X-Git-Url: https://git.octo.it/?p=rrdtool.git;a=blobdiff_plain;f=src%2Frrd_daemon.c;h=dcca55d755f53d52513d3e1ba1f9c3a08e5d76bd;hp=5c47ac1fd7b3908edce814a15631c1768e571442;hb=18ff159c38b7d919bafbdcc46d03c3f83c17476f;hpb=001e472141c96137de1b5267f8de7fb91e204ab7

diff --git a/src/rrd_daemon.c b/src/rrd_daemon.c
index 5c47ac1..dcca55d 100644
--- a/src/rrd_daemon.c
+++ b/src/rrd_daemon.c
@@ -62,23 +62,37 @@
 /*
  * Now for some includes..
  */
-#include "rrd.h" /* {{{ */
+/* {{{ */
+#if defined(_WIN32) && !defined(__CYGWIN__) && !defined(__CYGWIN32__) && !defined(HAVE_CONFIG_H)
+#include "../win32/config.h"
+#else
+#ifdef HAVE_CONFIG_H
+#include "../rrd_config.h"
+#endif
+#endif
+
+#include "rrd.h"
 #include "rrd_client.h"
 
 #include <stdlib.h>
+
+#ifndef WIN32
 #include <stdint.h>
-#include <stdio.h>
 #include <unistd.h>
-#include <string.h>
 #include <strings.h>
-#include <stdint.h>
 #include <inttypes.h>
+#	include <sys/socket.h>
+
+#else
+
+#endif
+#include <stdio.h>
+#include <string.h>
 
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 #include <signal.h>
-#include <sys/socket.h>
 #include <sys/un.h>
 #include <netdb.h>
 #include <poll.h>
@@ -130,6 +144,31 @@ struct listen_socket_s
 };
 typedef struct listen_socket_s listen_socket_t;
 
+struct command;
+/* note: guard against "unused" warnings in the handlers */
+#define DISPATCH_PROTO	listen_socket_t *sock	__attribute__((unused)),\
+			time_t now		__attribute__((unused)),\
+			char  *buffer		__attribute__((unused)),\
+			size_t buffer_size	__attribute__((unused))
+
+#define HANDLER_PROTO	struct command *cmd	__attribute__((unused)),\
+			DISPATCH_PROTO
+
+struct command {
+  char   *cmd;
+  int (*handler)(HANDLER_PROTO);
+  socket_privilege min_priv;
+
+  char  context;		/* where we expect to see it */
+#define CMD_CONTEXT_CLIENT	(1<<0)
+#define CMD_CONTEXT_BATCH	(1<<1)
+#define CMD_CONTEXT_JOURNAL	(1<<2)
+#define CMD_CONTEXT_ANY		(0x7f)
+
+  char *syntax;
+  char *help;
+};
+
 struct cache_item_s;
 typedef struct cache_item_s cache_item_t;
 struct cache_item_s
@@ -178,10 +217,15 @@ static size_t listen_fds_num = 0;
 
 static int do_shutdown = 0;
 
-static pthread_t queue_thread;
+static pthread_t *queue_threads;
+static pthread_cond_t queue_cond = PTHREAD_COND_INITIALIZER;
+static int config_queue_threads = 4;
+
+static pthread_t flush_thread;
+static pthread_cond_t flush_cond = PTHREAD_COND_INITIALIZER;
 
-static pthread_t *connection_threads = NULL;
 static pthread_mutex_t connection_threads_lock = PTHREAD_MUTEX_INITIALIZER;
+static pthread_cond_t  connection_threads_done = PTHREAD_COND_INITIALIZER;
 static int connection_threads_num = 0;
 
 /* Cache stuff */
@@ -189,7 +233,6 @@ static GTree          *cache_tree = NULL;
 static cache_item_t   *cache_queue_head = NULL;
 static cache_item_t   *cache_queue_tail = NULL;
 static pthread_mutex_t cache_lock = PTHREAD_MUTEX_INITIALIZER;
-static pthread_cond_t  cache_cond = PTHREAD_COND_INITIALIZER;
 
 static int config_write_interval = 300;
 static int config_write_jitter   = 0;
@@ -221,6 +264,9 @@ static int journal_write(char *cmd, char *args);
 static void journal_done(void);
 static void journal_rotate(void);
 
+/* prototypes for forward refernces */
+static int handle_request_help (HANDLER_PROTO);
+
 /* 
  * Functions
  */
@@ -228,7 +274,8 @@ static void sig_common (const char *sig) /* {{{ */
 {
   RRDD_LOG(LOG_NOTICE, "caught SIG%s", sig);
   do_shutdown++;
-  pthread_cond_broadcast(&cache_cond);
+  pthread_cond_broadcast(&flush_cond);
+  pthread_cond_broadcast(&queue_cond);
 } /* }}} void sig_common */
 
 static void sig_int_handler (int s __attribute__((unused))) /* {{{ */
@@ -286,7 +333,7 @@ static void install_signal_handlers(void) /* {{{ */
 
 } /* }}} void install_signal_handlers */
 
-static int open_pidfile(void) /* {{{ */
+static int open_pidfile(char *action, int oflag) /* {{{ */
 {
   int fd;
   char *file;
@@ -295,14 +342,52 @@ static int open_pidfile(void) /* {{{ */
     ? config_pid_file
     : LOCALSTATEDIR "/run/rrdcached.pid";
 
-  fd = open(file, O_CREAT|O_EXCL|O_WRONLY, S_IRUSR|S_IRGRP|S_IROTH);
+  fd = open(file, oflag, S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH);
   if (fd < 0)
-    fprintf(stderr, "FATAL: cannot create '%s' (%s)\n",
-            file, rrd_strerror(errno));
+    fprintf(stderr, "rrdcached: can't %s pid file '%s' (%s)\n",
+            action, file, rrd_strerror(errno));
 
   return(fd);
 } /* }}} static int open_pidfile */
 
+/* check existing pid file to see whether a daemon is running */
+static int check_pidfile(void)
+{
+  int pid_fd;
+  pid_t pid;
+  char pid_str[16];
+
+  pid_fd = open_pidfile("open", O_RDWR);
+  if (pid_fd < 0)
+    return pid_fd;
+
+  if (read(pid_fd, pid_str, sizeof(pid_str)) <= 0)
+    return -1;
+
+  pid = atoi(pid_str);
+  if (pid <= 0)
+    return -1;
+
+  /* another running process that we can signal COULD be
+   * a competing rrdcached */
+  if (pid != getpid() && kill(pid, 0) == 0)
+  {
+    fprintf(stderr,
+            "FATAL: Another rrdcached daemon is running?? (pid %d)\n", pid);
+    close(pid_fd);
+    return -1;
+  }
+
+  lseek(pid_fd, 0, SEEK_SET);
+  ftruncate(pid_fd, 0);
+
+  fprintf(stderr,
+          "rrdcached: removed stale PID file (no rrdcached on pid %d)\n"
+          "rrdcached: starting normally.\n", pid);
+
+  return pid_fd;
+} /* }}} static int check_pidfile */
+
 static int write_pidfile (int fd) /* {{{ */
 {
   pid_t pid;
@@ -382,7 +467,7 @@ static int add_to_wbuf(listen_socket_t *sock, char *str, size_t len) /* {{{ */
 
   assert(sock != NULL);
 
