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)
293 char *config_file = get_config_file_name();
294 int ret = get_dss_pid(config_file, &pid);
299 if (OPT_GIVEN(DSS, DRY_RUN)) {
300 dss_msg("%d\n", (int)pid);
303 ret = kill(pid, sig);
305 return -ERRNO_TO_DSS_ERROR(errno);
310 const char * const name;
315 * The table below was taken 2016 from proc/sig.c of procps-3.2.8. Copyright
316 * 1998-2003 by Albert Cahalan, GPLv2.
318 static const struct signal_info signal_table[] = {
319 {"ABRT", SIGABRT}, /* IOT */
322 {"CHLD", SIGCHLD}, /* CLD */
331 {"POLL", SIGPOLL}, /* IO */
340 {"STKFLT", SIGSTKFLT},
343 {"SYS", SIGSYS}, /* UNUSED */
352 {"VTALRM", SIGVTALRM},
358 #define SIGNAL_TABLE_SIZE (sizeof(signal_table) / sizeof(signal_table[0]))
363 static int com_kill(void)
365 const char *arg = OPT_STRING_VAL(KILL, SIGNAL);
368 if (*arg >= '0' && *arg <= '9') {
370 ret = dss_atoi64(arg, &val);
373 if (val < 0 || val > SIGRTMAX)
374 return -ERRNO_TO_DSS_ERROR(EINVAL);
375 return send_signal(val);
377 if (strncasecmp(arg, "sig", 3) == 0)
379 if (strcasecmp(arg, "CLD") == 0)
380 return send_signal(SIGCHLD);
381 if (strcasecmp(arg, "IOT") == 0)
382 return send_signal(SIGABRT);
383 for (i = 0; i < SIGNAL_TABLE_SIZE; i++)
384 if (strcasecmp(arg, signal_table[i].name) == 0)
385 return send_signal(signal_table[i].num);
386 DSS_ERROR_LOG(("invalid sigspec: %s\n", arg));
387 return -ERRNO_TO_DSS_ERROR(EINVAL);
389 EXPORT_CMD_HANDLER(kill);
391 static void dss_get_snapshot_list(struct snapshot_list *sl)
393 get_snapshot_list(sl, OPT_UINT32_VAL(DSS, UNIT_INTERVAL),
394 OPT_UINT32_VAL(DSS, NUM_INTERVALS));
397 static int64_t compute_next_snapshot_time(void)
399 int64_t x = 0, now = get_current_time(), unit_interval
400 = 24 * 3600 * OPT_UINT32_VAL(DSS, UNIT_INTERVAL), ret;
401 unsigned wanted = desired_number_of_snapshots(0,
402 OPT_UINT32_VAL(DSS, NUM_INTERVALS)),
405 struct snapshot *s = NULL;
406 struct snapshot_list sl;
408 dss_get_snapshot_list(&sl);
409 FOR_EACH_SNAPSHOT(s, i, &sl) {
410 if (!(s->flags & SS_COMPLETE))
413 x += s->completion_time - s->creation_time;
418 if (num_complete == 0)
420 x /= num_complete; /* avg time to create one snapshot */
421 if (unit_interval < x * wanted) /* oops, no sleep at all */
423 ret = s->completion_time + unit_interval / wanted - x;
425 free_snapshot_list(&sl);
429 static inline void invalidate_next_snapshot_time(void)
431 next_snapshot_time = 0;
434 static inline int next_snapshot_time_is_valid(void)
436 return next_snapshot_time != 0;
439 static int next_snapshot_is_due(void)
441 int64_t now = get_current_time();
443 if (!next_snapshot_time_is_valid())
444 next_snapshot_time = compute_next_snapshot_time();
445 if (next_snapshot_time <= now) {
446 DSS_DEBUG_LOG(("next snapshot: now\n"));
449 DSS_DEBUG_LOG(("next snapshot due in %" PRId64 " seconds\n",
450 next_snapshot_time - now));
454 static void pre_create_hook(void)
456 assert(snapshot_creation_status == HS_READY);
457 /* make sure that the next snapshot time will be recomputed */
458 invalidate_next_snapshot_time();
459 DSS_DEBUG_LOG(("executing %s\n", OPT_STRING_VAL(DSS, PRE_CREATE_HOOK)));
460 dss_exec_cmdline_pid(&create_pid, OPT_STRING_VAL(DSS, PRE_CREATE_HOOK));
461 snapshot_creation_status = HS_PRE_RUNNING;
464 static void pre_remove_hook(struct snapshot *s, const char *why)
470 DSS_DEBUG_LOG(("%s snapshot %s\n", why, s->name));
471 assert(snapshot_removal_status == HS_READY);
472 assert(remove_pid == 0);
473 assert(!snapshot_currently_being_removed);
475 snapshot_currently_being_removed = dss_malloc(sizeof(struct snapshot));
476 *snapshot_currently_being_removed = *s;
477 snapshot_currently_being_removed->name = dss_strdup(s->name);
479 cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, PRE_REMOVE_HOOK),
480 OPT_STRING_VAL(DSS, DEST_DIR), s->name);
481 DSS_DEBUG_LOG(("executing %s\n", cmd));
482 dss_exec_cmdline_pid(&remove_pid, cmd);
484 snapshot_removal_status = HS_PRE_RUNNING;
487 static int exec_rm(void)
489 struct snapshot *s = snapshot_currently_being_removed;
490 char *new_name = being_deleted_name(s);
499 assert(snapshot_removal_status == HS_PRE_SUCCESS);
500 assert(remove_pid == 0);
502 DSS_NOTICE_LOG(("removing %s (interval = %i)\n", s->name, s->interval));
503 ret = dss_rename(s->name, new_name);
506 dss_exec(&remove_pid, argv[0], argv);
507 snapshot_removal_status = HS_RUNNING;
513 static int snapshot_is_being_created(struct snapshot *s)
515 return s->creation_time == current_snapshot_creation_time;
518 static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
523 DSS_DEBUG_LOG(("looking for orphaned snapshots\n"));
524 FOR_EACH_SNAPSHOT(s, i, sl) {
525 if (snapshot_is_being_created(s))
528 * We know that no rm is currently running, so if s is marked
529 * as being deleted, a previously started rm must have failed.
531 if (s->flags & SS_BEING_DELETED)
534 if (s->flags & SS_COMPLETE) /* good snapshot */
537 * This snapshot is incomplete and it is not the snapshot
538 * currently being created. However, we must not remove it if
539 * rsync is about to be restarted. As only the newest snapshot
540 * can be restarted, this snapshot is orphaned if it is not the
541 * newest snapshot or if we are not about to restart rsync.
543 if (get_newest_snapshot(sl) != s)
545 if (snapshot_creation_status != HS_NEEDS_RESTART)
548 /* no orphaned snapshots */
552 static int is_reference_snapshot(struct snapshot *s)
554 if (!name_of_reference_snapshot)
556 return strcmp(s->name, name_of_reference_snapshot)? 0 : 1;
560 * return: 0: no redundant snapshots, 1: rm process started, negative: error
562 static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
566 unsigned missing = 0;
567 uint32_t N = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
569 DSS_DEBUG_LOG(("looking for intervals containing too many snapshots\n"));
570 for (interval = N - 1; interval >= 0; interval--) {
571 unsigned keep = desired_number_of_snapshots(interval, N);
572 unsigned num = sl->interval_count[interval];
573 struct snapshot *victim = NULL, *prev = NULL;
574 int64_t score = LONG_MAX;
577 missing += keep - num;
578 if (keep + missing >= num)
580 /* redundant snapshot in this interval, pick snapshot with lowest score */
581 FOR_EACH_SNAPSHOT(s, i, sl) {
584 if (snapshot_is_being_created(s))
586 if (is_reference_snapshot(s))
588 if (s->interval > interval) {
592 if (s->interval < interval)
600 /* check if s is a better victim */
601 this_score = s->creation_time - prev->creation_time;
602 assert(this_score >= 0);
603 if (this_score < score) {
615 static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
620 DSS_DEBUG_LOG(("looking for snapshots belonging to intervals >= %d\n",
621 OPT_UINT32_VAL(DSS, NUM_INTERVALS)));
622 FOR_EACH_SNAPSHOT(s, i, sl) {
623 if (snapshot_is_being_created(s))
625 if (is_reference_snapshot(s))
627 if (s->interval < OPT_UINT32_VAL(DSS, NUM_INTERVALS))
634 static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
637 struct snapshot *s, *ref = NULL;
639 num_complete = num_complete_snapshots(sl);
640 if (num_complete <= OPT_UINT32_VAL(DSS, MIN_COMPLETE))
642 FOR_EACH_SNAPSHOT(s, i, sl) {
643 if (snapshot_is_being_created(s))
645 if (is_reference_snapshot(s)) { /* avoid this one */
649 DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
653 DSS_WARNING_LOG(("removing reference snapshot %s\n", ref->name));
657 static int rename_incomplete_snapshot(int64_t start)
664 * We don't want the dss_rename() below to fail with EEXIST because the
665 * last complete snapshot was created (and completed) in the same
666 * second as this one.
668 while ((now = get_current_time()) == start)
670 free(path_to_last_complete_snapshot);
671 ret = complete_name(start, now, &path_to_last_complete_snapshot);
674 old_name = incomplete_name(start);
675 ret = dss_rename(old_name, path_to_last_complete_snapshot);
677 DSS_NOTICE_LOG(("%s -> %s\n", old_name,
678 path_to_last_complete_snapshot));
683 static int try_to_free_disk_space(void)
686 struct snapshot_list sl;
687 struct snapshot *victim;
692 ret = disk_space_low(NULL);
695 low_disk_space = ret;
696 gettimeofday(&now, NULL);
697 if (tv_diff(&next_removal_check, &now, NULL) > 0)
699 if (!low_disk_space) {
700 if (OPT_GIVEN(DSS, KEEP_REDUNDANT))
702 if (snapshot_creation_status != HS_READY)
704 if (next_snapshot_is_due())
708 * Idle and --keep_redundant not given, or low disk space. Look at
709 * existing snapshots.
711 dss_get_snapshot_list(&sl);
714 * Don't remove anything if there is free space and we have fewer
715 * snapshots than configured, plus one. This way there is always one
716 * snapshot that can be recycled.
718 if (!low_disk_space && sl.num_snapshots <=
719 1 << OPT_UINT32_VAL(DSS, NUM_INTERVALS))
722 victim = find_outdated_snapshot(&sl);
726 victim = find_redundant_snapshot(&sl);
730 victim = find_orphaned_snapshot(&sl);
733 /* try harder only if disk space is low */
736 DSS_WARNING_LOG(("disk space low and nothing obvious to remove\n"));
737 victim = find_oldest_removable_snapshot(&sl);
740 DSS_CRIT_LOG(("uhuhu: disk space low and nothing to remove\n"));
741 ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
744 pre_remove_hook(victim, why);
746 free_snapshot_list(&sl);
750 static void post_create_hook(void)
752 char *cmd = make_message("%s %s/%s",
753 OPT_STRING_VAL(DSS, POST_CREATE_HOOK),
754 OPT_STRING_VAL(DSS, DEST_DIR), path_to_last_complete_snapshot);
755 DSS_NOTICE_LOG(("executing %s\n", cmd));
756 dss_exec_cmdline_pid(&create_pid, cmd);
758 snapshot_creation_status = HS_POST_RUNNING;
761 static void post_remove_hook(void)
764 struct snapshot *s = snapshot_currently_being_removed;
768 cmd = make_message("%s %s/%s", OPT_STRING_VAL(DSS, POST_REMOVE_HOOK),
769 OPT_STRING_VAL(DSS, DEST_DIR), s->name);
770 DSS_NOTICE_LOG(("executing %s\n", cmd));
771 dss_exec_cmdline_pid(&remove_pid, cmd);
773 snapshot_removal_status = HS_POST_RUNNING;
776 static void dss_kill(pid_t pid, int sig, const char *msg)
778 const char *signame, *process_name;
783 case SIGTERM: signame = "TERM"; break;
784 case SIGSTOP: signame = "STOP"; break;
785 case SIGCONT: signame = "CONT"; break;
786 default: signame = "????";
789 if (pid == create_pid)
790 process_name = "create";
791 else if (pid == remove_pid)
792 process_name = "remove";
793 else process_name = "??????";
796 DSS_INFO_LOG(("%s\n", msg));
797 DSS_DEBUG_LOG(("sending signal %d (%s) to pid %d (%s process)\n",
798 sig, signame, (int)pid, process_name));
799 if (kill(pid, sig) >= 0)
801 DSS_INFO_LOG(("failed to send signal %d (%s) to pid %d (%s process)\n",
802 sig, signame, (int)pid, process_name));
805 static void stop_create_process(void)
807 if (create_process_stopped)
809 dss_kill(create_pid, SIGSTOP, "suspending create process");
810 create_process_stopped = 1;
813 static void restart_create_process(void)
815 if (!create_process_stopped)
817 dss_kill(create_pid, SIGCONT, "resuming create process");
818 create_process_stopped = 0;
822 * Print a log message about the exit status of a child.
824 static void log_termination_msg(pid_t pid, int status)
826 if (WIFEXITED(status))
827 DSS_INFO_LOG(("child %i exited. Exit status: %i\n", (int)pid,
828 WEXITSTATUS(status)));
829 else if (WIFSIGNALED(status))
830 DSS_NOTICE_LOG(("child %i was killed by signal %i\n", (int)pid,
833 DSS_WARNING_LOG(("child %i terminated abormally\n", (int)pid));
836 static int wait_for_process(pid_t pid, int *status)
840 DSS_DEBUG_LOG(("Waiting for process %d to terminate\n", (int)pid));
845 FD_SET(signal_pipe, &rfds);
846 ret = dss_select(signal_pipe + 1, &rfds, NULL, NULL);
852 if (ret == SIGCHLD) {
853 ret = waitpid(pid, status, 0);
856 if (errno != EINTR) { /* error */
857 ret = -ERRNO_TO_DSS_ERROR(errno);
861 /* SIGINT or SIGTERM */
862 dss_kill(pid, SIGTERM, "killing child process");
865 DSS_ERROR_LOG(("failed to wait for process %d\n", (int)pid));
867 log_termination_msg(pid, *status);
871 static void handle_pre_remove_exit(int status)
873 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
874 snapshot_removal_status = HS_READY;
875 gettimeofday(&next_removal_check, NULL);
876 next_removal_check.tv_sec += 60;
879 snapshot_removal_status = HS_PRE_SUCCESS;
882 static int handle_rm_exit(int status)
884 if (!WIFEXITED(status)) {
885 snapshot_removal_status = HS_READY;
886 return -E_INVOLUNTARY_EXIT;
888 if (WEXITSTATUS(status)) {
889 snapshot_removal_status = HS_READY;
890 return -E_BAD_EXIT_CODE;
892 snapshot_removal_status = HS_SUCCESS;
896 static void handle_post_remove_exit(void)
898 snapshot_removal_status = HS_READY;
901 static int handle_remove_exit(int status)
904 struct snapshot *s = snapshot_currently_being_removed;
907 switch (snapshot_removal_status) {
909 handle_pre_remove_exit(status);
913 ret = handle_rm_exit(status);
915 case HS_POST_RUNNING:
916 handle_post_remove_exit();
922 if (snapshot_removal_status == HS_READY) {
925 snapshot_currently_being_removed = NULL;
931 static int wait_for_remove_process(void)
937 snapshot_removal_status == HS_PRE_RUNNING ||
938 snapshot_removal_status == HS_RUNNING ||
939 snapshot_removal_status == HS_POST_RUNNING
941 ret = wait_for_process(remove_pid, &status);
944 return handle_remove_exit(status);
947 static int handle_rsync_exit(int status)
951 if (!WIFEXITED(status)) {
952 DSS_ERROR_LOG(("rsync process %d died involuntary\n", (int)create_pid));
953 ret = -E_INVOLUNTARY_EXIT;
954 snapshot_creation_status = HS_READY;
957 es = WEXITSTATUS(status);
959 * Restart rsync on non-fatal errors:
960 * 24: Partial transfer due to vanished source files
962 if (es != 0 && es != 24) {
963 DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
964 es, ++num_consecutive_rsync_errors));
965 if (!logfile) { /* called by com_run() */
966 ret = -E_BAD_EXIT_CODE;
969 if (num_consecutive_rsync_errors >
970 OPT_UINT32_VAL(RUN, MAX_RSYNC_ERRORS)) {
971 ret = -E_TOO_MANY_RSYNC_ERRORS;
972 snapshot_creation_status = HS_READY;
975 DSS_WARNING_LOG(("restarting rsync process\n"));
976 snapshot_creation_status = HS_NEEDS_RESTART;
977 next_snapshot_time = get_current_time() + 60;
981 num_consecutive_rsync_errors = 0;
982 ret = rename_incomplete_snapshot(current_snapshot_creation_time);
985 snapshot_creation_status = HS_SUCCESS;
986 free(name_of_reference_snapshot);
987 name_of_reference_snapshot = NULL;
989 create_process_stopped = 0;
993 static int handle_pre_create_hook_exit(int status)
996 static int warn_count;
998 if (!WIFEXITED(status)) {
999 snapshot_creation_status = HS_READY;
1000 ret = -E_INVOLUNTARY_EXIT;
1003 es = WEXITSTATUS(status);
1005 if (!warn_count--) {
1006 DSS_NOTICE_LOG(("pre_create_hook %s returned %d\n",
1007 OPT_STRING_VAL(DSS, PRE_CREATE_HOOK), es));
1008 DSS_NOTICE_LOG(("deferring snapshot creation...\n"));
1009 warn_count = 60; /* warn only once per hour */
1011 next_snapshot_time = get_current_time() + 60;
1012 snapshot_creation_status = HS_READY;
1017 snapshot_creation_status = HS_PRE_SUCCESS;
1023 static int handle_sigchld(void)
1026 int status, ret = reap_child(&pid, &status);
1031 if (pid == create_pid) {
1032 switch (snapshot_creation_status) {
1033 case HS_PRE_RUNNING:
1034 ret = handle_pre_create_hook_exit(status);
1037 ret = handle_rsync_exit(status);
1039 case HS_POST_RUNNING:
1040 snapshot_creation_status = HS_READY;
1044 DSS_EMERG_LOG(("BUG: create can't die in status %d\n",
1045 snapshot_creation_status));
1051 if (pid == remove_pid) {
1052 ret = handle_remove_exit(status);
1057 DSS_EMERG_LOG(("BUG: unknown process %d died\n", (int)pid));
1061 static int change_to_dest_dir(void)
1064 const char *dd = OPT_STRING_VAL(DSS, DEST_DIR);
1066 DSS_INFO_LOG(("changing cwd to %s\n", dd));
1069 ret = -ERRNO_TO_DSS_ERROR(errno);
1070 DSS_ERROR_LOG(("could not change cwd to %s\n", dd));
1074 static int check_config(void)
1077 uint32_t unit_interval = OPT_UINT32_VAL(DSS, UNIT_INTERVAL);
1078 uint32_t num_intervals = OPT_UINT32_VAL(DSS, NUM_INTERVALS);
1080 if (unit_interval == 0) {
1081 DSS_ERROR_LOG(("bad unit interval: %i\n", unit_interval));
1082 return -E_INVALID_NUMBER;
1084 DSS_DEBUG_LOG(("unit interval: %i day(s)\n", unit_interval));
1086 if (num_intervals == 0 || num_intervals > 30) {
1087 DSS_ERROR_LOG(("bad number of intervals: %i\n", num_intervals));
1088 return -E_INVALID_NUMBER;
1090 if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE))
1091 if (!OPT_GIVEN(DSS, SOURCE_DIR)) {
1092 DSS_ERROR_LOG(("--source-dir required\n"));
1095 if (subcmd == CMD_PTR(RUN) || subcmd == CMD_PTR(CREATE)
1096 || subcmd == CMD_PTR(LS) || subcmd == CMD_PTR(PRUNE)) {
1097 if (!OPT_GIVEN(DSS, DEST_DIR)) {
1098 DSS_ERROR_LOG(("--dest-dir required\n"));
1101 ret = change_to_dest_dir();
1105 DSS_DEBUG_LOG(("number of intervals: %i\n", num_intervals));
1109 static int lopsub_error(int lopsub_ret, char **errctx)
1111 const char *msg = lls_strerror(-lopsub_ret);
1113 DSS_ERROR_LOG(("%s: %s\n", *errctx, msg));
1115 DSS_ERROR_LOG(("%s\n", msg));
1121 static int parse_config_file(bool sighup, const struct lls_command *cmd)
1124 char *config_file = get_config_file_name();
1125 struct stat statbuf;
1129 char **cf_argv, *errctx = NULL;
1130 struct lls_parse_result *cf_lpr, *merged_lpr, *clpr;
1131 const char *subcmd_name;
1133 ret = open(config_file, O_RDONLY);
1135 if (errno != ENOENT || OPT_GIVEN(DSS, CONFIG_FILE)) {
1136 ret = -ERRNO_TO_DSS_ERROR(errno);
1137 DSS_ERROR_LOG(("config file %s can not be opened\n",
1141 /* no config file -- nothing to do */
1146 ret = fstat(fd, &statbuf);
1148 ret = -ERRNO_TO_DSS_ERROR(errno);
1149 DSS_ERROR_LOG(("failed to stat config file %s\n", config_file));
1152 sz = statbuf.st_size;
1153 if (sz == 0) { /* config file is empty -- nothing to do */
1157 map = mmap(NULL, sz, PROT_READ, MAP_PRIVATE, fd, 0);
1158 if (map == MAP_FAILED) {
1159 ret = -ERRNO_TO_DSS_ERROR(errno);
1160 DSS_ERROR_LOG(("failed to mmap config file %s\n",
1164 if (cmd == CMD_PTR(DSS))
1167 subcmd_name = lls_command_name(cmd);
1168 ret = lls_convert_config(map, sz, subcmd_name, &cf_argv, &errctx);
1171 DSS_ERROR_LOG(("failed to convert config file %s\n",
1173 ret = lopsub_error(ret, &errctx);
1177 ret = lls_parse(cf_argc, cf_argv, cmd, &cf_lpr, &errctx);
1178 lls_free_argv(cf_argv);
1180 ret = lopsub_error(ret, &errctx);
1183 clpr = cmd == CMD_PTR(DSS)? cmdline_lpr : cmdline_sublpr;
1184 if (sighup) /* config file overrides command line */
1185 ret = lls_merge(cf_lpr, clpr, cmd, &merged_lpr, &errctx);
1186 else /* command line options overrride config file options */
1187 ret = lls_merge(clpr, cf_lpr, cmd, &merged_lpr, &errctx);
1188 lls_free_parse_result(cf_lpr, cmd);
1190 ret = lopsub_error(ret, &errctx);
1196 DSS_DEBUG_LOG(("loglevel: %d\n", OPT_UINT32_VAL(DSS, LOGLEVEL)));
1197 if (cmd != CMD_PTR(DSS)) {
1199 if (sublpr != cmdline_sublpr)
1200 lls_free_parse_result(sublpr, cmd);
1201 sublpr = merged_lpr;
1203 sublpr = cmdline_sublpr;
1206 if (lpr != cmdline_lpr)
1207 lls_free_parse_result(lpr, cmd);
1218 DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
1222 static int handle_sighup(void)
1226 DSS_NOTICE_LOG(("SIGHUP, re-reading config\n"));
1227 dump_dss_config("old");
1228 ret = parse_config_file(true /* SIGHUP */, CMD_PTR(DSS));
1231 ret = parse_config_file(true /* SIGHUP */, CMD_PTR(RUN));
1234 ret = check_config();
1239 if (OPT_GIVEN(RUN, DAEMON) || daemonized) {
1240 logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
1241 log_welcome(OPT_UINT32_VAL(DSS, LOGLEVEL));
1244 dump_dss_config("reloaded");
1245 invalidate_next_snapshot_time();
1249 static void kill_children(void)
1251 restart_create_process();
1252 dss_kill(create_pid, SIGTERM, NULL);
1253 dss_kill(remove_pid, SIGTERM, NULL);
1256 static int handle_signal(void)
1258 int sig, ret = next_signal();
1270 ret = handle_sighup();
1273 ret = handle_sigchld();
1278 DSS_ERROR_LOG(("%s\n", dss_strerror(-ret)));
1283 * We can not use rsync locally if the local user is different from the remote
1284 * user or if the src dir is not on the local host (or both).
