From: David Goulet Date: Thu, 14 Jul 2011 17:19:49 +0000 (-0400) Subject: Add CPU hotplug support X-Git-Tag: v2.0-pre1~36 X-Git-Url: https://git.lttng.org./?a=commitdiff_plain;h=7a485870536daad1f6c5aeb1ae18783a8862cbee;p=lttng-tools.git Add CPU hotplug support This introduce a new thread to the session daemon that poll() on all channel's fd for an event which indicate a new CPU. If an even is catched, new streams are created for all channels of all tracing sessions and kernel consumers are updated. Signed-off-by: David Goulet --- diff --git a/ltt-sessiond/main.c b/ltt-sessiond/main.c index 10e13e387..3426b5cde 100644 --- a/ltt-sessiond/main.c +++ b/ltt-sessiond/main.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include @@ -74,6 +75,7 @@ static int opt_daemon; static int is_root; /* Set to 1 if the daemon is running as root */ static pid_t ppid; /* Parent PID for --sig-parent option */ static pid_t kconsumerd_pid; +static struct pollfd *kernel_pollfd; static char apps_unix_sock_path[PATH_MAX]; /* Global application Unix socket path */ static char client_unix_sock_path[PATH_MAX]; /* Global client Unix socket path */ @@ -86,11 +88,13 @@ static int apps_sock; static int kconsumerd_err_sock; static int kconsumerd_cmd_sock; static int kernel_tracer_fd; +static int kernel_poll_pipe[2]; /* Pthread, Mutexes and Semaphores */ static pthread_t kconsumerd_thread; static pthread_t apps_thread; static pthread_t client_thread; +static pthread_t kernel_thread; static sem_t kconsumerd_sem; static pthread_mutex_t kconsumerd_pid_mutex; /* Mutex to control kconsumerd pid assignation */ @@ -133,6 +137,7 @@ static void cleanup() DBG("Terminating all threads"); pthread_cancel(client_thread); pthread_cancel(apps_thread); + pthread_cancel(kernel_thread); if (kconsumerd_pid != 0) { pthread_cancel(kconsumerd_thread); } @@ -163,6 +168,8 @@ static void cleanup() DBG("Closing kernel fd"); close(kernel_tracer_fd); + close(kernel_poll_pipe[0]); + close(kernel_poll_pipe[1]); } /* @@ -203,23 +210,24 @@ static void clean_command_ctx(struct command_ctx *cmd_ctx) } /* - * send_kconsumerd_fds + * send_kconsumerd_channel_fds * - * Send all stream fds of the kernel session to the consumer. + * Send all stream fds of kernel channel to the consumer. */ -static int send_kconsumerd_fds(int sock, struct ltt_kernel_session *session) +static int send_kconsumerd_channel_fds(int sock, struct ltt_kernel_channel *channel) { int ret; size_t nb_fd; struct ltt_kernel_stream *stream; - struct ltt_kernel_channel *chan; struct lttcomm_kconsumerd_header lkh; struct lttcomm_kconsumerd_msg lkm; - nb_fd = session->stream_count_global; + DBG("Sending fds of channel %s to kernel consumer", channel->channel->name); + + nb_fd = channel->stream_count; /* Setup header */ - lkh.payload_size = (nb_fd + 1) * sizeof(struct lttcomm_kconsumerd_msg); + lkh.payload_size = nb_fd * sizeof(struct lttcomm_kconsumerd_msg); lkh.cmd_type = ADD_STREAM; DBG("Sending kconsumerd header"); @@ -230,25 +238,11 @@ static int send_kconsumerd_fds(int sock, struct ltt_kernel_session *session) goto error; } - DBG("Sending metadata stream fd"); - - /* Send metadata stream fd first */ - lkm.fd = session->metadata_stream_fd; - lkm.state = ACTIVE_FD; - lkm.max_sb_size = session->metadata->conf->attr.subbuf_size; - strncpy(lkm.path_name, session->metadata->pathname, PATH_MAX); - - ret = lttcomm_send_fds_unix_sock(sock, &lkm, &lkm.fd, 1, sizeof(lkm)); - if (ret < 0) { - perror("send kconsumerd fd"); - goto error; - } - - cds_list_for_each_entry(chan, &session->channel_list.head, list) { - cds_list_for_each_entry(stream, &chan->stream_list.head, list) { + cds_list_for_each_entry(stream, &channel->stream_list.head, list) { + if (stream->fd != 0) { lkm.fd = stream->fd; lkm.state = stream->state; - lkm.max_sb_size = chan->channel->attr.subbuf_size; + lkm.max_sb_size = channel->channel->attr.subbuf_size; strncpy(lkm.path_name, stream->pathname, PATH_MAX); DBG("Sending fd %d to kconsumerd", lkm.fd); @@ -261,7 +255,7 @@ static int send_kconsumerd_fds(int sock, struct ltt_kernel_session *session) } } - DBG("Kconsumerd fds sent"); + DBG("Kconsumerd channel fds sent"); return 0; @@ -270,27 +264,54 @@ error: } /* - * create_trace_dir + * send_kconsumerd_fds * - * Create the trace output directory. + * Send all stream fds of the kernel session to the consumer. */ -static int create_trace_dir(struct ltt_kernel_session *session) +static int send_kconsumerd_fds(int sock, struct ltt_kernel_session *session) { int ret; struct ltt_kernel_channel *chan; + struct lttcomm_kconsumerd_header lkh; + struct lttcomm_kconsumerd_msg lkm; + + /* Setup header */ + lkh.payload_size = sizeof(struct lttcomm_kconsumerd_msg); + lkh.cmd_type = ADD_STREAM; + + DBG("Sending kconsumerd header for metadata"); + + ret = lttcomm_send_unix_sock(sock, &lkh, sizeof(struct lttcomm_kconsumerd_header)); + if (ret < 0) { + perror("send kconsumerd header"); + goto error; + } + + DBG("Sending metadata stream fd"); + + if (session->metadata_stream_fd != 0) { + /* Send metadata stream fd first */ + lkm.fd = session->metadata_stream_fd; + lkm.state = ACTIVE_FD; + lkm.max_sb_size = session->metadata->conf->attr.subbuf_size; + strncpy(lkm.path_name, session->metadata->pathname, PATH_MAX); + + ret = lttcomm_send_fds_unix_sock(sock, &lkm, &lkm.fd, 1, sizeof(lkm)); + if (ret < 0) { + perror("send kconsumerd fd"); + goto error; + } + } - /* Create all channel directories */ cds_list_for_each_entry(chan, &session->channel_list.head, list) { - DBG("Creating trace directory at %s", chan->pathname); - ret = mkdir_recursive(chan->pathname, S_IRWXU | S_IRWXG ); + ret = send_kconsumerd_channel_fds(sock, chan); if (ret < 0) { - if (ret != EEXIST) { - ERR("Trace directory creation error"); - goto error; - } + goto error; } } + DBG("Kconsumerd fds (metadata and channel streams) sent"); + return 0; error: @@ -395,6 +416,181 @@ error: return ret; } +/* + * update_kernel_pollfd + * + * Update the kernel pollfd set of all channel fd available over + * all tracing session. Add the wakeup pipe at the end of the set. + */ +static int update_kernel_pollfd(void) +{ + int i = 0; + unsigned int nb_fd = 1; + struct ltt_session *session; + struct ltt_kernel_channel *channel; + + DBG("Updating kernel_pollfd"); + + /* Get the number of channel of all kernel session */ + cds_list_for_each_entry(session, <t_session_list.head, list) { + if (session->kernel_session == NULL) { + continue; + } + nb_fd += session->kernel_session->channel_count; + } + + DBG("Resizing kernel_pollfd to size %d", nb_fd); + + kernel_pollfd = realloc(kernel_pollfd, nb_fd * sizeof(struct pollfd)); + if (kernel_pollfd == NULL) { + perror("malloc kernel_pollfd"); + goto error; + } + + cds_list_for_each_entry(session, <t_session_list.head, list) { + if (session->kernel_session == NULL) { + continue; + } + if (i >= nb_fd) { + ERR("To much channel for kernel_pollfd size"); + break; + } + cds_list_for_each_entry(channel, &session->kernel_session->channel_list.head, list) { + kernel_pollfd[i].fd = channel->fd; + kernel_pollfd[i].events = POLLIN | POLLRDNORM; + i++; + } + } + + /* Adding wake up pipe */ + kernel_pollfd[nb_fd - 1].fd = kernel_poll_pipe[0]; + kernel_pollfd[nb_fd - 1].events = POLLIN; + + return nb_fd; + +error: + return -1; +} + +/* + * update_kernel_stream + * + * Find the channel fd from 'fd' over all tracing session. When found, check + * for new channel stream and send those stream fds to the kernel consumer. + * + * Useful for CPU hotplug feature. + */ +static int update_kernel_stream(int fd) +{ + int ret = 0; + struct ltt_session *session; + struct ltt_kernel_channel *channel; + + DBG("Updating kernel streams for channel fd %d", fd); + + cds_list_for_each_entry(session, <t_session_list.head, list) { + if (session->kernel_session == NULL) { + continue; + } + cds_list_for_each_entry(channel, &session->kernel_session->channel_list.head, list) { + if (channel->fd == fd) { + DBG("Channel found, updating kernel streams"); + ret = kernel_open_channel_stream(channel); + if (ret < 0) { + goto end; + } + /* + * Have we already sent fds to the consumer? If yes, it means that + * tracing is started so it is safe to send our updated stream fds. + */ + if (session->kernel_session->kconsumer_fds_sent == 1) { + ret = send_kconsumerd_channel_fds(kconsumerd_cmd_sock, channel); + if (ret < 0) { + goto end; + } + } + goto end; + } + } + } + +end: + return ret; +} + +/* + * thread_manage_kernel + * + * This thread manage event coming from the kernel. + * + * Features supported in this thread: + * -) CPU Hotplug + */ +static void *thread_manage_kernel(void *data) +{ + int ret, i, nb_fd = 0; + char tmp; + int update_poll_flag = 1; + + DBG("Thread manage kernel started"); + + while (1) { + if (update_poll_flag == 1) { + nb_fd = update_kernel_pollfd(); + if (nb_fd < 0) { + goto error; + } + update_poll_flag = 0; + } + + DBG("Polling on %d fds", nb_fd); + + /* Poll infinite value of time */ + ret = poll(kernel_pollfd, nb_fd, -1); + if (ret < 0) { + perror("poll kernel thread"); + goto error; + } else if (ret == 0) { + /* Should not happen since timeout is infinite */ + continue; + } + + DBG("Kernel poll event triggered"); + + /* + * Check if the wake up pipe was triggered. If so, the kernel_pollfd + * must be updated. + */ + if (kernel_pollfd[nb_fd - 1].revents == POLLIN) { + ret = read(kernel_poll_pipe[0], &tmp, 1); + update_poll_flag = 1; + continue; + } + + for (i = 0; i < nb_fd; i++) { + switch (kernel_pollfd[i].revents) { + /* + * New CPU detected by the kernel. Adding kernel stream to kernel + * session and updating the kernel consumer + */ + case POLLIN | POLLRDNORM: + ret = update_kernel_stream(kernel_pollfd[i].fd); + if (ret < 0) { + continue; + } + break; + } + } + } + +error: + DBG("Kernel thread dying"); + if (kernel_pollfd) { + free(kernel_pollfd); + } + return NULL; +} + /* * thread_manage_kconsumerd * @@ -785,17 +981,6 @@ static int start_kernel_trace(struct ltt_kernel_session *session) { int ret; - /* Create trace directory */ - ret = create_trace_dir(session); - if (ret < 0) { - if (ret == -EEXIST) { - ret = LTTCOMM_KERN_DIR_EXIST; - } else { - ret = LTTCOMM_KERN_DIR_FAIL; - goto error; - } - } - if (session->kconsumer_fds_sent == 0) { ret = send_kconsumerd_fds(kconsumerd_cmd_sock, session); if (ret < 0) { @@ -811,6 +996,22 @@ error: return ret; } +/* + * Notify kernel thread to update it's pollfd. + */ +static int notify_kernel_pollfd(void) +{ + int ret; + + /* Inform kernel thread of the new kernel channel */ + ret = write(kernel_poll_pipe[1], "!", 1); + if (ret < 0) { + perror("write kernel poll pipe"); + } + + return ret; +} + /* * init_default_channel * @@ -866,6 +1067,14 @@ static int create_kernel_session(struct ltt_session *session) goto error; } + ret = mkdir_recursive(session->path, S_IRWXU | S_IRWXG ); + if (ret < 0) { + if (ret != EEXIST) { + ERR("Trace directory creation error"); + goto error; + } + } + DBG("Creating default kernel channel %s", DEFAULT_CHANNEL_NAME); ret = kernel_create_channel(session->kernel_session, chan, session->path); @@ -874,6 +1083,8 @@ static int create_kernel_session(struct ltt_session *session) goto error; } + ret = notify_kernel_pollfd(); + error: return ret; } @@ -1066,6 +1277,12 @@ static int process_client_msg(struct command_ctx *cmd_ctx) goto error; } + ret = notify_kernel_pollfd(); + if (ret < 0) { + ret = LTTCOMM_FATAL; + goto error; + } + ret = LTTCOMM_OK; break; } @@ -1478,6 +1695,16 @@ static int process_client_msg(struct command_ctx *cmd_ctx) goto error; } + /* + * Must notify the kernel thread here to update it's pollfd in order to + * remove the channel(s)' fd just destroyed. + */ + ret = notify_kernel_pollfd(); + if (ret < 0) { + ret = LTTCOMM_FATAL; + goto error; + } + ret = LTTCOMM_OK; break; } @@ -1928,6 +2155,16 @@ end: return ret; } +/* + * create_kernel_poll_pipe + * + * Create the pipe used to wake up the kernel thread. + */ +static int create_kernel_poll_pipe(void) +{ + return pipe2(kernel_poll_pipe, O_CLOEXEC); +} + /* * create_lttng_rundir * @@ -2191,6 +2428,11 @@ int main(int argc, char **argv) ppid = getppid(); } + /* Setup the kernel pipe for waking up the kernel thread */ + if (create_kernel_poll_pipe() < 0) { + goto error; + } + while (1) { /* Create thread to manage the client socket */ ret = pthread_create(&client_thread, NULL, thread_manage_clients, (void *) NULL); @@ -2206,6 +2448,13 @@ int main(int argc, char **argv) goto error; } + /* Create kernel thread to manage kernel event */ + ret = pthread_create(&kernel_thread, NULL, thread_manage_kernel, (void *) NULL); + if (ret != 0) { + perror("pthread_create"); + goto error; + } + ret = pthread_join(client_thread, &status); if (ret != 0) { perror("pthread_join");