Skip unnecessary check whether next snapshot is due.
[dss.git] / dss.c
1 /*
2 * Copyright (C) 2008-2009 Andre Noll <maan@systemlinux.org>
3 *
4 * Licensed under the GPL v2. For licencing details see COPYING.
5 */
6 #include <string.h>
7 #include <stdlib.h>
8 #include <stdarg.h>
9 #include <assert.h>
10 #include <errno.h>
11 #include <sys/types.h>
12 #include <signal.h>
13 #include <ctype.h>
14 #include <sys/stat.h>
15 #include <unistd.h>
16 #include <inttypes.h>
17 #include <sys/time.h>
18 #include <time.h>
19 #include <sys/wait.h>
20 #include <fnmatch.h>
21 #include <limits.h>
22
23
24 #include "gcc-compat.h"
25 #include "cmdline.h"
26 #include "log.h"
27 #include "string.h"
28 #include "error.h"
29 #include "fd.h"
30 #include "exec.h"
31 #include "daemon.h"
32 #include "signal.h"
33 #include "df.h"
34 #include "time.h"
35 #include "snap.h"
36
37 /** Command line and config file options. */
38 static struct gengetopt_args_info conf;
39 /** Non-NULL if we log to a file. */
40 static FILE *logfile;
41 /** The read end of the signal pipe */
42 static int signal_pipe;
43 /** Process id of current pre-create-hook/rsync/post-create-hook process. */
44 static pid_t create_pid;
45 /** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
46 static int create_process_stopped;
47 /** Process id of current pre-remove/rm/post-remove process. */
48 static pid_t remove_pid;
49 /** When the next snapshot is due. */
50 static int64_t next_snapshot_time;
51 /** When to try to remove something. */
52 static struct timeval next_removal_check;
53 /** Creation time of the snapshot currently being created. */
54 static int64_t current_snapshot_creation_time;
55 /** The snapshot currently being removed. */
56 struct snapshot *snapshot_currently_being_removed;
57 /** Needed by the post-create hook. */
58 static char *path_to_last_complete_snapshot;
59 static char *name_of_reference_snapshot;
60 /** \sa \ref snap.h for details. */
61 enum hook_status snapshot_creation_status;
62 /** \sa \ref snap.h for details. */
63 enum hook_status snapshot_removal_status;
64
65
66 DEFINE_DSS_ERRLIST;
67
68
69 /* a litte cpp magic helps to DRY */
70 #define COMMANDS \
71 COMMAND(ls) \
72 COMMAND(create) \
73 COMMAND(prune) \
74 COMMAND(run)
75 #define COMMAND(x) static int com_ ##x(void);
76 COMMANDS
77 #undef COMMAND
78 #define COMMAND(x) if (conf.x ##_given) return com_ ##x();
79 static int call_command_handler(void)
80 {
81 COMMANDS
82 DSS_EMERG_LOG("BUG: did not find command handler\n");
83 return -E_BUG;
84 }
85 #undef COMMAND
86 #undef COMMANDS
87
88 /**
89 * The log function of dss.
90 *
91 * \param ll Loglevel.
92 * \param fml Usual format string.
93 *
94 * All DSS_XXX_LOG() macros use this function.
95 */
96 __printf_2_3 void dss_log(int ll, const char* fmt,...)
97 {
98 va_list argp;
99 FILE *outfd;
100 struct tm *tm;
101 time_t t1;
102 char str[255] = "";
103
104 if (ll < conf.loglevel_arg)
105 return;
106 outfd = logfile? logfile : stderr;
107 time(&t1);
108 tm = localtime(&t1);
109 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
110 fprintf(outfd, "%s ", str);
111 if (conf.loglevel_arg <= INFO)
112 fprintf(outfd, "%i: ", ll);
113 va_start(argp, fmt);
114 vfprintf(outfd, fmt, argp);
115 va_end(argp);
116 }
117
118 /**
119 * Print a message either to stdout or to the log file.
120 */
121 static __printf_1_2 void dss_msg(const char* fmt,...)