1286 static int use_rsync_locally(char *logname)
1288 const char *h = OPT_STRING_VAL(DSS, REMOTE_HOST);
1290 if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
1292 if (OPT_GIVEN(DSS, REMOTE_USER) &&
1293 strcmp(OPT_STRING_VAL(DSS, REMOTE_USER), logname))
1298 static int rename_resume_snap(int64_t creation_time)
1300 struct snapshot_list sl;
1301 struct snapshot *s = NULL;
1302 char *new_name = incomplete_name(creation_time);
1306 sl.num_snapshots = 0;
1309 dss_get_snapshot_list(&sl);
1311 * Snapshot recycling: We first look at the newest snapshot. If this
1312 * snapshot happens to be incomplete, the last rsync process was
1313 * aborted and we reuse this one. Otherwise we look at snapshots which
1314 * could be removed (outdated and redundant snapshots) as candidates
1315 * for recycling. If no outdated/redundant snapshot exists, we check if
1316 * there is an orphaned snapshot, which likely is useless anyway.
1318 * Only if no existing snapshot is suitable for recycling, we bite the
1319 * bullet and create a new one.
1321 s = get_newest_snapshot(&sl);
1322 if (!s) /* no snapshots at all */
1324 /* re-use last snapshot if it is incomplete */
1326 if ((s->flags & SS_COMPLETE) == 0)
1329 s = find_outdated_snapshot(&sl);
1333 s = find_redundant_snapshot(&sl);
1337 s = find_orphaned_snapshot(&sl);
1340 DSS_NOTICE_LOG(("recycling %s snapshot %s\n", why, s->name));
1341 ret = dss_rename(s->name, new_name);
1344 DSS_NOTICE_LOG(("creating %s\n", new_name));
1346 free_snapshot_list(&sl);
1350 static void create_rsync_argv(char ***argv, int64_t *num)
1353 int i = 0, j, N = OPT_GIVEN(DSS, RSYNC_OPTION);
1354 struct snapshot_list sl;
1357 dss_get_snapshot_list(&sl);
1358 assert(!name_of_reference_snapshot);
1359 name_of_reference_snapshot = name_of_newest_complete_snapshot(&sl);
1360 free_snapshot_list(&sl);
1362 *argv = dss_malloc((15 + N) * sizeof(char *));
1363 (*argv)[i++] = dss_strdup("rsync");
1364 (*argv)[i++] = dss_strdup("-a");
1365 (*argv)[i++] = dss_strdup("--delete");
1367 srandom((unsigned)time(NULL)); /* no need to be fancy here */
1370 if (1000 * (random() / (RAND_MAX + 1.0)) < OPT_UINT32_VAL(DSS, CHECKSUM)) {
1371 DSS_NOTICE_LOG(("adding --checksum to rsync options\n"));
1372 (*argv)[i++] = dss_strdup("--checksum");
1374 for (j = 0; j < N; j++)
1375 (*argv)[i++] = dss_strdup(lls_string_val(j,
1376 OPT_RESULT(DSS, RSYNC_OPTION)));
1377 if (name_of_reference_snapshot) {
1378 DSS_INFO_LOG(("using %s as reference\n", name_of_reference_snapshot));
1379 (*argv)[i++] = make_message("--link-dest=../%s",
1380 name_of_reference_snapshot);
1382 DSS_INFO_LOG(("no suitable reference snapshot found\n"));
1383 logname = dss_logname();
1384 if (use_rsync_locally(logname))
1385 (*argv)[i++] = dss_strdup(OPT_STRING_VAL(DSS, SOURCE_DIR));
