4bf01e961ad1d2b3d32313edd70668511f899b37
[dss.git] / dss.c
1 #include <string.h>
2 #include <stdlib.h>
3 #include <stdarg.h>
4 #include <assert.h>
5 #include <errno.h>
6 #include <sys/types.h>
7 #include <signal.h>
8 #include <ctype.h>
9 #include <sys/stat.h>
10 #include <unistd.h>
11 #include <inttypes.h>
12 #include <sys/time.h>
13 #include <time.h>
14 #include <sys/wait.h>
15 #include <fnmatch.h>
16 #include <limits.h>
17
18
19 #include "gcc-compat.h"
20 #include "cmdline.h"
21 #include "log.h"
22 #include "string.h"
23 #include "error.h"
24 #include "fd.h"
25 #include "exec.h"
26 #include "daemon.h"
27 #include "signal.h"
28 #include "df.h"
29 #include "time.h"
30
31
32 struct gengetopt_args_info conf;
33 char *dss_error_txt = NULL;
34 static FILE *logfile;
35 static int signal_pipe;
36
37 /** Process id of current rsync process. */
38 static pid_t rsync_pid;
39 /** Whether the rsync process is currently stopped */
40 static int rsync_stopped;
41 /** Process id of current rm process. */
42 static pid_t rm_pid;
43 /** When the next snapshot is due. */
44 struct timeval next_snapshot_time;
45 /* Creation time of the snapshot currently being created. */
46 int64_t current_snapshot_creation_time;
47
48
49 DEFINE_DSS_ERRLIST;
50
51
52 /* a litte cpp magic helps to DRY */
53 #define COMMANDS \
54 COMMAND(ls) \
55 COMMAND(create) \
56 COMMAND(prune) \
57 COMMAND(run)
58 #define COMMAND(x) int com_ ##x(void);
59 COMMANDS
60 #undef COMMAND
61 #define COMMAND(x) if (conf.x ##_given) return com_ ##x();
62 int call_command_handler(void)
63 {
64 COMMANDS
65 DSS_EMERG_LOG("BUG: did not find command handler\n");
66 exit(EXIT_FAILURE);
67 }
68 #undef COMMAND
69 #undef COMMANDS
70
71 /*
72 * complete, not being deleted: 1204565370-1204565371.Sun_Mar_02_2008_14_33-Sun_Mar_02_2008_14_43
73 * complete, being deleted: 1204565370-1204565371.being_deleted
74 * incomplete, not being deleted: 1204565370-incomplete
75 * incomplete, being deleted: 1204565370-incomplete.being_deleted
76 */
77 enum snapshot_status_flags {
78 /** The rsync process terminated successfully. */
79 SS_COMPLETE = 1,
80 /** The rm process is running to remove this snapshot. */
81 SS_BEING_DELETED = 2,
82 };
83
84 struct snapshot {
85 char *name;
86 int64_t creation_time;
87 int64_t completion_time;
88 enum snapshot_status_flags flags;
89 unsigned interval;
90 };
91
92 /*
93 * An edge snapshot is either the oldest one or the newest one.
94 *
95 * We need to find either of them occasionally: The create code
96 * needs to know the newest snapshot because that is the one
97 * used as the link destination dir. The pruning code needs to
98 * find the oldest one in case disk space becomes low.
99 */
100 struct edge_snapshot_data {
101 int64_t now;
102 struct snapshot snap;
103 };
104
105 __printf_2_3 void dss_log(int ll, const char* fmt,...)
106 {
107 va_list argp;
108 FILE *outfd;
109 struct tm *tm;
110 time_t t1;
111 char str[255] = "";
112
113 if (ll < conf.loglevel_arg)
114 return;
115 outfd = logfile? logfile : stderr;
116 time(&t1);
117 tm = localtime(&t1);
118 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
119 fprintf(outfd, "%s ", str);
120 if (conf.loglevel_arg <= INFO)
121 fprintf(outfd, "%i: ", ll);
122 va_start(argp, fmt);
123 vfprintf(outfd, fmt, argp);
124 va_end(argp);
125 }
126
127 /**
128 * Print a message either to stdout or to the log file.
129 */
130 __printf_1_2 void dss_msg(const char* fmt,...)
131 {
132 FILE *outfd = conf.daemon_given? logfile : stdout;
133 va_list argp;
134 va_start(argp, fmt);
135 vfprintf(outfd, fmt, argp);
136 va_end(argp);
137 }
138
139 /**
140 * Return the desired number of snapshots of an interval.
