dss.c: Remove pointless comment for main().
[dss.git] / dss.c
diff --git a/dss.c b/dss.c
index 2a0fb9edce9cb6af9ee9eecc880f9148bb2546bc..bdbdd043ef8c4651bf921a72050526b84c968a6d 100644 (file)
--- a/dss.c
+++ b/dss.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 Andre Noll <maan@systemlinux.org>
+ * Copyright (C) 2008-2011 Andre Noll <maan@tuebingen.mpg.de>
  *
  * Licensed under the GPL v2. For licencing details see COPYING.
  */
 #include "gcc-compat.h"
 #include "cmdline.h"
 #include "log.h"
-#include "string.h"
-#include "error.h"
-#include "fd.h"
+#include "str.h"
+#include "err.h"
+#include "file.h"
 #include "exec.h"
 #include "daemon.h"
-#include "signal.h"
+#include "sig.h"
 #include "df.h"
-#include "time.h"
+#include "tv.h"
 #include "snap.h"
+#include "ipc.h"
 
 /** Command line and config file options. */
 static struct gengetopt_args_info conf;
@@ -44,29 +45,140 @@ static int signal_pipe;
 static pid_t create_pid;
 /** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
 static int create_process_stopped;
-/** Process id of current rm process. */
-static pid_t rm_pid;
+/** How many times in a row the rsync command failed. */
+static int num_consecutive_rsync_errors;
+/** Process id of current pre-remove/rm/post-remove process. */
+static pid_t remove_pid;
 /** When the next snapshot is due. */
-static struct timeval next_snapshot_time;
+static int64_t next_snapshot_time;
+/** When to try to remove something. */
+static struct timeval next_removal_check;
 /** Creation time of the snapshot currently being created. */
 static int64_t current_snapshot_creation_time;
+/** The snapshot currently being removed. */
+struct snapshot *snapshot_currently_being_removed;
 /** Needed by the post-create hook. */
 static char *path_to_last_complete_snapshot;
+static char *name_of_reference_snapshot;
 /** \sa \ref snap.h for details. */
-static unsigned snapshot_creation_status;
+enum hook_status snapshot_creation_status;
 /** \sa \ref snap.h for details. */
-static unsigned snapshot_removal_status;
+enum hook_status snapshot_removal_status;
 
 
 DEFINE_DSS_ERRLIST;
+static const char *hook_status_description[] = {HOOK_STATUS_ARRAY};
 
+/* may be called with ds == NULL. */
+static int disk_space_low(struct disk_space *ds)
+{
+       struct disk_space ds_struct;
+
+       if (!ds) {
+               int ret = get_disk_space(".", &ds_struct);
+               if (ret < 0)
+                       return ret;
+               ds = &ds_struct;
+       }
+       if (conf.min_free_mb_arg)
+               if (ds->free_mb < conf.min_free_mb_arg)
+                       return 1;
+       if (conf.min_free_percent_arg)
+               if (ds->percent_free < conf.min_free_percent_arg)
+                       return 1;
+       if (conf.min_free_percent_inodes_arg)
+               if (ds->percent_free_inodes < conf.min_free_percent_inodes_arg)
+                       return 1;
+       return 0;
+}
+
+static void dump_dss_config(const char *msg)
+{
+       const char dash[] = "-----------------------------";
+       int ret;
+       FILE *log = logfile? logfile : stderr;
+       struct disk_space ds;
+       int64_t now = get_current_time();
+
+       if (conf.loglevel_arg > INFO)
+               return;
+
+       fprintf(log, "%s <%s config> %s\n", dash, msg, dash);
+       fprintf(log, "\n*** disk space ***\n\n");
+       ret = get_disk_space(".", &ds);
+       if (ret >= 0) {
+               DSS_INFO_LOG(("disk space low: %s\n", disk_space_low(&ds)?
+                       "yes" : "no"));
+               log_disk_space(&ds);
+       } else
+               DSS_ERROR_LOG(("can not get free disk space: %s\n",
+                       dss_strerror(-ret)));
+
+       /* we continue on errors from get_disk_space */
+
+       fprintf(log, "\n*** command line and config file options ***\n\n");
+       cmdline_parser_dump(log, &conf);
+       fprintf(log, "\n*** internal state ***\n\n");
+       fprintf(log,
+               "pid: %d\n"
+               "logile: %s\n"
+               "snapshot_currently_being_removed: %s\n"
+               "path_to_last_complete_snapshot: %s\n"
+               "reference_snapshot: %s\n"
+               "snapshot_creation_status: %s\n"
+               "snapshot_removal_status: %s\n"
+               "num_consecutive_rsync_errors: %d\n"
+               ,
+               (int) getpid(),
+               logfile? conf.logfile_arg : "stderr",
+               snapshot_currently_being_removed?
+                       snapshot_currently_being_removed->name : "(none)",
+               path_to_last_complete_snapshot?
+                       path_to_last_complete_snapshot : "(none)",
+               name_of_reference_snapshot?
+                       name_of_reference_snapshot : "(none)",
+               hook_status_description[snapshot_creation_status],
+               hook_status_description[snapshot_removal_status],
+               num_consecutive_rsync_errors
+       );
+       if (create_pid != 0)
+               fprintf(log,
+                       "create_pid: %" PRId32 "\n"
+                       "create process is %sstopped\n"
+                       ,
+                       create_pid,
+                       create_process_stopped? "" : "not "
+               );
+       if (remove_pid != 0)
+               fprintf(log, "remove_pid: %" PRId32 "\n", remove_pid);
+       if (next_snapshot_time != 0)
+               fprintf(log, "next snapshot due in %" PRId64 " seconds\n",
+                       next_snapshot_time - now);
+       if (current_snapshot_creation_time != 0)
+               fprintf(log, "current_snapshot_creation_time: %"
+                       PRId64 " (%" PRId64 " seconds ago)\n",
+                       current_snapshot_creation_time,
+                       now - current_snapshot_creation_time
+               );
+       if (next_removal_check.tv_sec != 0) {
+               fprintf(log, "next removal check: %llu (%llu seconds ago)\n",
+                       (long long unsigned)next_removal_check.tv_sec,
+                       now - (long long unsigned)next_removal_check.tv_sec
+               );
+
+       }
+       fprintf(log, "%s </%s config> %s\n", dash, msg, dash);
+}
 
 /* a litte cpp magic helps to DRY */
 #define COMMANDS \
        COMMAND(ls) \
        COMMAND(create) \
        COMMAND(prune) \
-       COMMAND(run)
+       COMMAND(run) \
+       COMMAND(kill) \
+       COMMAND(reload) \
+
 #define COMMAND(x) static int com_ ##x(void);
 COMMANDS
 #undef COMMAND
@@ -74,12 +186,25 @@ COMMANDS
 static int call_command_handler(void)
 {
        COMMANDS
-       DSS_EMERG_LOG("BUG: did not find command handler\n");
+       DSS_EMERG_LOG(("BUG: did not find command handler\n"));
        return -E_BUG;
 }
 #undef COMMAND
 #undef COMMANDS
 