-  new_buf = realloc(sock->wbuf, sock->wbuf_len + len + 1);
+  new_buf = rrd_realloc(sock->wbuf, sock->wbuf_len + len + 1);
   if (new_buf == NULL)
   {
     RRDD_LOG(LOG_ERR, "add_to_wbuf: realloc failed");
@@ -409,7 +494,7 @@ static int add_response_info(listen_socket_t *sock, char *fmt, ...) /* {{{ */
 
   va_start(argp, fmt);
 #ifdef HAVE_VSNPRINTF
-  len = vsnprintf(buffer, sizeof(buffer)-1, fmt, argp);
+  len = vsnprintf(buffer, sizeof(buffer), fmt, argp);
 #else
   len = vsprintf(buffer, fmt, argp);
 #endif
@@ -467,7 +552,7 @@ static int send_response (listen_socket_t *sock, response_code rc,
   rclen = sprintf(buffer, "%d ", lines);
   va_start(argp, fmt);
 #ifdef HAVE_VSNPRINTF
-  len = vsnprintf(buffer+rclen, sizeof(buffer)-rclen-1, fmt, argp);
+  len = vsnprintf(buffer+rclen, sizeof(buffer)-rclen, fmt, argp);
 #else
   len = vsprintf(buffer+rclen, fmt, argp);
 #endif
@@ -526,6 +611,7 @@ static void wipe_ci_values(cache_item_t *ci, time_t when)
 static void remove_from_queue(cache_item_t *ci) /* {{{ */
 {
   if (ci == NULL) return;
+  if ((ci->flags & CI_FLAGS_IN_QUEUE) == 0) return; /* not queued */
 
   if (ci->prev == NULL)
     cache_queue_head = ci->next; /* reset head */
@@ -539,20 +625,21 @@ static void remove_from_queue(cache_item_t *ci) /* {{{ */
 
   ci->next = ci->prev = NULL;
   ci->flags &= ~CI_FLAGS_IN_QUEUE;
+
+  pthread_mutex_lock (&stats_lock);
+  assert (stats_queue_length > 0);
+  stats_queue_length--;
+  pthread_mutex_unlock (&stats_lock);
+
 } /* }}} static void remove_from_queue */
 
-/* remove an entry from the tree and free all its resources.
- * must hold 'cache lock' while calling this.
- * returns 0 on success, otherwise errno */
-static int forget_file(const char *file)
+/* free the resources associated with the cache_item_t
+ * must hold cache_lock when calling this function
+ */
+static void *free_cache_item(cache_item_t *ci) /* {{{ */
 {
-  cache_item_t *ci;
+  if (ci == NULL) return NULL;
 
-  ci = g_tree_lookup(cache_tree, file);
-  if (ci == NULL)
-    return ENOENT;
-
-  g_tree_remove (cache_tree, file);
   remove_from_queue(ci);
 
   for (int i=0; i < ci->values_num; i++)
@@ -566,8 +653,8 @@ static int forget_file(const char *file)
 
   free (ci);
 
-  return 0;
-} /* }}} static int forget_file */
+  return NULL;
+} /* }}} static void *free_cache_item */
 
 /*
  * enqueue_cache_item:
@@ -587,9 +674,8 @@ static int enqueue_cache_item (cache_item_t *ci, /* {{{ */
     if (cache_queue_head == ci)
       return 0;
 
-    /* remove from the double linked list */
-    if (ci->flags & CI_FLAGS_IN_QUEUE)
-      remove_from_queue(ci);
+    /* remove if further down in queue */
+    remove_from_queue(ci);
 
     ci->prev = NULL;
     ci->next = cache_queue_head;
@@ -621,7 +707,7 @@ static int enqueue_cache_item (cache_item_t *ci, /* {{{ */
 
   ci->flags |= CI_FLAGS_IN_QUEUE;
 
-  pthread_cond_broadcast(&cache_cond);
+  pthread_cond_signal(&queue_cond);
   pthread_mutex_lock (&stats_lock);
   stats_queue_length++;
   pthread_mutex_unlock (&stats_lock);
@@ -631,7 +717,7 @@ static int enqueue_cache_item (cache_item_t *ci, /* {{{ */
 
 /*
  * tree_callback_flush:
- * Called via `g_tree_foreach' in `queue_thread_main'. `cache_lock' is held
+ * Called via `g_tree_foreach' in `flush_thread_main'. `cache_lock' is held
  * while this is in progress.
  */
 static gboolean tree_callback_flush (gpointer key, gpointer value, /* {{{ */
@@ -643,25 +729,25 @@ static gboolean tree_callback_flush (gpointer key, gpointer value, /* {{{ */
   ci = (cache_item_t *) value;
   cfd = (callback_flush_data_t *) data;
 
+  if (ci->flags & CI_FLAGS_IN_QUEUE)
+    return FALSE;
+
   if ((ci->last_flush_time <= cfd->abs_timeout)
-      && ((ci->flags & CI_FLAGS_IN_QUEUE) == 0)
       && (ci->values_num > 0))
   {
     enqueue_cache_item (ci, TAIL);
   }
   else if ((do_shutdown != 0)
-      && ((ci->flags & CI_FLAGS_IN_QUEUE) == 0)
       && (ci->values_num > 0))
   {
     enqueue_cache_item (ci, TAIL);
   }
   else if (((cfd->now - ci->last_flush_time) >= config_flush_interval)
-      && ((ci->flags & CI_FLAGS_IN_QUEUE) == 0)
       && (ci->values_num <= 0))
   {
     char **temp;
 
-    temp = (char **) realloc (cfd->keys,
+    temp = (char **) rrd_realloc (cfd->keys,
         sizeof (char *) * (cfd->keys_num + 1));
     if (temp == NULL)
     {
@@ -705,7 +791,7 @@ static int flush_old_values (int max_age)
   {
     /* should never fail, since we have held the cache_lock
      * the entire time */
-    assert( forget_file(cfd.keys[k]) == 0 );
+    assert( g_tree_remove(cache_tree, cfd.keys[k]) == TRUE );
   }
 
   if (cfd.keys != NULL)
@@ -717,27 +803,20 @@ static int flush_old_values (int max_age)
   return (0);
 } /* int flush_old_values */
 
-static void *queue_thread_main (void *args __attribute__((unused))) /* {{{ */
+static void *flush_thread_main (void *args __attribute__((unused))) /* {{{ */
 {
   struct timeval now;
   struct timespec next_flush;
-  int final_flush = 0; /* make sure we only flush once on shutdown */
+  int status;
 
   gettimeofday (&now, NULL);
   next_flush.tv_sec = now.tv_sec + config_flush_interval;
   next_flush.tv_nsec = 1000 * now.tv_usec;
 
-  pthread_mutex_lock (&cache_lock);
-  while ((do_shutdown == 0) || (cache_queue_head != NULL))
-  {
-    cache_item_t *ci;
-    char *file;
-    char **values;
-    int values_num;
-    int status;
-    int i;
+  pthread_mutex_lock(&cache_lock);
 
-    /* First, check if it's time to do the cache flush. */
+  while (!do_shutdown)
+  {
     gettimeofday (&now, NULL);
     if ((now.tv_sec > next_flush.tv_sec)
         || ((now.tv_sec == next_flush.tv_sec)
@@ -758,28 +837,48 @@ static void *queue_thread_main (void *args __attribute__((unused))) /* {{{ */
       pthread_mutex_lock(&cache_lock);
     }
 
+    status = pthread_cond_timedwait(&flush_cond, &cache_lock, &next_flush);
+    if (status != 0 && status != ETIMEDOUT)
+    {
+      RRDD_LOG (LOG_ERR, "flush_thread_main: "
+                "pthread_cond_timedwait returned %i.", status);
+    }
+  }
+
+  if (config_flush_at_shutdown)
+    flush_old_values (-1); /* flush everything */
+
+  pthread_mutex_unlock(&cache_lock);
+
+  return NULL;
+} /* void *flush_thread_main */
+
+static void *queue_thread_main (void *args __attribute__((unused))) /* {{{ */
+{
+  pthread_mutex_lock (&cache_lock);
+
+  while (!do_shutdown
+         || (cache_queue_head != NULL && config_flush_at_shutdown))
+  {
+    cache_item_t *ci;
+    char *file;
+    char **values;
+    int values_num;
+    int status;
+    int i;
+
     /* Now, check if there's something to store away. If not, wait until
-     * something comes in or it's time to do the cache flush.  if we are
-     * shutting down, do not wait around.  */
+     * something comes in.  if we are shutting down, do not wait around.  */
     if (cache_queue_head == NULL && !do_shutdown)
     {
-      status = pthread_cond_timedwait (&cache_cond, &cache_lock, &next_flush);
+      status = pthread_cond_wait (&queue_cond, &cache_lock);
       if ((status != 0) && (status != ETIMEDOUT))
       {
         RRDD_LOG (LOG_ERR, "queue_thread_main: "
-            "pthread_cond_timedwait returned %i.", status);
+            "pthread_cond_wait returned %i.", status);
       }
     }
 