141 */
142 unsigned num_snapshots(int interval)
143 {
144 unsigned n;
145
146 assert(interval >= 0);
147
148 if (interval >= conf.num_intervals_arg)
149 return 0;
150 n = conf.num_intervals_arg - interval - 1;
151 return 1 << n;
152 }
153
154 /* return: Whether dirname is a snapshot directory (0: no, 1: yes) */
155 int is_snapshot(const char *dirname, int64_t now, struct snapshot *s)
156 {
157 int i, ret;
158 char *dash, *dot, *tmp;
159 int64_t num;
160
161 assert(dirname);
162 dash = strchr(dirname, '-');
163 if (!dash || !dash[1] || dash == dirname)
164 return 0;
165 for (i = 0; dirname[i] != '-'; i++)
166 if (!isdigit(dirname[i]))
167 return 0;
168 tmp = dss_strdup(dirname);
169 tmp[i] = '\0';
170 ret = dss_atoi64(tmp, &num);
171 free(tmp);
172 if (ret < 0) {
173 free(dss_error_txt);
174 return 0;
175 }
176 assert(num >= 0);
177 if (num > now)
178 return 0;
179 s->creation_time = num;
180 //DSS_DEBUG_LOG("%s start time: %lli\n", dirname, (long long)s->creation_time);
181 s->interval = (long long) ((now - s->creation_time)
182 / conf.unit_interval_arg / 24 / 3600);
183 if (!strcmp(dash + 1, "incomplete")) {
184 s->completion_time = -1;
185 s->flags = 0; /* neither complete, nor being deleted */
186 goto success;
187 }
188 if (!strcmp(dash + 1, "incomplete.being_deleted")) {
189 s->completion_time = -1;
190 s->flags = SS_BEING_DELETED; /* mot cpmplete, being deleted */
191 goto success;
192 }
193 tmp = dash + 1;
194 dot = strchr(tmp, '.');
195 if (!dot || !dot[1] || dot == tmp)
196 return 0;
197 for (i = 0; tmp[i] != '.'; i++)
198 if (!isdigit(tmp[i]))
199 return 0;
200 tmp = dss_strdup(dash + 1);
201 tmp[i] = '\0';
202 ret = dss_atoi64(tmp, &num);
203 free(tmp);
204 if (ret < 0) {
205 free(dss_error_txt);
206 return 0;
207 }
208 if (num > now)
209 return 0;
210 s->completion_time = num;
211 s->flags = SS_COMPLETE;
212 if (!strcmp(dot + 1, "being_deleted"))
213 s->flags |= SS_BEING_DELETED;
214 success:
215 s->name = dss_strdup(dirname);
216 return 1;
217 }
218
219 int64_t get_current_time(void)
220 {
221 time_t now;
222 time(&now);
223 DSS_DEBUG_LOG("now: %lli\n", (long long) now);
224 return (int64_t)now;
225 }
226
227 char *incomplete_name(int64_t start)
228 {
229 return make_message("%lli-incomplete", (long long)start);
230 }
231
232 char *being_deleted_name(struct snapshot *s)
233 {
234 if (s->flags & SS_COMPLETE)
235 return make_message("%lli-%lli.being_deleted",
236 (long long)s->creation_time,
237 (long long)s->completion_time);
238 return make_message("%lli-incomplete.being_deleted",
239 (long long)s->creation_time);
240 }
241
242 int complete_name(int64_t start, int64_t end, char **result)
243 {
244 struct tm start_tm, end_tm;
245 time_t *start_seconds = (time_t *) (uint64_t *)&start; /* STFU, gcc */
246 time_t *end_seconds = (time_t *) (uint64_t *)&end; /* STFU, gcc */
247 char start_str[200], end_str[200];
248
249 if (!localtime_r(start_seconds, &start_tm)) {
250 make_err_msg("%lli", (long long)start);
251 return -E_LOCALTIME;
252 }
253 if (!localtime_r(end_seconds, &end_tm)) {
254 make_err_msg("%lli", (long long)end);
255 return -E_LOCALTIME;
256 }
257 if (!strftime(start_str, sizeof(start_str), "%a_%b_%d_%Y_%H_%M_%S", &start_tm)) {
258 make_err_msg("%lli", (long long)start);
259 return -E_STRFTIME;
260 }
261 if (!strftime(end_str, sizeof(end_str), "%a_%b_%d_%Y_%H_%M_%S", &end_tm)) {
262 make_err_msg("%lli", (long long)end);
263 return -E_STRFTIME;
264 }
265 *result = make_message("%lli-%lli.%s-%s", (long long) start, (long long) end,
266 start_str, end_str);
267 return 1;
268 }
269
270 struct snapshot_list {
271 int64_t now;
272 unsigned num_snapshots;
273 unsigned array_size;
274 struct snapshot **snapshots;
275 /**
276 * Array of size num_intervals + 1
277 *
278 * It contains the number of snapshots in each interval. interval_count[num_intervals]
279 * is the number of snapshots which belong to any interval greater than num_intervals.
