]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/journal/sd-journal.c
Merge pull request #653 from dvdhrm/bus-gold
[thirdparty/systemd.git] / src / journal / sd-journal.c
1 /***
2 This file is part of systemd.
3
4 Copyright 2011 Lennart Poettering
5
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
10
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
18 ***/
19
20 #include <errno.h>
21 #include <fcntl.h>
22 #include <inttypes.h>
23 #include <linux/magic.h>
24 #include <poll.h>
25 #include <stddef.h>
26 #include <sys/inotify.h>
27 #include <sys/vfs.h>
28 #include <unistd.h>
29
30 #include "sd-journal.h"
31
32 #include "alloc-util.h"
33 #include "catalog.h"
34 #include "compress.h"
35 #include "dirent-util.h"
36 #include "fd-util.h"
37 #include "fileio.h"
38 #include "formats-util.h"
39 #include "fs-util.h"
40 #include "hashmap.h"
41 #include "hostname-util.h"
42 #include "io-util.h"
43 #include "journal-def.h"
44 #include "journal-file.h"
45 #include "journal-internal.h"
46 #include "list.h"
47 #include "lookup3.h"
48 #include "missing.h"
49 #include "path-util.h"
50 #include "replace-var.h"
51 #include "stat-util.h"
52 #include "stdio-util.h"
53 #include "string-util.h"
54 #include "strv.h"
55
56 #define JOURNAL_FILES_MAX 7168
57
58 #define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC)
59
60 #define REPLACE_VAR_MAX 256
61
62 #define DEFAULT_DATA_THRESHOLD (64*1024)
63
64 static void remove_file_real(sd_journal *j, JournalFile *f);
65
66 static bool journal_pid_changed(sd_journal *j) {
67 assert(j);
68
69 /* We don't support people creating a journal object and
70 * keeping it around over a fork(). Let's complain. */
71
72 return j->original_pid != getpid();
73 }
74
75 static int journal_put_error(sd_journal *j, int r, const char *path) {
76 char *copy;
77 int k;
78
79 /* Memorize an error we encountered, and store which
80 * file/directory it was generated from. Note that we store
81 * only *one* path per error code, as the error code is the
82 * key into the hashmap, and the path is the value. This means
83 * we keep track only of all error kinds, but not of all error
84 * locations. This has the benefit that the hashmap cannot
85 * grow beyond bounds.
86 *
87 * We return an error here only if we didn't manage to
88 * memorize the real error. */
89
90 if (r >= 0)
91 return r;
92
93 k = hashmap_ensure_allocated(&j->errors, NULL);
94 if (k < 0)
95 return k;
96
97 if (path) {
98 copy = strdup(path);
99 if (!copy)
100 return -ENOMEM;
101 } else
102 copy = NULL;
103
104 k = hashmap_put(j->errors, INT_TO_PTR(r), copy);
105 if (k < 0) {
106 free(copy);
107
108 if (k == -EEXIST)
109 return 0;
110
111 return k;
112 }
113
114 return 0;
115 }
116
117 static void detach_location(sd_journal *j) {
118 Iterator i;
119 JournalFile *f;
120
121 assert(j);
122
123 j->current_file = NULL;
124 j->current_field = 0;
125
126 ORDERED_HASHMAP_FOREACH(f, j->files, i)
127 journal_file_reset_location(f);
128 }
129
130 static void reset_location(sd_journal *j) {
131 assert(j);
132
133 detach_location(j);
134 zero(j->current_location);
135 }
136
137 static void init_location(Location *l, LocationType type, JournalFile *f, Object *o) {
138 assert(l);
139 assert(type == LOCATION_DISCRETE || type == LOCATION_SEEK);
140 assert(f);
141 assert(o->object.type == OBJECT_ENTRY);
142
143 l->type = type;
144 l->seqnum = le64toh(o->entry.seqnum);
145 l->seqnum_id = f->header->seqnum_id;
146 l->realtime = le64toh(o->entry.realtime);
147 l->monotonic = le64toh(o->entry.monotonic);
148 l->boot_id = o->entry.boot_id;
149 l->xor_hash = le64toh(o->entry.xor_hash);
150
151 l->seqnum_set = l->realtime_set = l->monotonic_set = l->xor_hash_set = true;
152 }
153
154 static void set_location(sd_journal *j, JournalFile *f, Object *o) {
155 assert(j);
156 assert(f);
157 assert(o);
158
159 init_location(&j->current_location, LOCATION_DISCRETE, f, o);
160
161 j->current_file = f;
162 j->current_field = 0;
163
164 /* Let f know its candidate entry was picked. */
165 assert(f->location_type == LOCATION_SEEK);
166 f->location_type = LOCATION_DISCRETE;
167 }
168
169 static int match_is_valid(const void *data, size_t size) {
170 const char *b, *p;
171
172 assert(data);
173
174 if (size < 2)
175 return false;
176
177 if (startswith(data, "__"))
178 return false;
179
180 b = data;
181 for (p = b; p < b + size; p++) {
182
183 if (*p == '=')
184 return p > b;
185
186 if (*p == '_')
187 continue;
188
189 if (*p >= 'A' && *p <= 'Z')
190 continue;
191
192 if (*p >= '0' && *p <= '9')
193 continue;
194
195 return false;
196 }
197
198 return false;
199 }
200
201 static bool same_field(const void *_a, size_t s, const void *_b, size_t t) {
202 const uint8_t *a = _a, *b = _b;
203 size_t j;
204
205 for (j = 0; j < s && j < t; j++) {
206
207 if (a[j] != b[j])
208 return false;
209
210 if (a[j] == '=')
211 return true;
212 }
213
214 assert_not_reached("\"=\" not found");
215 }
216
217 static Match *match_new(Match *p, MatchType t) {
218 Match *m;
219
220 m = new0(Match, 1);
221 if (!m)
222 return NULL;
223
224 m->type = t;
225
226 if (p) {
227 m->parent = p;
228 LIST_PREPEND(matches, p->matches, m);
229 }
230
231 return m;
232 }
233
234 static void match_free(Match *m) {
235 assert(m);
236
237 while (m->matches)
238 match_free(m->matches);
239
240 if (m->parent)
241 LIST_REMOVE(matches, m->parent->matches, m);
242
243 free(m->data);
244 free(m);
245 }
246
247 static void match_free_if_empty(Match *m) {
248 if (!m || m->matches)
249 return;
250
251 match_free(m);
252 }
253
254 _public_ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
255 Match *l3, *l4, *add_here = NULL, *m;
256 le64_t le_hash;
257
258 assert_return(j, -EINVAL);
259 assert_return(!journal_pid_changed(j), -ECHILD);
260 assert_return(data, -EINVAL);
261
262 if (size == 0)
263 size = strlen(data);
264
265 assert_return(match_is_valid(data, size), -EINVAL);
266
267 /* level 0: AND term
268 * level 1: OR terms
269 * level 2: AND terms
270 * level 3: OR terms
271 * level 4: concrete matches */
272
273 if (!j->level0) {
274 j->level0 = match_new(NULL, MATCH_AND_TERM);
275 if (!j->level0)
276 return -ENOMEM;
277 }
278
279 if (!j->level1) {
280 j->level1 = match_new(j->level0, MATCH_OR_TERM);
281 if (!j->level1)
282 return -ENOMEM;
283 }
284
285 if (!j->level2) {
286 j->level2 = match_new(j->level1, MATCH_AND_TERM);
287 if (!j->level2)
288 return -ENOMEM;
289 }
290
291 assert(j->level0->type == MATCH_AND_TERM);
292 assert(j->level1->type == MATCH_OR_TERM);
293 assert(j->level2->type == MATCH_AND_TERM);
294
295 le_hash = htole64(hash64(data, size));
296
297 LIST_FOREACH(matches, l3, j->level2->matches) {
298 assert(l3->type == MATCH_OR_TERM);
299
300 LIST_FOREACH(matches, l4, l3->matches) {
301 assert(l4->type == MATCH_DISCRETE);
302
303 /* Exactly the same match already? Then ignore
304 * this addition */
305 if (l4->le_hash == le_hash &&
306 l4->size == size &&
307 memcmp(l4->data, data, size) == 0)
308 return 0;
309
310 /* Same field? Then let's add this to this OR term */
311 if (same_field(data, size, l4->data, l4->size)) {
312 add_here = l3;
313 break;
314 }
315 }
316
317 if (add_here)
318 break;
319 }
320
321 if (!add_here) {
322 add_here = match_new(j->level2, MATCH_OR_TERM);
323 if (!add_here)
324 goto fail;
325 }
326
327 m = match_new(add_here, MATCH_DISCRETE);
328 if (!m)
329 goto fail;
330
331 m->le_hash = le_hash;
332 m->size = size;
333 m->data = memdup(data, size);
334 if (!m->data)
335 goto fail;
336
337 detach_location(j);
338
339 return 0;
340
341 fail:
342 match_free_if_empty(add_here);
343 match_free_if_empty(j->level2);
344 match_free_if_empty(j->level1);
345 match_free_if_empty(j->level0);
346
347 return -ENOMEM;
348 }
349
350 _public_ int sd_journal_add_conjunction(sd_journal *j) {
351 assert_return(j, -EINVAL);
352 assert_return(!journal_pid_changed(j), -ECHILD);
353
354 if (!j->level0)
355 return 0;
356
357 if (!j->level1)
358 return 0;
359
360 if (!j->level1->matches)
361 return 0;
362
363 j->level1 = NULL;
364 j->level2 = NULL;
365
366 return 0;
367 }
368
369 _public_ int sd_journal_add_disjunction(sd_journal *j) {
370 assert_return(j, -EINVAL);
371 assert_return(!journal_pid_changed(j), -ECHILD);
372
373 if (!j->level0)
374 return 0;
375
376 if (!j->level1)
377 return 0;
378
379 if (!j->level2)
380 return 0;
381
382 if (!j->level2->matches)
383 return 0;
384
385 j->level2 = NULL;
386 return 0;
387 }
388
389 static char *match_make_string(Match *m) {
390 char *p, *r;
391 Match *i;
392 bool enclose = false;
393
394 if (!m)
395 return strdup("none");
396
397 if (m->type == MATCH_DISCRETE)
398 return strndup(m->data, m->size);
399
400 p = NULL;
401 LIST_FOREACH(matches, i, m->matches) {
402 char *t, *k;
403
404 t = match_make_string(i);
405 if (!t) {
406 free(p);
407 return NULL;
408 }
409
410 if (p) {
411 k = strjoin(p, m->type == MATCH_OR_TERM ? " OR " : " AND ", t, NULL);
412 free(p);
413 free(t);
414
415 if (!k)
416 return NULL;
417
418 p = k;
419
420 enclose = true;
421 } else
422 p = t;
423 }
424
425 if (enclose) {
