Add ust basic tracing implementation
authorMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mon, 26 Sep 2011 15:28:57 +0000 (11:28 -0400)
committerMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mon, 26 Sep 2011 15:28:57 +0000 (11:28 -0400)
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
tests/ust-basic-tracing/Makefile.am [new file with mode: 0644]
tests/ust-basic-tracing/ust-basic-tracing.c [new file with mode: 0644]

diff --git a/tests/ust-basic-tracing/Makefile.am b/tests/ust-basic-tracing/Makefile.am
new file mode 100644 (file)
index 0000000..676ae2e
--- /dev/null
@@ -0,0 +1,6 @@
+AM_CPPFLAGS = -I$(top_srcdir)/include -I$(top_srcdir)/libust
+
+noinst_PROGRAMS = ust-basic-tracing
+ust_basic_tracing_SOURCES = ust-basic-tracing.c
+ust_basic_tracing_LDADD = $(top_builddir)/liblttng-ust-comm/liblttng-ust-comm.la \
+       $(top_builddir)/libringbuffer/libringbuffer.la
diff --git a/tests/ust-basic-tracing/ust-basic-tracing.c b/tests/ust-basic-tracing/ust-basic-tracing.c
new file mode 100644 (file)
index 0000000..437b9c2
--- /dev/null
@@ -0,0 +1,978 @@
+/*
+ * ust-basic-tracing.c - Basic single-session, single-channel, single-process UST tracing
+ *
+ * Copyright (C) 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ *
+ * 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 _LARGEFILE64_SOURCE
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <signal.h>
+#include <unistd.h>
+#include <sys/wait.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <sys/mman.h>
+#include <limits.h>
+#include <urcu/futex.h>
+#include <urcu/uatomic.h>
+#include <assert.h>
+#include <sys/socket.h>
+
+#include "lttng-ust-comm.h"
+#include "../../libringbuffer/backend.h"
+#include "../../libringbuffer/frontend.h"
+
+#define MAX_NR_STREAMS 64
+#define MAX_NR_EVENTS  128
+
+struct object_data {
+       int handle;
+       int shm_fd;
+       int wait_fd;
+       uint64_t memory_map_size;
+};
+
+static int session_handle;
+static struct object_data metadata_stream_data;
+static struct object_data metadata_data;
+static struct object_data channel_data;
+static struct object_data stream_data[MAX_NR_STREAMS];
+static int event_handle[MAX_NR_EVENTS];
+
+static int apps_socket = -1;
+static char apps_sock_path[PATH_MAX];
+static char local_apps_wait_shm_path[PATH_MAX];
+
+static volatile int quit_program;
+
+static void handle_signals(int signo)
+{
+       quit_program = 1;
+}
+
+static int send_app_msg(int sock, struct lttcomm_ust_msg *lum)
+{
+       ssize_t len;
+
+       len = lttcomm_send_unix_sock(sock, lum, sizeof(*lum));
+       switch (len) {
+       case sizeof(*lum):
+               printf("message successfully sent\n");
+               break;
+       case -1:
+               if (errno == ECONNRESET) {
+                       printf("remote end closed connection\n");
+                       return 0;
+               }
+               return -1;
+       default:
+               printf("incorrect message size: %zd\n", len);
+               return -1;
+       }
+       return 0;
+}
+
+static int recv_app_reply(int sock, struct lttcomm_ust_reply *lur,
+                         uint32_t expected_handle, uint32_t expected_cmd)
+{
+       ssize_t len;
+
+       memset(lur, 0, sizeof(*lur));
+       len = lttcomm_recv_unix_sock(sock, lur, sizeof(*lur));
+       switch (len) {
+       case 0: /* orderly shutdown */
+               printf("Application has performed an orderly shutdown\n");
+               return -EINVAL;
+       case sizeof(*lur):
+               printf("result message received\n");
+               if (lur->handle != expected_handle) {
+                       printf("Unexpected result message handle\n");
+                       return -EINVAL;
+               }
+
+               if (lur->cmd != expected_cmd) {
+                       printf("Unexpected result message command\n");
+                       return -EINVAL;
+               }
+               if (lur->ret_code != LTTCOMM_OK) {
+                       printf("remote operation failed with code %d.\n",
+                               lur->ret_code);
+                       return lur->ret_code;
+               }
+               return 0;
+       case -1:
+               if (errno == ECONNRESET) {
+                       printf("remote end closed connection\n");
+                       return -EINVAL;
+               }
+               return -1;
+       default:
+               printf("incorrect message size: %zd\n", len);
+               return len > 0 ? -1 : len;
+       }
+}
+
+static int send_app_cmd(int sock,
+                       struct lttcomm_ust_msg *lum,
+                       struct lttcomm_ust_reply *lur)
+{
+       int ret;
+
+       ret = send_app_msg(sock, lum);
+       if (ret)
+               return ret;
+       ret = recv_app_reply(sock, lur, lum->handle, lum->cmd);
+       if (ret)
+               return ret;
+       return 0;
+}
+
+
+/*
+ * Receives a single fd from socket.
