb4af..27e96007c920a94ef94aaae8fefa21d7aa771ab8 100644 (file)
--- a/src/common/consumer/consumer.c
+++ b/src/common/consumer/consumer.c
@@ -3673,34 +3673,6 @@ error_nosignal:
        }
 }
 
-/*
- * Try to lock the stream mutex.
- *
- * On success, 1 is returned else 0 indicating that the mutex is NOT lock.
- */
-static int stream_try_lock(struct lttng_consumer_stream *stream)
-{
-       int ret;
-
-       assert(stream);
-
-       /*
-        * Try to lock the stream mutex. On failure, we know that the stream is
-        * being used else where hence there is data still being extracted.
-        */
-       ret = pthread_mutex_trylock(&stream->lock);
-       if (ret) {
-               /* For both EBUSY and EINVAL error, the mutex is NOT locked. */
-               ret = 0;
-               goto end;
-       }
-
-       ret = 1;
-
-end:
-       return ret;
-}
-
 /*
  * Search for a relayd associated to the session id and return the reference.
  *
@@ -3767,30 +3739,11 @@ int consumer_data_pending(uint64_t id)
        /* Ease our life a bit */
        ht = consumer_data.stream_list_ht;
 
-       relayd = find_relayd_by_session_id(id);
-       if (relayd) {
-               /* Send init command for data pending. */
-               pthread_mutex_lock(&relayd->ctrl_sock_mutex);
-               ret = relayd_begin_data_pending(&relayd->control_sock,
-                               relayd->relayd_session_id);
-               pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
-               if (ret < 0) {
-                       /* Communication error thus the relayd so no data pending. */
-                       ERR("Relayd begin data pending failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
-                       lttng_consumer_cleanup_relayd(relayd);
-                       goto data_not_pending;
-               }
-       }
-
        cds_lfht_for_each_entry_duplicate(ht->ht,
                        ht->hash_fct(&id, lttng_ht_seed),
                        ht->match_fct, &id,
                        &iter.iter, stream, node_session_id.node) {
-               /* If this call fails, the stream is being used hence data pending. */
-               ret = stream_try_lock(stream);
-               if (!ret) {
-                       goto data_pending;
-               }
+               pthread_mutex_lock(&stream->lock);
 
                /*
                 * A removed node from the hash table indicates that the stream has
@@ -3808,9 +3761,27 @@ int consumer_data_pending(uint64_t id)
                        }
                }
 
-               /* Relayd check */
-               if (relayd) {
-                       pthread_mutex_lock(&relayd->ctrl_sock_mutex);
+               pthread_mutex_unlock(&stream->lock);
+       }
+
+       relayd = find_relayd_by_session_id(id);
+       if (relayd) {
+               unsigned int is_data_inflight = 0;
+
+               /* Send init command for data pending. */
+               pthread_mutex_lock(&relayd->ctrl_sock_mutex);
+               ret = relayd_begin_data_pending(&relayd->control_sock,
+                               relayd->relayd_session_id);
+               if (ret < 0) {
+                       pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
+                       /* Communication error thus the relayd so no data pending. */
+                       goto data_not_pending;
+               }
+
+               cds_lfht_for_each_entry_duplicate(ht->ht,
+                               ht->hash_fct(&id, lttng_ht_seed),
+                               ht->match_fct, &id,
+                               &iter.iter, stream, node_session_id.node) {
                        if (stream->metadata_flag) {
                                ret = relayd_quiescent_control(&relayd->control_sock,
                                                stream->relayd_stream_id);
@@ -3819,27 +3790,19 @@ int consumer_data_pending(uint64_t id)
                                                stream->relayd_stream_id,
                                                stream->next_net_seq_num - 1);
                        }
-                       if (ret < 0) {
+
+                       if (ret == 1) {
+                               pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
+                               goto data_pending;
+                       } else if (ret < 0) {
                                ERR("Relayd data pending failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
                                lttng_consumer_cleanup_relayd(relayd);
                                pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
-                               pthread_mutex_unlock(&stream->lock);
                                goto data_not_pending;
                        }
-                       pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
-                       if (ret == 1) {
-                               pthread_mutex_unlock(&stream->lock);
-                               goto data_pending;
-                       }
                }
-               pthread_mutex_unlock(&stream->lock);
-       }
 
-       if (relayd) {
-               unsigned int is_data_inflight = 0;
-
-               /* Send init command for data pending. */
-               pthread_mutex_lock(&relayd->ctrl_sock_mutex);
+               /* Send end command for data pending. */
                ret = relayd_end_data_pending(&relayd->control_sock,
                                relayd->relayd_session_id, &is_data_inflight);
                pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
@@ -4216,6 +4179,10 @@ int rotate_relay_stream(struct lttng_consumer_local_data *ctx,
                        stream->chan->current_chunk_id,
                        stream->last_sequence_number);
        pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
+       if (ret < 0) {
+               ERR("Relayd rotate stream failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
+               lttng_consumer_cleanup_relayd(relayd);
+       }
        if (ret) {
                ERR("Rotate relay stream");
        }
@@ -4389,6 +4356,10 @@ int rotate_rename_relay(const char *old_path, const char *new_path,
 
        pthread_mutex_lock(&relayd->ctrl_sock_mutex);
        ret = relayd_rotate_rename(&relayd->control_sock, old_path, new_path);
+       if (ret < 0) {
+               ERR("Relayd rotate rename failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
+               lttng_consumer_cleanup_relayd(relayd);
+       }
        pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
 end:
        return ret;
@@ -4419,6 +4390,10 @@ int lttng_consumer_rotate_pending_relay(uint64_t session_id,
 
        pthread_mutex_lock(&relayd->ctrl_sock_mutex);
        ret = relayd_rotate_pending(&relayd->control_sock, chunk_id);
+       if (ret < 0) {
+               ERR("Relayd rotate pending failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
+               lttng_consumer_cleanup_relayd(relayd);
+       }
        pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
 
 end:
@@ -4456,6 +4431,10 @@ int mkdir_relay(const char *path, uint64_t relayd_id)
 
        pthread_mutex_lock(&relayd->ctrl_sock_mutex);
        ret = relayd_mkdir(&relayd->control_sock, path);
+       if (ret < 0) {
+               ERR("Relayd mkdir failed. Cleaning up relayd %" PRIu64".", relayd->net_seq_idx);
+               lttng_consumer_cleanup_relayd(relayd);
+       }
        pthread_mutex_unlock(&relayd->ctrl_sock_mutex);
 
 end: