Merge branch 'master' into benchmark
authorDavid Goulet <dgoulet@efficios.com>
Fri, 27 Jan 2012 16:19:56 +0000 (11:19 -0500)
committerDavid Goulet <dgoulet@efficios.com>
Fri, 27 Jan 2012 16:19:56 +0000 (11:19 -0500)
Signed-off-by: David Goulet <dgoulet@efficios.com>
1  2 
.gitignore
Makefile.am
configure.ac
src/bin/lttng-sessiond/Makefile.am
src/bin/lttng-sessiond/main.c
src/bin/lttng-sessiond/shm.c

diff --cc .gitignore
index 79d0a02c5f587c3a00748b5a5063b1ef3bc9abce,0eacb03bb3cadf1245e14f455afe5af67a61b746..776aa78efb241e8a68bdbd3f79ab2fcae0d8de6f
@@@ -37,7 -36,12 +36,12 @@@ src/bin/lttng-consumerd/lttng-consumer
  
  tests/test_sessions
  tests/test_kernel_data_trace
+ tests/test_ust_data_trace
  tests/kernel_all_events_basic
  tests/kernel_event_basic
+ tests/ust_global_all_events_basic
+ tests/ust_global_event_basic
+ tests/ust-nevents/gen-nevents
+ tests/ust-nprocesses/gen-events-time
  
 -benchmark/
 +benchmark/bench_sessions
diff --cc Makefile.am
index fa7f5ec565a6a4944f55ac413826d488307efdd8,7184340a90f2e6c24fcf81600354decb309a3c22..35828f2db8c7bf8c1da4f7d6db5896aca1eb80a2
@@@ -1,24 -1,6 +1,10 @@@
  ACLOCAL_AMFLAGS = -I config
  
- SUBDIRS = benchmark \
-                 common \
-                 liblttng-sessiond-comm \
-                 libkernelctl \
-                 liblttng-ht \
-                 liblttng-kconsumer \
-                 liblttng-ustconsumer \
-                 liblttng-consumer \
-                 lttng-consumerd
- if ! BUILD_CONSUMERD_ONLY
- SUBDIRS += liblttngctl \
-            lttng \
-            lttng-sessiond
- endif
- SUBDIRS += tests \
-            include \
-            doc
+ SUBDIRS = src \
+                 tests \
+                 include \
++                benchmark \
+                 doc
 +
 +bench:
 +      ./benchmark/runall.sh
diff --cc configure.ac
index 68845123cc103b3976a77dfd06a53cdd9f936065,53f21ab098181b3e47f8aaa7c514e6a1a5f3fe49..bd65c55ee24b033f0595a9d9a77ff25c8f8c23f3
@@@ -136,21 -131,24 +136,25 @@@ AC_SUBST(DEFAULT_INCLUDES
  
  AC_CONFIG_FILES([
        Makefile
+       doc/Makefile
        include/Makefile
-       common/Makefile
-       libkernelctl/Makefile
-       liblttng-consumer/Makefile
-       liblttng-kconsumer/Makefile
-       liblttng-ustconsumer/Makefile
-       liblttngctl/Makefile
-       liblttng-sessiond-comm/Makefile
-       liblttng-ht/Makefile
-       lttng-consumerd/Makefile
-       lttng-sessiond/Makefile
-       lttng/Makefile
 +      benchmark/Makefile
+       src/Makefile
+       src/common/Makefile
+       src/common/kernel-ctl/Makefile
+       src/common/kernel-consumer/Makefile
+       src/common/ust-consumer/Makefile
+       src/common/hashtable/Makefile
+       src/common/sessiond-comm/Makefile
+       src/lib/Makefile
+       src/lib/lttng-ctl/Makefile
+       src/bin/Makefile
+       src/bin/lttng-consumerd/Makefile
+       src/bin/lttng-sessiond/Makefile
+       src/bin/lttng/Makefile
        tests/Makefile
-       doc/Makefile
+       tests/ust-nevents/Makefile
+       tests/ust-nprocesses/Makefile
  ])
  
  AC_OUTPUT
index 0000000000000000000000000000000000000000,88b7c6974f35acf77f733700a5b7b7544842bfa8..1e449476c6323b1cddf31b11f1b5735d76373985
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,39 +1,41 @@@
 -AM_CPPFLAGS = -DINSTALL_BIN_PATH=\""$(bindir)"\" \
++AM_CPPFLAGS = -I$(top_srcdir)/benchmark \
++                        -DINSTALL_BIN_PATH=\""$(bindir)"\" \
+                         -DINSTALL_LIB_PATH=\""$(libdir)"\"
+ AM_CFLAGS = -fno-strict-aliasing
+ bin_PROGRAMS = lttng-sessiond
+ lttng_sessiond_SOURCES = utils.c utils.h \
+                        trace-kernel.c trace-kernel.h \
+                        kernel.c kernel.h \
+                        ust-ctl.h ust-app.h trace-ust.h \
+                        context.c context.h \
+                        channel.c channel.h \
+                        event.c event.h \
+                        futex.c futex.h \
+                        shm.c shm.h \
+                        session.c session.h \
+                        modprobe.c modprobe.h kern-modules.h \
+                        lttng-ust-ctl.h lttng-ust-abi.h
+ if HAVE_LIBLTTNG_UST_CTL
+ lttng_sessiond_SOURCES += trace-ust.c ust-app.c ust-consumer.c ust-consumer.h
+ endif
+ # Add main.c at the end for compile order
+ lttng_sessiond_SOURCES += lttng-sessiond.h main.c
+ # link on liblttngctl for check if sessiond is already alive.
+ lttng_sessiond_LDADD = -lrt -lurcu-common -lurcu \
+               $(top_builddir)/src/lib/lttng-ctl/liblttng-ctl.la \
+               $(top_builddir)/src/common/sessiond-comm/libsessiond-comm.la \
+               $(top_builddir)/src/common/kernel-ctl/libkernel-ctl.la \
+               $(top_builddir)/src/common/hashtable/libhashtable.la \
+               $(top_builddir)/src/common/libcommon.la \
++              $(top_builddir)/benchmark/liblttng-benchmark.la \
+               $(top_builddir)/src/common/libcompat.la
+ if HAVE_LIBLTTNG_UST_CTL
+ lttng_sessiond_LDADD += -llttng-ust-ctl
+ endif
index 0000000000000000000000000000000000000000,631aa6b2d92e5e0d13aaa5a2d59072327b31eb69..acade53db2f3fa9b13d90f5245ef8fed77f07847
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,4474 +1,4548 @@@
 -
+ /*
+  * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+  *                      Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+  *
+  * This program is free software; you can redistribute it and/or modify it
+  * under the terms of the GNU General Public License as published by the Free
+  * Software Foundation; only version 2 of the License.
+  *
+  * This program is distributed in the hope that it will be useful, but WITHOUT
+  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+  * more details.
+  *
+  * You should have received a copy of the GNU General Public License along with
+  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+  * Place - Suite 330, Boston, MA  02111-1307, USA.
+  */
+ #define _GNU_SOURCE
+ #include <fcntl.h>
+ #include <getopt.h>
+ #include <grp.h>
+ #include <limits.h>
+ #include <pthread.h>
+ #include <semaphore.h>
+ #include <signal.h>
+ #include <stdio.h>
+ #include <stdlib.h>
+ #include <string.h>
+ #include <sys/mman.h>
+ #include <sys/mount.h>
+ #include <sys/resource.h>
+ #include <sys/socket.h>
+ #include <sys/stat.h>
+ #include <sys/types.h>
+ #include <sys/wait.h>
+ #include <urcu/futex.h>
+ #include <unistd.h>
+ #include <config.h>
+ #include <common/common.h>
+ #include <common/compat/poll.h>
+ #include <common/defaults.h>
+ #include <common/kernel-consumer/kernel-consumer.h>
+ #include <common/ust-consumer/ust-consumer.h>
+ #include "lttng-sessiond.h"
+ #include "channel.h"
+ #include "context.h"
+ #include "event.h"
+ #include "futex.h"
+ #include "kernel.h"
+ #include "modprobe.h"
+ #include "shm.h"
+ #include "ust-ctl.h"
+ #include "utils.h"
+ #define CONSUMERD_FILE        "lttng-consumerd"
+ struct consumer_data {
+       enum lttng_consumer_type type;
+       pthread_t thread;       /* Worker thread interacting with the consumer */
+       sem_t sem;
+       /* Mutex to control consumerd pid assignation */
+       pthread_mutex_t pid_mutex;
+       pid_t pid;
+       int err_sock;
+       int cmd_sock;
+       /* consumer error and command Unix socket path */
+       char err_unix_sock_path[PATH_MAX];
+       char cmd_unix_sock_path[PATH_MAX];
+ };
++#include "benchmark.h"
++
+ /* Const values */
+ const char default_home_dir[] = DEFAULT_HOME_DIR;
+ const char default_tracing_group[] = DEFAULT_TRACING_GROUP;
+ const char default_ust_sock_dir[] = DEFAULT_UST_SOCK_DIR;
+ const char default_global_apps_pipe[] = DEFAULT_GLOBAL_APPS_PIPE;
+ /* Variables */
+ int opt_verbose;    /* Not static for lttngerr.h */
+ int opt_verbose_consumer;    /* Not static for lttngerr.h */
+ int opt_quiet;      /* Not static for lttngerr.h */
+ const char *progname;
+ const char *opt_tracing_group;
+ static int opt_sig_parent;
+ static int opt_daemon;
+ static int opt_no_kernel;
+ 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 char *rundir;
+ /* Consumer daemon specific control data */
+ static struct consumer_data kconsumer_data = {
+       .type = LTTNG_CONSUMER_KERNEL,
+       .err_unix_sock_path = DEFAULT_KCONSUMERD_ERR_SOCK_PATH,
+       .cmd_unix_sock_path = DEFAULT_KCONSUMERD_CMD_SOCK_PATH,
+ };
+ static struct consumer_data ustconsumer64_data = {
+       .type = LTTNG_CONSUMER64_UST,
+       .err_unix_sock_path = DEFAULT_USTCONSUMERD64_ERR_SOCK_PATH,
+       .cmd_unix_sock_path = DEFAULT_USTCONSUMERD64_CMD_SOCK_PATH,
+ };
+ static struct consumer_data ustconsumer32_data = {
+       .type = LTTNG_CONSUMER32_UST,
+       .err_unix_sock_path = DEFAULT_USTCONSUMERD32_ERR_SOCK_PATH,
+       .cmd_unix_sock_path = DEFAULT_USTCONSUMERD32_CMD_SOCK_PATH,
+ };
+ static int dispatch_thread_exit;
+ /* Global application Unix socket path */
+ static char apps_unix_sock_path[PATH_MAX];
+ /* Global client Unix socket path */
+ static char client_unix_sock_path[PATH_MAX];
+ /* global wait shm path for UST */
+ static char wait_shm_path[PATH_MAX];
+ /* Sockets and FDs */
+ static int client_sock;
+ static int apps_sock;
+ static int kernel_tracer_fd;
+ static int kernel_poll_pipe[2];
+ /*
+  * Quit pipe for all threads. This permits a single cancellation point
+  * for all threads when receiving an event on the pipe.
+  */
+ static int thread_quit_pipe[2];
+ /*
+  * This pipe is used to inform the thread managing application communication
+  * that a command is queued and ready to be processed.
+  */
+ static int apps_cmd_pipe[2];
+ /* Pthread, Mutexes and Semaphores */
+ static pthread_t apps_thread;
+ static pthread_t reg_apps_thread;
+ static pthread_t client_thread;
+ static pthread_t kernel_thread;
+ static pthread_t dispatch_thread;
+ /*
+  * UST registration command queue. This queue is tied with a futex and uses a N
+  * wakers / 1 waiter implemented and detailed in futex.c/.h
+  *
+  * The thread_manage_apps and thread_dispatch_ust_registration interact with
+  * this queue and the wait/wake scheme.
+  */
+ static struct ust_cmd_queue ust_cmd_queue;
+ /*
+  * Pointer initialized before thread creation.
+  *
+  * This points to the tracing session list containing the session count and a
+  * mutex lock. The lock MUST be taken if you iterate over the list. The lock
+  * MUST NOT be taken if you call a public function in session.c.
+  *
+  * The lock is nested inside the structure: session_list_ptr->lock. Please use
+  * session_lock_list and session_unlock_list for lock acquisition.
+  */
+ static struct ltt_session_list *session_list_ptr;
+ int ust_consumerd64_fd = -1;
+ int ust_consumerd32_fd = -1;
+ static const char *consumerd32_bin = CONFIG_CONSUMERD32_BIN;
+ static const char *consumerd64_bin = CONFIG_CONSUMERD64_BIN;
+ static const char *consumerd32_libdir = CONFIG_CONSUMERD32_LIBDIR;
+ static const char *consumerd64_libdir = CONFIG_CONSUMERD64_LIBDIR;
+ static
+ void setup_consumerd_path(void)
+ {
+       const char *bin, *libdir;
+       /*
+        * Allow INSTALL_BIN_PATH to be used as a target path for the
+        * native architecture size consumer if CONFIG_CONSUMER*_PATH
+        * has not been defined.
+        */
+ #if (CAA_BITS_PER_LONG == 32)
+       if (!consumerd32_bin[0]) {
+               consumerd32_bin = INSTALL_BIN_PATH "/" CONSUMERD_FILE;
+       }
+       if (!consumerd32_libdir[0]) {
+               consumerd32_libdir = INSTALL_LIB_PATH;
+       }
+ #elif (CAA_BITS_PER_LONG == 64)
+       if (!consumerd64_bin[0]) {
+               consumerd64_bin = INSTALL_BIN_PATH "/" CONSUMERD_FILE;
+       }
+       if (!consumerd64_libdir[0]) {
+               consumerd64_libdir = INSTALL_LIB_PATH;
+       }
+ #else
+ #error "Unknown bitness"
+ #endif
+       /*
+        * runtime env. var. overrides the build default.
+        */
+       bin = getenv("LTTNG_CONSUMERD32_BIN");
+       if (bin) {
+               consumerd32_bin = bin;
+       }
+       bin = getenv("LTTNG_CONSUMERD64_BIN");
+       if (bin) {
+               consumerd64_bin = bin;
+       }
+       libdir = getenv("LTTNG_CONSUMERD32_LIBDIR");
+       if (libdir) {
+               consumerd32_libdir = libdir;
+       }
+       libdir = getenv("LTTNG_CONSUMERD64_LIBDIR");
+       if (libdir) {
+               consumerd64_libdir = libdir;
+       }
+ }
+ /*
+  * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set.
+  */
+ static int create_thread_poll_set(struct lttng_poll_event *events,
+               unsigned int size)
+ {
+       int ret;
+       if (events == NULL || size == 0) {
+               ret = -1;
+               goto error;
+       }
+       ret = lttng_poll_create(events, size, LTTNG_CLOEXEC);
+       if (ret < 0) {
+               goto error;
+       }
+       /* Add quit pipe */
+       ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN);
+       if (ret < 0) {
+               goto error;
+       }
+       return 0;
+ error:
+       return ret;
+ }
+ /*
+  * Check if the thread quit pipe was triggered.
+  *
+  * Return 1 if it was triggered else 0;
+  */
+ static int check_thread_quit_pipe(int fd, uint32_t events)
+ {
+       if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) {
+               return 1;
+       }
+       return 0;
+ }
+ /*
+  * Return group ID of the tracing group or -1 if not found.
+  */
+ static gid_t allowed_group(void)
+ {
+       struct group *grp;
+       if (opt_tracing_group) {
+               grp = getgrnam(opt_tracing_group);
+       } else {
+               grp = getgrnam(default_tracing_group);
+       }
+       if (!grp) {
+               return -1;
+       } else {
+               return grp->gr_gid;
+       }
+ }
+ /*
+  * Init thread quit pipe.
+  *
+  * Return -1 on error or 0 if all pipes are created.
+  */
+ static int init_thread_quit_pipe(void)
+ {
+       int ret;
+       ret = pipe2(thread_quit_pipe, O_CLOEXEC);
+       if (ret < 0) {
+               perror("thread quit pipe");
+               goto error;
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Complete teardown of a kernel session. This free all data structure related
+  * to a kernel session and update counter.
+  */
+ static void teardown_kernel_session(struct ltt_session *session)
+ {
+       if (!session->kernel_session) {
+               DBG3("No kernel session when tearingdown session");
+               return;
+       }
+       DBG("Tearing down kernel session");
+       /*
+        * If a custom kernel consumer was registered, close the socket before
+        * tearing down the complete kernel session structure
+        */
+       if (session->kernel_session->consumer_fd != kconsumer_data.cmd_sock) {
+               lttcomm_close_unix_sock(session->kernel_session->consumer_fd);
+       }
+       trace_kernel_destroy_session(session->kernel_session);
+ }
+ /*
+  * Complete teardown of all UST sessions. This will free everything on his path
+  * and destroy the core essence of all ust sessions :)
+  */
+ static void teardown_ust_session(struct ltt_session *session)
+ {
+       int ret;
+       if (!session->ust_session) {
+               DBG3("No UST session when tearingdown session");
+               return;
+       }
+       DBG("Tearing down UST session(s)");
+       ret = ust_app_destroy_trace_all(session->ust_session);
+       if (ret) {
+               ERR("Error in ust_app_destroy_trace_all");
+       }
+       trace_ust_destroy_session(session->ust_session);
+ }
+ /*
+  * Stop all threads by closing the thread quit pipe.
+  */
+ static void stop_threads(void)
+ {
+       int ret;
+       /* Stopping all threads */
+       DBG("Terminating all threads");
+       ret = notify_thread_pipe(thread_quit_pipe[1]);
+       if (ret < 0) {
+               ERR("write error on thread quit pipe");
+       }
+       /* Dispatch thread */
+       dispatch_thread_exit = 1;
+       futex_nto1_wake(&ust_cmd_queue.futex);
+ }
+ /*
+  * Cleanup the daemon
+  */
+ static void cleanup(void)
+ {
+       int ret;
+       char *cmd;
+       struct ltt_session *sess, *stmp;
+       DBG("Cleaning up");
+       DBG("Removing %s directory", rundir);
+       ret = asprintf(&cmd, "rm -rf %s", rundir);
+       if (ret < 0) {
+               ERR("asprintf failed. Something is really wrong!");
+       }
+       /* Remove lttng run directory */
+       ret = system(cmd);
+       if (ret < 0) {
+               ERR("Unable to clean %s", rundir);
+       }
+       free(cmd);
+       DBG("Cleaning up all session");
+       /* Destroy session list mutex */
+       if (session_list_ptr != NULL) {
+               pthread_mutex_destroy(&session_list_ptr->lock);
+               /* Cleanup ALL session */
+               cds_list_for_each_entry_safe(sess, stmp,
+                               &session_list_ptr->head, list) {
+                       teardown_kernel_session(sess);
+                       teardown_ust_session(sess);
+                       free(sess);
+               }
+       }
+       DBG("Closing all UST sockets");
+       ust_app_clean_list();
+       pthread_mutex_destroy(&kconsumer_data.pid_mutex);
+       if (is_root && !opt_no_kernel) {
+               DBG2("Closing kernel fd");
+               close(kernel_tracer_fd);
+               DBG("Unloading kernel modules");
+               modprobe_remove_lttng_all();
+       }
+       close(thread_quit_pipe[0]);
+       close(thread_quit_pipe[1]);
++      /* OUTPUT BENCHMARK RESULTS */
++      bench_init();
++
++      if (getenv("BENCH_UST_NOTIFY")) {
++              bench_print_ust_notification();
++      }
++
++      if (getenv("BENCH_UST_REGISTER")) {
++              bench_print_ust_register();
++              bench_print_ust_unregister();
++      }
++
++      if (getenv("BENCH_BOOT_PROCESS")) {
++              bench_print_boot_process();
++      }
++
++      bench_close();
++      /* END BENCHMARK */
++
+       /* <fun> */
+       DBG("%c[%d;%dm*** assert failed :-) *** ==> %c[%dm%c[%d;%dm"
+                       "Matthew, BEET driven development works!%c[%dm",
+                       27, 1, 31, 27, 0, 27, 1, 33, 27, 0);
+       /* </fun> */
+ }
+ /*
+  * Send data on a unix socket using the liblttsessiondcomm API.
+  *
+  * Return lttcomm error code.
+  */
+ static int send_unix_sock(int sock, void *buf, size_t len)
+ {
+       /* Check valid length */
+       if (len <= 0) {
+               return -1;
+       }
+       return lttcomm_send_unix_sock(sock, buf, len);
+ }
+ /*
+  * Free memory of a command context structure.
+  */
+ static void clean_command_ctx(struct command_ctx **cmd_ctx)
+ {
+       DBG("Clean command context structure");
+       if (*cmd_ctx) {
+               if ((*cmd_ctx)->llm) {
+                       free((*cmd_ctx)->llm);
+               }
+               if ((*cmd_ctx)->lsm) {
+                       free((*cmd_ctx)->lsm);
+               }
+               free(*cmd_ctx);
+               *cmd_ctx = NULL;
+       }
+ }
+ /*
+  * Send all stream fds of kernel channel to the consumer.
+  */
+ static int send_kconsumer_channel_streams(struct consumer_data *consumer_data,
+               int sock, struct ltt_kernel_channel *channel,
+               uid_t uid, gid_t gid)
+ {
+       int ret;
+       struct ltt_kernel_stream *stream;
+       struct lttcomm_consumer_msg lkm;
+       DBG("Sending streams of channel %s to kernel consumer",
+                       channel->channel->name);
+       /* Send channel */
+       lkm.cmd_type = LTTNG_CONSUMER_ADD_CHANNEL;
+       lkm.u.channel.channel_key = channel->fd;
+       lkm.u.channel.max_sb_size = channel->channel->attr.subbuf_size;
+       lkm.u.channel.mmap_len = 0;     /* for kernel */
+       DBG("Sending channel %d to consumer", lkm.u.channel.channel_key);
+       ret = lttcomm_send_unix_sock(sock, &lkm, sizeof(lkm));
+       if (ret < 0) {
+               perror("send consumer channel");
+               goto error;
+       }
+       /* Send streams */
+       cds_list_for_each_entry(stream, &channel->stream_list.head, list) {
+               if (!stream->fd) {
+                       continue;
+               }
+               lkm.cmd_type = LTTNG_CONSUMER_ADD_STREAM;
+               lkm.u.stream.channel_key = channel->fd;
+               lkm.u.stream.stream_key = stream->fd;
+               lkm.u.stream.state = stream->state;
+               lkm.u.stream.output = channel->channel->attr.output;
+               lkm.u.stream.mmap_len = 0;      /* for kernel */
+               lkm.u.stream.uid = uid;
+               lkm.u.stream.gid = gid;
+               strncpy(lkm.u.stream.path_name, stream->pathname, PATH_MAX - 1);
+               lkm.u.stream.path_name[PATH_MAX - 1] = '\0';
+               DBG("Sending stream %d to consumer", lkm.u.stream.stream_key);
+               ret = lttcomm_send_unix_sock(sock, &lkm, sizeof(lkm));
+               if (ret < 0) {
+                       perror("send consumer stream");
+                       goto error;
+               }
+               ret = lttcomm_send_fds_unix_sock(sock, &stream->fd, 1);
+               if (ret < 0) {
+                       perror("send consumer stream ancillary data");
+                       goto error;
+               }
+       }
+       DBG("consumer channel streams sent");
+       return 0;
+ error:
+       return ret;
+ }
+ /*
+  * Send all stream fds of the kernel session to the consumer.
+  */
+ static int send_kconsumer_session_streams(struct consumer_data *consumer_data,
+               struct ltt_kernel_session *session)
+ {
+       int ret;
+       struct ltt_kernel_channel *chan;
+       struct lttcomm_consumer_msg lkm;
+       int sock = session->consumer_fd;
+       DBG("Sending metadata stream fd");
+       /* Extra protection. It's NOT supposed to be set to 0 at this point */
+       if (session->consumer_fd == 0) {
+               session->consumer_fd = consumer_data->cmd_sock;
+       }
+       if (session->metadata_stream_fd != 0) {
+               /* Send metadata channel fd */
+               lkm.cmd_type = LTTNG_CONSUMER_ADD_CHANNEL;
+               lkm.u.channel.channel_key = session->metadata->fd;
+               lkm.u.channel.max_sb_size = session->metadata->conf->attr.subbuf_size;
+               lkm.u.channel.mmap_len = 0;     /* for kernel */
+               DBG("Sending metadata channel %d to consumer", lkm.u.stream.stream_key);
+               ret = lttcomm_send_unix_sock(sock, &lkm, sizeof(lkm));
+               if (ret < 0) {
+                       perror("send consumer channel");
+                       goto error;
+               }
+               /* Send metadata stream fd */
+               lkm.cmd_type = LTTNG_CONSUMER_ADD_STREAM;
+               lkm.u.stream.channel_key = session->metadata->fd;
+               lkm.u.stream.stream_key = session->metadata_stream_fd;
+               lkm.u.stream.state = LTTNG_CONSUMER_ACTIVE_STREAM;
+               lkm.u.stream.output = DEFAULT_KERNEL_CHANNEL_OUTPUT;
+               lkm.u.stream.mmap_len = 0;      /* for kernel */
+               lkm.u.stream.uid = session->uid;
+               lkm.u.stream.gid = session->gid;
+               strncpy(lkm.u.stream.path_name, session->metadata->pathname, PATH_MAX - 1);
+               lkm.u.stream.path_name[PATH_MAX - 1] = '\0';
+               DBG("Sending metadata stream %d to consumer", lkm.u.stream.stream_key);
+               ret = lttcomm_send_unix_sock(sock, &lkm, sizeof(lkm));
+               if (ret < 0) {
+                       perror("send consumer stream");
+                       goto error;
+               }
+               ret = lttcomm_send_fds_unix_sock(sock, &session->metadata_stream_fd, 1);
+               if (ret < 0) {
+                       perror("send consumer stream");
+                       goto error;
+               }
+       }
+       cds_list_for_each_entry(chan, &session->channel_list.head, list) {
+               ret = send_kconsumer_channel_streams(consumer_data, sock, chan,
+                               session->uid, session->gid);
+               if (ret < 0) {
+                       goto error;
+               }
+       }
+       DBG("consumer fds (metadata and channel streams) sent");
+       return 0;
+ error:
+       return ret;
+ }
+ /*
+  * Notify UST applications using the shm mmap futex.
+  */
+ static int notify_ust_apps(int active)
+ {
+       char *wait_shm_mmap;
+       DBG("Notifying applications of session daemon state: %d", active);
++      tracepoint(ust_notify_apps_start);
++
+       /* See shm.c for this call implying mmap, shm and futex calls */
+       wait_shm_mmap = shm_ust_get_mmap(wait_shm_path, is_root);
+       if (wait_shm_mmap == NULL) {
+               goto error;
+       }
+       /* Wake waiting process */
+       futex_wait_update((int32_t *) wait_shm_mmap, active);
++      tracepoint(ust_notify_apps_stop);
++
+       /* Apps notified successfully */
+       return 0;
+ error:
+       return -1;
+ }
+ /*
+  * Setup the outgoing data buffer for the response (llm) by allocating the
+  * right amount of memory and copying the original information from the lsm
+  * structure.
+  *
+  * Return total size of the buffer pointed by buf.
+  */
+ static int setup_lttng_msg(struct command_ctx *cmd_ctx, size_t size)
+ {
+       int ret, buf_size;
+       buf_size = size;
+       cmd_ctx->llm = zmalloc(sizeof(struct lttcomm_lttng_msg) + buf_size);
+       if (cmd_ctx->llm == NULL) {
+               perror("zmalloc");
+               ret = -ENOMEM;
+               goto error;
+       }
+       /* Copy common data */
+       cmd_ctx->llm->cmd_type = cmd_ctx->lsm->cmd_type;
+       cmd_ctx->llm->pid = cmd_ctx->lsm->domain.attr.pid;
+       cmd_ctx->llm->data_size = size;
+       cmd_ctx->lttng_msg_size = sizeof(struct lttcomm_lttng_msg) + buf_size;
+       return buf_size;
+ error:
+       return ret;
+ }
+ /*
+  * Update the kernel poll set of all channel fd available over all tracing
+  * session. Add the wakeup pipe at the end of the set.
+  */
+ static int update_kernel_poll(struct lttng_poll_event *events)
+ {
+       int ret;
+       struct ltt_session *session;
+       struct ltt_kernel_channel *channel;
+       DBG("Updating kernel poll set");
+       session_lock_list();
+       cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+               session_lock(session);
+               if (session->kernel_session == NULL) {
+                       session_unlock(session);
+                       continue;
+               }
+               cds_list_for_each_entry(channel,
+                               &session->kernel_session->channel_list.head, list) {
+                       /* Add channel fd to the kernel poll set */
+                       ret = lttng_poll_add(events, channel->fd, LPOLLIN | LPOLLRDNORM);
+                       if (ret < 0) {
+                               session_unlock(session);
+                               goto error;
+                       }
+                       DBG("Channel fd %d added to kernel set", channel->fd);
+               }
+               session_unlock(session);
+       }
+       session_unlock_list();
+       return 0;
+ error:
+       session_unlock_list();
+       return -1;
+ }
+ /*
+  * 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(struct consumer_data *consumer_data, int fd)
+ {
+       int ret = 0;
+       struct ltt_session *session;
+       struct ltt_kernel_channel *channel;
+       DBG("Updating kernel streams for channel fd %d", fd);
+       session_lock_list();
+       cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+               session_lock(session);
+               if (session->kernel_session == NULL) {
+                       session_unlock(session);
+                       continue;
+               }
+               /* This is not suppose to be 0 but this is an extra security check */
+               if (session->kernel_session->consumer_fd == 0) {
+                       session->kernel_session->consumer_fd = consumer_data->cmd_sock;
+               }
+               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 error;
+                               }
+                               /*
+                                * 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->consumer_fds_sent == 1) {
+                                       ret = send_kconsumer_channel_streams(consumer_data,
+                                                       session->kernel_session->consumer_fd, channel,
+                                                       session->uid, session->gid);
+                                       if (ret < 0) {
+                                               goto error;
+                                       }
+                               }
+                               goto error;
+                       }
+               }
+               session_unlock(session);
+       }
+       session_unlock_list();
+       return ret;
+ error:
+       session_unlock(session);
+       session_unlock_list();
+       return ret;
+ }
+ /*
+  * For each tracing session, update newly registered apps.
+  */
+ static void update_ust_app(int app_sock)
+ {
+       struct ltt_session *sess, *stmp;
+       /* For all tracing session(s) */
+       cds_list_for_each_entry_safe(sess, stmp, &session_list_ptr->head, list) {
+               if (sess->ust_session) {
+                       ust_app_global_update(sess->ust_session, app_sock);
+               }
+       }
+ }
+ /*
+  * 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, pollfd, update_poll_flag = 1;
+       uint32_t revents, nb_fd;
+       char tmp;
+       struct lttng_poll_event events;
++      tracepoint(sessiond_th_kern_start);
++
+       DBG("Thread manage kernel started");
+       ret = create_thread_poll_set(&events, 2);
+       if (ret < 0) {
+               goto error;
+       }
+       ret = lttng_poll_add(&events, kernel_poll_pipe[0], LPOLLIN);
+       if (ret < 0) {
+               goto error;
+       }
+       while (1) {
+               if (update_poll_flag == 1) {
+                       /*
+                        * Reset number of fd in the poll set. Always 2 since there is the thread
+                        * quit pipe and the kernel pipe.
+                        */
+                       events.nb_fd = 2;
+                       ret = update_kernel_poll(&events);
+                       if (ret < 0) {
+                               goto error;
+                       }
+                       update_poll_flag = 0;
+               }
+               nb_fd = LTTNG_POLL_GETNB(&events);
+               DBG("Thread kernel polling on %d fds", nb_fd);
+               /* Zeroed the poll events */
+               lttng_poll_reset(&events);
++              tracepoint(sessiond_th_kern_poll);
++
+               /* Poll infinite value of time */
+               ret = lttng_poll_wait(&events, -1);
+               if (ret < 0) {
+                       goto error;
+               } else if (ret == 0) {
+                       /* Should not happen since timeout is infinite */
+                       ERR("Return value of poll is 0 with an infinite timeout.\n"
+                               "This should not have happened! Continuing...");
+                       continue;
+               }
+               for (i = 0; i < nb_fd; i++) {
+                       /* Fetch once the poll data */
+                       revents = LTTNG_POLL_GETEV(&events, i);
+                       pollfd = LTTNG_POLL_GETFD(&events, i);
+                       /* Thread quit pipe has been closed. Killing thread. */
+                       ret = check_thread_quit_pipe(pollfd, revents);
+                       if (ret) {
+                               goto error;
+                       }
+                       /* Check for data on kernel pipe */
+                       if (pollfd == kernel_poll_pipe[0] && (revents & LPOLLIN)) {
+                               ret = read(kernel_poll_pipe[0], &tmp, 1);
+                               update_poll_flag = 1;
+                               continue;
+                       } else {
+                               /*
+                                * New CPU detected by the kernel. Adding kernel stream to
+                                * kernel session and updating the kernel consumer
+                                */
+                               if (revents & LPOLLIN) {
+                                       ret = update_kernel_stream(&kconsumer_data, pollfd);
+                                       if (ret < 0) {
+                                               continue;
+                                       }
+                                       break;
+                                       /*
+                                        * TODO: We might want to handle the LPOLLERR | LPOLLHUP
+                                        * and unregister kernel stream at this point.
+                                        */
+                               }
+                       }
+               }
+       }
+ error:
+       DBG("Kernel thread dying");
+       close(kernel_poll_pipe[0]);
+       close(kernel_poll_pipe[1]);
+       lttng_poll_clean(&events);
+       return NULL;
+ }
+ /*
+  * This thread manage the consumer error sent back to the session daemon.
+  */
+ static void *thread_manage_consumer(void *data)
+ {
+       int sock = 0, i, ret, pollfd;
+       uint32_t revents, nb_fd;
+       enum lttcomm_return_code code;
+       struct lttng_poll_event events;
+       struct consumer_data *consumer_data = data;
++      tracepoint(sessiond_th_kcon_start);
++
+       DBG("[thread] Manage consumer started");
+       ret = lttcomm_listen_unix_sock(consumer_data->err_sock);
+       if (ret < 0) {
+               goto error;
+       }
+       /*
+        * Pass 2 as size here for the thread quit pipe and kconsumerd_err_sock.
+        * Nothing more will be added to this poll set.