280 */
281 unsigned *interval_count;
282 };
283
284 #define FOR_EACH_SNAPSHOT(s, i, sl) \
285 for ((i) = 0; (i) < (sl)->num_snapshots && ((s) = (sl)->snapshots[(i)]); (i)++)
286
287
288
289 #define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
290
291 static int compare_snapshots(const void *a, const void *b)
292 {
293 struct snapshot *s1 = *(struct snapshot **)a;
294 struct snapshot *s2 = *(struct snapshot **)b;
295 return NUM_COMPARE(s2->creation_time, s1->creation_time);
296 }
297
298 /** Compute the minimum of \a a and \a b. */
299 #define DSS_MIN(a,b) ((a) < (b) ? (a) : (b))
300
301 int add_snapshot(const char *dirname, void *private)
302 {
303 struct snapshot_list *sl = private;
304 struct snapshot s;
305 int ret = is_snapshot(dirname, sl->now, &s);
306
307 if (!ret)
308 return 1;
309 if (sl->num_snapshots >= sl->array_size) {
310 sl->array_size = 2 * sl->array_size + 1;
311 sl->snapshots = dss_realloc(sl->snapshots,
312 sl->array_size * sizeof(struct snapshot *));
313 }
314 sl->snapshots[sl->num_snapshots] = dss_malloc(sizeof(struct snapshot));
315 *(sl->snapshots[sl->num_snapshots]) = s;
316 sl->interval_count[DSS_MIN(s.interval, conf.num_intervals_arg)]++;
317 sl->num_snapshots++;
318 return 1;
319 }
320
321 void get_snapshot_list(struct snapshot_list *sl)
322 {
323 sl->now = get_current_time();
324 sl->num_snapshots = 0;
325 sl->array_size = 0;
326 sl->snapshots = NULL;
327 sl->interval_count = dss_calloc((conf.num_intervals_arg + 1) * sizeof(unsigned));
328 for_each_subdir(add_snapshot, sl);
329 qsort(sl->snapshots, sl->num_snapshots, sizeof(struct snapshot *),
330 compare_snapshots);
331 }
332
333 void free_snapshot_list(struct snapshot_list *sl)
334 {
335 int i;
336 struct snapshot *s;
337
338 FOR_EACH_SNAPSHOT(s, i, sl) {
339 free(s->name);
340 free(s);
341 }
342 free(sl->interval_count);
343 free(sl->snapshots);
344 }
345
346 void stop_rsync_process(void)
347 {
348 if (!rsync_pid || rsync_stopped)
349 return;
350 kill(SIGSTOP, rsync_pid);
351 rsync_stopped = 1;
352 }
353
354 void restart_rsync_process(void)
355 {
356 if (!rsync_pid || !rsync_stopped)
357 return;
358 kill (SIGCONT, rsync_pid);
359 rsync_stopped = 0;
360 }
361
362 /**
363 * Print a log message about the exit status of a child.
