Move the code from lttd to liblttd and adapt everything so it works
[ltt-control.git] / liblttd / liblttd.c
diff --git a/liblttd/liblttd.c b/liblttd/liblttd.c
new file mode 100644 (file)
index 0000000..0f0e5b5
--- /dev/null
@@ -0,0 +1,747 @@
+/* lttd
+ *
+ * Linux Trace Toolkit Daemon
+ *
+ * This is a simple daemon that reads a few relay+debugfs channels and save
+ * them in a trace.
+ *
+ * CPU hot-plugging is supported using inotify.
+ *
+ * Copyright 2005 -
+ *     Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "liblttd.h"
+
+#define _REENTRANT
+#define _GNU_SOURCE
+#include <features.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <errno.h>
+#include <sys/types.h>
+#include <stdlib.h>
+#include <dirent.h>
+#include <string.h>
+#include <fcntl.h>
+#include <sys/stat.h>
+#include <sys/poll.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+#include <unistd.h>
+#include <asm/ioctls.h>
+
+#include <linux/version.h>
+
+/* Relayfs IOCTL */
+#include <asm/ioctl.h>
+#include <asm/types.h>
+
+/* Get the next sub buffer that can be read. */
+#define RELAY_GET_SB           _IOR(0xF5, 0x00,__u32)
+/* Release the oldest reserved (by "get") sub buffer. */
+#define RELAY_PUT_SB           _IOW(0xF5, 0x01,__u32)
+/* returns the number of sub buffers in the per cpu channel. */
+#define RELAY_GET_N_SB         _IOR(0xF5, 0x02,__u32)
+/* returns the size of the current sub buffer. */
+#define RELAY_GET_SB_SIZE      _IOR(0xF5, 0x03, __u32)
+/* returns the size of data to consume in the current sub-buffer. */
+#define RELAY_GET_MAX_SB_SIZE  _IOR(0xF5, 0x04, __u32)
+
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
+#include <sys/inotify.h>
+
+#define HAS_INOTIFY
+#else
+static inline int inotify_init (void)
+{
+       return -1;
+}
+
+static inline int inotify_add_watch (int fd, const char *name, __u32 mask)
+{
+       return 0;
+}
+
+static inline int inotify_rm_watch (int fd, __u32 wd)
+{
+       return 0;
+}
+#undef HAS_INOTIFY
+#endif
+
+struct liblttd_callbacks *callbacks;
+
+struct channel_trace_fd {
+       struct fd_pair *pair;
+       int num_pairs;
+};
+
+struct inotify_watch {
+       int wd;
+       char path_channel[PATH_MAX];
+       char *base_path_channel;
+};
+
+struct inotify_watch_array {
+       struct inotify_watch *elem;
+       int num;
+};
+
+struct channel_trace_fd fd_pairs = { NULL, 0 };
+int inotify_fd = -1;
+struct inotify_watch_array inotify_watch_array = { NULL, 0 };
+
+/* protects fd_pairs and inotify_watch_array */
+pthread_rwlock_t fd_pairs_lock = PTHREAD_RWLOCK_INITIALIZER;
+
+static char            *channel_name = NULL;
+static unsigned long   num_threads = 1;
+volatile static int    quit_program = 0;       /* For signal handler */
+static int             dump_flight_only = 0;
+static int             dump_normal_only = 0;
+static int             verbose_mode = 0;
+
+#define printf_verbose(fmt, args...) \
+  do {                               \
+    if (verbose_mode)                \
+      printf(fmt, ##args);           \
+  } while (0)
+
+
+int open_buffer_file(char *filename, char *path_channel,
+       char *base_path_channel, struct channel_trace_fd *fd_pairs)
+{
+       int open_ret = 0;
+       int ret = 0;
+
+       if(strncmp(filename, "flight-", sizeof("flight-")-1) != 0) {
+               if(dump_flight_only) {
+                       printf_verbose("Skipping normal channel %s\n",
+                               path_channel);
+                       return 0;
+               }
+       } else {
+               if(dump_normal_only) {
+                       printf_verbose("Skipping flight channel %s\n",
+                               path_channel);
+                       return 0;
+               }
+       }
+       printf_verbose("Opening file.\n");
+
+       fd_pairs->pair = realloc(fd_pairs->pair,
+                       ++fd_pairs->num_pairs * sizeof(struct fd_pair));
+
+       /* Open the channel in read mode */
+       fd_pairs->pair[fd_pairs->num_pairs-1].channel =
+               open(path_channel, O_RDONLY | O_NONBLOCK);
+       if(fd_pairs->pair[fd_pairs->num_pairs-1].channel == -1) {
+               perror(path_channel);
+               fd_pairs->num_pairs--;
+               return 0;       /* continue */
+       }
+
+       if(callbacks->on_open_channel) ret = callbacks->on_open_channel(
+                       callbacks, &fd_pairs->pair[fd_pairs->num_pairs-1],
+                       base_path_channel);
+
+       if(ret != 0) {
+               open_ret = -1;