+        */
+       ret = create_thread_poll_set(&events, 2);
+       if (ret < 0) {
+               goto error;
+       }
+       ret = lttng_poll_add(&events, consumer_data->err_sock, LPOLLIN | LPOLLRDHUP);
+       if (ret < 0) {
+               goto error;
+       }
+       nb_fd = LTTNG_POLL_GETNB(&events);
++      tracepoint(sessiond_th_kcon_poll);
++
+       /* Inifinite blocking call, waiting for transmission */
+       ret = lttng_poll_wait(&events, -1);
+       if (ret < 0) {
+               goto error;
+       }
+       for (i = 0; i < nb_fd; i++) {
+               /* Fetch once the poll data */
+               revents = LTTNG_POLL_GETEV(&events, i);
+               pollfd = LTTNG_POLL_GETFD(&events, i);
+               /* Thread quit pipe has been closed. Killing thread. */
+               ret = check_thread_quit_pipe(pollfd, revents);
+               if (ret) {
+                       goto error;
+               }
+               /* Event on the registration socket */
+               if (pollfd == consumer_data->err_sock) {
+                       if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                               ERR("consumer err socket poll error");
+                               goto error;
+                       }
+               }
+       }
+       sock = lttcomm_accept_unix_sock(consumer_data->err_sock);
+       if (sock < 0) {
+               goto error;
+       }
+       DBG2("Receiving code from consumer err_sock");
+       /* Getting status code from kconsumerd */
+       ret = lttcomm_recv_unix_sock(sock, &code,
+                       sizeof(enum lttcomm_return_code));
+       if (ret <= 0) {
+               goto error;
+       }
+       if (code == CONSUMERD_COMMAND_SOCK_READY) {
+               consumer_data->cmd_sock =
+                       lttcomm_connect_unix_sock(consumer_data->cmd_unix_sock_path);
+               if (consumer_data->cmd_sock < 0) {
+                       sem_post(&consumer_data->sem);
+                       PERROR("consumer connect");
+                       goto error;
+               }
+               /* Signal condition to tell that the kconsumerd is ready */
+               sem_post(&consumer_data->sem);
+               DBG("consumer command socket ready");
+       } else {
+               ERR("consumer error when waiting for SOCK_READY : %s",
+                               lttcomm_get_readable_code(-code));
+               goto error;
+       }
+       /* Remove the kconsumerd error sock since we've established a connexion */
+       ret = lttng_poll_del(&events, consumer_data->err_sock);
+       if (ret < 0) {
+               goto error;
+       }
+       ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLRDHUP);
+       if (ret < 0) {
+               goto error;
+       }
+       /* Update number of fd */
+       nb_fd = LTTNG_POLL_GETNB(&events);
+       /* Inifinite blocking call, waiting for transmission */
+       ret = lttng_poll_wait(&events, -1);
+       if (ret < 0) {
+               goto error;
+       }
+       for (i = 0; i < nb_fd; i++) {
+               /* Fetch once the poll data */
+               revents = LTTNG_POLL_GETEV(&events, i);
+               pollfd = LTTNG_POLL_GETFD(&events, i);
+               /* Thread quit pipe has been closed. Killing thread. */
+               ret = check_thread_quit_pipe(pollfd, revents);
+               if (ret) {
+                       goto error;
+               }
+               /* Event on the kconsumerd socket */
+               if (pollfd == sock) {
+                       if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                               ERR("consumer err socket second poll error");
+                               goto error;
+                       }
+               }
+       }
+       /* Wait for any kconsumerd error */
+       ret = lttcomm_recv_unix_sock(sock, &code,
+                       sizeof(enum lttcomm_return_code));
+       if (ret <= 0) {
+               ERR("consumer closed the command socket");
+               goto error;
+       }
+       ERR("consumer return code : %s", lttcomm_get_readable_code(-code));
+ error:
+       DBG("consumer thread dying");
+       close(consumer_data->err_sock);
+       close(consumer_data->cmd_sock);
+       close(sock);
+       unlink(consumer_data->err_unix_sock_path);
+       unlink(consumer_data->cmd_unix_sock_path);
+       consumer_data->pid = 0;
+       lttng_poll_clean(&events);
+       return NULL;
+ }
+ /*
+  * This thread manage application communication.
+  */
+ static void *thread_manage_apps(void *data)
+ {
+       int i, ret, pollfd;
+       uint32_t revents, nb_fd;
+       struct ust_command ust_cmd;
+       struct lttng_poll_event events;
++      tracepoint(sessiond_th_apps_start);
++
+       DBG("[thread] Manage application started");
+       rcu_register_thread();
+       rcu_thread_online();
+       ret = create_thread_poll_set(&events, 2);
+       if (ret < 0) {
+               goto error;
+       }
+       ret = lttng_poll_add(&events, apps_cmd_pipe[0], LPOLLIN | LPOLLRDHUP);
+       if (ret < 0) {
+               goto error;
+       }
+       while (1) {
+               /* Zeroed the events structure */
+               lttng_poll_reset(&events);
+               nb_fd = LTTNG_POLL_GETNB(&events);
+               DBG("Apps thread polling on %d fds", nb_fd);
++              tracepoint(sessiond_th_apps_poll);
++
+               /* Inifinite blocking call, waiting for transmission */
+               ret = lttng_poll_wait(&events, -1);
+               if (ret < 0) {
+                       goto error;
+               }
+               for (i = 0; i < nb_fd; i++) {
+                       /* Fetch once the poll data */
+                       revents = LTTNG_POLL_GETEV(&events, i);
+                       pollfd = LTTNG_POLL_GETFD(&events, i);
+                       /* Thread quit pipe has been closed. Killing thread. */
+                       ret = check_thread_quit_pipe(pollfd, revents);
+                       if (ret) {
+                               goto error;
+                       }
+                       /* Inspect the apps cmd pipe */
+                       if (pollfd == apps_cmd_pipe[0]) {
+                               if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                                       ERR("Apps command pipe error");
+                                       goto error;
+                               } else if (revents & LPOLLIN) {
++                                      tracepoint(ust_register_read_start);
+                                       /* Empty pipe */
+                                       ret = read(apps_cmd_pipe[0], &ust_cmd, sizeof(ust_cmd));
+                                       if (ret < 0 || ret < sizeof(ust_cmd)) {
+                                               perror("read apps cmd pipe");
+                                               goto error;
+                                       }
++                                      tracepoint(ust_register_read_stop);
++                                      tracepoint(ust_register_add_start);
+                                       /* Register applicaton to the session daemon */
+                                       ret = ust_app_register(&ust_cmd.reg_msg,
+                                                       ust_cmd.sock);
+                                       if (ret == -ENOMEM) {
+                                               goto error;
+                                       } else if (ret < 0) {
+                                               break;
+                                       }
++                                      tracepoint(ust_register_add_stop);
++                                      tracepoint(ust_register_done_start);
+                                       /*
+                                        * Validate UST version compatibility.
+                                        */
+                                       ret = ust_app_validate_version(ust_cmd.sock);
+                                       if (ret >= 0) {
+                                               /*
+                                                * Add channel(s) and event(s) to newly registered apps
+                                                * from lttng global UST domain.
+                                                */
+                                               update_ust_app(ust_cmd.sock);
+                                       }
+                                       ret = ust_app_register_done(ust_cmd.sock);
+                                       if (ret < 0) {
+                                               /*
+                                                * If the registration is not possible, we simply
+                                                * unregister the apps and continue
+                                                */
+                                               ust_app_unregister(ust_cmd.sock);
+                                       } else {
+                                               /*
+                                                * We just need here to monitor the close of the UST
+                                                * socket and poll set monitor those by default.
+                                                */
+                                               ret = lttng_poll_add(&events, ust_cmd.sock, 0);
+                                               if (ret < 0) {
+                                                       goto error;
+                                               }
+                                               DBG("Apps with sock %d added to poll set",
+                                                               ust_cmd.sock);
+                                       }
++                                      tracepoint(ust_register_done_stop);
+                                       break;
+                               }
+                       } else {
+                               /*
+                                * At this point, we know that a registered application made
+                                * the event at poll_wait.
+                                */
+                               if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
++                                      tracepoint(ust_unregister_start);
++
+                                       /* Removing from the poll set */
+                                       ret = lttng_poll_del(&events, pollfd);
+                                       if (ret < 0) {
+                                               goto error;
+                                       }
+                                       /* Socket closed on remote end. */
+                                       ust_app_unregister(pollfd);
++
++                                      tracepoint(ust_unregister_stop);
+                                       break;
+                               }
+                       }
+               }
+       }
+ error:
+       DBG("Application communication apps dying");
+       close(apps_cmd_pipe[0]);
+       close(apps_cmd_pipe[1]);
+       lttng_poll_clean(&events);
+       rcu_thread_offline();
+       rcu_unregister_thread();
+       return NULL;
+ }
+ /*
+  * Dispatch request from the registration threads to the application
+  * communication thread.
+  */
+ static void *thread_dispatch_ust_registration(void *data)
+ {
+       int ret;
+       struct cds_wfq_node *node;
+       struct ust_command *ust_cmd = NULL;
++      tracepoint(sessiond_th_dispatch_start);
++
+       DBG("[thread] Dispatch UST command started");
+       while (!dispatch_thread_exit) {
+               /* Atomically prepare the queue futex */
+               futex_nto1_prepare(&ust_cmd_queue.futex);
+               do {
++                      tracepoint(sessiond_th_dispatch_block);
++
+                       /* Dequeue command for registration */
+                       node = cds_wfq_dequeue_blocking(&ust_cmd_queue.queue);
+                       if (node == NULL) {
+                               DBG("Woken up but nothing in the UST command queue");
+                               /* Continue thread execution */
+                               break;
+                       }
++                      tracepoint(ust_dispatch_register_start);
++
+                       ust_cmd = caa_container_of(node, struct ust_command, node);
+                       DBG("Dispatching UST registration pid:%d ppid:%d uid:%d"
+                                       " gid:%d sock:%d name:%s (version %d.%d)",
+                                       ust_cmd->reg_msg.pid, ust_cmd->reg_msg.ppid,
+                                       ust_cmd->reg_msg.uid, ust_cmd->reg_msg.gid,
+                                       ust_cmd->sock, ust_cmd->reg_msg.name,
+                                       ust_cmd->reg_msg.major, ust_cmd->reg_msg.minor);
+                       /*
+                        * Inform apps thread of the new application registration. This
+                        * call is blocking so we can be assured that the data will be read
+                        * at some point in time or wait to the end of the world :)
+                        */
+                       ret = write(apps_cmd_pipe[1], ust_cmd,
+                                       sizeof(struct ust_command));
+                       if (ret < 0) {
+                               perror("write apps cmd pipe");
+                               if (errno == EBADF) {
+                                       /*
+                                        * We can't inform the application thread to process
+                                        * registration. We will exit or else application
+                                        * registration will not occur and tracing will never
+                                        * start.
+                                        */
+                                       goto error;
+                               }
+                       }
+                       free(ust_cmd);
+               } while (node != NULL);
++              tracepoint(ust_dispatch_register_stop);
++
+               /* Futex wait on queue. Blocking call on futex() */
+               futex_nto1_wait(&ust_cmd_queue.futex);
+       }
+ error:
+       DBG("Dispatch thread dying");
+       return NULL;
+ }
+ /*
+  * This thread manage application registration.
+  */
+ static void *thread_registration_apps(void *data)
+ {
+       int sock = 0, i, ret, pollfd;
+       uint32_t revents, nb_fd;
+       struct lttng_poll_event events;
+       /*
+        * Get allocated in this thread, enqueued to a global queue, dequeued and
+        * freed in the manage apps thread.
+        */
+       struct ust_command *ust_cmd = NULL;
++      tracepoint(sessiond_th_reg_start);
++
+       DBG("[thread] Manage application registration started");
+       ret = lttcomm_listen_unix_sock(apps_sock);
+       if (ret < 0) {
+               goto error;
+       }
+       /*
+        * Pass 2 as size here for the thread quit pipe and apps socket. Nothing
+        * more will be added to this poll set.
+        */
+       ret = create_thread_poll_set(&events, 2);
+       if (ret < 0) {
+               goto error;
+       }
+       /* Add the application registration socket */
+       ret = lttng_poll_add(&events, apps_sock, LPOLLIN | LPOLLRDHUP);
+       if (ret < 0) {
+               goto error;
+       }
+       /* Notify all applications to register */
+       ret = notify_ust_apps(1);
+       if (ret < 0) {
+               ERR("Failed to notify applications or create the wait shared memory.\n"
+                       "Execution continues but there might be problem for already\n"
+                       "running applications that wishes to register.");
+       }
+       while (1) {
+               DBG("Accepting application registration");
++              tracepoint(sessiond_th_reg_poll);
++
+               nb_fd = LTTNG_POLL_GETNB(&events);
+               /* Inifinite blocking call, waiting for transmission */
+               ret = lttng_poll_wait(&events, -1);
+               if (ret < 0) {
+                       goto error;
+               }
+               for (i = 0; i < nb_fd; i++) {
+                       /* Fetch once the poll data */
+                       revents = LTTNG_POLL_GETEV(&events, i);
+                       pollfd = LTTNG_POLL_GETFD(&events, i);
+                       /* Thread quit pipe has been closed. Killing thread. */
+                       ret = check_thread_quit_pipe(pollfd, revents);
+                       if (ret) {
+                               goto error;
+                       }
+                       /* Event on the registration socket */
+                       if (pollfd == apps_sock) {
+                               if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                                       ERR("Register apps socket poll error");
+                                       goto error;
+                               } else if (revents & LPOLLIN) {
++                                      /* Registration starts here. Recording cycles */
++                                      tracepoint(ust_register_start);
++
+                                       sock = lttcomm_accept_unix_sock(apps_sock);
+                                       if (sock < 0) {
+                                               goto error;
+                                       }
+                                       /* Create UST registration command for enqueuing */
+                                       ust_cmd = zmalloc(sizeof(struct ust_command));
+                                       if (ust_cmd == NULL) {
+                                               perror("ust command zmalloc");
+                                               goto error;
+                                       }
+                                       /*
+                                        * Using message-based transmissions to ensure we don't
+                                        * have to deal with partially received messages.
+                                        */
+                                       ret = lttcomm_recv_unix_sock(sock, &ust_cmd->reg_msg,
+                                                       sizeof(struct ust_register_msg));
+                                       if (ret < 0 || ret < sizeof(struct ust_register_msg)) {
+                                               if (ret < 0) {
+                                                       perror("lttcomm_recv_unix_sock register apps");
+                                               } else {
+                                                       ERR("Wrong size received on apps register");
+                                               }
+                                               free(ust_cmd);
+                                               close(sock);
+                                               continue;
+                                       }
+                                       ust_cmd->sock = sock;
+                                       DBG("UST registration received with pid:%d ppid:%d uid:%d"
+                                                       " gid:%d sock:%d name:%s (version %d.%d)",
+                                                       ust_cmd->reg_msg.pid, ust_cmd->reg_msg.ppid,
+                                                       ust_cmd->reg_msg.uid, ust_cmd->reg_msg.gid,
+                                                       ust_cmd->sock, ust_cmd->reg_msg.name,
+                                                       ust_cmd->reg_msg.major, ust_cmd->reg_msg.minor);
+                                       /*
+                                        * Lock free enqueue the registration request. The red pill
+                                        * has been taken! This apps will be part of the *system*.
+                                        */
+                                       cds_wfq_enqueue(&ust_cmd_queue.queue, &ust_cmd->node);
+                                       /*
+                                        * Wake the registration queue futex. Implicit memory
+                                        * barrier with the exchange in cds_wfq_enqueue.
+                                        */
+                                       futex_nto1_wake(&ust_cmd_queue.futex);
++
++                                      tracepoint(ust_register_stop);
+                               }
+                       }
+               }
+       }
+ error:
+       DBG("UST Registration thread dying");
+       /* Notify that the registration thread is gone */
+       notify_ust_apps(0);
+       close(apps_sock);
+       close(sock);
+       unlink(apps_unix_sock_path);
+       lttng_poll_clean(&events);
+       return NULL;
+ }
+ /*
+  * Start the thread_manage_consumer. This must be done after a lttng-consumerd
+  * exec or it will fails.
+  */
+ static int spawn_consumer_thread(struct consumer_data *consumer_data)
+ {
+       int ret;
+       struct timespec timeout;
+       timeout.tv_sec = DEFAULT_SEM_WAIT_TIMEOUT;
+       timeout.tv_nsec = 0;
+       /* Setup semaphore */
+       ret = sem_init(&consumer_data->sem, 0, 0);
+       if (ret < 0) {
+               PERROR("sem_init consumer semaphore");
+               goto error;
+       }
+       ret = pthread_create(&consumer_data->thread, NULL,
+                       thread_manage_consumer, consumer_data);
+       if (ret != 0) {
+               PERROR("pthread_create consumer");
+               ret = -1;
+               goto error;
+       }
+       /* Get time for sem_timedwait absolute timeout */
+       ret = clock_gettime(CLOCK_REALTIME, &timeout);
+       if (ret < 0) {
+               PERROR("clock_gettime spawn consumer");
+               /* Infinite wait for the kconsumerd thread to be ready */
+               ret = sem_wait(&consumer_data->sem);
+       } else {
+               /* Normal timeout if the gettime was successful */
+               timeout.tv_sec += DEFAULT_SEM_WAIT_TIMEOUT;
+               ret = sem_timedwait(&consumer_data->sem, &timeout);
+       }
+       if (ret < 0) {
+               if (errno == ETIMEDOUT) {
+                       /*
+                        * Call has timed out so we kill the kconsumerd_thread and return
+                        * an error.
+                        */
+                       ERR("The consumer thread was never ready. Killing it");
+                       ret = pthread_cancel(consumer_data->thread);
+                       if (ret < 0) {
+                               PERROR("pthread_cancel consumer thread");
+                       }
+               } else {
+                       PERROR("semaphore wait failed consumer thread");
+               }
+               goto error;
+       }
+       pthread_mutex_lock(&consumer_data->pid_mutex);
+       if (consumer_data->pid == 0) {
+               ERR("Kconsumerd did not start");
+               pthread_mutex_unlock(&consumer_data->pid_mutex);
+               goto error;
+       }
+       pthread_mutex_unlock(&consumer_data->pid_mutex);
+       return 0;
+ error:
+       return ret;
+ }
+ /*
+  * Join consumer thread
+  */
+ static int join_consumer_thread(struct consumer_data *consumer_data)
+ {
+       void *status;
+       int ret;
+       if (consumer_data->pid != 0) {
+               ret = kill(consumer_data->pid, SIGTERM);
+               if (ret) {
+                       ERR("Error killing consumer daemon");
+                       return ret;
+               }
+               return pthread_join(consumer_data->thread, &status);
+       } else {
+               return 0;
+       }
+ }
+ /*
+  * Fork and exec a consumer daemon (consumerd).
+  *
+  * Return pid if successful else -1.
+  */
+ static pid_t spawn_consumerd(struct consumer_data *consumer_data)
+ {
+       int ret;
+       pid_t pid;
+       const char *consumer_to_use;
+       const char *verbosity;
+       struct stat st;
+       DBG("Spawning consumerd");
+       pid = fork();
+       if (pid == 0) {
+               /*
+                * Exec consumerd.