364 */
365 void log_termination_msg(pid_t pid, int status)
366 {
367 if (WIFEXITED(status))
368 DSS_INFO_LOG("child %i exited. Exit status: %i\n", (int)pid,
369 WEXITSTATUS(status));
370 else if (WIFSIGNALED(status))
371 DSS_NOTICE_LOG("child %i was killed by signal %i\n", (int)pid,
372 WTERMSIG(status));
373 else
374 DSS_WARNING_LOG("child %i terminated abormally\n", (int)pid);
375 }
376
377 int wait_for_process(pid_t pid, int *status)
378 {
379 int ret;
380
381 DSS_DEBUG_LOG("Waiting for process %d to terminate\n", (int)pid);
382 for (;;) {
383 pause();
384 ret = next_signal();
385 if (ret < 0)
386 break;
387 if (!ret)
388 continue;
389 if (ret == SIGCHLD) {
390 ret = waitpid(pid, status, 0);
391 if (ret >= 0)
392 break;
393 if (errno != EINTR) { /* error */
394 ret = -ERRNO_TO_DSS_ERROR(errno);
395 break;
396 }
397 }
398 /* SIGINT or SIGTERM */
399 DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
400 kill(pid, SIGTERM);
401 }
402 if (ret < 0)
403 make_err_msg("failed to wait for process %d", (int)pid);
404 else
405 log_termination_msg(pid, *status);
406 return ret;
407 }
408
409 int remove_snapshot(struct snapshot *s)
410 {
411 int fds[3] = {0, 0, 0};
412 assert(!rm_pid);
413 char *new_name = being_deleted_name(s);
414 int ret = dss_rename(s->name, new_name);
415 char *argv[] = {"rm", "-rf", new_name, NULL};
416
417 if (ret < 0)
418 goto out;
419 DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
420 stop_rsync_process();
421 ret = dss_exec(&rm_pid, argv[0], argv, fds);
422 out:
423 free(new_name);
424 return ret;
425 }
426
427 /*
428 * return: 0: no redundant snapshots, 1: rm process started, negative: error
429 */
430 int remove_redundant_snapshot(struct snapshot_list *sl)
431 {
432 int ret, i, interval;
433 struct snapshot *s;
434 unsigned missing = 0;
435
436 DSS_INFO_LOG("looking for intervals containing too many snapshots\n");
437 for (interval = conf.num_intervals_arg - 1; interval >= 0; interval--) {
438 unsigned keep = num_snapshots(interval);
439 unsigned num = sl->interval_count[interval];
440 struct snapshot *victim = NULL, *prev = NULL;
441 int64_t score = LONG_MAX;
442
443 if (keep >= num)
444 missing += keep - num;
445 DSS_DEBUG_LOG("interval %i: keep: %u, have: %u, missing: %u\n",
446 interval, keep, num, missing);
447 if (keep + missing >= num)
448 continue;
449 /* redundant snapshot in this interval, pick snapshot with lowest score */
450 FOR_EACH_SNAPSHOT(s, i, sl) {
451 int64_t this_score;
452
453 //DSS_DEBUG_LOG("checking %s\n", s->name);
454 if (s->interval > interval) {
455 prev = s;
456 continue;
457 }
458 if (s->interval < interval)
459 break;
460 if (!victim) {
461 victim = s;
462 prev = s;
463 continue;
464 }
465 assert(prev);
466 /* check if s is a better victim */
467 this_score = s->creation_time - prev->creation_time;
468 assert(this_score >= 0);
469 //DSS_DEBUG_LOG("%s: score %lli\n", s->name, (long long)score);
470 if (this_score < score) {
471 score = this_score;
472 victim = s;
473 }
474 prev = s;
475 }
476 assert(victim);
477 if (conf.dry_run_given) {
478 dss_msg("%s would be removed (interval = %i)\n",
479 victim->name, victim->interval);
480 continue;
481 }
482 ret = remove_snapshot(victim);
483 return ret < 0? ret : 1;
484 }
485 return 0;
486 }
487
488 int remove_outdated_snapshot(struct snapshot_list *sl)
489 {
490 int i, ret;
491 struct snapshot *s;
492
493 DSS_INFO_LOG("looking for snapshots belonging to intervals greater than %d\n",
494 conf.num_intervals_arg);
495 FOR_EACH_SNAPSHOT(s, i, sl) {
496 if (s->interval <= conf.num_intervals_arg)
497 continue;
498 if (conf.dry_run_given) {
499 dss_msg("%s would be removed (interval = %i)\n",
500 s->name, s->interval);
501 continue;
502 }