+               close(fd_pairs->pair[fd_pairs->num_pairs-1].channel);
+               fd_pairs->num_pairs--;
+               goto end;
+       }
+
+end:
+       return open_ret;
+}
+
+int open_channel_trace_pairs(char *subchannel_name,
+               char *base_subchannel_name,
+               struct channel_trace_fd *fd_pairs, int *inotify_fd,
+               struct inotify_watch_array *iwatch_array)
+{
+       DIR *channel_dir = opendir(subchannel_name);
+       struct dirent *entry;
+       struct stat stat_buf;
+       int ret;
+       char path_channel[PATH_MAX];
+       int path_channel_len;
+       char *path_channel_ptr;
+       char *base_subchannel_ptr;
+
+       int open_ret = 0;
+
+       if(channel_dir == NULL) {
+               perror(subchannel_name);
+               open_ret = ENOENT;
+               goto end;
+       }
+
+       printf_verbose("Calling on new channels folder");
+       if(callbacks->on_new_channels_folder) ret = callbacks->
+                       on_new_channels_folder(callbacks,
+                       base_subchannel_name);
+       if(ret == -1) {
+               open_ret = -1;
+               goto end;
+       }
+
+       strncpy(path_channel, subchannel_name, PATH_MAX-1);
+       path_channel_len = strlen(path_channel);
+       path_channel[path_channel_len] = '/';
+       path_channel_len++;
+       path_channel_ptr = path_channel + path_channel_len;
+       base_subchannel_ptr = path_channel +
+               (base_subchannel_name - subchannel_name);
+
+#ifdef HAS_INOTIFY
+       iwatch_array->elem = realloc(iwatch_array->elem,
+               ++iwatch_array->num * sizeof(struct inotify_watch));
+
+       printf_verbose("Adding inotify for channel %s\n", path_channel);
+       iwatch_array->elem[iwatch_array->num-1].wd = inotify_add_watch(*inotify_fd, path_channel, IN_CREATE);
+       strcpy(iwatch_array->elem[iwatch_array->num-1].path_channel, path_channel);
+       iwatch_array->elem[iwatch_array->num-1].base_path_channel =
+               iwatch_array->elem[iwatch_array->num-1].path_channel +
+               (base_subchannel_name - subchannel_name);
+       printf_verbose("Added inotify for channel %s, wd %u\n",
+               iwatch_array->elem[iwatch_array->num-1].path_channel,
+               iwatch_array->elem[iwatch_array->num-1].wd);
+#endif
+
+       while((entry = readdir(channel_dir)) != NULL) {
+
+               if(entry->d_name[0] == '.') continue;
+
+               strncpy(path_channel_ptr, entry->d_name, PATH_MAX - path_channel_len);
+
+               ret = stat(path_channel, &stat_buf);
+               if(ret == -1) {
+                       perror(path_channel);
+                       continue;
+               }
+
+               printf_verbose("Channel file : %s\n", path_channel);
+
+               if(S_ISDIR(stat_buf.st_mode)) {
+
+                       printf_verbose("Entering channel subdirectory...\n");
+                       ret = open_channel_trace_pairs(path_channel, base_subchannel_ptr, fd_pairs,
+                               inotify_fd, iwatch_array);
+                       if(ret < 0) continue;
+               } else if(S_ISREG(stat_buf.st_mode)) {
+                       open_ret = open_buffer_file(entry->d_name, path_channel, base_subchannel_ptr,
+                               fd_pairs);
+                       if(open_ret)
+                               goto end;
+               }
+       }
+
+end:
+       closedir(channel_dir);
+
+       return open_ret;
+}
+
+
+int read_subbuffer(struct fd_pair *pair)
+{
+       unsigned int consumed_old, len;
+       int err;
+       long ret;
+       off_t offset;
+
+
+       err = ioctl(pair->channel, RELAY_GET_SB, &consumed_old);
+       printf_verbose("cookie : %u\n", consumed_old);
+       if(err != 0) {
+               ret = errno;
+               perror("Reserving sub buffer failed (everything is normal, it is due to concurrency)");
+               goto get_error;
+       }
+
+       err = ioctl(pair->channel, RELAY_GET_SB_SIZE, &len);
+       if(err != 0) {
+               ret = errno;
+               perror("Getting sub-buffer len failed.");
+               goto get_error;
+       }
+
+       if(callbacks->on_read_subbuffer) ret = callbacks->on_read_subbuffer(
+               callbacks, pair, len);
+
+write_error:
+       ret = 0;
+       err = ioctl(pair->channel, RELAY_PUT_SB, &consumed_old);
+       if(err != 0) {
+               ret = errno;
+               if(errno == EFAULT) {
+                       perror("Error in unreserving sub buffer\n");
