Bump version number to 0.1.2.
[dss.git] / dss.c
diff --git a/dss.c b/dss.c
index b27bdf6604d14f44dfa34d402e1c22f6e37055ae..fd876e5ddc1d340dec79dee84f9706dc9a09912f 100644 (file)
--- a/dss.c
+++ b/dss.c
@@ -40,18 +40,14 @@ static struct gengetopt_args_info conf;
 static FILE *logfile;
 /** The read end of the signal pipe */
 static int signal_pipe;
-/** Process id of current rsync process. */
-static pid_t rsync_pid;
-/** Whether the rsync process is currently stopped */
-static int rsync_stopped;
+/** Process id of current pre-create-hook/rsync/post-create-hook process. */
+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;
 /** When the next snapshot is due. */
 static struct timeval next_snapshot_time;
-/** The pid of the pre-create hook. */
-static pid_t pre_create_hook_pid;
-/** The pid of the post-create hook. */
-static pid_t post_create_hook_pid;
 /** Creation time of the snapshot currently being created. */
 static int64_t current_snapshot_creation_time;
 /** Needed by the post-create hook. */
@@ -175,13 +171,13 @@ static void compute_next_snapshot_time(void)
        tmp.tv_sec = x;
        tmp.tv_usec = 0;
        ret = tv_diff(&unit_interval, &tmp, &diff); /* total sleep time per unit interval */
-       gettimeofday(&now, NULL);
-       if (ret < 0 || !s)
+       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;
+       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)
                goto out;
 min_sleep:
@@ -192,7 +188,7 @@ out:
 }
 
 
-static int remove_snapshot(struct snapshot *s)
+static int remove_snapshot(struct snapshot *s, char *why)
 {
        int fds[3] = {0, 0, 0};
        assert(!rm_pid);
@@ -202,13 +198,19 @@ static int remove_snapshot(struct snapshot *s)
 
        if (ret < 0)
                goto out;
-       DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
+       DSS_NOTICE_LOG("removing %s snapshot %s (interval = %i)\n",
+               why, s->name, s->interval);
        ret = dss_exec(&rm_pid, argv[0], argv, fds);
 out:
        free(new_name);
        return ret;
 }
 
+static int snapshot_is_being_created(struct snapshot *s)
+{
+       return s->creation_time == current_snapshot_creation_time;
+}
+
 /*
  * return: 0: no redundant snapshots, 1: rm process started, negative: error
  */
@@ -235,6 +237,8 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
                FOR_EACH_SNAPSHOT(s, i, sl) {
                        int64_t this_score;
 
+                       if (snapshot_is_being_created(s))
+                               continue;
                        //DSS_DEBUG_LOG("checking %s\n", s->name);
                        if (s->interval > interval) {
                                prev = s;
@@ -264,7 +268,7 @@ static int remove_redundant_snapshot(struct snapshot_list *sl)
                                victim->name, victim->interval);
                        continue;
                }
-               ret = remove_snapshot(victim);
+               ret = remove_snapshot(victim, "redundant");
                return ret < 0? ret : 1;
        }
        return 0;
@@ -278,6 +282,8 @@ static int remove_outdated_snapshot(struct snapshot_list *sl)
        DSS_DEBUG_LOG("looking for snapshots belonging to intervals greater than %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)
                        continue;
                if (conf.dry_run_given) {
@@ -285,7 +291,7 @@ static int remove_outdated_snapshot(struct snapshot_list *sl)
                                s->name, s->interval);
                        continue;
                }
-               ret = remove_snapshot(s);
+               ret = remove_snapshot(s, "outdated");
                if (ret < 0)
                        return ret;
                return 1;
@@ -300,9 +306,9 @@ static int remove_oldest_snapshot(struct snapshot_list *sl)
        if (!s) /* no snapshot found */
                return 0;
        DSS_INFO_LOG("oldest snapshot: %s\n", s->name);
-       if (s->creation_time == current_snapshot_creation_time)
-               return 0; /* do not remove the snapshot currently being created */
-       return remove_snapshot(s);
+       if (snapshot_is_being_created(s))
+               return 0;
+       return remove_snapshot(s, "oldest");
 }
 
 static int rename_incomplete_snapshot(int64_t start)
@@ -362,7 +368,7 @@ static int pre_create_hook(void)
                return 0;
        }
        DSS_NOTICE_LOG("executing %s\n", conf.pre_create_hook_arg);
-       ret = dss_exec_cmdline_pid(&pre_create_hook_pid,
+       ret = dss_exec_cmdline_pid(&create_pid,
                conf.pre_create_hook_arg, fds);
        if (ret < 0)
                return ret;
@@ -380,10 +386,10 @@ static int post_create_hook(void)
                compute_next_snapshot_time();
                return 0;
        }
-       cmd = make_message("%s %s", conf.post_create_hook_arg,
-               path_to_last_complete_snapshot);
+       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(&post_create_hook_pid, cmd, fds);
+       ret = dss_exec_cmdline_pid(&create_pid, cmd, fds);
        free(cmd);
        if (ret < 0)
                return ret;
@@ -399,23 +405,22 @@ static void kill_process(pid_t pid)
        kill(pid, SIGTERM);
 }
 