503 ret = remove_snapshot(s);
504 if (ret < 0)
505 return ret;
506 return 1;
507 }
508 return 0;
509 }
510
511 int handle_rm_exit(int status)
512 {
513 int es, ret;
514
515 if (!WIFEXITED(status)) {
516 make_err_msg("rm process %d died involuntary", (int)rm_pid);
517 ret = -E_INVOLUNTARY_EXIT;
518 goto out;
519 }
520 es = WEXITSTATUS(status);
521 if (es) {
522 make_err_msg("rm process %d returned %d", (int)rm_pid, es);
523 ret = -E_BAD_EXIT_CODE;
524 goto out;
525 }
526 ret = 1;
527 rm_pid = 0;
528 out:
529 return ret;
530 }
531
532 int wait_for_rm_process(void)
533 {
534 int status, ret = wait_for_process(rm_pid, &status);
535
536 if (ret < 0)
537 return ret;
538 return handle_rm_exit(status);
539 }
540
541 void kill_process(pid_t pid)
542 {
543 if (!pid)
544 return;
545 DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
546 kill(pid, SIGTERM);
547 }
548
549 void handle_sighup()
550 {
551 DSS_INFO_LOG("FIXME: no sighup handling yet\n");
552 }
553
554 int rename_incomplete_snapshot(int64_t start)
555 {
556 char *old_name, *new_name;
557 int ret;
558
559 ret = complete_name(start, get_current_time(), &new_name);
560 if (ret < 0)
561 return ret;
562 old_name = incomplete_name(start);
563 ret = dss_rename(old_name, new_name);
564 if (ret >= 0)
565 DSS_NOTICE_LOG("%s -> %s\n", old_name, new_name);
566 free(old_name);
567 free(new_name);
568 return ret;
569 }
570
571 int handle_rsync_exit(int status)
572 {
573 int es, ret;
574
575 if (!WIFEXITED(status)) {
576 make_err_msg("rsync process %d died involuntary", (int)rsync_pid);
577 ret = -E_INVOLUNTARY_EXIT;
578 goto out;
579 }
580 es = WEXITSTATUS(status);
581 if (es != 0 && es != 23 && es != 24) {
582 make_err_msg("rsync process %d returned %d", (int)rsync_pid, es);
583 ret = -E_BAD_EXIT_CODE;
584 goto out;
585 }
586 ret = rename_incomplete_snapshot(current_snapshot_creation_time);
587 out:
588 rsync_pid = 0;
589 current_snapshot_creation_time = 0;
590 rsync_stopped = 0;
591 return ret;
592 }
593
594 int get_newest_complete(const char *dirname, void *private)
595 {
596 struct edge_snapshot_data *esd = private;
597 struct snapshot s;
598 int ret = is_snapshot(dirname, esd->now, &s);
599
600 if (ret <= 0)
601 return 1;
602 if (s.flags != SS_COMPLETE) /* incomplete or being deleted */
603 return 1;
604 if (s.creation_time < esd->snap.creation_time)
605 return 1;
606 free(esd->snap.name);
607 esd->snap = s;
608 return 1;
609 }
610
611 __malloc char *name_of_newest_complete_snapshot(void)
612 {
613 struct edge_snapshot_data esd = {
614 .now = get_current_time(),
615 .snap = {.creation_time = -1}
616 };
617 for_each_subdir(get_newest_complete, &esd);
618 return esd.snap.name;
619 }
620
621 void create_rsync_argv(char ***argv, int64_t *num)
622 {
623 char *logname, *newest = name_of_newest_complete_snapshot();
624 int i = 0, j;
625
626 *argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
627 (*argv)[i++] = dss_strdup("rsync");
628 (*argv)[i++] = dss_strdup("-aq");
629 (*argv)[i++] = dss_strdup("--delete");
630 for (j = 0; j < conf.rsync_option_given; j++)
631 (*argv)[i++] = dss_strdup(conf.rsync_option_arg[j]);
632 if (newest) {
633 DSS_INFO_LOG("using %s as reference snapshot\n", newest);
634 (*argv)[i++] = make_message("--link-dest=../%s", newest);
635 free(newest);
636 } else
637 DSS_INFO_LOG("no previous snapshot found\n");
638 if (conf.exclude_patterns_given) {
639 (*argv)[i++] = dss_strdup("--exclude-from");
640 (*argv)[i++] = dss_strdup(conf.exclude_patterns_arg);
641
642 }
643 logname = dss_logname();
644 if (conf.remote_user_given && !strcmp(conf.remote_user_arg, logname))
645 (*argv)[i++] = dss_strdup(conf.source_dir_arg);
646 else
647 (*argv)[i++] = make_message("%s@%s:%s/", conf.remote_user_given?