122 {
123 FILE *outfd = conf.daemon_given? logfile : stdout;
124 va_list argp;
125 va_start(argp, fmt);
126 vfprintf(outfd, fmt, argp);
127 va_end(argp);
128 }
129
130 static int disk_space_low(void)
131 {
132 struct disk_space ds;
133 int ret = get_disk_space(".", &ds);
134
135 if (ret < 0)
136 return ret;
137 if (conf.min_free_mb_arg)
138 if (ds.free_mb < conf.min_free_mb_arg)
139 return 1;
140 if (conf.min_free_percent_arg)
141 if (ds.percent_free < conf.min_free_percent_arg)
142 return 1;
143 if (conf.min_free_percent_inodes_arg)
144 if (ds.percent_free_inodes < conf.min_free_percent_inodes_arg)
145 return 1;
146 return 0;
147 }
148
149 static void dss_get_snapshot_list(struct snapshot_list *sl)
150 {
151 get_snapshot_list(sl, conf.unit_interval_arg, conf.num_intervals_arg);
152 }
153
154 static int64_t compute_next_snapshot_time(void)
155 {
156 int64_t x = 0, now = get_current_time(), unit_interval
157 = 24 * 3600 * conf.unit_interval_arg, ret;
158 unsigned wanted = desired_number_of_snapshots(0, conf.num_intervals_arg),
159 num_complete_snapshots = 0;
160 int i;
161 struct snapshot *s = NULL;
162 struct snapshot_list sl;
163
164 dss_get_snapshot_list(&sl);
165 FOR_EACH_SNAPSHOT(s, i, &sl) {
166 if (!(s->flags & SS_COMPLETE))
167 continue;
168 num_complete_snapshots++;
169 x += s->completion_time - s->creation_time;
170 }
171 assert(x >= 0);
172
173 ret = now;
174 if (num_complete_snapshots == 0)
175 goto out;
176 x /= num_complete_snapshots; /* avg time to create one snapshot */
177 if (unit_interval < x * wanted) /* oops, no sleep at all */
178 goto out;
179 ret = s->completion_time + unit_interval / wanted - x;
180 out:
181 free_snapshot_list(&sl);
182 return ret;
183 }
184
185 static inline void invalidate_next_snapshot_time(void)
186 {
187 next_snapshot_time = 0;
188 }
189
190 static inline int next_snapshot_time_is_valid(void)
191 {
192 return next_snapshot_time != 0;
193 }
194
195 static int next_snapshot_is_due(void)
196 {
197 int64_t now = get_current_time();
198
199 if (!next_snapshot_time_is_valid())
200 next_snapshot_time = compute_next_snapshot_time();
201 if (next_snapshot_time <= now) {
202 DSS_DEBUG_LOG("next snapshot: now\n");
203 return 1;
204 }
205 DSS_DEBUG_LOG("next snapshot due in %" PRId64 " seconds\n",
206 next_snapshot_time - now);
207 return 0;
208 }
209
210 static int pre_create_hook(void)
211 {
212 int ret, fds[3] = {0, 0, 0};
213
214 assert(snapshot_creation_status == HS_READY);
215 /* make sure that the next snapshot time will be recomputed */
216 invalidate_next_snapshot_time();
217 if (!conf.pre_create_hook_given) {
218 snapshot_creation_status = HS_PRE_SUCCESS;
219 return 0;
220 }
221 DSS_DEBUG_LOG("executing %s\n", conf.pre_create_hook_arg);
222 ret = dss_exec_cmdline_pid(&create_pid,
223 conf.pre_create_hook_arg, fds);
224 if (ret < 0)
225 return ret;
226 snapshot_creation_status = HS_PRE_RUNNING;
227 return ret;
228 }
229
230 static int pre_remove_hook(struct snapshot *s, const char *why)
231 {
232 int ret, fds[3] = {0, 0, 0};
233 char *cmd;
234
235 if (!s)
236 return 0;
237 DSS_DEBUG_LOG("%s snapshot %s\n", why, s->name);
238 assert(snapshot_removal_status == HS_READY);
239 assert(remove_pid == 0);
240 assert(!snapshot_currently_being_removed);
241
242 snapshot_currently_being_removed = dss_malloc(sizeof(struct snapshot));
243 *snapshot_currently_being_removed = *s;
244 snapshot_currently_being_removed->name = dss_strdup(s->name);
245
246 if (!conf.pre_remove_hook_given) {
247 snapshot_removal_status = HS_PRE_SUCCESS;
248 return 0;
249 }
250 cmd = make_message("%s %s/%s", conf.pre_remove_hook_arg,
251 conf.dest_dir_arg, s->name);
252 DSS_DEBUG_LOG("executing %s\n", cmd);
253 ret = dss_exec_cmdline_pid(&remove_pid, cmd, fds);
254 free(cmd);
255 if (ret < 0)
256 return ret;
257 snapshot_removal_status = HS_PRE_RUNNING;
258 return ret;
259 }
260
261 static int exec_rm(void)
262 {
263 struct snapshot *s = snapshot_currently_being_removed;
264 int fds[3] = {0, 0, 0};
265 char *new_name = being_deleted_name(s);
266 char *argv[] = {"rm", "-rf", new_name, NULL};
267 int ret;
268
269 assert(snapshot_removal_status == HS_PRE_SUCCESS);
270 assert(remove_pid == 0);
271
272 DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
273 ret = dss_rename(s->name, new_name);
274 if (ret < 0)
275 goto out;
276 ret = dss_exec(&remove_pid, argv[0], argv, fds);
277 if (ret < 0)
278 goto out;
279 snapshot_removal_status = HS_RUNNING;
280 out:
281 free(new_name);
282 return ret;
283 }
284
285 static int snapshot_is_being_created(struct snapshot *s)
286 {
287 return s->creation_time == current_snapshot_creation_time;
288 }
289
290 static struct snapshot *find_orphaned_snapshot(struct snapshot_list *sl)
291 {
292 struct snapshot *s;
293 int i;
294
295 DSS_DEBUG_LOG("looking for orphaned snapshots\n");
296 FOR_EACH_SNAPSHOT(s, i, sl) {
297 if (snapshot_is_being_created(s))
298 continue;
299 /*
300 * We know that no rm is currently running, so if s is marked
301 * as being deleted, a previously started rm must have failed.
302 */
303 if (s->flags & SS_BEING_DELETED)
304 return s;
305
306 if (s->flags & SS_COMPLETE) /* good snapshot */
307 continue;
308 /*
309 * This snapshot is incomplete and it is not the snapshot
310 * currently being created. However, we must not remove it if
311 * rsync is about to be restarted. As only the newest snapshot
312 * can be restarted, this snapshot is orphaned if it is not the
313 * newest snapshot or if we are not about to restart rsync.