426 r = strjoin("(", p, ")", NULL);
427 free(p);
428 return r;
429 }
430
431 return p;
432 }
433
434 char *journal_make_match_string(sd_journal *j) {
435 assert(j);
436
437 return match_make_string(j->level0);
438 }
439
440 _public_ void sd_journal_flush_matches(sd_journal *j) {
441 if (!j)
442 return;
443
444 if (j->level0)
445 match_free(j->level0);
446
447 j->level0 = j->level1 = j->level2 = NULL;
448
449 detach_location(j);
450 }
451
452 _pure_ static int compare_with_location(JournalFile *f, Location *l) {
453 assert(f);
454 assert(l);
455 assert(f->location_type == LOCATION_SEEK);
456 assert(l->type == LOCATION_DISCRETE || l->type == LOCATION_SEEK);
457
458 if (l->monotonic_set &&
459 sd_id128_equal(f->current_boot_id, l->boot_id) &&
460 l->realtime_set &&
461 f->current_realtime == l->realtime &&
462 l->xor_hash_set &&
463 f->current_xor_hash == l->xor_hash)
464 return 0;
465
466 if (l->seqnum_set &&
467 sd_id128_equal(f->header->seqnum_id, l->seqnum_id)) {
468
469 if (f->current_seqnum < l->seqnum)
470 return -1;
471 if (f->current_seqnum > l->seqnum)
472 return 1;
473 }
474
475 if (l->monotonic_set &&
476 sd_id128_equal(f->current_boot_id, l->boot_id)) {
477
478 if (f->current_monotonic < l->monotonic)
479 return -1;
480 if (f->current_monotonic > l->monotonic)
481 return 1;
482 }
483
484 if (l->realtime_set) {
485
486 if (f->current_realtime < l->realtime)
487 return -1;
488 if (f->current_realtime > l->realtime)
489 return 1;
490 }
491
492 if (l->xor_hash_set) {
493
494 if (f->current_xor_hash < l->xor_hash)
495 return -1;
496 if (f->current_xor_hash > l->xor_hash)
497 return 1;
498 }
499
500 return 0;
501 }
502
503 static int next_for_match(
504 sd_journal *j,
505 Match *m,
506 JournalFile *f,
507 uint64_t after_offset,
508 direction_t direction,
509 Object **ret,
510 uint64_t *offset) {
511
512 int r;
513 uint64_t np = 0;
514 Object *n;
515
516 assert(j);
517 assert(m);
518 assert(f);
519
520 if (m->type == MATCH_DISCRETE) {
521 uint64_t dp;
522
523 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
524 if (r <= 0)
525 return r;
526
527 return journal_file_move_to_entry_by_offset_for_data(f, dp, after_offset, direction, ret, offset);
528
529 } else if (m->type == MATCH_OR_TERM) {
530 Match *i;
531
532 /* Find the earliest match beyond after_offset */
533
534 LIST_FOREACH(matches, i, m->matches) {
535 uint64_t cp;
536
537 r = next_for_match(j, i, f, after_offset, direction, NULL, &cp);
538 if (r < 0)
539 return r;
540 else if (r > 0) {
541 if (np == 0 || (direction == DIRECTION_DOWN ? cp < np : cp > np))
542 np = cp;
543 }
544 }
545
546 if (np == 0)
547 return 0;
548
549 } else if (m->type == MATCH_AND_TERM) {
550 Match *i, *last_moved;
551
552 /* Always jump to the next matching entry and repeat
553 * this until we find an offset that matches for all
554 * matches. */
555
556 if (!m->matches)
557 return 0;
558
559 r = next_for_match(j, m->matches, f, after_offset, direction, NULL, &np);
560 if (r <= 0)
561 return r;
562
563 assert(direction == DIRECTION_DOWN ? np >= after_offset : np <= after_offset);
564 last_moved = m->matches;
565
566 LIST_LOOP_BUT_ONE(matches, i, m->matches, last_moved) {
567 uint64_t cp;
568
569 r = next_for_match(j, i, f, np, direction, NULL, &cp);
570 if (r <= 0)
571 return r;
572
573 assert(direction == DIRECTION_DOWN ? cp >= np : cp <= np);
574 if (direction == DIRECTION_DOWN ? cp > np : cp < np) {
575 np = cp;
576 last_moved = i;
577 }
578 }
579 }
580
581 assert(np > 0);
582
583 r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
584 if (r < 0)
585 return r;
586
587 if (ret)
588 *ret = n;
589 if (offset)
590 *offset = np;
591
592 return 1;
593 }
594
595 static int find_location_for_match(
596 sd_journal *j,
597 Match *m,
598 JournalFile *f,
599 direction_t direction,
600 Object **ret,
601 uint64_t *offset) {
602
603 int r;
604
605 assert(j);
606 assert(m);
607 assert(f);
608
609 if (m->type == MATCH_DISCRETE) {
610 uint64_t dp;
611
612 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
613 if (r <= 0)
614 return r;
615
616 /* FIXME: missing: find by monotonic */
617
618 if (j->current_location.type == LOCATION_HEAD)
619 return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_DOWN, ret, offset);
620 if (j->current_location.type == LOCATION_TAIL)
621 return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_UP, ret, offset);
622 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
623 return journal_file_move_to_entry_by_seqnum_for_data(f, dp, j->current_location.seqnum, direction, ret, offset);
624 if (j->current_location.monotonic_set) {
625 r = journal_file_move_to_entry_by_monotonic_for_data(f, dp, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
626 if (r != -ENOENT)
627 return r;
628 }
629 if (j->current_location.realtime_set)
630 return journal_file_move_to_entry_by_realtime_for_data(f, dp, j->current_location.realtime, direction, ret, offset);
631
632 return journal_file_next_entry_for_data(f, NULL, 0, dp, direction, ret, offset);
633
634 } else if (m->type == MATCH_OR_TERM) {
635 uint64_t np = 0;
636 Object *n;
637 Match *i;
638
639 /* Find the earliest match */
640
641 LIST_FOREACH(matches, i, m->matches) {
642 uint64_t cp;
643
644 r = find_location_for_match(j, i, f, direction, NULL, &cp);
645 if (r < 0)
646 return r;
647 else if (r > 0) {
648 if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp))
649 np = cp;
650 }
651 }
652
653 if (np == 0)
654 return 0;
655
656 r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
657 if (r < 0)
658 return r;
659
660 if (ret)
661 *ret = n;
662 if (offset)
663 *offset = np;
664
665 return 1;
666
667 } else {
668 Match *i;
669 uint64_t np = 0;
670
671 assert(m->type == MATCH_AND_TERM);
672
673 /* First jump to the last match, and then find the
674 * next one where all matches match */
675
676 if (!m->matches)
677 return 0;
678
679 LIST_FOREACH(matches, i, m->matches) {
680 uint64_t cp;
681
682 r = find_location_for_match(j, i, f, direction, NULL, &cp);
683 if (r <= 0)
684 return r;
685
686 if (np == 0 || (direction == DIRECTION_DOWN ? cp > np : cp < np))
687 np = cp;
688 }
689
690 return next_for_match(j, m, f, np, direction, ret, offset);
691 }
692 }
693
694 static int find_location_with_matches(
695 sd_journal *j,
696 JournalFile *f,
697 direction_t direction,
698 Object **ret,
699 uint64_t *offset) {
700
701 int r;
702
703 assert(j);
704 assert(f);
705 assert(ret);
706 assert(offset);
707
708 if (!j->level0) {
709 /* No matches is simple */
710
711 if (j->current_location.type == LOCATION_HEAD)
712 return journal_file_next_entry(f, 0, DIRECTION_DOWN, ret, offset);
713 if (j->current_location.type == LOCATION_TAIL)
714 return journal_file_next_entry(f, 0, DIRECTION_UP, ret, offset);
715 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
716 return journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, ret, offset);
717 if (j->current_location.monotonic_set) {
718 r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
719 if (r != -ENOENT)
720 return r;
721 }
722 if (j->current_location.realtime_set)
723 return journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, ret, offset);
724
725 return journal_file_next_entry(f, 0, direction, ret, offset);
726 } else
727 return find_location_for_match(j, j->level0, f, direction, ret, offset);
728 }
729
730 static int next_with_matches(
731 sd_journal *j,
732 JournalFile *f,
733 direction_t direction,
734 Object **ret,
735 uint64_t *offset) {
736
737 assert(j);
738 assert(f);
739 assert(ret);
740 assert(offset);
741
742 /* No matches is easy. We simple advance the file
743 * pointer by one. */
744 if (!j->level0)
745 return journal_file_next_entry(f, f->current_offset, direction, ret, offset);
746
747 /* If we have a match then we look for the next matching entry
748 * with an offset at least one step larger */
749 return next_for_match(j, j->level0, f,
750 direction == DIRECTION_DOWN ? f->current_offset + 1
751 : f->current_offset - 1,
752 direction, ret, offset);
753 }
754
755 static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction) {
756 Object *c;
757 uint64_t cp, n_entries;
758 int r;
759
760 assert(j);
761 assert(f);
762
763 n_entries = le64toh(f->header->n_entries);
764
765 /* If we hit EOF before, we don't need to look into this file again
766 * unless direction changed or new entries appeared. */
767 if (f->last_direction == direction && f->location_type == LOCATION_TAIL &&
768 n_entries == f->last_n_entries)
769 return 0;
770
771 f->last_n_entries = n_entries;
772
773 if (f->last_direction == direction && f->current_offset > 0) {
774 /* LOCATION_SEEK here means we did the work in a previous
775 * iteration and the current location already points to a
776 * candidate entry. */
777 if (f->location_type != LOCATION_SEEK) {
778 r = next_with_matches(j, f, direction, &c, &cp);
779 if (r <= 0)
780 return r;
781
782 journal_file_save_location(f, c, cp);
783 }
784 } else {
785 f->last_direction = direction;
786
787 r = find_location_with_matches(j, f, direction, &c, &cp);
788 if (r <= 0)