648 conf.remote_user_arg : logname,
649 conf.remote_host_arg, conf.source_dir_arg);
650 free(logname);
651 *num = get_current_time();
652 (*argv)[i++] = incomplete_name(*num);
653 (*argv)[i++] = NULL;
654 for (j = 0; j < i; j++)
655 DSS_DEBUG_LOG("argv[%d] = %s\n", j, (*argv)[j]);
656 }
657
658 void free_rsync_argv(char **argv)
659 {
660 int i;
661 for (i = 0; argv[i]; i++)
662 free(argv[i]);
663 free(argv);
664 }
665
666 int create_snapshot(char **argv)
667 {
668 int fds[3] = {0, 0, 0};
669 char *name = incomplete_name(current_snapshot_creation_time);
670
671 DSS_NOTICE_LOG("creating new snapshot %s\n", name);
672 free(name);
673 return dss_exec(&rsync_pid, argv[0], argv, fds);
674 }
675
676 void compute_next_snapshot_time(struct snapshot_list *sl)
677 {
678 struct timeval now, unit_interval = {.tv_sec = 24 * 3600 * conf.unit_interval_arg},
679 tmp, diff;
680 int64_t x = 0;
681 unsigned wanted = num_snapshots(0), num_complete_snapshots = 0;
682 int i, ret;
683 struct snapshot *s;
684
685 gettimeofday(&now, NULL);
686 FOR_EACH_SNAPSHOT(s, i, sl) {
687 if (!(s->flags & SS_COMPLETE))
688 continue;
689 num_complete_snapshots++;
690 x += s->completion_time - s->creation_time;
691 }
692 assert(x >= 0);
693 if (num_complete_snapshots)
694 x /= num_complete_snapshots; /* avg time to create one snapshot */
695 x *= wanted; /* time to create all snapshots in interval 0 */
696 tmp.tv_sec = x;
697 tmp.tv_usec = 0;
698 ret = tv_diff(&unit_interval, &tmp, &diff); /* time between creation */
699 if (ret < 0) {
700 next_snapshot_time = now;
701 return;
702 }
703 tv_divide(wanted, &diff, &tmp);
704 tv_add(&now, &tmp, &next_snapshot_time);
705 }
706
707 void handle_signal(struct snapshot_list *sl)
708 {
709 int sig, ret = next_signal();
710
711 if (ret <= 0)
712 goto out;
713 sig = ret;
714 switch (sig) {
715 int status;
716 pid_t pid;
717 case SIGINT:
718 case SIGTERM:
719 restart_rsync_process();
720 kill_process(rsync_pid);
721 kill_process(rm_pid);
722 exit(EXIT_FAILURE);
723 case SIGHUP:
724 handle_sighup();
725 break;
726 case SIGCHLD:
727 ret = reap_child(&pid, &status);
728 if (ret <= 0)
729 break;
730 assert(pid == rsync_pid || pid == rm_pid);
731 if (pid == rsync_pid)
732 ret = handle_rsync_exit(status);
733 else
734 ret = handle_rm_exit(status);
735 free_snapshot_list(sl);
736 get_snapshot_list(sl);
737 compute_next_snapshot_time(sl);
738 }
739 out:
740 if (ret < 0)
741 log_err_msg(ERROR, -ret);
742 }
743
744 int get_oldest(const char *dirname, void *private)
745 {
746 struct edge_snapshot_data *esd = private;
747 struct snapshot s;
748 int ret = is_snapshot(dirname, esd->now, &s);
749
750 if (ret <= 0)
751 return 1;
752 if (s.creation_time > esd->snap.creation_time)
753 return 1;
754 free(esd->snap.name);
755 esd->snap = s;
756 return 1;
757 }
758
759 int remove_oldest_snapshot()
760 {
761 int ret;
762 struct edge_snapshot_data esd = {
763 .now = get_current_time(),
764 .snap = {.creation_time = LLONG_MAX}
765 };
766 for_each_subdir(get_oldest, &esd);
767 if (!esd.snap.name) /* no snapshot found */
768 return 0;
769 DSS_INFO_LOG("oldest snapshot: %s\n", esd.snap.name);
770 ret = 0;
771 if (esd.snap.creation_time == current_snapshot_creation_time)
772 goto out; /* do not remove the snapshot currently being created */
773 ret = remove_snapshot(&esd.snap);
774 out:
775 free(esd.snap.name);
776 return ret;
777 }
778
779 /* TODO: Also consider number of inodes. */
780 int disk_space_low(void)
781 {
782 struct disk_space ds;
783 int ret = get_disk_space(".", &ds);
784
785 if (ret < 0)