+                */
+               if (opt_verbose_consumer) {
+                       verbosity = "--verbose";
+               } else {
+                       verbosity = "--quiet";
+               }
+               switch (consumer_data->type) {
+               case LTTNG_CONSUMER_KERNEL:
+                       /*
+                        * Find out which consumerd to execute. We will first try the
+                        * 64-bit path, then the sessiond's installation directory, and
+                        * fallback on the 32-bit one,
+                        */
+                       DBG3("Looking for a kernel consumer at these locations:");
+                       DBG3("  1) %s", consumerd64_bin);
+                       DBG3("  2) %s/%s", INSTALL_BIN_PATH, CONSUMERD_FILE);
+                       DBG3("  3) %s", consumerd32_bin);
+                       if (stat(consumerd64_bin, &st) == 0) {
+                               DBG3("Found location #1");
+                               consumer_to_use = consumerd64_bin;
+                       } else if (stat(INSTALL_BIN_PATH "/" CONSUMERD_FILE, &st) == 0) {
+                               DBG3("Found location #2");
+                               consumer_to_use = INSTALL_BIN_PATH "/" CONSUMERD_FILE;
+                       } else if (stat(consumerd32_bin, &st) == 0) {
+                               DBG3("Found location #3");
+                               consumer_to_use = consumerd32_bin;
+                       } else {
+                               DBG("Could not find any valid consumerd executable");
+                               break;
+                       }
+                       DBG("Using kernel consumer at: %s",  consumer_to_use);
+                       execl(consumer_to_use,
+                               "lttng-consumerd", verbosity, "-k",
+                               "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path,
+                               "--consumerd-err-sock", consumer_data->err_unix_sock_path,
+                               NULL);
+                       break;
+               case LTTNG_CONSUMER64_UST:
+               {
+                       char *tmpnew = NULL;
+                       if (consumerd64_libdir[0] != '\0') {
+                               char *tmp;
+                               size_t tmplen;
+                               tmp = getenv("LD_LIBRARY_PATH");
+                               if (!tmp) {
+                                       tmp = "";
+                               }
+                               tmplen = strlen("LD_LIBRARY_PATH=")
+                                       + strlen(consumerd64_libdir) + 1 /* : */ + strlen(tmp);
+                               tmpnew = zmalloc(tmplen + 1 /* \0 */);
+                               if (!tmpnew) {
+                                       ret = -ENOMEM;
+                                       goto error;
+                               }
+                               strcpy(tmpnew, "LD_LIBRARY_PATH=");
+                               strcat(tmpnew, consumerd64_libdir);
+                               if (tmp[0] != '\0') {
+                                       strcat(tmpnew, ":");
+                                       strcat(tmpnew, tmp);
+                               }
+                               ret = putenv(tmpnew);
+                               if (ret) {
+                                       ret = -errno;
+                                       goto error;
+                               }
+                       }
+                       DBG("Using 64-bit UST consumer at: %s",  consumerd64_bin);
+                       ret = execl(consumerd64_bin, "lttng-consumerd", verbosity, "-u",
+                                       "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path,
+                                       "--consumerd-err-sock", consumer_data->err_unix_sock_path,
+                                       NULL);
+                       if (consumerd64_libdir[0] != '\0') {
+                               free(tmpnew);
+                       }
+                       if (ret) {
+                               goto error;
+                       }
+                       break;
+               }
+               case LTTNG_CONSUMER32_UST:
+               {
+                       char *tmpnew = NULL;
+                       if (consumerd32_libdir[0] != '\0') {
+                               char *tmp;
+                               size_t tmplen;
+                               tmp = getenv("LD_LIBRARY_PATH");
+                               if (!tmp) {
+                                       tmp = "";
+                               }
+                               tmplen = strlen("LD_LIBRARY_PATH=")
+                                       + strlen(consumerd32_libdir) + 1 /* : */ + strlen(tmp);
+                               tmpnew = zmalloc(tmplen + 1 /* \0 */);
+                               if (!tmpnew) {
+                                       ret = -ENOMEM;
+                                       goto error;
+                               }
+                               strcpy(tmpnew, "LD_LIBRARY_PATH=");
+                               strcat(tmpnew, consumerd32_libdir);
+                               if (tmp[0] != '\0') {
+                                       strcat(tmpnew, ":");
+                                       strcat(tmpnew, tmp);
+                               }
+                               ret = putenv(tmpnew);
+                               if (ret) {
+                                       ret = -errno;
+                                       goto error;
+                               }
+                       }
+                       DBG("Using 32-bit UST consumer at: %s",  consumerd32_bin);
+                       ret = execl(consumerd32_bin, "lttng-consumerd", verbosity, "-u",
+                                       "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path,
+                                       "--consumerd-err-sock", consumer_data->err_unix_sock_path,
+                                       NULL);
+                       if (consumerd32_libdir[0] != '\0') {
+                               free(tmpnew);
+                       }
+                       if (ret) {
+                               goto error;
+                       }
+                       break;
+               }
+               default:
+                       perror("unknown consumer type");
+                       exit(EXIT_FAILURE);
+               }
+               if (errno != 0) {
+                       perror("kernel start consumer exec");
+               }
+               exit(EXIT_FAILURE);
+       } else if (pid > 0) {
+               ret = pid;
+       } else {
+               perror("start consumer fork");
+               ret = -errno;
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Spawn the consumerd daemon and session daemon thread.
+  */
+ static int start_consumerd(struct consumer_data *consumer_data)
+ {
+       int ret;
+       pthread_mutex_lock(&consumer_data->pid_mutex);
+       if (consumer_data->pid != 0) {
+               pthread_mutex_unlock(&consumer_data->pid_mutex);
+               goto end;
+       }
+       ret = spawn_consumerd(consumer_data);
+       if (ret < 0) {
+               ERR("Spawning consumerd failed");
+               pthread_mutex_unlock(&consumer_data->pid_mutex);
+               goto error;
+       }
+       /* Setting up the consumer_data pid */
+       consumer_data->pid = ret;
+       DBG2("Consumer pid %d", consumer_data->pid);
+       pthread_mutex_unlock(&consumer_data->pid_mutex);
+       DBG2("Spawning consumer control thread");
+       ret = spawn_consumer_thread(consumer_data);
+       if (ret < 0) {
+               ERR("Fatal error spawning consumer control thread");
+               goto error;
+       }
+ end:
+       return 0;
+ error:
+       return ret;
+ }
+ /*
+  * Check version of the lttng-modules.
+  */
+ static int validate_lttng_modules_version(void)
+ {
+       return kernel_validate_version(kernel_tracer_fd);
+ }
+ /*
+  * Setup necessary data for kernel tracer action.
+  */
+ static int init_kernel_tracer(void)
+ {
+       int ret;
+       /* Modprobe lttng kernel modules */
+       ret = modprobe_lttng_control();
+       if (ret < 0) {
+               goto error;
+       }
+       /* Open debugfs lttng */
+       kernel_tracer_fd = open(module_proc_lttng, O_RDWR);
+       if (kernel_tracer_fd < 0) {
+               DBG("Failed to open %s", module_proc_lttng);
+               ret = -1;
+               goto error_open;
+       }
+       /* Validate kernel version */
+       ret = validate_lttng_modules_version();
+       if (ret < 0) {
+               goto error_version;
+       }
+       ret = modprobe_lttng_data();
+       if (ret < 0) {
+               goto error_modules;
+       }
+       DBG("Kernel tracer fd %d", kernel_tracer_fd);
+       return 0;
+ error_version:
+       modprobe_remove_lttng_control();
+       close(kernel_tracer_fd);
+       kernel_tracer_fd = 0;
+       return LTTCOMM_KERN_VERSION;
+ error_modules:
+       close(kernel_tracer_fd);
+ error_open:
+       modprobe_remove_lttng_control();
+ error:
+       WARN("No kernel tracer available");
+       kernel_tracer_fd = 0;
+       return LTTCOMM_KERN_NA;
+ }
+ /*
+  * Init tracing by creating trace directory and sending fds kernel consumer.
+  */
+ static int init_kernel_tracing(struct ltt_kernel_session *session)
+ {
+       int ret = 0;
+       if (session->consumer_fds_sent == 0) {
+               /*
+                * Assign default kernel consumer socket if no consumer assigned to the
+                * kernel session. At this point, it's NOT suppose to be 0 but this is
+                * an extra security check.
+                */
+               if (session->consumer_fd == 0) {
+                       session->consumer_fd = kconsumer_data.cmd_sock;
+               }
+               ret = send_kconsumer_session_streams(&kconsumer_data, session);
+               if (ret < 0) {
+                       ret = LTTCOMM_KERN_CONSUMER_FAIL;
+                       goto error;
+               }
+               session->consumer_fds_sent = 1;
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Create an UST session and add it to the session ust list.
+  */
+ static int create_ust_session(struct ltt_session *session,
+               struct lttng_domain *domain)
+ {
+       struct ltt_ust_session *lus = NULL;
+       int ret;
+       switch (domain->type) {
+       case LTTNG_DOMAIN_UST:
+               break;
+       default:
+               ret = LTTCOMM_UNKNOWN_DOMAIN;
+               goto error;
+       }
+       DBG("Creating UST session");
+       lus = trace_ust_create_session(session->path, session->id, domain);
+       if (lus == NULL) {
+               ret = LTTCOMM_UST_SESS_FAIL;
+               goto error;
+       }
+       ret = run_as_mkdir_recursive(lus->pathname, S_IRWXU | S_IRWXG,
+                       session->uid, session->gid);
+       if (ret < 0) {
+               if (ret != -EEXIST) {
+                       ERR("Trace directory creation error");
+                       ret = LTTCOMM_UST_SESS_FAIL;
+                       goto error;
+               }
+       }
+       /* The domain type dictate different actions on session creation */
+       switch (domain->type) {
+       case LTTNG_DOMAIN_UST:
+               /* No ustctl for the global UST domain */
+               break;
+       default:
+               ERR("Unknown UST domain on create session %d", domain->type);
+               goto error;
+       }
+       lus->uid = session->uid;
+       lus->gid = session->gid;
+       session->ust_session = lus;
+       return LTTCOMM_OK;
+ error:
+       free(lus);
+       return ret;
+ }
+ /*
+  * Create a kernel tracer session then create the default channel.
+  */
+ static int create_kernel_session(struct ltt_session *session)
+ {
+       int ret;
+       DBG("Creating kernel session");
+       ret = kernel_create_session(session, kernel_tracer_fd);
+       if (ret < 0) {
+               ret = LTTCOMM_KERN_SESS_FAIL;
+               goto error;
+       }
+       /* Set kernel consumer socket fd */
+       if (kconsumer_data.cmd_sock) {
+               session->kernel_session->consumer_fd = kconsumer_data.cmd_sock;
+       }
+       ret = run_as_mkdir_recursive(session->kernel_session->trace_path,
+                       S_IRWXU | S_IRWXG, session->uid, session->gid);
+       if (ret < 0) {
+               if (ret != -EEXIST) {
+                       ERR("Trace directory creation error");
+                       goto error;
+               }
+       }
+       session->kernel_session->uid = session->uid;
+       session->kernel_session->gid = session->gid;
+ error:
+       return ret;
+ }
+ /*
+  * Check if the UID or GID match the session. Root user has access to all
+  * sessions.
+  */
+ static int session_access_ok(struct ltt_session *session, uid_t uid, gid_t gid)
+ {
+       if (uid != session->uid && gid != session->gid && uid != 0) {
+               return 0;
+       } else {
+               return 1;
+       }
+ }
+ static unsigned int lttng_sessions_count(uid_t uid, gid_t gid)
+ {
+       unsigned int i = 0;
+       struct ltt_session *session;
+       DBG("Counting number of available session for UID %d GID %d",
+               uid, gid);
+       cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+               /*
+                * Only list the sessions the user can control.
+                */
+               if (!session_access_ok(session, uid, gid)) {
+                       continue;
+               }
+               i++;
+       }
+       return i;
+ }
+ /*
+  * Using the session list, filled a lttng_session array to send back to the
+  * client for session listing.
+  *
+  * The session list lock MUST be acquired before calling this function. Use
+  * session_lock_list() and session_unlock_list().
+  */
+ static void list_lttng_sessions(struct lttng_session *sessions, uid_t uid,
+               gid_t gid)
+ {
+       unsigned int i = 0;
+       struct ltt_session *session;
+       DBG("Getting all available session for UID %d GID %d",
+               uid, gid);
+       /*
+        * Iterate over session list and append data after the control struct in
+        * the buffer.
+        */
+       cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+               /*
+                * Only list the sessions the user can control.
+                */
+               if (!session_access_ok(session, uid, gid)) {
+                       continue;
+               }
+               strncpy(sessions[i].path, session->path, PATH_MAX);
+               sessions[i].path[PATH_MAX - 1] = '\0';
+               strncpy(sessions[i].name, session->name, NAME_MAX);
+               sessions[i].name[NAME_MAX - 1] = '\0';
+               sessions[i].enabled = session->enabled;
+               i++;
+       }
+ }
+ /*
+  * Fill lttng_channel array of all channels.
+  */
+ static void list_lttng_channels(int domain, struct ltt_session *session,
+               struct lttng_channel *channels)
+ {
+       int i = 0;
+       struct ltt_kernel_channel *kchan;
+       DBG("Listing channels for session %s", session->name);
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               /* Kernel channels */
+               if (session->kernel_session != NULL) {
+                       cds_list_for_each_entry(kchan,
+                                       &session->kernel_session->channel_list.head, list) {
+                               /* Copy lttng_channel struct to array */
+                               memcpy(&channels[i], kchan->channel, sizeof(struct lttng_channel));
+                               channels[i].enabled = kchan->enabled;
+                               i++;
+                       }
+               }
+               break;
+       case LTTNG_DOMAIN_UST:
+       {
+               struct lttng_ht_iter iter;
+               struct ltt_ust_channel *uchan;
+               cds_lfht_for_each_entry(session->ust_session->domain_global.channels->ht,
+                               &iter.iter, uchan, node.node) {
+                       strncpy(channels[i].name, uchan->name, LTTNG_SYMBOL_NAME_LEN);
+                       channels[i].attr.overwrite = uchan->attr.overwrite;
+                       channels[i].attr.subbuf_size = uchan->attr.subbuf_size;
+                       channels[i].attr.num_subbuf = uchan->attr.num_subbuf;
+                       channels[i].attr.switch_timer_interval =
+                               uchan->attr.switch_timer_interval;
+                       channels[i].attr.read_timer_interval =
+                               uchan->attr.read_timer_interval;
+                       channels[i].enabled = uchan->enabled;
+                       switch (uchan->attr.output) {
+                       case LTTNG_UST_MMAP:
+                       default:
+                               channels[i].attr.output = LTTNG_EVENT_MMAP;
+                               break;
+                       }
+                       i++;
+               }
+               break;
+       }
+       default:
+               break;
+       }
+ }
+ /*
+  * Create a list of ust global domain events.
+  */
+ static int list_lttng_ust_global_events(char *channel_name,
+               struct ltt_ust_domain_global *ust_global, struct lttng_event **events)
+ {
+       int i = 0, ret = 0;
+       unsigned int nb_event = 0;
+       struct lttng_ht_iter iter;
+       struct lttng_ht_node_str *node;
+       struct ltt_ust_channel *uchan;
+       struct ltt_ust_event *uevent;
+       struct lttng_event *tmp;
+       DBG("Listing UST global events for channel %s", channel_name);
+       rcu_read_lock();
+       lttng_ht_lookup(ust_global->channels, (void *)channel_name, &iter);
+       node = lttng_ht_iter_get_node_str(&iter);
+       if (node == NULL) {
+               ret = -LTTCOMM_UST_CHAN_NOT_FOUND;
+               goto error;
+       }
+       uchan = caa_container_of(&node->node, struct ltt_ust_channel, node.node);
+       nb_event += lttng_ht_get_count(uchan->events);
+       if (nb_event == 0) {
+               ret = nb_event;
+               goto error;
+       }
+       DBG3("Listing UST global %d events", nb_event);
+       tmp = zmalloc(nb_event * sizeof(struct lttng_event));
+       if (tmp == NULL) {
+               ret = -LTTCOMM_FATAL;
+               goto error;
+       }
+       cds_lfht_for_each_entry(uchan->events->ht, &iter.iter, uevent, node.node) {
+               strncpy(tmp[i].name, uevent->attr.name, LTTNG_SYMBOL_NAME_LEN);
+               tmp[i].name[LTTNG_SYMBOL_NAME_LEN - 1] = '\0';
+               tmp[i].enabled = uevent->enabled;
+               switch (uevent->attr.instrumentation) {
+               case LTTNG_UST_TRACEPOINT:
+                       tmp[i].type = LTTNG_EVENT_TRACEPOINT;
+                       break;
+               case LTTNG_UST_PROBE:
+                       tmp[i].type = LTTNG_EVENT_PROBE;
+                       break;
+               case LTTNG_UST_FUNCTION:
+                       tmp[i].type = LTTNG_EVENT_FUNCTION;
+                       break;
+               case LTTNG_UST_TRACEPOINT_LOGLEVEL:
+                       tmp[i].type = LTTNG_EVENT_TRACEPOINT_LOGLEVEL;
+                       break;
+               }
+               i++;
+       }
+       ret = nb_event;
+       *events = tmp;
+ error:
+       rcu_read_unlock();
+       return ret;
+ }
+ /*
+  * Fill lttng_event array of all kernel events in the channel.