+ *
+ * Returns the size of received data
+ */
+static int lttcomm_recv_fd(int sock)
+{
+       struct iovec iov[1];
+       int ret = 0;
+       int data_fd;
+       struct cmsghdr *cmsg;
+       char recv_fd[CMSG_SPACE(sizeof(int))];
+       struct msghdr msg = { 0 };
+
+       /* Prepare to receive the structures */
+       iov[0].iov_base = &data_fd;
+       iov[0].iov_len = sizeof(data_fd);
+       msg.msg_iov = iov;
+       msg.msg_iovlen = 1;
+       msg.msg_control = recv_fd;
+       msg.msg_controllen = sizeof(recv_fd);
+
+       printf("Waiting to receive fd\n");
+       if ((ret = recvmsg(sock, &msg, 0)) < 0) {
+               perror("recvmsg");
+               goto end;
+       }
+       if (ret != sizeof(data_fd)) {
+               printf("Received %d bytes, expected %ld", ret, sizeof(data_fd));
+               goto end;
+       }
+       cmsg = CMSG_FIRSTHDR(&msg);
+       if (!cmsg) {
+               printf("Invalid control message header\n");
+               ret = -1;
+               goto end;
+       }
+       if (cmsg->cmsg_level != SOL_SOCKET || cmsg->cmsg_type != SCM_RIGHTS) {
+               printf("Didn't received any fd\n");
+               ret = -1;
+               goto end;
+       }
+       /* this is our fd */
+       ret = ((int *) CMSG_DATA(cmsg))[0];
+       printf("received fd %d\n", ret);
+end:
+       return ret;
+}
+
+
+static
+int open_streams(int sock, int channel_handle, struct object_data *stream_datas,
+               int nr_check)
+{
+       int ret, k = 0;
+
+       for (;;) {
+               struct lttcomm_ust_msg lum;
+               struct lttcomm_ust_reply lur;
+
+               memset(&lum, 0, sizeof(lum));
+               lum.handle = channel_handle;
+               lum.cmd = LTTNG_UST_STREAM;
+               ret = send_app_cmd(sock, &lum, &lur);
+               if (!ret) {
+                       assert(k < nr_check);
+                       stream_datas[k].handle = lur.ret_val;
+                       printf("received stream handle %u\n",
+                               stream_datas[k].handle);
+                       if (lur.ret_code == LTTCOMM_OK) {
+                               ssize_t len;
+
+                               stream_datas[k].memory_map_size = lur.u.stream.memory_map_size;
+                               /* get shm fd */
+                               len = lttcomm_recv_fd(sock);
+                               if (len < 0)
+                                       return -EINVAL;
+                               stream_datas[k].shm_fd = len;
+                               /* get wait fd */
+                               len = lttcomm_recv_fd(sock);
+                               if (len < 0)
+                                       return -EINVAL;
+                               stream_datas[k].wait_fd = len;
+                       }
+                       k++;
+               }
+               if (ret == -ENOENT)
+                       break;
+               if (ret)
+                       return ret;
+       }
+       return 0;
+}
+
+static
+int close_streams(int sock, struct object_data *stream_datas, int nr_check)
+{
+       int ret, k;
+
+       for (k = 0; k < nr_check; k++) {
+               struct lttcomm_ust_msg lum;
+               struct lttcomm_ust_reply lur;
+
+               if (!stream_datas[k].handle)
+                       continue;
+               memset(&lum, 0, sizeof(lum));
+               lum.handle = stream_datas[k].handle;
+               lum.cmd = LTTNG_UST_RELEASE;
+               ret = send_app_cmd(sock, &lum, &lur);
+               if (ret) {
+                       printf("Error closing stream\n");
+                       return ret;
+               }
+               if (stream_datas[k].shm_fd >= 0) {
+                       ret = close(stream_datas[k].shm_fd);
+                       if (ret) {
+                               printf("Error closing stream shm_fd\n");
+                               return ret;
+                       }
+               }
