2 * Copyright (C) 2008-2011 Andre Noll <maan@tuebingen.mpg.de>
4 * Licensed under the GPL v2. For licencing details see COPYING.
12 #include <sys/types.h>
28 #include "gcc-compat.h"
42 #define CMD_PTR(_cname) lls_cmd(LSG_DSS_CMD_ ## _cname, dss_suite)
43 #define OPT_RESULT(_cname, _oname) (lls_opt_result(\
44 LSG_DSS_ ## _cname ## _OPT_ ## _oname, (CMD_PTR(_cname) == CMD_PTR(DSS))? lpr : sublpr))
45 #define OPT_GIVEN(_cname, _oname) (lls_opt_given(OPT_RESULT(_cname, _oname)))
46 #define OPT_STRING_VAL(_cname, _oname) (lls_string_val(0, \
47 OPT_RESULT(_cname, _oname)))
48 #define OPT_UINT32_VAL(_cname, _oname) (lls_uint32_val(0, \
49 OPT_RESULT(_cname, _oname)))
51 struct dss_user_data {int (*handler)(void);};
52 #define EXPORT_CMD_HANDLER(_cmd) const struct dss_user_data \
53 lsg_dss_com_ ## _cmd ## _user_data = { \
54 .handler = com_ ## _cmd \
58 * Command line and active options. We need to keep a copy of the parsed
59 * command line options for the SIGHUP case where we merge the command line
60 * options and the new config file options.
62 static struct lls_parse_result *cmdline_lpr, *lpr;
64 /** Parsed subcommand options. */
65 static struct lls_parse_result *cmdline_sublpr, *sublpr;
66 /* The executing subcommand (NULL at startup). */
67 static const struct lls_command *subcmd;
68 /** Wether daemon_init() was called. */
69 static bool daemonized;
70 /** Non-NULL if we log to a file. */
72 /** The read end of the signal pipe */
73 static int signal_pipe;
74 /** Process id of current pre-create-hook/rsync/post-create-hook process. */
75 static pid_t create_pid;
76 /** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
77 static int create_process_stopped;
78 /** How many times in a row the rsync command failed. */
79 static int num_consecutive_rsync_errors;
80 /** Process id of current pre-remove/rm/post-remove process. */
81 static pid_t remove_pid;
82 /** When the next snapshot is due. */
83 static int64_t next_snapshot_time;
84 /** When to try to remove something. */
85 static struct timeval next_removal_check;
86 /** Creation time of the snapshot currently being created. */
87 static int64_t current_snapshot_creation_time;
88 /** The snapshot currently being removed. */
89 struct snapshot *snapshot_currently_being_removed;
90 /** Needed by the post-create hook. */
91 static char *path_to_last_complete_snapshot;
92 static char *name_of_reference_snapshot;
93 /** \sa \ref snap.h for details. */
94 enum hook_status snapshot_creation_status;
95 /** \sa \ref snap.h for details. */
96 enum hook_status snapshot_removal_status;
100 static const char *hook_status_description[] = {HOOK_STATUS_ARRAY};
102 /* may be called with ds == NULL. */
103 static int disk_space_low(struct disk_space *ds)
105 struct disk_space ds_struct;
109 int ret = get_disk_space(".", &ds_struct);
114 val = OPT_UINT32_VAL(DSS, MIN_FREE_MB);
116 if (ds->free_mb < val)
118 val = OPT_UINT32_VAL(DSS, MIN_FREE_PERCENT);
120 if (ds->percent_free < val)
122 val = OPT_UINT32_VAL(DSS, MIN_FREE_PERCENT_INODES);
124 if (ds->percent_free_inodes < val)
129 static void dump_dss_config(const char *msg)
131 const char dash[] = "-----------------------------";
134 FILE *log = logfile? logfile : stderr;
135 struct disk_space ds;
136 int64_t now = get_current_time();
138 if (OPT_UINT32_VAL(DSS, LOGLEVEL) > INFO)
141 fprintf(log, "%s <%s config> %s\n", dash, msg, dash);
142 fprintf(log, "\n*** disk space ***\n\n");
143 ret = get_disk_space(".", &ds);
145 DSS_INFO_LOG(("disk space low: %s\n", disk_space_low(&ds)?
149 DSS_ERROR_LOG(("can not get free disk space: %s\n",
150 dss_strerror(-ret)));
152 /* we continue on errors from get_disk_space */
154 fprintf(log, "\n*** non-default options ***\n\n");
155 lopsub_dump = lls_dump_parse_result(lpr, CMD_PTR(DSS), true);
156 fprintf(log, "%s", lopsub_dump);
158 fprintf(log, "\n*** non-default options for \"run\" ***\n\n");
159 lopsub_dump = lls_dump_parse_result(lpr, CMD_PTR(RUN), true);
160 fprintf(log, "%s", lopsub_dump);
162 fprintf(log, "\n*** internal state ***\n\n");
166 "snapshot_currently_being_removed: %s\n"
167 "path_to_last_complete_snapshot: %s\n"
168 "reference_snapshot: %s\n"
169 "snapshot_creation_status: %s\n"
170 "snapshot_removal_status: %s\n"
171 "num_consecutive_rsync_errors: %d\n"
174 logfile? OPT_STRING_VAL(RUN, LOGFILE) : "stderr",
175 snapshot_currently_being_removed?
176 snapshot_currently_being_removed->name : "(none)",
177 path_to_last_complete_snapshot?
178 path_to_last_complete_snapshot : "(none)",
179 name_of_reference_snapshot?
180 name_of_reference_snapshot : "(none)",
181 hook_status_description[snapshot_creation_status],
182 hook_status_description[snapshot_removal_status],
183 num_consecutive_rsync_errors
187 "create_pid: %" PRId32 "\n"
188 "create process is %sstopped\n"
191 create_process_stopped? "" : "not "
194 fprintf(log, "remove_pid: %" PRId32 "\n", remove_pid);
195 if (next_snapshot_time != 0)
196 fprintf(log, "next snapshot due in %" PRId64 " seconds\n",
197 next_snapshot_time - now);
198 if (current_snapshot_creation_time != 0)
199 fprintf(log, "current_snapshot_creation_time: %"
200 PRId64 " (%" PRId64 " seconds ago)\n",
201 current_snapshot_creation_time,
202 now - current_snapshot_creation_time
204 if (next_removal_check.tv_sec != 0) {
205 fprintf(log, "next removal check: %llu (%llu seconds ago)\n",
206 (long long unsigned)next_removal_check.tv_sec,
207 now - (long long unsigned)next_removal_check.tv_sec
211 fprintf(log, "%s </%s config> %s\n", dash, msg, dash);
214 static int loglevel = -1;
215 static const char *location_file = NULL;
216 static int location_line = -1;
217 static const char *location_func = NULL;
219 void dss_log_set_params(int ll, const char *file, int line, const char *func)
222 location_file = file;
223 location_line = line;
224 location_func = func;
228 * The log function of dss.
230 * \param ll Loglevel.
231 * \param fml Usual format string.
233 * All DSS_XXX_LOG() macros use this function.
235 __printf_1_2 void dss_log(const char* fmt,...)
242 int lpr_ll = lpr? OPT_UINT32_VAL(DSS, LOGLEVEL) : WARNING;
244 if (loglevel < lpr_ll)
246 outfd = logfile? logfile : stderr;
247 if (subcmd == CMD_PTR(RUN)) {
250 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
251 fprintf(outfd, "%s ", str);
253 fprintf(outfd, "%i: ", loglevel);
255 if (subcmd == CMD_PTR(RUN))
256 #ifdef DSS_NO_FUNC_NAMES
257 fprintf(outfd, "%s:%d: ", location_file, location_line);
259 fprintf(outfd, "%s: ", location_func);
262 vfprintf(outfd, fmt, argp);
267 * Print a message either to stdout or to the log file.
269 static __printf_1_2 void dss_msg(const char* fmt,...)
271 FILE *outfd = logfile? logfile : stdout;
274 vfprintf(outfd, fmt, argp);
278 static char *get_config_file_name(void)
280 char *home, *config_file;
282 if (OPT_GIVEN(DSS, CONFIG_FILE))
283 return dss_strdup(OPT_STRING_VAL(DSS, CONFIG_FILE));
284 home = get_homedir();
285 config_file = make_message("%s/.dssrc", home);
290 static int send_signal(int sig, bool wait)
293 char *config_file = get_config_file_name();
294 int ret = get_dss_pid(config_file, &pid);
301 if (OPT_GIVEN(DSS, DRY_RUN)) {
302 dss_msg("%d\n", (int)pid);
305 DSS_NOTICE_LOG(("sending signal %d to pid %d\n", sig, (int)pid));
306 ret = kill(pid, sig);
308 return -ERRNO_TO_DSS_ERROR(errno);
312 ts.tv_sec = ms / 1000;
313 ts.tv_nsec = (ms % 1000) * 1000 * 1000;
314 ret = nanosleep(&ts, NULL);
316 return -ERRNO_TO_DSS_ERROR(errno);
320 return -ERRNO_TO_DSS_ERROR(errno);
325 return -E_KILL_TIMEOUT;
329 const char * const name;
334 * The table below was taken 2016 from proc/sig.c of procps-3.2.8. Copyright
335 * 1998-2003 by Albert Cahalan, GPLv2.
337 static const struct signal_info signal_table[] = {
338 {"ABRT", SIGABRT}, /* IOT */
341 {"CHLD", SIGCHLD}, /* CLD */
350 {"POLL", SIGPOLL}, /* IO */
359 {"STKFLT", SIGSTKFLT},
362 {"SYS", SIGSYS}, /* UNUSED */
371 {"VTALRM", SIGVTALRM},
377 #define SIGNAL_TABLE_SIZE (sizeof(signal_table) / sizeof(signal_table[0]))
382 static int com_kill(void)
384 bool w_given = OPT_GIVEN(KILL, WAIT);
385 const char *arg = OPT_STRING_VAL(KILL, SIGNAL);
388 if (*arg >= '0' && *arg <= '9') {
390 ret = dss_atoi64(arg, &val);
393 if (val < 0 || val > SIGRTMAX)
394 return -ERRNO_TO_DSS_ERROR(EINVAL);
395 return send_signal(val, w_given);
397 if (strncasecmp(arg, "sig", 3) == 0)
399 if (strcasecmp(arg, "CLD") == 0)
400 return send_signal(SIGCHLD, w_given);
401 if (strcasecmp(arg, "IOT") == 0)
402 return send_signal(SIGABRT, w_given);
403 for (i = 0; i < SIGNAL_TABLE_SIZE; i++)
404 if (strcasecmp(arg, signal_table[i].name) == 0)
405 return send_signal(signal_table[i].num, w_given);
406 DSS_ERROR_LOG(("invalid sigspec: %s\n", arg));
407 return -ERRNO_TO_DSS_ERROR(EINVAL);
409 EXPORT_CMD_HANDLER(kill);
411 static void dss_get_snapshot_list(struct snapshot_list *sl)
413 get_snapshot_list(sl, OPT_UINT32_VAL(DSS, UNIT_INTERVAL),
414 OPT_UINT32_VAL(DSS, NUM_INTERVALS));
417 static int64_t compute_next_snapshot_time(void)
419 int64_t x = 0, now = get_current_time(), unit_interval
420 = 24 * 3600 * OPT_UINT32_VAL(DSS, UNIT_INTERVAL), ret;
421 unsigned wanted = desired_number_of_snapshots(0,
422 OPT_UINT32_VAL(DSS, NUM_INTERVALS)),
425 struct snapshot *s = NULL;
426 struct snapshot_list sl;
428 dss_get_snapshot_list(&sl);
429 FOR_EACH_SNAPSHOT(s, i, &sl) {
430 if (!(s->flags & SS_COMPLETE))
433 x += s->completion_time - s->creation_time;
438 if (num_complete == 0)
440 x /= num_complete; /* avg time to create one snapshot */
441 if (unit_interval < x * wanted) /* oops, no sleep at all */
443 ret = s->completion_time + unit_interval / wanted - x;
445 free_snapshot_list(&sl);
449 static inline void invalidate_next_snapshot_time(void)
451 next_snapshot_time = 0;