314 */
315 if (get_newest_snapshot(sl) != s)
316 return s;
317 if (snapshot_creation_status != HS_NEEDS_RESTART)
318 return s;
319 }
320 /* no orphaned snapshots */
321 return NULL;
322 }
323
324 static int is_reference_snapshot(struct snapshot *s)
325 {
326 if (!name_of_reference_snapshot)
327 return 0;
328 return strcmp(s->name, name_of_reference_snapshot)? 0 : 1;
329 }
330
331 /*
332 * return: 0: no redundant snapshots, 1: rm process started, negative: error
333 */
334 static struct snapshot *find_redundant_snapshot(struct snapshot_list *sl)
335 {
336 int i, interval;
337 struct snapshot *s;
338 unsigned missing = 0;
339
340 DSS_DEBUG_LOG("looking for intervals containing too many snapshots\n");
341 for (interval = conf.num_intervals_arg - 1; interval >= 0; interval--) {
342 unsigned keep = desired_number_of_snapshots(interval, conf.num_intervals_arg);
343 unsigned num = sl->interval_count[interval];
344 struct snapshot *victim = NULL, *prev = NULL;
345 int64_t score = LONG_MAX;
346
347 if (keep >= num)
348 missing += keep - num;
349 // DSS_DEBUG_LOG("interval %i: keep: %u, have: %u, missing: %u\n",
350 // interval, keep, num, missing);
351 if (keep + missing >= num)
352 continue;
353 /* redundant snapshot in this interval, pick snapshot with lowest score */
354 FOR_EACH_SNAPSHOT(s, i, sl) {
355 int64_t this_score;
356
357 if (snapshot_is_being_created(s))
358 continue;
359 if (is_reference_snapshot(s))
360 continue;
361 //DSS_DEBUG_LOG("checking %s\n", s->name);
362 if (s->interval > interval) {
363 prev = s;
364 continue;
365 }
366 if (s->interval < interval)
367 break;
368 if (!victim) {
369 victim = s;
370 prev = s;
371 continue;
372 }
373 assert(prev);
374 /* check if s is a better victim */
375 this_score = s->creation_time - prev->creation_time;
376 assert(this_score >= 0);
377 //DSS_DEBUG_LOG("%s: score %lli\n", s->name, (long long)score);
378 if (this_score < score) {
379 score = this_score;
380 victim = s;
381 }
382 prev = s;
383 }
384 assert(victim);
385 return victim;
386 }
387 return NULL;
388 }
389
390 static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
391 {
392 int i;
393 struct snapshot *s;
394
395 DSS_DEBUG_LOG("looking for snapshots belonging to intervals >= %d\n",
396 conf.num_intervals_arg);
397 FOR_EACH_SNAPSHOT(s, i, sl) {
398 if (snapshot_is_being_created(s))
399 continue;
400 if (is_reference_snapshot(s))
401 continue;
402 if (s->interval < conf.num_intervals_arg)
403 continue;
404 return s;
405 }
406 return NULL;
407 }
408
409 struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
410 {
411 int i;
412 struct snapshot *s;
413 FOR_EACH_SNAPSHOT(s, i, sl) {
414 if (snapshot_is_being_created(s))
415 continue;
416 if (is_reference_snapshot(s))
417 continue;
418 DSS_INFO_LOG("oldest removable snapshot: %s\n", s->name);
419 return s;
420 }
421 return NULL;
422 }
423
424 static int rename_incomplete_snapshot(int64_t start)
425 {
426 char *old_name;
427 int ret;
428
429 free(path_to_last_complete_snapshot);
430 ret = complete_name(start, get_current_time(),
431 &path_to_last_complete_snapshot);
432 if (ret < 0)
433 return ret;
434 old_name = incomplete_name(start);
435 ret = dss_rename(old_name, path_to_last_complete_snapshot);
436 if (ret >= 0)
437 DSS_NOTICE_LOG("%s -> %s\n", old_name,
438 path_to_last_complete_snapshot);
439 free(old_name);
440 return ret;
441 }
442
443 static int try_to_free_disk_space(void)
444 {
445 int ret;
446 struct snapshot_list sl;
447 struct snapshot *victim;
448 struct timeval now;
449 const char *why;
450 int low_disk_space;
451
452 ret = disk_space_low();
453 if (ret < 0)
454 return ret;
455 low_disk_space = ret;
456 gettimeofday(&now, NULL);
457 if (tv_diff(&next_removal_check, &now, NULL) > 0)
458 return 0;
459 if (!low_disk_space) {
460 if (conf.keep_redundant_given)
461 return 0;
462 if (snapshot_creation_status != HS_READY)
463 return 0;
464 if (next_snapshot_is_due())
465 return 0;
466 }
467 dss_get_snapshot_list(&sl);
468 ret = 0;
469 if (!low_disk_space && sl.num_snapshots <= 1)
470 goto out;
471 why = "outdated";
472 victim = find_outdated_snapshot(&sl);
473 if (victim)
474 goto remove;
475 why = "redundant";
476 victim = find_redundant_snapshot(&sl);
477 if (victim)
478 goto remove;
479 /* try harder only if disk space is low */
480 if (!low_disk_space)
481 goto out;
482 why = "orphaned";
483 victim = find_orphaned_snapshot(&sl);