+               if (stream_datas[k].wait_fd >= 0) {
+                       ret = close(stream_datas[k].wait_fd);
+                       if (ret) {
+                               printf("Error closing stream wait_fd\n");
+                               return ret;
+                       }
+               }
+       }
+       return 0;
+}
+
+static
+struct shm_handle *map_channel(struct object_data *chan_data,
+               struct object_data *stream_datas, int nr_check)
+{
+       struct shm_handle *handle;
+       struct channel *chan;
+       int k, ret;
+
+       /* map metadata channel */
+       handle = channel_handle_create(chan_data->shm_fd,
+               chan_data->wait_fd,
+               chan_data->memory_map_size);
+       if (!handle) {
+               printf("create handle error\n");
+               return NULL;
+       }
+       chan_data->shm_fd = -1;
+       chan_data->wait_fd = -1;
+       chan = shmp(handle, handle->chan);
+
+       for (k = 0; k < nr_check; k++) {
+               struct object_data *stream_data = &stream_datas[k];
+
+               if (!stream_data->handle)
+                       break;
+               /* map stream */
+               ret = channel_handle_add_stream(handle,
+                       stream_data->shm_fd,
+                       stream_data->wait_fd,
+                       stream_data->memory_map_size);
+               if (ret) {
+                       printf("add stream error\n");
+                       goto error_destroy;
+               }
+               stream_data->shm_fd = -1;
+               stream_data->wait_fd = -1;
+       }
+       return handle;
+
+error_destroy:
+       channel_destroy(chan, handle, 1);
+       return NULL;
+}
+
+static
+void unmap_channel(struct shm_handle *handle)
+{
+       struct channel *chan;
+
+       chan = shmp(handle, handle->chan);
+       /* unmap channel */
+       channel_destroy(chan, handle, 1);
+}
+
+static
+int consume_stream(struct shm_handle *handle, int cpu, char *outfile)
+{
+       struct channel *chan;
+       struct lib_ring_buffer *buf;
+       int outfd, ret;
+       int shm_fd, wait_fd;
+       uint64_t memory_map_size;
+
+       chan = shmp(handle, handle->chan);
+
+       /* open stream */
+       buf = channel_get_ring_buffer(&chan->backend.config,
+               chan, cpu, handle, &shm_fd, &wait_fd, &memory_map_size);
+       if (!buf)
+               return -ENOENT;
+       ret = lib_ring_buffer_open_read(buf, handle, 1);
+       if (ret) {
+               return -1;
+       }
+
+       /* copy */
+       outfd = open(outfile, O_WRONLY | O_CREAT | O_LARGEFILE | O_TRUNC,
+                       S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
+       if (outfd < 0) {
+               perror("open output");
+               return -1;
+       }
+
+       printf("Waiting for buffer data for %s\n", outfile);
+       for (;;) {
+               unsigned long read_size;
+               unsigned long copy_size;
+               char *ptr;
+
+               ret = lib_ring_buffer_get_next_subbuf(buf, handle);
+               printf("get next ret %d\n", ret);
+               if (ret == -ENODATA)
+                       break;
+               if (ret == -EAGAIN) {
+                       sleep(1);
+                       continue;
+               }
+               if (ret) {
+                       printf("Error %d in lib_ring_buffer_get_next_subbuf\n", ret);
+                       return -1;
+               }
+               read_size = lib_ring_buffer_get_read_data_size(
+                       &chan->backend.config, buf, handle);
+               read_size = PAGE_ALIGN(read_size);
+               ptr = lib_ring_buffer_read_offset_address(
+                       &buf->backend, 0, handle);
+               printf("WRITE: copy %lu bytes\n", read_size);
+               copy_size = write(outfd, ptr, read_size);
+               if (copy_size < read_size) {
+                       printf("write issue: copied %zd, expected %lu\n", copy_size, read_size);