786 return ret;
787 if (conf.min_free_mb_arg)
788 if (ds.free_mb < conf.min_free_mb_arg)
789 return 1;
790 if (conf.min_free_percent_arg)
791 if (ds.percent_free < conf.min_free_percent_arg)
792 return 1;
793 return 0;
794 }
795
796 int try_to_free_disk_space(int low_disk_space, struct snapshot_list *sl)
797 {
798 int ret;
799
800 ret = remove_outdated_snapshot(sl);
801 if (ret) /* error, or we are removing something */
802 return ret;
803 /* no outdated snapshot */
804 ret = remove_redundant_snapshot(sl);
805 if (ret)
806 return ret;
807 if (!low_disk_space)
808 return 0;
809 DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
810 ret = remove_oldest_snapshot();
811 if (ret)
812 return ret;
813 make_err_msg("uhuhu: not enough disk space for a single snapshot");
814 return -ENOSPC;
815 }
816
817 int select_loop(void)
818 {
819 int ret;
820 struct timeval tv = {.tv_sec = 0, .tv_usec = 0};
821 struct snapshot_list sl = {.num_snapshots = 0};
822
823 get_snapshot_list(&sl);
824 compute_next_snapshot_time(&sl);
825 for (;;) {
826 struct timeval now, *tvp = &tv;
827 fd_set rfds;
828 int low_disk_space;
829 char **rsync_argv;
830
831 FD_ZERO(&rfds);
832 FD_SET(signal_pipe, &rfds);
833 if (rsync_pid)
834 tv.tv_sec = 60;
835 else if (rm_pid)
836 tvp = NULL;
837 ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
838 if (ret < 0)
839 return ret;
840 if (FD_ISSET(signal_pipe, &rfds))
841 handle_signal(&sl);
842 if (rm_pid)
843 continue;
844 ret = disk_space_low();
845 if (ret < 0)
846 break;
847 low_disk_space = ret;
848 if (low_disk_space)
849 stop_rsync_process();
850 ret = try_to_free_disk_space(low_disk_space, &sl);
851 if (ret < 0)
852 break;
853 if (rm_pid)
854 continue;
855 if (rsync_pid) {
856 restart_rsync_process();
857 continue;
858 }
859 /* neither rsync nor rm are running. Start rsync? */
860 gettimeofday(&now, NULL);
861 if (tv_diff(&next_snapshot_time, &now, &tv) > 0)
862 continue;
863 create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
864 ret = create_snapshot(rsync_argv);
865 free_rsync_argv(rsync_argv);
866 if (ret < 0)
867 break;
868 }
869 free_snapshot_list(&sl);
870 return ret;
871 }
872
873 int com_run(void)
874 {
875 int ret;
876
877 if (conf.dry_run_given) {
878 make_err_msg("dry_run not supported by this command");
879 return -E_SYNTAX;
880 }
881 ret = install_sighandler(SIGHUP);
882 if (ret < 0)
883 return ret;
884 return select_loop();
885 }
886
887 void log_disk_space(struct disk_space *ds)
888 {
889 DSS_INFO_LOG("free: %uM/%uM (%u%%), %u%% inodes unused\n",
890 ds->free_mb, ds->total_mb, ds->percent_free,
891 ds->percent_free_inodes);
892 }
893
894 int com_prune(void)
895 {
896 int ret;
897 struct snapshot_list sl;
898 struct disk_space ds;
899
900 ret = get_disk_space(".", &ds);
901 if (ret < 0)
902 return ret;
903 log_disk_space(&ds);
904 for (;;) {
905 get_snapshot_list(&sl);
906 ret = remove_outdated_snapshot(&sl);
907 free_snapshot_list(&sl);
908 if (ret < 0)
909 return ret;
910 if (!ret)
911 break;
912 ret = wait_for_rm_process();
913 if (ret < 0)
914 goto out;
915 }
916 for (;;) {
917 get_snapshot_list(&sl);
918 ret = remove_redundant_snapshot(&sl);
919 free_snapshot_list(&sl);
920 if (ret < 0)
921 return ret;
922 if (!ret)
923 break;
924 ret = wait_for_rm_process();
925 if (ret < 0)
926 goto out;
927 }
928 return 1;
929 out:
930 return ret;
931 }
932
933 int com_create(void)
934 {
935 int ret, status;
936 char **rsync_argv;
937
938 create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
939 if (conf.dry_run_given) {
940 int i;
