Fix: consumer: snapshot: assertion on subsequent snapshot
[lttng-tools.git] / src / common / kernel-consumer / kernel-consumer.c
index dd5cf1761a4992c9e7384d48675038c5438ece0e..9db9d085001f5ca1c6979bf58125301c255f4266 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011 Julien Desfossez <julien.desfossez@polymtl.ca>
+ * Copyright (C) 2011 EfficiOS Inc.
  * Copyright (C) 2011 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
  * Copyright (C) 2017 Jérémie Galarneau <jeremie.galarneau@efficios.com>
  *
@@ -41,7 +41,7 @@
 
 #include "kernel-consumer.h"
 
-extern struct lttng_consumer_global_data consumer_data;
+extern struct lttng_consumer_global_data the_consumer_data;
 extern int consumer_poll_timeout;
 
 /*
@@ -136,10 +136,34 @@ error:
        return ret;
 }
 
+static void finalize_snapshot_stream(
+               struct lttng_consumer_stream *stream, uint64_t relayd_id)
+{
+       ASSERT_LOCKED(stream->lock);
+
+       if (relayd_id == (uint64_t) -1ULL) {
+               if (stream->out_fd >= 0) {
+                       const int ret = close(stream->out_fd);
+
+                       if (ret < 0) {
+                               PERROR("Failed to close stream snapshot output file descriptor");
+                       }
+
+                       stream->out_fd = -1;
+               }
+       } else {
+               close_relayd_stream(stream);
+               stream->net_seq_idx = (uint64_t) -1ULL;
+       }
+
+       lttng_trace_chunk_put(stream->trace_chunk);
+       stream->trace_chunk = NULL;
+}
+
 /*
  * Take a snapshot of all the stream of a channel
  * RCU read-side lock must be held across this function to ensure existence of
- * channel. The channel lock must be held by the caller.
+ * channel.
  *
  * Returns 0 on success, < 0 on error
  */
@@ -154,6 +178,9 @@ static int lttng_kconsumer_snapshot_channel(
 
        DBG("Kernel consumer snapshot channel %" PRIu64, key);
 
+       /* Prevent channel modifications while we perform the snapshot.*/
+       pthread_mutex_lock(&channel->lock);
+
        rcu_read_lock();
 
        /* Splice is not supported yet for channel snapshot. */
@@ -197,13 +224,13 @@ static int lttng_kconsumer_snapshot_channel(
                        ret = consumer_send_relayd_stream(stream, path);
                        if (ret < 0) {
                                ERR("sending stream to relayd");
-                               goto end_unlock;
+                               goto error_finalize_stream;
                        }
                } else {
                        ret = consumer_stream_create_output_files(stream,
                                        false);
                        if (ret < 0) {
-                               goto end_unlock;
+                               goto error_finalize_stream;
                        }
                        DBG("Kernel consumer snapshot stream (%" PRIu64 ")",
                                        stream->key);
@@ -221,7 +248,7 @@ static int lttng_kconsumer_snapshot_channel(
                        ret = kernctl_buffer_flush(stream->wait_fd);
                        if (ret < 0) {
                                ERR("Failed to flush kernel stream");
-                               goto end_unlock;
+                               goto error_finalize_stream;
                        }
                        goto end_unlock;
                }
@@ -229,19 +256,19 @@ static int lttng_kconsumer_snapshot_channel(
                ret = lttng_kconsumer_take_snapshot(stream);
                if (ret < 0) {
                        ERR("Taking kernel snapshot");
-                       goto end_unlock;
+                       goto error_finalize_stream;
                }
 
                ret = lttng_kconsumer_get_produced_snapshot(stream, &produced_pos);
                if (ret < 0) {
                        ERR("Produced kernel snapshot position");
-                       goto end_unlock;
+                       goto error_finalize_stream;
                }
 
                ret = lttng_kconsumer_get_consumed_snapshot(stream, &consumed_pos);
                if (ret < 0) {
                        ERR("Consumerd kernel snapshot position");
-                       goto end_unlock;
+                       goto error_finalize_stream;
                }
 
                consumed_pos = consumer_get_consume_start_pos(consumed_pos,
@@ -261,7 +288,7 @@ static int lttng_kconsumer_snapshot_channel(
                        if (ret < 0) {
                                if (ret != -EAGAIN) {
                                        PERROR("kernctl_get_subbuf snapshot");
-                                       goto end_unlock;
+                                       goto error_finalize_stream;
                                }
                                DBG("Kernel consumer get subbuf failed. Skipping it.");
                                consumed_pos += stream->max_sb_size;
@@ -311,26 +338,12 @@ static int lttng_kconsumer_snapshot_channel(
                        ret = kernctl_put_subbuf(stream->wait_fd);
                        if (ret < 0) {
                                ERR("Snapshot kernctl_put_subbuf");
-                               goto end_unlock;
+                               goto error_finalize_stream;
                        }
                        consumed_pos += stream->max_sb_size;
                }
 