+static int loglevel = -1;
+static const char *location_file = NULL;
+static int         location_line = -1;
+static const char *location_func = NULL;
+
+void dss_log_set_params(int ll, const char *file, int line, const char *func)
+{
+       loglevel = ll;
+       location_file = file;
+       location_line = line;
+       location_func = func;
+}
+
 /**
  * The log function of dss.
  *
@@ -88,7 +213,7 @@ static int call_command_handler(void)
  *
  * All DSS_XXX_LOG() macros use this function.
  */
-__printf_2_3 void dss_log(int ll, const char* fmt,...)
+__printf_1_2 void dss_log(const char* fmt,...)
 {
        va_list argp;
        FILE *outfd;
@@ -96,7 +221,7 @@ __printf_2_3 void dss_log(int ll, const char* fmt,...)
        time_t t1;
        char str[255] = "";
 
-       if (ll < conf.loglevel_arg)
+       if (loglevel < conf.loglevel_arg)
                return;
        outfd = logfile? logfile : stderr;
        time(&t1);
@@ -104,7 +229,12 @@ __printf_2_3 void dss_log(int ll, const char* fmt,...)
        strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
        fprintf(outfd, "%s ", str);
        if (conf.loglevel_arg <= INFO)
-               fprintf(outfd, "%i: ", ll);
+               fprintf(outfd, "%i: ", loglevel);
+#ifdef DSS_NO_FUNC_NAMES
+       fprintf(outfd, "%s:%d: ", location_file, location_line);
+#else
+       fprintf(outfd, "%s: ", location_func);
+#endif
        va_start(argp, fmt);
        vfprintf(outfd, fmt, argp);
        va_end(argp);
@@ -122,23 +252,45 @@ static __printf_1_2 void dss_msg(const char* fmt,...)
        va_end(argp);
 }
 
-static int disk_space_low(void)
+static char *get_config_file_name(void)
 {
-       struct disk_space ds;
-       int ret = get_disk_space(".", &ds);
+       char *home, *config_file;
+
+       if (conf.config_file_given)
+               return dss_strdup(conf.config_file_arg);
+       home = get_homedir();
+       config_file = make_message("%s/.dssrc", home);
+       free(home);
+       return config_file;
+}
+
+static int send_signal(int sig)
+{
+       pid_t pid;
+       char *config_file = get_config_file_name();
+       int ret = get_dss_pid(config_file, &pid);
 
+       free(config_file);
        if (ret < 0)
                return ret;
-       if (conf.min_free_mb_arg)
-               if (ds.free_mb < conf.min_free_mb_arg)
-                       return 1;
-       if (conf.min_free_percent_arg)
-               if (ds.percent_free < conf.min_free_percent_arg)
-                       return 1;
-       if (conf.min_free_percent_inodes_arg)
-               if (ds.percent_free_inodes < conf.min_free_percent_inodes_arg)
-                       return 1;
-       return 0;
+       if (conf.dry_run_given) {
+               dss_msg("%d\n", (int)pid);
+               return 0;
+       }
+       ret = kill(pid, sig);
+       if (ret < 0)
+               return -ERRNO_TO_DSS_ERROR(errno);
+       return 1;
+}
+
+static int com_kill(void)
+{
+       return send_signal(SIGTERM);
+}
+
+static int com_reload(void)
+{
+       return send_signal(SIGHUP);
 }
 
 static void dss_get_snapshot_list(struct snapshot_list *sl)
@@ -146,19 +298,16 @@ static void dss_get_snapshot_list(struct snapshot_list *sl)
        get_snapshot_list(sl, conf.unit_interval_arg, conf.num_intervals_arg);
 }
 
-static void compute_next_snapshot_time(void)
+static int64_t compute_next_snapshot_time(void)
 {
-       struct timeval now, unit_interval = {.tv_sec = 24 * 3600 * conf.unit_interval_arg},
-               tmp, diff;
-       int64_t x = 0;
+       int64_t x = 0, now = get_current_time(), unit_interval
+               = 24 * 3600 * conf.unit_interval_arg, ret;
        unsigned wanted = desired_number_of_snapshots(0, conf.num_intervals_arg),
                num_complete_snapshots = 0;
-       int i, ret;
+       int i;
        struct snapshot *s = NULL;
        struct snapshot_list sl;
 
-       assert(snapshot_creation_status == SCS_READY);
-       current_snapshot_creation_time = 0;
        dss_get_snapshot_list(&sl);
        FOR_EACH_SNAPSHOT(s, i, &sl) {
                if (!(s->flags & SS_COMPLETE))
@@ -167,42 +316,98 @@ static void compute_next_snapshot_time(void)
                x += s->completion_time - s->creation_time;
        }
        assert(x >= 0);
-       if (num_complete_snapshots)
-               x /= num_complete_snapshots; /* avg time to create one snapshot */
-       x *= wanted; /* time to create all snapshots in interval 0 */
-       tmp.tv_sec = x;
-       tmp.tv_usec = 0;
-       ret = tv_diff(&unit_interval, &tmp, &diff); /* total sleep time per unit interval */
-       if (ret < 0 || !s) /* unit_interval < tmp or no snapshot */
-               goto min_sleep;
-       tv_divide(wanted, &diff, &tmp); /* sleep time betweeen two snapshots */
-       diff.tv_sec = s->completion_time; /* completion time of the the latest snaphot */
-       diff.tv_usec = 0;
-       tv_add(&diff, &tmp, &next_snapshot_time);
-       gettimeofday(&now, NULL);
-       if (tv_diff(&now, &next_snapshot_time, NULL) < 0)
+
+       ret = now;
+       if (num_complete_snapshots == 0)
+               goto out;
+       x /= num_complete_snapshots; /* avg time to create one snapshot */
+       if (unit_interval < x * wanted) /* oops, no sleep at all */
                goto out;
-min_sleep:
-       next_snapshot_time = now;
-       next_snapshot_time.tv_sec += 60;
+       ret = s->completion_time + unit_interval / wanted - x;
 out:
        free_snapshot_list(&sl);
+       return ret;
+}
+
+static inline void invalidate_next_snapshot_time(void)
+{
+       next_snapshot_time = 0;
+}
+
+static inline int next_snapshot_time_is_valid(void)
+{
+       return next_snapshot_time != 0;
+}
+
+static int next_snapshot_is_due(void)
+{
+       int64_t now = get_current_time();
+
+       if (!next_snapshot_time_is_valid())
+               next_snapshot_time = compute_next_snapshot_time();
+       if (next_snapshot_time <= now) {
+               DSS_DEBUG_LOG(("next snapshot: now\n"));
+               return 1;
+       }
+       DSS_DEBUG_LOG(("next snapshot due in %" PRId64 " seconds\n",
+               next_snapshot_time - now));
+       return 0;
 }
 
+static void pre_create_hook(void)
+{
+       assert(snapshot_creation_status == HS_READY);
+       /* make sure that the next snapshot time will be recomputed */
+       invalidate_next_snapshot_time();
+       DSS_DEBUG_LOG(("executing %s\n", conf.pre_create_hook_arg));
+       dss_exec_cmdline_pid(&create_pid, conf.pre_create_hook_arg);
+       snapshot_creation_status = HS_PRE_RUNNING;
+}
+
+static void pre_remove_hook(struct snapshot *s, const char *why)
+{
+       char *cmd;
+
+       if (!s)
+               return;
+       DSS_DEBUG_LOG(("%s snapshot %s\n", why, s->name));
+       assert(snapshot_removal_status == HS_READY);
+       assert(remove_pid == 0);
+       assert(!snapshot_currently_being_removed);
+
+       snapshot_currently_being_removed = dss_malloc(sizeof(struct snapshot));
+       *snapshot_currently_being_removed = *s;
+       snapshot_currently_being_removed->name = dss_strdup(s->name);
+
+       cmd = make_message("%s %s/%s", conf.pre_remove_hook_arg,
+               conf.dest_dir_arg, s->name);
+       DSS_DEBUG_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&remove_pid, cmd);
+       free(cmd);
+       snapshot_removal_status = HS_PRE_RUNNING;
+}
 