1387 (*argv)[i++] = make_message("%s@%s:%s/",
1388 OPT_GIVEN(DSS, REMOTE_USER)?
1389 OPT_STRING_VAL(DSS, REMOTE_USER) : logname,
1390 OPT_STRING_VAL(DSS, REMOTE_HOST),
1391 OPT_STRING_VAL(DSS, SOURCE_DIR));
1393 *num = get_current_time();
1394 (*argv)[i++] = incomplete_name(*num);
1395 (*argv)[i++] = NULL;
1396 for (j = 0; j < i; j++)
1397 DSS_DEBUG_LOG(("argv[%d] = %s\n", j, (*argv)[j]));
1400 static void free_rsync_argv(char **argv)
1406 for (i = 0; argv[i]; i++)
1411 static int create_snapshot(char **argv)
1415 ret = rename_resume_snap(current_snapshot_creation_time);
1418 dss_exec(&create_pid, argv[0], argv);
1419 snapshot_creation_status = HS_RUNNING;
1423 static int select_loop(void)
1426 /* check every 60 seconds for free disk space */
1428 char **rsync_argv = NULL;
1432 struct timeval *tvp;
1435 tvp = NULL; /* sleep until rm hook/process dies */
1436 else { /* sleep one minute */
1442 FD_SET(signal_pipe, &rfds);
1443 ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
1446 if (FD_ISSET(signal_pipe, &rfds)) {
1447 ret = handle_signal();
1453 if (snapshot_removal_status == HS_PRE_SUCCESS) {
1459 if (snapshot_removal_status == HS_SUCCESS) {
1463 ret = try_to_free_disk_space();
1466 if (snapshot_removal_status != HS_READY) {
1467 stop_create_process();
1470 restart_create_process();
1471 switch (snapshot_creation_status) {
1473 if (!next_snapshot_is_due())
1477 case HS_PRE_RUNNING:
1479 case HS_POST_RUNNING:
1481 case HS_PRE_SUCCESS:
1482 if (!name_of_reference_snapshot) {
1483 free_rsync_argv(rsync_argv);
1484 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1486 ret = create_snapshot(rsync_argv);
1490 case HS_NEEDS_RESTART:
1491 if (!next_snapshot_is_due())
1493 ret = create_snapshot(rsync_argv);
1506 static void exit_hook(int exit_code)
1508 const char *argv[3];
1511 argv[0] = OPT_STRING_VAL(DSS, EXIT_HOOK);
1512 argv[1] = dss_strerror(-exit_code);
1515 DSS_NOTICE_LOG(("executing %s %s\n", argv[0], argv[1]));
1516 dss_exec(&pid, argv[0], (char **)argv);
1519 static void lock_dss_or_die(void)
1521 char *config_file = get_config_file_name();
1522 int ret = lock_dss(config_file);
1526 DSS_EMERG_LOG(("failed to lock: %s\n", dss_strerror(-ret)));
1531 static int com_run(void)
1537 if (OPT_GIVEN(DSS, DRY_RUN)) {
1538 DSS_ERROR_LOG(("dry run not supported by this command\n"));
1541 config_file = get_config_file_name();
1542 ret = get_dss_pid(config_file, &pid);
1545 DSS_ERROR_LOG(("pid %d\n", (int)pid));
1546 return -E_ALREADY_RUNNING;
1548 if (OPT_GIVEN(RUN, DAEMON)) {
1551 logfile = open_log(OPT_STRING_VAL(RUN, LOGFILE));
1554 dump_dss_config("startup");
1555 ret = install_sighandler(SIGHUP);
1559 ret = write(fd, "\0", 1);
1561 DSS_ERROR_LOG(("write to daemon pipe returned %d\n",
1564 return -ERRNO_TO_DSS_ERROR(errno);
1568 ret = select_loop();
1569 if (ret >= 0) /* impossible */
1575 EXPORT_CMD_HANDLER(run);
1577 static int com_prune(void)
1580 struct snapshot_list sl;
1581 struct snapshot *victim;
1582 struct disk_space ds;
1586 ret = get_disk_space(".", &ds);
1589 log_disk_space(&ds);
1590 dss_get_snapshot_list(&sl);
1592 victim = find_outdated_snapshot(&sl);
1596 victim = find_redundant_snapshot(&sl);
1602 if (OPT_GIVEN(DSS, DRY_RUN)) {
1603 dss_msg("%s snapshot %s (interval = %i)\n",
1604 why, victim->name, victim->interval);
1608 pre_remove_hook(victim, why);
1609 if (snapshot_removal_status == HS_PRE_RUNNING) {
1610 ret = wait_for_remove_process();
1613 if (snapshot_removal_status != HS_PRE_SUCCESS)
1619 ret = wait_for_remove_process();
1622 if (snapshot_removal_status != HS_SUCCESS)
