X-Git-Url: http://git.tuebingen.mpg.de/?p=dss.git;a=blobdiff_plain;f=dss.c;h=fd876e5ddc1d340dec79dee84f9706dc9a09912f;hp=02d0c8bd5d9e1cc5ef3c5051088a8c99bbe60021;hb=d184cc0fa0368ca089b1ae6f2162ccfc224bf28c;hpb=e83760cba6aba516c3de4bb4ef4df9fa4020ccfa diff --git a/dss.c b/dss.c index 02d0c8b..fd876e5 100644 --- a/dss.c +++ b/dss.c @@ -1,3 +1,8 @@ +/* + * Copyright (C) 2008 Andre Noll + * + * Licensed under the GPL v2. For licencing details see COPYING. + */ #include #include #include @@ -35,18 +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. */ @@ -72,7 +73,7 @@ static int call_command_handler(void) { COMMANDS DSS_EMERG_LOG("BUG: did not find command handler\n"); - exit(EXIT_FAILURE); + return -E_BUG; } #undef COMMAND #undef COMMANDS @@ -119,7 +120,6 @@ static __printf_1_2 void dss_msg(const char* fmt,...) va_end(argp); } -/* TODO: Also consider number of inodes. */ static int disk_space_low(void) { struct disk_space ds; @@ -171,13 +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: @@ -188,7 +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); @@ -198,13 +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 */ @@ -214,7 +220,7 @@ static int remove_redundant_snapshot(struct snapshot_list *sl) struct snapshot *s; unsigned missing = 0; - DSS_INFO_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]; @@ -231,6 +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; @@ -260,7 +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; @@ -271,9 +279,11 @@ static int remove_outdated_snapshot(struct snapshot_list *sl) int i, ret; struct snapshot *s; - DSS_INFO_LOG("looking for snapshots belonging to intervals greater than %d\n", + 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) { @@ -281,7 +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; @@ -296,9 +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) @@ -325,6 +335,8 @@ static int try_to_free_disk_space(int low_disk_space) int ret; struct snapshot_list sl; + if (!low_disk_space && conf.keep_redundant_given) + return 0; dss_get_snapshot_list(&sl); ret = remove_outdated_snapshot(&sl); if (ret) /* error, or we are removing something */ @@ -341,7 +353,7 @@ static int try_to_free_disk_space(int low_disk_space) if (ret) goto out; DSS_CRIT_LOG("uhuhu: not enough disk space for a single snapshot\n"); - ret= -ENOSPC; + ret = -ERRNO_TO_DSS_ERROR(ENOSPC); out: free_snapshot_list(&sl); return ret; @@ -356,7 +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; @@ -374,10 +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; @@ -393,23 +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. */ @@ -485,15 +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(); @@ -504,8 +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; } @@ -529,7 +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; } @@ -540,19 +560,27 @@ 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); - exit(EXIT_FAILURE); + return -E_BUG; } static int check_config(void) @@ -570,8 +598,7 @@ static int check_config(void) return 1; } -/* exits on errors */ -static void parse_config_file(int override) +static int parse_config_file(int override) { int ret; char *config_file; @@ -606,6 +633,15 @@ static void 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, ¶ms); } ret = check_config(); @@ -627,24 +663,33 @@ static void parse_config_file(int override) logfile = open_log(conf.logfile_arg); log_welcome(conf.loglevel_arg); } - DSS_EMERG_LOG("loglevel: %d\n", conf.loglevel_arg); + DSS_DEBUG_LOG("loglevel: %d\n", conf.loglevel_arg); // cmdline_parser_dump(logfile? logfile : stdout, &conf); - ret = dss_chdir(conf.dest_dir_arg); out: free(config_file); - if (ret >= 0) - return; - DSS_EMERG_LOG("%s\n", dss_strerror(-ret)); - exit(EXIT_FAILURE); + if (ret < 0) + 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); + return dss_chdir(conf.dest_dir_arg); } -static void handle_sighup(void) +static int handle_sighup(void) { + int ret; + DSS_NOTICE_LOG("SIGHUP\n"); - parse_config_file(1); + ret = parse_config_file(1); + if (ret < 0) + return ret; + return change_to_dest_dir(); } -static void handle_signal(void) +static int handle_signal(void) { int sig, ret = next_signal(); @@ -654,13 +699,13 @@ static void 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); - exit(EXIT_FAILURE); + ret = -E_SIGNAL; + break; case SIGHUP: - handle_sighup(); - ret = 1; + ret = handle_sighup(); break; case SIGCHLD: ret = handle_sigchld(); @@ -669,6 +714,22 @@ static void handle_signal(void) out: if (ret < 0) DSS_ERROR_LOG("%s\n", dss_strerror(-ret)); + 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) @@ -693,13 +754,8 @@ static void create_rsync_argv(char ***argv, int64_t *num) free(newest); } else DSS_INFO_LOG("no previous snapshot found\n"); - if (conf.exclude_patterns_given) { - (*argv)[i++] = dss_strdup("--exclude-from"); - (*argv)[i++] = dss_strdup(conf.exclude_patterns_arg); - - } 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? @@ -716,6 +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); @@ -729,7 +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; @@ -739,42 +798,49 @@ static int create_snapshot(char **argv) static int select_loop(void) { int ret; - struct timeval tv = {.tv_sec = 0, .tv_usec = 0}; + /* 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 = &tv; + struct timeval now, *tvp; - if (rsync_pid) - tv.tv_sec = 60; /* check every 60 seconds for free disk space */ - else if (rm_pid) + if (rm_pid) tvp = NULL; /* sleep until rm process dies */ + else { /* sleep one minute */ + tv.tv_sec = 60; + tv.tv_usec = 0; + tvp = &tv; + } FD_ZERO(&rfds); FD_SET(signal_pipe, &rfds); - DSS_DEBUG_LOG("tvp: %p, tv_sec: %lu\n", tvp, (long unsigned) tv.tv_sec); + 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) - return ret; - if (FD_ISSET(signal_pipe, &rfds)) - handle_signal(); + goto out; + if (FD_ISSET(signal_pipe, &rfds)) { + ret = handle_signal(); + if (ret < 0) + goto out; + } if (rm_pid) continue; ret = disk_space_low(); if (ret < 0) - break; + goto out; low_disk_space = ret; ret = try_to_free_disk_space(low_disk_space); if (ret < 0) - break; + 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, &tv) > 0) + if (tv_diff(&next_snapshot_time, &now, NULL) > 0) continue; switch (snapshot_creation_status) { case SCS_READY: @@ -784,10 +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, ¤t_snapshot_creation_time); ret = create_snapshot(rsync_argv); - free_rsync_argv(rsync_argv); if (ret < 0) goto out; continue; @@ -806,6 +877,18 @@ out: return ret; } +static void exit_hook(int exit_code) +{ + int fds[3] = {0, 0, 0}; + char *argv[] = {conf.exit_hook_arg, dss_strerror(-exit_code), NULL}; + 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); +} + static int com_run(void) { int ret; @@ -818,7 +901,11 @@ static int com_run(void) if (ret < 0) return ret; compute_next_snapshot_time(); - return select_loop(); + ret = select_loop(); + if (ret >= 0) /* impossible */ + ret = -E_BUG; + exit_hook(ret); + return ret; } static int com_prune(void) @@ -883,8 +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); @@ -895,15 +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; @@ -916,13 +1003,17 @@ 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; } -static void setup_signal_handling(void) +static int setup_signal_handling(void) { int ret; @@ -930,17 +1021,11 @@ static void setup_signal_handling(void) signal_pipe = signal_init(); /* always successful */ ret = install_sighandler(SIGINT); if (ret < 0) - goto err; + return ret; ret = install_sighandler(SIGTERM); if (ret < 0) - goto err; - ret = install_sighandler(SIGCHLD); - if (ret < 0) - goto err; - return; -err: - DSS_EMERG_LOG("could not install signal handlers\n"); - exit(EXIT_FAILURE); + return ret; + return install_sighandler(SIGCHLD); } /** @@ -961,12 +1046,19 @@ int main(int argc, char **argv) }; cmdline_parser_ext(argc, argv, &conf, ¶ms); /* aborts on errors */ - parse_config_file(0); - + ret = parse_config_file(0); + if (ret < 0) + goto out; if (conf.daemon_given) daemon_init(); - setup_signal_handling(); + ret = change_to_dest_dir(); + if (ret < 0) + goto out; + ret = setup_signal_handling(); + if (ret < 0) + goto out; ret = call_command_handler(); +out: if (ret < 0) DSS_EMERG_LOG("%s\n", dss_strerror(-ret)); exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);