dss-0.1.7.
[dss.git] / dss.c
diff --git a/dss.c b/dss.c
index db065c689b95dc0274231c9923cb9be6dd852839..aa8c0ea3860086bd4ca9f22b8e8c2354c3f2cc3b 100644 (file)
--- a/dss.c
+++ b/dss.c
@@ -1,16 +1,18 @@
 /*
- * Copyright (C) 2008-2010 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 <string.h>
 #include <stdlib.h>
+#include <stdio.h>
 #include <stdarg.h>
 #include <assert.h>
 #include <errno.h>
 #include <sys/types.h>
 #include <signal.h>
 #include <ctype.h>
+#include <stdbool.h>
 #include <sys/stat.h>
 #include <unistd.h>
 #include <inttypes.h>
 #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;
@@ -44,6 +47,8 @@ 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;
+/** 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. */
@@ -64,7 +69,7 @@ enum hook_status snapshot_removal_status;
 
 
 DEFINE_DSS_ERRLIST;
-static const char const *hook_status_description[] = {HOOK_STATUS_ARRAY};
+static const char *hook_status_description[] = {HOOK_STATUS_ARRAY};
 
 /* may be called with ds == NULL. */
 static int disk_space_low(struct disk_space *ds)
@@ -104,12 +109,12 @@ static void dump_dss_config(const char *msg)
        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");
+               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));
+               DSS_ERROR_LOG(("can not get free disk space: %s\n",
+                       dss_strerror(-ret)));
 
        /* we continue on errors from get_disk_space */
 
@@ -124,6 +129,7 @@ static void dump_dss_config(const char *msg)
                "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",
@@ -134,7 +140,8 @@ static void dump_dss_config(const char *msg)
                name_of_reference_snapshot?
                        name_of_reference_snapshot : "(none)",
                hook_status_description[snapshot_creation_status],
-               hook_status_description[snapshot_removal_status]
+               hook_status_description[snapshot_removal_status],
+               num_consecutive_rsync_errors
        );
        if (create_pid != 0)
                fprintf(log,
@@ -170,7 +177,10 @@ static void dump_dss_config(const char *msg)
        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
@@ -178,12 +188,25 @@ 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.
  *
@@ -192,7 +215,7 @@ 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;
@@ -200,7 +223,7 @@ __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);
@@ -208,7 +231,12 @@ __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);
@@ -226,6 +254,47 @@ static __printf_1_2 void dss_msg(const char* fmt,...)
        va_end(argp);
 }
 
+static char *get_config_file_name(void)
+{
+       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.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)
 {
        get_snapshot_list(sl, conf.unit_interval_arg, conf.num_intervals_arg);
@@ -236,7 +305,7 @@ static int64_t compute_next_snapshot_time(void)
        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;
+               num_complete = 0;
        int i;
        struct snapshot *s = NULL;
        struct snapshot_list sl;
@@ -245,15 +314,15 @@ static int64_t compute_next_snapshot_time(void)
        FOR_EACH_SNAPSHOT(s, i, &sl) {
                if (!(s->flags & SS_COMPLETE))
                        continue;
-               num_complete_snapshots++;
+               num_complete++;
                x += s->completion_time - s->creation_time;
        }
        assert(x >= 0);
 
        ret = now;
-       if (num_complete_snapshots == 0)
+       if (num_complete == 0)
                goto out;
-       x /= num_complete_snapshots; /* avg time to create one snapshot */
+       x /= num_complete; /* avg time to create one snapshot */
        if (unit_interval < x * wanted) /* oops, no sleep at all */
                goto out;
        ret = s->completion_time + unit_interval / wanted - x;
@@ -279,38 +348,31 @@ static int next_snapshot_is_due(void)
        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");
+               DSS_DEBUG_LOG(("next snapshot: now\n"));
                return 1;
        }
-       DSS_DEBUG_LOG("next snapshot due in %" PRId64 " seconds\n",
-               next_snapshot_time - now);
+       DSS_DEBUG_LOG(("next snapshot due in %" PRId64 " seconds\n",
+               next_snapshot_time - now));
        return 0;
 }
 