-    /* We're about to shut down */
-    if (do_shutdown != 0 && !final_flush++)
-    {
-      if (config_flush_at_shutdown)
-        flush_old_values (-1); /* flush everything */
-      else
-        break;
-    }
-
     /* Check if a value has arrived. This may be NULL if we timed out or there
      * was an interrupt such as a signal. */
     if (cache_queue_head == NULL)
@@ -804,11 +903,6 @@ static void *queue_thread_main (void *args __attribute__((unused))) /* {{{ */
     wipe_ci_values(ci, time(NULL));
     remove_from_queue(ci);
 
-    pthread_mutex_lock (&stats_lock);
-    assert (stats_queue_length > 0);
-    stats_queue_length--;
-    pthread_mutex_unlock (&stats_lock);
-
     pthread_mutex_unlock (&cache_lock);
 
     rrd_clear_error ();
@@ -838,25 +932,8 @@ static void *queue_thread_main (void *args __attribute__((unused))) /* {{{ */
     }
 
     pthread_mutex_lock (&cache_lock);
-
-    /* We're about to shut down */
-    if (do_shutdown != 0 && !final_flush++)
-    {
-      if (config_flush_at_shutdown)
-          flush_old_values (-1); /* flush everything */
-      else
-        break;
-    }
-  } /* while ((do_shutdown == 0) || (cache_queue_head != NULL)) */
-  pthread_mutex_unlock (&cache_lock);
-
-  if (config_flush_at_shutdown)
-  {
-    assert(cache_queue_head == NULL);
-    RRDD_LOG(LOG_INFO, "clean shutdown; all RRDs flushed");
   }
-
-  journal_done();
+  pthread_mutex_unlock (&cache_lock);
 
   return (NULL);
 } /* }}} void *queue_thread_main */
@@ -956,6 +1033,26 @@ err:
   return 0;
 } /* }}} static int check_file_access */
 
+/* when using a base dir, convert relative paths to absolute paths.
+ * if necessary, modifies the "filename" pointer to point
+ * to the new path created in "tmp".  "tmp" is provided
+ * by the caller and sizeof(tmp) must be >= PATH_MAX.
+ *
+ * this allows us to optimize for the expected case (absolute path)
+ * with a no-op.
+ */
+static void get_abs_path(char **filename, char *tmp)
+{
+  assert(tmp != NULL);
+  assert(filename != NULL && *filename != NULL);
+
+  if (config_base_dir == NULL || **filename == '/')
+    return;
+
+  snprintf(tmp, PATH_MAX, "%s/%s", config_base_dir, *filename);
+  *filename = tmp;
+} /* }}} static int get_abs_path */
+
 /* returns 1 if we have the required privilege level,
  * otherwise issue an error to the user on sock */
 static int has_privilege (listen_socket_t *sock, /* {{{ */
@@ -998,141 +1095,17 @@ static int flush_file (const char *filename) /* {{{ */
   return (0);
 } /* }}} int flush_file */
 
-static int handle_request_help (listen_socket_t *sock, /* {{{ */
-    char *buffer, size_t buffer_size)
+static int syntax_error(listen_socket_t *sock, struct command *cmd) /* {{{ */
 {
-  int status;
-  char **help_text;
-  char *command;
-
-  char *help_help[2] =
-  {
-    "Command overview\n"
-    ,
-    "HELP [<command>]\n"
-    "FLUSH <filename>\n"
-    "FLUSHALL\n"
-    "PENDING <filename>\n"
-    "FORGET <filename>\n"
-    "UPDATE <filename> <values> [<values> ...]\n"
-    "BATCH\n"
-    "STATS\n"
-  };
-
-  char *help_flush[2] =
-  {
-    "Help for FLUSH\n"
-    ,
-    "Usage: FLUSH <filename>\n"
-    "\n"
-    "Adds the given filename to the head of the update queue and returns\n"
-    "after is has been dequeued.\n"
-  };
-
-  char *help_flushall[2] =
-  {
-    "Help for FLUSHALL\n"
-    ,
-    "Usage: FLUSHALL\n"
-    "\n"
-    "Triggers writing of all pending updates.  Returns immediately.\n"
-  };
-
-  char *help_pending[2] =
-  {
-    "Help for PENDING\n"
-    ,
-    "Usage: PENDING <filename>\n"
-    "\n"
-    "Shows any 'pending' updates for a file, in order.\n"
-    "The updates shown have not yet been written to the underlying RRD file.\n"
-  };
-
-  char *help_forget[2] =
-  {
-    "Help for FORGET\n"
-    ,
-    "Usage: FORGET <filename>\n"
-    "\n"
-    "Removes the file completely from the cache.\n"
-    "Any pending updates for the file will be lost.\n"
-  };
-
-  char *help_update[2] =
-  {
-    "Help for UPDATE\n"
-    ,
-    "Usage: UPDATE <filename> <values> [<values> ...]\n"
-    "\n"
-    "Adds the given file to the internal cache if it is not yet known and\n"
-    "appends the given value(s) to the entry. See the rrdcached(1) manpage\n"
-    "for details.\n"
-    "\n"
-    "Each <values> has the following form:\n"
-    "  <values> = <time>:<value>[:<value>[...]]\n"
-    "See the rrdupdate(1) manpage for details.\n"
-  };
-
-  char *help_stats[2] =
-  {
-    "Help for STATS\n"
-    ,
-    "Usage: STATS\n"
-    "\n"
-    "Returns some performance counters, see the rrdcached(1) manpage for\n"
-    "a description of the values.\n"
-  };
-
-  char *help_batch[2] =
-  {
-    "Help for BATCH\n"
-    ,
-    "The 'BATCH' command permits the client to initiate a bulk load\n"
-    "   of commands to rrdcached.\n"
-    "\n"
-    "Usage:\n"
-    "\n"
-    "    client: BATCH\n"
-    "    server: 0 Go ahead.  End with dot '.' on its own line.\n"
-    "    client: command #1\n"
-    "    client: command #2\n"
-    "    client: ... and so on\n"
-    "    client: .\n"
-    "    server: 2 errors\n"
-    "    server: 7 message for command #7\n"
-    "    server: 9 message for command #9\n"
-    "\n"
-    "For more information, consult the rrdcached(1) documentation.\n"
-  };
+  char *err = "Syntax error.\n";
 
-  status = buffer_get_field (&buffer, &buffer_size, &command);
-  if (status != 0)
-    help_text = help_help;
-  else
-  {
-    if (strcasecmp (command, "update") == 0)
-      help_text = help_update;
-    else if (strcasecmp (command, "flush") == 0)
-      help_text = help_flush;
-    else if (strcasecmp (command, "flushall") == 0)
-      help_text = help_flushall;
-    else if (strcasecmp (command, "pending") == 0)
-      help_text = help_pending;
-    else if (strcasecmp (command, "forget") == 0)
-      help_text = help_forget;
-    else if (strcasecmp (command, "stats") == 0)
-      help_text = help_stats;
-    else if (strcasecmp (command, "batch") == 0)
-      help_text = help_batch;
-    else
-      help_text = help_help;
-  }
+  if (cmd && cmd->syntax)
+    err = cmd->syntax;
 
-  add_response_info(sock, help_text[1]);
-  return send_response(sock, RESP_OK, help_text[0]);
-} /* }}} int handle_request_help */
+  return send_response(sock, RESP_ERR, "Usage: %s", err);
+} /* }}} static int syntax_error() */
 