-static int remove_snapshot(struct snapshot *s, char *why)
+static int exec_rm(void)
 {
-       int fds[3] = {0, 0, 0};
-       assert(!rm_pid);
+       struct snapshot *s = snapshot_currently_being_removed;
        char *new_name = being_deleted_name(s);
-       int ret = dss_rename(s->name, new_name);
-       char *argv[] = {"rm", "-rf", new_name, NULL};
+       char *argv[4];
+       int ret;
 
+       argv[0] = "rm";
+       argv[1] = "-rf";
+       argv[2] = new_name;
+       argv[3] = NULL;
+
+       assert(snapshot_removal_status == HS_PRE_SUCCESS);
+       assert(remove_pid == 0);
+
+       DSS_NOTICE_LOG(("removing %s (interval = %i)\n", s->name, s->interval));
+       ret = dss_rename(s->name, new_name);
        if (ret < 0)
                goto out;
-       DSS_NOTICE_LOG("removing %s snapshot %s (interval = %i)\n",
-               why, s->name, s->interval);
-       ret = dss_exec(&rm_pid, argv[0], argv, fds);
+       dss_exec(&remove_pid, argv[0], argv);
+       snapshot_removal_status = HS_RUNNING;
 out:
        free(new_name);
        return ret;
@@ -213,16 +418,57 @@ static int snapshot_is_being_created(struct snapshot *s)
        return s->creation_time == current_snapshot_creation_time;
 }
 
+static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
+{
+       struct snapshot *s;
+       int i;
+
+       DSS_DEBUG_LOG(("looking for orphaned snapshots\n"));
+       FOR_EACH_SNAPSHOT(s, i, sl) {
+               if (snapshot_is_being_created(s))
+                       continue;
+               /*
+                * We know that no rm is currently running, so if s is marked
+                * as being deleted, a previously started rm must have failed.
+                */
+               if (s->flags & SS_BEING_DELETED)
+                       return s;
+
+               if (s->flags & SS_COMPLETE) /* good snapshot */
+                       continue;
+               /*
+                * This snapshot is incomplete and it is not the snapshot
+                * currently being created. However, we must not remove it if
+                * rsync is about to be restarted. As only the newest snapshot
+                * can be restarted, this snapshot is orphaned if it is not the
+                * newest snapshot or if we are not about to restart rsync.
+                */
+               if (get_newest_snapshot(sl) != s)
+                       return s;
+               if (snapshot_creation_status != HS_NEEDS_RESTART)
+                       return s;
+       }
+       /* no orphaned snapshots */
+       return NULL;
+}
+
+static int is_reference_snapshot(struct snapshot *s)
+{
+       if (!name_of_reference_snapshot)
+               return 0;
+       return strcmp(s->name, name_of_reference_snapshot)? 0 : 1;
+}
+
 /*
  * return: 0: no redundant snapshots, 1: rm process started, negative: error
  */
-static int remove_redundant_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
 {
-       int ret, i, interval;
+       int i, interval;
        struct snapshot *s;
        unsigned missing = 0;
 
-       DSS_DEBUG_LOG("looking for intervals containing too many snapshots\n");
+       DSS_DEBUG_LOG(("looking for intervals containing too many snapshots\n"));
        for (interval = conf.num_intervals_arg - 1; interval >= 0; interval--) {
                unsigned keep = desired_number_of_snapshots(interval, conf.num_intervals_arg);
                unsigned num = sl->interval_count[interval];
@@ -231,8 +477,6 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
 
                if (keep >= num)
                        missing += keep - num;
-//             DSS_DEBUG_LOG("interval %i: keep: %u, have: %u, missing: %u\n",
-//                     interval, keep, num, missing);
                if (keep + missing >= num)
                        continue;
                /* redundant snapshot in this interval, pick snapshot with lowest score */
@@ -241,7 +485,8 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
 
                        if (snapshot_is_being_created(s))
                                continue;
-                       //DSS_DEBUG_LOG("checking %s\n", s->name);
+                       if (is_reference_snapshot(s))
+                               continue;
                        if (s->interval > interval) {
                                prev = s;
                                continue;
@@ -257,7 +502,6 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
                        /* check if s is a better victim */
                        this_score = s->creation_time - prev->creation_time;
                        assert(this_score >= 0);
-                       //DSS_DEBUG_LOG("%s: score %lli\n", s->name, (long long)score);
                        if (this_score < score) {
                                score = this_score;
                                victim = s;
@@ -265,161 +509,212 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
                        prev = s;
                }
                assert(victim);
-               if (conf.dry_run_given) {
-                       dss_msg("%s would be removed (interval = %i)\n",
-                               victim->name, victim->interval);
-                       continue;
-               }
-               ret = remove_snapshot(victim, "redundant");
-               return ret < 0? ret : 1;
+               return victim;
        }
-       return 0;
+       return NULL;
 }
 
-static int remove_outdated_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
 {
-       int i, ret;
+       int i;
        struct snapshot *s;
 
-       DSS_DEBUG_LOG("looking for snapshots belonging to intervals greater than %d\n",
-               conf.num_intervals_arg);
+       DSS_DEBUG_LOG(("looking for snapshots belonging to intervals >= %d\n",
+               conf.num_intervals_arg));
        FOR_EACH_SNAPSHOT(s, i, sl) {
                if (snapshot_is_being_created(s))
                        continue;
-               if (s->interval <= conf.num_intervals_arg)
+               if (is_reference_snapshot(s))
                        continue;
-               if (conf.dry_run_given) {
-                       dss_msg("%s would be removed (interval = %i)\n",
-                               s->name, s->interval);
+               if (s->interval < conf.num_intervals_arg)
                        continue;
-               }
-               ret = remove_snapshot(s, "outdated");
-               if (ret < 0)
-                       return ret;
-               return 1;
+               return s;
        }
-       return 0;
+       return NULL;
 }
 
-static int remove_oldest_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
 {
-       struct snapshot *s = get_oldest_snapshot(sl);
+       int i, num_complete;
+       struct snapshot *s, *ref = NULL;
 
-       if (!s) /* no snapshot found */
-               return 0;
-       DSS_INFO_LOG("oldest snapshot: %s\n", s->name);
-       if (snapshot_is_being_created(s))
-               return 0;
-       return remove_snapshot(s, "oldest");
+       num_complete = num_complete_snapshots(sl);
+       if (num_complete <= conf.min_complete_arg)
+               return NULL;
+       FOR_EACH_SNAPSHOT(s, i, sl) {
+               if (snapshot_is_being_created(s))
+                       continue;
+               if (is_reference_snapshot(s)) { /* avoid this one */
+                       ref = s;
+                       continue;
+               }
+               DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
+               return s;
+       }
+       assert(ref);
+       DSS_WARNING_LOG(("removing reference snapshot %s\n", ref->name));
+       return ref;
 }
 
 static int rename_incomplete_snapshot(int64_t start)
 {
        char *old_name;
        int ret;
-
+       int64_t now;
+
+       /*
+        * We don't want the dss_rename() below to fail with EEXIST because the
+        * last complete snapshot was created (and completed) in the same
+        * second as this one.
+        */
+       while ((now = get_current_time()) == start)
+               sleep(1);
        free(path_to_last_complete_snapshot);
-       ret = complete_name(start, get_current_time(),
-               &path_to_last_complete_snapshot);
+       ret = complete_name(start, now, &path_to_last_complete_snapshot);
        if (ret < 0)
                return ret;
        old_name = incomplete_name(start);
        ret = dss_rename(old_name, path_to_last_complete_snapshot);
        if (ret >= 0)
-               DSS_NOTICE_LOG("%s -> %s\n", old_name,
-                       path_to_last_complete_snapshot);
+               DSS_NOTICE_LOG(("%s -> %s\n", old_name,
+                       path_to_last_complete_snapshot));
        free(old_name);
        return ret;
 }
 