+               } else if(errno == EIO) {
+                       /* Should never happen with newer LTTng versions */
+                       perror("Reader has been pushed by the writer, last sub-buffer corrupted.");
+               }
+               goto get_error;
+       }
+
+get_error:
+       return ret;
+}
+
+
+int map_channels(struct channel_trace_fd *fd_pairs,
+       int idx_begin, int idx_end)
+{
+       int i,j;
+       int ret=0;
+
+       if(fd_pairs->num_pairs <= 0) {
+               printf("No channel to read\n");
+               goto end;
+       }
+
+       /* Get the subbuf sizes and number */
+
+       for(i=idx_begin;i<idx_end;i++) {
+               struct fd_pair *pair = &fd_pairs->pair[i];
+
+               ret = ioctl(pair->channel, RELAY_GET_N_SB, &pair->n_sb);
+               if(ret != 0) {
+                       perror("Error in getting the number of sub-buffers");
+                       goto end;
+               }
+               ret = ioctl(pair->channel, RELAY_GET_MAX_SB_SIZE,
+                           &pair->max_sb_size);
+               if(ret != 0) {
+                       perror("Error in getting the max sub-buffer size");
+                       goto end;
+               }
+               ret = pthread_mutex_init(&pair->mutex, NULL);   /* Fast mutex */
+               if(ret != 0) {
+                       perror("Error in mutex init");
+                       goto end;
+               }
+       }
+
+end:
+       return ret;
+}
+
+int unmap_channels(struct channel_trace_fd *fd_pairs)
+{
+       int j;
+       int ret=0;
+
+       /* Munmap each FD */
+       for(j=0;j<fd_pairs->num_pairs;j++) {
+               struct fd_pair *pair = &fd_pairs->pair[j];
+               int err_ret;
+
+               err_ret = pthread_mutex_destroy(&pair->mutex);
+               if(err_ret != 0) {
+                       perror("Error in mutex destroy");
+               }
+               ret |= err_ret;
+       }
+
+       return ret;
+}
+
+#ifdef HAS_INOTIFY
+/* Inotify event arrived.
+ *
+ * Only support add file for now.
+ */
+
+int read_inotify(int inotify_fd,
+       struct channel_trace_fd *fd_pairs,
+       struct inotify_watch_array *iwatch_array)
+{
+       char buf[sizeof(struct inotify_event) + PATH_MAX];
+       char path_channel[PATH_MAX];
+       ssize_t len;
+       struct inotify_event *ievent;
+       size_t offset;
+       unsigned int i;
+       int ret;
+       int old_num;
+
+       offset = 0;
+       len = read(inotify_fd, buf, sizeof(struct inotify_event) + PATH_MAX);
+       if(len < 0) {
+
+               if(errno == EAGAIN)
+                       return 0;  /* another thread got the data before us */
+
+               printf("Error in read from inotify FD %s.\n", strerror(len));
+               return -1;
+       }
+       while(offset < len) {
+               ievent = (struct inotify_event *)&(buf[offset]);
+               for(i=0; i<iwatch_array->num; i++) {
+                       if(iwatch_array->elem[i].wd == ievent->wd &&
+                               ievent->mask == IN_CREATE) {
+                               printf_verbose(
+                                       "inotify wd %u event mask : %u for %s%s\n",
+                                       ievent->wd, ievent->mask,
+                                       iwatch_array->elem[i].path_channel,
+                                       ievent->name);
+                               old_num = fd_pairs->num_pairs;
+                               strcpy(path_channel, iwatch_array->elem[i].path_channel);
+                               strcat(path_channel, ievent->name);
+                               if(ret = open_buffer_file(ievent->name, path_channel,
+                                       path_channel + (iwatch_array->elem[i].base_path_channel -
+                                       iwatch_array->elem[i].path_channel), fd_pairs)) {
+                                       printf("Error opening buffer file\n");
+                                       return -1;
+                               }
+                               if(ret = map_channels(fd_pairs, old_num, fd_pairs->num_pairs)) {
+                                       printf("Error mapping channel\n");
+                                       return -1;
+                               }
+
+                       }
+               }
+               offset += sizeof(*ievent) + ievent->len;
+       }
+}
+#endif //HAS_INOTIFY
+
+/* read_channels
+ *
+ * Thread worker.
