From: Jérémie Galarneau Date: Mon, 23 Apr 2018 23:03:16 +0000 (-0400) Subject: Fix: failure to launch agent thread is not reported X-Git-Url: http://git.efficios.com/?p=lttng-tools.git;a=commitdiff_plain;h=25258405ed9590fcbcfb02c10dd8c018dba3a7bd Fix: failure to launch agent thread is not reported A session daemon may fail to launch its agent thread. In such a case, the tracing of agent domains fails silently as events never get enabled through the agent. The problem that was reported was caused by a second session daemon being already bound on the agent TCP socket port, which prevented the launch of the agent thread. While in this situation tracing is still not possible, the user will at least get an error indicating as such when enabling an event in those domains. Reported-by: Deborah Barnard Signed-off-by: Jérémie Galarneau --- diff --git a/include/lttng/lttng-error.h b/include/lttng/lttng-error.h index 1b5ea699a..e6ad3d262 100644 --- a/include/lttng/lttng-error.h +++ b/include/lttng/lttng-error.h @@ -149,6 +149,7 @@ enum lttng_error_code { LTTNG_ERR_TRIGGER_EXISTS = 126, /* Trigger already registered. */ LTTNG_ERR_TRIGGER_NOT_FOUND = 127, /* Trigger not found. */ LTTNG_ERR_COMMAND_CANCELLED = 128, /* Command cancelled. */ + LTTNG_ERR_AGENT_TRACING_DISABLED = 139, /* Agent tracing disabled. */ /* MUST be last element */ LTTNG_ERR_NR, /* Last element */ diff --git a/src/bin/lttng-sessiond/agent-thread.c b/src/bin/lttng-sessiond/agent-thread.c index ddc6c7e62..d53e0b1c7 100644 --- a/src/bin/lttng-sessiond/agent-thread.c +++ b/src/bin/lttng-sessiond/agent-thread.c @@ -32,6 +32,8 @@ #include "session.h" #include "utils.h" +static int agent_tracing_enabled = -1; + /* * Note that there is not port here. It's set after this URI is parsed so we * can let the user define a custom one. However, localhost is ALWAYS the @@ -223,6 +225,15 @@ error: return ret; } +bool agent_tracing_is_enabled(void) +{ + int enabled; + + enabled = uatomic_read(&agent_tracing_enabled); + assert(enabled != -1); + return enabled == 1; +} + /* * This thread manage application notify communication. */ @@ -248,6 +259,12 @@ void *agent_thread_manage_registration(void *data) } reg_sock = init_tcp_socket(); + uatomic_set(&agent_tracing_enabled, !!reg_sock); + + /* + * Signal that the agent thread is ready. The command thread + * may start to query whether or not agent tracing is enabled. + */ sessiond_notify_ready(); if (!reg_sock) { goto error_tcp_socket; @@ -348,6 +365,7 @@ restart: } exit: + uatomic_set(&agent_tracing_enabled, 0); /* Whatever happens, try to delete it and exit. */ (void) lttng_poll_del(&events, reg_sock->fd); error: diff --git a/src/bin/lttng-sessiond/agent-thread.h b/src/bin/lttng-sessiond/agent-thread.h index 0a0b3682f..6bdfdcd5e 100644 --- a/src/bin/lttng-sessiond/agent-thread.h +++ b/src/bin/lttng-sessiond/agent-thread.h @@ -15,12 +15,15 @@ * Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. */ +#include + #ifndef LTTNG_SESSIOND_AGENT_THREAD_H #define LTTNG_SESSIOND_AGENT_THREAD_H #ifdef HAVE_LIBLTTNG_UST_CTL void *agent_thread_manage_registration(void *data); +bool agent_tracing_is_enabled(void); #else /* HAVE_LIBLTTNG_UST_CTL */ @@ -29,6 +32,11 @@ void *agent_thread_manage_registration(void *data) { return NULL; } +static inline +bool agent_tracing_is_enabled(void) +{ + return false; +} #endif /* HAVE_LIBLTTNG_UST_CTL */ diff --git a/src/bin/lttng-sessiond/cmd.c b/src/bin/lttng-sessiond/cmd.c index 081648889..91040ec1e 100644 --- a/src/bin/lttng-sessiond/cmd.c +++ b/src/bin/lttng-sessiond/cmd.c @@ -51,6 +51,7 @@ #include "buffer-registry.h" #include "notification-thread.h" #include "notification-thread-commands.h" +#include "agent-thread.h" #include "cmd.h" @@ -1380,9 +1381,15 @@ int cmd_enable_channel(struct ltt_session *session, break; } case LTTNG_DOMAIN_UST: + break; case LTTNG_DOMAIN_JUL: case LTTNG_DOMAIN_LOG4J: case LTTNG_DOMAIN_PYTHON: + if (!agent_tracing_is_enabled()) { + DBG("Attempted to enable a channel in an agent domain but the agent thread is not running"); + ret = LTTNG_ERR_AGENT_TRACING_DISABLED; + goto error; + } break; default: ret = LTTNG_ERR_UNKNOWN_DOMAIN; @@ -2090,6 +2097,12 @@ static int _cmd_enable_event(struct ltt_session *session, assert(usess); + if (!agent_tracing_is_enabled()) { + DBG("Attempted to enable an event in an agent domain but the agent thread is not running"); + ret = LTTNG_ERR_AGENT_TRACING_DISABLED; + goto error; + } + agt = trace_ust_find_agent(usess, domain->type); if (!agt) { agt = agent_create(domain->type); diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c index 602db1ee4..3feb32989 100644 --- a/src/bin/lttng-sessiond/main.c +++ b/src/bin/lttng-sessiond/main.c @@ -4287,12 +4287,41 @@ static void *thread_manage_clients(void *data) } sessiond_notify_ready(); + ret = sem_post(&load_info->message_thread_ready); if (ret) { PERROR("sem_post message_thread_ready"); goto error; } + /* + * Wait until all support threads are initialized before accepting + * commands. + */ + while (uatomic_read(<tng_sessiond_ready) != 0) { + fd_set read_fds; + struct timeval timeout; + + FD_ZERO(&read_fds); + FD_SET(thread_quit_pipe[0], &read_fds); + memset(&timeout, 0, sizeof(timeout)); + timeout.tv_usec = 1000; + + /* + * If a support thread failed to launch, it may signal that + * we must exit and the sessiond would never be marked as + * "ready". + * + * The timeout is set to 1ms, which serves as a way to + * pace down this check. + */ + ret = select(thread_quit_pipe[0] + 1, &read_fds, NULL, NULL, + &timeout); + if (ret > 0 || (ret < 0 && errno != EINTR)) { + goto exit; + } + } + /* This testpoint is after we signal readiness to the parent. */ if (testpoint(sessiond_thread_manage_clients)) { goto error; diff --git a/src/common/error.c b/src/common/error.c index db3042318..eed6aecaf 100644 --- a/src/common/error.c +++ b/src/common/error.c @@ -190,6 +190,7 @@ static const char *error_string_array[] = { [ ERROR_INDEX(LTTNG_ERR_TRIGGER_EXISTS) ] = "Trigger already registered", [ ERROR_INDEX(LTTNG_ERR_TRIGGER_NOT_FOUND) ] = "Trigger not found", [ ERROR_INDEX(LTTNG_ERR_COMMAND_CANCELLED) ] = "Command cancelled", + [ ERROR_INDEX(LTTNG_ERR_AGENT_TRACING_DISABLED) ] = "Session daemon agent tracing is disabled", /* Last element */ [ ERROR_INDEX(LTTNG_ERR_NR) ] = "Unknown error code"