-               if (relayd_id == (uint64_t) -1ULL) {
-                       if (stream->out_fd >= 0) {
-                               ret = close(stream->out_fd);
-                               if (ret < 0) {
-                                       PERROR("Kernel consumer snapshot close out_fd");
-                                       goto end_unlock;
-                               }
-                               stream->out_fd = -1;
-                       }
-               } else {
-                       close_relayd_stream(stream);
-                       stream->net_seq_idx = (uint64_t) -1ULL;
-               }
-               lttng_trace_chunk_put(stream->trace_chunk);
-               stream->trace_chunk = NULL;
+               finalize_snapshot_stream(stream, relayd_id);
                pthread_mutex_unlock(&stream->lock);
        }
 
@@ -343,17 +356,20 @@ error_put_subbuf:
        if (ret < 0) {
                ERR("Snapshot kernctl_put_subbuf error path");
        }
+error_finalize_stream:
+       finalize_snapshot_stream(stream, relayd_id);
 end_unlock:
        pthread_mutex_unlock(&stream->lock);
 end:
        rcu_read_unlock();
+       pthread_mutex_unlock(&channel->lock);
        return ret;
 }
 
 /*
  * Read the whole metadata available for a snapshot.
  * RCU read-side lock must be held across this function to ensure existence of
- * metadata_channel. The channel lock must be held by the caller.
+ * metadata_channel.
  *
  * Returns 0 on success, < 0 on error
  */
@@ -376,7 +392,7 @@ static int lttng_kconsumer_snapshot_metadata(
        metadata_stream = metadata_channel->metadata_stream;
        assert(metadata_stream);
 
-       pthread_mutex_lock(&metadata_stream->lock);
+       metadata_stream->read_subbuffer_ops.lock(metadata_stream);
        assert(metadata_channel->trace_chunk);
        assert(metadata_stream->trace_chunk);
 
@@ -403,16 +419,12 @@ static int lttng_kconsumer_snapshot_metadata(
 
                ret_read = lttng_consumer_read_subbuffer(metadata_stream, ctx, true);
                if (ret_read < 0) {
-                       if (ret_read != -EAGAIN) {
-                               ERR("Kernel snapshot reading metadata subbuffer (ret: %zd)",
-                                               ret_read);
-                               ret = ret_read;
-                               goto error_snapshot;
-                       }
-                       /* ret_read is negative at this point so we will exit the loop. */
-                       continue;
+                       ERR("Kernel snapshot reading metadata subbuffer (ret: %zd)",
+                                       ret_read);
+                       ret = ret_read;
+                       goto error_snapshot;
                }
-       } while (ret_read >= 0);
+       } while (ret_read > 0);
 
        if (use_relayd) {
                close_relayd_stream(metadata_stream);
@@ -435,8 +447,7 @@ static int lttng_kconsumer_snapshot_metadata(
 
        ret = 0;
 error_snapshot:
-       pthread_mutex_unlock(&metadata_stream->lock);
-       cds_list_del(&metadata_stream->send_node);
+       metadata_stream->read_subbuffer_ops.unlock(metadata_stream);
        consumer_stream_destroy(metadata_stream, NULL);
        metadata_channel->metadata_stream = NULL;
        rcu_read_unlock();
@@ -451,19 +462,24 @@ error_snapshot:
 int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                int sock, struct pollfd *consumer_sockpoll)
 {
-       ssize_t ret;
+       int ret_func;
        enum lttcomm_return_code ret_code = LTTCOMM_CONSUMERD_SUCCESS;
        struct lttcomm_consumer_msg msg;
 
        health_code_update();
 
-       ret = lttcomm_recv_unix_sock(sock, &msg, sizeof(msg));
-       if (ret != sizeof(msg)) {
-               if (ret > 0) {
-                       lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_ERROR_RECV_CMD);
-                       ret = -1;
+       {
+               ssize_t ret_recv;
+
+               ret_recv = lttcomm_recv_unix_sock(sock, &msg, sizeof(msg));
+               if (ret_recv != sizeof(msg)) {
+                       if (ret_recv > 0) {
+                               lttng_consumer_send_error(ctx,
+                                               LTTCOMM_CONSUMERD_ERROR_RECV_CMD);
+                               ret_recv = -1;
+                       }
+                       return ret_recv;
                }
-               return ret;
        }
 
        health_code_update();
@@ -479,24 +495,30 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
        switch (msg.cmd_type) {
        case LTTNG_CONSUMER_ADD_RELAYD_SOCKET:
        {
+               uint32_t major = msg.u.relayd_sock.major;
+               uint32_t minor = msg.u.relayd_sock.minor;
+               enum lttcomm_sock_proto protocol = (enum lttcomm_sock_proto)
+                               msg.u.relayd_sock.relayd_socket_protocol;
+
                /* Session daemon status message are handled in the following call. */
                consumer_add_relayd_socket(msg.u.relayd_sock.net_index,
-                               msg.u.relayd_sock.type, ctx, sock, consumer_sockpoll,
-                               &msg.u.relayd_sock.sock, msg.u.relayd_sock.session_id,
-                               msg.u.relayd_sock.relayd_session_id);
+                               msg.u.relayd_sock.type, ctx, sock,
+                               consumer_sockpoll, msg.u.relayd_sock.session_id,
+                               msg.u.relayd_sock.relayd_session_id, major,
+                               minor, protocol);
                goto end_nosignal;
        }
        case LTTNG_CONSUMER_ADD_CHANNEL:
        {
                struct lttng_consumer_channel *new_channel;
-               int ret_recv;
+               int ret_send_status, ret_add_channel = 0;
                const uint64_t chunk_id = msg.u.channel.chunk_id.value;
 
                health_code_update();
 