-static int pre_create_hook(void)
+static void pre_create_hook(void)
 {
-       int ret, fds[3] = {0, 0, 0};
-
        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);
-       ret = dss_exec_cmdline_pid(&create_pid,
-               conf.pre_create_hook_arg, fds);
-       if (ret < 0)
-               return ret;
+       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;
-       return ret;
 }
 
-static int pre_remove_hook(struct snapshot *s, const char *why)
+static void pre_remove_hook(struct snapshot *s, const char *why)
 {
-       int ret, fds[3] = {0, 0, 0};
        char *cmd;
 
        if (!s)
-               return 0;
-       DSS_DEBUG_LOG("%s snapshot %s\n", why, s->name);
+               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);
@@ -321,33 +383,32 @@ static int pre_remove_hook(struct snapshot *s, const char *why)
 
        cmd = make_message("%s %s/%s", conf.pre_remove_hook_arg,
                conf.dest_dir_arg, s->name);
-       DSS_DEBUG_LOG("executing %s\n", cmd);
-       ret = dss_exec_cmdline_pid(&remove_pid, cmd, fds);
+       DSS_DEBUG_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&remove_pid, cmd);
        free(cmd);
-       if (ret < 0)
-               return ret;
        snapshot_removal_status = HS_PRE_RUNNING;
-       return ret;
 }
 
 static int exec_rm(void)
 {
        struct snapshot *s = snapshot_currently_being_removed;
-       int fds[3] = {0, 0, 0};
        char *new_name = being_deleted_name(s);
-       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);
+       DSS_NOTICE_LOG(("removing %s (interval = %i)\n", s->name, s->interval));
        ret = dss_rename(s->name, new_name);
        if (ret < 0)
                goto out;
-       ret = dss_exec(&remove_pid, argv[0], argv, fds);
-       if (ret < 0)
-               goto out;
+       dss_exec(&remove_pid, argv[0], argv);
        snapshot_removal_status = HS_RUNNING;
 out:
        free(new_name);
@@ -364,7 +425,7 @@ static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
        struct snapshot *s;
        int i;
 
-       DSS_DEBUG_LOG("looking for orphaned snapshots\n");
+       DSS_DEBUG_LOG(("looking for orphaned snapshots\n"));
        FOR_EACH_SNAPSHOT(s, i, sl) {
                if (snapshot_is_being_created(s))
                        continue;
@@ -409,7 +470,7 @@ static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
        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];
@@ -418,8 +479,6 @@ static struct snapshot *find_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 */
@@ -430,7 +489,6 @@ static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
                                continue;
                        if (is_reference_snapshot(s))
                                continue;
-                       //DSS_DEBUG_LOG("checking %s\n", s->name);
                        if (s->interval > interval) {
                                prev = s;
                                continue;
@@ -446,7 +504,6 @@ static struct snapshot *find_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;
@@ -464,8 +521,8 @@ static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
        int i;
        struct snapshot *s;
 
-       DSS_DEBUG_LOG("looking for snapshots belonging to intervals >= %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;
@@ -478,36 +535,51 @@ static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
        return NULL;
 }
 
-struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
+static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
 {
-       int i;
-       struct snapshot *s;
+       int i, num_complete;
+       struct snapshot *s, *ref = NULL;
+
+       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))
+               if (is_reference_snapshot(s)) { /* avoid this one */
+                       ref = s;
                        continue;
-               DSS_INFO_LOG("oldest removable snapshot: %s\n", s->name);
+               }
+               DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
                return s;
        }
-       return NULL;
+       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;
 }
@@ -536,9 +608,18 @@ static int try_to_free_disk_space(void)
                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 = 0;
-       if (!low_disk_space && sl.num_snapshots <= 1)
+       /*
+        * 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;
        why = "outdated";
        victim = find_outdated_snapshot(&sl);
@@ -548,46 +629,39 @@ static int try_to_free_disk_space(void)
        victim = find_redundant_snapshot(&sl);
        if (victim)
                goto remove;
-       /* try harder only if disk space is low */
-       if (!low_disk_space)
-               goto out;
        why = "orphaned";
        victim = find_orphaned_snapshot(&sl);
        if (victim)
                goto remove;
-       DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
+       /* try harder only if disk space is low */
+       if (!low_disk_space)
+               goto out;
+       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");
+       DSS_CRIT_LOG(("uhuhu: disk space low and nothing to remove\n"));
        ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
        goto out;
 remove:
-       ret = pre_remove_hook(victim, why);
+       pre_remove_hook(victim, why);
 out:
        free_snapshot_list(&sl);
        return ret;
 }
 
-static int post_create_hook(void)
+static void post_create_hook(void)
 {
-       int ret, fds[3] = {0, 0, 0};
-       char *cmd;
-
-       cmd = make_message("%s %s/%s", conf.post_create_hook_arg,
+       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);
-       ret = dss_exec_cmdline_pid(&create_pid, cmd, fds);
+       DSS_NOTICE_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&create_pid, cmd);
        free(cmd);
-       if (ret < 0)
-               return ret;
        snapshot_creation_status = HS_POST_RUNNING;
-       return ret;
 }
 
-static int post_remove_hook(void)
+static void post_remove_hook(void)
 {
-       int ret, fds[3] = {0, 0, 0};
        char *cmd;
        struct snapshot *s = snapshot_currently_being_removed;
 
@@ -595,13 +669,10 @@ static int post_remove_hook(void)
 
        cmd = make_message("%s %s/%s", conf.post_remove_hook_arg,
                conf.dest_dir_arg, s->name);
-       DSS_NOTICE_LOG("executing %s\n", cmd);
-       ret = dss_exec_cmdline_pid(&remove_pid, cmd, fds);
+       DSS_NOTICE_LOG(("executing %s\n", cmd));
+       dss_exec_cmdline_pid(&remove_pid, cmd);
        free(cmd);
-       if (ret < 0)
-               return ret;
        snapshot_removal_status = HS_POST_RUNNING;
-       return ret;
 }
 
 static void dss_kill(pid_t pid, int sig, const char *msg)
@@ -624,13 +695,13 @@ static void dss_kill(pid_t pid, int sig, const char *msg)
        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);
+               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);
+       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)
@@ -655,20 +726,20 @@ 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;
 
@@ -693,7 +764,7 @@ static int wait_for_process(pid_t pid, int *status)
                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;
@@ -780,7 +851,7 @@ 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 = HS_READY;
                goto out;
@@ -788,23 +859,27 @@ static int handle_rsync_exit(int status)
        es = WEXITSTATUS(status);
        /*
         * Restart rsync on non-fatal errors:
-        * 12: Error in rsync protocol data stream
-        * 13: Errors with program diagnostics
+        * 24: Partial transfer due to vanished source files
         */
-       if (es == 12 || es == 13) {
-               DSS_WARNING_LOG("rsync process %d returned %d -- restarting\n",
-                       (int)create_pid, es);
+       if (es != 0 && es != 24) {
+               DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
+                       es, ++num_consecutive_rsync_errors));
+               if (conf.create_given) {
+                       ret = -E_BAD_EXIT_CODE;
+                       goto out;
+               }
+               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 = HS_READY;
-               goto out;
-       }
+       num_consecutive_rsync_errors = 0;
        ret = rename_incomplete_snapshot(current_snapshot_creation_time);
        if (ret < 0)
                goto out;
@@ -829,9 +904,9 @@ static int handle_pre_create_hook_exit(int status)
        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 */
                }
                next_snapshot_time = get_current_time() + 60;
@@ -867,8 +942,8 @@ static int handle_sigchld(void)
                        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;
@@ -880,22 +955,23 @@ static int handle_sigchld(void)
                        return ret;
                return ret;
        }
-       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;
 }
 