-static int handle_request_stats (listen_socket_t *sock) /* {{{ */
+static int handle_request_stats (HANDLER_PROTO) /* {{{ */
 {
   uint64_t copy_queue_length;
   uint64_t copy_updates_received;
@@ -1180,16 +1153,15 @@ static int handle_request_stats (listen_socket_t *sock) /* {{{ */
   return (0);
 } /* }}} int handle_request_stats */
 
-static int handle_request_flush (listen_socket_t *sock, /* {{{ */
-    char *buffer, size_t buffer_size)
+static int handle_request_flush (HANDLER_PROTO) /* {{{ */
 {
-  char *file;
+  char *file, file_tmp[PATH_MAX];
   int status;
 
   status = buffer_get_field (&buffer, &buffer_size, &file);
   if (status != 0)
   {
-    return send_response(sock, RESP_ERR, "Usage: flush <filename>\n");
+    return syntax_error(sock,cmd);
   }
   else
   {
@@ -1197,6 +1169,7 @@ static int handle_request_flush (listen_socket_t *sock, /* {{{ */
     stats_flush_received++;
     pthread_mutex_unlock(&stats_lock);
 
+    get_abs_path(&file, file_tmp);
     if (!check_file_access(file, sock)) return 0;
 
     status = flush_file (file);
@@ -1223,14 +1196,8 @@ static int handle_request_flush (listen_socket_t *sock, /* {{{ */
   assert(1==0);
 } /* }}} int handle_request_flush */
 
-static int handle_request_flushall(listen_socket_t *sock) /* {{{ */
+static int handle_request_flushall(HANDLER_PROTO) /* {{{ */
 {
-  int status;
-
-  status = has_privilege(sock, PRIV_HIGH);
-  if (status <= 0)
-    return status;
-
   RRDD_LOG(LOG_DEBUG, "Received FLUSHALL");
 
   pthread_mutex_lock(&cache_lock);
@@ -1240,21 +1207,17 @@ static int handle_request_flushall(listen_socket_t *sock) /* {{{ */
   return send_response(sock, RESP_OK, "Started flush.\n");
 } /* }}} static int handle_request_flushall */
 
-static int handle_request_pending(listen_socket_t *sock, /* {{{ */
-                                  char *buffer, size_t buffer_size)
+static int handle_request_pending(HANDLER_PROTO) /* {{{ */
 {
   int status;
-  char *file;
+  char *file, file_tmp[PATH_MAX];
   cache_item_t *ci;
 
   status = buffer_get_field(&buffer, &buffer_size, &file);
   if (status != 0)
-    return send_response(sock, RESP_ERR,
-                         "Usage: PENDING <filename>\n");
+    return syntax_error(sock,cmd);
 
-  status = has_privilege(sock, PRIV_HIGH);
-  if (status <= 0)
-    return status;
+  get_abs_path(&file, file_tmp);
 
   pthread_mutex_lock(&cache_lock);
   ci = g_tree_lookup(cache_tree, file);
@@ -1271,28 +1234,24 @@ static int handle_request_pending(listen_socket_t *sock, /* {{{ */
   return send_response(sock, RESP_OK, "updates pending\n");
 } /* }}} static int handle_request_pending */
 
-static int handle_request_forget(listen_socket_t *sock, /* {{{ */
-                                 char *buffer, size_t buffer_size)
+static int handle_request_forget(HANDLER_PROTO) /* {{{ */
 {
   int status;
-  char *file;
+  gboolean found;
+  char *file, file_tmp[PATH_MAX];
 
   status = buffer_get_field(&buffer, &buffer_size, &file);
   if (status != 0)
-    return send_response(sock, RESP_ERR,
-                         "Usage: FORGET <filename>\n");
-
-  status = has_privilege(sock, PRIV_HIGH);
-  if (status <= 0)
-    return status;
+    return syntax_error(sock,cmd);
 
+  get_abs_path(&file, file_tmp);
   if (!check_file_access(file, sock)) return 0;
 
   pthread_mutex_lock(&cache_lock);
-  status = forget_file(file);
+  found = g_tree_remove(cache_tree, file);
   pthread_mutex_unlock(&cache_lock);
 
-  if (status == 0)
+  if (found == TRUE)
   {
     if (sock != NULL)
       journal_write("forget", file);
@@ -1300,41 +1259,51 @@ static int handle_request_forget(listen_socket_t *sock, /* {{{ */
     return send_response(sock, RESP_OK, "Gone!\n");
   }
   else
-    return send_response(sock, RESP_ERR, "cannot forget: %s\n",
-                         status < 0 ? "Internal error" : rrd_strerror(status));
+    return send_response(sock, RESP_ERR, "%s\n", rrd_strerror(ENOENT));
 
   /* NOTREACHED */
   assert(1==0);
 } /* }}} static int handle_request_forget */
 
-static int handle_request_update (listen_socket_t *sock, /* {{{ */
-                                  time_t now,
-                                  char *buffer, size_t buffer_size)
+static int handle_request_queue (HANDLER_PROTO) /* {{{ */
 {
-  char *file;
+  cache_item_t *ci;
+
+  pthread_mutex_lock(&cache_lock);
+
+  ci = cache_queue_head;
+  while (ci != NULL)
+  {
+    add_response_info(sock, "%d %s\n", ci->values_num, ci->file);
+    ci = ci->next;
+  }
+
+  pthread_mutex_unlock(&cache_lock);
+
+  return send_response(sock, RESP_OK, "in queue.\n");
+} /* }}} int handle_request_queue */
+
+static int handle_request_update (HANDLER_PROTO) /* {{{ */
+{
+  char *file, file_tmp[PATH_MAX];
   int values_num = 0;
-  int bad_timestamps = 0;
   int status;
   char orig_buf[CMD_MAX];
 
   cache_item_t *ci;
 
-  status = has_privilege(sock, PRIV_HIGH);
-  if (status <= 0)
-    return status;
-
   /* save it for the journal later */
   strncpy(orig_buf, buffer, sizeof(orig_buf)-1);
 
   status = buffer_get_field (&buffer, &buffer_size, &file);
   if (status != 0)
-    return send_response(sock, RESP_ERR,
-                         "Usage: UPDATE <filename> <values> [<values> ...]\n");
+    return syntax_error(sock,cmd);
 
   pthread_mutex_lock(&stats_lock);
   stats_updates_received++;
   pthread_mutex_unlock(&stats_lock);
 
+  get_abs_path(&file, file_tmp);
   if (!check_file_access(file, sock)) return 0;
 
   pthread_mutex_lock (&cache_lock);
@@ -1387,9 +1356,10 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */
 
     wipe_ci_values(ci, now);
     ci->flags = CI_FLAGS_IN_TREE;
+    pthread_cond_init(&ci->flushed, NULL);
 
     pthread_mutex_lock(&cache_lock);
-    g_tree_insert (cache_tree, (void *) ci->file, (void *) ci);
+    g_tree_replace (cache_tree, (void *) ci->file, (void *) ci);
   } /* }}} */
   assert (ci != NULL);
 
@@ -1415,23 +1385,22 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */
     stamp = strtol(value, &eostamp, 10);
     if (eostamp == value || eostamp == NULL || *eostamp != ':')
     {
-      ++bad_timestamps;
-      add_response_info(sock, "Cannot find timestamp in '%s'!\n", value);
-      continue;
+      pthread_mutex_unlock(&cache_lock);
+      return send_response(sock, RESP_ERR,
+                           "Cannot find timestamp in '%s'!\n", value);
     }
     else if (stamp <= ci->last_update_stamp)
     {
-      ++bad_timestamps;
-      add_response_info(sock,
-                        "illegal attempt to update using time %ld when"
-                        " last update time is %ld (minimum one second step)\n",
-                        stamp, ci->last_update_stamp);
-      continue;
+      pthread_mutex_unlock(&cache_lock);
+      return send_response(sock, RESP_ERR,
+                           "illegal attempt to update using time %ld when last"
+                           " update time is %ld (minimum one second step)\n",
+                           stamp, ci->last_update_stamp);
     }
     else
       ci->last_update_stamp = stamp;
 