454 static inline int next_snapshot_time_is_valid(void)
456 return next_snapshot_time != 0;
459 static int next_snapshot_is_due(void)
461 int64_t now = get_current_time();
463 if (!next_snapshot_time_is_valid())
464 next_snapshot_time = compute_next_snapshot_time();
465 if (next_snapshot_time <= now) {
466 DSS_DEBUG_LOG(("next snapshot: now\n"));
469 DSS_DEBUG_LOG(("next snapshot due in %" PRId64 " seconds\n",
470 next_snapshot_time - now));
474 static void pre_create_hook(void)
476 assert(snapshot_creation_status == HS_READY);
477 /* make sure that the next snapshot time will be recomputed */
478 invalidate_next_snapshot_time();
479 DSS_DEBUG_LOG(("executing %s\n", OPT_STRING_VAL(DSS, PRE_CREATE_HOOK)));
480 dss_exec_cmdline_pid(&create_pid, OPT_STRING_VAL(DSS, PRE_CREATE_HOOK));
481 snapshot_creation_status = HS_PRE_RUNNING;
484 static void pre_remove_hook(struct snapshot *s, const char *why)
490 DSS_DEBUG_LOG(("%s snapshot %s\n", why, s->name));
491 assert(snapshot_removal_status == HS_READY);
492 assert(remove_pid == 0);
493 assert(!snapshot_currently_being_removed);
495 snapshot_currently_being_removed = dss_malloc(sizeof(struct snapshot));
496 *snapshot_currently_being_removed = *s;
497 snapshot_currently_being_removed->name = dss_strdup(s->name);
499 cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, PRE_REMOVE_HOOK),
500 OPT_STRING_VAL(DSS, DEST_DIR), s->name);
501 DSS_DEBUG_LOG(("executing %s\n", cmd));
502 dss_exec_cmdline_pid(&remove_pid, cmd);
504 snapshot_removal_status = HS_PRE_RUNNING;
507 static int exec_rm(void)
509 struct snapshot *s = snapshot_currently_being_removed;
510 char *new_name = being_deleted_name(s);
519 assert(snapshot_removal_status == HS_PRE_SUCCESS);
520 assert(remove_pid == 0);
522 DSS_NOTICE_LOG(("removing %s (interval = %i)\n", s->name, s->interval));
523 ret = dss_rename(s->name, new_name);
526 dss_exec(&remove_pid, argv[0], argv);
527 snapshot_removal_status = HS_RUNNING;
533 static int snapshot_is_being_created(struct snapshot *s)
535 return s->creation_time == current_snapshot_creation_time;
538 static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
543 DSS_DEBUG_LOG(("looking for old incomplete snapshots\n"));
544 FOR_EACH_SNAPSHOT(s, i, sl) {
545 if (snapshot_is_being_created(s))
548 * We know that no rm is currently running, so if s is marked
549 * as being deleted, a previously started rm must have failed.
551 if (s->flags & SS_BEING_DELETED)
554 if (s->flags & SS_COMPLETE) /* good snapshot */
557 * This snapshot is incomplete and it is not the snapshot
558 * currently being created. However, we must not remove it if
559 * rsync is about to be restarted. As only the newest snapshot
560 * can be restarted, this snapshot is orphaned if it is not the
561 * newest snapshot or if we are not about to restart rsync.
563 if (get_newest_snapshot(sl) != s)
565 if (snapshot_creation_status != HS_NEEDS_RESTART)
568 /* no orphaned snapshots */
572 static int is_reference_snapshot(struct snapshot *s)
574 if (!name_of_reference_snapshot)
576 return strcmp(s->name, name_of_reference_snapshot)? 0 : 1;
580 * return: 0: no redundant snapshots, 1: rm process started, negative: error
582 static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
586 unsigned missing = 0;
587 uint32_t N = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
589 DSS_DEBUG_LOG(("looking for intervals containing too many snapshots\n"));
590 for (interval = N - 1; interval >= 0; interval--) {
591 unsigned keep = desired_number_of_snapshots(interval, N);
592 unsigned num = sl->interval_count[interval];
593 struct snapshot *victim = NULL, *prev = NULL;
594 int64_t score = LONG_MAX;
597 missing += keep - num;
598 if (keep + missing >= num)
600 /* redundant snapshot in this interval, pick snapshot with lowest score */
601 FOR_EACH_SNAPSHOT(s, i, sl) {
604 if (snapshot_is_being_created(s))
606 if (is_reference_snapshot(s))
608 if (s->interval > interval) {
612 if (s->interval < interval)
620 /* check if s is a better victim */
621 this_score = s->creation_time - prev->creation_time;
622 assert(this_score >= 0);
623 if (this_score < score) {
635 static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
640 DSS_DEBUG_LOG(("looking for snapshots belonging to intervals >= %d\n",
641 OPT_UINT32_VAL(DSS, NUM_INTERVALS)));
642 FOR_EACH_SNAPSHOT(s, i, sl) {
643 if (snapshot_is_being_created(s))
645 if (is_reference_snapshot(s))
647 if (s->interval < OPT_UINT32_VAL(DSS, NUM_INTERVALS))
654 static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
657 struct snapshot *s, *ref = NULL;
659 num_complete = num_complete_snapshots(sl);
660 if (num_complete <= OPT_UINT32_VAL(DSS, MIN_COMPLETE))
662 FOR_EACH_SNAPSHOT(s, i, sl) {
663 if (snapshot_is_being_created(s))
665 if (is_reference_snapshot(s)) { /* avoid this one */
669 DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
673 DSS_WARNING_LOG(("removing reference snapshot %s\n", ref->name));
677 static int rename_incomplete_snapshot(int64_t start)
684 * We don't want the dss_rename() below to fail with EEXIST because the
685 * last complete snapshot was created (and completed) in the same
686 * second as this one.
688 while ((now = get_current_time()) == start)
690 free(path_to_last_complete_snapshot);
691 ret = complete_name(start, now, &path_to_last_complete_snapshot);
694 old_name = incomplete_name(start);
695 ret = dss_rename(old_name, path_to_last_complete_snapshot);
697 DSS_NOTICE_LOG(("%s -> %s\n", old_name,
698 path_to_last_complete_snapshot));
703 static int try_to_free_disk_space(void)
706 struct snapshot_list sl;
707 struct snapshot *victim;
712 ret = disk_space_low(NULL);
715 low_disk_space = ret;
716 gettimeofday(&now, NULL);
717 if (tv_diff(&next_removal_check, &now, NULL) > 0)
719 if (!low_disk_space) {
720 if (OPT_GIVEN(DSS, KEEP_REDUNDANT))
722 if (snapshot_creation_status != HS_READY)
724 if (next_snapshot_is_due())
728 * Idle and --keep_redundant not given, or low disk space. Look at
729 * existing snapshots.
731 dss_get_snapshot_list(&sl);
734 * Don't remove anything if there is free space and we have fewer
735 * snapshots than configured, plus one. This way there is always one
736 * snapshot that can be recycled.
738 if (!low_disk_space && sl.num_snapshots <=
739 1 << OPT_UINT32_VAL(DSS, NUM_INTERVALS))
742 victim = find_outdated_snapshot(&sl);
746 victim = find_redundant_snapshot(&sl);
750 victim = find_orphaned_snapshot(&sl);
753 /* try harder only if disk space is low */
756 DSS_WARNING_LOG(("disk space low and nothing obvious to remove\n"));
758 victim = find_oldest_removable_snapshot(&sl);
761 DSS_CRIT_LOG(("uhuhu: disk space low and nothing to remove\n"));
762 ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
765 pre_remove_hook(victim, why);
767 free_snapshot_list(&sl);
771 static void post_create_hook(void)
773 char *cmd = make_message("%s %s/%s",
774 OPT_STRING_VAL(DSS, POST_CREATE_HOOK),
775 OPT_STRING_VAL(DSS, DEST_DIR), path_to_last_complete_snapshot);
776 DSS_NOTICE_LOG(("executing %s\n", cmd));
777 dss_exec_cmdline_pid(&create_pid, cmd);
779 snapshot_creation_status = HS_POST_RUNNING;
782 static void post_remove_hook(void)
785 struct snapshot *s = snapshot_currently_being_removed;
789 cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, POST_REMOVE_HOOK),
790 OPT_STRING_VAL(DSS, DEST_DIR), s->name);
791 DSS_NOTICE_LOG(("executing %s\n", cmd));
792 dss_exec_cmdline_pid(&remove_pid, cmd);
794 snapshot_removal_status = HS_POST_RUNNING;
797 static void dss_kill(pid_t pid, int sig, const char *msg)
799 const char *signame, *process_name;
804 case SIGTERM: signame = "TERM"; break;
805 case SIGSTOP: signame = "STOP"; break;
806 case SIGCONT: signame = "CONT"; break;
807 default: signame = "????";
810 if (pid == create_pid)
811 process_name = "create";
812 else if (pid == remove_pid)
813 process_name = "remove";
814 else process_name = "??????";
817 DSS_INFO_LOG(("%s\n", msg));
818 DSS_DEBUG_LOG(("sending signal %d (%s) to pid %d (%s process)\n",
819 sig, signame, (int)pid, process_name));
820 if (kill(pid, sig) >= 0)
822 DSS_INFO_LOG(("failed to send signal %d (%s) to pid %d (%s process)\n",
823 sig, signame, (int)pid, process_name));
826 static void stop_create_process(void)
828 if (create_process_stopped)
830 dss_kill(create_pid, SIGSTOP, "suspending create process");
831 create_process_stopped = 1;
834 static void restart_create_process(void)
836 if (!create_process_stopped)
838 dss_kill(create_pid, SIGCONT, "resuming create process");
839 create_process_stopped = 0;
843 * Print a log message about the exit status of a child.