                /* First send a status message before receiving the fds. */
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_fatal;
                }
@@ -548,16 +570,20 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                health_code_update();
 
                if (ctx->on_recv_channel != NULL) {
-                       ret_recv = ctx->on_recv_channel(new_channel);
-                       if (ret_recv == 0) {
-                               ret = consumer_add_channel(new_channel, ctx);
-                       } else if (ret_recv < 0) {
+                       int ret_recv_channel =
+                                       ctx->on_recv_channel(new_channel);
+                       if (ret_recv_channel == 0) {
+                               ret_add_channel = consumer_add_channel(
+                                               new_channel, ctx);
+                       } else if (ret_recv_channel < 0) {
                                goto end_nosignal;
                        }
                } else {
-                       ret = consumer_add_channel(new_channel, ctx);
+                       ret_add_channel =
+                                       consumer_add_channel(new_channel, ctx);
                }
-               if (msg.u.channel.type == CONSUMER_CHANNEL_TYPE_DATA && !ret) {
+               if (msg.u.channel.type == CONSUMER_CHANNEL_TYPE_DATA &&
+                               !ret_add_channel) {
                        int monitor_start_ret;
 
                        DBG("Consumer starting monitor timer");
@@ -570,15 +596,15 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                                ERR("Starting channel monitoring timer failed");
                                goto end_nosignal;
                        }
-
                }
 
                health_code_update();
 
                /* If we received an error in add_channel, we need to report it. */
-               if (ret < 0) {
-                       ret = consumer_send_status_msg(sock, ret);
-                       if (ret < 0) {
+               if (ret_add_channel < 0) {
+                       ret_send_status = consumer_send_status_msg(
+                                       sock, ret_add_channel);
+                       if (ret_send_status < 0) {
                                goto error_fatal;
                        }
                        goto end_nosignal;
@@ -593,6 +619,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                struct lttng_consumer_stream *new_stream;
                struct lttng_consumer_channel *channel;
                int alloc_ret = 0;
+               int ret_send_status, ret_poll, ret_get_max_subbuf_size;
+               ssize_t ret_pipe_write, ret_recv;
 
                /*
                 * Get stream's channel reference. Needed when adding the stream to the
@@ -611,8 +639,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                health_code_update();
 
                /* First send a status message before receiving the fds. */
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_add_stream_fatal;
                }
@@ -626,18 +654,19 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                /* Blocking call */
                health_poll_entry();
-               ret = lttng_consumer_poll_socket(consumer_sockpoll);
+               ret_poll = lttng_consumer_poll_socket(consumer_sockpoll);
                health_poll_exit();
-               if (ret) {
+               if (ret_poll) {
                        goto error_add_stream_fatal;
                }
 
                health_code_update();
 
                /* Get stream file descriptor from socket */
-               ret = lttcomm_recv_fds_unix_sock(sock, &fd, 1);
-               if (ret != sizeof(fd)) {
+               ret_recv = lttcomm_recv_fds_unix_sock(sock, &fd, 1);
+               if (ret_recv != sizeof(fd)) {
                        lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_ERROR_RECV_FD);
+                       ret_func = ret_recv;
                        goto end;
                }
 
@@ -648,8 +677,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                 * above recv() failed, the session daemon is notified through the
                 * error socket and the teardown is eventually done.
                 */
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_add_stream_nosignal;
                }
@@ -682,9 +711,9 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                }
 