-    temp = (char **) realloc (ci->values,
+    temp = (char **) rrd_realloc (ci->values,
         sizeof (char *) * (ci->values_num + 1));
     if (temp == NULL)
     {
@@ -1461,18 +1430,7 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */
   pthread_mutex_unlock (&cache_lock);
 
   if (values_num < 1)
-  {
-    /* if we had only one update attempt, then return the full
-       error message... try to get the most information out
-       of the limited error space allowed by the protocol
-    */
-    if (bad_timestamps == 1)
-      return send_response(sock, RESP_ERR, "%s", sock->wbuf);
-    else
-      return send_response(sock, RESP_ERR,
-                           "No values updated (%d bad timestamps).\n",
-                           bad_timestamps);
-  }
+    return send_response(sock, RESP_ERR, "No values updated.\n");
   else
     return send_response(sock, RESP_OK,
                          "errors, enqueued %i value(s).\n", values_num);
@@ -1485,7 +1443,7 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */
 /* we came across a "WROTE" entry during journal replay.
  * throw away any values that we have accumulated for this file
  */
-static int handle_request_wrote (const char *buffer, time_t now) /* {{{ */
+static int handle_request_wrote (HANDLER_PROTO) /* {{{ */
 {
   int i;
   cache_item_t *ci;
@@ -1516,7 +1474,7 @@ static int handle_request_wrote (const char *buffer, time_t now) /* {{{ */
 } /* }}} int handle_request_wrote */
 
 /* start "BATCH" processing */
-static int batch_start (listen_socket_t *sock) /* {{{ */
+static int batch_start (HANDLER_PROTO) /* {{{ */
 {
   int status;
   if (sock->batch_start)
@@ -1531,7 +1489,7 @@ static int batch_start (listen_socket_t *sock) /* {{{ */
 } /* }}} static int batch_start */
 
 /* finish "BATCH" processing and return results to the client */
-static int batch_done (listen_socket_t *sock) /* {{{ */
+static int batch_done (HANDLER_PROTO) /* {{{ */
 {
   assert(sock->batch_start);
   sock->batch_start = 0;
@@ -1539,20 +1497,231 @@ static int batch_done (listen_socket_t *sock) /* {{{ */
   return send_response(sock, RESP_OK, "errors\n");
 } /* }}} static int batch_done */
 
+static int handle_request_quit (HANDLER_PROTO) /* {{{ */
+{
+  return -1;
+} /* }}} static int handle_request_quit */
+
+struct command COMMANDS[] = {
+  {
+    "UPDATE",
+    handle_request_update,
+    PRIV_HIGH,
+    CMD_CONTEXT_ANY,
+    "UPDATE <filename> <values> [<values> ...]\n"
+    ,
+    "Adds the given file to the internal cache if it is not yet known and\n"
+    "appends the given value(s) to the entry. See the rrdcached(1) manpage\n"
+    "for details.\n"
+    "\n"
+    "Each <values> has the following form:\n"
+    "  <values> = <time>:<value>[:<value>[...]]\n"
+    "See the rrdupdate(1) manpage for details.\n"
+  },
+  {
+    "WROTE",
+    handle_request_wrote,
+    PRIV_HIGH,
+    CMD_CONTEXT_JOURNAL,
+    NULL,
+    NULL
+  },
+  {
+    "FLUSH",
+    handle_request_flush,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT | CMD_CONTEXT_BATCH,
+    "FLUSH <filename>\n"
+    ,
+    "Adds the given filename to the head of the update queue and returns\n"
+    "after it has been dequeued.\n"
+  },
+  {
+    "FLUSHALL",
+    handle_request_flushall,
+    PRIV_HIGH,
+    CMD_CONTEXT_CLIENT,
+    "FLUSHALL\n"
+    ,
+    "Triggers writing of all pending updates.  Returns immediately.\n"
+  },
+  {
+    "PENDING",
+    handle_request_pending,
+    PRIV_HIGH,
+    CMD_CONTEXT_CLIENT,
+    "PENDING <filename>\n"
+    ,
+    "Shows any 'pending' updates for a file, in order.\n"
+    "The updates shown have not yet been written to the underlying RRD file.\n"
+  },
+  {
+    "FORGET",
+    handle_request_forget,
+    PRIV_HIGH,
+    CMD_CONTEXT_ANY,
+    "FORGET <filename>\n"
+    ,
+    "Removes the file completely from the cache.\n"
+    "Any pending updates for the file will be lost.\n"
+  },
+  {
+    "QUEUE",
+    handle_request_queue,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT,
+    "QUEUE\n"
+    ,
+        "Shows all files in the output queue.\n"
+    "The output is zero or more lines in the following format:\n"
+    "(where <num_vals> is the number of values to be written)\n"
+    "\n"
+    "<num_vals> <filename>\n"
+  },
+  {
+    "STATS",
+    handle_request_stats,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT,
+    "STATS\n"
+    ,
+    "Returns some performance counters, see the rrdcached(1) manpage for\n"
+    "a description of the values.\n"
+  },
+  {
+    "HELP",
+    handle_request_help,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT,
+    "HELP [<command>]\n",
+    NULL, /* special! */
+  },
+  {
+    "BATCH",
+    batch_start,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT,
+    "BATCH\n"
+    ,
+    "The 'BATCH' command permits the client to initiate a bulk load\n"
+    "   of commands to rrdcached.\n"
+    "\n"
+    "Usage:\n"
+    "\n"
+    "    client: BATCH\n"
+    "    server: 0 Go ahead.  End with dot '.' on its own line.\n"
+    "    client: command #1\n"
+    "    client: command #2\n"
+    "    client: ... and so on\n"
+    "    client: .\n"
+    "    server: 2 errors\n"
+    "    server: 7 message for command #7\n"
+    "    server: 9 message for command #9\n"
+    "\n"
+    "For more information, consult the rrdcached(1) documentation.\n"
+  },
+  {
+    ".",   /* BATCH terminator */
+    batch_done,
+    PRIV_LOW,
+    CMD_CONTEXT_BATCH,
+    NULL,
+    NULL
+  },
+  {
+    "QUIT",
+    handle_request_quit,
+    PRIV_LOW,
+    CMD_CONTEXT_CLIENT | CMD_CONTEXT_BATCH,
+    "QUIT\n"
+    ,
+    "Disconnect from rrdcached.\n"
+  },
+  {NULL,NULL,0,0,NULL,NULL}  /* LAST ENTRY */
+};
+
+static struct command *find_command(char *cmd)
+{
+  struct command *c = COMMANDS;
+
+  while (c->cmd != NULL)
+  {
+    if (strcasecmp(cmd, c->cmd) == 0)
+      break;
+    c++;
+  }
+
+  if (c->cmd == NULL)
+    return NULL;
+  else
+    return c;
+}
+
+/* check whether commands are received in the expected context */
+static int command_check_context(listen_socket_t *sock, struct command *cmd)
+{
+  if (sock == NULL)
+    return (cmd->context & CMD_CONTEXT_JOURNAL);
+  else if (sock->batch_start)
+    return (cmd->context & CMD_CONTEXT_BATCH);
+  else
+    return (cmd->context & CMD_CONTEXT_CLIENT);
+
+  /* NOTREACHED */
+  assert(1==0);
+}
+
+static int handle_request_help (HANDLER_PROTO) /* {{{ */
+{
+  int status;
+  char *cmd_str;
+  char *resp_txt;
+  struct command *help = NULL;
+
+  status = buffer_get_field (&buffer, &buffer_size, &cmd_str);
+  if (status == 0)
+    help = find_command(cmd_str);
+
+  if (help && (help->syntax || help->help))
+  {
+    char tmp[CMD_MAX];
+
+    snprintf(tmp, sizeof(tmp)-1, "Help for %s\n", help->cmd);
+    resp_txt = tmp;
+
+    if (help->syntax)
+      add_response_info(sock, "Usage: %s\n", help->syntax);
+
+    if (help->help)
+      add_response_info(sock, "%s\n", help->help);
+  }
+  else
+  {
+    help = COMMANDS;
+    resp_txt = "Command overview\n";
+
+    while (help->cmd)
+    {
+      if (help->syntax)
+        add_response_info(sock, "%s", help->syntax);
+      help++;
+    }
+  }
+
+  return send_response(sock, RESP_OK, resp_txt);
+} /* }}} int handle_request_help */
+
 /* if sock==NULL, we are in journal replay mode */
-static int handle_request (listen_socket_t *sock, /* {{{ */
-                           time_t now,
-                           char *buffer, size_t buffer_size)
+static int handle_request (DISPATCH_PROTO) /* {{{ */
 {
-  char *buffer_ptr;
-  char *command;
+  char *buffer_ptr = buffer;
+  char *cmd_str = NULL;
+  struct command *cmd = NULL;
   int status;
 