484 if (victim)
485 goto remove;
486 DSS_WARNING_LOG("disk space low and nothing obvious to remove\n");
487 victim = find_oldest_removable_snapshot(&sl);
488 if (victim)
489 goto remove;
490 DSS_CRIT_LOG("uhuhu: disk space low and nothing to remove\n");
491 ret = -ERRNO_TO_DSS_ERROR(ENOSPC);
492 goto out;
493 remove:
494 ret = pre_remove_hook(victim, why);
495 out:
496 free_snapshot_list(&sl);
497 return ret;
498 }
499
500 static int post_create_hook(void)
501 {
502 int ret, fds[3] = {0, 0, 0};
503 char *cmd;
504
505 if (!conf.post_create_hook_given) {
506 create_pid = 0;
507 snapshot_creation_status = HS_READY;
508 return 0;
509 }
510 cmd = make_message("%s %s/%s", conf.post_create_hook_arg,
511 conf.dest_dir_arg, path_to_last_complete_snapshot);
512 DSS_NOTICE_LOG("executing %s\n", cmd);
513 ret = dss_exec_cmdline_pid(&create_pid, cmd, fds);
514 free(cmd);
515 if (ret < 0)
516 return ret;
517 snapshot_creation_status = HS_POST_RUNNING;
518 return ret;
519 }
520
521 static int post_remove_hook(void)
522 {
523 int ret, fds[3] = {0, 0, 0};
524 char *cmd;
525 struct snapshot *s = snapshot_currently_being_removed;
526
527 assert(s);
528
529 if (!conf.post_remove_hook_given) {
530 snapshot_removal_status = HS_READY;
531 return 0;
532 }
533 cmd = make_message("%s %s/%s", conf.post_remove_hook_arg,
534 conf.dest_dir_arg, s->name);
535 DSS_NOTICE_LOG("executing %s\n", cmd);
536 ret = dss_exec_cmdline_pid(&remove_pid, cmd, fds);
537 free(cmd);
538 if (ret < 0)
539 return ret;
540 snapshot_removal_status = HS_POST_RUNNING;
541 return ret;
542 }
543
544 static void kill_process(pid_t pid)
545 {
546 if (!pid)
547 return;
548 DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
549 kill(pid, SIGTERM);
550 }
551
552 static void stop_create_process(void)
553 {
554 if (!create_pid || create_process_stopped)
555 return;
556 DSS_INFO_LOG("suspending create process %d\n", (int)create_pid);
557 kill(SIGSTOP, create_pid);
558 create_process_stopped = 1;
559 }
560
561 static void restart_create_process(void)
562 {
563 if (!create_pid || !create_process_stopped)
564 return;
565 DSS_INFO_LOG("resuming create process %d\n", (int)create_pid);
566 kill (SIGCONT, create_pid);
567 create_process_stopped = 0;
568 }
569
570 /**
571 * Print a log message about the exit status of a child.
572 */
573 static void log_termination_msg(pid_t pid, int status)
574 {
575 if (WIFEXITED(status))
576 DSS_INFO_LOG("child %i exited. Exit status: %i\n", (int)pid,
577 WEXITSTATUS(status));
578 else if (WIFSIGNALED(status))
579 DSS_NOTICE_LOG("child %i was killed by signal %i\n", (int)pid,
580 WTERMSIG(status));
581 else
582 DSS_WARNING_LOG("child %i terminated abormally\n", (int)pid);
583 }
584
585 static int wait_for_process(pid_t pid, int *status)
586 {
587 int ret;
588
589 DSS_DEBUG_LOG("Waiting for process %d to terminate\n", (int)pid);
590 for (;;) {
591 fd_set rfds;
592
593 FD_ZERO(&rfds);
594 FD_SET(signal_pipe, &rfds);
595 ret = dss_select(signal_pipe + 1, &rfds, NULL, NULL);
596 if (ret < 0)
597 break;
598 ret = next_signal();
599 if (!ret)
600 continue;
601 if (ret == SIGCHLD) {
602 ret = waitpid(pid, status, 0);
603 if (ret >= 0)
604 break;
605 if (errno != EINTR) { /* error */
606 ret = -ERRNO_TO_DSS_ERROR(errno);
607 break;
608 }
609 }
610 /* SIGINT or SIGTERM */
611 DSS_WARNING_LOG("sending SIGTERM to pid %d\n", (int)pid);
612 kill(pid, SIGTERM);
613 }
614 if (ret < 0)
615 DSS_ERROR_LOG("failed to wait for process %d\n", (int)pid);
616 else
617 log_termination_msg(pid, *status);
618 return ret;
619 }
620
621 static void handle_pre_remove_exit(int status)
622 {
623 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
624 snapshot_removal_status = HS_READY;
625 gettimeofday(&next_removal_check, NULL);
626 next_removal_check.tv_sec += 60;
627 return;
628 }
629 snapshot_removal_status = HS_PRE_SUCCESS;
630 }
631
632 static int handle_rm_exit(int status)
633 {
634 if (!WIFEXITED(status)) {
635 snapshot_removal_status = HS_READY;
636 return -E_INVOLUNTARY_EXIT;
637 }
638 if (WEXITSTATUS(status)) {
639 snapshot_removal_status = HS_READY;
640 return -E_BAD_EXIT_CODE;
641 }
642 if (conf.post_remove_hook_given)
643 snapshot_removal_status = HS_SUCCESS;
644 else
645 snapshot_removal_status = HS_READY;
646 return 1;
647 }
648
649 static void handle_post_remove_exit(void)
650 {
651 snapshot_removal_status = HS_READY;
652 }
653
654 static int handle_remove_exit(int status)