+  */
+ static int list_lttng_kernel_events(char *channel_name,
+               struct ltt_kernel_session *kernel_session, struct lttng_event **events)
+ {
+       int i = 0, ret;
+       unsigned int nb_event;
+       struct ltt_kernel_event *event;
+       struct ltt_kernel_channel *kchan;
+       kchan = trace_kernel_get_channel_by_name(channel_name, kernel_session);
+       if (kchan == NULL) {
+               ret = LTTCOMM_KERN_CHAN_NOT_FOUND;
+               goto error;
+       }
+       nb_event = kchan->event_count;
+       DBG("Listing events for channel %s", kchan->channel->name);
+       if (nb_event == 0) {
+               ret = nb_event;
+               goto error;
+       }
+       *events = zmalloc(nb_event * sizeof(struct lttng_event));
+       if (*events == NULL) {
+               ret = LTTCOMM_FATAL;
+               goto error;
+       }
+       /* Kernel channels */
+       cds_list_for_each_entry(event, &kchan->events_list.head , list) {
+               strncpy((*events)[i].name, event->event->name, LTTNG_SYMBOL_NAME_LEN);
+               (*events)[i].name[LTTNG_SYMBOL_NAME_LEN - 1] = '\0';
+               (*events)[i].enabled = event->enabled;
+               switch (event->event->instrumentation) {
+                       case LTTNG_KERNEL_TRACEPOINT:
+                               (*events)[i].type = LTTNG_EVENT_TRACEPOINT;
+                               break;
+                       case LTTNG_KERNEL_KPROBE:
+                       case LTTNG_KERNEL_KRETPROBE:
+                               (*events)[i].type = LTTNG_EVENT_PROBE;
+                               memcpy(&(*events)[i].attr.probe, &event->event->u.kprobe,
+                                               sizeof(struct lttng_kernel_kprobe));
+                               break;
+                       case LTTNG_KERNEL_FUNCTION:
+                               (*events)[i].type = LTTNG_EVENT_FUNCTION;
+                               memcpy(&((*events)[i].attr.ftrace), &event->event->u.ftrace,
+                                               sizeof(struct lttng_kernel_function));
+                               break;
+                       case LTTNG_KERNEL_NOOP:
+                               (*events)[i].type = LTTNG_EVENT_NOOP;
+                               break;
+                       case LTTNG_KERNEL_SYSCALL:
+                               (*events)[i].type = LTTNG_EVENT_SYSCALL;
+                               break;
+                       case LTTNG_KERNEL_ALL:
+                               assert(0);
+                               break;
+               }
+               i++;
+       }
+       return nb_event;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_DISABLE_CHANNEL processed by the client thread.
+  */
+ static int cmd_disable_channel(struct ltt_session *session,
+               int domain, char *channel_name)
+ {
+       int ret;
+       struct ltt_ust_session *usess;
+       usess = session->ust_session;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               ret = channel_kernel_disable(session->kernel_session,
+                               channel_name);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct ltt_ust_channel *uchan;
+               struct lttng_ht *chan_ht;
+               chan_ht = usess->domain_global.channels;
+               uchan = trace_ust_find_channel_by_name(chan_ht, channel_name);
+               if (uchan == NULL) {
+                       ret = LTTCOMM_UST_CHAN_NOT_FOUND;
+                       goto error;
+               }
+               ret = channel_ust_disable(usess, domain, uchan);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+ #endif
+       default:
+               ret = LTTCOMM_UNKNOWN_DOMAIN;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_ENABLE_CHANNEL processed by the client thread.
+  */
+ static int cmd_enable_channel(struct ltt_session *session,
+               int domain, struct lttng_channel *attr)
+ {
+       int ret;
+       struct ltt_ust_session *usess = session->ust_session;
+       struct lttng_ht *chan_ht;
+       DBG("Enabling channel %s for session %s", attr->name, session->name);
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               struct ltt_kernel_channel *kchan;
+               kchan = trace_kernel_get_channel_by_name(attr->name,
+                               session->kernel_session);
+               if (kchan == NULL) {
+                       ret = channel_kernel_create(session->kernel_session,
+                                       attr, kernel_poll_pipe[1]);
+               } else {
+                       ret = channel_kernel_enable(session->kernel_session, kchan);
+               }
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct ltt_ust_channel *uchan;
+               chan_ht = usess->domain_global.channels;
+               uchan = trace_ust_find_channel_by_name(chan_ht, attr->name);
+               if (uchan == NULL) {
+                       ret = channel_ust_create(usess, domain, attr);
+               } else {
+                       ret = channel_ust_enable(usess, domain, uchan);
+               }
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+ #endif
+       default:
+               ret = LTTCOMM_UNKNOWN_DOMAIN;
+               goto error;
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_DISABLE_EVENT processed by the client thread.
+  */
+ static int cmd_disable_event(struct ltt_session *session, int domain,
+               char *channel_name, char *event_name)
+ {
+       int ret;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               struct ltt_kernel_channel *kchan;
+               struct ltt_kernel_session *ksess;
+               ksess = session->kernel_session;
+               kchan = trace_kernel_get_channel_by_name(channel_name, ksess);
+               if (kchan == NULL) {
+                       ret = LTTCOMM_KERN_CHAN_NOT_FOUND;
+                       goto error;
+               }
+               ret = event_kernel_disable_tracepoint(ksess, kchan, event_name);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct ltt_ust_channel *uchan;
+               struct ltt_ust_session *usess;
+               usess = session->ust_session;
+               uchan = trace_ust_find_channel_by_name(usess->domain_global.channels,
+                               channel_name);
+               if (uchan == NULL) {
+                       ret = LTTCOMM_UST_CHAN_NOT_FOUND;
+                       goto error;
+               }
+               ret = event_ust_disable_tracepoint(usess, domain, uchan, event_name);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               DBG3("Disable UST event %s in channel %s completed", event_name,
+                               channel_name);
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+ #endif
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_DISABLE_ALL_EVENT processed by the client thread.
+  */
+ static int cmd_disable_event_all(struct ltt_session *session, int domain,
+               char *channel_name)
+ {
+       int ret;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               struct ltt_kernel_session *ksess;
+               struct ltt_kernel_channel *kchan;
+               ksess = session->kernel_session;
+               kchan = trace_kernel_get_channel_by_name(channel_name, ksess);
+               if (kchan == NULL) {
+                       ret = LTTCOMM_KERN_CHAN_NOT_FOUND;
+                       goto error;
+               }
+               ret = event_kernel_disable_all(ksess, kchan);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct ltt_ust_session *usess;
+               struct ltt_ust_channel *uchan;
+               usess = session->ust_session;
+               uchan = trace_ust_find_channel_by_name(usess->domain_global.channels,
+                               channel_name);
+               if (uchan == NULL) {
+                       ret = LTTCOMM_UST_CHAN_NOT_FOUND;
+                       goto error;
+               }
+               ret = event_ust_disable_all_tracepoints(usess, domain, uchan);
+               if (ret != 0) {
+                       goto error;
+               }
+               DBG3("Disable all UST events in channel %s completed", channel_name);
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+ #endif
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_ADD_CONTEXT processed by the client thread.
+  */
+ static int cmd_add_context(struct ltt_session *session, int domain,
+               char *channel_name, char *event_name, struct lttng_event_context *ctx)
+ {
+       int ret;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               /* Add kernel context to kernel tracer */
+               ret = context_kernel_add(session->kernel_session, ctx,
+                               event_name, channel_name);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               break;
+       case LTTNG_DOMAIN_UST:
+       {
+               struct ltt_ust_session *usess = session->ust_session;
+               ret = context_ust_add(usess, domain, ctx, event_name, channel_name);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+ #endif
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_ENABLE_EVENT processed by the client thread.
+  *
+  * TODO: currently, both events and loglevels are kept within the same
+  * namespace for UST global registry/app registery, so if an event
+  * happen to have the same name as the loglevel (very unlikely though),
+  * and an attempt is made to enable/disable both in the same session,
+  * the first to be created will be the only one allowed to exist.
+  */
+ static int cmd_enable_event(struct ltt_session *session, int domain,
+               char *channel_name, struct lttng_event *event)
+ {
+       int ret;
+       struct lttng_channel *attr;
+       struct ltt_ust_session *usess = session->ust_session;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               struct ltt_kernel_channel *kchan;
+               kchan = trace_kernel_get_channel_by_name(channel_name,
+                               session->kernel_session);
+               if (kchan == NULL) {
+                       attr = channel_new_default_attr(domain);
+                       if (attr == NULL) {
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+                       snprintf(attr->name, NAME_MAX, "%s", channel_name);
+                       /* This call will notify the kernel thread */
+                       ret = channel_kernel_create(session->kernel_session,
+                                       attr, kernel_poll_pipe[1]);
+                       if (ret != LTTCOMM_OK) {
+                               free(attr);
+                               goto error;
+                       }
+                       free(attr);
+               }
+               /* Get the newly created kernel channel pointer */
+               kchan = trace_kernel_get_channel_by_name(channel_name,
+                               session->kernel_session);
+               if (kchan == NULL) {
+                       /* This sould not happen... */
+                       ret = LTTCOMM_FATAL;
+                       goto error;
+               }
+               ret = event_kernel_enable_tracepoint(session->kernel_session, kchan,
+                               event);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct lttng_channel *attr;
+               struct ltt_ust_channel *uchan;
+               /* Get channel from global UST domain */
+               uchan = trace_ust_find_channel_by_name(usess->domain_global.channels,
+                               channel_name);
+               if (uchan == NULL) {
+                       /* Create default channel */
+                       attr = channel_new_default_attr(domain);
+                       if (attr == NULL) {
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+                       snprintf(attr->name, NAME_MAX, "%s", channel_name);
+                       attr->name[NAME_MAX - 1] = '\0';
+                       ret = channel_ust_create(usess, domain, attr);
+                       if (ret != LTTCOMM_OK) {
+                               free(attr);
+                               goto error;
+                       }
+                       free(attr);
+                       /* Get the newly created channel reference back */
+                       uchan = trace_ust_find_channel_by_name(
+                                       usess->domain_global.channels, channel_name);
+                       if (uchan == NULL) {
+                               /* Something is really wrong */
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+               }
+               /* At this point, the session and channel exist on the tracer */
+               ret = event_ust_enable_tracepoint(usess, domain, uchan, event);
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+ #endif
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_ENABLE_ALL_EVENT processed by the client thread.
+  */
+ static int cmd_enable_event_all(struct ltt_session *session, int domain,
+               char *channel_name, int event_type)
+ {
+       int ret;
+       struct ltt_kernel_channel *kchan;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               kchan = trace_kernel_get_channel_by_name(channel_name,
+                               session->kernel_session);
+               if (kchan == NULL) {
+                       /* This call will notify the kernel thread */
+                       ret = channel_kernel_create(session->kernel_session, NULL,
+                                       kernel_poll_pipe[1]);
+                       if (ret != LTTCOMM_OK) {
+                               goto error;
+                       }
+                       /* Get the newly created kernel channel pointer */
+                       kchan = trace_kernel_get_channel_by_name(channel_name,
+                                       session->kernel_session);
+                       if (kchan == NULL) {
+                               /* This sould not happen... */
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+               }
+               switch (event_type) {
+               case LTTNG_EVENT_SYSCALL:
+                       ret = event_kernel_enable_all_syscalls(session->kernel_session,
+                                       kchan, kernel_tracer_fd);
+                       break;
+               case LTTNG_EVENT_TRACEPOINT:
+                       /*
+                        * This call enables all LTTNG_KERNEL_TRACEPOINTS and
+                        * events already registered to the channel.
+                        */
+                       ret = event_kernel_enable_all_tracepoints(session->kernel_session,
+                                       kchan, kernel_tracer_fd);
+                       break;
+               case LTTNG_EVENT_ALL:
+                       /* Enable syscalls and tracepoints */
+                       ret = event_kernel_enable_all(session->kernel_session,
+                                       kchan, kernel_tracer_fd);
+                       break;
+               default:
+                       ret = LTTCOMM_KERN_ENABLE_FAIL;
+                       goto error;
+               }
+               /* Manage return value */
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+               break;
+       case LTTNG_DOMAIN_UST:
+       {
+               struct lttng_channel *attr;
+               struct ltt_ust_channel *uchan;
+               struct ltt_ust_session *usess = session->ust_session;
+               /* Get channel from global UST domain */
+               uchan = trace_ust_find_channel_by_name(usess->domain_global.channels,
+                               channel_name);
+               if (uchan == NULL) {
+                       /* Create default channel */
+                       attr = channel_new_default_attr(domain);
+                       if (attr == NULL) {
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+                       snprintf(attr->name, NAME_MAX, "%s", channel_name);
+                       attr->name[NAME_MAX - 1] = '\0';
+                       /* Use the internal command enable channel */
+                       ret = channel_ust_create(usess, domain, attr);
+                       if (ret != LTTCOMM_OK) {
+                               free(attr);
+                               goto error;
+                       }
+                       free(attr);
+                       /* Get the newly created channel reference back */
+                       uchan = trace_ust_find_channel_by_name(
+                                       usess->domain_global.channels, channel_name);
+                       if (uchan == NULL) {
+                               /* Something is really wrong */
+                               ret = LTTCOMM_FATAL;
+                               goto error;
+                       }
+               }
+               /* At this point, the session and channel exist on the tracer */
+               switch (event_type) {
+               case LTTNG_EVENT_ALL:
+               case LTTNG_EVENT_TRACEPOINT:
+                       ret = event_ust_enable_all_tracepoints(usess, domain, uchan);
+                       if (ret != LTTCOMM_OK) {
+                               goto error;
+                       }
+                       break;
+               default:
+                       ret = LTTCOMM_UST_ENABLE_FAIL;
+                       goto error;
+               }
+               /* Manage return value */
+               if (ret != LTTCOMM_OK) {
+                       goto error;
+               }
+               break;
+       }
+ #if 0
+       case LTTNG_DOMAIN_UST_EXEC_NAME:
+       case LTTNG_DOMAIN_UST_PID:
+       case LTTNG_DOMAIN_UST_PID_FOLLOW_CHILDREN:
+ #endif
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_LIST_TRACEPOINTS processed by the client thread.
+  */
+ static ssize_t cmd_list_tracepoints(int domain, struct lttng_event **events)
+ {
+       int ret;
+       ssize_t nb_events = 0;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               nb_events = kernel_list_events(kernel_tracer_fd, events);
+               if (nb_events < 0) {
+                       ret = LTTCOMM_KERN_LIST_FAIL;
+                       goto error;
+               }
+               break;
+       case LTTNG_DOMAIN_UST:
+               nb_events = ust_app_list_events(events);
+               if (nb_events < 0) {
+                       ret = LTTCOMM_UST_LIST_FAIL;
+                       goto error;
+               }
+               break;
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       return nb_events;
+ error:
+       /* Return negative value to differentiate return code */
+       return -ret;
+ }
+ /*
+  * Command LTTNG_START_TRACE processed by the client thread.
+  */
+ static int cmd_start_trace(struct ltt_session *session)
+ {
+       int ret;
+       struct ltt_kernel_session *ksession;
+       struct ltt_ust_session *usess;
+       /* Short cut */
+       ksession = session->kernel_session;
+       usess = session->ust_session;
+       if (session->enabled) {
+               ret = LTTCOMM_UST_START_FAIL;
+               goto error;
+       }
+       session->enabled = 1;
+       /* Kernel tracing */
+       if (ksession != NULL) {
+               struct ltt_kernel_channel *kchan;
+               /* Open kernel metadata */
+               if (ksession->metadata == NULL) {
+                       ret = kernel_open_metadata(ksession, ksession->trace_path);
+                       if (ret < 0) {
+                               ret = LTTCOMM_KERN_META_FAIL;
+                               goto error;
+                       }
+               }
+               /* Open kernel metadata stream */
+               if (ksession->metadata_stream_fd == 0) {
+                       ret = kernel_open_metadata_stream(ksession);
+                       if (ret < 0) {
+                               ERR("Kernel create metadata stream failed");
+                               ret = LTTCOMM_KERN_STREAM_FAIL;
+                               goto error;
+                       }
+               }
+               /* For each channel */
+               cds_list_for_each_entry(kchan, &ksession->channel_list.head, list) {
+                       if (kchan->stream_count == 0) {
+                               ret = kernel_open_channel_stream(kchan);
+                               if (ret < 0) {
+                                       ret = LTTCOMM_KERN_STREAM_FAIL;
+                                       goto error;
+                               }
+                               /* Update the stream global counter */
+                               ksession->stream_count_global += ret;
+                       }
+               }
+               /* Setup kernel consumer socket and send fds to it */
+               ret = init_kernel_tracing(ksession);
+               if (ret < 0) {
+                       ret = LTTCOMM_KERN_START_FAIL;
+                       goto error;
+               }
+               /* This start the kernel tracing */
+               ret = kernel_start_session(ksession);
+               if (ret < 0) {
+                       ret = LTTCOMM_KERN_START_FAIL;
+                       goto error;
+               }
+               /* Quiescent wait after starting trace */
+               kernel_wait_quiescent(kernel_tracer_fd);
+       }
+       /* Flag session that trace should start automatically */
+       if (usess) {
+               usess->start_trace = 1;
+               ret = ust_app_start_trace_all(usess);
+               if (ret < 0) {
+                       ret = LTTCOMM_UST_START_FAIL;
+                       goto error;
+               }
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_STOP_TRACE processed by the client thread.
+  */
+ static int cmd_stop_trace(struct ltt_session *session)
+ {
+       int ret;
+       struct ltt_kernel_channel *kchan;
+       struct ltt_kernel_session *ksession;
+       struct ltt_ust_session *usess;
+       /* Short cut */
+       ksession = session->kernel_session;
+       usess = session->ust_session;
+       if (!session->enabled) {
+               ret = LTTCOMM_UST_STOP_FAIL;
+               goto error;
+       }
+       session->enabled = 0;
+       /* Kernel tracer */
+       if (ksession != NULL) {
+               DBG("Stop kernel tracing");
+               /* Flush all buffers before stopping */
+               ret = kernel_metadata_flush_buffer(ksession->metadata_stream_fd);
+               if (ret < 0) {
+                       ERR("Kernel metadata flush failed");
+               }
+               cds_list_for_each_entry(kchan, &ksession->channel_list.head, list) {
+                       ret = kernel_flush_buffer(kchan);
+                       if (ret < 0) {
+                               ERR("Kernel flush buffer error");
+                       }
+               }
+               ret = kernel_stop_session(ksession);
+               if (ret < 0) {
+                       ret = LTTCOMM_KERN_STOP_FAIL;
+                       goto error;
+               }
+               kernel_wait_quiescent(kernel_tracer_fd);
+       }
+       if (usess) {
+               usess->start_trace = 0;
+               ret = ust_app_stop_trace_all(usess);
+               if (ret < 0) {
+                       ret = LTTCOMM_UST_STOP_FAIL;
+                       goto error;
+               }
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_CREATE_SESSION processed by the client thread.
+  */
+ static int cmd_create_session(char *name, char *path, struct ucred *creds)
+ {
+       int ret;
+       ret = session_create(name, path, creds->uid, creds->gid);
+       if (ret != LTTCOMM_OK) {
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_DESTROY_SESSION processed by the client thread.