-static int try_to_free_disk_space(int low_disk_space)
+static int try_to_free_disk_space(void)
 {
        int ret;
        struct snapshot_list sl;
+       struct snapshot *victim;
+       struct timeval now;
+       const char *why;
+       int low_disk_space;
 
-       if (!low_disk_space && conf.keep_redundant_given)
+       ret = disk_space_low(NULL);
+       if (ret < 0)
+               return ret;
+       low_disk_space = ret;
+       gettimeofday(&now, NULL);
+       if (tv_diff(&next_removal_check, &now, NULL) > 0)
                return 0;
+       if (!low_disk_space) {
+               if (conf.keep_redundant_given)
+                       return 0;
+               if (snapshot_creation_status != HS_READY)
+                       return 0;
+               if (next_snapshot_is_due())
+                       return 0;
+       }
+       /*
+        * Idle and --keep_redundant not given, or low disk space. Look at
+        * existing snapshots.
+        */
        dss_get_snapshot_list(&sl);
-       ret = remove_outdated_snapshot(&sl);
-       if (ret) /* error, or we are removing something */
-               goto out;
-       /* no outdated snapshot */
-       ret = remove_redundant_snapshot(&sl);
-       if (ret)
-               goto out;
        ret = 0;
-       if (!low_disk_space)
+       /*
+        * Don't remove anything if there is free space and we have fewer
+        * snapshots than configured, plus one. This way there is always one
+        * snapshot that can be recycled.
+        */
+       if (!low_disk_space && sl.num_snapshots <= 1 << conf.num_intervals_arg)
                goto out;
-       DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
-       ret = remove_oldest_snapshot(&sl);
-       if (ret)
+       why = "outdated";
+       victim = find_outdated_snapshot(&sl);
+       if (victim)
+               goto remove;
+       why = "redundant";
+       victim = find_redundant_snapshot(&sl);
+       if (victim)
+               goto remove;
+       why = "orphaned";
+       victim = find_orphaned_snapshot(&sl);
+       if (victim)
+               goto remove;
+       /* try harder only if disk space is low */
+       if (!low_disk_space)
                goto out;
-       DSS_CRIT_LOG("uhuhu: not enough disk space for a single snapshot\n");
+       DSS_WARNING_LOG(("disk space low and nothing obvious to remove\n"));
+       victim = find_oldest_removable_snapshot(&sl);
+       if (victim)
+               goto remove;
+       DSS_CRIT_LOG(("uhuhu: disk space low and nothing to remove\n"));
        ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
+       goto out;
+remove:
+       pre_remove_hook(victim, why);
 out:
        free_snapshot_list(&sl);
        return ret;
 }
 
-static int pre_create_hook(void)
+static void post_create_hook(void)
 {
-       int ret, fds[3] = {0, 0, 0};
-
-       if (!conf.pre_create_hook_given) {
-               snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
-               return 0;
-       }
-       DSS_DEBUG_LOG("executing %s\n", conf.pre_create_hook_arg);
-       ret = dss_exec_cmdline_pid(&create_pid,
-               conf.pre_create_hook_arg, fds);
-       if (ret < 0)
-               return ret;
-       snapshot_creation_status = SCS_PRE_HOOK_RUNNING;
-       return ret;
+       char *cmd = make_message("%s %s/%s", conf.post_create_hook_arg,
+               conf.dest_dir_arg, path_to_last_complete_snapshot);
+       DSS_NOTICE_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&create_pid, cmd);
+       free(cmd);
+       snapshot_creation_status = HS_POST_RUNNING;
 }
 
-static int post_create_hook(void)
+static void post_remove_hook(void)
 {
-       int ret, fds[3] = {0, 0, 0};
        char *cmd;
+       struct snapshot *s = snapshot_currently_being_removed;
 
-       if (!conf.post_create_hook_given) {
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
-               return 0;
-       }
-       cmd = make_message("%s %s/%s", conf.post_create_hook_arg,
-               conf.dest_dir_arg, path_to_last_complete_snapshot);
-       DSS_NOTICE_LOG("executing %s\n", cmd);
-       ret = dss_exec_cmdline_pid(&create_pid, cmd, fds);
+       assert(s);
+
+       cmd = make_message("%s %s/%s", conf.post_remove_hook_arg,
+               conf.dest_dir_arg, s->name);
+       DSS_NOTICE_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&remove_pid, cmd);
        free(cmd);
-       if (ret < 0)
-               return ret;
-       snapshot_creation_status = SCS_POST_HOOK_RUNNING;
-       return ret;
+       snapshot_removal_status = HS_POST_RUNNING;
 }
 
-static void kill_process(pid_t pid)
+static void dss_kill(pid_t pid, int sig, const char *msg)
 {
-       if (!pid)
+       const char *signame, *process_name;
+
+       if (pid == 0)
                return;
-       DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
-       kill(pid, SIGTERM);
+       switch (sig) {
+       case SIGTERM: signame = "TERM"; break;
+       case SIGSTOP: signame = "STOP"; break;
+       case SIGCONT: signame = "CONT"; break;
+       default: signame = "????";
+       }
+
+       if (pid == create_pid)
+               process_name = "create";
+       else if (pid == remove_pid)
+               process_name = "remove";
+       else process_name = "??????";
+
+       if (msg)
+               DSS_INFO_LOG(("%s\n", msg));
+       DSS_DEBUG_LOG(("sending signal %d (%s) to pid %d (%s process)\n",
+               sig, signame, (int)pid, process_name));
+       if (kill(pid, sig) >= 0)
+               return;
+       DSS_INFO_LOG(("failed to send signal %d (%s) to pid %d (%s process)\n",
+               sig, signame, (int)pid, process_name));
 }
 
 static void stop_create_process(void)
 {
-       if (!create_pid || create_process_stopped)
+       if (create_process_stopped)
                return;
-       kill(SIGSTOP, create_pid);
+       dss_kill(create_pid, SIGSTOP, "suspending create process");
        create_process_stopped = 1;
 }
 
 static void restart_create_process(void)
 {
-       if (!create_pid || !create_process_stopped)
+       if (!create_process_stopped)
                return;
-       kill (SIGCONT, create_pid);
+       dss_kill(create_pid, SIGCONT, "resuming create process");
        create_process_stopped = 0;
 }
 