845 static void log_termination_msg(pid_t pid, int status)
847 if (WIFEXITED(status))
848 DSS_INFO_LOG(("child %i exited. Exit status: %i\n", (int)pid,
849 WEXITSTATUS(status)));
850 else if (WIFSIGNALED(status))
851 DSS_NOTICE_LOG(("child %i was killed by signal %i\n", (int)pid,
854 DSS_WARNING_LOG(("child %i terminated abormally\n", (int)pid));
857 static int wait_for_process(pid_t pid, int *status)
861 DSS_DEBUG_LOG(("Waiting for process %d to terminate\n", (int)pid));
866 FD_SET(signal_pipe, &rfds);
867 ret = dss_select(signal_pipe + 1, &rfds, NULL, NULL);
873 if (ret == SIGCHLD) {
874 ret = waitpid(pid, status, 0);
877 if (errno != EINTR) { /* error */
878 ret = -ERRNO_TO_DSS_ERROR(errno);
882 /* SIGINT or SIGTERM */
883 dss_kill(pid, SIGTERM, "killing child process");
886 DSS_ERROR_LOG(("failed to wait for process %d\n", (int)pid));
888 log_termination_msg(pid, *status);
892 static void handle_pre_remove_exit(int status)
894 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
895 snapshot_removal_status = HS_READY;
896 gettimeofday(&next_removal_check, NULL);
897 next_removal_check.tv_sec += 60;
900 snapshot_removal_status = HS_PRE_SUCCESS;
903 static int handle_rm_exit(int status)
905 if (!WIFEXITED(status)) {
906 snapshot_removal_status = HS_READY;
907 return -E_INVOLUNTARY_EXIT;
909 if (WEXITSTATUS(status)) {
910 snapshot_removal_status = HS_READY;
911 return -E_BAD_EXIT_CODE;
913 snapshot_removal_status = HS_SUCCESS;
917 static void handle_post_remove_exit(void)
919 snapshot_removal_status = HS_READY;
922 static int handle_remove_exit(int status)
925 struct snapshot *s = snapshot_currently_being_removed;
928 switch (snapshot_removal_status) {
930 handle_pre_remove_exit(status);
934 ret = handle_rm_exit(status);
936 case HS_POST_RUNNING:
937 handle_post_remove_exit();
943 if (snapshot_removal_status == HS_READY) {
946 snapshot_currently_being_removed = NULL;
952 static int wait_for_remove_process(void)
958 snapshot_removal_status == HS_PRE_RUNNING ||
959 snapshot_removal_status == HS_RUNNING ||
960 snapshot_removal_status == HS_POST_RUNNING
962 ret = wait_for_process(remove_pid, &status);
965 return handle_remove_exit(status);
968 static int handle_rsync_exit(int status)
972 if (!WIFEXITED(status)) {
973 DSS_ERROR_LOG(("rsync process %d died involuntary\n", (int)create_pid));
974 ret = -E_INVOLUNTARY_EXIT;
975 snapshot_creation_status = HS_READY;
978 es = WEXITSTATUS(status);
980 * Restart rsync on non-fatal errors:
981 * 24: Partial transfer due to vanished source files
983 if (es != 0 && es != 24) {
984 DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
985 es, ++num_consecutive_rsync_errors));
986 if (!logfile) { /* called by com_run() */
987 ret = -E_BAD_EXIT_CODE;
990 if (num_consecutive_rsync_errors >
991 OPT_UINT32_VAL(RUN, MAX_RSYNC_ERRORS)) {
992 ret = -E_TOO_MANY_RSYNC_ERRORS;
993 snapshot_creation_status = HS_READY;
996 DSS_WARNING_LOG(("restarting rsync process\n"));
997 snapshot_creation_status = HS_NEEDS_RESTART;
998 next_snapshot_time = get_current_time() + 60;
1002 num_consecutive_rsync_errors = 0;
1003 ret = rename_incomplete_snapshot(current_snapshot_creation_time);
1006 snapshot_creation_status = HS_SUCCESS;
1007 free(name_of_reference_snapshot);
1008 name_of_reference_snapshot = NULL;
1010 create_process_stopped = 0;
1014 static int handle_pre_create_hook_exit(int status)
1017 static int warn_count;
1019 if (!WIFEXITED(status)) {
1020 snapshot_creation_status = HS_READY;
1021 ret = -E_INVOLUNTARY_EXIT;
1024 es = WEXITSTATUS(status);
1026 if (!warn_count--) {
1027 DSS_NOTICE_LOG(("pre_create_hook %s returned %d\n",
1028 OPT_STRING_VAL(DSS, PRE_CREATE_HOOK), es));
1029 DSS_NOTICE_LOG(("deferring snapshot creation...\n"));
1030 warn_count = 60; /* warn only once per hour */
1032 next_snapshot_time = get_current_time() + 60;
1033 snapshot_creation_status = HS_READY;
1038 snapshot_creation_status = HS_PRE_SUCCESS;
1044 static int handle_sigchld(void)
1047 int status, ret = reap_child(&pid, &status);
1052 if (pid == create_pid) {
1053 switch (snapshot_creation_status) {
1054 case HS_PRE_RUNNING:
1055 ret = handle_pre_create_hook_exit(status);
1058 ret = handle_rsync_exit(status);
1060 case HS_POST_RUNNING:
1061 snapshot_creation_status = HS_READY;
1065 DSS_EMERG_LOG(("BUG: create can't die in status %d\n",
1066 snapshot_creation_status));
1072 if (pid == remove_pid) {
1073 ret = handle_remove_exit(status);
1078 DSS_EMERG_LOG(("BUG: unknown process %d died\n", (int)pid));
1082 /* also checks if . is a mountpoint, if --mountpoint was given */
1083 static int change_to_dest_dir(void)
1086 const char *dd = OPT_STRING_VAL(DSS, DEST_DIR);
1087 struct stat dot, dotdot;
1089 DSS_INFO_LOG(("changing cwd to %s\n", dd));
1090 if (chdir(dd) < 0) {
1091 ret = -ERRNO_TO_DSS_ERROR(errno);
1092 DSS_ERROR_LOG(("could not change cwd to %s\n", dd));
1095 if (!OPT_GIVEN(DSS, MOUNTPOINT))
1097 if (stat(".", &dot) < 0) {
1098 ret = -ERRNO_TO_DSS_ERROR(errno);
1099 DSS_ERROR_LOG(("could not stat .\n"));
1102 if (stat("..", &dotdot) < 0) {
1103 ret = -ERRNO_TO_DSS_ERROR(errno);
1104 DSS_ERROR_LOG(("could not stat ..\n"));
1107 if (dot.st_dev == dotdot.st_dev && dot.st_ino != dotdot.st_ino) {
1108 DSS_ERROR_LOG(("mountpoint check failed for %s\n", dd));
1109 return -E_MOUNTPOINT;
