X-Git-Url: https://git.lttng.org/?a=blobdiff_plain;f=src%2Fbin%2Flttng-relayd%2Fmain.c;h=238d20431618677943f8197434f65343fca60a03;hb=695f8ee100ee631ac103456c78dc77af8aff3f3b;hp=ea46ec72276117359b32fa499b584357b5aeb401;hpb=0072e5e28fb5f18daffb930be7efd55d2474e6ad;p=lttng-tools.git diff --git a/src/bin/lttng-relayd/main.c b/src/bin/lttng-relayd/main.c index ea46ec722..238d20431 100644 --- a/src/bin/lttng-relayd/main.c +++ b/src/bin/lttng-relayd/main.c @@ -70,6 +70,7 @@ #include "stream.h" #include "connection.h" #include "tracefile-array.h" +#include "tcp_keep_alive.h" /* command line options */ char *opt_output_path; @@ -161,10 +162,11 @@ static struct option long_options[] = { { "output", 1, 0, 'o', }, { "verbose", 0, 0, 'v', }, { "config", 1, 0, 'f' }, + { "version", 0, 0, 'V' }, { NULL, 0, 0, 0, }, }; -static const char *config_ignore_options[] = { "help", "config" }; +static const char *config_ignore_options[] = { "help", "config", "version" }; /* * Take an option from the getopt output and set it in the right variable to be @@ -255,6 +257,9 @@ static int set_option(int opt, const char *arg, const char *optname) perror("exec"); } exit(EXIT_FAILURE); + case 'V': + fprintf(stdout, "%s\n", VERSION); + exit(EXIT_SUCCESS); case 'o': if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", @@ -887,6 +892,15 @@ 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); @@ -965,12 +979,16 @@ 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(); @@ -1575,6 +1593,7 @@ 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; @@ -1599,9 +1618,7 @@ 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; @@ -1620,6 +1637,11 @@ 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); @@ -1942,6 +1964,7 @@ 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); @@ -1953,9 +1976,12 @@ 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); @@ -2019,8 +2045,13 @@ 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; } @@ -2179,41 +2210,36 @@ 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; } @@ -2225,8 +2251,13 @@ 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: @@ -2676,6 +2707,11 @@ 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. @@ -2928,6 +2964,12 @@ 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. */