@@ -429,20 +724,20 @@ static void restart_create_process(void)
 static void log_termination_msg(pid_t pid, int status)
 {
        if (WIFEXITED(status))
-               DSS_INFO_LOG("child %i exited. Exit status: %i\n", (int)pid,
-                       WEXITSTATUS(status));
+               DSS_INFO_LOG(("child %i exited. Exit status: %i\n", (int)pid,
+                       WEXITSTATUS(status)));
        else if (WIFSIGNALED(status))
-               DSS_NOTICE_LOG("child %i was killed by signal %i\n", (int)pid,
-                       WTERMSIG(status));
+               DSS_NOTICE_LOG(("child %i was killed by signal %i\n", (int)pid,
+                       WTERMSIG(status)));
        else
-               DSS_WARNING_LOG("child %i terminated abormally\n", (int)pid);
+               DSS_WARNING_LOG(("child %i terminated abormally\n", (int)pid));
 }
 
 static int wait_for_process(pid_t pid, int *status)
 {
        int ret;
 
-       DSS_DEBUG_LOG("Waiting for process %d to terminate\n", (int)pid);
+       DSS_DEBUG_LOG(("Waiting for process %d to terminate\n", (int)pid));
        for (;;) {
                fd_set rfds;
 
@@ -464,33 +759,89 @@ static int wait_for_process(pid_t pid, int *status)
                        }
                }
                /* SIGINT or SIGTERM */
-               DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
-               kill(pid, SIGTERM);
+               dss_kill(pid, SIGTERM, "killing child process");
        }
        if (ret < 0)
-               DSS_ERROR_LOG("failed to wait for process %d\n", (int)pid);
+               DSS_ERROR_LOG(("failed to wait for process %d\n", (int)pid));
        else
                log_termination_msg(pid, *status);
        return ret;
 }
 
+static void handle_pre_remove_exit(int status)
+{
+       if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+               snapshot_removal_status = HS_READY;
+               gettimeofday(&next_removal_check, NULL);
+               next_removal_check.tv_sec += 60;
+               return;
+       }
+       snapshot_removal_status = HS_PRE_SUCCESS;
+}
+
 static int handle_rm_exit(int status)
 {
-       rm_pid = 0;
-       if (!WIFEXITED(status))
+       if (!WIFEXITED(status)) {
+               snapshot_removal_status = HS_READY;
                return -E_INVOLUNTARY_EXIT;
-       if (WEXITSTATUS(status))
+       }
+       if (WEXITSTATUS(status)) {
+               snapshot_removal_status = HS_READY;
                return -E_BAD_EXIT_CODE;
+       }
+       snapshot_removal_status = HS_SUCCESS;
        return 1;
 }
 
-static int wait_for_rm_process(void)
+static void handle_post_remove_exit(void)
 {
-       int status, ret = wait_for_process(rm_pid, &status);
+       snapshot_removal_status = HS_READY;
+}
 
+static int handle_remove_exit(int status)
+{
+       int ret;
+       struct snapshot *s = snapshot_currently_being_removed;
+
+       assert(s);
+       switch (snapshot_removal_status) {
+       case HS_PRE_RUNNING:
+               handle_pre_remove_exit(status);
+               ret = 1;
+               break;
+       case HS_RUNNING:
+               ret = handle_rm_exit(status);
+               break;
+       case HS_POST_RUNNING:
+               handle_post_remove_exit();
+               ret = 1;
+               break;
+       default:
+               ret = -E_BUG;
+       }
+       if (snapshot_removal_status == HS_READY) {
+               free(s->name);
+               free(s);
+               snapshot_currently_being_removed = NULL;
+       }
+       remove_pid = 0;
+       return ret;
+}
+
+static int wait_for_remove_process(void)
+{
+       int status, ret;
+
+       assert(remove_pid);
+       assert(
+               snapshot_removal_status == HS_PRE_RUNNING ||
+               snapshot_removal_status == HS_RUNNING ||
+               snapshot_removal_status == HS_POST_RUNNING
+       );
+       ret = wait_for_process(remove_pid, &status);
        if (ret < 0)
                return ret;
-       return handle_rm_exit(status);
+       return handle_remove_exit(status);
 }
 
 static int handle_rsync_exit(int status)
@@ -498,35 +849,38 @@ static int handle_rsync_exit(int status)
        int es, ret;
 
        if (!WIFEXITED(status)) {
-               DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)create_pid);
+               DSS_ERROR_LOG(("rsync process %d died involuntary\n", (int)create_pid));
                ret = -E_INVOLUNTARY_EXIT;
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
+               snapshot_creation_status = HS_READY;
                goto out;
        }
        es = WEXITSTATUS(status);
-       if (es == 13) { /* Errors with program diagnostics */
-               DSS_WARNING_LOG("rsync process %d returned %d -- restarting\n",
-                       (int)create_pid, es);
-               snapshot_creation_status = SCS_RSYNC_NEEDS_RESTART;
-               gettimeofday(&next_snapshot_time, NULL);
-               next_snapshot_time.tv_sec += 60;
+       /*
+        * Restart rsync on non-fatal errors:
+        * 24: Partial transfer due to vanished source files
+        */
+       if (es != 0 && es != 24) {
+               DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
+                       es, ++num_consecutive_rsync_errors));
+               if (num_consecutive_rsync_errors > conf.max_rsync_errors_arg) {
+                       ret = -E_TOO_MANY_RSYNC_ERRORS;
+                       snapshot_creation_status = HS_READY;
+                       goto out;
+               }
+               DSS_WARNING_LOG(("restarting rsync process\n"));
+               snapshot_creation_status = HS_NEEDS_RESTART;
+               next_snapshot_time = get_current_time() + 60;
                ret = 1;
                goto out;
        }
-       if (es != 0 && es != 23 && es != 24) {
-               DSS_ERROR_LOG("rsync process %d returned %d\n", (int)create_pid, es);
-               ret = -E_BAD_EXIT_CODE;
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
-               goto out;
-       }
+       num_consecutive_rsync_errors = 0;
        ret = rename_incomplete_snapshot(current_snapshot_creation_time);
        if (ret < 0)
                goto out;
-       snapshot_creation_status = SCS_RSYNC_SUCCESS;
+       snapshot_creation_status = HS_SUCCESS;
+       free(name_of_reference_snapshot);
+       name_of_reference_snapshot = NULL;
 out:
-       create_pid = 0;
        create_process_stopped = 0;
        return ret;
 }
@@ -537,29 +891,27 @@ static int handle_pre_create_hook_exit(int status)
        static int warn_count;
 
        if (!WIFEXITED(status)) {
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
+               snapshot_creation_status = HS_READY;
                ret = -E_INVOLUNTARY_EXIT;
                goto out;
        }
        es = WEXITSTATUS(status);
        if (es) {
                if (!warn_count--) {
-                       DSS_NOTICE_LOG("pre_create_hook %s returned %d\n",
-                               conf.pre_create_hook_arg, es);
-                       DSS_NOTICE_LOG("deferring snapshot creation...\n");
+                       DSS_NOTICE_LOG(("pre_create_hook %s returned %d\n",
+                               conf.pre_create_hook_arg, es));
+                       DSS_NOTICE_LOG(("deferring snapshot creation...\n"));
                        warn_count = 60; /* warn only once per hour */
                }
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
+               next_snapshot_time = get_current_time() + 60;
+               snapshot_creation_status = HS_READY;
                ret = 0;
                goto out;
        }
        warn_count = 0;
-       snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
+       snapshot_creation_status = HS_PRE_SUCCESS;
        ret = 1;
 out:
-       create_pid = 0;
        return ret;
 }
 