                new_stream->wait_fd = fd;
-               ret = kernctl_get_max_subbuf_size(new_stream->wait_fd,
-                               &new_stream->max_sb_size);
-               if (ret < 0) {
+               ret_get_max_subbuf_size = kernctl_get_max_subbuf_size(
+                               new_stream->wait_fd, &new_stream->max_sb_size);
+               if (ret_get_max_subbuf_size < 0) {
                        pthread_mutex_unlock(&channel->lock);
                        ERR("Failed to get kernel maximal subbuffer size");
                        goto error_add_stream_nosignal;
@@ -717,8 +746,8 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                pthread_mutex_lock(&new_stream->lock);
                if (ctx->on_recv_stream) {
-                       ret = ctx->on_recv_stream(new_stream);
-                       if (ret < 0) {
+                       int ret_recv_stream = ctx->on_recv_stream(new_stream);
+                       if (ret_recv_stream < 0) {
                                pthread_mutex_unlock(&new_stream->lock);
                                pthread_mutex_unlock(&channel->lock);
                                consumer_stream_free(new_stream);
@@ -744,9 +773,11 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                /* Send stream to relayd if the stream has an ID. */
                if (new_stream->net_seq_idx != (uint64_t) -1ULL) {
-                       ret = consumer_send_relayd_stream(new_stream,
-                                       new_stream->chan->pathname);
-                       if (ret < 0) {
+                       int ret_send_relayd_stream;
+
+                       ret_send_relayd_stream = consumer_send_relayd_stream(
+                                       new_stream, new_stream->chan->pathname);
+                       if (ret_send_relayd_stream < 0) {
                                pthread_mutex_unlock(&new_stream->lock);
                                pthread_mutex_unlock(&channel->lock);
                                consumer_stream_free(new_stream);
@@ -759,9 +790,12 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
                         * to send the "streams_sent" command to relayd.
                         */
                        if (channel->streams_sent_to_relayd) {
-                               ret = consumer_send_relayd_streams_sent(
-                                               new_stream->net_seq_idx);
-                               if (ret < 0) {
+                               int ret_send_relayd_streams_sent;
+
+                               ret_send_relayd_streams_sent =
+                                               consumer_send_relayd_streams_sent(
+                                                               new_stream->net_seq_idx);
+                               if (ret_send_relayd_streams_sent < 0) {
                                        pthread_mutex_unlock(&new_stream->lock);
                                        pthread_mutex_unlock(&channel->lock);
                                        goto error_add_stream_nosignal;
@@ -785,8 +819,9 @@ int lttng_kconsumer_recv_cmd(struct lttng_consumer_local_data *ctx,
 
                health_code_update();
 
-               ret = lttng_pipe_write(stream_pipe, &new_stream, sizeof(new_stream));
-               if (ret < 0) {
+               ret_pipe_write = lttng_pipe_write(
+                               stream_pipe, &new_stream, sizeof(new_stream));
+               if (ret_pipe_write < 0) {
                        ERR("Consumer write %s stream to pipe %d",
                                        new_stream->metadata_flag ? "metadata" : "data",
                                        lttng_pipe_get_writefd(stream_pipe));
@@ -810,6 +845,7 @@ error_add_stream_fatal:
        case LTTNG_CONSUMER_STREAMS_SENT:
        {
                struct lttng_consumer_channel *channel;
+               int ret_send_status;
 
                /*
                 * Get stream's channel reference. Needed when adding the stream to the
@@ -831,8 +867,9 @@ error_add_stream_fatal:
                /*
                 * Send status code to session daemon.
                 */
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0 || ret_code != LTTCOMM_CONSUMERD_SUCCESS) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0 ||
+                               ret_code != LTTCOMM_CONSUMERD_SUCCESS) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_streams_sent_nosignal;
                }
@@ -850,9 +887,11 @@ error_add_stream_fatal:
                health_code_update();
                /* Send stream to relayd if the stream has an ID. */
                if (msg.u.sent_streams.net_seq_idx != (uint64_t) -1ULL) {
-                       ret = consumer_send_relayd_streams_sent(
+                       int ret_send_relay_streams;
+
+                       ret_send_relay_streams = consumer_send_relayd_streams_sent(
                                        msg.u.sent_streams.net_seq_idx);
-                       if (ret < 0) {
+                       if (ret_send_relay_streams < 0) {
                                goto error_streams_sent_nosignal;
                        }
                        channel->streams_sent_to_relayd = true;
@@ -871,6 +910,7 @@ error_streams_sent_nosignal:
        {
                uint64_t index = msg.u.destroy_relayd.net_seq_idx;
                struct consumer_relayd_sock_pair *relayd;
+               int ret_send_status;
 
                DBG("Kernel consumer destroying relayd %" PRIu64, index);
 
@@ -897,8 +937,8 @@ error_streams_sent_nosignal:
 
                health_code_update();
 
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_fatal;
                }
@@ -907,18 +947,20 @@ error_streams_sent_nosignal:
        }
        case LTTNG_CONSUMER_DATA_PENDING:
        {
-               int32_t ret;
+               int32_t ret_data_pending;
                uint64_t id = msg.u.data_pending.session_id;
+               ssize_t ret_send;
 
                DBG("Kernel consumer data pending command for id %" PRIu64, id);
 
-               ret = consumer_data_pending(id);
+               ret_data_pending = consumer_data_pending(id);
 
                health_code_update();
 
                /* Send back returned value to session daemon */
-               ret = lttcomm_send_unix_sock(sock, &ret, sizeof(ret));
-               if (ret < 0) {
+               ret_send = lttcomm_send_unix_sock(sock, &ret_data_pending,
+                               sizeof(ret_data_pending));
+               if (ret_send < 0) {
                        PERROR("send data pending ret code");
                        goto error_fatal;
                }
@@ -933,38 +975,45 @@ error_streams_sent_nosignal:
        {
                struct lttng_consumer_channel *channel;
                uint64_t key = msg.u.snapshot_channel.key;
+               int ret_send_status;
 