655 {
656 int ret;
657 struct snapshot *s = snapshot_currently_being_removed;
658
659 assert(s);
660 switch (snapshot_removal_status) {
661 case HS_PRE_RUNNING:
662 handle_pre_remove_exit(status);
663 ret = 1;
664 break;
665 case HS_RUNNING:
666 ret = handle_rm_exit(status);
667 break;
668 case HS_POST_RUNNING:
669 handle_post_remove_exit();
670 ret = 1;
671 break;
672 default:
673 ret = -E_BUG;
674 }
675 if (snapshot_removal_status == HS_READY) {
676 free(s->name);
677 free(s);
678 snapshot_currently_being_removed = NULL;
679 }
680 remove_pid = 0;
681 return ret;
682 }
683
684 static int wait_for_remove_process(void)
685 {
686 int status, ret;
687
688 assert(remove_pid);
689 assert(
690 snapshot_removal_status == HS_PRE_RUNNING ||
691 snapshot_removal_status == HS_RUNNING ||
692 snapshot_removal_status == HS_POST_RUNNING
693 );
694 ret = wait_for_process(remove_pid, &status);
695 if (ret < 0)
696 return ret;
697 return handle_remove_exit(status);
698 }
699
700 static int handle_rsync_exit(int status)
701 {
702 int es, ret;
703
704 if (!WIFEXITED(status)) {
705 DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)create_pid);
706 ret = -E_INVOLUNTARY_EXIT;
707 snapshot_creation_status = HS_READY;
708 goto out;
709 }
710 es = WEXITSTATUS(status);
711 /*
712 * Restart rsync on non-fatal errors:
713 * 12: Error in rsync protocol data stream
714 * 13: Errors with program diagnostics
715 */
716 if (es == 12 || es == 13) {
717 DSS_WARNING_LOG("rsync process %d returned %d -- restarting\n",
718 (int)create_pid, es);
719 snapshot_creation_status = HS_NEEDS_RESTART;
720 next_snapshot_time = get_current_time() + 60;
721 ret = 1;
722 goto out;
723 }
724 if (es != 0 && es != 23 && es != 24) {
725 DSS_ERROR_LOG("rsync process %d returned %d\n", (int)create_pid, es);
726 ret = -E_BAD_EXIT_CODE;
727 snapshot_creation_status = HS_READY;
728 goto out;
729 }
730 ret = rename_incomplete_snapshot(current_snapshot_creation_time);
731 if (ret < 0)
732 goto out;
733 snapshot_creation_status = HS_SUCCESS;
734 free(name_of_reference_snapshot);
735 name_of_reference_snapshot = NULL;
736 out:
737 create_process_stopped = 0;
738 return ret;
739 }
740
741 static int handle_pre_create_hook_exit(int status)
742 {
743 int es, ret;
744 static int warn_count;
745
746 if (!WIFEXITED(status)) {
747 snapshot_creation_status = HS_READY;
748 ret = -E_INVOLUNTARY_EXIT;
749 goto out;
750 }
751 es = WEXITSTATUS(status);
752 if (es) {
753 if (!warn_count--) {
754 DSS_NOTICE_LOG("pre_create_hook %s returned %d\n",
755 conf.pre_create_hook_arg, es);
756 DSS_NOTICE_LOG("deferring snapshot creation...\n");
757 warn_count = 60; /* warn only once per hour */
758 }
759 next_snapshot_time = get_current_time() + 60;
760 snapshot_creation_status = HS_READY;
761 ret = 0;
762 goto out;
763 }
764 warn_count = 0;
765 snapshot_creation_status = HS_PRE_SUCCESS;
766 ret = 1;
767 out:
768 return ret;
769 }
770
771 static int handle_sigchld(void)
772 {
773 pid_t pid;
774 int status, ret = reap_child(&pid, &status);
775
776 if (ret <= 0)
777 return ret;
778
779 if (pid == create_pid) {
780 switch (snapshot_creation_status) {
781 case HS_PRE_RUNNING:
782 ret = handle_pre_create_hook_exit(status);
783 break;
784 case HS_RUNNING:
785 ret = handle_rsync_exit(status);
786 break;
787 case HS_POST_RUNNING:
788 snapshot_creation_status = HS_READY;
789 ret = 1;
790 break;
791 default:
792 DSS_EMERG_LOG("BUG: create can't die in status %d\n",
793 snapshot_creation_status);
794 return -E_BUG;
795 }
796 create_pid = 0;
797 return ret;
798 }
799 if (pid == remove_pid) {
800 ret = handle_remove_exit(status);
801 if (ret < 0)
802 return ret;
803 return ret;
804 }
805 DSS_EMERG_LOG("BUG: unknown process %d died\n", (int)pid);
806 return -E_BUG;
807 }
808
809 static int check_config(void)
810 {
811 if (conf.unit_interval_arg <= 0) {
812 DSS_ERROR_LOG("bad unit interval: %i\n", conf.unit_interval_arg);
813 return -E_INVALID_NUMBER;
814 }
815 DSS_DEBUG_LOG("unit interval: %i day(s)\n", conf.unit_interval_arg);
816 if (conf.num_intervals_arg <= 0) {
817 DSS_ERROR_LOG("bad number of intervals %i\n", conf.num_intervals_arg);
818 return -E_INVALID_NUMBER;
819 }
820 DSS_DEBUG_LOG("number of intervals: %i\n", conf.num_intervals_arg);
821 return 1;
822 }
823
824 /*
825 * Returns < 0 on errors, 0 if no config file is given and > 0 if the config
826 * file was read successfully.