+ *
+ * Read the debugfs channels and write them in the paired tracefiles.
+ *
+ * @fd_pairs : paired channels and trace files.
+ *
+ * returns 0 on success, -1 on error.
+ *
+ * Note that the high priority polled channels are consumed first. We then poll
+ * again to see if these channels are still in priority. Only when no
+ * high priority channel is left, we start reading low priority channels.
+ *
+ * Note that a channel is considered high priority when the buffer is almost
+ * full.
+ */
+
+int read_channels(unsigned long thread_num, struct channel_trace_fd *fd_pairs,
+       int inotify_fd, struct inotify_watch_array *iwatch_array)
+{
+       struct pollfd *pollfd = NULL;
+       int num_pollfd;
+       int i,j;
+       int num_rdy, num_hup;
+       int high_prio;
+       int ret = 0;
+       int inotify_fds;
+       unsigned int old_num;
+
+#ifdef HAS_INOTIFY
+       inotify_fds = 1;
+#else
+       inotify_fds = 0;
+#endif
+
+       pthread_rwlock_rdlock(&fd_pairs_lock);
+
+       /* Start polling the FD. Keep one fd for inotify */
+       pollfd = malloc((inotify_fds + fd_pairs->num_pairs) * sizeof(struct pollfd));
+
+#ifdef HAS_INOTIFY
+       pollfd[0].fd = inotify_fd;
+       pollfd[0].events = POLLIN|POLLPRI;
+#endif
+
+       for(i=0;i<fd_pairs->num_pairs;i++) {
+               pollfd[inotify_fds+i].fd = fd_pairs->pair[i].channel;
+               pollfd[inotify_fds+i].events = POLLIN|POLLPRI;
+       }
+       num_pollfd = inotify_fds + fd_pairs->num_pairs;
+
+
+       pthread_rwlock_unlock(&fd_pairs_lock);
+
+       while(1) {
+               high_prio = 0;
+               num_hup = 0;
+#ifdef DEBUG
+               printf("Press a key for next poll...\n");
+               char buf[1];
+               read(STDIN_FILENO, &buf, 1);
+               printf("Next poll (polling %d fd) :\n", num_pollfd);
+#endif //DEBUG
+
+               /* Have we received a signal ? */
+               if(quit_program) break;
+
+               num_rdy = poll(pollfd, num_pollfd, -1);
+
+               if(num_rdy == -1) {
+                       perror("Poll error");
+                       goto free_fd;
+               }
+
+               printf_verbose("Data received\n");
+#ifdef HAS_INOTIFY
+               switch(pollfd[0].revents) {
+                       case POLLERR:
+                               printf_verbose(
+                                       "Error returned in polling inotify fd %d.\n",
+                                       pollfd[0].fd);
+                               break;
+                       case POLLHUP:
+                               printf_verbose(
+                                       "Polling inotify fd %d tells it has hung up.\n",
+                                       pollfd[0].fd);
+                               break;
+                       case POLLNVAL:
+                               printf_verbose(
+                                       "Polling inotify fd %d tells fd is not open.\n",
+                                       pollfd[0].fd);
+                               break;
+                       case POLLPRI:
+                       case POLLIN:
+                               printf_verbose(
+                                       "Polling inotify fd %d : data ready.\n",
+                                       pollfd[0].fd);
+
+                               pthread_rwlock_wrlock(&fd_pairs_lock);
+                               read_inotify(inotify_fd, fd_pairs, iwatch_array);
+                               pthread_rwlock_unlock(&fd_pairs_lock);
+
+                       break;
+               }
+#endif
+
+               for(i=inotify_fds;i<num_pollfd;i++) {
+                       switch(pollfd[i].revents) {
+                               case POLLERR:
+                                       printf_verbose(
+                                               "Error returned in polling fd %d.\n",
+                                               pollfd[i].fd);
+                                       num_hup++;
+                                       break;
+                               case POLLHUP:
+                                       printf_verbose(
+                                               "Polling fd %d tells it has hung up.\n",
+                                               pollfd[i].fd);
+                                       num_hup++;
+                                       break;
+                               case POLLNVAL:
+                                       printf_verbose(
