X-Git-Url: https://git.lttng.org/?p=lttng-tools.git;a=blobdiff_plain;f=src%2Fcommon%2Fust-consumer%2Fust-consumer.c;h=73d27b57b04effa5ad3307576e3b1461be63699c;hp=76eee2ad9f56d448a9f5c9913cd86685b0279e8c;hb=3eb928aad0c1f249830d16d6744a5a23b145a26c;hpb=602766ecff63f7ea4267336ba85fe969283bb212 diff --git a/src/common/ust-consumer/ust-consumer.c b/src/common/ust-consumer/ust-consumer.c index 76eee2ad9..73d27b57b 100644 --- a/src/common/ust-consumer/ust-consumer.c +++ b/src/common/ust-consumer/ust-consumer.c @@ -1,6 +1,7 @@ /* * Copyright (C) 2011 - Julien Desfossez * Mathieu Desnoyers + * Copyright (C) 2017 - Jérémie Galarneau * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License, version 2 only, @@ -31,6 +32,7 @@ #include #include #include +#include #include #include @@ -50,7 +52,6 @@ extern struct lttng_consumer_global_data consumer_data; extern int consumer_poll_timeout; -extern volatile int consumer_quit; /* * Free channel object and all streams associated with it. This MUST be used @@ -146,7 +147,8 @@ static struct lttng_consumer_channel *allocate_channel(uint64_t session_id, */ static struct lttng_consumer_stream *allocate_stream(int cpu, int key, struct lttng_consumer_channel *channel, - struct lttng_consumer_local_data *ctx, int *_alloc_ret) + struct lttng_consumer_local_data *ctx, int *_alloc_ret, + uint64_t trace_archive_id) { int alloc_ret; struct lttng_consumer_stream *stream = NULL; @@ -165,7 +167,8 @@ static struct lttng_consumer_stream *allocate_stream(int cpu, int key, cpu, &alloc_ret, channel->type, - channel->monitor); + channel->monitor, + trace_archive_id); if (stream == NULL) { switch (alloc_ret) { case -ENOENT: @@ -184,6 +187,7 @@ static struct lttng_consumer_stream *allocate_stream(int cpu, int key, goto error; } + consumer_stream_update_channel_attributes(stream, channel); stream->chan = channel; error: @@ -206,28 +210,21 @@ static int send_stream_to_thread(struct lttng_consumer_stream *stream, /* Get the right pipe where the stream will be sent. */ if (stream->metadata_flag) { - ret = consumer_add_metadata_stream(stream); - if (ret) { - ERR("Consumer add metadata stream %" PRIu64 " failed.", - stream->key); - goto error; - } + consumer_add_metadata_stream(stream); stream_pipe = ctx->consumer_metadata_pipe; } else { - ret = consumer_add_data_stream(stream); - if (ret) { - ERR("Consumer add stream %" PRIu64 " failed.", - stream->key); - goto error; - } + consumer_add_data_stream(stream); stream_pipe = ctx->consumer_data_pipe; } /* * From this point on, the stream's ownership has been moved away from - * the channel and becomes globally visible. + * the channel and it becomes globally visible. Hence, remove it from + * the local stream list to prevent the stream from being both local and + * global. */ stream->globally_visible = 1; + cds_list_del(&stream->send_node); ret = lttng_pipe_write(stream_pipe, &stream, sizeof(stream)); if (ret < 0) { @@ -239,7 +236,9 @@ static int send_stream_to_thread(struct lttng_consumer_stream *stream, } else { consumer_del_stream_for_data(stream); } + goto error; } + error: return ret; } @@ -270,7 +269,8 @@ end: * Return 0 on success else a negative value. */ static int create_ust_streams(struct lttng_consumer_channel *channel, - struct lttng_consumer_local_data *ctx) + struct lttng_consumer_local_data *ctx, + uint64_t trace_archive_id) { int ret, cpu = 0; struct ustctl_consumer_stream *ustream; @@ -301,7 +301,8 @@ static int create_ust_streams(struct lttng_consumer_channel *channel, } /* Allocate consumer stream object. */ - stream = allocate_stream(cpu, wait_fd, channel, ctx, &ret); + stream = allocate_stream(cpu, wait_fd, channel, ctx, &ret, + trace_archive_id); if (!stream) { goto error_alloc; } @@ -410,9 +411,7 @@ error_shm_open: return -1; } -static int open_ust_stream_fd(struct lttng_consumer_channel *channel, - struct ustctl_consumer_channel_attr *attr, - int cpu) +static int open_ust_stream_fd(struct lttng_consumer_channel *channel, int cpu) { char shm_path[PATH_MAX]; int ret; @@ -467,7 +466,7 @@ static int create_ust_channel(struct lttng_consumer_channel *channel, goto error_alloc; } for (i = 0; i < nr_stream_fds; i++) { - stream_fds[i] = open_ust_stream_fd(channel, attr, i); + stream_fds[i] = open_ust_stream_fd(channel, i); if (stream_fds[i] < 0) { ret = -1; goto error_open; @@ -543,11 +542,11 @@ error: } /* - * Send channel to sessiond. + * Send channel to sessiond and relayd if applicable. * * Return 0 on success or else a negative value. */ -static int send_sessiond_channel(int sock, +static int send_channel_to_sessiond_and_relayd(int sock, struct lttng_consumer_channel *channel, struct lttng_consumer_local_data *ctx, int *relayd_error) { @@ -567,6 +566,8 @@ static int send_sessiond_channel(int sock, health_code_update(); /* Try to send the stream to the relayd if one is available. */ + DBG("Sending stream %" PRIu64 " of channel \"%s\" to relayd", + stream->key, channel->name); ret = consumer_send_relayd_stream(stream, stream->chan->pathname); if (ret < 0) { /* @@ -642,9 +643,10 @@ error: * Return 0 on success or else, a negative value is returned and the channel * MUST be destroyed by consumer_del_channel(). */ -static int ask_channel(struct lttng_consumer_local_data *ctx, int sock, +static int ask_channel(struct lttng_consumer_local_data *ctx, struct lttng_consumer_channel *channel, - struct ustctl_consumer_channel_attr *attr) + struct ustctl_consumer_channel_attr *attr, + uint64_t trace_archive_id) { int ret; @@ -685,7 +687,7 @@ static int ask_channel(struct lttng_consumer_local_data *ctx, int sock, } /* Open all streams for this channel. */ - ret = create_ust_streams(channel, ctx); + ret = create_ust_streams(channel, ctx, trace_archive_id); if (ret < 0) { goto end; } @@ -721,14 +723,8 @@ static int send_streams_to_thread(struct lttng_consumer_channel *channel, * If we are unable to send the stream to the thread, there is * a big problem so just stop everything. */ - /* Remove node from the channel stream list. */ - cds_list_del(&stream->send_node); goto error; } - - /* Remove node from the channel stream list. */ - cds_list_del(&stream->send_node); - } error: @@ -831,6 +827,7 @@ static int close_metadata(uint64_t chan_key) { int ret = 0; struct lttng_consumer_channel *channel; + unsigned int channel_monitor; DBG("UST consumer close metadata key %" PRIu64, chan_key); @@ -849,13 +846,48 @@ static int close_metadata(uint64_t chan_key) pthread_mutex_lock(&consumer_data.lock); pthread_mutex_lock(&channel->lock); - + channel_monitor = channel->monitor; if (cds_lfht_is_node_deleted(&channel->node.node)) { goto error_unlock; } lttng_ustconsumer_close_metadata(channel); + pthread_mutex_unlock(&channel->lock); + pthread_mutex_unlock(&consumer_data.lock); + /* + * The ownership of a metadata channel depends on the type of + * session to which it belongs. In effect, the monitor flag is checked + * to determine if this metadata channel is in "snapshot" mode or not. + * + * In the non-snapshot case, the metadata channel is created along with + * a single stream which will remain present until the metadata channel + * is destroyed (on the destruction of its session). In this case, the + * metadata stream in "monitored" by the metadata poll thread and holds + * the ownership of its channel. + * + * Closing the metadata will cause the metadata stream's "metadata poll + * pipe" to be closed. Closing this pipe will wake-up the metadata poll + * thread which will teardown the metadata stream which, in return, + * deletes the metadata channel. + * + * In the snapshot case, the metadata stream is created and destroyed + * on every snapshot record. Since the channel doesn't have an owner + * other than the session daemon, it is safe to destroy it immediately + * on reception of the CLOSE_METADATA command. + */ + if (!channel_monitor) { + /* + * The channel and consumer_data locks must be + * released before this call since consumer_del_channel + * re-acquires the channel and consumer_data locks to teardown + * the channel and queue its reclamation by the "call_rcu" + * worker thread. + */ + consumer_del_channel(channel); + } + + return ret; error_unlock: pthread_mutex_unlock(&channel->lock); pthread_mutex_unlock(&consumer_data.lock); @@ -918,6 +950,10 @@ static int setup_metadata(struct lttng_consumer_local_data *ctx, uint64_t key) } } + /* + * Ownership of metadata stream is passed along. Freeing is handled by + * the callee. + */ ret = send_streams_to_thread(metadata, ctx); if (ret < 0) { /* @@ -925,7 +961,7 @@ static int setup_metadata(struct lttng_consumer_local_data *ctx, uint64_t key) * a big problem so just stop everything. */ ret = LTTCOMM_CONSUMERD_FATAL; - goto error; + goto send_streams_error; } /* List MUST be empty after or else it could be reused. */ assert(cds_list_empty(&metadata->streams.head)); @@ -943,6 +979,7 @@ error: consumer_stream_destroy(metadata->metadata_stream, NULL); cds_list_del(&metadata->metadata_stream->send_node); metadata->metadata_stream = NULL; +send_streams_error: error_no_stream: end: return ret; @@ -950,14 +987,17 @@ end: /* * Snapshot the whole metadata. + * RCU read-side lock must be held across this function to ensure existence of + * metadata_channel. * * Returns 0 on success, < 0 on error */ -static int snapshot_metadata(uint64_t key, char *path, uint64_t relayd_id, - struct lttng_consumer_local_data *ctx) +static int snapshot_metadata(struct lttng_consumer_channel *metadata_channel, + uint64_t key, char *path, uint64_t relayd_id, + struct lttng_consumer_local_data *ctx, + uint64_t trace_archive_id) { int ret = 0; - struct lttng_consumer_channel *metadata_channel; struct lttng_consumer_stream *metadata_stream; assert(path); @@ -968,13 +1008,6 @@ static int snapshot_metadata(uint64_t key, char *path, uint64_t relayd_id, rcu_read_lock(); - metadata_channel = consumer_find_channel(key); - if (!metadata_channel) { - ERR("UST snapshot metadata channel not found for key %" PRIu64, - key); - ret = -1; - goto error; - } assert(!metadata_channel->monitor); health_code_update(); @@ -994,7 +1027,7 @@ static int snapshot_metadata(uint64_t key, char *path, uint64_t relayd_id, * The metadata stream is NOT created in no monitor mode when the channel * is created on a sessiond ask channel command. */ - ret = create_ust_streams(metadata_channel, ctx); + ret = create_ust_streams(metadata_channel, ctx, trace_archive_id); if (ret < 0) { goto error; } @@ -1045,16 +1078,19 @@ error: /* * 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. * * Returns 0 on success, < 0 on error */ -static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, - uint64_t nb_packets_per_stream, struct