1625 if (snapshot_removal_status != HS_POST_RUNNING)
1627 ret = wait_for_remove_process();
1632 free_snapshot_list(&sl);
1635 EXPORT_CMD_HANDLER(prune);
1637 static int com_create(void)
1643 if (OPT_GIVEN(DSS, DRY_RUN)) {
1646 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1647 for (i = 0; rsync_argv[i]; i++) {
1649 msg = make_message("%s%s%s", tmp? tmp : "",
1650 tmp? " " : "", rsync_argv[i]);
1653 free_rsync_argv(rsync_argv);
1654 dss_msg("%s\n", msg);
1660 ret = wait_for_process(create_pid, &status);
1663 ret = handle_pre_create_hook_exit(status);
1664 if (ret <= 0) /* error, or pre-create failed */
1667 create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
1668 ret = create_snapshot(rsync_argv);
1671 ret = wait_for_process(create_pid, &status);
1674 ret = handle_rsync_exit(status);
1679 ret = wait_for_process(create_pid, &status);
1681 free_rsync_argv(rsync_argv);
1684 EXPORT_CMD_HANDLER(create);
1686 static int com_ls(void)
1689 struct snapshot_list sl;
1692 dss_get_snapshot_list(&sl);
1693 FOR_EACH_SNAPSHOT(s, i, &sl) {
1695 if (s->flags & SS_COMPLETE)
1696 d = (s->completion_time - s->creation_time) / 60;
1697 dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval, s->name, d/60, d%60);
1699 free_snapshot_list(&sl);
1702 EXPORT_CMD_HANDLER(ls);
1704 static int setup_signal_handling(void)
1708 DSS_INFO_LOG(("setting up signal handlers\n"));
1709 signal_pipe = signal_init(); /* always successful */
1710 ret = install_sighandler(SIGINT);
1713 ret = install_sighandler(SIGTERM);
1716 return install_sighandler(SIGCHLD);
1719 static void handle_version_and_help(void)
1723 if (OPT_GIVEN(DSS, DETAILED_HELP))
1724 txt = lls_long_help(CMD_PTR(DSS));
1725 else if (OPT_GIVEN(DSS, HELP))
1726 txt = lls_short_help(CMD_PTR(DSS));
1727 else if (OPT_GIVEN(DSS, VERSION))
1728 txt = dss_strdup(VERSION_STRING);
1736 static void show_subcommand_summary(void)
1738 const struct lls_command *cmd;
1741 printf("Available subcommands:\n");
1742 for (i = 1; (cmd = lls_cmd(i, dss_suite)); i++) {
1743 const char *name = lls_command_name(cmd);
1744 const char *purpose = lls_purpose(cmd);
1745 printf("%-10s%s\n", name, purpose);
1750 int main(int argc, char **argv)
1753 char *errctx = NULL;
1754 unsigned num_inputs;
1755 const struct dss_user_data *ud;
1757 ret = lls_parse(argc, argv, CMD_PTR(DSS), &cmdline_lpr, &errctx);
1759 ret = lopsub_error(ret, &errctx);
1763 ret = parse_config_file(false /* no SIGHUP */, CMD_PTR(DSS));
1766 handle_version_and_help();
1767 num_inputs = lls_num_inputs(lpr);
1768 if (num_inputs == 0)
1769 show_subcommand_summary();
1770 ret = lls_lookup_subcmd(argv[argc - num_inputs], dss_suite, &errctx);
1772 ret = lopsub_error(ret, &errctx);
1775 subcmd = lls_cmd(ret, dss_suite);
1776 ret = lls_parse(num_inputs, argv + argc - num_inputs, subcmd,
1777 &cmdline_sublpr, &errctx);
1779 ret = lopsub_error(ret, &errctx);
1782 sublpr = cmdline_sublpr;
1783 ret = parse_config_file(false /* no SIGHUP */, subcmd);
1786 ret = check_config();
1789 ret = setup_signal_handling();
1792 ud = lls_user_data(subcmd);
1793 ret = ud->handler();
1798 DSS_ERROR_LOG(("%s\n", errctx));
1799 DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));
1802 lls_free_parse_result(lpr, CMD_PTR(DSS));
1803 if (lpr != cmdline_lpr)
1804 lls_free_parse_result(cmdline_lpr, CMD_PTR(DSS));
1805 lls_free_parse_result(sublpr, subcmd);
1806 if (sublpr != cmdline_sublpr)
1807 lls_free_parse_result(cmdline_sublpr, subcmd);
1808 exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);