+                                               "Polling fd %d tells fd is not open.\n",
+                                               pollfd[i].fd);
+                                       num_hup++;
+                                       break;
+                               case POLLPRI:
+                                       pthread_rwlock_rdlock(&fd_pairs_lock);
+                                       if(pthread_mutex_trylock(&fd_pairs->pair[i-inotify_fds].mutex) == 0) {
+                                               printf_verbose(
+                                                       "Urgent read on fd %d\n",
+                                                       pollfd[i].fd);
+                                               /* Take care of high priority channels first. */
+                                               high_prio = 1;
+                                               /* it's ok to have an unavailable sub-buffer */
+                                               ret = read_subbuffer(&fd_pairs->pair[i-inotify_fds]);
+                                               if(ret == EAGAIN) ret = 0;
+
+                                               ret = pthread_mutex_unlock(&fd_pairs->pair[i-inotify_fds].mutex);
+                                               if(ret)
+                                                       printf("Error in mutex unlock : %s\n", strerror(ret));
+                                       }
+                                       pthread_rwlock_unlock(&fd_pairs_lock);
+                                       break;
+                       }
+               }
+               /* If every buffer FD has hung up, we end the read loop here */
+               if(num_hup == num_pollfd - inotify_fds) break;
+
+               if(!high_prio) {
+                       for(i=inotify_fds;i<num_pollfd;i++) {
+                               switch(pollfd[i].revents) {
+                                       case POLLIN:
+                                               pthread_rwlock_rdlock(&fd_pairs_lock);
+                                               if(pthread_mutex_trylock(&fd_pairs->pair[i-inotify_fds].mutex) == 0) {
+                                                       /* Take care of low priority channels. */
+                                                       printf_verbose(
+                                                               "Normal read on fd %d\n",
+                                                               pollfd[i].fd);
+                                                       /* it's ok to have an unavailable subbuffer */
+                                                       ret = read_subbuffer(&fd_pairs->pair[i-inotify_fds]);
+                                                       if(ret == EAGAIN) ret = 0;
+
+                                                       ret = pthread_mutex_unlock(&fd_pairs->pair[i-inotify_fds].mutex);
+                                                       if(ret)
+                                                               printf("Error in mutex unlock : %s\n", strerror(ret));
+                                               }
+                                               pthread_rwlock_unlock(&fd_pairs_lock);
+                                               break;
+                               }
+                       }
+               }
+
+               /* Update pollfd array if an entry was added to fd_pairs */
+               pthread_rwlock_rdlock(&fd_pairs_lock);
+               if((inotify_fds + fd_pairs->num_pairs) != num_pollfd) {
+                       pollfd = realloc(pollfd,
+                                       (inotify_fds + fd_pairs->num_pairs) * sizeof(struct pollfd));
+                       for(i=num_pollfd-inotify_fds;i<fd_pairs->num_pairs;i++) {
+                               pollfd[inotify_fds+i].fd = fd_pairs->pair[i].channel;
+                               pollfd[inotify_fds+i].events = POLLIN|POLLPRI;
+                       }
+                       num_pollfd = fd_pairs->num_pairs + inotify_fds;
+               }
+               pthread_rwlock_unlock(&fd_pairs_lock);
+
+               /* NB: If the fd_pairs structure is updated by another thread from this
+                *     point forward, the current thread will wait in the poll without
+                *     monitoring the new channel. However, this thread will add the
+                *     new channel on next poll (and this should not take too much time
+                *     on a loaded system).