lttng_consumer_local_data *ctx) +static int snapshot_channel(struct lttng_consumer_channel *channel, + uint64_t key, char *path, uint64_t relayd_id, + uint64_t nb_packets_per_stream, + struct lttng_consumer_local_data *ctx) { int ret; unsigned use_relayd = 0; unsigned long consumed_pos, produced_pos; - struct lttng_consumer_channel *channel; struct lttng_consumer_stream *stream; assert(path); @@ -1066,19 +1102,10 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, use_relayd = 1; } - channel = consumer_find_channel(key); - if (!channel) { - ERR("UST snapshot channel not found for key %" PRIu64, key); - ret = -1; - goto error; - } assert(!channel->monitor); DBG("UST consumer snapshot channel %" PRIu64, key); cds_list_for_each_entry(stream, &channel->streams.head, send_node) { - /* Are we at a position _before_ the first available packet ? */ - bool before_first_packet = true; - health_code_update(); /* Lock stream because we are about to change its state. */ @@ -1104,12 +1131,6 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, DBG("UST consumer snapshot stream %s/%s (%" PRIu64 ")", path, stream->name, stream->key); } - if (relayd_id != -1ULL) { - ret = consumer_send_relayd_streams_sent(relayd_id); - if (ret < 0) { - goto error_unlock; - } - } /* * If tracing is active, we want to perform a "full" buffer flush. @@ -1150,7 +1171,6 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, while (consumed_pos < produced_pos) { ssize_t read_len; unsigned long len, padded_len; - int lost_packet = 0; health_code_update(); @@ -1164,15 +1184,7 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, } DBG("UST consumer get subbuf failed. Skipping it."); consumed_pos += stream->max_sb_size; - - /* - * Start accounting lost packets only when we - * already have extracted packets (to match the - * content of the final snapshot). - */ - if (!before_first_packet) { - lost_packet = 1; - } + stream->chan->lost_packets++; continue; } @@ -1208,16 +1220,6 @@ static int snapshot_channel(uint64_t key, char *path, uint64_t relayd_id, goto error_close_stream; } consumed_pos += stream->max_sb_size; - - /* - * Only account lost packets located between - * succesfully extracted packets (do not account before - * and after since they are not visible in the - * resulting snapshot). - */ - stream->chan->lost_packets += lost_packet; - lost_packet = 0; - before_first_packet = false; } /* Simply close the stream so we can use it on the next snapshot. */ @@ -1236,7 +1238,6 @@ error_close_stream: consumer_stream_close(stream); error_unlock: pthread_mutex_unlock(&stream->lock); -error: rcu_read_unlock(); return ret; } @@ -1354,7 +1355,7 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, case LTTNG_CONSUMER_ADD_RELAYD_SOCKET: { /* Session daemon status message are handled in the following call. */ - ret = consumer_add_relayd_socket(msg.u.relayd_sock.net_index, + 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); @@ -1455,6 +1456,7 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, attr.read_timer_interval = msg.u.ask_channel.read_timer_interval; attr.chan_id = msg.u.ask_channel.chan_id; memcpy(attr.uuid, msg.u.ask_channel.uuid, sizeof(attr.uuid)); + attr.blocking_timeout= msg.u.ask_channel.blocking_timeout; /* Match channel buffer type to the UST abi. */ switch (msg.u.ask_channel.output) { @@ -1487,7 +1489,8 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, health_code_update(); - ret = ask_channel(ctx, sock, channel, &attr); + ret = ask_channel(ctx, channel, &attr, + msg.u.ask_channel.trace_archive_id); if (ret < 0) { goto end_channel_error; } @@ -1501,8 +1504,17 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, consumer_timer_switch_start(channel, attr.switch_timer_interval); attr.switch_timer_interval = 0; } else { + int monitor_start_ret; + consumer_timer_live_start(channel, msg.u.ask_channel.live_timer_interval); + monitor_start_ret = consumer_timer_monitor_start( + channel, + msg.u.ask_channel.monitor_timer_interval); + if (monitor_start_ret < 0) { + ERR("Starting channel monitoring timer failed"); + goto end_channel_error; + } } health_code_update(); @@ -1525,6 +1537,9 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, if (channel->live_timer_enabled == 1) { consumer_timer_live_stop(channel); } + if (channel->monitor_timer_enabled == 1) { + consumer_timer_monitor_stop(channel); + } goto end_channel_error; } @@ -1560,8 +1575,9 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, health_code_update(); - /* Send everything to sessiond. */ - ret = send_sessiond_channel(sock, channel, ctx, &relayd_err); + /* Send the channel to sessiond (and relayd, if applicable). */ + ret = send_channel_to_sessiond_and_relayd(sock, channel, ctx, + &relayd_err); if (ret < 0) { if (relayd_err) { /* @@ -1725,27 +1741,36 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, } case LTTNG_CONSUMER_SNAPSHOT_CHANNEL: { - if (msg.u.snapshot_channel.metadata) { - ret = snapshot_metadata(msg.u.snapshot_channel.key, - msg.u.snapshot_channel.pathname, - msg.u.snapshot_channel.relayd_id, - ctx); - if (ret < 0) { - ERR("Snapshot metadata failed"); - ret_code = LTTCOMM_CONSUMERD_ERROR_METADATA; - } + struct lttng_consumer_channel *channel; + uint64_t key = msg.u.snapshot_channel.key; + + channel = consumer_find_channel(key); + if (!channel) { + DBG("UST snapshot channel not found for key %" PRIu64, key); + ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND; } else { - ret = snapshot_channel(msg.u.snapshot_channel.key, - msg.u.snapshot_channel.pathname, - msg.u.snapshot_channel.relayd_id, - msg.u.snapshot_channel.nb_packets_per_stream, - ctx); - if (ret < 0) { - ERR("Snapshot channel failed"); - ret_code = LTTCOMM_CONSUMERD_CHANNEL_FAIL; + if (msg.u.snapshot_channel.metadata) { + ret = snapshot_metadata(channel, key, + msg.u.snapshot_channel.pathname, + msg.u.snapshot_channel.relayd_id, + ctx, + msg.u.snapshot_channel.trace_archive_id); + if (ret < 0) { + ERR("Snapshot metadata failed"); + ret_code = LTTCOMM_CONSUMERD_SNAPSHOT_FAILED; + } + } else { + ret = snapshot_channel(channel, key, + msg.u.snapshot_channel.pathname, + msg.u.snapshot_channel.relayd_id, + msg.u.snapshot_channel.nb_packets_per_stream, + ctx); + if (ret < 0) { + ERR("Snapshot channel failed"); + ret_code = LTTCOMM_CONSUMERD_SNAPSHOT_FAILED; + } } } - health_code_update(); ret = consumer_send_status_msg(sock, ret_code); if (ret < 0) { @@ -1857,6 +1882,233 @@ int lttng_ustconsumer_recv_cmd(struct lttng_consumer_local_data *ctx, break; } + case LTTNG_CONSUMER_SET_CHANNEL_MONITOR_PIPE: + { + int channel_monitor_pipe; + + ret_code = LTTCOMM_CONSUMERD_SUCCESS; + /* Successfully received the command's type. */ + ret = consumer_send_status_msg(sock, ret_code); + if (ret < 0) { + goto error_fatal; + } + + ret = lttcomm_recv_fds_unix_sock(sock, &channel_monitor_pipe, + 1); + if (ret != 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) { + int flags; + + ret_code = LTTCOMM_CONSUMERD_SUCCESS; + /* Set the pipe as non-blocking. */ + ret = fcntl(channel_monitor_pipe, F_GETFL, 0); + if (ret == -1) { + PERROR("fcntl get flags of the channel monitoring