Fix: grab more than one packet for snapshots

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

diff --git a/src/common/kernel-consumer/kernel-consumer.c b/src/common/kernel-consumer/kernel-consumer.c
index c95355e9a472e35970ab7b734b18611d14bc356c..bca9ad2b9b867ab59acd5842eb1421eddced7f66 100644 (file)
--- a/src/common/kernel-consumer/kernel-consumer.c
+++ b/src/common/kernel-consumer/kernel-consumer.c
@@ -113,7 +113,7 @@ int lttng_kconsumer_get_consumed_snapshot(struct lttng_consumer_stream *stream,
  * Returns 0 on success, < 0 on error
  */
 int lttng_kconsumer_snapshot_channel(uint64_t key, char *path,
-               uint64_t relayd_id, uint64_t max_stream_size,
+               uint64_t relayd_id, uint64_t nb_packets_per_stream,
                struct lttng_consumer_local_data *ctx)
 {
        int ret;
@@ -176,10 +176,12 @@ int lttng_kconsumer_snapshot_channel(uint64_t key, char *path,
                        DBG("Kernel consumer snapshot stream %s/%s (%" PRIu64 ")",
                                        path, stream->name, stream->key);
                }
-               ret = consumer_send_relayd_streams_sent(relayd_id);
-               if (ret < 0) {
-                       ERR("sending streams sent to relayd");
-                       goto end_unlock;
+               if (relayd_id != -1ULL) {
+                       ret = consumer_send_relayd_streams_sent(relayd_id);
+                       if (ret < 0) {
+                               ERR("sending streams sent to relayd");
+                               goto end_unlock;
+                       }
                }
 
                ret = kernctl_buffer_flush(stream->wait_fd);
@@ -217,14 +219,9 @@ int lttng_kconsumer_snapshot_channel(uint64_t key, char *path,
                        }
                }
 
-               /*
-                * 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
-                * 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;
@@ -439,17 +436,8 @@ 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();
 
@@ -568,7 +556,7 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                         * happens while tearing down.
                         */
                        ERR("Unable to find channel key %" PRIu64, msg.u.stream.channel_key);
-                       ret_code = LTTNG_ERR_KERN_CHAN_NOT_FOUND;
+                       ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                }
 
                health_code_update();
@@ -591,9 +579,8 @@ 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();
@@ -649,6 +636,10 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                switch (channel->output) {
                case CONSUMER_CHANNEL_SPLICE:
                        new_stream->output = LTTNG_EVENT_SPLICE;
+                       ret = utils_create_pipe(new_stream->splice_pipe);
+                       if (ret < 0) {
+                               goto end_nosignal;
+                       }
                        break;
                case CONSUMER_CHANNEL_MMAP:
                        new_stream->output = LTTNG_EVENT_MMAP;
@@ -772,7 +763,7 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                         */
                        ERR("Unable to find channel key %" PRIu64,
                                        msg.u.sent_streams.channel_key);
-                       ret_code = LTTNG_ERR_KERN_CHAN_NOT_FOUND;
+                       ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                }
 
                health_code_update();
@@ -781,7 +772,7 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                 * Send status code to session daemon.
                 */
                ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               if (ret < 0 || ret_code != LTTCOMM_CONSUMERD_SUCCESS) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto end_nosignal;
                }
@@ -823,7 +814,7 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                relayd = consumer_find_relayd(index);
                if (relayd == NULL) {
                        DBG("Unable to find relayd %" PRIu64, index);
-                       ret_code = LTTNG_ERR_NO_CONSUMER;
+                       ret_code = LTTCOMM_CONSUMERD_RELAYD_FAIL;
                }
 
                /*
@@ -882,17 +873,17 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                                        msg.u.snapshot_channel.relayd_id, ctx);
                        if (ret < 0) {
                                ERR("Snapshot metadata failed");
-                               ret_code = LTTNG_ERR_KERN_META_FAIL;
+                               ret_code = LTTCOMM_CONSUMERD_ERROR_METADATA;
                        }
                } else {
                        ret = lttng_kconsumer_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");
-                               ret_code = LTTNG_ERR_KERN_CHAN_FAIL;
+                               ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                        }
                }
 
@@ -913,7 +904,7 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                channel = consumer_find_channel(key);
                if (!channel) {
                        ERR("Kernel consumer destroy channel %" PRIu64 " not found", key);
-                       ret_code = LTTNG_ERR_KERN_CHAN_NOT_FOUND;
+                       ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                }
 
                health_code_update();
@@ -926,6 +917,11 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                health_code_update();
 
+               /* Stop right now if no channel was found. */
+               if (!channel) {
+                       goto end_nosignal;
+               }
+
                /*
                 * This command should ONLY be issued for channel with streams set in
                 * no monitor mode.
@@ -1152,9 +1148,10 @@ ssize_t lttng_kconsumer_read_subbuffer(struct lttng_consumer_stream *stream,
                                (ret != len && stream->net_seq_idx == (uint64_t) -1ULL)) {
                        /*
                         * Display the error but continue processing to try to release the
-                        * subbuffer
+                        * subbuffer. This is a DBG statement since this is possible to
+                        * happen without being a critical error.
                         */
-                       ERR("Error writing to tracefile "
+                       DBG("Error writing to tracefile "
                                        "(ret: %zd != len: %lu != subbuf_size: %lu)",
                                        ret, len, subbuf_size);
                        write_index = 0;