789 return r;
790
791 journal_file_save_location(f, c, cp);
792 }
793
794 /* OK, we found the spot, now let's advance until an entry
795 * that is actually different from what we were previously
796 * looking at. This is necessary to handle entries which exist
797 * in two (or more) journal files, and which shall all be
798 * suppressed but one. */
799
800 for (;;) {
801 bool found;
802
803 if (j->current_location.type == LOCATION_DISCRETE) {
804 int k;
805
806 k = compare_with_location(f, &j->current_location);
807
808 found = direction == DIRECTION_DOWN ? k > 0 : k < 0;
809 } else
810 found = true;
811
812 if (found)
813 return 1;
814
815 r = next_with_matches(j, f, direction, &c, &cp);
816 if (r <= 0)
817 return r;
818
819 journal_file_save_location(f, c, cp);
820 }
821 }
822
823 static int real_journal_next(sd_journal *j, direction_t direction) {
824 JournalFile *f, *new_file = NULL;
825 Iterator i;
826 Object *o;
827 int r;
828
829 assert_return(j, -EINVAL);
830 assert_return(!journal_pid_changed(j), -ECHILD);
831
832 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
833 bool found;
834
835 r = next_beyond_location(j, f, direction);
836 if (r < 0) {
837 log_debug_errno(r, "Can't iterate through %s, ignoring: %m", f->path);
838 remove_file_real(j, f);
839 continue;
840 } else if (r == 0) {
841 f->location_type = LOCATION_TAIL;
842 continue;
843 }
844
845 if (!new_file)
846 found = true;
847 else {
848 int k;
849
850 k = journal_file_compare_locations(f, new_file);
851
852 found = direction == DIRECTION_DOWN ? k < 0 : k > 0;
853 }
854
855 if (found)
856 new_file = f;
857 }
858
859 if (!new_file)
860 return 0;
861
862 r = journal_file_move_to_object(new_file, OBJECT_ENTRY, new_file->current_offset, &o);
863 if (r < 0)
864 return r;
865
866 set_location(j, new_file, o);
867
868 return 1;
869 }
870
871 _public_ int sd_journal_next(sd_journal *j) {
872 return real_journal_next(j, DIRECTION_DOWN);
873 }
874
875 _public_ int sd_journal_previous(sd_journal *j) {
876 return real_journal_next(j, DIRECTION_UP);
877 }
878
879 static int real_journal_next_skip(sd_journal *j, direction_t direction, uint64_t skip) {
880 int c = 0, r;
881
882 assert_return(j, -EINVAL);
883 assert_return(!journal_pid_changed(j), -ECHILD);
884
885 if (skip == 0) {
886 /* If this is not a discrete skip, then at least
887 * resolve the current location */
888 if (j->current_location.type != LOCATION_DISCRETE)
889 return real_journal_next(j, direction);
890
891 return 0;
892 }
893
894 do {
895 r = real_journal_next(j, direction);
896 if (r < 0)
897 return r;
898
899 if (r == 0)
900 return c;
901
902 skip--;
903 c++;
904 } while (skip > 0);
905
906 return c;
907 }
908
909 _public_ int sd_journal_next_skip(sd_journal *j, uint64_t skip) {
910 return real_journal_next_skip(j, DIRECTION_DOWN, skip);
911 }
912
913 _public_ int sd_journal_previous_skip(sd_journal *j, uint64_t skip) {
914 return real_journal_next_skip(j, DIRECTION_UP, skip);
915 }
916
917 _public_ int sd_journal_get_cursor(sd_journal *j, char **cursor) {
918 Object *o;
919 int r;
920 char bid[33], sid[33];
921
922 assert_return(j, -EINVAL);
923 assert_return(!journal_pid_changed(j), -ECHILD);
924 assert_return(cursor, -EINVAL);
925
926 if (!j->current_file || j->current_file->current_offset <= 0)
927 return -EADDRNOTAVAIL;
928
929 r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
930 if (r < 0)
931 return r;
932
933 sd_id128_to_string(j->current_file->header->seqnum_id, sid);
934 sd_id128_to_string(o->entry.boot_id, bid);
935
936 if (asprintf(cursor,
937 "s=%s;i=%"PRIx64";b=%s;m=%"PRIx64";t=%"PRIx64";x=%"PRIx64,
938 sid, le64toh(o->entry.seqnum),
939 bid, le64toh(o->entry.monotonic),
940 le64toh(o->entry.realtime),
941 le64toh(o->entry.xor_hash)) < 0)
942 return -ENOMEM;
943
944 return 0;
945 }
946
947 _public_ int sd_journal_seek_cursor(sd_journal *j, const char *cursor) {
948 const char *word, *state;
949 size_t l;
950 unsigned long long seqnum, monotonic, realtime, xor_hash;
951 bool
952 seqnum_id_set = false,
953 seqnum_set = false,
954 boot_id_set = false,
955 monotonic_set = false,
956 realtime_set = false,
957 xor_hash_set = false;
958 sd_id128_t seqnum_id, boot_id;
959
960 assert_return(j, -EINVAL);
961 assert_return(!journal_pid_changed(j), -ECHILD);
962 assert_return(!isempty(cursor), -EINVAL);
963
964 FOREACH_WORD_SEPARATOR(word, l, cursor, ";", state) {
965 char *item;
966 int k = 0;
967
968 if (l < 2 || word[1] != '=')
969 return -EINVAL;
970
971 item = strndup(word, l);
972 if (!item)
973 return -ENOMEM;
974
975 switch (word[0]) {
976
977 case 's':
978 seqnum_id_set = true;
979 k = sd_id128_from_string(item+2, &seqnum_id);
980 break;
981
982 case 'i':
983 seqnum_set = true;
984 if (sscanf(item+2, "%llx", &seqnum) != 1)
985 k = -EINVAL;
986 break;
987
988 case 'b':
989 boot_id_set = true;
990 k = sd_id128_from_string(item+2, &boot_id);
991 break;
992
993 case 'm':
994 monotonic_set = true;
995 if (sscanf(item+2, "%llx", &monotonic) != 1)
996 k = -EINVAL;
997 break;
998
999 case 't':
1000 realtime_set = true;
1001 if (sscanf(item+2, "%llx", &realtime) != 1)
1002 k = -EINVAL;
1003 break;
1004
1005 case 'x':
1006 xor_hash_set = true;
1007 if (sscanf(item+2, "%llx", &xor_hash) != 1)
1008 k = -EINVAL;
1009 break;
1010 }
1011
1012 free(item);
1013
1014 if (k < 0)
1015 return k;
1016 }
1017
1018 if ((!seqnum_set || !seqnum_id_set) &&
1019 (!monotonic_set || !boot_id_set) &&
1020 !realtime_set)
1021 return -EINVAL;
1022
1023 reset_location(j);
1024
1025 j->current_location.type = LOCATION_SEEK;
1026
1027 if (realtime_set) {
1028 j->current_location.realtime = (uint64_t) realtime;
1029 j->current_location.realtime_set = true;
1030 }
1031
1032 if (seqnum_set && seqnum_id_set) {
1033 j->current_location.seqnum = (uint64_t) seqnum;
1034 j->current_location.seqnum_id = seqnum_id;
1035 j->current_location.seqnum_set = true;
1036 }
1037
1038 if (monotonic_set && boot_id_set) {
1039 j->current_location.monotonic = (uint64_t) monotonic;
1040 j->current_location.boot_id = boot_id;
1041 j->current_location.monotonic_set = true;
1042 }
1043
1044 if (xor_hash_set) {
1045 j->current_location.xor_hash = (uint64_t) xor_hash;
1046 j->current_location.xor_hash_set = true;
1047 }
1048
1049 return 0;
1050 }
1051
1052 _public_ int sd_journal_test_cursor(sd_journal *j, const char *cursor) {
1053 int r;
1054 Object *o;
1055
1056 assert_return(j, -EINVAL);
1057 assert_return(!journal_pid_changed(j), -ECHILD);
1058 assert_return(!isempty(cursor), -EINVAL);
1059
1060 if (!j->current_file || j->current_file->current_offset <= 0)
1061 return -EADDRNOTAVAIL;
1062
1063 r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
1064 if (r < 0)
1065 return r;
1066
1067 for (;;) {
1068 _cleanup_free_ char *item = NULL;
1069 unsigned long long ll;
1070 sd_id128_t id;
1071 int k = 0;
1072
1073 r = extract_first_word(&cursor, &item, ";", EXTRACT_DONT_COALESCE_SEPARATORS);
1074 if (r < 0)
1075 return r;
1076
1077 if (r == 0)
1078 break;
1079
1080 if (strlen(item) < 2 || item[1] != '=')
1081 return -EINVAL;
1082
1083 switch (item[0]) {
1084
1085 case 's':
1086 k = sd_id128_from_string(item+2, &id);
1087 if (k < 0)
1088 return k;
1089 if (!sd_id128_equal(id, j->current_file->header->seqnum_id))
1090 return 0;
1091 break;
1092
1093 case 'i':
1094 if (sscanf(item+2, "%llx", &ll) != 1)
1095 return -EINVAL;
1096 if (ll != le64toh(o->entry.seqnum))
1097 return 0;
1098 break;
1099
1100 case 'b':
1101 k = sd_id128_from_string(item+2, &id);
1102 if (k < 0)
1103 return k;
1104 if (!sd_id128_equal(id, o->entry.boot_id))
1105 return 0;
1106 break;
1107
1108 case 'm':
1109 if (sscanf(item+2, "%llx", &ll) != 1)
1110 return -EINVAL;
1111 if (ll != le64toh(o->entry.monotonic))
1112 return 0;
1113 break;
1114
1115 case 't':
1116 if (sscanf(item+2, "%llx", &ll) != 1)
1117 return -EINVAL;
1118 if (ll != le64toh(o->entry.realtime))
1119 return 0;
1120 break;
1121
1122 case 'x':
1123 if (sscanf(item+2, "%llx", &ll) != 1)
1124 return -EINVAL;
1125 if (ll != le64toh(o->entry.xor_hash))
1126 return 0;
1127 break;
1128 }
1129 }
1130
1131 return 1;
1132 }
1133
1134
1135 _public_ int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) {
1136 assert_return(j, -EINVAL);
1137 assert_return(!journal_pid_changed(j), -ECHILD);
1138
1139 reset_location(j);
1140 j->current_location.type = LOCATION_SEEK;
1141 j->current_location.boot_id = boot_id;
1142 j->current_location.monotonic = usec;
1143 j->current_location.monotonic_set = true;
1144
1145 return 0;
1146 }
1147
1148 _public_ int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) {
1149 assert_return(j, -EINVAL);
1150 assert_return(!journal_pid_changed(j), -ECHILD);
1151
1152 reset_location(j);
1153 j->current_location.type = LOCATION_SEEK;
1154 j->current_location.realtime = usec;
1155 j->current_location.realtime_set = true;
1156
1157 return 0;
1158 }
1159
1160 _public_ int sd_journal_seek_head(sd_journal *j) {
1161 assert_return(j, -EINVAL);
1162 assert_return(!journal_pid_changed(j), -ECHILD);
1163
1164 reset_location(j);
