1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2011 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
26 #include <sys/inotify.h>
29 #include <linux/magic.h>
31 #include "sd-journal.h"
32 #include "journal-def.h"
33 #include "journal-file.h"
36 #include "path-util.h"
39 #include "journal-internal.h"
42 #include "replace-var.h"
44 #define JOURNAL_FILES_MAX 1024
46 #define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC)
48 #define REPLACE_VAR_MAX 256
50 #define DEFAULT_DATA_THRESHOLD (64*1024)
52 /* We return an error here only if we didn't manage to
53 memorize the real error. */
54 static int set_put_error(sd_journal *j, int r) {
60 k = set_ensure_allocated(&j->errors, trivial_hash_func, trivial_compare_func);
64 return set_put(j->errors, INT_TO_PTR(r));
67 static void detach_location(sd_journal *j) {
73 j->current_file = NULL;
76 HASHMAP_FOREACH(f, j->files, i)
77 f->current_offset = 0;
80 static void reset_location(sd_journal *j) {
84 zero(j->current_location);
87 static void init_location(Location *l, LocationType type, JournalFile *f, Object *o) {
89 assert(type == LOCATION_DISCRETE || type == LOCATION_SEEK);
91 assert(o->object.type == OBJECT_ENTRY);
94 l->seqnum = le64toh(o->entry.seqnum);
95 l->seqnum_id = f->header->seqnum_id;
96 l->realtime = le64toh(o->entry.realtime);
97 l->monotonic = le64toh(o->entry.monotonic);
98 l->boot_id = o->entry.boot_id;
99 l->xor_hash = le64toh(o->entry.xor_hash);
101 l->seqnum_set = l->realtime_set = l->monotonic_set = l->xor_hash_set = true;
104 static void set_location(sd_journal *j, LocationType type, JournalFile *f, Object *o, uint64_t offset) {
106 assert(type == LOCATION_DISCRETE || type == LOCATION_SEEK);
110 init_location(&j->current_location, type, f, o);
113 j->current_file->current_offset = 0;
116 j->current_field = 0;
118 f->current_offset = offset;
121 static int match_is_valid(const void *data, size_t size) {
129 if (startswith(data, "__"))
133 for (p = b; p < b + size; p++) {
141 if (*p >= 'A' && *p <= 'Z')
144 if (*p >= '0' && *p <= '9')
153 static bool same_field(const void *_a, size_t s, const void *_b, size_t t) {
154 const uint8_t *a = _a, *b = _b;
157 for (j = 0; j < s && j < t; j++) {
169 static Match *match_new(Match *p, MatchType t) {
180 LIST_PREPEND(Match, matches, p->matches, m);
186 static void match_free(Match *m) {
190 match_free(m->matches);
193 LIST_REMOVE(Match, matches, m->parent->matches, m);
199 static void match_free_if_empty(Match *m) {
200 if (!m || m->matches)
206 _public_ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
207 Match *l3, *l4, *add_here = NULL, *m;
219 if (!match_is_valid(data, size))
226 * level 4: concrete matches */
229 j->level0 = match_new(NULL, MATCH_AND_TERM);
235 j->level1 = match_new(j->level0, MATCH_OR_TERM);
241 j->level2 = match_new(j->level1, MATCH_AND_TERM);
246 assert(j->level0->type == MATCH_AND_TERM);
247 assert(j->level1->type == MATCH_OR_TERM);
248 assert(j->level2->type == MATCH_AND_TERM);
250 le_hash = htole64(hash64(data, size));
252 LIST_FOREACH(matches, l3, j->level2->matches) {
253 assert(l3->type == MATCH_OR_TERM);
255 LIST_FOREACH(matches, l4, l3->matches) {
256 assert(l4->type == MATCH_DISCRETE);
258 /* Exactly the same match already? Then ignore
260 if (l4->le_hash == le_hash &&
262 memcmp(l4->data, data, size) == 0)
265 /* Same field? Then let's add this to this OR term */
266 if (same_field(data, size, l4->data, l4->size)) {
277 add_here = match_new(j->level2, MATCH_OR_TERM);
282 m = match_new(add_here, MATCH_DISCRETE);
286 m->le_hash = le_hash;
288 m->data = memdup(data, size);
297 match_free_if_empty(add_here);
298 match_free_if_empty(j->level2);
299 match_free_if_empty(j->level1);
300 match_free_if_empty(j->level0);
305 _public_ int sd_journal_add_conjunction(sd_journal *j) {
314 if (!j->level1->matches)
323 _public_ int sd_journal_add_disjunction(sd_journal *j) {