+               }
+               lib_ring_buffer_put_next_subbuf(buf, handle);
+       }
+
+       ret = close(outfd);
+       if (ret) {
+               perror("close");
+               return -1;
+       }
+
+       /* close stream */
+       lib_ring_buffer_release_read(buf, handle, 1);
+       return 0;
+}
+
+static
+int consume_buffers(const char *outputpath)
+{
+       int k, ret;
+       mode_t old_umask;
+       char pathname[PATH_MAX];
+       struct shm_handle *handle;
+
+       snprintf(pathname, PATH_MAX - 1, "%s", outputpath);
+       old_umask = umask(0);
+       ret = mkdir(pathname, S_IRWXU | S_IRWXG);
+       if (ret && errno != EEXIST) {
+               perror("mkdir");
+               umask(old_umask);
+               return -1;
+       }
+       umask(old_umask);
+
+       /* copy metadata */
+       handle = map_channel(&metadata_data,
+                       &metadata_stream_data, 1);
+       if (!handle)
+               return -1;
+       snprintf(pathname, PATH_MAX - 1,
+               "%s/metadata", outputpath);
+       ret = consume_stream(handle, -1, pathname);
+       if (ret && ret != -ENOENT) {
+               printf("Error in consume_stream\n");
+               return ret;
+       }
+       unmap_channel(handle);
+
+       /* copy binary data */
+       handle = map_channel(&channel_data,
+                       stream_data, MAX_NR_STREAMS);
+       if (!handle)
+               return -1;
+       for (k = 0; k < MAX_NR_STREAMS; k++) {
+               snprintf(pathname, PATH_MAX - 1,
+                       "%s/data_%u", outputpath, k);
+               ret = consume_stream(handle, k, pathname);
+               if (ret && ret != -ENOENT) {
+                       printf("Error in consume_stream\n");
+                       return ret;
+               }
+       }
+       unmap_channel(handle);
+
+       return 0;
+}
+
+int send_app_msgs(int sock, const char *outputpath,
+               unsigned int nr_events, const char **event_names)
+{
+       struct lttcomm_ust_msg lum;
+       struct lttcomm_ust_reply lur;
+       int ret, k;
+
+       /* Create session */
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = LTTNG_UST_ROOT_HANDLE;
+       lum.cmd = LTTNG_UST_SESSION;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       session_handle = lur.ret_val;
+       printf("received session handle %u\n", session_handle);
+
+       /* Create metadata channel */
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = session_handle;
+       lum.cmd = LTTNG_UST_METADATA;
+       lum.u.channel.overwrite = 0;
+       lum.u.channel.subbuf_size = 32768;
+       lum.u.channel.num_subbuf = 4;
+       lum.u.channel.switch_timer_interval = 0;
+       lum.u.channel.read_timer_interval = 0;
+       lum.u.channel.output = LTTNG_UST_MMAP;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       metadata_data.handle = lur.ret_val;
+       printf("received metadata handle %u\n", metadata_data.handle);
+       if (lur.ret_code == LTTCOMM_OK) {
+               ssize_t len;
+
+               metadata_data.memory_map_size = lur.u.channel.memory_map_size;
+               /* get shm fd */
+               len = lttcomm_recv_fd(sock);
+               if (len < 0)
+                       return -EINVAL;
+               metadata_data.shm_fd = len;
+               /* get wait fd */
+               len = lttcomm_recv_fd(sock);
+               if (len < 0)
+                       return -EINVAL;
+               metadata_data.wait_fd = len;
+       }
+
+       ret = open_streams(sock, metadata_data.handle,
+                       &metadata_stream_data, 1);
+       if (ret) {
+               printf("Error in open_streams\n");
+               return ret;
+       }
+
+       /* Create data channel */
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = session_handle;
+       lum.cmd = LTTNG_UST_CHANNEL;
+       //lum.u.channel.overwrite = 0;
+       lum.u.channel.overwrite = 1;
+       lum.u.channel.subbuf_size = 32768;