+  */
+ static int cmd_destroy_session(struct ltt_session *session, char *name)
+ {
+       int ret;
+       /* Clean kernel session teardown */
+       teardown_kernel_session(session);
+       /* UST session teardown */
+       teardown_ust_session(session);
+       /*
+        * Must notify the kernel thread here to update it's poll setin order
+        * to remove the channel(s)' fd just destroyed.
+        */
+       ret = notify_thread_pipe(kernel_poll_pipe[1]);
+       if (ret < 0) {
+               perror("write kernel poll pipe");
+       }
+       ret = session_destroy(session);
+       return ret;
+ }
+ /*
+  * Command LTTNG_CALIBRATE processed by the client thread.
+  */
+ static int cmd_calibrate(int domain, struct lttng_calibrate *calibrate)
+ {
+       int ret;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+       {
+               struct lttng_kernel_calibrate kcalibrate;
+               kcalibrate.type = calibrate->type;
+               ret = kernel_calibrate(kernel_tracer_fd, &kcalibrate);
+               if (ret < 0) {
+                       ret = LTTCOMM_KERN_ENABLE_FAIL;
+                       goto error;
+               }
+               break;
+       }
+       case LTTNG_DOMAIN_UST:
+       {
+               struct lttng_ust_calibrate ucalibrate;
+               ucalibrate.type = calibrate->type;
+               ret = ust_app_calibrate_glb(&ucalibrate);
+               if (ret < 0) {
+                       ret = LTTCOMM_UST_CALIBRATE_FAIL;
+                       goto error;
+               }
+               break;
+       }
+       default:
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_REGISTER_CONSUMER processed by the client thread.
+  */
+ static int cmd_register_consumer(struct ltt_session *session, int domain,
+               char *sock_path)
+ {
+       int ret, sock;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               /* Can't register a consumer if there is already one */
+               if (session->kernel_session->consumer_fds_sent != 0) {
+                       ret = LTTCOMM_KERN_CONSUMER_FAIL;
+                       goto error;
+               }
+               sock = lttcomm_connect_unix_sock(sock_path);
+               if (sock < 0) {
+                       ret = LTTCOMM_CONNECT_FAIL;
+                       goto error;
+               }
+               session->kernel_session->consumer_fd = sock;
+               break;
+       default:
+               /* TODO: Userspace tracing */
+               ret = LTTCOMM_UND;
+               goto error;
+       }
+       ret = LTTCOMM_OK;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_LIST_DOMAINS processed by the client thread.
+  */
+ static ssize_t cmd_list_domains(struct ltt_session *session,
+               struct lttng_domain **domains)
+ {
+       int ret, index = 0;
+       ssize_t nb_dom = 0;
+       if (session->kernel_session != NULL) {
+               DBG3("Listing domains found kernel domain");
+               nb_dom++;
+       }
+       if (session->ust_session != NULL) {
+               DBG3("Listing domains found UST global domain");
+               nb_dom++;
+       }
+       *domains = zmalloc(nb_dom * sizeof(struct lttng_domain));
+       if (*domains == NULL) {
+               ret = -LTTCOMM_FATAL;
+               goto error;
+       }
+       if (session->kernel_session != NULL) {
+               (*domains)[index].type = LTTNG_DOMAIN_KERNEL;
+               index++;
+       }
+       if (session->ust_session != NULL) {
+               (*domains)[index].type = LTTNG_DOMAIN_UST;
+               index++;
+       }
+       return nb_dom;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_LIST_CHANNELS processed by the client thread.
+  */
+ static ssize_t cmd_list_channels(int domain, struct ltt_session *session,
+               struct lttng_channel **channels)
+ {
+       int ret;
+       ssize_t nb_chan = 0;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               if (session->kernel_session != NULL) {
+                       nb_chan = session->kernel_session->channel_count;
+               }
+               DBG3("Number of kernel channels %zd", nb_chan);
+               break;
+       case LTTNG_DOMAIN_UST:
+               if (session->ust_session != NULL) {
+                       nb_chan = lttng_ht_get_count(
+                                       session->ust_session->domain_global.channels);
+               }
+               DBG3("Number of UST global channels %zd", nb_chan);
+               break;
+       default:
+               *channels = NULL;
+               ret = -LTTCOMM_UND;
+               goto error;
+       }
+       if (nb_chan > 0) {
+               *channels = zmalloc(nb_chan * sizeof(struct lttng_channel));
+               if (*channels == NULL) {
+                       ret = -LTTCOMM_FATAL;
+                       goto error;
+               }
+               list_lttng_channels(domain, session, *channels);
+       } else {
+               *channels = NULL;
+       }
+       return nb_chan;
+ error:
+       return ret;
+ }
+ /*
+  * Command LTTNG_LIST_EVENTS processed by the client thread.
+  */
+ static ssize_t cmd_list_events(int domain, struct ltt_session *session,
+               char *channel_name, struct lttng_event **events)
+ {
+       int ret = 0;
+       ssize_t nb_event = 0;
+       switch (domain) {
+       case LTTNG_DOMAIN_KERNEL:
+               if (session->kernel_session != NULL) {
+                       nb_event = list_lttng_kernel_events(channel_name,
+                                       session->kernel_session, events);
+               }
+               break;
+       case LTTNG_DOMAIN_UST:
+       {
+               if (session->ust_session != NULL) {
+                       nb_event = list_lttng_ust_global_events(channel_name,
+                                       &session->ust_session->domain_global, events);
+               }
+               break;
+       }
+       default:
+               ret = -LTTCOMM_UND;
+               goto error;
+       }
+       ret = nb_event;
+ error:
+       return ret;
+ }
+ /*
+  * Process the command requested by the lttng client within the command
+  * context structure. This function make sure that the return structure (llm)
+  * is set and ready for transmission before returning.
+  *
+  * Return any error encountered or 0 for success.
+  */
+ static int process_client_msg(struct command_ctx *cmd_ctx)
+ {
+       int ret = LTTCOMM_OK;
+       int need_tracing_session = 1;
+       DBG("Processing client command %d", cmd_ctx->lsm->cmd_type);
+       if (opt_no_kernel && cmd_ctx->lsm->domain.type == LTTNG_DOMAIN_KERNEL) {
+               ret = LTTCOMM_KERN_NA;
+               goto error;
+       }
+       /*
+        * Check for command that don't needs to allocate a returned payload. We do
+        * this here so we don't have to make the call for no payload at each
+        * command.
+        */
+       switch(cmd_ctx->lsm->cmd_type) {
+       case LTTNG_LIST_SESSIONS:
+       case LTTNG_LIST_TRACEPOINTS:
+       case LTTNG_LIST_DOMAINS:
+       case LTTNG_LIST_CHANNELS:
+       case LTTNG_LIST_EVENTS:
+               break;
+       default:
+               /* Setup lttng message with no payload */
+               ret = setup_lttng_msg(cmd_ctx, 0);
+               if (ret < 0) {
+                       /* This label does not try to unlock the session */
+                       goto init_setup_error;
+               }
+       }
+       /* Commands that DO NOT need a session. */
+       switch (cmd_ctx->lsm->cmd_type) {
+       case LTTNG_CALIBRATE:
+       case LTTNG_CREATE_SESSION:
+       case LTTNG_LIST_SESSIONS:
+       case LTTNG_LIST_TRACEPOINTS:
+               need_tracing_session = 0;
+               break;
+       default:
+               DBG("Getting session %s by name", cmd_ctx->lsm->session.name);
+               session_lock_list();
+               cmd_ctx->session = session_find_by_name(cmd_ctx->lsm->session.name);
+               session_unlock_list();
+               if (cmd_ctx->session == NULL) {
+                       if (cmd_ctx->lsm->session.name != NULL) {
+                               ret = LTTCOMM_SESS_NOT_FOUND;
+                       } else {
+                               /* If no session name specified */
+                               ret = LTTCOMM_SELECT_SESS;
+                       }
+                       goto error;
+               } else {
+                       /* Acquire lock for the session */
+                       session_lock(cmd_ctx->session);
+               }
+               break;
+       }
+       /*
+        * Check domain type for specific "pre-action".
+        */
+       switch (cmd_ctx->lsm->domain.type) {
+       case LTTNG_DOMAIN_KERNEL:
+               if (!is_root) {
+                       ret = LTTCOMM_KERN_NA;
+                       goto error;
+               }
+               /* Kernel tracer check */
+               if (kernel_tracer_fd == 0) {
+                       /* Basically, load kernel tracer modules */
+                       ret = init_kernel_tracer();
+                       if (ret != 0) {
+                               goto error;
+                       }
+               }
+               /* Need a session for kernel command */
+               if (need_tracing_session) {
+                       if (cmd_ctx->session->kernel_session == NULL) {
+                               ret = create_kernel_session(cmd_ctx->session);
+                               if (ret < 0) {
+                                       ret = LTTCOMM_KERN_SESS_FAIL;
+                                       goto error;
+                               }
+                       }
+                       /* Start the kernel consumer daemon */
+                       pthread_mutex_lock(&kconsumer_data.pid_mutex);
+                       if (kconsumer_data.pid == 0 &&
+                                       cmd_ctx->lsm->cmd_type != LTTNG_REGISTER_CONSUMER) {
+                               pthread_mutex_unlock(&kconsumer_data.pid_mutex);
+                               ret = start_consumerd(&kconsumer_data);
+                               if (ret < 0) {
+                                       ret = LTTCOMM_KERN_CONSUMER_FAIL;
+                                       goto error;
+                               }
+                       } else {
+                               pthread_mutex_unlock(&kconsumer_data.pid_mutex);
+                       }
+               }
+               break;
+       case LTTNG_DOMAIN_UST:
+       {
+               if (need_tracing_session) {
+                       if (cmd_ctx->session->ust_session == NULL) {
+                               ret = create_ust_session(cmd_ctx->session,
+                                               &cmd_ctx->lsm->domain);
+                               if (ret != LTTCOMM_OK) {
+                                       goto error;
+                               }
+                       }
+                       /* Start the UST consumer daemons */
+                       /* 64-bit */
+                       pthread_mutex_lock(&ustconsumer64_data.pid_mutex);
+                       if (consumerd64_bin[0] != '\0' &&
+                                       ustconsumer64_data.pid == 0 &&
+                                       cmd_ctx->lsm->cmd_type != LTTNG_REGISTER_CONSUMER) {
+                               pthread_mutex_unlock(&ustconsumer64_data.pid_mutex);
+                               ret = start_consumerd(&ustconsumer64_data);
+                               if (ret < 0) {
+                                       ret = LTTCOMM_UST_CONSUMER64_FAIL;
+                                       ust_consumerd64_fd = -EINVAL;
+                                       goto error;
+                               }
+                               ust_consumerd64_fd = ustconsumer64_data.cmd_sock;
+                       } else {
+                               pthread_mutex_unlock(&ustconsumer64_data.pid_mutex);
+                       }
+                       /* 32-bit */
+                       if (consumerd32_bin[0] != '\0' &&
+                                       ustconsumer32_data.pid == 0 &&
+                                       cmd_ctx->lsm->cmd_type != LTTNG_REGISTER_CONSUMER) {
+                               pthread_mutex_unlock(&ustconsumer32_data.pid_mutex);
+                               ret = start_consumerd(&ustconsumer32_data);
+                               if (ret < 0) {
+                                       ret = LTTCOMM_UST_CONSUMER32_FAIL;
+                                       ust_consumerd32_fd = -EINVAL;
+                                       goto error;
+                               }
+                               ust_consumerd32_fd = ustconsumer32_data.cmd_sock;
+                       } else {
+                               pthread_mutex_unlock(&ustconsumer32_data.pid_mutex);
+                       }
+               }
+               break;
+       }
+       default:
+               break;
+       }
+       /*
+        * Check that the UID or GID match that of the tracing session.
+        * The root user can interact with all sessions.
+        */
+       if (need_tracing_session) {
+               if (!session_access_ok(cmd_ctx->session,
+                               cmd_ctx->creds.uid, cmd_ctx->creds.gid)) {
+                       ret = LTTCOMM_EPERM;
+                       goto error;
+               }
+       }
+       /* Process by command type */
+       switch (cmd_ctx->lsm->cmd_type) {
+       case LTTNG_ADD_CONTEXT:
+       {
+               ret = cmd_add_context(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.context.channel_name,
+                               cmd_ctx->lsm->u.context.event_name,
+                               &cmd_ctx->lsm->u.context.ctx);
+               break;
+       }
+       case LTTNG_DISABLE_CHANNEL:
+       {
+               ret = cmd_disable_channel(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.disable.channel_name);
+               break;
+       }
+       case LTTNG_DISABLE_EVENT:
+       {
+               ret = cmd_disable_event(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.disable.channel_name,
+                               cmd_ctx->lsm->u.disable.name);
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_DISABLE_ALL_EVENT:
+       {
+               DBG("Disabling all events");
+               ret = cmd_disable_event_all(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.disable.channel_name);
+               break;
+       }
+       case LTTNG_ENABLE_CHANNEL:
+       {
+               ret = cmd_enable_channel(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               &cmd_ctx->lsm->u.channel.chan);
+               break;
+       }
+       case LTTNG_ENABLE_EVENT:
+       {
+               ret = cmd_enable_event(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.enable.channel_name,
+                               &cmd_ctx->lsm->u.enable.event);
+               break;
+       }
+       case LTTNG_ENABLE_ALL_EVENT:
+       {
+               DBG("Enabling all events");
+               ret = cmd_enable_event_all(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.enable.channel_name,
+                               cmd_ctx->lsm->u.enable.event.type);
+               break;
+       }
+       case LTTNG_LIST_TRACEPOINTS:
+       {
+               struct lttng_event *events;
+               ssize_t nb_events;
+               nb_events = cmd_list_tracepoints(cmd_ctx->lsm->domain.type, &events);
+               if (nb_events < 0) {
+                       ret = -nb_events;
+                       goto error;
+               }
+               /*
+                * Setup lttng message with payload size set to the event list size in
+                * bytes and then copy list into the llm payload.
+                */
+               ret = setup_lttng_msg(cmd_ctx, sizeof(struct lttng_event) * nb_events);
+               if (ret < 0) {
+                       free(events);
+                       goto setup_error;
+               }
+               /* Copy event list into message payload */
+               memcpy(cmd_ctx->llm->payload, events,
+                               sizeof(struct lttng_event) * nb_events);
+               free(events);
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_START_TRACE:
+       {
+               ret = cmd_start_trace(cmd_ctx->session);
+               break;
+       }
+       case LTTNG_STOP_TRACE:
+       {
+               ret = cmd_stop_trace(cmd_ctx->session);
+               break;
+       }
+       case LTTNG_CREATE_SESSION:
+       {
++              tracepoint(create_session_start);
+               ret = cmd_create_session(cmd_ctx->lsm->session.name,
+                               cmd_ctx->lsm->session.path, &cmd_ctx->creds);
++              tracepoint(create_session_end);
+               break;
+       }
+       case LTTNG_DESTROY_SESSION:
+       {
++              tracepoint(destroy_session_start);
+               ret = cmd_destroy_session(cmd_ctx->session,
+                               cmd_ctx->lsm->session.name);
++              tracepoint(destroy_session_end);
+               break;
+       }
+       case LTTNG_LIST_DOMAINS:
+       {
+               ssize_t nb_dom;
+               struct lttng_domain *domains;
+               nb_dom = cmd_list_domains(cmd_ctx->session, &domains);
+               if (nb_dom < 0) {
+                       ret = -nb_dom;
+                       goto error;
+               }
+               ret = setup_lttng_msg(cmd_ctx, nb_dom * sizeof(struct lttng_domain));
+               if (ret < 0) {
+                       goto setup_error;
+               }
+               /* Copy event list into message payload */
+               memcpy(cmd_ctx->llm->payload, domains,
+                               nb_dom * sizeof(struct lttng_domain));
+               free(domains);
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_LIST_CHANNELS:
+       {
+               size_t nb_chan;
+               struct lttng_channel *channels;
+               nb_chan = cmd_list_channels(cmd_ctx->lsm->domain.type,
+                               cmd_ctx->session, &channels);
+               if (nb_chan < 0) {
+                       ret = -nb_chan;
+                       goto error;
+               }
+               ret = setup_lttng_msg(cmd_ctx, nb_chan * sizeof(struct lttng_channel));
+               if (ret < 0) {
+                       goto setup_error;
+               }
+               /* Copy event list into message payload */
+               memcpy(cmd_ctx->llm->payload, channels,
+                               nb_chan * sizeof(struct lttng_channel));
+               free(channels);
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_LIST_EVENTS:
+       {
+               ssize_t nb_event;
+               struct lttng_event *events = NULL;
+               nb_event = cmd_list_events(cmd_ctx->lsm->domain.type, cmd_ctx->session,
+                               cmd_ctx->lsm->u.list.channel_name, &events);
+               if (nb_event < 0) {
+                       ret = -nb_event;
+                       goto error;
+               }
+               ret = setup_lttng_msg(cmd_ctx, nb_event * sizeof(struct lttng_event));
+               if (ret < 0) {
+                       goto setup_error;
+               }
+               /* Copy event list into message payload */
+               memcpy(cmd_ctx->llm->payload, events,
+                               nb_event * sizeof(struct lttng_event));
+               free(events);
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_LIST_SESSIONS:
+       {
+               unsigned int nr_sessions;
+               session_lock_list();
+               nr_sessions = lttng_sessions_count(cmd_ctx->creds.uid, cmd_ctx->creds.gid);
+               if (nr_sessions == 0) {
+                       ret = LTTCOMM_NO_SESSION;
+                       session_unlock_list();
+                       goto error;
+               }
+               ret = setup_lttng_msg(cmd_ctx, sizeof(struct lttng_session) * nr_sessions);
+               if (ret < 0) {
+                       session_unlock_list();
+                       goto setup_error;
+               }
+               /* Filled the session array */
+               list_lttng_sessions((struct lttng_session *)(cmd_ctx->llm->payload),
+                       cmd_ctx->creds.uid, cmd_ctx->creds.gid);
+               session_unlock_list();
+               ret = LTTCOMM_OK;
+               break;
+       }
+       case LTTNG_CALIBRATE:
+       {
+               ret = cmd_calibrate(cmd_ctx->lsm->domain.type,
+                               &cmd_ctx->lsm->u.calibrate);
+               break;
+       }
+       case LTTNG_REGISTER_CONSUMER:
+       {
+               ret = cmd_register_consumer(cmd_ctx->session, cmd_ctx->lsm->domain.type,
+                               cmd_ctx->lsm->u.reg.path);
+               break;
+       }
+       default:
+               ret = LTTCOMM_UND;
+               break;
+       }
+ error:
+       if (cmd_ctx->llm == NULL) {
+               DBG("Missing llm structure. Allocating one.");
+               if (setup_lttng_msg(cmd_ctx, 0) < 0) {
+                       goto setup_error;
+               }
+       }
+       /* Set return code */
+       cmd_ctx->llm->ret_code = ret;
+ setup_error:
+       if (cmd_ctx->session) {
+               session_unlock(cmd_ctx->session);
+       }
+ init_setup_error:
+       return ret;
+ }
+ /*
+  * This thread manage all clients request using the unix client socket for
+  * communication.