335 if (!j->level2->matches)
342 static char *match_make_string(Match *m) {
345 bool enclose = false;
350 if (m->type == MATCH_DISCRETE)
351 return strndup(m->data, m->size);
354 LIST_FOREACH(matches, i, m->matches) {
357 t = match_make_string(i);
364 k = strjoin(p, m->type == MATCH_OR_TERM ? " OR " : " AND ", t, NULL);
381 r = strjoin("(", p, ")", NULL);
389 char *journal_make_match_string(sd_journal *j) {
392 return match_make_string(j->level0);
395 _public_ void sd_journal_flush_matches(sd_journal *j) {
401 match_free(j->level0);
403 j->level0 = j->level1 = j->level2 = NULL;
408 static int compare_entry_order(JournalFile *af, Object *_ao,
409 JournalFile *bf, uint64_t bp) {
419 /* The mmap cache might invalidate the object from the first
420 * file if we look at the one from the second file. Hence
421 * temporarily copy the header of the first one, and look at
423 ao = alloca(offsetof(EntryObject, items));
424 memcpy(ao, _ao, offsetof(EntryObject, items));
426 r = journal_file_move_to_object(bf, OBJECT_ENTRY, bp, &bo);
428 return strcmp(af->path, bf->path);
430 /* We operate on two different files here, hence we can access
431 * two objects at the same time, which we normally can't.
433 * If contents and timestamps match, these entries are
434 * identical, even if the seqnum does not match */
436 if (sd_id128_equal(ao->entry.boot_id, bo->entry.boot_id) &&
437 ao->entry.monotonic == bo->entry.monotonic &&
438 ao->entry.realtime == bo->entry.realtime &&
439 ao->entry.xor_hash == bo->entry.xor_hash)
442 if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) {
444 /* If this is from the same seqnum source, compare
446 a = le64toh(ao->entry.seqnum);
447 b = le64toh(bo->entry.seqnum);
454 /* Wow! This is weird, different data but the same
455 * seqnums? Something is borked, but let's make the
456 * best of it and compare by time. */
459 if (sd_id128_equal(ao->entry.boot_id, bo->entry.boot_id)) {
461 /* If the boot id matches compare monotonic time */
462 a = le64toh(ao->entry.monotonic);
463 b = le64toh(bo->entry.monotonic);
471 /* Otherwise compare UTC time */
472 a = le64toh(ao->entry.realtime);
473 b = le64toh(bo->entry.realtime);
480 /* Finally, compare by contents */
481 a = le64toh(ao->entry.xor_hash);
482 b = le64toh(bo->entry.xor_hash);
492 _pure_ static int compare_with_location(JournalFile *af, Object *ao, Location *l) {
498 assert(l->type == LOCATION_DISCRETE || l->type == LOCATION_SEEK);
500 if (l->monotonic_set &&
501 sd_id128_equal(ao->entry.boot_id, l->boot_id) &&
503 le64toh(ao->entry.realtime) == l->realtime &&
505 le64toh(ao->entry.xor_hash) == l->xor_hash)
509 sd_id128_equal(af->header->seqnum_id, l->seqnum_id)) {
511 a = le64toh(ao->entry.seqnum);
519 if (l->monotonic_set &&
520 sd_id128_equal(ao->entry.boot_id, l->boot_id)) {
522 a = le64toh(ao->entry.monotonic);
524 if (a < l->monotonic)
526 if (a > l->monotonic)
530 if (l->realtime_set) {
532 a = le64toh(ao->entry.realtime);
540 if (l->xor_hash_set) {
541 a = le64toh(ao->entry.xor_hash);
552 static int next_for_match(
556 uint64_t after_offset,
557 direction_t direction,
569 if (m->type == MATCH_DISCRETE) {
572 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
576 return journal_file_move_to_entry_by_offset_for_data(f, dp, after_offset, direction, ret, offset);
578 } else if (m->type == MATCH_OR_TERM) {
581 /* Find the earliest match beyond after_offset */
583 LIST_FOREACH(matches, i, m->matches) {
586 r = next_for_match(j, i, f, after_offset, direction, NULL, &cp);
590 if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp))
595 } else if (m->type == MATCH_AND_TERM) {
596 Match *i, *last_moved;
598 /* Always jump to the next matching entry and repeat
599 * this until we find an offset that matches for all
605 r = next_for_match(j, m->matches, f, after_offset, direction, NULL, &np);