   assert (buffer[buffer_size - 1] == '\0');
 
-  buffer_ptr = buffer;
-  command = NULL;
-  status = buffer_get_field (&buffer_ptr, &buffer_size, &command);
+  status = buffer_get_field (&buffer_ptr, &buffer_size, &cmd_str);
   if (status != 0)
   {
     RRDD_LOG (LOG_INFO, "handle_request: Unable parse command.");
@@ -1562,34 +1731,18 @@ static int handle_request (listen_socket_t *sock, /* {{{ */
   if (sock != NULL && sock->batch_start)
     sock->batch_cmd++;
 
-  if (strcasecmp (command, "update") == 0)
-    return (handle_request_update (sock, now, buffer_ptr, buffer_size));
-  else if (strcasecmp (command, "wrote") == 0 && sock == NULL)
-  {
-    /* this is only valid in replay mode */
-    return (handle_request_wrote (buffer_ptr, now));
-  }
-  else if (strcasecmp (command, "flush") == 0)
-    return (handle_request_flush (sock, buffer_ptr, buffer_size));
-  else if (strcasecmp (command, "flushall") == 0)
-    return (handle_request_flushall(sock));
-  else if (strcasecmp (command, "pending") == 0)
-    return (handle_request_pending(sock, buffer_ptr, buffer_size));
-  else if (strcasecmp (command, "forget") == 0)
-    return (handle_request_forget(sock, buffer_ptr, buffer_size));
-  else if (strcasecmp (command, "stats") == 0)
-    return (handle_request_stats (sock));
-  else if (strcasecmp (command, "help") == 0)
-    return (handle_request_help (sock, buffer_ptr, buffer_size));
-  else if (strcasecmp (command, "batch") == 0 && sock != NULL)
-    return batch_start(sock);
-  else if (strcasecmp (command, ".") == 0 && sock != NULL && sock->batch_start)
-    return batch_done(sock);
-  else
-    return send_response(sock, RESP_ERR, "Unknown command: %s\n", command);
+  cmd = find_command(cmd_str);
+  if (!cmd)
+    return send_response(sock, RESP_ERR, "Unknown command: %s\n", cmd_str);
 
-  /* NOTREACHED */
-  assert(1==0);
+  status = has_privilege(sock, cmd->min_priv);
+  if (status <= 0)
+    return status;
+
+  if (!command_check_context(sock, cmd))
+    return send_response(sock, RESP_ERR, "Can't use '%s' here.\n", cmd_str);
+
+  return cmd->handler(cmd, sock, now, buffer_ptr, buffer_size);
 } /* }}} int handle_request */
 
 /* MUST NOT hold journal_lock before calling this */
@@ -1704,7 +1857,7 @@ static int journal_replay (const char *file) /* {{{ */
   if (file == NULL) return 0;
 
   {
-    char *reason;
+    char *reason = "unknown error";
     int status = 0;
     struct stat statbuf;
 
@@ -1813,20 +1966,30 @@ static void journal_init(void) /* {{{ */
 
 } /* }}} static void journal_init */
 
-static void close_connection(listen_socket_t *sock)
+static void free_listen_socket(listen_socket_t *sock) /* {{{ */
 {
-  close(sock->fd) ;  sock->fd   = -1;
+  assert(sock != NULL);
+
   free(sock->rbuf);  sock->rbuf = NULL;
   free(sock->wbuf);  sock->wbuf = NULL;
-
   free(sock);
-}
+} /* }}} void free_listen_socket */
+
+static void close_connection(listen_socket_t *sock) /* {{{ */
+{
+  if (sock->fd >= 0)
+  {
+    close(sock->fd);
+    sock->fd = -1;
+  }
+
+  free_listen_socket(sock);
+
+} /* }}} void close_connection */
 
 static void *connection_thread_main (void *args) /* {{{ */
 {
-  pthread_t self;
   listen_socket_t *sock;
-  int i;
   int fd;
 
   sock = (listen_socket_t *) args;
@@ -1843,22 +2006,7 @@ static void *connection_thread_main (void *args) /* {{{ */
   }
 
   pthread_mutex_lock (&connection_threads_lock);
-  {
-    pthread_t *temp;
-
-    temp = (pthread_t *) realloc (connection_threads,
-        sizeof (pthread_t) * (connection_threads_num + 1));
-    if (temp == NULL)
-    {
-      RRDD_LOG (LOG_ERR, "connection_thread_main: realloc failed.");
-    }
-    else
-    {
-      connection_threads = temp;
-      connection_threads[connection_threads_num] = pthread_self ();
-      connection_threads_num++;
-    }
-  }
+  connection_threads_num++;
   pthread_mutex_unlock (&connection_threads_lock);
 
   while (do_shutdown == 0)
@@ -1926,24 +2074,11 @@ static void *connection_thread_main (void *args) /* {{{ */
 out_close:
   close_connection(sock);
 
-  self = pthread_self ();
   /* Remove this thread from the connection threads list */
   pthread_mutex_lock (&connection_threads_lock);
-  /* Find out own index in the array */
-  for (i = 0; i < connection_threads_num; i++)
-    if (pthread_equal (connection_threads[i], self) != 0)
-      break;
-  assert (i < connection_threads_num);
-
-  /* Move the trailing threads forward. */
-  if (i < (connection_threads_num - 1))
-  {
-    memmove (connection_threads + i,
-        connection_threads + i + 1,
-        sizeof (pthread_t) * (connection_threads_num - i - 1));
-  }
-
   connection_threads_num--;
+  if (connection_threads_num <= 0)
+    pthread_cond_broadcast(&connection_threads_done);
   pthread_mutex_unlock (&connection_threads_lock);
 
   return (NULL);
@@ -1961,11 +2096,11 @@ static int open_listen_socket_unix (const listen_socket_t *sock) /* {{{ */
   if (strncmp(path, "unix:", strlen("unix:")) == 0)
     path += strlen("unix:");
 
-  temp = (listen_socket_t *) realloc (listen_fds,
+  temp = (listen_socket_t *) rrd_realloc (listen_fds,
       sizeof (listen_fds[0]) * (listen_fds_num + 1));
   if (temp == NULL)
   {
-    RRDD_LOG (LOG_ERR, "open_listen_socket_unix: realloc failed.");
+    fprintf (stderr, "rrdcached: open_listen_socket_unix: realloc failed.\n");
     return (-1);
   }
   listen_fds = temp;
@@ -1974,7 +2109,8 @@ static int open_listen_socket_unix (const listen_socket_t *sock) /* {{{ */
   fd = socket (PF_UNIX, SOCK_STREAM, /* protocol = */ 0);
   if (fd < 0)
   {
-    RRDD_LOG (LOG_ERR, "open_listen_socket_unix: socket(2) failed.");
+    fprintf (stderr, "rrdcached: unix socket(2) failed: %s\n",
+             rrd_strerror(errno));
     return (-1);
   }
 