+                *
+                *     This event is quite unlikely and can only occur if a CPU is
+                *     hot-plugged while multple lttd threads are running.
+                */
+       }
+
+free_fd:
+       free(pollfd);
+
+end:
+       return ret;
+}
+
+
+void close_channel_trace_pairs(struct channel_trace_fd *fd_pairs, int inotify_fd,
+       struct inotify_watch_array *iwatch_array)
+{
+       int i;
+       int ret;
+
+       for(i=0;i<fd_pairs->num_pairs;i++) {
+               ret = close(fd_pairs->pair[i].channel);
+               if(ret == -1) perror("Close error on channel");
+               if(callbacks->on_close_channel) {
+                       ret = callbacks->on_close_channel(
+                               callbacks, &fd_pairs->pair[i]);
+                       if(ret != 0) perror("Error on close channel callback");
+               }
+       }
+       free(fd_pairs->pair);
+       free(iwatch_array->elem);
+}
+
+/* Thread worker */
+void * thread_main(void *arg)
+{
+       long ret = 0;
+       unsigned long thread_num = (unsigned long)arg;
+
+       if(callbacks->on_new_thread)
+               ret = callbacks->on_new_thread(callbacks, thread_num);
+
+       if (ret < 0) {
+               return (void*)ret;
+       }
+       ret = read_channels(thread_num, &fd_pairs, inotify_fd, &inotify_watch_array);
+
+       if(callbacks->on_close_thread)
+               callbacks->on_close_thread(callbacks, thread_num);
+
+       return (void*)ret;
+}
+
+/*on_close_thread has to be reentrant, it'll be called by many threads*/
+int(*on_close_thread)(struct liblttd_callbacks *data, unsigned long thread_num);
+
+int channels_init()
+{
+       int ret = 0;
+
+       inotify_fd = inotify_init();
+       fcntl(inotify_fd, F_SETFL, O_NONBLOCK);
+
+       if(ret = open_channel_trace_pairs(channel_name,
+                       channel_name + strlen(channel_name), &fd_pairs,
+                       &inotify_fd, &inotify_watch_array))
+               goto close_channel;
+       if (fd_pairs.num_pairs == 0) {
+               printf("No channel available for reading, exiting\n");
+               ret = -ENOENT;
+               goto close_channel;
+       }
+       if(ret = map_channels(&fd_pairs, 0, fd_pairs.num_pairs))
+               goto close_channel;
+       return 0;
+
+close_channel:
+       close_channel_trace_pairs(&fd_pairs, inotify_fd, &inotify_watch_array);
+       if(inotify_fd >= 0)
+               close(inotify_fd);
+       return ret;
+}
+
+int liblttd_start(char *channel_path, unsigned long n_threads,
+               int flight_only, int normal_only, int verbose,
+               struct liblttd_callbacks *user_data){
+       int ret = 0;
+       pthread_t *tids;
+       unsigned long i;
+       void *tret;
+
+       channel_name = channel_path;
+       num_threads = n_threads;
+       dump_flight_only = flight_only;
+       dump_normal_only = normal_only;
+       verbose_mode = verbose;
+       callbacks = user_data;
+
+       if(ret = channels_init())
+               return ret;
+
+       tids = malloc(sizeof(pthread_t) * num_threads);
+       for(i=0; i<num_threads; i++) {
+
+               ret = pthread_create(&tids[i], NULL, thread_main, (void*)i);
+               if(ret) {
+                       perror("Error creating thread");
+                       break;
+               }
+       }
+
+       for(i=0; i<num_threads; i++) {
+               ret = pthread_join(tids[i], &tret);
+               if(ret) {
+                       perror("Error joining thread");
+                       break;
+               }
+               if((long)tret != 0) {
+                       printf("Error %s occured in thread %ld\n",
+                               strerror((long)tret), i);
+               }
+       }
+
+       free(tids);
+       ret = unmap_channels(&fd_pairs);
+       close_channel_trace_pairs(&fd_pairs, inotify_fd, &inotify_watch_array);
+       if(inotify_fd >= 0)
+               close(inotify_fd);
+
+       if(callbacks->on_trace_end) callbacks->on_trace_end(callbacks);
+
+       return ret;
+}
+
+int liblttd_stop() {
+       quit_program = 1;
+       return 0;
+}
+
This page took 0.031026 seconds and 4 git commands to generate.