1114 static int check_config(void)
1117 uint32_t unit_interval = OPT_UINT32_VAL(DSS, UNIT_INTERVAL);
1118 uint32_t num_intervals = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
1120 if (unit_interval == 0) {
1121 DSS_ERROR_LOG(("bad unit interval: %i\n", unit_interval));
1122 return -E_INVALID_NUMBER;
1124 DSS_DEBUG_LOG(("unit interval: %i day(s)\n", unit_interval));
1126 if (num_intervals == 0 || num_intervals > 30) {
1127 DSS_ERROR_LOG(("bad number of intervals: %i\n", num_intervals));
1128 return -E_INVALID_NUMBER;
1130 if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE))
1131 if (!OPT_GIVEN(DSS, SOURCE_DIR)) {
1132 DSS_ERROR_LOG(("--source-dir required\n"));
1135 if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE)
1136 || subcmd == CMD_PTR(LS) || subcmd == CMD_PTR(PRUNE)) {
1137 if (!OPT_GIVEN(DSS, DEST_DIR)) {
1138 DSS_ERROR_LOG(("--dest-dir required\n"));
1141 ret = change_to_dest_dir();
1145 DSS_DEBUG_LOG(("number of intervals: %i\n", num_intervals));
1149 static int lopsub_error(int lopsub_ret, char **errctx)
1151 const char *msg = lls_strerror(-lopsub_ret);
1153 DSS_ERROR_LOG(("%s: %s\n", *errctx, msg));
1155 DSS_ERROR_LOG(("%s\n", msg));
1161 static int parse_config_file(bool sighup, const struct lls_command *cmd)
1164 char *config_file = get_config_file_name();
1165 struct stat statbuf;
1169 char **cf_argv, *errctx = NULL;
1170 struct lls_parse_result *cf_lpr, *merged_lpr, *clpr;
1171 const char *subcmd_name;
1173 ret = open(config_file, O_RDONLY);
1175 if (errno != ENOENT || OPT_GIVEN(DSS, CONFIG_FILE)) {
1176 ret = -ERRNO_TO_DSS_ERROR(errno);
1177 DSS_ERROR_LOG(("config file %s can not be opened\n",
1181 /* no config file -- nothing to do */
1186 ret = fstat(fd, &statbuf);
1188 ret = -ERRNO_TO_DSS_ERROR(errno);
1189 DSS_ERROR_LOG(("failed to stat config file %s\n", config_file));
1192 sz = statbuf.st_size;
1193 if (sz == 0) { /* config file is empty -- nothing to do */
1197 map = mmap(NULL, sz, PROT_READ, MAP_PRIVATE, fd, 0);
1198 if (map == MAP_FAILED) {
1199 ret = -ERRNO_TO_DSS_ERROR(errno);
1200 DSS_ERROR_LOG(("failed to mmap config file %s\n",
1204 if (cmd == CMD_PTR(DSS))
1207 subcmd_name = lls_command_name(cmd);
1208 ret = lls_convert_config(map, sz, subcmd_name, &cf_argv, &errctx);
1211 DSS_ERROR_LOG(("failed to convert config file %s\n",
1213 ret = lopsub_error(ret, &errctx);
1217 ret = lls_parse(cf_argc, cf_argv, cmd, &cf_lpr, &errctx);
1218 lls_free_argv(cf_argv);
1220 ret = lopsub_error(ret, &errctx);
1223 clpr = cmd == CMD_PTR(DSS)? cmdline_lpr : cmdline_sublpr;
1224 if (sighup) /* config file overrides command line */
1225 ret = lls_merge(cf_lpr, clpr, cmd, &merged_lpr, &errctx);
1226 else /* command line options overrride config file options */
1227 ret = lls_merge(clpr, cf_lpr, cmd, &merged_lpr, &errctx);
1228 lls_free_parse_result(cf_lpr, cmd);
1230 ret = lopsub_error(ret, &errctx);
1236 DSS_DEBUG_LOG(("loglevel: %d\n", OPT_UINT32_VAL(DSS, LOGLEVEL)));
1237 if (cmd != CMD_PTR(DSS)) {
1239 if (sublpr != cmdline_sublpr)
1240 lls_free_parse_result(sublpr, cmd);
1241 sublpr = merged_lpr;
1243 sublpr = cmdline_sublpr;
1246 if (lpr != cmdline_lpr)
1247 lls_free_parse_result(lpr, cmd);
1258 DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
1262 static int handle_sighup(void)
1266 DSS_NOTICE_LOG(("SIGHUP, re-reading config\n"));
1267 dump_dss_config("old");
1268 ret = parse_config_file(true /* SIGHUP */, CMD_PTR(DSS));
1271 ret = parse_config_file(true /* SIGHUP */, CMD_PTR(RUN));
1274 ret = check_config();
1279 if (OPT_GIVEN(RUN, DAEMON) || daemonized) {
1280 logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
1281 log_welcome(OPT_UINT32_VAL(DSS, LOGLEVEL));
1284 dump_dss_config("reloaded");
1285 invalidate_next_snapshot_time();
1289 static void kill_children(void)
1291 restart_create_process();
1292 dss_kill(create_pid, SIGTERM, NULL);
1293 dss_kill(remove_pid, SIGTERM, NULL);
1296 static int handle_signal(void)
1298 int sig, ret = next_signal();
1310 ret = handle_sighup();
1313 ret = handle_sigchld();
1318 DSS_ERROR_LOG(("%s\n", dss_strerror(-ret)));
1323 * We can not use rsync locally if the local user is different from the remote
1324 * user or if the src dir is not on the local host (or both).
1326 static int use_rsync_locally(char *logname)
1328 const char *h = OPT_STRING_VAL(DSS, REMOTE_HOST);
1330 if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
1332 if (OPT_GIVEN(DSS, REMOTE_USER) &&
1333 strcmp(OPT_STRING_VAL(DSS, REMOTE_USER), logname))
1338 static int rename_resume_snap(int64_t creation_time)
1340 struct snapshot_list sl;
1341 struct snapshot *s = NULL;
1342 char *new_name = incomplete_name(creation_time);
1346 sl.num_snapshots = 0;
1349 dss_get_snapshot_list(&sl);
1351 * Snapshot recycling: We first look at the newest snapshot. If this
1352 * snapshot happens to be incomplete, the last rsync process was
1353 * aborted and we reuse this one. Otherwise we look at snapshots which
1354 * could be removed (outdated and redundant snapshots) as candidates
1355 * for recycling. If no outdated/redundant snapshot exists, we check if
1356 * there is an orphaned snapshot, which likely is useless anyway.
1358 * Only if no existing snapshot is suitable for recycling, we bite the
1359 * bullet and create a new one.