@@ -573,76 +925,81 @@ static int handle_sigchld(void)
 
        if (pid == create_pid) {
                switch (snapshot_creation_status) {
-               case SCS_PRE_HOOK_RUNNING:
-                       return handle_pre_create_hook_exit(status);
-               case SCS_RSYNC_RUNNING:
-                       return handle_rsync_exit(status);
-               case SCS_POST_HOOK_RUNNING:
-                       snapshot_creation_status = SCS_READY;
-                       compute_next_snapshot_time();
-                       return 1;
+               case HS_PRE_RUNNING:
+                       ret = handle_pre_create_hook_exit(status);
+                       break;
+               case HS_RUNNING:
+                       ret = handle_rsync_exit(status);
+                       break;
+               case HS_POST_RUNNING:
+                       snapshot_creation_status = HS_READY;
+                       ret = 1;
+                       break;
                default:
-                       DSS_EMERG_LOG("BUG: create can't die in status %d\n",
-                               snapshot_creation_status);
+                       DSS_EMERG_LOG(("BUG: create can't die in status %d\n",
+                               snapshot_creation_status));
                        return -E_BUG;
                }
+               create_pid = 0;
+               return ret;
+       }
+       if (pid == remove_pid) {
+               ret = handle_remove_exit(status);
+               if (ret < 0)
+                       return ret;
+               return ret;
        }
-       if (pid == rm_pid)
-               return handle_rm_exit(status);
-       DSS_EMERG_LOG("BUG: unknown process %d died\n", (int)pid);
+       DSS_EMERG_LOG(("BUG: unknown process %d died\n", (int)pid));
        return -E_BUG;
 }
 
 static int check_config(void)
 {
        if (conf.unit_interval_arg <= 0) {
-               DSS_ERROR_LOG("bad unit interval: %i\n", conf.unit_interval_arg);
+               DSS_ERROR_LOG(("bad unit interval: %i\n", conf.unit_interval_arg));
                return -E_INVALID_NUMBER;
        }
-       DSS_DEBUG_LOG("unit interval: %i day(s)\n", conf.unit_interval_arg);
-       if (conf.num_intervals_arg <= 0) {
-               DSS_ERROR_LOG("bad number of intervals  %i\n", conf.num_intervals_arg);
+       DSS_DEBUG_LOG(("unit interval: %i day(s)\n", conf.unit_interval_arg));
+       if (conf.num_intervals_arg <= 0 || conf.num_intervals_arg > 30) {
+               DSS_ERROR_LOG(("bad number of intervals: %i\n",
+                       conf.num_intervals_arg));
                return -E_INVALID_NUMBER;
        }
-       DSS_DEBUG_LOG("number of intervals: %i\n", conf.num_intervals_arg);
+       DSS_DEBUG_LOG(("number of intervals: %i\n", conf.num_intervals_arg));
        return 1;
 }
 
+/*
+ * Returns < 0 on errors, 0 if no config file is given and > 0 if the config
+ * file was read successfully.
+ */
 static int parse_config_file(int override)
 {
-       int ret;
-       char *config_file;
+       int ret, config_file_exists;
+       char *config_file = get_config_file_name();
        struct stat statbuf;
        char *old_logfile_arg = NULL;
        int old_daemon_given = 0;
 
-       if (conf.config_file_given)
-               config_file = dss_strdup(conf.config_file_arg);
-       else {
-               char *home = get_homedir();
-               config_file = make_message("%s/.dssrc", home);
-               free(home);
-       }
        if (override) { /* SIGHUP */
                if (conf.logfile_given)
                        old_logfile_arg = dss_strdup(conf.logfile_arg);
                old_daemon_given = conf.daemon_given;
        }
 
-       ret = stat(config_file, &statbuf);
-       if (ret && conf.config_file_given) {
+       config_file_exists = !stat(config_file, &statbuf);
+       if (!config_file_exists && conf.config_file_given) {
                ret = -ERRNO_TO_DSS_ERROR(errno);
-               DSS_ERROR_LOG("failed to stat config file %s\n", config_file);
+               DSS_ERROR_LOG(("failed to stat config file %s\n", config_file));
                goto out;
        }
-       if (!ret) {
-               struct cmdline_parser_params params = {
-                       .override = override,
-                       .initialize = 0,
-                       .check_required = 1,
-                       .check_ambiguity = 0,
-                       .print_errors = 1
-               };
+       if (config_file_exists) {
+               struct cmdline_parser_params params;
+               params.override = override;
+               params.initialize = 0;
+               params.check_required = 1;
+               params.check_ambiguity = 0;
+               params.print_errors = 1;
                if (override) { /* invalidate all rsync options */
                        int i;
 
@@ -669,22 +1026,22 @@ static int parse_config_file(int override)
                        conf.logfile_given = 1;
                }
        }
-       if (conf.logfile_given) {
+       if (conf.logfile_given && conf.run_given && conf.daemon_given) {
                logfile = open_log(conf.logfile_arg);
                log_welcome(conf.loglevel_arg);
        }
-       DSS_DEBUG_LOG("loglevel: %d\n", conf.loglevel_arg);
-//     cmdline_parser_dump(logfile? logfile : stdout, &conf);
+       DSS_DEBUG_LOG(("loglevel: %d\n", conf.loglevel_arg));
+       ret = config_file_exists;
 out:
        free(config_file);
        if (ret < 0)
-               DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
+               DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
        return ret;
 }
 
 static int change_to_dest_dir(void)
 {
-       DSS_INFO_LOG("changing cwd to %s\n", conf.dest_dir_arg);
+       DSS_INFO_LOG(("changing cwd to %s\n", conf.dest_dir_arg));
        return dss_chdir(conf.dest_dir_arg);
 }
 
@@ -692,13 +1049,23 @@ static int handle_sighup(void)
 {
        int ret;
 
-       DSS_NOTICE_LOG("SIGHUP\n");
+       DSS_NOTICE_LOG(("SIGHUP, re-reading config\n"));
+       dump_dss_config("old");
        ret = parse_config_file(1);
        if (ret < 0)
                return ret;
+       dump_dss_config("reloaded");
+       invalidate_next_snapshot_time();
        return change_to_dest_dir();
 }
 
+static void kill_children(void)
+{
+       restart_create_process();
+       dss_kill(create_pid, SIGTERM, NULL);
+       dss_kill(remove_pid, SIGTERM, NULL);
+}
+
 static int handle_signal(void)
 {
        int sig, ret = next_signal();
@@ -709,9 +1076,7 @@ static int handle_signal(void)
        switch (sig) {
        case SIGINT:
        case SIGTERM:
-               restart_create_process();
-               kill_process(create_pid);
-               kill_process(rm_pid);
+               kill_children();
                ret = -E_SIGNAL;
                break;
        case SIGHUP:
@@ -723,7 +1088,7 @@ static int handle_signal(void)
        }
 out:
        if (ret < 0)
-               DSS_ERROR_LOG("%s\n", dss_strerror(-ret));
+               DSS_ERROR_LOG(("%s\n", dss_strerror(-ret)));
        return ret;
 }
 
@@ -742,28 +1107,83 @@ static int use_rsync_locally(char *logname)
        return 1;
 }
 