                channel = consumer_find_channel(key);
                if (!channel) {
                        ERR("Channel %" PRIu64 " not found", key);
                        ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                } else {
-                       pthread_mutex_lock(&channel->lock);
                        if (msg.u.snapshot_channel.metadata == 1) {
-                               ret = lttng_kconsumer_snapshot_metadata(channel, key,
+                               int ret_snapshot;
+
+                               ret_snapshot = lttng_kconsumer_snapshot_metadata(
+                                               channel, key,
                                                msg.u.snapshot_channel.pathname,
-                                               msg.u.snapshot_channel.relayd_id, ctx);
-                               if (ret < 0) {
+                                               msg.u.snapshot_channel.relayd_id,
+                                               ctx);
+                               if (ret_snapshot < 0) {
                                        ERR("Snapshot metadata failed");
                                        ret_code = LTTCOMM_CONSUMERD_SNAPSHOT_FAILED;
                                }
                        } else {
-                               ret = lttng_kconsumer_snapshot_channel(channel, key,
+                               int ret_snapshot;
+
+                               ret_snapshot = lttng_kconsumer_snapshot_channel(
+                                               channel, key,
                                                msg.u.snapshot_channel.pathname,
                                                msg.u.snapshot_channel.relayd_id,
-                                               msg.u.snapshot_channel.nb_packets_per_stream,
+                                               msg.u.snapshot_channel
+                                                               .nb_packets_per_stream,
                                                ctx);
-                               if (ret < 0) {
+                               if (ret_snapshot < 0) {
                                        ERR("Snapshot channel failed");
                                        ret_code = LTTCOMM_CONSUMERD_SNAPSHOT_FAILED;
                                }
                        }
-                       pthread_mutex_unlock(&channel->lock);
                }
                health_code_update();
 
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto end_nosignal;
                }
@@ -974,6 +1023,7 @@ error_streams_sent_nosignal:
        {
                uint64_t key = msg.u.destroy_channel.key;
                struct lttng_consumer_channel *channel;
+               int ret_send_status;
 
                channel = consumer_find_channel(key);
                if (!channel) {
@@ -983,8 +1033,8 @@ error_streams_sent_nosignal:
 
                health_code_update();
 
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto end_destroy_channel;
                }
@@ -1077,39 +1127,43 @@ end_destroy_channel:
        case LTTNG_CONSUMER_SET_CHANNEL_MONITOR_PIPE:
        {
                int channel_monitor_pipe;
+               int ret_send_status, ret_set_channel_monitor_pipe;
+               ssize_t ret_recv;
 
                ret_code = LTTCOMM_CONSUMERD_SUCCESS;
                /* Successfully received the command's type. */
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        goto error_fatal;
                }
 
-               ret = lttcomm_recv_fds_unix_sock(sock, &channel_monitor_pipe,
-                               1);
-               if (ret != sizeof(channel_monitor_pipe)) {
+               ret_recv = lttcomm_recv_fds_unix_sock(
+                               sock, &channel_monitor_pipe, 1);
+               if (ret_recv != sizeof(channel_monitor_pipe)) {
                        ERR("Failed to receive channel monitor pipe");
                        goto error_fatal;
                }
 
                DBG("Received channel monitor pipe (%d)", channel_monitor_pipe);
-               ret = consumer_timer_thread_set_channel_monitor_pipe(
-                               channel_monitor_pipe);
-               if (!ret) {
+               ret_set_channel_monitor_pipe =
+                               consumer_timer_thread_set_channel_monitor_pipe(
+                                               channel_monitor_pipe);
+               if (!ret_set_channel_monitor_pipe) {
                        int flags;
+                       int ret_fcntl;
 
                        ret_code = LTTCOMM_CONSUMERD_SUCCESS;
                        /* Set the pipe as non-blocking. */
-                       ret = fcntl(channel_monitor_pipe, F_GETFL, 0);
-                       if (ret == -1) {
+                       ret_fcntl = fcntl(channel_monitor_pipe, F_GETFL, 0);
+                       if (ret_fcntl == -1) {
                                PERROR("fcntl get flags of the channel monitoring pipe");
                                goto error_fatal;
                        }
-                       flags = ret;
+                       flags = ret_fcntl;
 
-                       ret = fcntl(channel_monitor_pipe, F_SETFL,
+                       ret_fcntl = fcntl(channel_monitor_pipe, F_SETFL,
                                        flags | O_NONBLOCK);
-                       if (ret == -1) {
+                       if (ret_fcntl == -1) {
                                PERROR("fcntl set O_NONBLOCK flag of the channel monitoring pipe");
                                goto error_fatal;
                        }
@@ -1117,8 +1171,8 @@ end_destroy_channel:
                } else {
                        ret_code = LTTCOMM_CONSUMERD_ALREADY_SET;
                }
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        goto error_fatal;
                }
                break;
@@ -1127,6 +1181,7 @@ end_destroy_channel:
        {
                struct lttng_consumer_channel *channel;
                uint64_t key = msg.u.rotate_channel.key;
+               int ret_send_status;
 
                DBG("Consumer rotate channel %" PRIu64, key);
 