+  */
+ static void *thread_manage_clients(void *data)
+ {
+       int sock = 0, ret, i, pollfd;
+       uint32_t revents, nb_fd;
+       struct command_ctx *cmd_ctx = NULL;
+       struct lttng_poll_event events;
++      tracepoint(sessiond_th_cli_start);
++
+       DBG("[thread] Manage client started");
+       rcu_register_thread();
+       ret = lttcomm_listen_unix_sock(client_sock);
+       if (ret < 0) {
+               goto error;
+       }
+       /*
+        * Pass 2 as size here for the thread quit pipe and client_sock. Nothing
+        * more will be added to this poll set.
+        */
+       ret = create_thread_poll_set(&events, 2);
+       if (ret < 0) {
+               goto error;
+       }
+       /* Add the application registration socket */
+       ret = lttng_poll_add(&events, client_sock, LPOLLIN | LPOLLPRI);
+       if (ret < 0) {
+               goto error;
+       }
+       /*
+        * Notify parent pid that we are ready to accept command for client side.
+        */
+       if (opt_sig_parent) {
+               kill(ppid, SIGUSR1);
+       }
+       while (1) {
+               DBG("Accepting client command ...");
++              tracepoint(sessiond_th_cli_poll);
++
+               nb_fd = LTTNG_POLL_GETNB(&events);
+               /* Inifinite blocking call, waiting for transmission */
+               ret = lttng_poll_wait(&events, -1);
+               if (ret < 0) {
+                       goto error;
+               }
+               for (i = 0; i < nb_fd; i++) {
+                       /* Fetch once the poll data */
+                       revents = LTTNG_POLL_GETEV(&events, i);
+                       pollfd = LTTNG_POLL_GETFD(&events, i);
+                       /* Thread quit pipe has been closed. Killing thread. */
+                       ret = check_thread_quit_pipe(pollfd, revents);
+                       if (ret) {
+                               goto error;
+                       }
+                       /* Event on the registration socket */
+                       if (pollfd == client_sock) {
+                               if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                                       ERR("Client socket poll error");
+                                       goto error;
+                               }
+                       }
+               }
+               DBG("Wait for client response");
+               sock = lttcomm_accept_unix_sock(client_sock);
+               if (sock < 0) {
+                       goto error;
+               }
+               /* Set socket option for credentials retrieval */
+               ret = lttcomm_setsockopt_creds_unix_sock(sock);
+               if (ret < 0) {
+                       goto error;
+               }
+               /* Allocate context command to process the client request */
+               cmd_ctx = zmalloc(sizeof(struct command_ctx));
+               if (cmd_ctx == NULL) {
+                       perror("zmalloc cmd_ctx");
+                       goto error;
+               }
+               /* Allocate data buffer for reception */
+               cmd_ctx->lsm = zmalloc(sizeof(struct lttcomm_session_msg));
+               if (cmd_ctx->lsm == NULL) {
+                       perror("zmalloc cmd_ctx->lsm");
+                       goto error;
+               }
+               cmd_ctx->llm = NULL;
+               cmd_ctx->session = NULL;
+               /*
+                * Data is received from the lttng client. The struct
+                * lttcomm_session_msg (lsm) contains the command and data request of
+                * the client.
+                */
+               DBG("Receiving data from client ...");
+               ret = lttcomm_recv_creds_unix_sock(sock, cmd_ctx->lsm,
+                               sizeof(struct lttcomm_session_msg), &cmd_ctx->creds);
+               if (ret <= 0) {
+                       DBG("Nothing recv() from client... continuing");
+                       close(sock);
+                       free(cmd_ctx);
+                       continue;
+               }
+               // TODO: Validate cmd_ctx including sanity check for
+               // security purpose.
+               rcu_thread_online();
+               /*
+                * This function dispatch the work to the kernel or userspace tracer
+                * libs and fill the lttcomm_lttng_msg data structure of all the needed
+                * informations for the client. The command context struct contains
+                * everything this function may needs.
+                */
+               ret = process_client_msg(cmd_ctx);
+               rcu_thread_offline();
+               if (ret < 0) {
+                       /*
+                        * TODO: Inform client somehow of the fatal error. At
+                        * this point, ret < 0 means that a zmalloc failed
+                        * (ENOMEM). Error detected but still accept command.
+                        */
+                       clean_command_ctx(&cmd_ctx);
+                       continue;
+               }
+               DBG("Sending response (size: %d, retcode: %s)",
+                               cmd_ctx->lttng_msg_size,
+                               lttng_strerror(-cmd_ctx->llm->ret_code));
+               ret = send_unix_sock(sock, cmd_ctx->llm, cmd_ctx->lttng_msg_size);
+               if (ret < 0) {
+                       ERR("Failed to send data back to client");
+               }
+               /* End of transmission */
+               close(sock);
+               clean_command_ctx(&cmd_ctx);
+       }
+ error:
+       DBG("Client thread dying");
+       unlink(client_unix_sock_path);
+       close(client_sock);
+       close(sock);
+       lttng_poll_clean(&events);
+       clean_command_ctx(&cmd_ctx);
+       rcu_unregister_thread();
+       return NULL;
+ }
+ /*
+  * usage function on stderr
+  */
+ static void usage(void)
+ {
+       fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname);
+       fprintf(stderr, "  -h, --help                         Display this usage.\n");
+       fprintf(stderr, "  -c, --client-sock PATH             Specify path for the client unix socket\n");
+       fprintf(stderr, "  -a, --apps-sock PATH               Specify path for apps unix socket\n");
+       fprintf(stderr, "      --kconsumerd-err-sock PATH     Specify path for the kernel consumer error socket\n");
+       fprintf(stderr, "      --kconsumerd-cmd-sock PATH     Specify path for the kernel consumer command socket\n");
+       fprintf(stderr, "      --ustconsumerd32-err-sock PATH Specify path for the 32-bit UST consumer error socket\n");
+       fprintf(stderr, "      --ustconsumerd64-err-sock PATH Specify path for the 64-bit UST consumer error socket\n");
+       fprintf(stderr, "      --ustconsumerd32-cmd-sock PATH Specify path for the 32-bit UST consumer command socket\n");
+       fprintf(stderr, "      --ustconsumerd64-cmd-sock PATH Specify path for the 64-bit UST consumer command socket\n");
+       fprintf(stderr, "      --consumerd32-path PATH     Specify path for the 32-bit UST consumer daemon binary\n");
+       fprintf(stderr, "      --consumerd32-libdir PATH   Specify path for the 32-bit UST consumer daemon libraries\n");
+       fprintf(stderr, "      --consumerd64-path PATH     Specify path for the 64-bit UST consumer daemon binary\n");
+       fprintf(stderr, "      --consumerd64-libdir PATH   Specify path for the 64-bit UST consumer daemon libraries\n");
+       fprintf(stderr, "  -d, --daemonize                    Start as a daemon.\n");
+       fprintf(stderr, "  -g, --group NAME                   Specify the tracing group name. (default: tracing)\n");
+       fprintf(stderr, "  -V, --version                      Show version number.\n");
+       fprintf(stderr, "  -S, --sig-parent                   Send SIGCHLD to parent pid to notify readiness.\n");
+       fprintf(stderr, "  -q, --quiet                        No output at all.\n");
+       fprintf(stderr, "  -v, --verbose                      Verbose mode. Activate DBG() macro.\n");
+       fprintf(stderr, "      --verbose-consumer             Verbose mode for consumer. Activate DBG() macro.\n");
+       fprintf(stderr, "      --no-kernel                    Disable kernel tracer\n");
+ }
+ /*
+  * daemon argument parsing
+  */
+ static int parse_args(int argc, char **argv)
+ {
+       int c;
+       static struct option long_options[] = {
+               { "client-sock", 1, 0, 'c' },
+               { "apps-sock", 1, 0, 'a' },
+               { "kconsumerd-cmd-sock", 1, 0, 'C' },
+               { "kconsumerd-err-sock", 1, 0, 'E' },
+               { "ustconsumerd32-cmd-sock", 1, 0, 'G' },
+               { "ustconsumerd32-err-sock", 1, 0, 'H' },
+               { "ustconsumerd64-cmd-sock", 1, 0, 'D' },
+               { "ustconsumerd64-err-sock", 1, 0, 'F' },
+               { "consumerd32-path", 1, 0, 'u' },
+               { "consumerd32-libdir", 1, 0, 'U' },
+               { "consumerd64-path", 1, 0, 't' },
+               { "consumerd64-libdir", 1, 0, 'T' },
+               { "daemonize", 0, 0, 'd' },
+               { "sig-parent", 0, 0, 'S' },
+               { "help", 0, 0, 'h' },
+               { "group", 1, 0, 'g' },
+               { "version", 0, 0, 'V' },
+               { "quiet", 0, 0, 'q' },
+               { "verbose", 0, 0, 'v' },
+               { "verbose-consumer", 0, 0, 'Z' },
+               { "no-kernel", 0, 0, 'N' },
+               { NULL, 0, 0, 0 }
+       };
+       while (1) {
+               int option_index = 0;
+               c = getopt_long(argc, argv, "dhqvVSN" "a:c:g:s:C:E:D:F:Z:u:t",
+                               long_options, &option_index);
+               if (c == -1) {
+                       break;
+               }
+               switch (c) {
+               case 0:
+                       fprintf(stderr, "option %s", long_options[option_index].name);
+                       if (optarg) {
+                               fprintf(stderr, " with arg %s\n", optarg);
+                       }
+                       break;
+               case 'c':
+                       snprintf(client_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'a':
+                       snprintf(apps_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'd':
+                       opt_daemon = 1;
+                       break;
+               case 'g':
+                       opt_tracing_group = optarg;
+                       break;
+               case 'h':
+                       usage();
+                       exit(EXIT_FAILURE);
+               case 'V':
+                       fprintf(stdout, "%s\n", VERSION);
+                       exit(EXIT_SUCCESS);
+               case 'S':
+                       opt_sig_parent = 1;
+                       break;
+               case 'E':
+                       snprintf(kconsumer_data.err_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'C':
+                       snprintf(kconsumer_data.cmd_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'F':
+                       snprintf(ustconsumer64_data.err_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'D':
+                       snprintf(ustconsumer64_data.cmd_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'H':
+                       snprintf(ustconsumer32_data.err_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'G':
+                       snprintf(ustconsumer32_data.cmd_unix_sock_path, PATH_MAX, "%s", optarg);
+                       break;
+               case 'N':
+                       opt_no_kernel = 1;
+                       break;
+               case 'q':
+                       opt_quiet = 1;
+                       break;
+               case 'v':
+                       /* Verbose level can increase using multiple -v */
+                       opt_verbose += 1;
+                       break;
+               case 'Z':
+                       opt_verbose_consumer += 1;
+                       break;
+               case 'u':
+                       consumerd32_bin= optarg;
+                       break;
+               case 'U':
+                       consumerd32_libdir = optarg;
+                       break;
+               case 't':
+                       consumerd64_bin = optarg;
+                       break;
+               case 'T':
+                       consumerd64_libdir = optarg;
+                       break;
+               default:
+                       /* Unknown option or other error.
+                        * Error is printed by getopt, just return */
+                       return -1;
+               }
+       }
+       return 0;
+ }
+ /*
+  * Creates the two needed socket by the daemon.
+  *        apps_sock - The communication socket for all UST apps.
+  *        client_sock - The communication of the cli tool (lttng).
+  */
+ static int init_daemon_socket(void)
+ {
+       int ret = 0;
+       mode_t old_umask;
+       old_umask = umask(0);
+       /* Create client tool unix socket */
+       client_sock = lttcomm_create_unix_sock(client_unix_sock_path);
+       if (client_sock < 0) {
+               ERR("Create unix sock failed: %s", client_unix_sock_path);
+               ret = -1;
+               goto end;
+       }
+       /* File permission MUST be 660 */
+       ret = chmod(client_unix_sock_path, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
+       if (ret < 0) {
+               ERR("Set file permissions failed: %s", client_unix_sock_path);
+               perror("chmod");
+               goto end;
+       }
+       /* Create the application unix socket */
+       apps_sock = lttcomm_create_unix_sock(apps_unix_sock_path);
+       if (apps_sock < 0) {
+               ERR("Create unix sock failed: %s", apps_unix_sock_path);
+               ret = -1;
+               goto end;
+       }
+       /* File permission MUST be 666 */
+       ret = chmod(apps_unix_sock_path,
+                       S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH);
+       if (ret < 0) {
+               ERR("Set file permissions failed: %s", apps_unix_sock_path);
+               perror("chmod");
+               goto end;
+       }
+ end:
+       umask(old_umask);
+       return ret;
+ }
+ /*
+  * Check if the global socket is available, and if a daemon is answering at the
+  * other side. If yes, error is returned.
+  */
+ static int check_existing_daemon(void)
+ {
+       if (access(client_unix_sock_path, F_OK) < 0 &&
+                       access(apps_unix_sock_path, F_OK) < 0) {
+               return 0;
+       }
+       /* Is there anybody out there ? */
+       if (lttng_session_daemon_alive()) {
+               return -EEXIST;
+       } else {
+               return 0;
+       }
+ }
+ /*
+  * Set the tracing group gid onto the client socket.
+  *
+  * Race window between mkdir and chown is OK because we are going from more
+  * permissive (root.root) to les permissive (root.tracing).
+  */
+ static int set_permissions(char *rundir)
+ {
+       int ret;
+       gid_t gid;
+       gid = allowed_group();
+       if (gid < 0) {
+               WARN("No tracing group detected");
+               ret = 0;
+               goto end;
+       }
+       /* Set lttng run dir */
+       ret = chown(rundir, 0, gid);
+       if (ret < 0) {
+               ERR("Unable to set group on %s", rundir);
+               perror("chown");
+       }
+       /* lttng client socket path */
+       ret = chown(client_unix_sock_path, 0, gid);
+       if (ret < 0) {
+               ERR("Unable to set group on %s", client_unix_sock_path);
+               perror("chown");
+       }
+       /* kconsumer error socket path */
+       ret = chown(kconsumer_data.err_unix_sock_path, 0, gid);
+       if (ret < 0) {
+               ERR("Unable to set group on %s", kconsumer_data.err_unix_sock_path);
+               perror("chown");
+       }
+       /* 64-bit ustconsumer error socket path */
+       ret = chown(ustconsumer64_data.err_unix_sock_path, 0, gid);
+       if (ret < 0) {
+               ERR("Unable to set group on %s", ustconsumer64_data.err_unix_sock_path);
+               perror("chown");
+       }
+       /* 32-bit ustconsumer compat32 error socket path */
+       ret = chown(ustconsumer32_data.err_unix_sock_path, 0, gid);
+       if (ret < 0) {
+               ERR("Unable to set group on %s", ustconsumer32_data.err_unix_sock_path);
+               perror("chown");
+       }
+       DBG("All permissions are set");
+ end:
+       return ret;
+ }
+ /*
+  * 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 the application command pipe to wake thread_manage_apps.
+  */
+ static int create_apps_cmd_pipe(void)
+ {
+       return pipe2(apps_cmd_pipe, O_CLOEXEC);
+ }
+ /*
+  * Create the lttng run directory needed for all global sockets and pipe.
+  */
+ static int create_lttng_rundir(const char *rundir)
+ {
+       int ret;
+       DBG3("Creating LTTng run directory: %s", rundir);
+       ret = mkdir(rundir, S_IRWXU | S_IRWXG );
+       if (ret < 0) {
+               if (errno != EEXIST) {
+                       ERR("Unable to create %s", rundir);
+                       goto error;
+               } else {
+                       ret = 0;
+               }
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Setup sockets and directory needed by the kconsumerd communication with the
+  * session daemon.
+  */
+ static int set_consumer_sockets(struct consumer_data *consumer_data,
+               const char *rundir)
+ {
+       int ret;
+       char path[PATH_MAX];
+     switch (consumer_data->type) {
+       case LTTNG_CONSUMER_KERNEL:
+               snprintf(path, PATH_MAX, DEFAULT_KCONSUMERD_PATH, rundir);
+               break;
+       case LTTNG_CONSUMER64_UST:
+               snprintf(path, PATH_MAX, DEFAULT_USTCONSUMERD64_PATH, rundir);
+               break;
+       case LTTNG_CONSUMER32_UST:
+               snprintf(path, PATH_MAX, DEFAULT_USTCONSUMERD32_PATH, rundir);
+               break;
+       default:
+               ERR("Consumer type unknown");
+               ret = -EINVAL;
+               goto error;
+       }
+       DBG2("Creating consumer directory: %s", path);
+       ret = mkdir(path, S_IRWXU | S_IRWXG);
+       if (ret < 0) {
+               if (errno != EEXIST) {
+                       ERR("Failed to create %s", path);
+                       goto error;
+               }
+               ret = 0;
+       }
+       /* Create the kconsumerd error unix socket */
+       consumer_data->err_sock =
+               lttcomm_create_unix_sock(consumer_data->err_unix_sock_path);
+       if (consumer_data->err_sock < 0) {
+               ERR("Create unix sock failed: %s", consumer_data->err_unix_sock_path);
+               ret = -1;
+               goto error;
+       }
+       /* File permission MUST be 660 */
+       ret = chmod(consumer_data->err_unix_sock_path,
+                       S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
+       if (ret < 0) {
+               ERR("Set file permissions failed: %s", consumer_data->err_unix_sock_path);
+               PERROR("chmod");
+               goto error;
+       }
+ error:
+       return ret;
+ }
+ /*
+  * Signal handler for the daemon
+  *
+  * Simply stop all worker threads, leaving main() return gracefully after
+  * joining all threads and calling cleanup().