1165 j->current_location.type = LOCATION_HEAD;
1166
1167 return 0;
1168 }
1169
1170 _public_ int sd_journal_seek_tail(sd_journal *j) {
1171 assert_return(j, -EINVAL);
1172 assert_return(!journal_pid_changed(j), -ECHILD);
1173
1174 reset_location(j);
1175 j->current_location.type = LOCATION_TAIL;
1176
1177 return 0;
1178 }
1179
1180 static void check_network(sd_journal *j, int fd) {
1181 struct statfs sfs;
1182
1183 assert(j);
1184
1185 if (j->on_network)
1186 return;
1187
1188 if (fstatfs(fd, &sfs) < 0)
1189 return;
1190
1191 j->on_network =
1192 F_TYPE_EQUAL(sfs.f_type, CIFS_MAGIC_NUMBER) ||
1193 F_TYPE_EQUAL(sfs.f_type, CODA_SUPER_MAGIC) ||
1194 F_TYPE_EQUAL(sfs.f_type, NCP_SUPER_MAGIC) ||
1195 F_TYPE_EQUAL(sfs.f_type, NFS_SUPER_MAGIC) ||
1196 F_TYPE_EQUAL(sfs.f_type, SMB_SUPER_MAGIC);
1197 }
1198
1199 static bool file_has_type_prefix(const char *prefix, const char *filename) {
1200 const char *full, *tilded, *atted;
1201
1202 full = strjoina(prefix, ".journal");
1203 tilded = strjoina(full, "~");
1204 atted = strjoina(prefix, "@");
1205
1206 return streq(filename, full) ||
1207 streq(filename, tilded) ||
1208 startswith(filename, atted);
1209 }
1210
1211 static bool file_type_wanted(int flags, const char *filename) {
1212 assert(filename);
1213
1214 if (!endswith(filename, ".journal") && !endswith(filename, ".journal~"))
1215 return false;
1216
1217 /* no flags set → every type is OK */
1218 if (!(flags & (SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)))
1219 return true;
1220
1221 if (flags & SD_JOURNAL_SYSTEM && file_has_type_prefix("system", filename))
1222 return true;
1223
1224 if (flags & SD_JOURNAL_CURRENT_USER) {
1225 char prefix[5 + DECIMAL_STR_MAX(uid_t) + 1];
1226
1227 xsprintf(prefix, "user-"UID_FMT, getuid());
1228
1229 if (file_has_type_prefix(prefix, filename))
1230 return true;
1231 }
1232
1233 return false;
1234 }
1235
1236 static bool path_has_prefix(sd_journal *j, const char *path, const char *prefix) {
1237 assert(j);
1238 assert(path);
1239 assert(prefix);
1240
1241 if (j->toplevel_fd >= 0)
1242 return false;
1243
1244 return path_startswith(path, prefix);
1245 }
1246
1247 static const char *skip_slash(const char *p) {
1248
1249 if (!p)
1250 return NULL;
1251
1252 while (*p == '/')
1253 p++;
1254
1255 return p;
1256 }
1257
1258 static int add_any_file(sd_journal *j, int fd, const char *path) {
1259 JournalFile *f = NULL;
1260 bool close_fd = false;
1261 int r, k;
1262
1263 assert(j);
1264 assert(fd >= 0 || path);
1265
1266 if (path && ordered_hashmap_get(j->files, path))
1267 return 0;
1268
1269 if (ordered_hashmap_size(j->files) >= JOURNAL_FILES_MAX) {
1270 log_debug("Too many open journal files, not adding %s.", path);
1271 r = -ETOOMANYREFS;
1272 goto fail;
1273 }
1274
1275 if (fd < 0 && j->toplevel_fd >= 0) {
1276
1277 /* If there's a top-level fd defined, open the file relative to this now. (Make the path relative,
1278 * explicitly, since otherwise openat() ignores the first argument.) */
1279
1280 fd = openat(j->toplevel_fd, skip_slash(path), O_RDONLY|O_CLOEXEC);
1281 if (fd < 0) {
1282 r = log_debug_errno(errno, "Failed to open journal file %s: %m", path);
1283 goto fail;
1284 }
1285
1286 close_fd = true;
1287 }
1288
1289 r = journal_file_open(fd, path, O_RDONLY, 0, false, false, NULL, j->mmap, NULL, NULL, &f);
1290 if (r < 0) {
1291 if (close_fd)
1292 safe_close(fd);
1293 log_debug_errno(r, "Failed to open journal file %s: %m", path);
1294 goto fail;
1295 }
1296
1297 /* journal_file_dump(f); */
1298
1299 r = ordered_hashmap_put(j->files, f->path, f);
1300 if (r < 0) {
1301 f->close_fd = close_fd;
1302 (void) journal_file_close(f);
1303 goto fail;
1304 }
1305
1306 if (!j->has_runtime_files && path_has_prefix(j, f->path, "/run"))
1307 j->has_runtime_files = true;
1308 else if (!j->has_persistent_files && path_has_prefix(j, f->path, "/var"))
1309 j->has_persistent_files = true;
1310
1311 log_debug("File %s added.", f->path);
1312
1313 check_network(j, f->fd);
1314
1315 j->current_invalidate_counter++;
1316
1317 return 0;
1318
1319 fail:
1320 k = journal_put_error(j, r, path);
1321 if (k < 0)
1322 return k;
1323
1324 return r;
1325 }
1326
1327 static int add_file(sd_journal *j, const char *prefix, const char *filename) {
1328 const char *path;
1329
1330 assert(j);
1331 assert(prefix);
1332 assert(filename);
1333
1334 if (j->no_new_files)
1335 return 0;
1336
1337 if (!file_type_wanted(j->flags, filename))
1338 return 0;
1339
1340 path = strjoina(prefix, "/", filename);
1341 return add_any_file(j, -1, path);
1342 }
1343
1344 static void remove_file(sd_journal *j, const char *prefix, const char *filename) {
1345 const char *path;
1346 JournalFile *f;
1347
1348 assert(j);
1349 assert(prefix);
1350 assert(filename);
1351
1352 path = strjoina(prefix, "/", filename);
1353 f = ordered_hashmap_get(j->files, path);
1354 if (!f)
1355 return;
1356
1357 remove_file_real(j, f);
1358 }
1359
1360 static void remove_file_real(sd_journal *j, JournalFile *f) {
1361 assert(j);
1362 assert(f);
1363
1364 ordered_hashmap_remove(j->files, f->path);
1365
1366 log_debug("File %s removed.", f->path);
1367
1368 if (j->current_file == f) {
1369 j->current_file = NULL;
1370 j->current_field = 0;
1371 }
1372
1373 if (j->unique_file == f) {
1374 /* Jump to the next unique_file or NULL if that one was last */
1375 j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path);
1376 j->unique_offset = 0;
1377 if (!j->unique_file)
1378 j->unique_file_lost = true;
1379 }
1380
1381 if (j->fields_file == f) {
1382 j->fields_file = ordered_hashmap_next(j->files, j->fields_file->path);
1383 j->fields_offset = 0;
1384 if (!j->fields_file)
1385 j->fields_file_lost = true;
1386 }
1387
1388 (void) journal_file_close(f);
1389
1390 j->current_invalidate_counter++;
1391 }
1392
1393 static int dirname_is_machine_id(const char *fn) {
1394 sd_id128_t id, machine;
1395 int r;
1396
1397 r = sd_id128_get_machine(&machine);
1398 if (r < 0)
1399 return r;
1400
1401 r = sd_id128_from_string(fn, &id);
1402 if (r < 0)
1403 return r;
1404
1405 return sd_id128_equal(id, machine);
1406 }
1407
1408 static int add_directory(sd_journal *j, const char *prefix, const char *dirname) {
1409 _cleanup_free_ char *path = NULL;
1410 _cleanup_closedir_ DIR *d = NULL;
1411 struct dirent *de = NULL;
1412 Directory *m;
1413 int r, k;
1414
1415 assert(j);
1416 assert(prefix);
1417
1418 /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch
1419 * and reenumerates directory contents */
1420
1421 if (dirname)
1422 path = strjoin(prefix, "/", dirname, NULL);
1423 else
1424 path = strdup(prefix);
1425 if (!path) {
1426 r = -ENOMEM;
1427 goto fail;
1428 }
1429
1430 log_debug("Considering directory %s.", path);
1431
1432 /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */
1433 if ((j->flags & SD_JOURNAL_LOCAL_ONLY) &&
1434 !((dirname && dirname_is_machine_id(dirname) > 0) || path_has_prefix(j, path, "/run")))
1435 return 0;
1436
1437
1438 if (j->toplevel_fd < 0)
1439 d = opendir(path);
1440 else
1441 /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is
1442 * relative, by dropping the initial slash */
1443 d = xopendirat(j->toplevel_fd, skip_slash(path), 0);
1444 if (!d) {
1445 r = log_debug_errno(errno, "Failed to open directory %s: %m", path);
1446 goto fail;
1447 }
1448
1449 m = hashmap_get(j->directories_by_path, path);
1450 if (!m) {
1451 m = new0(Directory, 1);
1452 if (!m) {
1453 r = -ENOMEM;
1454 goto fail;
1455 }
1456
1457 m->is_root = false;
1458 m->path = path;
1459
1460 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1461 free(m);
1462 r = -ENOMEM;
1463 goto fail;
1464 }
1465
1466 path = NULL; /* avoid freeing in cleanup */
1467 j->current_invalidate_counter++;
1468
1469 log_debug("Directory %s added.", m->path);
1470
1471 } else if (m->is_root)
1472 return 0;
1473
1474 if (m->wd <= 0 && j->inotify_fd >= 0) {
1475 /* Watch this directory, if it not being watched yet. */
1476
1477 m->wd = inotify_add_watch_fd(j->inotify_fd, dirfd(d),
1478 IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1479 IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT|IN_MOVED_FROM|
1480 IN_ONLYDIR);
1481
1482 if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0)
1483 inotify_rm_watch(j->inotify_fd, m->wd);
1484 }
1485
1486 FOREACH_DIRENT_ALL(de, d, r = log_debug_errno(errno, "Failed to read directory %s: %m", m->path); goto fail) {
1487
1488 if (dirent_is_file_with_suffix(de, ".journal") ||
1489 dirent_is_file_with_suffix(de, ".journal~"))
1490 (void) add_file(j, m->path, de->d_name);
1491 }
1492
1493 check_network(j, dirfd(d));
1494
1495 return 0;
1496
1497 fail:
1498 k = journal_put_error(j, r, path ?: prefix);
1499 if (k < 0)
1500 return k;
1501
1502 return r;
1503 }
1504
1505 static int add_root_directory(sd_journal *j, const char *p, bool missing_ok) {
1506
1507 _cleanup_closedir_ DIR *d = NULL;
1508 struct dirent *de;
1509 Directory *m;
1510 int r, k;
1511
1512 assert(j);
1513
1514 /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we
1515 * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially
1516 * populate the set, as well as to update it later. */
1517
1518 if (p) {
1519 /* If there's a path specified, use it. */
1520
1521 if ((j->flags & SD_JOURNAL_RUNTIME_ONLY) &&
1522 !path_has_prefix(j, p, "/run"))
1523 return -EINVAL;
1524
1525 if (j->prefix)
1526 p = strjoina(j->prefix, p);
1527
1528 if (j->toplevel_fd < 0)
1529 d = opendir(p);
1530 else
1531 d = xopendirat(j->toplevel_fd, skip_slash(p), 0);
1532
1533 if (!d) {
1534 if (errno == ENOENT && missing_ok)
1535 return 0;
1536
1537 r = log_debug_errno(errno, "Failed to open root directory %s: %m", p);
1538 goto fail;
1539 }
1540 } else {
1541 int dfd;
1542
1543 /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since
1544 * opendir() will take possession of the fd, and close it, which we don't want. */
1545
1546 p = "."; /* store this as "." in the directories hashmap */
1547
1548 dfd = fcntl(j->toplevel_fd, F_DUPFD_CLOEXEC, 3);
1549 if (dfd < 0) {
1550 r = -errno;
1551 goto fail;
1552 }
1553
1554 d = fdopendir(dfd);
1555 if (!d) {
1556 r = -errno;
1557 safe_close(dfd);
1558 goto fail;
1559 }
1560
1561 rewinddir(d);
1562 }
1563
1564 m = hashmap_get(j->directories_by_path, p);
1565 if (!m) {
1566 m = new0(Directory, 1);
1567 if (!m) {
1568 r = -ENOMEM;
1569 goto fail;
1570 }
1571
1572 m->is_root = true;
1573
1574 m->path = strdup(p);
1575 if (!m->path) {
1576 free(m);
1577 r = -ENOMEM;
1578 goto fail;
1579 }
1580
1581 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1582 free(m->path);
1583 free(m);
1584 r = -ENOMEM;
1585 goto fail;
1586 }
1587
1588 j->current_invalidate_counter++;
1589
1590 log_debug("Root directory %s added.", m->path);
1591
1592 } else if (!m->is_root)
1593 return 0;
1594
1595 if (m->wd <= 0 && j->inotify_fd >= 0) {
1596
1597 m->wd = inotify_add_watch_fd(j->inotify_fd, dirfd(d),
1598 IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1599 IN_ONLYDIR);
1600
1601 if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0)
1602 inotify_rm_watch(j->inotify_fd, m->wd);
1603 }
1604
1605 if (j->no_new_files)
1606 return 0;
1607
1608 FOREACH_DIRENT_ALL(de, d, r = log_debug_errno(errno, "Failed to read directory %s: %m", m->path); goto fail) {
1609 sd_id128_t id;
1610
1611 if (dirent_is_file_with_suffix(de, ".journal") ||
1612 dirent_is_file_with_suffix(de, ".journal~"))
1613 (void) add_file(j, m->path, de->d_name);
1614 else if (IN_SET(de->d_type, DT_DIR, DT_LNK, DT_UNKNOWN) &&
1615 sd_id128_from_string(de->d_name, &id) >= 0)
1616 (void) add_directory(j, m->path, de->d_name);
1617 }
1618
1619 check_network(j, dirfd(d));
1620
1621 return 0;
1622
1623 fail:
1624 k = journal_put_error(j, r, p);
1625 if (k < 0)
1626 return k;
1627
1628 return r;
1629 }
1630
1631 static void remove_directory(sd_journal *j, Directory *d) {
1632 assert(j);
1633
1634 if (d->wd > 0) {
1635 hashmap_remove(j->directories_by_wd, INT_TO_PTR(d->wd));
1636
1637 if (j->inotify_fd >= 0)
1638 inotify_rm_watch(j->inotify_fd, d->wd);
1639 }
1640
1641 hashmap_remove(j->directories_by_path, d->path);
1642
1643 if (d->is_root)
1644 log_debug("Root directory %s removed.", d->path);
1645 else
1646 log_debug("Directory %s removed.", d->path);
1647
1648 free(d->path);
1649 free(d);
1650 }
1651
1652 static int add_search_paths(sd_journal *j) {
1653
1654 static const char search_paths[] =
1655 "/run/log/journal\0"
1656 "/var/log/journal\0";
1657 const char *p;
1658
1659 assert(j);
1660
1661 /* We ignore most errors here, since the idea is to only open
1662 * what's actually accessible, and ignore the rest. */
1663
1664 NULSTR_FOREACH(p, search_paths)
1665 (void) add_root_directory(j, p, true);
1666
1667 return 0;
1668 }
1669
1670 static int add_current_paths(sd_journal *j) {
1671 Iterator i;
1672 JournalFile *f;
1673
1674 assert(j);
1675 assert(j->no_new_files);
1676
1677 /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we
1678 * treat them as fatal. */
1679
1680 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
1681 _cleanup_free_ char *dir;
1682 int r;
1683
1684 dir = dirname_malloc(f->path);
1685 if (!dir)
1686 return -ENOMEM;
1687
1688 r = add_directory(j, dir, NULL);
1689 if (r < 0)
1690 return r;
1691 }
1692
1693 return 0;
1694 }
1695
1696 static int allocate_inotify(sd_journal *j) {
1697 assert(j);
1698
1699 if (j->inotify_fd < 0) {
1700 j->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC);
1701 if (j->inotify_fd < 0)
1702 return -errno;
1703 }
1704
1705 return hashmap_ensure_allocated(&j->directories_by_wd, NULL);
1706 }
1707
1708 static sd_journal *journal_new(int flags, const char *path) {
1709 sd_journal *j;
1710
1711 j = new0(sd_journal, 1);
1712 if (!j)
1713 return NULL;
1714
1715 j->original_pid = getpid();
1716 j->toplevel_fd = -1;
1717 j->inotify_fd = -1;
1718 j->flags = flags;
1719 j->data_threshold = DEFAULT_DATA_THRESHOLD;
1720
1721 if (path) {
1722 char *t;
1723
1724 t = strdup(path);
1725 if (!t)
1726 goto fail;
1727
1728 if (flags & SD_JOURNAL_OS_ROOT)
1729 j->prefix = t;
1730 else
1731 j->path = t;
1732 }
1733
1734 j->files = ordered_hashmap_new(&string_hash_ops);
1735 j->directories_by_path = hashmap_new(&string_hash_ops);
1736 j->mmap = mmap_cache_new();
1737 if (!j->files || !j->directories_by_path || !j->mmap)
1738 goto fail;
1739
1740 return j;
1741
1742 fail:
1743 sd_journal_close(j);
1744 return NULL;
1745 }
1746
1747 #define OPEN_ALLOWED_FLAGS \
1748 (SD_JOURNAL_LOCAL_ONLY | \
1749 SD_JOURNAL_RUNTIME_ONLY | \
1750 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)
1751
1752 _public_ int sd_journal_open(sd_journal **ret, int flags) {
1753 sd_journal *j;
1754 int r;
1755
1756 assert_return(ret, -EINVAL);
1757 assert_return((flags & ~OPEN_ALLOWED_FLAGS) == 0, -EINVAL);
1758
1759 j = journal_new(flags, NULL);
1760 if (!j)
1761 return -ENOMEM;
1762
1763 r = add_search_paths(j);
1764 if (r < 0)
1765 goto fail;
1766
1767 *ret = j;
1768 return 0;
1769
1770 fail:
1771 sd_journal_close(j);
1772
1773 return r;
1774 }
1775
1776 #define OPEN_CONTAINER_ALLOWED_FLAGS \
1777 (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM)
1778
1779 _public_ int sd_journal_open_container(sd_journal **ret, const char *machine, int flags) {
1780 _cleanup_free_ char *root = NULL, *class = NULL;
1781 sd_journal *j;
1782 char *p;
1783 int r;
1784
1785 /* This is pretty much deprecated, people should use machined's OpenMachineRootDirectory() call instead in
1786 * combination with sd_journal_open_directory_fd(). */
1787
1788 assert_return(machine, -EINVAL);
1789 assert_return(ret, -EINVAL);
1790 assert_return((flags & ~OPEN_CONTAINER_ALLOWED_FLAGS) == 0, -EINVAL);
1791 assert_return(machine_name_is_valid(machine), -EINVAL);
1792
1793 p = strjoina("/run/systemd/machines/", machine);
1794 r = parse_env_file(p, NEWLINE, "ROOT", &root, "CLASS", &class, NULL);
1795 if (r == -ENOENT)
1796 return -EHOSTDOWN;
1797 if (r < 0)
1798 return r;
1799 if (!root)
1800 return -ENODATA;
1801
1802 if (!streq_ptr(class, "container"))
1803 return -EIO;
1804
1805 j = journal_new(flags, root);
1806 if (!j)
1807 return -ENOMEM;
1808
1809 r = add_search_paths(j);
1810 if (r < 0)
1811 goto fail;
1812
1813 *ret = j;
1814 return 0;
1815
1816 fail:
1817 sd_journal_close(j);
1818 return r;
1819 }
1820
1821 #define OPEN_DIRECTORY_ALLOWED_FLAGS \
1822 (SD_JOURNAL_OS_ROOT | \
1823 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1824
1825 _public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) {
1826 sd_journal *j;
1827 int r;
1828
1829 assert_return(ret, -EINVAL);
1830 assert_return(path, -EINVAL);
1831 assert_return((flags & ~OPEN_DIRECTORY_ALLOWED_FLAGS) == 0, -EINVAL);
1832
1833 j = journal_new(flags, path);
1834 if (!j)
1835 return -ENOMEM;
1836
1837 if (flags & SD_JOURNAL_OS_ROOT)
1838 r = add_search_paths(j);
1839 else
1840 r = add_root_directory(j, path, false);
1841 if (r < 0)
1842 goto fail;
1843
1844 *ret = j;
1845 return 0;
1846
1847 fail:
1848 sd_journal_close(j);
1849 return r;
1850 }
1851
1852 _public_ int sd_journal_open_files(sd_journal **ret, const char **paths, int flags) {
1853 sd_journal *j;
1854 const char **path;
1855 int r;
1856
1857 assert_return(ret, -EINVAL);
1858 assert_return(flags == 0, -EINVAL);
1859
1860 j = journal_new(flags, NULL);
1861 if (!j)
1862 return -ENOMEM;
1863
1864 STRV_FOREACH(path, paths) {
1865 r = add_any_file(j, -1, *path);
1866 if (r < 0)
1867 goto fail;
1868 }
1869
1870 j->no_new_files = true;
1871
1872 *ret = j;
1873 return 0;
1874
1875 fail:
1876 sd_journal_close(j);
1877 return r;
1878 }
1879
1880 #define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \
1881 (SD_JOURNAL_OS_ROOT | \
1882 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1883
1884 _public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) {
1885 sd_journal *j;
1886 struct stat st;
1887 int r;
1888
1889 assert_return(ret, -EINVAL);
1890 assert_return(fd >= 0, -EBADF);
1891 assert_return((flags & ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS) == 0, -EINVAL);
1892
1893 if (fstat(fd, &st) < 0)
1894 return -errno;
1895
1896 if (!S_ISDIR(st.st_mode))
1897 return -EBADFD;