609 assert(direction == DIRECTION_DOWN ? np >= after_offset : np <= after_offset);
610 last_moved = m->matches;
612 LIST_LOOP_BUT_ONE(matches, i, m->matches, last_moved) {
615 r = next_for_match(j, i, f, np, direction, NULL, &cp);
619 assert(direction == DIRECTION_DOWN ? cp >= np : cp <= np);
620 if (direction == DIRECTION_DOWN ? cp > np : cp < np) {
630 r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
642 static int find_location_for_match(
646 direction_t direction,
656 if (m->type == MATCH_DISCRETE) {
659 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
663 /* FIXME: missing: find by monotonic */
665 if (j->current_location.type == LOCATION_HEAD)
666 return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_DOWN, ret, offset);
667 if (j->current_location.type == LOCATION_TAIL)
668 return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_UP, ret, offset);
669 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
670 return journal_file_move_to_entry_by_seqnum_for_data(f, dp, j->current_location.seqnum, direction, ret, offset);
671 if (j->current_location.monotonic_set) {
672 r = journal_file_move_to_entry_by_monotonic_for_data(f, dp, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
676 if (j->current_location.realtime_set)
677 return journal_file_move_to_entry_by_realtime_for_data(f, dp, j->current_location.realtime, direction, ret, offset);
679 return journal_file_next_entry_for_data(f, NULL, 0, dp, direction, ret, offset);
681 } else if (m->type == MATCH_OR_TERM) {
686 /* Find the earliest match */
688 LIST_FOREACH(matches, i, m->matches) {
691 r = find_location_for_match(j, i, f, direction, NULL, &cp);
695 if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp))
703 r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
718 assert(m->type == MATCH_AND_TERM);
720 /* First jump to the last match, and then find the
721 * next one where all matches match */
726 LIST_FOREACH(matches, i, m->matches) {
729 r = find_location_for_match(j, i, f, direction, NULL, &cp);
733 if (np == 0 || (direction == DIRECTION_DOWN ? np < cp : np > cp))
737 return next_for_match(j, m, f, np, direction, ret, offset);
741 static int find_location_with_matches(
744 direction_t direction,
756 /* No matches is simple */
758 if (j->current_location.type == LOCATION_HEAD)
759 return journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, ret, offset);
760 if (j->current_location.type == LOCATION_TAIL)
761 return journal_file_next_entry(f, NULL, 0, DIRECTION_UP, ret, offset);
762 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
763 return journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, ret, offset);
764 if (j->current_location.monotonic_set) {
765 r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
769 if (j->current_location.realtime_set)
770 return journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, ret, offset);
772 return journal_file_next_entry(f, NULL, 0, direction, ret, offset);
774 return find_location_for_match(j, j->level0, f, direction, ret, offset);
777 static int next_with_matches(
780 direction_t direction,
795 /* No matches is easy. We simple advance the file
798 return journal_file_next_entry(f, c, cp, direction, ret, offset);
800 /* If we have a match then we look for the next matching entry
801 * with an offset at least one step larger */
802 return next_for_match(j, j->level0, f, direction == DIRECTION_DOWN ? cp+1 : cp-1, direction, ret, offset);
805 static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction, Object **ret, uint64_t *offset) {
813 if (f->current_offset > 0) {
814 cp = f->current_offset;
816 r = journal_file_move_to_object(f, OBJECT_ENTRY, cp, &c);
820 r = next_with_matches(j, f, direction, &c, &cp);
824 r = find_location_with_matches(j, f, direction, &c, &cp);
829 /* OK, we found the spot, now let's advance until to an entry