827 */
828 static int parse_config_file(int override)
829 {
830 int ret, config_file_exists;
831 char *config_file;
832 struct stat statbuf;
833 char *old_logfile_arg = NULL;
834 int old_daemon_given = 0;
835
836 if (conf.config_file_given)
837 config_file = dss_strdup(conf.config_file_arg);
838 else {
839 char *home = get_homedir();
840 config_file = make_message("%s/.dssrc", home);
841 free(home);
842 }
843 if (override) { /* SIGHUP */
844 if (conf.logfile_given)
845 old_logfile_arg = dss_strdup(conf.logfile_arg);
846 old_daemon_given = conf.daemon_given;
847 }
848
849 config_file_exists = !stat(config_file, &statbuf);
850 if (!config_file_exists && conf.config_file_given) {
851 ret = -ERRNO_TO_DSS_ERROR(errno);
852 DSS_ERROR_LOG("failed to stat config file %s\n", config_file);
853 goto out;
854 }
855 if (config_file_exists) {
856 struct cmdline_parser_params params = {
857 .override = override,
858 .initialize = 0,
859 .check_required = 1,
860 .check_ambiguity = 0,
861 .print_errors = 1
862 };
863 if (override) { /* invalidate all rsync options */
864 int i;
865
866 for (i = 0; i < conf.rsync_option_given; i++) {
867 free(conf.rsync_option_arg[i]);
868 conf.rsync_option_arg[i] = NULL;
869 }
870 conf.rsync_option_given = 0;
871 }
872 cmdline_parser_config_file(config_file, &conf, &params);
873 }
874 ret = check_config();
875 if (ret < 0)
876 goto out;
877 if (override) {
878 /* don't change daemon mode on SIGHUP */
879 conf.daemon_given = old_daemon_given;
880 close_log(logfile);
881 logfile = NULL;
882 if (conf.logfile_given)
883 free(old_logfile_arg);
884 else if (conf.daemon_given) { /* re-use old logfile */
885 conf.logfile_arg = old_logfile_arg;
886 conf.logfile_given = 1;
887 }
888 }
889 if (conf.logfile_given) {
890 logfile = open_log(conf.logfile_arg);
891 log_welcome(conf.loglevel_arg);
892 }
893 DSS_DEBUG_LOG("loglevel: %d\n", conf.loglevel_arg);
894 ret = config_file_exists;
895 out:
896 free(config_file);
897 if (ret < 0)
898 DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
899 return ret;
900 }
901
902 static int change_to_dest_dir(void)
903 {
904 DSS_INFO_LOG("changing cwd to %s\n", conf.dest_dir_arg);
905 return dss_chdir(conf.dest_dir_arg);
906 }
907
908 static void dump_dss_config(const char *msg)
909 {
910 if (conf.loglevel_arg > INFO)
911 return;
912 DSS_INFO_LOG("%s\n", msg);
913 cmdline_parser_dump(logfile? logfile : stderr, &conf);
914 }
915
916 static int handle_sighup(void)
917 {
918 int ret;
919
920 DSS_NOTICE_LOG("SIGHUP, re-reading config\n");
921 dump_dss_config("current config");
922 ret = parse_config_file(1);
923 if (ret < 0)
924 return ret;
925 dump_dss_config("new config");
926 invalidate_next_snapshot_time();
927 return change_to_dest_dir();
928 }
929
930 static int handle_signal(void)
931 {
932 int sig, ret = next_signal();
933
934 if (ret <= 0)
935 goto out;
936 sig = ret;
937 switch (sig) {
938 case SIGINT:
939 case SIGTERM:
940 restart_create_process();
941 kill_process(create_pid);
942 kill_process(remove_pid);
943 ret = -E_SIGNAL;
944 break;
945 case SIGHUP:
946 ret = handle_sighup();
947 break;
948 case SIGCHLD:
949 ret = handle_sigchld();
950 break;
951 }
952 out:
953 if (ret < 0)
954 DSS_ERROR_LOG("%s\n", dss_strerror(-ret));
955 return ret;
956 }
957
958 /*
959 * We can not use rsync locally if the local user is different from the remote
960 * user or if the src dir is not on the local host (or both).
961 */
962 static int use_rsync_locally(char *logname)
963 {
964 char *h = conf.remote_host_arg;
965
966 if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
967 return 0;
968 if (conf.remote_user_given && strcmp(conf.remote_user_arg, logname))
969 return 0;
970 return 1;
971 }
972
973 static int rename_resume_snap(int64_t creation_time)
974 {
975 struct snapshot_list sl = {.num_snapshots = 0};
976 struct snapshot *s;
977 char *new_name = incomplete_name(creation_time);
978 int ret;
979
980 ret = 0;
981 if (conf.no_resume_given)
982 goto out;
983 dss_get_snapshot_list(&sl);
984 s = get_newest_snapshot(&sl);
985 if (!s)
986 goto out;
987 if ((s->flags & SS_COMPLETE) != 0) /* complete */
988 goto out;
989 DSS_INFO_LOG("resuming: reusing %s as destination dir\n", s->name);
990 ret = dss_rename(s->name, new_name);
991 out:
992 if (ret >= 0)
993 DSS_NOTICE_LOG("creating new snapshot %s\n", new_name);
994 free(new_name);
995 free_snapshot_list(&sl);
996 return ret;
997 }
998
999 static void create_rsync_argv(char ***argv, int64_t *num)
1000 {
1001 char *logname;
1002 int i = 0, j;
1003 struct snapshot_list sl;
1004
1005 dss_get_snapshot_list(&sl);
1006 assert(!name_of_reference_snapshot);
1007 name_of_reference_snapshot = name_of_newest_complete_snapshot(&sl);
1008 free_snapshot_list(&sl);
1009
1010 *argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
1011 (*argv)[i++] = dss_strdup("rsync");
1012 (*argv)[i++] = dss_strdup("-aq");
1013 (*argv)[i++] = dss_strdup("--delete");
1014 for (j = 0; j < conf.rsync_option_given; j++)
1015 (*argv)[i++] = dss_strdup(conf.rsync_option_arg[j]);
1016 if (name_of_reference_snapshot) {
1017 DSS_INFO_LOG("using %s as reference\n", name_of_reference_snapshot);
1018 (*argv)[i++] = make_message("--link-dest=../%s",
1019 name_of_reference_snapshot);
1020 } else
1021 DSS_INFO_LOG("no suitable reference snapshot found\n");
1022 logname = dss_logname();
1023 if (use_rsync_locally(logname))
1024 (*argv)[i++] = dss_strdup(conf.source_dir_arg);
1025 else
1026 (*argv)[i++] = make_message("%s@%s:%s/", conf.remote_user_given?
