X-Git-Url: http://git.lttng.org/?a=blobdiff_plain;f=src%2Fbin%2Flttng-relayd%2Fmain.c;h=6f439176f830d3e4b45e9584cd094a4beb698951;hb=a86c0657eb37be08e07549a2444c5dc05b0dbda6;hp=9c0e2b1e37158d728ace4bae91a21a70d9d80a93;hpb=3a90409872c4a2f631b1efa1b37551b713a961ad;p=lttng-tools.git diff --git a/src/bin/lttng-relayd/main.c b/src/bin/lttng-relayd/main.c index 9c0e2b1e3..6f439176f 100644 --- a/src/bin/lttng-relayd/main.c +++ b/src/bin/lttng-relayd/main.c @@ -70,6 +70,15 @@ #include "stream.h" #include "connection.h" #include "tracefile-array.h" +#include "tcp_keep_alive.h" + +static const char *help_msg = +#ifdef LTTNG_EMBED_HELP +#include +#else +NULL +#endif +; /* command line options */ char *opt_output_path; @@ -250,9 +259,9 @@ static int set_option(int opt, const char *arg, const char *optname) } break; case 'h': - ret = utils_show_man_page(8, "lttng-relayd"); + ret = utils_show_help(8, "lttng-relayd", help_msg); if (ret) { - ERR("Cannot view man page lttng-relayd(8)"); + ERR("Cannot show --help for `lttng-relayd`"); perror("exec"); } exit(EXIT_FAILURE); @@ -740,6 +749,7 @@ static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) ret = sock->ops->bind(sock); if (ret < 0) { + PERROR("Failed to bind socket"); goto error; } @@ -891,6 +901,15 @@ restart: lttcomm_destroy_sock(newsock); goto error; } + + ret = socket_apply_keep_alive_config(newsock->fd); + if (ret < 0) { + ERR("Failed to apply TCP keep-alive configuration on socket (%i)", + newsock->fd); + lttcomm_destroy_sock(newsock); + goto error; + } + new_conn = connection_create(newsock, type); if (!new_conn) { lttcomm_destroy_sock(newsock); @@ -969,12 +988,16 @@ static void *relay_thread_dispatcher(void *data) health_code_update(); - while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { + for (;;) { health_code_update(); /* Atomically prepare the queue futex */ futex_nto1_prepare(&relay_conn_queue.futex); + if (CMM_LOAD_SHARED(dispatch_thread_exit)) { + break; + } + do { health_code_update(); @@ -1579,6 +1602,7 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, { int ret; struct lttcomm_relayd_version reply, msg; + bool compatible = true; conn->version_check_done = 1; @@ -1603,9 +1627,7 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, if (reply.major != be32toh(msg.major)) { DBG("Incompatible major versions (%u vs %u), deleting session", reply.major, be32toh(msg.major)); - connection_put(conn); - ret = 0; - goto end; + compatible = false; } conn->major = reply.major; @@ -1624,6 +1646,11 @@ static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, ERR("Relay sending version"); } + if (!compatible) { + ret = -1; + goto end; + } + DBG("Version check done using protocol %u.%u", conn->major, conn->minor); @@ -1946,6 +1973,7 @@ static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, struct lttcomm_relayd_generic_reply reply; struct relay_stream *stream; uint64_t net_seq_num; + size_t msg_len; assert(conn); @@ -1957,9 +1985,12 @@ static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, goto end_no_session; } + msg_len = lttcomm_relayd_index_len( + lttng_to_index_major(conn->major, conn->minor), + lttng_to_index_minor(conn->major, conn->minor)); ret = conn->sock->ops->recvmsg(conn->sock, &index_info, - sizeof(index_info), 0); - if (ret < sizeof(index_info)) { + msg_len, 0); + if (ret < msg_len) { if (ret == 0) { /* Orderly shutdown. Not necessary to print an error. */ DBG("Socket %d did an orderly shutdown", conn->sock->fd); @@ -2023,8 +2054,13 @@ static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, /* no flush. */ ret = 0; } else { + /* + * ret < 0 + * + * relay_index_try_flush is responsible for the self-reference + * put of the index object on error. + */ ERR("relay_index_try_flush error %d", ret); - relay_index_put(index); ret = -1; } @@ -2183,41 +2219,36 @@ static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, goto end; } - if (rotate_index || !stream->index_fd) { - int fd; + if (rotate_index || !stream->index_file) { + uint32_t major, minor; - /* Put ref on previous index_fd. */ - if (stream->index_fd) { - stream_fd_put(stream->index_fd); - stream->index_fd = NULL; + /* Put ref on previous index_file. */ + if (stream->index_file) { + lttng_index_file_put(stream->index_file); + stream->index_file = NULL; } - - fd = index_create_file(stream->path_name, stream->channel_name, + major = stream->trace->session->major; + minor = stream->trace->session->minor; + stream->index_file = lttng_index_file_create(stream->path_name, + stream->channel_name, -1, -1, stream->tracefile_size, - tracefile_array_get_file_index_head(stream->tfa)); - if (fd < 0) { - ret = -1; - /* Put self-ref for this index due to error. */ - relay_index_put(index); - goto end; - } - stream->index_fd = stream_fd_create(fd); - if (!stream->index_fd) { + tracefile_array_get_file_index_head(stream->tfa), + lttng_to_index_major(major, minor), + lttng_to_index_minor(major, minor)); + if (!stream->index_file) { ret = -1; - if (close(fd)) { - PERROR("Error closing FD %d", fd); - } /* Put self-ref for this index due to error. */ relay_index_put(index); - /* Will put the local ref. */ + index = NULL; goto end; } } - if (relay_index_set_fd(index, stream->index_fd, data_offset)) { + if (relay_index_set_file(index, stream->index_file, data_offset)) { ret = -1; /* Put self-ref for this index due to error. */ relay_index_put(index); + index = NULL; goto end; } @@ -2229,8 +2260,13 @@ static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, /* No flush. */ ret = 0; } else { - /* Put self-ref for this index due to error. */ - relay_index_put(index); + /* + * ret < 0 + * + * relay_index_try_flush is responsible for the self-reference + * put of the index object on error. + */ + ERR("relay_index_try_flush error %d", ret); ret = -1; } end: @@ -2680,6 +2716,11 @@ error: destroy_conn, sock_n.node) { health_code_update(); + + if (session_abort(destroy_conn->session)) { + assert(0); + } + /* * No need to grab another ref, because we own * destroy_conn. @@ -2932,6 +2973,12 @@ exit_init_data: health_app_destroy(health_relayd); exit_health_app_create: exit_options: + /* + * Wait for all pending call_rcu work to complete before tearing + * down data structures. call_rcu worker may be trying to + * perform lookups in those structures. + */ + rcu_barrier(); relayd_cleanup(); /* Ensure all prior call_rcu are done. */