-static void stop_rsync_process(void)
+static void stop_create_process(void)
 {
-       if (!rsync_pid || rsync_stopped)
+       if (!create_pid || create_process_stopped)
                return;
-       kill(SIGSTOP, rsync_pid);
-       rsync_stopped = 1;
+       kill(SIGSTOP, create_pid);
+       create_process_stopped = 1;
 }
 
-static void restart_rsync_process(void)
+static void restart_create_process(void)
 {
-       if (!rsync_pid || !rsync_stopped)
+       if (!create_pid || !create_process_stopped)
                return;
-       kill (SIGCONT, rsync_pid);
-       rsync_stopped = 0;
+       kill (SIGCONT, create_pid);
+       create_process_stopped = 0;
 }
 
-
 /**
  * Print a log message about the exit status of a child.
  */
@@ -491,15 +496,24 @@ static int handle_rsync_exit(int status)
        int es, ret;
 
        if (!WIFEXITED(status)) {
-               DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)rsync_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();
                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;
+               ret = 1;
+               goto out;
+       }
        if (es != 0 && es != 23 && es != 24) {
-               DSS_ERROR_LOG("rsync process %d returned %d\n", (int)rsync_pid, es);
+               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();
@@ -510,8 +524,8 @@ static int handle_rsync_exit(int status)
                goto out;
        snapshot_creation_status = SCS_RSYNC_SUCCESS;
 out:
-       rsync_pid = 0;
-       rsync_stopped = 0;
+       create_pid = 0;
+       create_process_stopped = 0;
        return ret;
 }
 
@@ -535,7 +549,7 @@ static int handle_pre_create_hook_exit(int status)
        snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
        ret = 1;
 out:
-       pre_create_hook_pid = 0;
+       create_pid = 0;
        return ret;
 }
 
@@ -546,17 +560,25 @@ static int handle_sigchld(void)
 
        if (ret <= 0)
                return ret;
-       if (pid == rsync_pid)
-               return handle_rsync_exit(status);
+
+       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;
+               default:
+                       DSS_EMERG_LOG("BUG: create can't die in status %d\n",
+                               snapshot_creation_status);
+                       return -E_BUG;
+               }
+       }
        if (pid == rm_pid)
                return handle_rm_exit(status);
-       if (pid == pre_create_hook_pid)
-               return handle_pre_create_hook_exit(status);
-       if (pid == post_create_hook_pid) {
-               snapshot_creation_status = SCS_READY;
-               compute_next_snapshot_time();
-               return 1;
-       }
        DSS_EMERG_LOG("BUG: unknown process %d died\n", (int)pid);
        return -E_BUG;
 }
@@ -611,6 +633,15 @@ static int parse_config_file(int override)
                        .check_ambiguity = 0,
                        .print_errors = 1
                };
+               if (override) { /* invalidate all rsync options */
+                       int i;
+
+                       for (i = 0; i < conf.rsync_option_given; i++) {
+                               free(conf.rsync_option_arg[i]);
+                               conf.rsync_option_arg[i] = NULL;
+                       }
+                       conf.rsync_option_given = 0;
+               }
                cmdline_parser_config_file(config_file, &conf, &params);
        }
        ret = check_config();
@@ -668,8 +699,8 @@ static int handle_signal(void)
        switch (sig) {
        case SIGINT:
        case SIGTERM:
-               restart_rsync_process();
-               kill_process(rsync_pid);
+               restart_create_process();
+               kill_process(create_pid);
                kill_process(rm_pid);
                ret = -E_SIGNAL;
                break;
@@ -686,6 +717,21 @@ out:
        return ret;
 }
 