@@ -1138,27 +1193,32 @@ end_destroy_channel:
                        /*
                         * Sample the rotate position of all the streams in this channel.
                         */
-                       ret = lttng_consumer_rotate_channel(channel, key,
+                       int ret_rotate_channel;
+
+                       ret_rotate_channel = lttng_consumer_rotate_channel(
+                                       channel, key,
                                        msg.u.rotate_channel.relayd_id,
-                                       msg.u.rotate_channel.metadata,
-                                       ctx);
-                       if (ret < 0) {
+                                       msg.u.rotate_channel.metadata, ctx);
+                       if (ret_rotate_channel < 0) {
                                ERR("Rotate channel failed");
                                ret_code = LTTCOMM_CONSUMERD_ROTATION_FAIL;
                        }
 
                        health_code_update();
                }
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto error_rotate_channel;
                }
                if (channel) {
                        /* Rotate the streams that are ready right now. */
-                       ret = lttng_consumer_rotate_ready_streams(
+                       int ret_rotate;
+
+                       ret_rotate = lttng_consumer_rotate_ready_streams(
                                        channel, key, ctx);
-                       if (ret < 0) {
+                       if (ret_rotate < 0) {
                                ERR("Rotate ready streams failed");
                        }
                }
@@ -1170,22 +1230,27 @@ error_rotate_channel:
        {
                struct lttng_consumer_channel *channel;
                uint64_t key = msg.u.clear_channel.key;
+               int ret_send_status;
 
                channel = consumer_find_channel(key);
                if (!channel) {
                        DBG("Channel %" PRIu64 " not found", key);
                        ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
                } else {
-                       ret = lttng_consumer_clear_channel(channel);
-                       if (ret) {
+                       int ret_clear_channel;
+
+                       ret_clear_channel =
+                                       lttng_consumer_clear_channel(channel);
+                       if (ret_clear_channel) {
                                ERR("Clear channel failed");
-                               ret_code = ret;
+                               ret_code = ret_clear_channel;
                        }
 
                        health_code_update();
                }
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto end_nosignal;
                }
@@ -1194,11 +1259,13 @@ error_rotate_channel:
        }
        case LTTNG_CONSUMER_INIT:
        {
+               int ret_send_status;
+
                ret_code = lttng_consumer_init_command(ctx,
                                msg.u.init.sessiond_uuid);
                health_code_update();
-               ret = consumer_send_status_msg(sock, ret_code);
-               if (ret < 0) {
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
                        /* Somehow, the session daemon is not responding anymore. */
                        goto end_nosignal;
                }
@@ -1207,8 +1274,8 @@ error_rotate_channel:
        case LTTNG_CONSUMER_CREATE_TRACE_CHUNK:
        {
                const struct lttng_credentials credentials = {
-                       .uid = msg.u.create_trace_chunk.credentials.value.uid,
-                       .gid = msg.u.create_trace_chunk.credentials.value.gid,
+                       .uid = LTTNG_OPTIONAL_INIT_VALUE(msg.u.create_trace_chunk.credentials.value.uid),
+                       .gid = LTTNG_OPTIONAL_INIT_VALUE(msg.u.create_trace_chunk.credentials.value.gid),
                };
                const bool is_local_trace =
                                !msg.u.create_trace_chunk.relayd_id.is_set;
@@ -1226,17 +1293,20 @@ error_rotate_channel:
                 */
                if (is_local_trace) {
                        int chunk_dirfd;
+                       int ret_send_status;
+                       ssize_t ret_recv;
 
                        /* Acnowledge the reception of the command. */
-                       ret = consumer_send_status_msg(sock,
-                                       LTTCOMM_CONSUMERD_SUCCESS);
-                       if (ret < 0) {
+                       ret_send_status = consumer_send_status_msg(
+                                       sock, LTTCOMM_CONSUMERD_SUCCESS);
+                       if (ret_send_status < 0) {
                                /* Somehow, the session daemon is not responding anymore. */
                                goto end_nosignal;
                        }
 
-                       ret = lttcomm_recv_fds_unix_sock(sock, &chunk_dirfd, 1);
-                       if (ret != sizeof(chunk_dirfd)) {
+                       ret_recv = lttcomm_recv_fds_unix_sock(
+                                       sock, &chunk_dirfd, 1);
+                       if (ret_recv != sizeof(chunk_dirfd)) {
                                ERR("Failed to receive trace chunk directory file descriptor");
                                goto error_fatal;
                        }
@@ -1276,6 +1346,7 @@ error_rotate_channel:
                                msg.u.close_trace_chunk.relayd_id.value;
                struct lttcomm_consumer_close_trace_chunk_reply reply;
                char path[LTTNG_PATH_MAX];
+               ssize_t ret_send;
 