+  */
+ static void sighandler(int sig)
+ {
+       switch (sig) {
+       case SIGPIPE:
+               DBG("SIGPIPE caught");
+               return;
+       case SIGINT:
+               DBG("SIGINT caught");
+               stop_threads();
+               break;
+       case SIGTERM:
+               DBG("SIGTERM caught");
+               stop_threads();
+               break;
+       default:
+               break;
+       }
+ }
+ /*
+  * Setup signal handler for :
+  *            SIGINT, SIGTERM, SIGPIPE
+  */
+ static int set_signal_handler(void)
+ {
+       int ret = 0;
+       struct sigaction sa;
+       sigset_t sigset;
+       if ((ret = sigemptyset(&sigset)) < 0) {
+               perror("sigemptyset");
+               return ret;
+       }
+       sa.sa_handler = sighandler;
+       sa.sa_mask = sigset;
+       sa.sa_flags = 0;
+       if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) {
+               perror("sigaction");
+               return ret;
+       }
+       if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) {
+               perror("sigaction");
+               return ret;
+       }
+       if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) {
+               perror("sigaction");
+               return ret;
+       }
+       DBG("Signal handler set for SIGTERM, SIGPIPE and SIGINT");
+       return ret;
+ }
+ /*
+  * Set open files limit to unlimited. This daemon can open a large number of
+  * file descriptors in order to consumer multiple kernel traces.
+  */
+ static void set_ulimit(void)
+ {
+       int ret;
+       struct rlimit lim;
+       /* The kernel does not allowed an infinite limit for open files */
+       lim.rlim_cur = 65535;
+       lim.rlim_max = 65535;
+       ret = setrlimit(RLIMIT_NOFILE, &lim);
+       if (ret < 0) {
+               perror("failed to set open files limit");
+       }
+ }
+ /*
+  * main
+  */
+ int main(int argc, char **argv)
+ {
+       int ret = 0;
+       void *status;
+       const char *home_path;
++      tracepoint(sessiond_boot_start);
+       rcu_register_thread();
+       /* Create thread quit pipe */
+       if ((ret = init_thread_quit_pipe()) < 0) {
+               goto error;
+       }
+       setup_consumerd_path();
+       /* Parse arguments */
+       progname = argv[0];
+       if ((ret = parse_args(argc, argv) < 0)) {
+               goto error;
+       }
+       /* Daemonize */
+       if (opt_daemon) {
+               ret = daemon(0, 0);
+               if (ret < 0) {
+                       perror("daemon");
+                       goto error;
+               }
+       }
+       /* Check if daemon is UID = 0 */
+       is_root = !getuid();
+       if (is_root) {
+               rundir = strdup(DEFAULT_LTTNG_RUNDIR);
+               /* Create global run dir with root access */
+               ret = create_lttng_rundir(rundir);
+               if (ret < 0) {
+                       goto error;
+               }
+               if (strlen(apps_unix_sock_path) == 0) {
+                       snprintf(apps_unix_sock_path, PATH_MAX,
+                                       DEFAULT_GLOBAL_APPS_UNIX_SOCK);
+               }
+               if (strlen(client_unix_sock_path) == 0) {
+                       snprintf(client_unix_sock_path, PATH_MAX,
+                                       DEFAULT_GLOBAL_CLIENT_UNIX_SOCK);
+               }
+               /* Set global SHM for ust */
+               if (strlen(wait_shm_path) == 0) {
+                       snprintf(wait_shm_path, PATH_MAX,
+                                       DEFAULT_GLOBAL_APPS_WAIT_SHM_PATH);
+               }
+               /* Setup kernel consumerd path */
+               snprintf(kconsumer_data.err_unix_sock_path, PATH_MAX,
+                               DEFAULT_KCONSUMERD_ERR_SOCK_PATH, rundir);
+               snprintf(kconsumer_data.cmd_unix_sock_path, PATH_MAX,
+                               DEFAULT_KCONSUMERD_CMD_SOCK_PATH, rundir);
+               DBG2("Kernel consumer err path: %s",
+                               kconsumer_data.err_unix_sock_path);
+               DBG2("Kernel consumer cmd path: %s",
+                               kconsumer_data.cmd_unix_sock_path);
+       } else {
+               home_path = get_home_dir();
+               if (home_path == NULL) {
+                       /* TODO: Add --socket PATH option */
+                       ERR("Can't get HOME directory for sockets creation.");
+                       ret = -EPERM;
+                       goto error;
+               }
+               /*
+                * Create rundir from home path. This will create something like
+                * $HOME/.lttng
+                */
+               ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
+               if (ret < 0) {
+                       ret = -ENOMEM;
+                       goto error;
+               }
+               ret = create_lttng_rundir(rundir);
+               if (ret < 0) {
+                       goto error;
+               }
+               if (strlen(apps_unix_sock_path) == 0) {
+                       snprintf(apps_unix_sock_path, PATH_MAX,
+                                       DEFAULT_HOME_APPS_UNIX_SOCK, home_path);
+               }
+               /* Set the cli tool unix socket path */
+               if (strlen(client_unix_sock_path) == 0) {
+                       snprintf(client_unix_sock_path, PATH_MAX,
+                                       DEFAULT_HOME_CLIENT_UNIX_SOCK, home_path);
+               }
+               /* Set global SHM for ust */
+               if (strlen(wait_shm_path) == 0) {
+                       snprintf(wait_shm_path, PATH_MAX,
+                                       DEFAULT_HOME_APPS_WAIT_SHM_PATH, geteuid());
+               }
+       }
+       DBG("Client socket path %s", client_unix_sock_path);
+       DBG("Application socket path %s", apps_unix_sock_path);
+       DBG("LTTng run directory path: %s", rundir);
+       /* 32 bits consumerd path setup */
+       snprintf(ustconsumer32_data.err_unix_sock_path, PATH_MAX,
+                       DEFAULT_USTCONSUMERD32_ERR_SOCK_PATH, rundir);
+       snprintf(ustconsumer32_data.cmd_unix_sock_path, PATH_MAX,
+                       DEFAULT_USTCONSUMERD32_CMD_SOCK_PATH, rundir);
+       DBG2("UST consumer 32 bits err path: %s",
+                       ustconsumer32_data.err_unix_sock_path);
+       DBG2("UST consumer 32 bits cmd path: %s",
+                       ustconsumer32_data.cmd_unix_sock_path);
+       /* 64 bits consumerd path setup */
+       snprintf(ustconsumer64_data.err_unix_sock_path, PATH_MAX,
+                       DEFAULT_USTCONSUMERD64_ERR_SOCK_PATH, rundir);
+       snprintf(ustconsumer64_data.cmd_unix_sock_path, PATH_MAX,
+                       DEFAULT_USTCONSUMERD64_CMD_SOCK_PATH, rundir);
+       DBG2("UST consumer 64 bits err path: %s",
+                       ustconsumer64_data.err_unix_sock_path);
+       DBG2("UST consumer 64 bits cmd path: %s",
+                       ustconsumer64_data.cmd_unix_sock_path);
+       /*
+        * See if daemon already exist.
+        */
+       if ((ret = check_existing_daemon()) < 0) {
+               ERR("Already running daemon.\n");
+               /*
+                * We do not goto exit because we must not cleanup()
+                * because a daemon is already running.
+                */
+               goto error;
+       }
+       /* After this point, we can safely call cleanup() with "goto exit" */
+       /*
+        * These actions must be executed as root. We do that *after* setting up
+        * the sockets path because we MUST make the check for another daemon using
+        * those paths *before* trying to set the kernel consumer sockets and init
+        * kernel tracer.
+        */
+       if (is_root) {
+               ret = set_consumer_sockets(&kconsumer_data, rundir);
+               if (ret < 0) {
+                       goto exit;
+               }
+               /* Setup kernel tracer */
+               if (!opt_no_kernel) {
+                       init_kernel_tracer();
+               }
+               /* Set ulimit for open files */
+               set_ulimit();
+       }
+       ret = set_consumer_sockets(&ustconsumer64_data, rundir);
+       if (ret < 0) {
+               goto exit;
+       }
+       ret = set_consumer_sockets(&ustconsumer32_data, rundir);
+       if (ret < 0) {
+               goto exit;
+       }
+       if ((ret = set_signal_handler()) < 0) {
+               goto exit;
+       }
+       /* Setup the needed unix socket */
+       if ((ret = init_daemon_socket()) < 0) {
+               goto exit;
+       }
+       /* Set credentials to socket */
+       if (is_root && ((ret = set_permissions(rundir)) < 0)) {
+               goto exit;
+       }
+       /* Get parent pid if -S, --sig-parent is specified. */
+       if (opt_sig_parent) {
+               ppid = getppid();
+       }
+       /* Setup the kernel pipe for waking up the kernel thread */
+       if ((ret = create_kernel_poll_pipe()) < 0) {
+               goto exit;
+       }
+       /* Setup the thread apps communication pipe. */
+       if ((ret = create_apps_cmd_pipe()) < 0) {
+               goto exit;
+       }
+       /* Init UST command queue. */
+       cds_wfq_init(&ust_cmd_queue.queue);
+       /* Init UST app hash table */
+       ust_app_ht_alloc();
+       /*
+        * Get session list pointer. This pointer MUST NOT be free(). This list is
+        * statically declared in session.c
+        */
+       session_list_ptr = session_get_list();
+       /* Set up max poll set size */
+       lttng_poll_set_max_size();
+       /* Create thread to manage the client socket */
+       ret = pthread_create(&client_thread, NULL,
+                       thread_manage_clients, (void *) NULL);
+       if (ret != 0) {
+               perror("pthread_create clients");
+               goto exit_client;
+       }
+       /* Create thread to dispatch registration */
+       ret = pthread_create(&dispatch_thread, NULL,
+                       thread_dispatch_ust_registration, (void *) NULL);
+       if (ret != 0) {
+               perror("pthread_create dispatch");
+               goto exit_dispatch;
+       }
+       /* Create thread to manage application registration. */
+       ret = pthread_create(&reg_apps_thread, NULL,
+                       thread_registration_apps, (void *) NULL);
+       if (ret != 0) {
+               perror("pthread_create registration");
+               goto exit_reg_apps;
+       }
+       /* Create thread to manage application socket */
+       ret = pthread_create(&apps_thread, NULL,
+                       thread_manage_apps, (void *) NULL);
+       if (ret != 0) {
+               perror("pthread_create apps");
+               goto exit_apps;
+       }
+       /* Create kernel thread to manage kernel event */
+       ret = pthread_create(&kernel_thread, NULL,
+                       thread_manage_kernel, (void *) NULL);
+       if (ret != 0) {
+               perror("pthread_create kernel");
+               goto exit_kernel;
+       }
++      tracepoint(sessiond_boot_end);
++
+       ret = pthread_join(kernel_thread, &status);
+       if (ret != 0) {
+               perror("pthread_join");
+               goto error;     /* join error, exit without cleanup */
+       }
+ exit_kernel:
+       ret = pthread_join(apps_thread, &status);
+       if (ret != 0) {
+               perror("pthread_join");
+               goto error;     /* join error, exit without cleanup */
+       }
+ exit_apps:
+       ret = pthread_join(reg_apps_thread, &status);
+       if (ret != 0) {
+               perror("pthread_join");
+               goto error;     /* join error, exit without cleanup */
+       }
+ exit_reg_apps:
+       ret = pthread_join(dispatch_thread, &status);
+       if (ret != 0) {
+               perror("pthread_join");
+               goto error;     /* join error, exit without cleanup */
+       }
+ exit_dispatch:
+       ret = pthread_join(client_thread, &status);
+       if (ret != 0) {
+               perror("pthread_join");
+               goto error;     /* join error, exit without cleanup */
+       }
+       ret = join_consumer_thread(&kconsumer_data);
+       if (ret != 0) {
+               perror("join_consumer");
+               goto error;     /* join error, exit without cleanup */
+       }
+ exit_client:
+ exit:
+       /*
+        * cleanup() is called when no other thread is running.
+        */
+       rcu_thread_online();
+       cleanup();
+       rcu_thread_offline();
+       rcu_unregister_thread();
+       if (!ret) {
+               exit(EXIT_SUCCESS);
+       }
+ error:
+       exit(EXIT_FAILURE);
+ }
index 0000000000000000000000000000000000000000,77c09995cf194d41620317c90e4ce6b961b8b1af..0bf3ff58f0dd138e553f9c1dafaa38d418b1d09c
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,160 +1,172 @@@
+ /*
+  * Copyright (C)  2011 - David Goulet <david.goulet@polymtl.ca>
+  *                       Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+  *
+  * This program is free software; you can redistribute it and/or modify it
+  * under the terms of the GNU General Public License as published by the Free
+  * Software Foundation; only version 2 of the License.
+  *
+  * This program is distributed in the hope that it will be useful, but WITHOUT
+  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+  * more details.
+  *
+  * You should have received a copy of the GNU General Public License along with
+  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+  * Place - Suite 330, Boston, MA  02111-1307, USA.
+  */
+ #define _GNU_SOURCE
+ #include <fcntl.h>
+ #include <limits.h>
+ #include <sys/mman.h>
+ #include <sys/stat.h>
+ #include <sys/types.h>
+ #include <sys/wait.h>
+ #include <unistd.h>
+ #include <urcu.h>
+ #include <common/error.h>
++#include "benchmark.h"
++#include "measures.h"
+ #include "shm.h"
+ /*
+  * Using fork to set umask in the child process (not multi-thread safe). We
+  * deal with the shm_open vs ftruncate race (happening when the sessiond owns
+  * the shm and does not let everybody modify it, to ensure safety against
+  * shm_unlink) by simply letting the mmap fail and retrying after a few
+  * seconds. For global shm, everybody has rw access to it until the sessiond
+  * starts.
+  */
+ static int get_wait_shm(char *shm_path, size_t mmap_size, int global)
+ {
+       int wait_shm_fd, ret;
+       mode_t mode;
++      tracepoint(ust_notify_perms_start);
++
+       /* Default permissions */
+       mode = S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP;
+       /* Change owner of the shm path */
+       if (global) {
+               ret = chown(shm_path, 0, 0);
+               if (ret < 0) {
+                       if (errno != ENOENT) {
+                               perror("chown wait shm");
+                               goto error;
+                       }
+               }
+               /*
+                * If global session daemon, any application can register so the shm
+                * needs to be set in read-only mode for others.
+                */
+               mode |= S_IROTH;
+       } else {
+               ret = chown(shm_path, getuid(), getgid());
+               if (ret < 0) {
+                       if (errno != ENOENT) {
+                               perror("chown wait shm");
+                               goto error;
+                       }
+               }
+       }
+       /*
+        * Set permissions to the shm even if we did not create the shm.
+        */
+       ret = chmod(shm_path, mode);
+       if (ret < 0) {
+               if (errno != ENOENT) {
+                       perror("chmod wait shm");
+                       goto error;
+               }
+       }
+       /*
+        * We're alone in a child process, so we can modify the process-wide
+        * umask.
+        */
+       umask(~mode);
++      tracepoint(ust_notify_perms_stop);
++
++      tracepoint(ust_notify_shm_start);
+       /*
+        * Try creating shm (or get rw access). We don't do an exclusive open,
+        * because we allow other processes to create+ftruncate it concurrently.
+        */
+       wait_shm_fd = shm_open(shm_path, O_RDWR | O_CREAT, mode);
+       if (wait_shm_fd < 0) {
+               perror("shm_open wait shm");
+               goto error;
+       }
+       ret = ftruncate(wait_shm_fd, mmap_size);
+       if (ret < 0) {
+               perror("ftruncate wait shm");
+               exit(EXIT_FAILURE);
+       }
+       ret = fchmod(wait_shm_fd, mode);
+       if (ret < 0) {
+               perror("fchmod");
+               exit(EXIT_FAILURE);
+       }
++      tracepoint(ust_notify_shm_stop);
++
+       DBG("Got the wait shm fd %d", wait_shm_fd);
+       return wait_shm_fd;
+ error:
+       DBG("Failing to get the wait shm fd");
+       return -1;
+ }
+ /*
+  * Return the wait shm mmap for UST application notification. The global
+  * variable is used to indicate if the the session daemon is global
+  * (root:tracing) or running with an unprivileged user.
+  *
+  * This returned value is used by futex_wait_update() in futex.c to WAKE all
+  * waiters which are UST application waiting for a session daemon.
+  */
+ char *shm_ust_get_mmap(char *shm_path, int global)
+ {
+       size_t mmap_size = sysconf(_SC_PAGE_SIZE);
+       int wait_shm_fd, ret;
+       char *wait_shm_mmap;
+       wait_shm_fd = get_wait_shm(shm_path, mmap_size, global);
+       if (wait_shm_fd < 0) {
+               goto error;
+       }
++      tracepoint(ust_notify_mmap_start);
++
+       wait_shm_mmap = mmap(NULL, mmap_size, PROT_WRITE | PROT_READ,
+                       MAP_SHARED, wait_shm_fd, 0);
++      tracepoint(ust_notify_mmap_stop);
+       /* close shm fd immediately after taking the mmap reference */
+       ret = close(wait_shm_fd);
+       if (ret) {
+               perror("Error closing fd");
+       }
+       if (wait_shm_mmap == MAP_FAILED) {
+               DBG("mmap error (can be caused by race with ust).");
+               goto error;
+       }
+       return wait_shm_mmap;
+ error:
+       return NULL;
+ }
This page took 0.077946 seconds and 4 git commands to generate.