pipe"); + goto error_fatal; + } + flags = ret; + + ret = fcntl(channel_monitor_pipe, F_SETFL, + flags | O_NONBLOCK); + if (ret == -1) { + PERROR("fcntl set O_NONBLOCK flag of the channel monitoring pipe"); + goto error_fatal; + } + DBG("Channel monitor pipe set as non-blocking"); + } else { + ret_code = LTTCOMM_CONSUMERD_ALREADY_SET; + } + goto end_msg_sessiond; + } + case LTTNG_CONSUMER_ROTATE_CHANNEL: + { + struct lttng_consumer_channel *channel; + uint64_t key = msg.u.rotate_channel.key; + + channel = consumer_find_channel(key); + if (!channel) { + DBG("Channel %" PRIu64 " not found", key); + ret_code = LTTCOMM_CONSUMERD_CHAN_NOT_FOUND; + } else { + /* + * Sample the rotate position of all the streams in + * this channel. + */ + ret = lttng_consumer_rotate_channel(channel, key, + msg.u.rotate_channel.pathname, + msg.u.rotate_channel.relayd_id, + msg.u.rotate_channel.metadata, + msg.u.rotate_channel.new_chunk_id, + ctx); + if (ret < 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) { + /* Somehow, the session daemon is not responding anymore. */ + goto end_nosignal; + } + + /* + * Rotate the streams that are ready right now. + * FIXME: this is a second consecutive iteration over the + * streams in a channel, there is probably a better way to + * handle this, but it needs to be after the + * consumer_send_status_msg() call. + */ + if (channel) { + ret = lttng_consumer_rotate_ready_streams( + channel, key, ctx); + if (ret < 0) { + ERR("Rotate channel failed"); + } + } + break; + } + case LTTNG_CONSUMER_ROTATE_RENAME: + { + DBG("Consumer rename session %" PRIu64 " after rotation", + msg.u.rotate_rename.session_id); + ret = lttng_consumer_rotate_rename(msg.u.rotate_rename.old_path, + msg.u.rotate_rename.new_path, + msg.u.rotate_rename.uid, + msg.u.rotate_rename.gid, + msg.u.rotate_rename.relayd_id); + if (ret < 0) { + ERR("Rotate rename failed"); + ret_code = LTTCOMM_CONSUMERD_ROTATE_RENAME_FAILED; + } + + health_code_update(); + + ret = consumer_send_status_msg(sock, ret_code); + if (ret < 0) { + /* Somehow, the session daemon is not responding anymore. */ + goto end_nosignal; + } + break; + } + case LTTNG_CONSUMER_CHECK_ROTATION_PENDING_LOCAL: + { + int pending; + uint32_t pending_reply; + + DBG("Perform local check of pending rotation for session id %" PRIu64, + msg.u.check_rotation_pending_local.session_id); + pending = lttng_consumer_check_rotation_pending_local( + msg.u.check_rotation_pending_local.session_id, + msg.u.check_rotation_pending_local.chunk_id); + if (pending < 0) { + ERR("Local rotation pending check failed with code %i", pending); + ret_code = LTTCOMM_CONSUMERD_ROTATION_PENDING_LOCAL_FAILED; + } else { + pending_reply = !!pending; + } + + health_code_update(); + + ret = consumer_send_status_msg(sock, ret_code); + if (ret < 0) { + /* Somehow, the session daemon is not responding anymore. */ + goto end_nosignal; + } + + if (pending < 0) { + /* + * An error occured while running the command; + * don't send the 'pending' flag as the sessiond + * will not read it. + */ + break; + } + + /* Send back returned value to session daemon */ + ret = lttcomm_send_unix_sock(sock, &pending_reply, + sizeof(pending_reply)); + if (ret < 0) { + PERROR("Failed to send rotation pending return code"); + goto error_fatal; + } + break; + } + case LTTNG_CONSUMER_CHECK_ROTATION_PENDING_RELAY: + { + int pending; + uint32_t pending_reply; + + DBG("Perform relayd check of pending rotation for session id %" PRIu64, + msg.u.check_rotation_pending_relay.session_id); + pending = lttng_consumer_check_rotation_pending_relay( + msg.u.check_rotation_pending_relay.session_id, + msg.u.check_rotation_pending_relay.relayd_id, + msg.u.check_rotation_pending_relay.chunk_id); + if (pending < 0) { + ERR("Relayd rotation pending check failed with code %i", pending); + ret_code = LTTCOMM_CONSUMERD_ROTATION_PENDING_RELAY_FAILED; + } else { + pending_reply = !!pending; + } + + health_code_update(); + + ret = consumer_send_status_msg(sock, ret_code); + if (ret < 0) { + /* Somehow, the session daemon is not responding anymore. */ + goto end_nosignal; + } + + if (pending < 0) { + /* + * An error occured while running the command; + * don't send the 'pending' flag as the sessiond + * will not read it. + */ + break; + } + + /* Send back returned value to session daemon */ + ret = lttcomm_send_unix_sock(sock, &pending_reply, + sizeof(pending_reply)); + if (ret < 0) { + PERROR("Failed to send rotation pending return code"); + goto error_fatal; + } + break; + } + case LTTNG_CONSUMER_MKDIR: + { + DBG("Consumer mkdir %s in session %" PRIu64, + msg.u.mkdir.path, + msg.u.mkdir.session_id); + ret = lttng_consumer_mkdir(msg.u.mkdir.path, + msg.u.mkdir.uid, + msg.u.mkdir.gid, + msg.u.mkdir.relayd_id); + if (ret < 0) { + ERR("consumer mkdir failed"); + ret_code = LTTCOMM_CONSUMERD_MKDIR_FAILED; + } + + health_code_update(); + + ret = consumer_send_status_msg(sock, ret_code); + if (ret < 0) { + /* Somehow, the session daemon is not responding anymore. */ + goto end_nosignal; + } + break; + } default: break; } @@ -1937,8 +2189,17 @@ void *lttng_ustctl_get_mmap_base(struct lttng_consumer_stream *stream) return ustctl_get_mmap_base(stream->ustream); } +void lttng_ustctl_flush_buffer(struct lttng_consumer_stream *stream, + int producer_active) +{ + assert(stream); + assert(stream->ustream); + + ustctl_flush_buffer(stream->ustream, producer_active); +} + /* - * Take a snapshot for a specific fd + * Take a snapshot for a specific stream. * * Returns 0 on success, < 0 on error */ @@ -1950,6 +2211,20 @@ int lttng_ustconsumer_take_snapshot(struct lttng_consumer_stream *stream) return ustctl_snapshot(stream->ustream); } +/* + * Sample consumed and produced positions for a specific stream. + * + * Returns 0 on success, < 0 on error. + */ +int lttng_ustconsumer_sample_snapshot_positions( + struct lttng_consumer_stream *stream) +{ + assert(stream); + assert(stream->ustream); + + return ustctl_snapshot_sample_positions(stream->ustream); +} + /* * Get the produced position * @@ -2273,10 +2548,10 @@ int lttng_ustconsumer_sync_metadata(struct lttng_consumer_local_data *ctx, * because we locked the metadata thread. */ ret = lttng_ustconsumer_request_metadata(ctx, metadata->chan, 0, 0); + pthread_mutex_lock(&metadata->lock); if (ret < 0) { goto end; } - pthread_mutex_lock(&metadata->lock); ret = commit_one_metadata_packet(metadata); if (ret <= 0) { @@ -2423,10 +2698,10 @@ end: * Return 0 on success else a negative value. */ int lttng_ustconsumer_read_subbuffer(struct lttng_consumer_stream *stream, - struct lttng_consumer_local_data *ctx) + struct lttng_consumer_local_data *ctx, bool *rotated) { unsigned long len, subbuf_size, padding; - int err, write_index = 1; + int err, write_index = 1, rotation_ret; long ret = 0; struct ustctl_consumer_stream *ustream; struct ctf_packet_index index; @@ -2458,7 +2733,21 @@ int lttng_ustconsumer_read_subbuffer(struct lttng_consumer_stream *stream, readlen = lttng_read(stream->wait_fd, &dummy, 1); if (readlen < 0 && errno != EAGAIN && errno != EWOULDBLOCK) { ret = readlen; - goto end; + goto