1361 s = get_newest_snapshot(&sl);
1362 if (!s) /* no snapshots at all */
1364 /* re-use last snapshot if it is incomplete */
1366 if ((s->flags & SS_COMPLETE) == 0)
1369 s = find_outdated_snapshot(&sl);
1373 s = find_redundant_snapshot(&sl);
1377 s = find_orphaned_snapshot(&sl);
1380 DSS_NOTICE_LOG(("recycling %s snapshot %s\n", why, s->name));
1381 ret = dss_rename(s->name, new_name);
1384 DSS_NOTICE_LOG(("creating %s\n", new_name));
1386 free_snapshot_list(&sl);
1390 static void create_rsync_argv(char ***argv, int64_t *num)
1393 int i = 0, j, N = OPT_GIVEN(DSS, RSYNC_OPTION);
1394 struct snapshot_list sl;
1397 dss_get_snapshot_list(&sl);
1398 assert(!name_of_reference_snapshot);
1399 name_of_reference_snapshot = name_of_newest_complete_snapshot(&sl);
1400 free_snapshot_list(&sl);
1402 *argv = dss_malloc((15 + N) * sizeof(char *));
1403 (*argv)[i++] = dss_strdup("rsync");
1404 (*argv)[i++] = dss_strdup("-a");
1405 (*argv)[i++] = dss_strdup("--delete");
1407 srandom((unsigned)time(NULL)); /* no need to be fancy here */
1410 if (1000 * (random() / (RAND_MAX + 1.0)) < OPT_UINT32_VAL(DSS, CHECKSUM)) {
1411 DSS_NOTICE_LOG(("adding --checksum to rsync options\n"));
1412 (*argv)[i++] = dss_strdup("--checksum");
1414 for (j = 0; j < N; j++)
1415 (*argv)[i++] = dss_strdup(lls_string_val(j,
1416 OPT_RESULT(DSS, RSYNC_OPTION)));
1417 if (name_of_reference_snapshot) {
1418 DSS_INFO_LOG(("using %s as reference\n", name_of_reference_snapshot));
1419 (*argv)[i++] = make_message("--link-dest=../%s",
1420 name_of_reference_snapshot);
1422 DSS_INFO_LOG(("no suitable reference snapshot found\n"));
1423 logname = dss_logname();
1424 if (use_rsync_locally(logname))
1425 (*argv)[i++] = dss_strdup(OPT_STRING_VAL(DSS, SOURCE_DIR));
1427 (*argv)[i++] = make_message("%s@%s:%s/",
1428 OPT_GIVEN(DSS, REMOTE_USER)?
1429 OPT_STRING_VAL(DSS, REMOTE_USER) : logname,
1430 OPT_STRING_VAL(DSS, REMOTE_HOST),
1431 OPT_STRING_VAL(DSS, SOURCE_DIR));
1433 *num = get_current_time();
1434 (*argv)[i++] = incomplete_name(*num);
1435 (*argv)[i++] = NULL;
1436 for (j = 0; j < i; j++)
1437 DSS_DEBUG_LOG(("argv[%d] = %s\n", j, (*argv)[j]));
1440 static void free_rsync_argv(char **argv)
1446 for (i = 0; argv[i]; i++)
1451 static int create_snapshot(char **argv)
1455 ret = rename_resume_snap(current_snapshot_creation_time);
1458 dss_exec(&create_pid, argv[0], argv);
1459 snapshot_creation_status = HS_RUNNING;
1463 static int select_loop(void)
1466 /* check every 60 seconds for free disk space */
1468 char **rsync_argv = NULL;
1472 struct timeval *tvp;
1475 tvp = NULL; /* sleep until rm hook/process dies */
1476 else { /* sleep one minute */
1482 FD_SET(signal_pipe, &rfds);
1483 ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
1486 if (FD_ISSET(signal_pipe, &rfds)) {
1487 ret = handle_signal();
1493 if (snapshot_removal_status == HS_PRE_SUCCESS) {
1499 if (snapshot_removal_status == HS_SUCCESS) {
1503 ret = try_to_free_disk_space();
1506 if (snapshot_removal_status != HS_READY) {
1507 stop_create_process();
1510 restart_create_process();
1511 switch (snapshot_creation_status) {
1513 if (!next_snapshot_is_due())
1517 case HS_PRE_RUNNING:
1519 case HS_POST_RUNNING:
1521 case HS_PRE_SUCCESS:
1522 if (!name_of_reference_snapshot) {
1523 free_rsync_argv(rsync_argv);
1524 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1526 ret = create_snapshot(rsync_argv);
1530 case HS_NEEDS_RESTART:
1531 if (!next_snapshot_is_due())
1533 ret = create_snapshot(rsync_argv);
1546 static void exit_hook(int exit_code)
1549 char **argv, *tmp = dss_strdup(OPT_STRING_VAL(DSS, EXIT_HOOK));
1550 unsigned n = split_args(tmp, &argv, " \t");
1553 argv = dss_realloc(argv, (n + 1) * sizeof(char *));
1554 argv[n - 1] = dss_strdup(dss_strerror(-exit_code));
1556 dss_exec(&pid, argv[0], argv);
1562 static void lock_dss_or_die(void)
1564 char *config_file = get_config_file_name();
1565 int ret = lock_dss(config_file);
1569 DSS_EMERG_LOG(("failed to lock: %s\n", dss_strerror(-ret)));
1574 static int com_run(void)
1580 if (OPT_GIVEN(DSS, DRY_RUN)) {
1581 DSS_ERROR_LOG(("dry run not supported by this command\n"));
1584 config_file = get_config_file_name();
1585 ret = get_dss_pid(config_file, &pid);
1588 DSS_ERROR_LOG(("pid %d\n", (int)pid));
1589 return -E_ALREADY_RUNNING;
1591 if (OPT_GIVEN(RUN, DAEMON)) {
1594 logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
1597 dump_dss_config("startup");
1598 ret = install_sighandler(SIGHUP);
1602 ret = write(fd, "\0", 1);
1604 DSS_ERROR_LOG(("write to daemon pipe returned %d\n",
1607 return -ERRNO_TO_DSS_ERROR(errno);