+       lum.u.channel.num_subbuf = 8;
+       //lum.u.channel.num_subbuf = 4;
+       //lum.u.channel.num_subbuf = 2;
+       lum.u.channel.switch_timer_interval = 0;
+       lum.u.channel.read_timer_interval = 0;
+       lum.u.channel.output = LTTNG_UST_MMAP;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       channel_data.handle = lur.ret_val;
+       printf("received channel handle %u\n", channel_data.handle);
+       if (lur.ret_code == LTTCOMM_OK) {
+               ssize_t len;
+
+               channel_data.memory_map_size = lur.u.channel.memory_map_size;
+               /* get shm fd */
+               len = lttcomm_recv_fd(sock);
+               if (len < 0)
+                       return -EINVAL;
+               channel_data.shm_fd = len;
+               /* get wait fd */
+               len = lttcomm_recv_fd(sock);
+               if (len < 0)
+                       return -EINVAL;
+               channel_data.wait_fd = len;
+       }
+
+       /* Create events */
+       for (k = 0; k < nr_events; k++) {
+               memset(&lum, 0, sizeof(lum));
+               lum.handle = channel_data.handle;
+               lum.cmd = LTTNG_UST_EVENT;
+               strncpy(lum.u.event.name, event_names[k],
+                       LTTNG_UST_SYM_NAME_LEN);
+               lum.u.event.instrumentation = LTTNG_UST_TRACEPOINT;
+               ret = send_app_cmd(sock, &lum, &lur);
+               if (ret)
+                       return ret;
+               event_handle[k] = lur.ret_val;
+               printf("received event handle %u\n", event_handle[k]);
+       }
+
+       /* Get references to channel streams */
+       ret = open_streams(sock, channel_data.handle,
+                       stream_data, MAX_NR_STREAMS);
+       if (ret) {
+               printf("Error in open_streams\n");
+               return ret;
+       }
+
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = session_handle;
+       lum.cmd = LTTNG_UST_SESSION_START;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       printf("Session handle %u started.\n", session_handle);
+
+       /* Tell application registration is done */
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = LTTNG_UST_ROOT_HANDLE;
+       lum.cmd = LTTNG_UST_REGISTER_DONE;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       printf("Registration done acknowledged.\n");
+
+       sleep(4);
+
+       ret = consume_buffers(outputpath);
+       if (ret) {
+               printf("Error in consume_buffers\n");
+               return ret;
+       }
+
+       /* Release data channel */
+       /* Release streams */
+       ret = close_streams(sock, stream_data,
+                       MAX_NR_STREAMS);
+       if (ret)
+               return ret;
+
+       /* Release events */
+       for (k = 0; k < nr_events; k++) {
+               memset(&lum, 0, sizeof(lum));
+               lum.handle = event_handle[k];
+               lum.cmd = LTTNG_UST_RELEASE;
+               ret = send_app_cmd(sock, &lum, &lur);
+               if (ret)
+                       return ret;
+       }
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = channel_data.handle;
+       lum.cmd = LTTNG_UST_RELEASE;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       if (channel_data.shm_fd >= 0) {
+               ret = close(channel_data.shm_fd);
+               if (ret)
+                       return ret;
+       }
+       if (channel_data.wait_fd >= 0) {
+               ret = close(channel_data.wait_fd);
+               if (ret)
+                       return ret;
+       }
+
+       /* Release metadata channel */
+       ret = close_streams(sock, &metadata_stream_data, 1);
+       if (ret)
+               return ret;
+
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = metadata_data.handle;
+       lum.cmd = LTTNG_UST_RELEASE;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+       if (metadata_data.shm_fd >= 0) {
+               ret = close(metadata_data.shm_fd);
+               if (ret)
+                       return ret;