1898
1899 j = journal_new(flags, NULL);
1900 if (!j)
1901 return -ENOMEM;
1902
1903 j->toplevel_fd = fd;
1904
1905 if (flags & SD_JOURNAL_OS_ROOT)
1906 r = add_search_paths(j);
1907 else
1908 r = add_root_directory(j, NULL, false);
1909 if (r < 0)
1910 goto fail;
1911
1912 *ret = j;
1913 return 0;
1914
1915 fail:
1916 sd_journal_close(j);
1917 return r;
1918 }
1919
1920 _public_ int sd_journal_open_files_fd(sd_journal **ret, int fds[], unsigned n_fds, int flags) {
1921 Iterator iterator;
1922 JournalFile *f;
1923 sd_journal *j;
1924 unsigned i;
1925 int r;
1926
1927 assert_return(ret, -EINVAL);
1928 assert_return(n_fds > 0, -EBADF);
1929 assert_return(flags == 0, -EINVAL);
1930
1931 j = journal_new(flags, NULL);
1932 if (!j)
1933 return -ENOMEM;
1934
1935 for (i = 0; i < n_fds; i++) {
1936 struct stat st;
1937
1938 if (fds[i] < 0) {
1939 r = -EBADF;
1940 goto fail;
1941 }
1942
1943 if (fstat(fds[i], &st) < 0) {
1944 r = -errno;
1945 goto fail;
1946 }
1947
1948 if (!S_ISREG(st.st_mode)) {
1949 r = -EBADFD;
1950 goto fail;
1951 }
1952
1953 r = add_any_file(j, fds[i], NULL);
1954 if (r < 0)
1955 goto fail;
1956 }
1957
1958 j->no_new_files = true;
1959 j->no_inotify = true;
1960
1961 *ret = j;
1962 return 0;
1963
1964 fail:
1965 /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they
1966 * remain open */
1967 ORDERED_HASHMAP_FOREACH(f, j->files, iterator)
1968 f->close_fd = false;
1969
1970 sd_journal_close(j);
1971 return r;
1972 }
1973
1974 _public_ void sd_journal_close(sd_journal *j) {
1975 Directory *d;
1976 JournalFile *f;
1977 char *p;
1978
1979 if (!j)
1980 return;
1981
1982 sd_journal_flush_matches(j);
1983
1984 while ((f = ordered_hashmap_steal_first(j->files)))
1985 (void) journal_file_close(f);
1986
1987 ordered_hashmap_free(j->files);
1988
1989 while ((d = hashmap_first(j->directories_by_path)))
1990 remove_directory(j, d);
1991
1992 while ((d = hashmap_first(j->directories_by_wd)))
1993 remove_directory(j, d);
1994
1995 hashmap_free(j->directories_by_path);
1996 hashmap_free(j->directories_by_wd);
1997
1998 safe_close(j->inotify_fd);
1999
2000 if (j->mmap) {
2001 log_debug("mmap cache statistics: %u hit, %u miss", mmap_cache_get_hit(j->mmap), mmap_cache_get_missed(j->mmap));
2002 mmap_cache_unref(j->mmap);
2003 }
2004
2005 while ((p = hashmap_steal_first(j->errors)))
2006 free(p);
2007 hashmap_free(j->errors);
2008
2009 free(j->path);
2010 free(j->prefix);
2011 free(j->unique_field);
2012 free(j->fields_buffer);
2013 free(j);
2014 }
2015
2016 _public_ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) {
2017 Object *o;
2018 JournalFile *f;
2019 int r;
2020
2021 assert_return(j, -EINVAL);
2022 assert_return(!journal_pid_changed(j), -ECHILD);
2023 assert_return(ret, -EINVAL);
2024
2025 f = j->current_file;
2026 if (!f)
2027 return -EADDRNOTAVAIL;
2028
2029 if (f->current_offset <= 0)
2030 return -EADDRNOTAVAIL;
2031
2032 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2033 if (r < 0)
2034 return r;
2035
2036 *ret = le64toh(o->entry.realtime);
2037 return 0;
2038 }
2039
2040 _public_ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) {
2041 Object *o;
2042 JournalFile *f;
2043 int r;
2044 sd_id128_t id;
2045
2046 assert_return(j, -EINVAL);
2047 assert_return(!journal_pid_changed(j), -ECHILD);
2048
2049 f = j->current_file;
2050 if (!f)
2051 return -EADDRNOTAVAIL;
2052
2053 if (f->current_offset <= 0)
2054 return -EADDRNOTAVAIL;
2055
2056 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2057 if (r < 0)
2058 return r;
2059
2060 if (ret_boot_id)
2061 *ret_boot_id = o->entry.boot_id;
2062 else {
2063 r = sd_id128_get_boot(&id);
2064 if (r < 0)
2065 return r;
2066
2067 if (!sd_id128_equal(id, o->entry.boot_id))
2068 return -ESTALE;
2069 }
2070
2071 if (ret)
2072 *ret = le64toh(o->entry.monotonic);
2073
2074 return 0;
2075 }
2076
2077 static bool field_is_valid(const char *field) {
2078 const char *p;
2079
2080 assert(field);
2081
2082 if (isempty(field))
2083 return false;
2084
2085 if (startswith(field, "__"))
2086 return false;
2087
2088 for (p = field; *p; p++) {
2089
2090 if (*p == '_')
2091 continue;
2092
2093 if (*p >= 'A' && *p <= 'Z')
2094 continue;
2095
2096 if (*p >= '0' && *p <= '9')
2097 continue;
2098
2099 return false;
2100 }
2101
2102 return true;
2103 }
2104
2105 _public_ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) {
2106 JournalFile *f;
2107 uint64_t i, n;
2108 size_t field_length;
2109 int r;
2110 Object *o;
2111
2112 assert_return(j, -EINVAL);
2113 assert_return(!journal_pid_changed(j), -ECHILD);
2114 assert_return(field, -EINVAL);
2115 assert_return(data, -EINVAL);
2116 assert_return(size, -EINVAL);
2117 assert_return(field_is_valid(field), -EINVAL);
2118
2119 f = j->current_file;
2120 if (!f)
2121 return -EADDRNOTAVAIL;
2122
2123 if (f->current_offset <= 0)
2124 return -EADDRNOTAVAIL;
2125
2126 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2127 if (r < 0)
2128 return r;
2129
2130 field_length = strlen(field);
2131
2132 n = journal_file_entry_n_items(o);
2133 for (i = 0; i < n; i++) {
2134 uint64_t p, l;
2135 le64_t le_hash;
2136 size_t t;
2137 int compression;
2138
2139 p = le64toh(o->entry.items[i].object_offset);
2140 le_hash = o->entry.items[i].hash;
2141 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
2142 if (r < 0)
2143 return r;
2144
2145 if (le_hash != o->data.hash)
2146 return -EBADMSG;
2147
2148 l = le64toh(o->object.size) - offsetof(Object, data.payload);
2149
2150 compression = o->object.flags & OBJECT_COMPRESSION_MASK;
2151 if (compression) {
2152 #if defined(HAVE_XZ) || defined(HAVE_LZ4)
2153 r = decompress_startswith(compression,
2154 o->data.payload, l,
2155 &f->compress_buffer, &f->compress_buffer_size,
2156 field, field_length, '=');
2157 if (r < 0)
2158 log_debug_errno(r, "Cannot decompress %s object of length %"PRIu64" at offset "OFSfmt": %m",
2159 object_compressed_to_string(compression), l, p);
2160 else if (r > 0) {
2161
2162 size_t rsize;
2163
2164 r = decompress_blob(compression,
2165 o->data.payload, l,
2166 &f->compress_buffer, &f->compress_buffer_size, &rsize,
2167 j->data_threshold);
2168 if (r < 0)
2169 return r;
2170
2171 *data = f->compress_buffer;
2172 *size = (size_t) rsize;
2173
2174 return 0;
2175 }
2176 #else
2177 return -EPROTONOSUPPORT;
2178 #endif
2179 } else if (l >= field_length+1 &&
2180 memcmp(o->data.payload, field, field_length) == 0 &&
2181 o->data.payload[field_length] == '=') {
2182
2183 t = (size_t) l;
2184
2185 if ((uint64_t) t != l)
2186 return -E2BIG;
2187
2188 *data = o->data.payload;
2189 *size = t;
2190
2191 return 0;
2192 }
2193
2194 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2195 if (r < 0)
2196 return r;
2197 }
2198
2199 return -ENOENT;
2200 }
2201
2202 static int return_data(sd_journal *j, JournalFile *f, Object *o, const void **data, size_t *size) {
2203 size_t t;
2204 uint64_t l;
2205 int compression;
2206
2207 l = le64toh(o->object.size) - offsetof(Object, data.payload);
2208 t = (size_t) l;
2209
2210 /* We can't read objects larger than 4G on a 32bit machine */
2211 if ((uint64_t) t != l)
2212 return -E2BIG;
2213
2214 compression = o->object.flags & OBJECT_COMPRESSION_MASK;
2215 if (compression) {
2216 #if defined(HAVE_XZ) || defined(HAVE_LZ4)
2217 size_t rsize;
2218 int r;
2219
2220 r = decompress_blob(compression,
2221 o->data.payload, l, &f->compress_buffer,
2222 &f->compress_buffer_size, &rsize, j->data_threshold);
2223 if (r < 0)
2224 return r;
2225
2226 *data = f->compress_buffer;
2227 *size = (size_t) rsize;
2228 #else
2229 return -EPROTONOSUPPORT;
2230 #endif
2231 } else {
2232 *data = o->data.payload;
2233 *size = t;
2234 }
2235
2236 return 0;
2237 }
2238
2239 _public_ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
2240 JournalFile *f;
2241 uint64_t p, n;
2242 le64_t le_hash;
2243 int r;
2244 Object *o;
2245
2246 assert_return(j, -EINVAL);
2247 assert_return(!journal_pid_changed(j), -ECHILD);
2248 assert_return(data, -EINVAL);
2249 assert_return(size, -EINVAL);
2250
2251 f = j->current_file;
2252 if (!f)
2253 return -EADDRNOTAVAIL;
2254
2255 if (f->current_offset <= 0)
2256 return -EADDRNOTAVAIL;
2257
2258 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2259 if (r < 0)
2260 return r;
2261
2262 n = journal_file_entry_n_items(o);
2263 if (j->current_field >= n)
2264 return 0;
2265
2266 p = le64toh(o->entry.items[j->current_field].object_offset);
2267 le_hash = o->entry.items[j->current_field].hash;
2268 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
2269 if (r < 0)
2270 return r;
2271
2272 if (le_hash != o->data.hash)
2273 return -EBADMSG;
2274
2275 r = return_data(j, f, o, data, size);
2276 if (r < 0)
2277 return r;
2278
2279 j->current_field++;
2280
2281 return 1;
2282 }
2283
2284 _public_ void sd_journal_restart_data(sd_journal *j) {
2285 if (!j)
2286 return;
2287
2288 j->current_field = 0;
2289 }
2290
2291 _public_ int sd_journal_get_fd(sd_journal *j) {
2292 int r;
2293
2294 assert_return(j, -EINVAL);
2295 assert_return(!journal_pid_changed(j), -ECHILD);
2296
2297 if (j->no_inotify)