                ret_code = lttng_consumer_close_trace_chunk(
                                msg.u.close_trace_chunk.relayd_id.is_set ?
@@ -1289,12 +1360,13 @@ error_rotate_channel:
                                                NULL, path);
                reply.ret_code = ret_code;
                reply.path_length = strlen(path) + 1;
-               ret = lttcomm_send_unix_sock(sock, &reply, sizeof(reply));
-               if (ret != sizeof(reply)) {
+               ret_send = lttcomm_send_unix_sock(sock, &reply, sizeof(reply));
+               if (ret_send != sizeof(reply)) {
                        goto error_fatal;
                }
-               ret = lttcomm_send_unix_sock(sock, path, reply.path_length);
-               if (ret != reply.path_length) {
+               ret_send = lttcomm_send_unix_sock(
+                               sock, path, reply.path_length);
+               if (ret_send != reply.path_length) {
                        goto error_fatal;
                }
                goto end_nosignal;
@@ -1311,6 +1383,24 @@ error_rotate_channel:
                                msg.u.trace_chunk_exists.chunk_id);
                goto end_msg_sessiond;
        }
+       case LTTNG_CONSUMER_OPEN_CHANNEL_PACKETS:
+       {
+               const uint64_t key = msg.u.open_channel_packets.key;
+               struct lttng_consumer_channel *channel =
+                               consumer_find_channel(key);
+
+               if (channel) {
+                       pthread_mutex_lock(&channel->lock);
+                       ret_code = lttng_consumer_open_channel_packets(channel);
+                       pthread_mutex_unlock(&channel->lock);
+               } else {
+                       WARN("Channel %" PRIu64 " not found", key);
+                       ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND;
+               }
+
+               health_code_update();
+               goto end_msg_sessiond;
+       }
        default:
                goto end_nosignal;
        }
@@ -1320,11 +1410,11 @@ end_nosignal:
         * Return 1 to indicate success since the 0 value can be a socket
         * shutdown during the recv() or send() call.
         */
-       ret = 1;
+       ret_func = 1;
        goto end;
 error_fatal:
        /* This will issue a consumer stop. */
-       ret = -1;
+       ret_func = -1;
        goto end;
 end_msg_sessiond:
        /*
@@ -1332,15 +1422,21 @@ end_msg_sessiond:
         * the caller because the session daemon socket management is done
         * elsewhere. Returning a negative code or 0 will shutdown the consumer.
         */
-       ret = consumer_send_status_msg(sock, ret_code);
-       if (ret < 0) {
-               goto error_fatal;
+       {
+               int ret_send_status;
+
+               ret_send_status = consumer_send_status_msg(sock, ret_code);
+               if (ret_send_status < 0) {
+                       goto error_fatal;
+               }
        }
-       ret = 1;
+
+       ret_func = 1;
+
 end:
        health_code_update();
        rcu_read_unlock();
-       return ret;
+       return ret_func;
 }
 
 /*
@@ -1348,36 +1444,38 @@ end:
  * metadata thread can consumer them.
  *
  * Metadata stream lock MUST be acquired.
- *
- * Return 0 if new metadatda is available, EAGAIN if the metadata stream
- * is empty or a negative value on error.
  */
-int lttng_kconsumer_sync_metadata(struct lttng_consumer_stream *metadata)
+enum sync_metadata_status lttng_kconsumer_sync_metadata(
+               struct lttng_consumer_stream *metadata)
 {
        int ret;
+       enum sync_metadata_status status;
 
        assert(metadata);
 
        ret = kernctl_buffer_flush(metadata->wait_fd);
        if (ret < 0) {
                ERR("Failed to flush kernel stream");
+               status = SYNC_METADATA_STATUS_ERROR;
                goto end;
        }
 
        ret = kernctl_snapshot(metadata->wait_fd);
        if (ret < 0) {
-               if (ret != -EAGAIN) {
+               if (errno == EAGAIN) {
+                       /* No new metadata, exit. */
+                       DBG("Sync metadata, no new kernel metadata");
+                       status = SYNC_METADATA_STATUS_NO_DATA;
+               } else {
                        ERR("Sync metadata, taking kernel snapshot failed.");
-                       goto end;
+                       status = SYNC_METADATA_STATUS_ERROR;
                }
-               DBG("Sync metadata, no new kernel metadata");
-               /* No new metadata, exit. */
-               ret = ENODATA;
-               goto end;
+       } else {
+               status = SYNC_METADATA_STATUS_NEW_DATA;
        }
 
 end:
-       return ret;
+       return status;
 }
 
 static
@@ -1504,68 +1602,94 @@ end:
 }
 
 static
-int get_subbuffer_common(struct lttng_consumer_stream *stream,
+enum get_next_subbuffer_status get_subbuffer_common(
+               struct lttng_consumer_stream *stream,
                struct stream_subbuffer *subbuffer)
 {
        int ret;
+       enum get_next_subbuffer_status status;
 
        ret = kernctl_get_next_subbuf(stream->wait_fd);
-       if (ret) {
+       switch (ret) {
+       case 0:
+               status = GET_NEXT_SUBBUFFER_STATUS_OK;
+               break;
+       case -ENODATA:
+               case -EAGAIN:
+               /*
+                * The caller only expects -ENODATA when there is no data to
+                * read, but the kernel tracer returns -EAGAIN when there is
+                * currently no data for a non-finalized stream, and -ENODATA
+                * when there is no data for a finalized stream. Those can be
+                * combined into a -ENODATA return value.
+                */
+               status = GET_NEXT_SUBBUFFER_STATUS_NO_DATA;
+               goto end;
+       default:
+               status = GET_NEXT_SUBBUFFER_STATUS_ERROR;
                goto end;
        }
 
        ret = stream->read_subbuffer_ops.extract_subbuffer_info(
-                       stream, subbuffer);
+               stream, subbuffer);
+       if (ret) {
+               status = GET_NEXT_SUBBUFFER_STATUS_ERROR;
+       }
 end:
-       return ret;
+       return status;
 }
 
 static
-int get_next_subbuffer_splice(struct lttng_consumer_stream *stream,
+enum get_next_subbuffer_status get_next_subbuffer_splice(
+               struct lttng_consumer_stream *stream,
                struct stream_subbuffer *subbuffer)
 {
-       int ret;
+       const enum get_next_subbuffer_status status =
+                       get_subbuffer_common(stream, subbuffer);
 