@@ -1982,19 +2118,26 @@ static int open_listen_socket_unix (const listen_socket_t *sock) /* {{{ */
   sa.sun_family = AF_UNIX;
   strncpy (sa.sun_path, path, sizeof (sa.sun_path) - 1);
 
+  /* if we've gotten this far, we own the pid file.  any daemon started
+   * with the same args must not be alive.  therefore, ensure that we can
+   * create the socket...
+   */
+  unlink(path);
+
   status = bind (fd, (struct sockaddr *) &sa, sizeof (sa));
   if (status != 0)
   {
-    RRDD_LOG (LOG_ERR, "open_listen_socket_unix: bind(2) failed.");
+    fprintf (stderr, "rrdcached: bind(%s) failed: %s.\n",
+             path, rrd_strerror(errno));
     close (fd);
-    unlink (path);
     return (-1);
   }
 
   status = listen (fd, /* backlog = */ 10);
   if (status != 0)
   {
-    RRDD_LOG (LOG_ERR, "open_listen_socket_unix: listen(2) failed.");
+    fprintf (stderr, "rrdcached: listen(%s) failed: %s.\n",
+             path, rrd_strerror(errno));
     close (fd);
     unlink (path);
     return (-1);
@@ -2019,7 +2162,7 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
   char *port;
   int status;
 
-  strncpy (addr_copy, sock->addr, sizeof (addr_copy));
+  strncpy (addr_copy, sock->addr, sizeof(addr_copy)-1);
   addr_copy[sizeof (addr_copy) - 1] = 0;
   addr = addr_copy;
 
@@ -2040,8 +2183,7 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     port = strchr (addr, ']');
     if (port == NULL)
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: Malformed address: %s",
-          sock->addr);
+      fprintf (stderr, "rrdcached: Malformed address: %s\n", sock->addr);
       return (-1);
     }
     *port = 0;
@@ -2053,8 +2195,7 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
       port = NULL;
     else
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: Garbage after address: %s",
-          port);
+      fprintf (stderr, "rrdcached: Garbage after address: %s\n", port);
       return (-1);
     }
   } /* if (*addr = ']') */
@@ -2073,8 +2214,8 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
                         &ai_hints, &ai_res);
   if (status != 0)
   {
-    RRDD_LOG (LOG_ERR, "open_listen_socket_network: getaddrinfo(%s) failed: "
-        "%s", addr, gai_strerror (status));
+    fprintf (stderr, "rrdcached: getaddrinfo(%s) failed: %s\n",
+             addr, gai_strerror (status));
     return (-1);
   }
 
@@ -2084,11 +2225,12 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     listen_socket_t *temp;
     int one = 1;
 
-    temp = (listen_socket_t *) realloc (listen_fds,
+    temp = (listen_socket_t *) rrd_realloc (listen_fds,
         sizeof (listen_fds[0]) * (listen_fds_num + 1));
     if (temp == NULL)
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: realloc failed.");
+      fprintf (stderr,
+               "rrdcached: open_listen_socket_network: realloc failed.\n");
       continue;
     }
     listen_fds = temp;
@@ -2097,7 +2239,8 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     fd = socket (ai_ptr->ai_family, ai_ptr->ai_socktype, ai_ptr->ai_protocol);
     if (fd < 0)
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: socket(2) failed.");
+      fprintf (stderr, "rrdcached: network socket(2) failed: %s.\n",
+               rrd_strerror(errno));
       continue;
     }
 
@@ -2106,7 +2249,8 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     status = bind (fd, ai_ptr->ai_addr, ai_ptr->ai_addrlen);
     if (status != 0)
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: bind(2) failed.");
+      fprintf (stderr, "rrdcached: bind(%s) failed: %s.\n",
+               sock->addr, rrd_strerror(errno));
       close (fd);
       continue;
     }
@@ -2114,8 +2258,10 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     status = listen (fd, /* backlog = */ 10);
     if (status != 0)
     {
-      RRDD_LOG (LOG_ERR, "open_listen_socket_network: listen(2) failed.");
+      fprintf (stderr, "rrdcached: listen(%s) failed: %s\n.",
+               sock->addr, rrd_strerror(errno));
       close (fd);
+      freeaddrinfo(ai_res);
       return (-1);
     }
 
@@ -2124,6 +2270,7 @@ static int open_listen_socket_network(const listen_socket_t *sock) /* {{{ */
     listen_fds_num++;
   } /* for (ai_ptr) */
 
+  freeaddrinfo(ai_res);
   return (0);
 } /* }}} static int open_listen_socket_network */
 
@@ -2165,21 +2312,9 @@ static void *listen_thread_main (void *args __attribute__((unused))) /* {{{ */
   int status;
   int i;
 
-  for (i = 0; i < config_listen_address_list_len; i++)
-    open_listen_socket (config_listen_address_list[i]);
-
-  if (config_listen_address_list_len < 1)
-  {
-    listen_socket_t sock;
-    memset(&sock, 0, sizeof(sock));
-    strncpy(sock.addr, RRDCACHED_DEFAULT_ADDRESS, sizeof(sock.addr));
-    open_listen_socket (&sock);
-  }
-
   if (listen_fds_num < 1)
   {
-    RRDD_LOG (LOG_ERR, "listen_thread_main: No listen sockets "
-        "could be opened. Sorry.");
+    RRDD_LOG(LOG_ERR, "listen_thread_main: no listen_fds !");
     return (NULL);
   }
 
@@ -2196,7 +2331,6 @@ static void *listen_thread_main (void *args __attribute__((unused))) /* {{{ */
 
   while (do_shutdown == 0)
   {
-    assert (pollfds_num == ((int) listen_fds_num));
     for (i = 0; i < pollfds_num; i++)
     {
       pollfds[i].fd = listen_fds[i].fd;
@@ -2276,30 +2410,51 @@ static void *listen_thread_main (void *args __attribute__((unused))) /* {{{ */
 
   pthread_mutex_lock (&connection_threads_lock);
   while (connection_threads_num > 0)
-  {
-    pthread_t wait_for;
-
-    wait_for = connection_threads[0];
-
-    pthread_mutex_unlock (&connection_threads_lock);
-    pthread_join (wait_for, /* retval = */ NULL);
-    pthread_mutex_lock (&connection_threads_lock);
-  }
+    pthread_cond_wait(&connection_threads_done, &connection_threads_lock);
   pthread_mutex_unlock (&connection_threads_lock);
 
+  free(pollfds);
+
   return (NULL);
 } /* }}} void *listen_thread_main */
 
 static int daemonize (void) /* {{{ */
 {
-  int status;
-  int fd;
+  int pid_fd;
   char *base_dir;
 
   daemon_uid = geteuid();
 
-  fd = open_pidfile();
-  if (fd < 0) return fd;
+  pid_fd = open_pidfile("create", O_CREAT|O_EXCL|O_WRONLY);
+  if (pid_fd < 0)
+    pid_fd = check_pidfile();
+  if (pid_fd < 0)
+    return pid_fd;
+
+  /* open all the listen sockets */
+  if (config_listen_address_list_len > 0)
+  {
+    for (int i = 0; i < config_listen_address_list_len; i++)
+    {
+      open_listen_socket (config_listen_address_list[i]);
+      free_listen_socket (config_listen_address_list[i]);
+    }
+
+    free(config_listen_address_list);
+  }
+  else
+  {
+    listen_socket_t sock;
+    memset(&sock, 0, sizeof(sock));
+    strncpy(sock.addr, RRDCACHED_DEFAULT_ADDRESS, sizeof(sock.addr)-1);
+    open_listen_socket (&sock);
+  }
+
+  if (listen_fds_num < 1)
+  {
+    fprintf (stderr, "rrdcached: FATAL: cannot open any listen sockets\n");
+    goto error;
+  }
 
   if (!stay_foreground)
   {
@@ -2309,12 +2464,10 @@ static int daemonize (void) /* {{{ */
     if (child < 0)
     {
       fprintf (stderr, "daemonize: fork(2) failed.\n");
-      return (-1);
+      goto error;
     }
     else if (child > 0)
-    {
-      return (1);
-    }
+      exit(0);
 