+static int rename_resume_snap(int64_t creation_time)
+{
+       struct snapshot_list sl;
+       struct snapshot *s = NULL;
+       char *new_name = incomplete_name(creation_time);
+       int ret;
+       const char *why;
+
+       sl.num_snapshots = 0;
+
+       ret = 0;
+       if (conf.no_resume_given)
+               goto out;
+       dss_get_snapshot_list(&sl);
+       /*
+        * Snapshot recycling: We first look at the newest snapshot. If this
+        * snapshot happens to be incomplete, the last rsync process was
+        * aborted and we reuse this one. Otherwise we look at snapshots which
+        * could be removed (outdated and redundant snapshots) as candidates
+        * for recycling. If no outdated/redundant snapshot exists, we check if
+        * there is an orphaned snapshot, which likely is useless anyway.
+        *
+        * Only if no existing snapshot is suitable for recycling, we bite the
+        * bullet and create a new one.
+        */
+       s = get_newest_snapshot(&sl);
+       if (!s) /* no snapshots at all */
+               goto out;
+       /* re-use last snapshot if it is incomplete */
+       why = "aborted";
+       if ((s->flags & SS_COMPLETE) == 0)
+               goto out;
+       why = "outdated";
+       s = find_outdated_snapshot(&sl);
+       if (s)
+               goto out;
+       why = "redundant";
+       s = find_redundant_snapshot(&sl);
+       if (s)
+               goto out;
+       why = "orphaned";
+       s = find_orphaned_snapshot(&sl);
+out:
+       if (s) {
+               DSS_INFO_LOG(("reusing %s snapshot %s\n", why, s->name));
+               ret = dss_rename(s->name, new_name);
+       }
+       if (ret >= 0)
+               DSS_NOTICE_LOG(("creating new snapshot %s\n", new_name));
+       free(new_name);
+       free_snapshot_list(&sl);
+       return ret;
+}
+
 static void create_rsync_argv(char ***argv, int64_t *num)
 {
-       char *logname, *newest;
+       char *logname;
        int i = 0, j;
        struct snapshot_list sl;
 
        dss_get_snapshot_list(&sl);
-       newest = name_of_newest_complete_snapshot(&sl);
+       assert(!name_of_reference_snapshot);
+       name_of_reference_snapshot = name_of_newest_complete_snapshot(&sl);
        free_snapshot_list(&sl);
 
        *argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
        (*argv)[i++] = dss_strdup("rsync");
-       (*argv)[i++] = dss_strdup("-aq");
+       (*argv)[i++] = dss_strdup("-a");
        (*argv)[i++] = dss_strdup("--delete");
        for (j = 0; j < conf.rsync_option_given; j++)
                (*argv)[i++] = dss_strdup(conf.rsync_option_arg[j]);
-       if (newest) {
-               DSS_INFO_LOG("using %s as reference snapshot\n", newest);
-               (*argv)[i++] = make_message("--link-dest=../%s", newest);
-               free(newest);
+       if (name_of_reference_snapshot) {
+               DSS_INFO_LOG(("using %s as reference\n", name_of_reference_snapshot));
+               (*argv)[i++] = make_message("--link-dest=../%s",
+                       name_of_reference_snapshot);
        } else
-               DSS_INFO_LOG("no previous snapshot found\n");
+               DSS_INFO_LOG(("no suitable reference snapshot found\n"));
        logname = dss_logname();
        if (use_rsync_locally(logname))
                (*argv)[i++] = dss_strdup(conf.source_dir_arg);
@@ -776,7 +1196,7 @@ static void create_rsync_argv(char ***argv, int64_t *num)
        (*argv)[i++] = incomplete_name(*num);
        (*argv)[i++] = NULL;
        for (j = 0; j < i; j++)
-               DSS_DEBUG_LOG("argv[%d] = %s\n", j, (*argv)[j]);
+               DSS_DEBUG_LOG(("argv[%d] = %s\n", j, (*argv)[j]));
 }
 
 static void free_rsync_argv(char **argv)
@@ -792,16 +1212,13 @@ static void free_rsync_argv(char **argv)
 
 static int create_snapshot(char **argv)
 {
-       int ret, fds[3] = {0, 0, 0};
-       char *name;
+       int ret;
 
-       name = incomplete_name(current_snapshot_creation_time);
-       DSS_NOTICE_LOG("creating new snapshot %s\n", name);
-       free(name);
-       ret = dss_exec(&create_pid, argv[0], argv, fds);
+       ret = rename_resume_snap(current_snapshot_creation_time);
        if (ret < 0)
                return ret;
-       snapshot_creation_status = SCS_RSYNC_RUNNING;
+       dss_exec(&create_pid, argv[0], argv);
+       snapshot_creation_status = HS_RUNNING;
        return ret;
 }
 
@@ -814,11 +1231,10 @@ static int select_loop(void)
 
        for (;;) {
                fd_set rfds;
-               int low_disk_space;
-               struct timeval now, *tvp;
+               struct timeval *tvp;
 
-               if (rm_pid)
-                       tvp = NULL; /* sleep until rm process dies */
+               if (remove_pid)
+                       tvp = NULL; /* sleep until rm hook/process dies */
                else { /* sleep one minute */
                        tv.tv_sec = 60;
                        tv.tv_usec = 0;
@@ -826,7 +1242,6 @@ static int select_loop(void)
                }
                FD_ZERO(&rfds);
                FD_SET(signal_pipe, &rfds);
-               DSS_DEBUG_LOG("tvp: %p, tv_sec : %lu\n", tvp, (long unsigned) tv.tv_sec);
                ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
                if (ret < 0)
                        goto out;
@@ -835,51 +1250,54 @@ static int select_loop(void)
                        if (ret < 0)
                                goto out;
                }
-               if (rm_pid)
+               if (remove_pid)
                        continue;
-               ret = disk_space_low();
-               if (ret < 0)
-                       goto out;
-               low_disk_space = ret;
-               ret = try_to_free_disk_space(low_disk_space);
+               if (snapshot_removal_status == HS_PRE_SUCCESS) {
+                       ret = exec_rm();
+                       if (ret < 0)
+                               goto out;
+                       continue;
+               }
+               if (snapshot_removal_status == HS_SUCCESS) {
+                       post_remove_hook();
+                       continue;
+               }
+               ret = try_to_free_disk_space();
                if (ret < 0)
                        goto out;
-               if (rm_pid) {
+               if (snapshot_removal_status != HS_READY) {
                        stop_create_process();
                        continue;
                }
                restart_create_process();
-               gettimeofday(&now, NULL);
-               if (tv_diff(&next_snapshot_time, &now, NULL) > 0)
-                       continue;
                switch (snapshot_creation_status) {
-               case SCS_READY:
-                       ret = pre_create_hook();
-                       if (ret < 0)
-                               goto out;
+               case HS_READY:
+                       if (!next_snapshot_is_due())
+                               continue;
+                       pre_create_hook();
                        continue;
-               case SCS_PRE_HOOK_RUNNING:
+               case HS_PRE_RUNNING:
+               case HS_RUNNING:
+               case HS_POST_RUNNING:
                        continue;
-               case SCS_RSYNC_NEEDS_RESTART:
+               case HS_PRE_SUCCESS:
+                       if (!name_of_reference_snapshot) {
+                               free_rsync_argv(rsync_argv);
+                               create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
+                       }
                        ret = create_snapshot(rsync_argv);
                        if (ret < 0)
                                goto out;
                        continue;
-               case SCS_PRE_HOOK_SUCCESS:
-                       free_rsync_argv(rsync_argv);
-                       create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
+               case HS_NEEDS_RESTART:
+                       if (!next_snapshot_is_due())
+                               continue;
                        ret = create_snapshot(rsync_argv);
                        if (ret < 0)
                                goto out;
                        continue;
-               case SCS_RSYNC_RUNNING:
-                       continue;
-               case SCS_RSYNC_SUCCESS:
-                       ret = post_create_hook();
-                       if (ret < 0)
-                               goto out;
-                       continue;
-               case SCS_POST_HOOK_RUNNING:
+               case HS_SUCCESS:
+                       post_create_hook();
                        continue;
                }
        }
@@ -889,31 +1307,45 @@ out:
 
 static void exit_hook(int exit_code)
 {
-       int fds[3] = {0, 0, 0};
-       char *argv[] = {conf.exit_hook_arg, dss_strerror(-exit_code), NULL};
+       char *argv[3];
        pid_t pid;
 