2298 return -EMEDIUMTYPE;
2299
2300 if (j->inotify_fd >= 0)
2301 return j->inotify_fd;
2302
2303 r = allocate_inotify(j);
2304 if (r < 0)
2305 return r;
2306
2307 log_debug("Reiterating files to get inotify watches established");
2308
2309 /* Iterate through all dirs again, to add them to the
2310 * inotify */
2311 if (j->no_new_files)
2312 r = add_current_paths(j);
2313 else if (j->flags & SD_JOURNAL_OS_ROOT)
2314 r = add_search_paths(j);
2315 else if (j->toplevel_fd >= 0)
2316 r = add_root_directory(j, NULL, false);
2317 else if (j->path)
2318 r = add_root_directory(j, j->path, true);
2319 else
2320 r = add_search_paths(j);
2321 if (r < 0)
2322 return r;
2323
2324 return j->inotify_fd;
2325 }
2326
2327 _public_ int sd_journal_get_events(sd_journal *j) {
2328 int fd;
2329
2330 assert_return(j, -EINVAL);
2331 assert_return(!journal_pid_changed(j), -ECHILD);
2332
2333 fd = sd_journal_get_fd(j);
2334 if (fd < 0)
2335 return fd;
2336
2337 return POLLIN;
2338 }
2339
2340 _public_ int sd_journal_get_timeout(sd_journal *j, uint64_t *timeout_usec) {
2341 int fd;
2342
2343 assert_return(j, -EINVAL);
2344 assert_return(!journal_pid_changed(j), -ECHILD);
2345 assert_return(timeout_usec, -EINVAL);
2346
2347 fd = sd_journal_get_fd(j);
2348 if (fd < 0)
2349 return fd;
2350
2351 if (!j->on_network) {
2352 *timeout_usec = (uint64_t) -1;
2353 return 0;
2354 }
2355
2356 /* If we are on the network we need to regularly check for
2357 * changes manually */
2358
2359 *timeout_usec = j->last_process_usec + JOURNAL_FILES_RECHECK_USEC;
2360 return 1;
2361 }
2362
2363 static void process_inotify_event(sd_journal *j, struct inotify_event *e) {
2364 Directory *d;
2365
2366 assert(j);
2367 assert(e);
2368
2369 /* Is this a subdirectory we watch? */
2370 d = hashmap_get(j->directories_by_wd, INT_TO_PTR(e->wd));
2371 if (d) {
2372 sd_id128_t id;
2373
2374 if (!(e->mask & IN_ISDIR) && e->len > 0 &&
2375 (endswith(e->name, ".journal") ||
2376 endswith(e->name, ".journal~"))) {
2377
2378 /* Event for a journal file */
2379
2380 if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB))
2381 (void) add_file(j, d->path, e->name);
2382 else if (e->mask & (IN_DELETE|IN_MOVED_FROM|IN_UNMOUNT))
2383 remove_file(j, d->path, e->name);
2384
2385 } else if (!d->is_root && e->len == 0) {
2386
2387 /* Event for a subdirectory */
2388
2389 if (e->mask & (IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT))
2390 remove_directory(j, d);
2391
2392 } else if (d->is_root && (e->mask & IN_ISDIR) && e->len > 0 && sd_id128_from_string(e->name, &id) >= 0) {
2393
2394 /* Event for root directory */
2395
2396 if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB))
2397 (void) add_directory(j, d->path, e->name);
2398 }
2399
2400 return;
2401 }
2402
2403 if (e->mask & IN_IGNORED)
2404 return;
2405
2406 log_debug("Unknown inotify event.");
2407 }
2408
2409 static int determine_change(sd_journal *j) {
2410 bool b;
2411
2412 assert(j);
2413
2414 b = j->current_invalidate_counter != j->last_invalidate_counter;
2415 j->last_invalidate_counter = j->current_invalidate_counter;
2416
2417 return b ? SD_JOURNAL_INVALIDATE : SD_JOURNAL_APPEND;
2418 }
2419
2420 _public_ int sd_journal_process(sd_journal *j) {
2421 bool got_something = false;
2422
2423 assert_return(j, -EINVAL);
2424 assert_return(!journal_pid_changed(j), -ECHILD);
2425
2426 j->last_process_usec = now(CLOCK_MONOTONIC);
2427
2428 for (;;) {
2429 union inotify_event_buffer buffer;
2430 struct inotify_event *e;
2431 ssize_t l;
2432
2433 l = read(j->inotify_fd, &buffer, sizeof(buffer));
2434 if (l < 0) {
2435 if (errno == EAGAIN || errno == EINTR)
2436 return got_something ? determine_change(j) : SD_JOURNAL_NOP;
2437
2438 return -errno;
2439 }
2440
2441 got_something = true;
2442
2443 FOREACH_INOTIFY_EVENT(e, buffer, l)
2444 process_inotify_event(j, e);
2445 }
2446 }
2447
2448 _public_ int sd_journal_wait(sd_journal *j, uint64_t timeout_usec) {
2449 int r;
2450 uint64_t t;
2451
2452 assert_return(j, -EINVAL);
2453 assert_return(!journal_pid_changed(j), -ECHILD);
2454
2455 if (j->inotify_fd < 0) {
2456
2457 /* This is the first invocation, hence create the
2458 * inotify watch */
2459 r = sd_journal_get_fd(j);
2460 if (r < 0)
2461 return r;
2462
2463 /* The journal might have changed since the context
2464 * object was created and we weren't watching before,
2465 * hence don't wait for anything, and return
2466 * immediately. */
2467 return determine_change(j);
2468 }
2469
2470 r = sd_journal_get_timeout(j, &t);
2471 if (r < 0)
2472 return r;
2473
2474 if (t != (uint64_t) -1) {
2475 usec_t n;
2476
2477 n = now(CLOCK_MONOTONIC);
2478 t = t > n ? t - n : 0;
2479
2480 if (timeout_usec == (uint64_t) -1 || timeout_usec > t)
2481 timeout_usec = t;
2482 }
2483
2484 do {
2485 r = fd_wait_for_event(j->inotify_fd, POLLIN, timeout_usec);
2486 } while (r == -EINTR);
2487
2488 if (r < 0)
2489 return r;
2490
2491 return sd_journal_process(j);
2492 }
2493
2494 _public_ int sd_journal_get_cutoff_realtime_usec(sd_journal *j, uint64_t *from, uint64_t *to) {
2495 Iterator i;
2496 JournalFile *f;
2497 bool first = true;
2498 uint64_t fmin = 0, tmax = 0;
2499 int r;
2500
2501 assert_return(j, -EINVAL);
2502 assert_return(!journal_pid_changed(j), -ECHILD);
2503 assert_return(from || to, -EINVAL);
2504 assert_return(from != to, -EINVAL);
2505
2506 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2507 usec_t fr, t;
2508
2509 r = journal_file_get_cutoff_realtime_usec(f, &fr, &t);
2510 if (r == -ENOENT)
2511 continue;
2512 if (r < 0)
2513 return r;
2514 if (r == 0)
2515 continue;
2516
2517 if (first) {
2518 fmin = fr;
2519 tmax = t;
2520 first = false;
2521 } else {
2522 fmin = MIN(fr, fmin);
2523 tmax = MAX(t, tmax);
2524 }
2525 }
2526
2527 if (from)
2528 *from = fmin;
2529 if (to)
2530 *to = tmax;
2531
2532 return first ? 0 : 1;
2533 }
2534
2535 _public_ int sd_journal_get_cutoff_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t *from, uint64_t *to) {
2536 Iterator i;
2537 JournalFile *f;
2538 bool found = false;
2539 int r;
2540
2541 assert_return(j, -EINVAL);
2542 assert_return(!journal_pid_changed(j), -ECHILD);
2543 assert_return(from || to, -EINVAL);
2544 assert_return(from != to, -EINVAL);
2545
2546 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2547 usec_t fr, t;
2548
2549 r = journal_file_get_cutoff_monotonic_usec(f, boot_id, &fr, &t);
2550 if (r == -ENOENT)
2551 continue;
2552 if (r < 0)
2553 return r;
2554 if (r == 0)
2555 continue;
2556
2557 if (found) {
2558 if (from)
2559 *from = MIN(fr, *from);
2560 if (to)
2561 *to = MAX(t, *to);
2562 } else {
2563 if (from)
2564 *from = fr;
2565 if (to)
2566 *to = t;
2567 found = true;
2568 }
2569 }
2570
2571 return found;
2572 }
2573
2574 void journal_print_header(sd_journal *j) {
2575 Iterator i;
2576 JournalFile *f;
2577 bool newline = false;
2578
2579 assert(j);
2580
2581 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2582 if (newline)
2583 putchar('\n');
2584 else
2585 newline = true;
2586
2587 journal_file_print_header(f);
2588 }
2589 }
2590
2591 _public_ int sd_journal_get_usage(sd_journal *j, uint64_t *bytes) {
2592 Iterator i;
2593 JournalFile *f;
2594 uint64_t sum = 0;
2595
2596 assert_return(j, -EINVAL);
2597 assert_return(!journal_pid_changed(j), -ECHILD);
2598 assert_return(bytes, -EINVAL);
2599
2600 ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2601 struct stat st;
2602
2603 if (fstat(f->fd, &st) < 0)
2604 return -errno;
2605
2606 sum += (uint64_t) st.st_blocks * 512ULL;
2607 }
2608
2609 *bytes = sum;
2610 return 0;
2611 }
2612
2613 _public_ int sd_journal_query_unique(sd_journal *j, const char *field) {
2614 char *f;
2615
2616 assert_return(j, -EINVAL);
2617 assert_return(!journal_pid_changed(j), -ECHILD);
2618 assert_return(!isempty(field), -EINVAL);
2619 assert_return(field_is_valid(field), -EINVAL);
2620
2621 f = strdup(field);
2622 if (!f)
2623 return -ENOMEM;
2624
2625 free(j->unique_field);
2626 j->unique_field = f;
2627 j->unique_file = NULL;
2628 j->unique_offset = 0;
2629 j->unique_file_lost = false;
2630
2631 return 0;
2632 }
2633
2634 _public_ int sd_journal_enumerate_unique(sd_journal *j, const void **data, size_t *l) {
2635 size_t k;
2636
2637 assert_return(j, -EINVAL);
2638 assert_return(!journal_pid_changed(j), -ECHILD);
2639 assert_return(data, -EINVAL);
2640 assert_return(l, -EINVAL);
2641 assert_return(j->unique_field, -EINVAL);
2642
2643 k = strlen(j->unique_field);
2644
2645 if (!j->unique_file) {
2646 if (j->unique_file_lost)
2647 return 0;
2648
2649 j->unique_file = ordered_hashmap_first(j->files);
2650 if (!j->unique_file)
2651 return 0;
2652
2653 j->unique_offset = 0;
2654 }
2655
2656 for (;;) {
2657 JournalFile *of;
2658 Iterator i;
2659 Object *o;
2660 const void *odata;
2661 size_t ol;
2662 bool found;
2663 int r;
2664
2665 /* Proceed to next data object in the field's linked list */
2666 if (j->unique_offset == 0) {
2667 r = journal_file_find_field_object(j->unique_file, j->unique_field, k, &o, NULL);
2668 if (r < 0)
2669 return r;
2670