941 char *msg = NULL;
942 for (i = 0; rsync_argv[i]; i++) {
943 char *tmp = msg;
944 msg = make_message("%s%s%s", tmp? tmp : "",
945 tmp? " " : "", rsync_argv[i]);
946 free(tmp);
947 }
948 dss_msg("%s\n", msg);
949 free(msg);
950 return 1;
951 }
952 ret = create_snapshot(rsync_argv);
953 if (ret < 0)
954 goto out;
955 ret = wait_for_process(rsync_pid, &status);
956 if (ret < 0)
957 goto out;
958 ret = handle_rsync_exit(status);
959 out:
960 free_rsync_argv(rsync_argv);
961 return ret;
962 }
963
964 int com_ls(void)
965 {
966 int i;
967 struct snapshot_list sl;
968 struct snapshot *s;
969 get_snapshot_list(&sl);
970 FOR_EACH_SNAPSHOT(s, i, &sl)
971 dss_msg("%u\t%s\n", s->interval, s->name);
972 free_snapshot_list(&sl);
973 return 1;
974 }
975
976 __noreturn void clean_exit(int status)
977 {
978 free(dss_error_txt);
979 exit(status);
980 }
981
982 int read_config_file(void)
983 {
984 int ret;
985 char *config_file;
986 struct stat statbuf;
987
988 if (conf.config_file_given)
989 config_file = dss_strdup(conf.config_file_arg);
990 else {
991 char *home = get_homedir();
992 config_file = make_message("%s/.dssrc", home);
993 free(home);
994 }
995 ret = stat(config_file, &statbuf);
996 if (ret && conf.config_file_given) {
997 ret = -ERRNO_TO_DSS_ERROR(errno);
998 make_err_msg("failed to stat config file %s", config_file);
999 goto out;
1000 }
1001 if (!ret) {
1002 struct cmdline_parser_params params = {
1003 .override = 0,
1004 .initialize = 0,
1005 .check_required = 0,
1006 .check_ambiguity = 0
1007 };
1008 cmdline_parser_config_file(config_file, &conf, &params);
1009 }
1010 if (!conf.source_dir_given || !conf.dest_dir_given) {
1011 ret = -E_SYNTAX;
1012 make_err_msg("you need to specify both source_dir and dest_dir");
1013 goto out;
1014 }
1015 ret = 1;
1016 out:
1017 free(config_file);
1018 return ret;
1019 }
1020
1021 int check_config(void)
1022 {
1023 if (conf.unit_interval_arg <= 0) {
1024 make_err_msg("bad unit interval: %i", conf.unit_interval_arg);
1025 return -E_INVALID_NUMBER;
1026 }
1027 DSS_DEBUG_LOG("unit interval: %i day(s)\n", conf.unit_interval_arg);
1028 if (conf.num_intervals_arg <= 0) {
1029 make_err_msg("bad number of intervals %i", conf.num_intervals_arg);
1030 return -E_INVALID_NUMBER;
1031 }
1032 DSS_DEBUG_LOG("number of intervals: %i\n", conf.num_intervals_arg);
1033 return 1;
1034 }
1035
1036 static void setup_signal_handling(void)
1037 {
1038 int ret;
1039
1040 DSS_INFO_LOG("setting up signal handlers\n");
1041 signal_pipe = signal_init(); /* always successful */
1042 ret = install_sighandler(SIGINT);
1043 if (ret < 0)
1044 goto err;
1045 ret = install_sighandler(SIGTERM);
1046 if (ret < 0)
1047 goto err;
1048 ret = install_sighandler(SIGCHLD);
1049 if (ret < 0)
1050 goto err;
1051 return;
1052 err:
1053 DSS_EMERG_LOG("could not install signal handlers\n");
1054 exit(EXIT_FAILURE);
1055 }
1056
1057 int main(int argc, char **argv)
1058 {
1059 int ret;
1060
1061 cmdline_parser(argc, argv, &conf); /* aborts on errors */
1062 ret = read_config_file();
1063 if (ret < 0)
1064 goto out;
1065 ret = check_config();
1066 if (ret < 0)
1067 goto out;
1068 if (conf.logfile_given) {
1069 logfile = open_log(conf.logfile_arg);
1070 log_welcome(conf.loglevel_arg);
1071 }
1072 ret = dss_chdir(conf.dest_dir_arg);
1073 if (ret < 0)
1074 goto out;
1075 if (conf.daemon_given)
1076 daemon_init();
1077 setup_signal_handling();
1078 ret = call_command_handler();
1079 out:
1080 if (ret < 0)
1081 log_err_msg(EMERG, -ret);
1082 clean_exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
1083 }