-       if (!conf.exit_hook_given)
-               return;
-       DSS_NOTICE_LOG("executing %s %s\n", argv[0], argv[1]);
-       dss_exec(&pid, conf.exit_hook_arg, argv, fds);
+       argv[0] = conf.exit_hook_arg;
+       argv[1] = dss_strerror(-exit_code);
+       argv[2] = NULL;
+
+       DSS_NOTICE_LOG(("executing %s %s\n", argv[0], argv[1]));
+       dss_exec(&pid, conf.exit_hook_arg, argv);
+}
+
+static void lock_dss_or_die(void)
+{
+       char *config_file = get_config_file_name();
+       int ret = lock_dss(config_file);
+
+       free(config_file);
+       if (ret < 0) {
+               DSS_EMERG_LOG(("failed to lock: %s\n", dss_strerror(-ret)));
+               exit(EXIT_FAILURE);
+       }
 }
 
 static int com_run(void)
 {
        int ret;
 
+       lock_dss_or_die();
        if (conf.dry_run_given) {
-               DSS_ERROR_LOG("dry_run not supported by this command\n");
+               DSS_ERROR_LOG(("dry_run not supported by this command\n"));
                return -E_SYNTAX;
        }
        ret = install_sighandler(SIGHUP);
        if (ret < 0)
                return ret;
-       compute_next_snapshot_time();
        ret = select_loop();
        if (ret >= 0) /* impossible */
                ret = -E_BUG;
+       kill_children();
        exit_hook(ret);
        return ret;
 }
@@ -922,38 +1354,58 @@ static int com_prune(void)
 {
        int ret;
        struct snapshot_list sl;
+       struct snapshot *victim;
        struct disk_space ds;
+       const char *why;
 
+       lock_dss_or_die();
        ret = get_disk_space(".", &ds);
        if (ret < 0)
                return ret;
        log_disk_space(&ds);
-       for (;;) {
-               dss_get_snapshot_list(&sl);
-               ret = remove_outdated_snapshot(&sl);
-               free_snapshot_list(&sl);
-               if (ret < 0)
-                       return ret;
-               if (!ret)
-                       break;
-               ret = wait_for_rm_process();
-               if (ret < 0)
-                       goto out;
+       dss_get_snapshot_list(&sl);
+       why = "outdated";
+       victim = find_outdated_snapshot(&sl);
+       if (victim)
+               goto rm;
+       why = "redundant";
+       victim = find_redundant_snapshot(&sl);
+       if (victim)
+               goto rm;
+       ret = 0;
+       goto out;
+rm:
+       if (conf.dry_run_given) {
+               dss_msg("%s snapshot %s (interval = %i)\n",
+                       why, victim->name, victim->interval);
+               ret = 0;
+               goto out;
        }
-       for (;;) {
-               dss_get_snapshot_list(&sl);
-               ret = remove_redundant_snapshot(&sl);
-               free_snapshot_list(&sl);
-               if (ret < 0)
-                       return ret;
-               if (!ret)
-                       break;
-               ret = wait_for_rm_process();
+       pre_remove_hook(victim, why);
+       if (snapshot_removal_status == HS_PRE_RUNNING) {
+               ret = wait_for_remove_process();
                if (ret < 0)
                        goto out;
+               if (snapshot_removal_status != HS_PRE_SUCCESS)
+                       goto out;
        }
-       return 1;
+       ret = exec_rm();
+       if (ret < 0)
+               goto out;
+       ret = wait_for_remove_process();
+       if (ret < 0)
+               goto out;
+       if (snapshot_removal_status != HS_SUCCESS)
+               goto out;
+       post_remove_hook();
+       if (snapshot_removal_status != HS_POST_RUNNING)
+               goto out;
+       ret = wait_for_remove_process();
+       if (ret < 0)
+               goto out;
+       ret = 1;
 out:
+       free_snapshot_list(&sl);
        return ret;
 }
 
@@ -962,6 +1414,7 @@ static int com_create(void)
        int ret, status;
        char **rsync_argv;
 
+       lock_dss_or_die();
        if (conf.dry_run_given) {
                int i;
                char *msg = NULL;
@@ -977,15 +1430,13 @@ static int com_create(void)
                free(msg);
                return 1;
        }
-       ret = pre_create_hook();
-       if (ret < 0)
-               return ret;
+       pre_create_hook();
        if (create_pid) {
                ret = wait_for_process(create_pid, &status);
                if (ret < 0)
                        return ret;
                ret = handle_pre_create_hook_exit(status);
-               if (ret < 0)
+               if (ret <= 0) /* error, or pre-create failed */
                        return ret;
        }
        create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
@@ -1018,7 +1469,7 @@ static int com_ls(void)
                if (s->flags & SS_COMPLETE)
                        d = (s->completion_time - s->creation_time) / 60;
                dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval, s->name, d/60, d%60);
-       };
+       }
        free_snapshot_list(&sl);
        return 1;
 }
@@ -1027,7 +1478,7 @@ static int setup_signal_handling(void)
 {
        int ret;
 
-       DSS_INFO_LOG("setting up signal handlers\n");
+       DSS_INFO_LOG(("setting up signal handlers\n"));
        signal_pipe = signal_init(); /* always successful */
        ret = install_sighandler(SIGINT);
        if (ret < 0)
@@ -1038,38 +1489,47 @@ static int setup_signal_handling(void)
        return install_sighandler(SIGCHLD);
 }
 
-/**
- * The main function of dss.
- *
- * \param argc Usual argument count.
- * \param argv Usual argument vector.
- */
 int main(int argc, char **argv)
 {
        int ret;
-       struct cmdline_parser_params params = {
-               .override = 0,
-               .initialize = 1,
-               .check_required = 0,
-               .check_ambiguity = 0,
-               .print_errors = 1
-       };
+       struct cmdline_parser_params params;
+
+       params.override = 0;
+       params.initialize = 1;
+       params.check_required = 0;
+       params.check_ambiguity = 0;
+       params.print_errors = 1;
 
        cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
        ret = parse_config_file(0);
        if (ret < 0)
                goto out;
+       if (ret == 0) { /* no config file given */
+               /*
+                * Parse the command line options again, but this time check
+                * that all required options are given.
+                */
+               struct cmdline_parser_params params;
+               params.override = 1;
+               params.initialize = 1;
+               params.check_required = 1;
+               params.check_ambiguity = 1;
+               params.print_errors = 1;
+               cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
+       }
        if (conf.daemon_given)
                daemon_init();
        ret = change_to_dest_dir();
        if (ret < 0)
                goto out;
+       dump_dss_config("startup");
        ret = setup_signal_handling();
        if (ret < 0)
                goto out;
        ret = call_command_handler();
+       signal_shutdown();
 out:
        if (ret < 0)
-               DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
+               DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
        exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
 }