     /* Become session leader */
     setsid ();
@@ -2333,11 +2486,11 @@ static int daemonize (void) /* {{{ */
   base_dir = (config_base_dir != NULL)
     ? config_base_dir
     : "/tmp";
-  status = chdir (base_dir);
-  if (status != 0)
+
+  if (chdir (base_dir) != 0)
   {
     fprintf (stderr, "daemonize: chdir (%s) failed.\n", base_dir);
-    return (-1);
+    goto error;
   }
 
   install_signal_handlers();
@@ -2345,26 +2498,50 @@ static int daemonize (void) /* {{{ */
   openlog ("rrdcached", LOG_PID, LOG_DAEMON);
   RRDD_LOG(LOG_INFO, "starting up");
 
-  cache_tree = g_tree_new ((GCompareFunc) strcmp);
+  cache_tree = g_tree_new_full ((GCompareDataFunc) strcmp, NULL, NULL,
+                                (GDestroyNotify) free_cache_item);
   if (cache_tree == NULL)
   {
     RRDD_LOG (LOG_ERR, "daemonize: g_tree_new failed.");
-    return (-1);
+    goto error;
   }
 
-  status = write_pidfile (fd);
-  return status;
+  return write_pidfile (pid_fd);
+
+error:
+  remove_pidfile();
+  return -1;
 } /* }}} int daemonize */
 
 static int cleanup (void) /* {{{ */
 {
   do_shutdown++;
 
-  pthread_cond_signal (&cache_cond);
-  pthread_join (queue_thread, /* return = */ NULL);
+  pthread_cond_broadcast (&flush_cond);
+  pthread_join (flush_thread, NULL);
+
+  pthread_cond_broadcast (&queue_cond);
+  for (int i = 0; i < config_queue_threads; i++)
+    pthread_join (queue_threads[i], NULL);
 
+  if (config_flush_at_shutdown)
+  {
+    assert(cache_queue_head == NULL);
+    RRDD_LOG(LOG_INFO, "clean shutdown; all RRDs flushed");
+  }
+
+  journal_done();
   remove_pidfile ();
 
+  free(queue_threads);
+  free(config_base_dir);
+  free(config_pid_file);
+  free(journal_cur);
+  free(journal_old);
+
+  pthread_mutex_lock(&cache_lock);
+  g_tree_destroy(cache_tree);
+
   RRDD_LOG(LOG_INFO, "goodbye");
   closelog ();
 
@@ -2376,7 +2553,7 @@ static int read_options (int argc, char **argv) /* {{{ */
   int option;
   int status = 0;
 
-  while ((option = getopt(argc, argv, "gl:L:f:w:b:Bz:p:j:h?F")) != -1)
+  while ((option = getopt(argc, argv, "gl:L:f:w:z:t:Bb:p:Fj:h?")) != -1)
   {
     switch (option)
     {
@@ -2398,7 +2575,7 @@ static int read_options (int argc, char **argv) /* {{{ */
         }
         memset(new, 0, sizeof(listen_socket_t));
 
-        temp = (listen_socket_t **) realloc (config_listen_address_list,
+        temp = (listen_socket_t **) rrd_realloc (config_listen_address_list,
             sizeof (listen_socket_t *) * (config_listen_address_list_len + 1));
         if (temp == NULL)
         {
@@ -2461,6 +2638,20 @@ static int read_options (int argc, char **argv) /* {{{ */
         break;
       }
 
+      case 't':
+      {
+        int threads;
+        threads = atoi(optarg);
+        if (threads >= 1)
+          config_queue_threads = threads;
+        else
+        {
+          fprintf (stderr, "Invalid thread count: -t %s\n", optarg);
+          return 1;
+        }
+      }
+      break;
+
       case 'B':
         config_write_base_only = 1;
         break;
@@ -2468,6 +2659,7 @@ static int read_options (int argc, char **argv) /* {{{ */
       case 'b':
       {
         size_t len;
+        char base_realpath[PATH_MAX];
 
         if (config_base_dir != NULL)
           free (config_base_dir);
@@ -2478,6 +2670,27 @@ static int read_options (int argc, char **argv) /* {{{ */
           return (3);
         }
 
+        /* make sure that the base directory is not resolved via
+         * symbolic links.  this makes some performance-enhancing
+         * assumptions possible (we don't have to resolve paths
+         * that start with a "/")
+         */
+        if (realpath(config_base_dir, base_realpath) == NULL)
+        {
+          fprintf (stderr, "Invalid base directory '%s'.\n", config_base_dir);
+          return 5;
+        }
+        else if (strncmp(config_base_dir,
+                         base_realpath, sizeof(base_realpath)) != 0)
+        {
+          fprintf(stderr,
+                  "Base directory (-b) resolved via file system links!\n"
+                  "Please consult rrdcached '-b' documentation!\n"
+                  "Consider specifying the real directory (%s)\n",
+                  base_realpath);
+          return 5;
+        }
+
         len = strlen (config_base_dir);
         while ((len > 0) && (config_base_dir[len - 1] == '/'))
         {
@@ -2558,6 +2771,7 @@ static int read_options (int argc, char **argv) /* {{{ */
             "  -L <address>  Socket address to listen to ('FLUSH' only).\n"
             "  -w <seconds>  Interval in which to write data.\n"
             "  -z <delay>    Delay writes up to <delay> seconds to spread load\n"
+            "  -t <threads>  Number of write threads.\n"
             "  -f <seconds>  Interval in which to flush dead data.\n"
             "  -p <file>     Location of the PID-file.\n"
             "  -b <dir>      Base directory to change to.\n"
@@ -2606,33 +2820,40 @@ int main (int argc, char **argv)
   }
 
   status = daemonize ();
-  if (status == 1)
+  if (status != 0)
   {
-    struct sigaction sigchld;
+    fprintf (stderr, "rrdcached: daemonize failed, exiting.\n");
+    return (1);
+  }
 
-    memset (&sigchld, 0, sizeof (sigchld));
-    sigchld.sa_handler = SIG_IGN;
-    sigaction (SIGCHLD, &sigchld, NULL);
+  journal_init();
 
-    return (0);
-  }
-  else if (status != 0)
+  /* start the queue threads */
+  queue_threads = calloc(config_queue_threads, sizeof(*queue_threads));
+  if (queue_threads == NULL)
   {
-    fprintf (stderr, "daemonize failed, exiting.\n");
+    RRDD_LOG (LOG_ERR, "FATAL: cannot calloc queue threads");
+    cleanup();
     return (1);
   }
+  for (int i = 0; i < config_queue_threads; i++)
+  {
+    memset (&queue_threads[i], 0, sizeof (*queue_threads));
+    status = pthread_create (&queue_threads[i], NULL, queue_thread_main, NULL);
+    if (status != 0)
+    {
+      RRDD_LOG (LOG_ERR, "FATAL: cannot create queue thread");
+      cleanup();
+      return (1);
+    }
+  }
 
-  journal_init();
-
-  /* start the queue thread */
-  memset (&queue_thread, 0, sizeof (queue_thread));
-  status = pthread_create (&queue_thread,
-                           NULL, /* attr */
-                           queue_thread_main,
-                           NULL); /* args */
+  /* start the flush thread */
+  memset(&flush_thread, 0, sizeof(flush_thread));
+  status = pthread_create (&flush_thread, NULL, flush_thread_main, NULL);
   if (status != 0)
   {
-    RRDD_LOG (LOG_ERR, "FATAL: cannot create queue thread");
+    RRDD_LOG (LOG_ERR, "FATAL: cannot create flush thread");
     cleanup();
     return (1);
   }