+       }
+       if (metadata_data.wait_fd >= 0) {
+               ret = close(metadata_data.wait_fd);
+               if (ret)
+                       return ret;
+       }
+
+       /* Release session */
+       memset(&lum, 0, sizeof(lum));
+       lum.handle = session_handle;
+       lum.cmd = LTTNG_UST_RELEASE;
+       ret = send_app_cmd(sock, &lum, &lur);
+       if (ret)
+               return ret;
+
+       return 0;
+}
+
+/*
+ * 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;
+
+       /* 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);
+
+       /*
+        * 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);
+       }
+
+       printf("Got the wait shm fd %d\n", wait_shm_fd);
+
+       return wait_shm_fd;
+
+error:
+       printf("Failing to get the wait shm fd\n");
+
+       return -1;
+}
+
+int update_futex(int fd, int active)
+{
+       size_t mmap_size = sysconf(_SC_PAGE_SIZE);
+       char *wait_shm_mmap;
+       int ret;
+
+       wait_shm_mmap = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE,
+                 MAP_SHARED, fd, 0);
+       if (wait_shm_mmap == MAP_FAILED) {
+               perror("mmap");
+               goto error;
+       }
+
+       if (active) {
+               uatomic_set((int32_t *) wait_shm_mmap, 1);
+               futex_async((int32_t *) wait_shm_mmap, FUTEX_WAKE,
+                               INT_MAX, NULL, NULL, 0);
+       } else {
+               uatomic_set((int32_t *) wait_shm_mmap, 0);
+       }
+       ret = munmap(wait_shm_mmap, mmap_size);
+       if (ret) {
+               perror("Error unmapping wait shm");
+               goto error;
+       }
+       return 0;
+error:
+       return -1;
+}
+
+/*
+ * 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;
+
+       /*
+        * If not root, we cannot increase our max open files. But our
+        * scope is then limited to processes from a single user.
+        */
+       if (getuid() != 0)
+               return;
+       /* 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");
+       }
+}
+
+/*
+ * Usage:
+ * ./ust-basic-tracing outputpath event_name1 event_name2 ....
+ */
+int main(int argc, const char **argv)
+{
+       const char *home_dir;
+       int ret, wait_shm_fd;
+       struct sigaction act;
+       mode_t old_umask;
+       const char *outputpath;
+       const char **event_names;
+       unsigned int nr_events;
+
+       if (argc < 2) {
+               printf("Usage:\n");
+               printf("%s outputpath event_name1 event_name2 ...\n",
+                       argv[0]);
+               exit(-1);
+       }
+       outputpath = argv[1];
+       event_names = &argv[2];
+       nr_events = argc - 2;
+
+       set_ulimit();
+
+       /* Ignore sigpipe */
+       memset(&act, 0, sizeof(act));
+       ret = sigemptyset(&act.sa_mask);
+       if (ret == -1) {
+               perror("sigemptyset");
+               return -1;
+       }
+
+       act.sa_handler = SIG_IGN;
+       ret = sigaction(SIGPIPE, &act, NULL);
+       if (ret == -1) {
+               perror("sigaction");
+               return -1;
+       }
+
+       /* Handle SIGTERM */
+       act.sa_handler = handle_signals;
+       ret = sigaction(SIGTERM, &act, NULL);
+       if (ret == -1) {
+               perror("sigaction");
+               return -1;
+       }
+       /* Handle SIGINT */
+       ret = sigaction(SIGINT, &act, NULL);
+       if (ret == -1) {
+               perror("sigaction");
+               return -1;
+       }
+
+       if (geteuid() == 0) {
+               ret = mkdir(LTTNG_RUNDIR, S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
+               if (ret && errno != EEXIST) {
+                       perror("mkdir");
+                       return -1;
+               }
+               wait_shm_fd = get_wait_shm(DEFAULT_GLOBAL_APPS_WAIT_SHM_PATH,
+                                       sysconf(_SC_PAGE_SIZE), 1);
+               if (wait_shm_fd < 0) {
+                       perror("global wait shm error");