@@ -903,22 +979,15 @@ static int check_config(void)
  * 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)
+static int parse_config_file(bool sighup)
 {
        int ret, config_file_exists;
-       char *config_file;
+       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 (sighup) {
                if (conf.logfile_given)
                        old_logfile_arg = dss_strdup(conf.logfile_arg);
                old_daemon_given = conf.daemon_given;
@@ -927,18 +996,17 @@ static int parse_config_file(int override)
        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 (config_file_exists) {
-               struct cmdline_parser_params params = {
-                       .override = override,
-                       .initialize = 0,
-                       .check_required = 1,
-                       .check_ambiguity = 0,
-                       .print_errors = 1
-               };
-               if (override) { /* invalidate all rsync options */
+               struct cmdline_parser_params params;
+               params.override = sighup;
+               params.initialize = 0;
+               params.check_required = 1;
+               params.check_ambiguity = 0;
+               params.print_errors = 1;
+               if (sighup) { /* invalidate all rsync options */
                        int i;
 
                        for (i = 0; i < conf.rsync_option_given; i++) {
@@ -952,7 +1020,7 @@ static int parse_config_file(int override)
        ret = check_config();
        if (ret < 0)
                goto out;
-       if (override) {
+       if (sighup) {
                /* don't change daemon mode on SIGHUP */
                conf.daemon_given = old_daemon_given;
                close_log(logfile);
@@ -964,22 +1032,22 @@ 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);
+       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);
 }
 
@@ -987,9 +1055,9 @@ static int handle_sighup(void)
 {
        int ret;
 
-       DSS_NOTICE_LOG("SIGHUP, re-reading config\n");
+       DSS_NOTICE_LOG(("SIGHUP, re-reading config\n"));
        dump_dss_config("old");
-       ret = parse_config_file(1);
+       ret = parse_config_file(true /* SIGHUP */);
        if (ret < 0)
                return ret;
        dump_dss_config("reloaded");
@@ -997,6 +1065,13 @@ static int handle_sighup(void)
        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();
@@ -1007,9 +1082,7 @@ static int handle_signal(void)
        switch (sig) {
        case SIGINT:
        case SIGTERM:
-               restart_create_process();
-               dss_kill(create_pid, SIGTERM, NULL);
-               dss_kill(remove_pid, SIGTERM, NULL);
+               kill_children();
                ret = -E_SIGNAL;
                break;
        case SIGHUP:
@@ -1021,7 +1094,7 @@ 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;
 }
 
@@ -1042,25 +1115,53 @@ static int use_rsync_locally(char *logname)
 
 static int rename_resume_snap(int64_t creation_time)
 {
-       struct snapshot_list sl = {.num_snapshots = 0};
-       struct snapshot *s;
+       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)
+       if (!s) /* no snapshots at all */
                goto out;
-       if ((s->flags & SS_COMPLETE) != 0) /* complete */
+       /* re-use last snapshot if it is incomplete */
+       why = "aborted";
+       if ((s->flags & SS_COMPLETE) == 0)
                goto out;
-       DSS_INFO_LOG("resuming: reusing %s as destination dir\n", s->name);
-       ret = dss_rename(s->name, new_name);
+       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_NOTICE_LOG(("recycling %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);
+               DSS_NOTICE_LOG(("creating %s\n", new_name));
        free(new_name);
        free_snapshot_list(&sl);
        return ret;
@@ -1079,16 +1180,16 @@ static void create_rsync_argv(char ***argv, int64_t *num)
 