+/*
+ * We can not use rsync locally if the local user is different from the remote
+ * user or if the src dir is not on the local host (or both).
+ */
+static int use_rsync_locally(char *logname)
+{
+       char *h = conf.remote_host_arg;
+
+       if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
+               return 0;
+       if (conf.remote_user_given && strcmp(conf.remote_user_arg, logname))
+               return 0;
+       return 1;
+}
+
 static void create_rsync_argv(char ***argv, int64_t *num)
 {
        char *logname, *newest;
@@ -709,7 +755,7 @@ static void create_rsync_argv(char ***argv, int64_t *num)
        } else
                DSS_INFO_LOG("no previous snapshot found\n");
        logname = dss_logname();
-       if (conf.remote_user_given && !strcmp(conf.remote_user_arg, logname))
+       if (use_rsync_locally(logname))
                (*argv)[i++] = dss_strdup(conf.source_dir_arg);
        else
                (*argv)[i++] = make_message("%s@%s:%s/", conf.remote_user_given?
@@ -726,6 +772,9 @@ static void create_rsync_argv(char ***argv, int64_t *num)
 static void free_rsync_argv(char **argv)
 {
        int i;
+
+       if (!argv)
+               return;
        for (i = 0; argv[i]; i++)
                free(argv[i]);
        free(argv);
@@ -739,7 +788,7 @@ static int create_snapshot(char **argv)
        name = incomplete_name(current_snapshot_creation_time);
        DSS_NOTICE_LOG("creating new snapshot %s\n", name);
        free(name);
-       ret = dss_exec(&rsync_pid, argv[0], argv, fds);
+       ret = dss_exec(&create_pid, argv[0], argv, fds);
        if (ret < 0)
                return ret;
        snapshot_creation_status = SCS_RSYNC_RUNNING;
@@ -751,11 +800,11 @@ static int select_loop(void)
        int ret;
        /* check every 60 seconds for free disk space */
        struct timeval tv;
+       char **rsync_argv = NULL;
 
        for (;;) {
                fd_set rfds;
                int low_disk_space;
-               char **rsync_argv;
                struct timeval now, *tvp;
 
                if (rm_pid)
@@ -786,10 +835,10 @@ static int select_loop(void)
                if (ret < 0)
                        goto out;
                if (rm_pid) {
-                       stop_rsync_process();
+                       stop_create_process();
                        continue;
                }
-               restart_rsync_process();
+               restart_create_process();
                gettimeofday(&now, NULL);
                if (tv_diff(&next_snapshot_time, &now, NULL) > 0)
                        continue;
@@ -801,10 +850,15 @@ static int select_loop(void)
                        continue;
                case SCS_PRE_HOOK_RUNNING:
                        continue;
+               case SCS_RSYNC_NEEDS_RESTART:
+                       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);
                        ret = create_snapshot(rsync_argv);
-                       free_rsync_argv(rsync_argv);
                        if (ret < 0)
                                goto out;
                        continue;
@@ -826,16 +880,13 @@ out:
 static void exit_hook(int exit_code)
 {
        int fds[3] = {0, 0, 0};
-       char *cmd;
+       char *argv[] = {conf.exit_hook_arg, dss_strerror(-exit_code), NULL};
        pid_t pid;
 
        if (!conf.exit_hook_given)
                return;
-       cmd = make_message("%s %s", conf.exit_hook_arg,
-               dss_strerror(-exit_code));
-       DSS_NOTICE_LOG("executing %s\n", cmd);
-       dss_exec_cmdline_pid(&pid, cmd, fds);
-       free(cmd);
+       DSS_NOTICE_LOG("executing %s %s\n", argv[0], argv[1]);
+       dss_exec(&pid, conf.exit_hook_arg, argv, fds);
 }
 
 static int com_run(void)
@@ -919,8 +970,8 @@ static int com_create(void)
        ret = pre_create_hook();
        if (ret < 0)
                return ret;
-       if (pre_create_hook_pid) {
-               ret = wait_for_process(pre_create_hook_pid, &status);
+       if (create_pid) {
+               ret = wait_for_process(create_pid, &status);
                if (ret < 0)
                        return ret;
                ret = handle_pre_create_hook_exit(status);
@@ -931,15 +982,15 @@ static int com_create(void)
        ret = create_snapshot(rsync_argv);
        if (ret < 0)
                goto out;
-       ret = wait_for_process(rsync_pid, &status);
+       ret = wait_for_process(create_pid, &status);
        if (ret < 0)
                goto out;
        ret = handle_rsync_exit(status);
        if (ret < 0)
                goto out;
        post_create_hook();
-       if (post_create_hook_pid)
-               ret = wait_for_process(post_create_hook_pid, &status);
+       if (create_pid)
+               ret = wait_for_process(create_pid, &status);
 out:
        free_rsync_argv(rsync_argv);
        return ret;
@@ -952,8 +1003,12 @@ static int com_ls(void)
        struct snapshot *s;
 
        dss_get_snapshot_list(&sl);
-       FOR_EACH_SNAPSHOT(s, i, &sl)
-               dss_msg("%u\t%s\n", s->interval, s->name);
+       FOR_EACH_SNAPSHOT(s, i, &sl) {
+               int64_t d = 0;
+               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;
 }