2671 j->unique_offset = r > 0 ? le64toh(o->field.head_data_offset) : 0;
2672 } else {
2673 r = journal_file_move_to_object(j->unique_file, OBJECT_DATA, j->unique_offset, &o);
2674 if (r < 0)
2675 return r;
2676
2677 j->unique_offset = le64toh(o->data.next_field_offset);
2678 }
2679
2680 /* We reached the end of the list? Then start again, with the next file */
2681 if (j->unique_offset == 0) {
2682 j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path);
2683 if (!j->unique_file)
2684 return 0;
2685
2686 continue;
2687 }
2688
2689 /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED
2690 * instead, so that we can look at this data object at the same
2691 * time as one on another file */
2692 r = journal_file_move_to_object(j->unique_file, OBJECT_UNUSED, j->unique_offset, &o);
2693 if (r < 0)
2694 return r;
2695
2696 /* Let's do the type check by hand, since we used 0 context above. */
2697 if (o->object.type != OBJECT_DATA) {
2698 log_debug("%s:offset " OFSfmt ": object has type %d, expected %d",
2699 j->unique_file->path, j->unique_offset,
2700 o->object.type, OBJECT_DATA);
2701 return -EBADMSG;
2702 }
2703
2704 r = return_data(j, j->unique_file, o, &odata, &ol);
2705 if (r < 0)
2706 return r;
2707
2708 /* Check if we have at least the field name and "=". */
2709 if (ol <= k) {
2710 log_debug("%s:offset " OFSfmt ": object has size %zu, expected at least %zu",
2711 j->unique_file->path, j->unique_offset,
2712 ol, k + 1);
2713 return -EBADMSG;
2714 }
2715
2716 if (memcmp(odata, j->unique_field, k) || ((const char*) odata)[k] != '=') {
2717 log_debug("%s:offset " OFSfmt ": object does not start with \"%s=\"",
2718 j->unique_file->path, j->unique_offset,
2719 j->unique_field);
2720 return -EBADMSG;
2721 }
2722
2723 /* OK, now let's see if we already returned this data
2724 * object by checking if it exists in the earlier
2725 * traversed files. */
2726 found = false;
2727 ORDERED_HASHMAP_FOREACH(of, j->files, i) {
2728 if (of == j->unique_file)
2729 break;
2730
2731 /* Skip this file it didn't have any fields indexed */
2732 if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0)
2733 continue;
2734
2735 r = journal_file_find_data_object_with_hash(of, odata, ol, le64toh(o->data.hash), NULL, NULL);
2736 if (r < 0)
2737 return r;
2738 if (r > 0) {
2739 found = true;
2740 break;
2741 }
2742 }
2743
2744 if (found)
2745 continue;
2746
2747 r = return_data(j, j->unique_file, o, data, l);
2748 if (r < 0)
2749 return r;
2750
2751 return 1;
2752 }
2753 }
2754
2755 _public_ void sd_journal_restart_unique(sd_journal *j) {
2756 if (!j)
2757 return;
2758
2759 j->unique_file = NULL;
2760 j->unique_offset = 0;
2761 j->unique_file_lost = false;
2762 }
2763
2764 _public_ int sd_journal_enumerate_fields(sd_journal *j, const char **field) {
2765 int r;
2766
2767 assert_return(j, -EINVAL);
2768 assert_return(!journal_pid_changed(j), -ECHILD);
2769 assert_return(field, -EINVAL);
2770
2771 if (!j->fields_file) {
2772 if (j->fields_file_lost)
2773 return 0;
2774
2775 j->fields_file = ordered_hashmap_first(j->files);
2776 if (!j->fields_file)
2777 return 0;
2778
2779 j->fields_hash_table_index = 0;
2780 j->fields_offset = 0;
2781 }
2782
2783 for (;;) {
2784 JournalFile *f, *of;
2785 Iterator i;
2786 uint64_t m;
2787 Object *o;
2788 size_t sz;
2789 bool found;
2790
2791 f = j->fields_file;
2792
2793 if (j->fields_offset == 0) {
2794 bool eof = false;
2795
2796 /* We are not yet positioned at any field. Let's pick the first one */
2797 r = journal_file_map_field_hash_table(f);
2798 if (r < 0)
2799 return r;
2800
2801 m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
2802 for (;;) {
2803 if (j->fields_hash_table_index >= m) {
2804 /* Reached the end of the hash table, go to the next file. */
2805 eof = true;
2806 break;
2807 }
2808
2809 j->fields_offset = le64toh(f->field_hash_table[j->fields_hash_table_index].head_hash_offset);
2810
2811 if (j->fields_offset != 0)
2812 break;
2813
2814 /* Empty hash table bucket, go to next one */
2815 j->fields_hash_table_index++;
2816 }
2817
2818 if (eof) {
2819 /* Proceed with next file */
2820 j->fields_file = ordered_hashmap_next(j->files, f->path);
2821 if (!j->fields_file) {
2822 *field = NULL;
2823 return 0;
2824 }
2825
2826 j->fields_offset = 0;
2827 j->fields_hash_table_index = 0;
2828 continue;
2829 }
2830
2831 } else {
2832 /* We are already positioned at a field. If so, let's figure out the next field from it */
2833
2834 r = journal_file_move_to_object(f, OBJECT_FIELD, j->fields_offset, &o);
2835 if (r < 0)
2836 return r;
2837
2838 j->fields_offset = le64toh(o->field.next_hash_offset);
2839 if (j->fields_offset == 0) {
2840 /* Reached the end of the hash table chain */
2841 j->fields_hash_table_index++;
2842 continue;
2843 }
2844 }
2845
2846 /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */
2847 r = journal_file_move_to_object(f, OBJECT_UNUSED, j->fields_offset, &o);
2848 if (r < 0)
2849 return r;
2850
2851 /* Because we used OBJECT_UNUSED above, we need to do our type check manually */
2852 if (o->object.type != OBJECT_FIELD) {
2853 log_debug("%s:offset " OFSfmt ": object has type %i, expected %i", f->path, j->fields_offset, o->object.type, OBJECT_FIELD);
2854 return -EBADMSG;
2855 }
2856
2857 sz = le64toh(o->object.size) - offsetof(Object, field.payload);
2858
2859 /* Let's see if we already returned this field name before. */
2860 found = false;
2861 ORDERED_HASHMAP_FOREACH(of, j->files, i) {
2862 if (of == f)
2863 break;
2864
2865 /* Skip this file it didn't have any fields indexed */
2866 if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0)
2867 continue;
2868
2869 r = journal_file_find_field_object_with_hash(of, o->field.payload, sz, le64toh(o->field.hash), NULL, NULL);
2870 if (r < 0)
2871 return r;
2872 if (r > 0) {
2873 found = true;
2874 break;
2875 }
2876 }
2877
2878 if (found)
2879 continue;
2880
2881 /* Check if this is really a valid string containing no NUL byte */
2882 if (memchr(o->field.payload, 0, sz))
2883 return -EBADMSG;
2884
2885 if (sz > j->data_threshold)
2886 sz = j->data_threshold;
2887
2888 if (!GREEDY_REALLOC(j->fields_buffer, j->fields_buffer_allocated, sz + 1))
2889 return -ENOMEM;
2890
2891 memcpy(j->fields_buffer, o->field.payload, sz);
2892 j->fields_buffer[sz] = 0;
2893
2894 if (!field_is_valid(j->fields_buffer))
2895 return -EBADMSG;
2896
2897 *field = j->fields_buffer;
2898 return 1;
2899 }
2900 }
2901
2902 _public_ void sd_journal_restart_fields(sd_journal *j) {
2903 if (!j)
2904 return;
2905
2906 j->fields_file = NULL;
2907 j->fields_hash_table_index = 0;
2908 j->fields_offset = 0;
2909 j->fields_file_lost = false;
2910 }
2911
2912 _public_ int sd_journal_reliable_fd(sd_journal *j) {
2913 assert_return(j, -EINVAL);
2914 assert_return(!journal_pid_changed(j), -ECHILD);
2915
2916 return !j->on_network;
2917 }
2918
2919 static char *lookup_field(const char *field, void *userdata) {
2920 sd_journal *j = userdata;
2921 const void *data;
2922 size_t size, d;
2923 int r;
2924
2925 assert(field);
2926 assert(j);
2927
2928 r = sd_journal_get_data(j, field, &data, &size);
2929 if (r < 0 ||
2930 size > REPLACE_VAR_MAX)
2931 return strdup(field);
2932
2933 d = strlen(field) + 1;
2934
2935 return strndup((const char*) data + d, size - d);
2936 }
2937
2938 _public_ int sd_journal_get_catalog(sd_journal *j, char **ret) {
2939 const void *data;
2940 size_t size;
2941 sd_id128_t id;
2942 _cleanup_free_ char *text = NULL, *cid = NULL;
2943 char *t;
2944 int r;
2945
2946 assert_return(j, -EINVAL);
2947 assert_return(!journal_pid_changed(j), -ECHILD);
2948 assert_return(ret, -EINVAL);
2949
2950 r = sd_journal_get_data(j, "MESSAGE_ID", &data, &size);
2951 if (r < 0)
2952 return r;
2953
2954 cid = strndup((const char*) data + 11, size - 11);
2955 if (!cid)
2956 return -ENOMEM;
2957
2958 r = sd_id128_from_string(cid, &id);
2959 if (r < 0)
2960 return r;
2961
2962 r = catalog_get(CATALOG_DATABASE, id, &text);
2963 if (r < 0)
2964 return r;
2965
2966 t = replace_var(text, lookup_field, j);
2967 if (!t)
2968 return -ENOMEM;
2969
2970 *ret = t;
2971 return 0;
2972 }
2973
2974 _public_ int sd_journal_get_catalog_for_message_id(sd_id128_t id, char **ret) {
2975 assert_return(ret, -EINVAL);
2976
2977 return catalog_get(CATALOG_DATABASE, id, ret);
2978 }
2979
2980 _public_ int sd_journal_set_data_threshold(sd_journal *j, size_t sz) {
2981 assert_return(j, -EINVAL);
2982 assert_return(!journal_pid_changed(j), -ECHILD);
2983
2984 j->data_threshold = sz;
2985 return 0;
2986 }
2987
2988 _public_ int sd_journal_get_data_threshold(sd_journal *j, size_t *sz) {
2989 assert_return(j, -EINVAL);
2990 assert_return(!journal_pid_changed(j), -ECHILD);
2991 assert_return(sz, -EINVAL);
2992
2993 *sz = j->data_threshold;
2994 return 0;
2995 }
2996
2997 _public_ int sd_journal_has_runtime_files(sd_journal *j) {
2998 assert_return(j, -EINVAL);
2999
3000 return j->has_runtime_files;
3001 }
3002
3003 _public_ int sd_journal_has_persistent_files(sd_journal *j) {
3004 assert_return(j, -EINVAL);
3005
3006 return j->has_persistent_files;
3007 }