error; + } + } + + /* + * If the stream was flagged to be ready for rotation before we extract the + * next packet, rotate it now. + */ + if (stream->rotate_ready) { + DBG("Rotate stream before extracting data"); + rotation_ret = lttng_consumer_rotate_stream(ctx, stream, rotated); + if (rotation_ret < 0) { + ERR("Stream rotation error"); + ret = -1; + goto error; } } @@ -2473,7 +2762,7 @@ retry: if (stream->metadata_flag) { ret = commit_one_metadata_packet(stream); if (ret <= 0) { - goto end; + goto error; } ustctl_flush_buffer(stream->ustream, 1); goto retry; @@ -2488,7 +2777,7 @@ retry: */ DBG("Reserving sub buffer failed (everything is normal, " "it is due to concurrency) [ret: %d]", err); - goto end; + goto error; } assert(stream->chan->output == CONSUMER_CHANNEL_MMAP); @@ -2496,14 +2785,18 @@ retry: index.offset = htobe64(stream->out_fd_offset); ret = get_index_values(&index, ustream); if (ret < 0) { - goto end; + err = ustctl_put_subbuf(ustream); + assert(err == 0); + goto error; } /* Update the stream's sequence and discarded events count. */ ret = update_stream_stats(stream); if (ret < 0) { PERROR("kernctl_get_events_discarded"); - goto end; + err = ustctl_put_subbuf(ustream); + assert(err == 0); + goto error; } } else { write_index = 0; @@ -2521,6 +2814,7 @@ retry: assert(len >= subbuf_size); padding = len - subbuf_size; + /* write the subbuffer to the tracefile */ ret = lttng_consumer_on_read_subbuffer_mmap(ctx, stream, subbuf_size, padding, &index); /* @@ -2552,13 +2846,13 @@ retry: if (!stream->metadata_flag) { ret = notify_if_more_data(stream, ctx); if (ret < 0) { - goto end; + goto error; } } /* Write index if needed. */ if (!write_index) { - goto end; + goto rotate; } if (stream->chan->live_timer_interval && !stream->metadata_flag) { @@ -2583,17 +2877,35 @@ retry: } if (err < 0) { - goto end; + goto error; } } assert(!stream->metadata_flag); err = consumer_stream_write_index(stream, &index); if (err < 0) { - goto end; + goto error; } -end: +rotate: + /* + * After extracting the packet, we check if the stream is now ready to be + * rotated and perform the action immediately. + */ + rotation_ret = lttng_consumer_stream_is_rotate_ready(stream); + if (rotation_ret == 1) { + rotation_ret = lttng_consumer_rotate_stream(ctx, stream, rotated); + if (rotation_ret < 0) { + ERR("Stream rotation error"); + ret = -1; + goto error; + } + } else if (rotation_ret < 0) { + ERR("Checking if stream is ready to rotate"); + ret = -1; + goto error; + } +error: return ret; } @@ -2620,14 +2932,18 @@ int lttng_ustconsumer_on_recv_stream(struct lttng_consumer_stream *stream) stream->tracefile_size_current = 0; if (!stream->metadata_flag) { - ret = index_create_file(stream->chan->pathname, + struct lttng_index_file *index_file; + + index_file = lttng_index_file_create(stream->chan->pathname, stream->name, stream->uid, stream->gid, stream->chan->tracefile_size, - stream->tracefile_count_current); - if (ret < 0) { + stream->tracefile_count_current, + CTF_INDEX_MAJOR, CTF_INDEX_MINOR); + if (!index_file) { goto error; } - stream->index_fd = ret; + assert(!stream->index_file); + stream->index_file = index_file; } } ret = 0; @@ -2834,7 +3150,7 @@ int lttng_ustconsumer_request_metadata(struct lttng_consumer_local_data *ctx, request.key = channel->key; DBG("Sending metadata request to sessiond, session id %" PRIu64 - ", per-pid %" PRIu64 ", app UID %u and channek key %" PRIu64, + ", per-pid %" PRIu64 ", app UID %u and channel key %" PRIu64, request.session_id, request.session_id_per_pid, request.uid, request.key);