        *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 (name_of_reference_snapshot) {
-               DSS_INFO_LOG("using %s as reference\n", 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 suitable reference 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);
@@ -1101,7 +1202,7 @@ 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)
@@ -1117,14 +1218,12 @@ static void free_rsync_argv(char **argv)
 
 static int create_snapshot(char **argv)
 {
-       int ret, fds[3] = {0, 0, 0};
+       int ret;
 
        ret = rename_resume_snap(current_snapshot_creation_time);
        if (ret < 0)
                return ret;
-       ret = dss_exec(&create_pid, argv[0], argv, fds);
-       if (ret < 0)
-               return ret;
+       dss_exec(&create_pid, argv[0], argv);
        snapshot_creation_status = HS_RUNNING;
        return ret;
 }
@@ -1166,9 +1265,7 @@ static int select_loop(void)
                        continue;
                }
                if (snapshot_removal_status == HS_SUCCESS) {
-                       ret = post_remove_hook();
-                       if (ret < 0)
-                               goto out;
+                       post_remove_hook();
                        continue;
                }
                ret = try_to_free_disk_space();
@@ -1183,9 +1280,7 @@ static int select_loop(void)
                case HS_READY:
                        if (!next_snapshot_is_due())
                                continue;
-                       ret = pre_create_hook();
-                       if (ret < 0)
-                               goto out;
+                       pre_create_hook();
                        continue;
                case HS_PRE_RUNNING:
                case HS_RUNNING:
@@ -1208,9 +1303,7 @@ static int select_loop(void)
                                goto out;
                        continue;
                case HS_SUCCESS:
-                       ret = post_create_hook();
-                       if (ret < 0)
-                               goto out;
+                       post_create_hook();
                        continue;
                }
        }
@@ -1220,20 +1313,36 @@ 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;
 
-       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);
@@ -1242,6 +1351,7 @@ static int com_run(void)
        ret = select_loop();
        if (ret >= 0) /* impossible */
                ret = -E_BUG;
+       kill_children();
        exit_hook(ret);
        return ret;
 }
@@ -1254,6 +1364,7 @@ static int com_prune(void)
        struct disk_space ds;
        const char *why;
 
+       lock_dss_or_die();
        ret = get_disk_space(".", &ds);
        if (ret < 0)
                return ret;
@@ -1276,9 +1387,7 @@ rm:
                ret = 0;
                goto out;
        }
-       ret = pre_remove_hook(victim, why);
-       if (ret < 0)
-               goto out;
+       pre_remove_hook(victim, why);
        if (snapshot_removal_status == HS_PRE_RUNNING) {
                ret = wait_for_remove_process();
                if (ret < 0)
@@ -1294,9 +1403,7 @@ rm:
                goto out;
        if (snapshot_removal_status != HS_SUCCESS)
                goto out;
-       ret = post_remove_hook();
-       if (ret < 0)
-               goto out;
+       post_remove_hook();
        if (snapshot_removal_status != HS_POST_RUNNING)
                goto out;
        ret = wait_for_remove_process();
@@ -1313,6 +1420,7 @@ 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;
@@ -1328,9 +1436,7 @@ 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)
@@ -1369,7 +1475,7 @@ 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;
 }
@@ -1378,7 +1484,7 @@ 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)
@@ -1389,25 +1495,20 @@ 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);
+       ret = parse_config_file(false /* no SIGHUP */);
        if (ret < 0)
                goto out;
        if (ret == 0) { /* no config file given */
@@ -1415,13 +1516,11 @@ int main(int argc, char **argv)
                 * Parse the command line options again, but this time check
                 * that all required options are given.
                 */
-               params = (struct cmdline_parser_params) {
-                       .override = 1,
-                       .initialize = 1,
-                       .check_required = 1,
-                       .check_ambiguity = 1,
-                       .print_errors = 1
-               };
+               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)
@@ -1434,8 +1533,9 @@ int main(int argc, char **argv)
        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);
 }