1027 conf.remote_user_arg : logname,
1028 conf.remote_host_arg, conf.source_dir_arg);
1029 free(logname);
1030 *num = get_current_time();
1031 (*argv)[i++] = incomplete_name(*num);
1032 (*argv)[i++] = NULL;
1033 for (j = 0; j < i; j++)
1034 DSS_DEBUG_LOG("argv[%d] = %s\n", j, (*argv)[j]);
1035 }
1036
1037 static void free_rsync_argv(char **argv)
1038 {
1039 int i;
1040
1041 if (!argv)
1042 return;
1043 for (i = 0; argv[i]; i++)
1044 free(argv[i]);
1045 free(argv);
1046 }
1047
1048 static int create_snapshot(char **argv)
1049 {
1050 int ret, fds[3] = {0, 0, 0};
1051
1052 ret = rename_resume_snap(current_snapshot_creation_time);
1053 if (ret < 0)
1054 return ret;
1055 ret = dss_exec(&create_pid, argv[0], argv, fds);
1056 if (ret < 0)
1057 return ret;
1058 snapshot_creation_status = HS_RUNNING;
1059 return ret;
1060 }
1061
1062 static int select_loop(void)
1063 {
1064 int ret;
1065 /* check every 60 seconds for free disk space */
1066 struct timeval tv;
1067 char **rsync_argv = NULL;
1068
1069 for (;;) {
1070 fd_set rfds;
1071 struct timeval *tvp;
1072
1073 if (remove_pid)
1074 tvp = NULL; /* sleep until rm hook/process dies */
1075 else { /* sleep one minute */
1076 tv.tv_sec = 60;
1077 tv.tv_usec = 0;
1078 tvp = &tv;
1079 }
1080 FD_ZERO(&rfds);
1081 FD_SET(signal_pipe, &rfds);
1082 ret = dss_select(signal_pipe + 1, &rfds, NULL, tvp);
1083 if (ret < 0)
1084 goto out;
1085 if (FD_ISSET(signal_pipe, &rfds)) {
1086 ret = handle_signal();
1087 if (ret < 0)
1088 goto out;
1089 }
1090 if (remove_pid)
1091 continue;
1092 if (snapshot_removal_status == HS_PRE_SUCCESS) {
1093 ret = exec_rm();
1094 if (ret < 0)
1095 goto out;
1096 continue;
1097 }
1098 if (snapshot_removal_status == HS_SUCCESS) {
1099 ret = post_remove_hook();
1100 if (ret < 0)
1101 goto out;
1102 continue;
1103 }
1104 ret = try_to_free_disk_space();
1105 if (ret < 0)
1106 goto out;
1107 if (snapshot_removal_status != HS_READY) {
1108 stop_create_process();
1109 continue;
1110 }
1111 restart_create_process();
1112 switch (snapshot_creation_status) {
1113 case HS_READY:
1114 if (!next_snapshot_is_due())
1115 continue;
1116 ret = pre_create_hook();
1117 if (ret < 0)
1118 goto out;
1119 continue;
1120 case HS_PRE_RUNNING:
1121 case HS_RUNNING:
1122 case HS_POST_RUNNING:
1123 continue;
1124 case HS_PRE_SUCCESS:
1125 if (!name_of_reference_snapshot) {
1126 free_rsync_argv(rsync_argv);
1127 create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
1128 }
1129 ret = create_snapshot(rsync_argv);
1130 if (ret < 0)
1131 goto out;
1132 continue;
1133 case HS_NEEDS_RESTART:
1134 if (!next_snapshot_is_due())
1135 continue;
1136 ret = create_snapshot(rsync_argv);
1137 if (ret < 0)
1138 goto out;
1139 continue;
1140 case HS_SUCCESS:
1141 ret = post_create_hook();
1142 if (ret < 0)
1143 goto out;
1144 continue;
1145 }
1146 }
1147 out:
1148 return ret;
1149 }
1150
1151 static void exit_hook(int exit_code)
1152 {
1153 int fds[3] = {0, 0, 0};
1154 char *argv[] = {conf.exit_hook_arg, dss_strerror(-exit_code), NULL};
1155 pid_t pid;
1156
1157 if (!conf.exit_hook_given)
1158 return;
1159 DSS_NOTICE_LOG("executing %s %s\n", argv[0], argv[1]);
1160 dss_exec(&pid, conf.exit_hook_arg, argv, fds);
1161 }
1162
1163 static int com_run(void)
1164 {
1165 int ret;
1166
1167 if (conf.dry_run_given) {
1168 DSS_ERROR_LOG("dry_run not supported by this command\n");
1169 return -E_SYNTAX;
1170 }
1171 ret = install_sighandler(SIGHUP);
1172 if (ret < 0)
1173 return ret;
1174 ret = select_loop();
1175 if (ret >= 0) /* impossible */
1176 ret = -E_BUG;
1177 exit_hook(ret);
1178 return ret;
1179 }
1180
1181 static int com_prune(void)
1182 {
1183 int ret;
1184 struct snapshot_list sl;
1185 struct snapshot *victim;
1186 struct disk_space ds;
1187 const char *why;
1188
1189 ret = get_disk_space(".", &ds);
1190 if (ret < 0)
1191 return ret;
1192 log_disk_space(&ds);
1193 dss_get_snapshot_list(&sl);
1194 why = "outdated";
1195 victim = find_outdated_snapshot(&sl);
1196 if (victim)
1197 goto rm;
1198 why = "redundant";
1199 victim = find_redundant_snapshot(&sl);
1200 if (victim)
1201 goto rm;
1202 ret = 0;
1203 goto out;
1204 rm:
1205 if (conf.dry_run_given) {
1206 dss_msg("%s snapshot %s (interval = %i)\n",