-       ret = get_subbuffer_common(stream, subbuffer);
-       if (ret) {
+       if (status != GET_NEXT_SUBBUFFER_STATUS_OK) {
                goto end;
        }
 
        subbuffer->buffer.fd = stream->wait_fd;
 end:
-       return ret;
+       return status;
 }
 
 static
-int get_next_subbuffer_mmap(struct lttng_consumer_stream *stream,
+enum get_next_subbuffer_status get_next_subbuffer_mmap(
+               struct lttng_consumer_stream *stream,
                struct stream_subbuffer *subbuffer)
 {
        int ret;
+       enum get_next_subbuffer_status status;
        const char *addr;
 
-       ret = get_subbuffer_common(stream, subbuffer);
-       if (ret) {
+       status = get_subbuffer_common(stream, subbuffer);
+       if (status != GET_NEXT_SUBBUFFER_STATUS_OK) {
                goto end;
        }
 
        ret = get_current_subbuf_addr(stream, &addr);
        if (ret) {
+               status = GET_NEXT_SUBBUFFER_STATUS_ERROR;
                goto end;
        }
 
        subbuffer->buffer.buffer = lttng_buffer_view_init(
                        addr, 0, subbuffer->info.data.padded_subbuf_size);
 end:
-       return ret;
+       return status;
 }
 
 static
-int get_next_subbuffer_metadata_check(struct lttng_consumer_stream *stream,
+enum get_next_subbuffer_status get_next_subbuffer_metadata_check(struct lttng_consumer_stream *stream,
                struct stream_subbuffer *subbuffer)
 {
        int ret;
        const char *addr;
        bool coherent;
+       enum get_next_subbuffer_status status;
 
        ret = kernctl_get_next_subbuf_metadata_check(stream->wait_fd,
                        &coherent);
@@ -1592,7 +1716,33 @@ int get_next_subbuffer_metadata_check(struct lttng_consumer_stream *stream,
                        subbuffer->info.metadata.padded_subbuf_size,
                        coherent ? "true" : "false");
 end:
-       return ret;
+       /*
+        * The caller only expects -ENODATA when there is no data to read, but
+        * the kernel tracer returns -EAGAIN when there is currently no data
+        * for a non-finalized stream, and -ENODATA when there is no data for a
+        * finalized stream. Those can be combined into a -ENODATA return value.
+        */
+       switch (ret) {
+       case 0:
+               status = GET_NEXT_SUBBUFFER_STATUS_OK;
+               break;
+       case -ENODATA:
+       case -EAGAIN:
+               /*
+                * The caller only expects -ENODATA when there is no data to
+                * read, but the kernel tracer returns -EAGAIN when there is
+                * currently no data for a non-finalized stream, and -ENODATA
+                * when there is no data for a finalized stream. Those can be
+                * combined into a -ENODATA return value.
+                */
+               status = GET_NEXT_SUBBUFFER_STATUS_NO_DATA;
+               break;
+       default:
+               status = GET_NEXT_SUBBUFFER_STATUS_ERROR;
+               break;
+       }
+
+       return status;
 }
 
 static
@@ -1616,8 +1766,23 @@ int put_next_subbuffer(struct lttng_consumer_stream *stream,
 static
 bool is_get_next_check_metadata_available(int tracer_fd)
 {
-       return kernctl_get_next_subbuf_metadata_check(tracer_fd, NULL) !=
-                       -ENOTTY;
+       const int ret = kernctl_get_next_subbuf_metadata_check(tracer_fd, NULL);
+       const bool available = ret != -ENOTTY;
+
+       if (ret == 0) {
+               /* get succeeded, make sure to put the subbuffer. */
+               kernctl_put_subbuf(tracer_fd);
+       }
+
+       return available;
+}
+
+static
+int signal_metadata(struct lttng_consumer_stream *stream,
+               struct lttng_consumer_local_data *ctx)
+{
+       ASSERT_LOCKED(stream->metadata_rdv_lock);
+       return pthread_cond_broadcast(&stream->metadata_rdv) ? -errno : 0;
 }
 
 static
@@ -1650,6 +1815,8 @@ int lttng_kconsumer_set_stream_ops(
                        metadata_bucket_destroy(stream->metadata_bucket);
                        stream->metadata_bucket = NULL;
                }
+
+               stream->read_subbuffer_ops.on_sleep = signal_metadata;
        }
 
        if (!stream->read_subbuffer_ops.get_next_subbuffer) {
This page took 0.038762 seconds and 4 git commands to generate.