+                       return -1;
+               }
+               strcpy(apps_sock_path, DEFAULT_GLOBAL_APPS_UNIX_SOCK);
+               old_umask = umask(0);
+       } else {
+               snprintf(local_apps_wait_shm_path, PATH_MAX,
+                        DEFAULT_HOME_APPS_WAIT_SHM_PATH, getuid());
+               wait_shm_fd = get_wait_shm(local_apps_wait_shm_path,
+                                       sysconf(_SC_PAGE_SIZE), 0);
+               if (wait_shm_fd < 0) {
+                       perror("local wait shm error");
+                       return -1;
+               }
+               home_dir = (const char *) getenv("HOME");
+               if (!home_dir) {
+                       perror("getenv error");
+                       return -ENOENT;
+               }
+               snprintf(apps_sock_path, PATH_MAX,
+                        DEFAULT_HOME_APPS_UNIX_SOCK, home_dir);
+       }
+
+       ret = lttcomm_create_unix_sock(apps_sock_path);
+       if (ret < 0) {
+               perror("create error");
+               return ret;
+       }
+       apps_socket = ret;
+
+       if (getuid() == 0) {
+               /* File permission MUST be 666 */
+               ret = chmod(apps_sock_path,
+                               S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH);
+               if (ret < 0) {
+                       printf("Set file permissions failed: %s\n", apps_sock_path);
+                       perror("chmod");
+                       goto end;
+               }
+               umask(old_umask);
+       }
+       ret = lttcomm_listen_unix_sock(apps_socket);
+       if (ret < 0) {
+               perror("listen error");
+               return ret;
+       }
+
+       /* wake up futexes */
+       ret = update_futex(wait_shm_fd, 1);
+       if (ret) {
+               fprintf(stderr, "Error wakeup futex\n");
+               return -1;
+       }
+
+       for (;;) {
+               int sock;
+               ssize_t len;
+               struct {
+                       uint32_t major;
+                       uint32_t minor;
+                       pid_t pid;
+                       pid_t ppid;
+                       uid_t uid;
+                       gid_t gid;
+                       char name[16];  /* Process name */
+               } reg_msg;
+               char bufname[17];
+
+               if (quit_program)
+                       break;
+
+               printf("Accepting application registration\n");
+               sock = lttcomm_accept_unix_sock(apps_socket);
+               if (sock < 0) {
+                       perror("accept error");
+                       goto end;
+               }
+
+               /*
+                * Basic recv here to handle the very simple data
+                * that the libust send to register (reg_msg).
+                */
+               len = lttcomm_recv_unix_sock(sock, &reg_msg, sizeof(reg_msg));
+               if (len < 0 || len != sizeof(reg_msg)) {
+                       perror("lttcomm_recv_unix_sock");
+                       continue;
+               }
+               memcpy(bufname, reg_msg.name, 16);
+               bufname[16] = '\0';
+               printf("Application %s pid %u ppid %u uid %u gid %u has registered (version : %u.%u)\n",
+                       bufname, reg_msg.pid, reg_msg.ppid, reg_msg.uid,
+                       reg_msg.gid, reg_msg.major, reg_msg.minor);
+               ret = send_app_msgs(sock, outputpath, nr_events, event_names);
+               if (ret) {
+                       printf("Error in send_app_msgs.\n");
+                       sleep(1);
+               }
+               close(sock);
+       }
+
+end:
+       printf("quitting.\n");
+       /* Let applications know we are not responding anymore */
+       ret = update_futex(wait_shm_fd, 0);
+       if (ret) {
+               fprintf(stderr, "Error wakeup futex\n");
+               return -1;
+       }
+
+       return 0;
+}
This page took 0.033966 seconds and 4 git commands to generate.