1207 why, victim->name, victim->interval);
1208 ret = 0;
1209 goto out;
1210 }
1211 ret = pre_remove_hook(victim, why);
1212 if (ret < 0)
1213 goto out;
1214 if (snapshot_removal_status == HS_PRE_RUNNING) {
1215 ret = wait_for_remove_process();
1216 if (ret < 0)
1217 goto out;
1218 if (snapshot_removal_status != HS_PRE_SUCCESS)
1219 goto out;
1220 }
1221 ret = exec_rm();
1222 if (ret < 0)
1223 goto out;
1224 ret = wait_for_remove_process();
1225 if (ret < 0)
1226 goto out;
1227 if (snapshot_removal_status != HS_SUCCESS)
1228 goto out;
1229 ret = post_remove_hook();
1230 if (ret < 0)
1231 goto out;
1232 if (snapshot_removal_status != HS_POST_RUNNING)
1233 goto out;
1234 ret = wait_for_remove_process();
1235 if (ret < 0)
1236 goto out;
1237 ret = 1;
1238 out:
1239 free_snapshot_list(&sl);
1240 return ret;
1241 }
1242
1243 static int com_create(void)
1244 {
1245 int ret, status;
1246 char **rsync_argv;
1247
1248 if (conf.dry_run_given) {
1249 int i;
1250 char *msg = NULL;
1251 create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
1252 for (i = 0; rsync_argv[i]; i++) {
1253 char *tmp = msg;
1254 msg = make_message("%s%s%s", tmp? tmp : "",
1255 tmp? " " : "", rsync_argv[i]);
1256 free(tmp);
1257 }
1258 free_rsync_argv(rsync_argv);
1259 dss_msg("%s\n", msg);
1260 free(msg);
1261 return 1;
1262 }
1263 ret = pre_create_hook();
1264 if (ret < 0)
1265 return ret;
1266 if (create_pid) {
1267 ret = wait_for_process(create_pid, &status);
1268 if (ret < 0)
1269 return ret;
1270 ret = handle_pre_create_hook_exit(status);
1271 if (ret <= 0) /* error, or pre-create failed */
1272 return ret;
1273 }
1274 create_rsync_argv(&rsync_argv, &current_snapshot_creation_time);
1275 ret = create_snapshot(rsync_argv);
1276 if (ret < 0)
1277 goto out;
1278 ret = wait_for_process(create_pid, &status);
1279 if (ret < 0)
1280 goto out;
1281 ret = handle_rsync_exit(status);
1282 if (ret < 0)
1283 goto out;
1284 post_create_hook();
1285 if (create_pid)
1286 ret = wait_for_process(create_pid, &status);
1287 out:
1288 free_rsync_argv(rsync_argv);
1289 return ret;
1290 }
1291
1292 static int com_ls(void)
1293 {
1294 int i;
1295 struct snapshot_list sl;
1296 struct snapshot *s;
1297
1298 dss_get_snapshot_list(&sl);
1299 FOR_EACH_SNAPSHOT(s, i, &sl) {
1300 int64_t d = 0;
1301 if (s->flags & SS_COMPLETE)
1302 d = (s->completion_time - s->creation_time) / 60;
1303 dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval, s->name, d/60, d%60);
1304 };
1305 free_snapshot_list(&sl);
1306 return 1;
1307 }
1308
1309 static int setup_signal_handling(void)
1310 {
1311 int ret;
1312
1313 DSS_INFO_LOG("setting up signal handlers\n");
1314 signal_pipe = signal_init(); /* always successful */
1315 ret = install_sighandler(SIGINT);
1316 if (ret < 0)
1317 return ret;
1318 ret = install_sighandler(SIGTERM);
1319 if (ret < 0)
1320 return ret;
1321 return install_sighandler(SIGCHLD);
1322 }
1323
1324 /**
1325 * The main function of dss.
1326 *
1327 * \param argc Usual argument count.
1328 * \param argv Usual argument vector.
1329 */
1330 int main(int argc, char **argv)
1331 {
1332 int ret;
1333 struct cmdline_parser_params params = {
1334 .override = 0,
1335 .initialize = 1,
1336 .check_required = 0,
1337 .check_ambiguity = 0,
1338 .print_errors = 1
1339 };
1340
1341 cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
1342 ret = parse_config_file(0);
1343 if (ret < 0)
1344 goto out;
1345 if (ret == 0) { /* no config file given */
1346 /*
1347 * Parse the command line options again, but this time check
1348 * that all required options are given.
1349 */
1350 params = (struct cmdline_parser_params) {
1351 .override = 1,
1352 .initialize = 1,
1353 .check_required = 1,
1354 .check_ambiguity = 1,
1355 .print_errors = 1
1356 };
1357 cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
1358 }
1359 if (conf.daemon_given)
1360 daemon_init();
1361 dump_dss_config("dss configuration");
1362 ret = change_to_dest_dir();
1363 if (ret < 0)
1364 goto out;
1365 ret = setup_signal_handling();
1366 if (ret < 0)
1367 goto out;
1368 ret = call_command_handler();
1369 out:
1370 if (ret < 0)
1371 DSS_EMERG_LOG("%s\n", dss_strerror(-ret));
1372 exit(ret >= 0? EXIT_SUCCESS : EXIT_FAILURE);
1373 }