1611 ret = select_loop();
1612 if (ret >= 0) /* impossible */
1616 while (wait(NULL) >= 0 || errno != ECHILD)
1617 ; /* still have children to wait for */
1620 EXPORT_CMD_HANDLER(run);
1622 static int com_prune(void)
1625 struct snapshot_list sl;
1626 struct snapshot *victim;
1627 struct disk_space ds;
1631 ret = get_disk_space(".", &ds);
1634 log_disk_space(&ds);
1635 dss_get_snapshot_list(&sl);
1637 victim = find_outdated_snapshot(&sl);
1641 victim = find_redundant_snapshot(&sl);
1647 if (OPT_GIVEN(DSS, DRY_RUN)) {
1648 dss_msg("%s snapshot %s (interval = %i)\n",
1649 why, victim->name, victim->interval);
1653 pre_remove_hook(victim, why);
1654 if (snapshot_removal_status == HS_PRE_RUNNING) {
1655 ret = wait_for_remove_process();
1658 if (snapshot_removal_status != HS_PRE_SUCCESS)
1664 ret = wait_for_remove_process();
1667 if (snapshot_removal_status != HS_SUCCESS)
1670 if (snapshot_removal_status != HS_POST_RUNNING)
1672 ret = wait_for_remove_process();
1677 free_snapshot_list(&sl);
1680 EXPORT_CMD_HANDLER(prune);
1682 static int com_create(void)
1688 if (OPT_GIVEN(DSS, DRY_RUN)) {
1691 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1692 for (i = 0; rsync_argv[i]; i++) {
1694 msg = make_message("%s%s%s", tmp? tmp : "",
1695 tmp? " " : "", rsync_argv[i]);
1698 free_rsync_argv(rsync_argv);
1699 dss_msg("%s\n", msg);
1705 ret = wait_for_process(create_pid, &status);
1708 ret = handle_pre_create_hook_exit(status);
1709 if (ret <= 0) /* error, or pre-create failed */
1712 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1713 ret = create_snapshot(rsync_argv);
1716 ret = wait_for_process(create_pid, &status);
1719 ret = handle_rsync_exit(status);
1724 ret = wait_for_process(create_pid, &status);
1726 free_rsync_argv(rsync_argv);
1729 EXPORT_CMD_HANDLER(create);
1731 static int com_ls(void)
1734 struct snapshot_list sl;
1736 int64_t now = get_current_time();
1738 dss_get_snapshot_list(&sl);
1739 FOR_EACH_SNAPSHOT(s, i, &sl) {
1741 if (s->flags & SS_COMPLETE)
1742 d = (s->completion_time - s->creation_time) / 60;
1744 d = (now - s->creation_time) / 60;
1745 dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval,
1746 s->name, d / 60, d % 60);
1748 free_snapshot_list(&sl);
1751 EXPORT_CMD_HANDLER(ls);
1753 static int com_configtest(void)
1755 printf("Syntax Ok\n");
1758 EXPORT_CMD_HANDLER(configtest);
1760 static int setup_signal_handling(void)
1764 DSS_INFO_LOG(("setting up signal handlers\n"));
1765 signal_pipe = signal_init(); /* always successful */
1766 ret = install_sighandler(SIGINT);
1769 ret = install_sighandler(SIGTERM);
1772 return install_sighandler(SIGCHLD);
1775 static void handle_version_and_help(void)
1779 if (OPT_GIVEN(DSS, DETAILED_HELP))
1780 txt = lls_long_help(CMD_PTR(DSS));
1781 else if (OPT_GIVEN(DSS, HELP))
1782 txt = lls_short_help(CMD_PTR(DSS));
1783 else if (OPT_GIVEN(DSS, VERSION))
1784 txt = dss_strdup(VERSION_STRING);
1792 static void show_subcommand_summary(void)
1794 const struct lls_command *cmd;
1797 printf("Available subcommands:\n");
1798 for (i = 1; (cmd = lls_cmd(i, dss_suite)); i++) {
1799 const char *name = lls_command_name(cmd);
1800 const char *purpose = lls_purpose(cmd);
1801 printf("%-11s%s\n", name, purpose);
1806 int main(int argc, char **argv)
1809 char *errctx = NULL;
1810 unsigned num_inputs;
1811 const struct dss_user_data *ud;
1813 ret = lls_parse(argc, argv, CMD_PTR(DSS), &cmdline_lpr, &errctx);
1815 ret = lopsub_error(ret, &errctx);
1819 ret = parse_config_file(false /* no SIGHUP */, CMD_PTR(DSS));
1822 handle_version_and_help();
1823 num_inputs = lls_num_inputs(lpr);
1824 if (num_inputs == 0)
1825 show_subcommand_summary();
1826 ret = lls_lookup_subcmd(argv[argc - num_inputs], dss_suite, &errctx);
1828 ret = lopsub_error(ret, &errctx);
1831 subcmd = lls_cmd(ret, dss_suite);
1832 ret = lls_parse(num_inputs, argv + argc - num_inputs, subcmd,
1833 &cmdline_sublpr, &errctx);
1835 ret = lopsub_error(ret, &errctx);
1838 sublpr = cmdline_sublpr;
1839 ret = parse_config_file(false /* no SIGHUP */, subcmd);
1842 ret = check_config();
1845 ret = setup_signal_handling();
1848 ud = lls_user_data(subcmd);
1849 ret = ud->handler();
1854 DSS_ERROR_LOG(("%s\n", errctx));
1855 DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
1858 lls_free_parse_result(lpr, CMD_PTR(DSS));
1859 if (lpr != cmdline_lpr)
1860 lls_free_parse_result(cmdline_lpr, CMD_PTR(DSS));
1861 lls_free_parse_result(sublpr, subcmd);
1862 if (sublpr != cmdline_sublpr)
1863 lls_free_parse_result(cmdline_sublpr, subcmd);
1864 exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);