830 * that is actually different from what we were previously
831 * looking at. This is necessary to handle entries which exist
832 * in two (or more) journal files, and which shall all be
833 * suppressed but one. */
838 if (j->current_location.type == LOCATION_DISCRETE) {
841 k = compare_with_location(f, c, &j->current_location);
842 if (direction == DIRECTION_DOWN)
857 r = next_with_matches(j, f, direction, &c, &cp);
863 static int real_journal_next(sd_journal *j, direction_t direction) {
864 JournalFile *f, *new_file = NULL;
865 uint64_t new_offset = 0;
874 HASHMAP_FOREACH(f, j->files, i) {
877 r = next_beyond_location(j, f, direction, &o, &p);
879 log_debug("Can't iterate through %s, ignoring: %s", f->path, strerror(-r));
889 k = compare_entry_order(f, o, new_file, new_offset);
891 if (direction == DIRECTION_DOWN)
906 r = journal_file_move_to_object(new_file, OBJECT_ENTRY, new_offset, &o);
910 set_location(j, LOCATION_DISCRETE, new_file, o, new_offset);
915 _public_ int sd_journal_next(sd_journal *j) {
916 return real_journal_next(j, DIRECTION_DOWN);
919 _public_ int sd_journal_previous(sd_journal *j) {
920 return real_journal_next(j, DIRECTION_UP);
923 static int real_journal_next_skip(sd_journal *j, direction_t direction, uint64_t skip) {
930 /* If this is not a discrete skip, then at least
931 * resolve the current location */
932 if (j->current_location.type != LOCATION_DISCRETE)
933 return real_journal_next(j, direction);
939 r = real_journal_next(j, direction);
953 _public_ int sd_journal_next_skip(sd_journal *j, uint64_t skip) {
954 return real_journal_next_skip(j, DIRECTION_DOWN, skip);
957 _public_ int sd_journal_previous_skip(sd_journal *j, uint64_t skip) {
958 return real_journal_next_skip(j, DIRECTION_UP, skip);
961 _public_ int sd_journal_get_cursor(sd_journal *j, char **cursor) {
964 char bid[33], sid[33];
971 if (!j->current_file || j->current_file->current_offset <= 0)
972 return -EADDRNOTAVAIL;
974 r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
978 sd_id128_to_string(j->current_file->header->seqnum_id, sid);
979 sd_id128_to_string(o->entry.boot_id, bid);
982 "s=%s;i=%llx;b=%s;m=%llx;t=%llx;x=%llx",
983 sid, (unsigned long long) le64toh(o->entry.seqnum),
984 bid, (unsigned long long) le64toh(o->entry.monotonic),
985 (unsigned long long) le64toh(o->entry.realtime),
986 (unsigned long long) le64toh(o->entry.xor_hash)) < 0)
992 _public_ int sd_journal_seek_cursor(sd_journal *j, const char *cursor) {
995 unsigned long long seqnum, monotonic, realtime, xor_hash;
997 seqnum_id_set = false,
1000 monotonic_set = false,
1001 realtime_set = false,
1002 xor_hash_set = false;
1003 sd_id128_t seqnum_id, boot_id;
1007 if (isempty(cursor))
1010 FOREACH_WORD_SEPARATOR(w, l, cursor, ";", state) {
1014 if (l < 2 || w[1] != '=')
1017 item = strndup(w, l);
1024 seqnum_id_set = true;
1025 k = sd_id128_from_string(item+2, &seqnum_id);
1030 if (sscanf(item+2, "%llx", &seqnum) != 1)
1036 k = sd_id128_from_string(item+2, &boot_id);
1040 monotonic_set = true;
1041 if (sscanf(item+2, "%llx", &monotonic) != 1)
1046 realtime_set = true;
1047 if (sscanf(item+2, "%llx", &realtime) != 1)
1052 xor_hash_set = true;
1053 if (sscanf(item+2, "%llx", &xor_hash) != 1)
1064 if ((!seqnum_set || !seqnum_id_set) &&
1065 (!monotonic_set || !boot_id_set) &&
1071 j->current_location.type = LOCATION_SEEK;
1074 j->current_location.realtime = (uint64_t) realtime;
1075 j->current_location.realtime_set = true;
1078 if (seqnum_set && seqnum_id_set) {
1079 j->current_location.seqnum = (uint64_t) seqnum;
1080 j->current_location.seqnum_id = seqnum_id;
1081 j->current_location.seqnum_set = true;
1084 if (monotonic_set && boot_id_set) {
1085 j->current_location.monotonic = (uint64_t) monotonic;
1086 j->current_location.boot_id = boot_id;
1087 j->current_location.monotonic_set = true;
1091 j->current_location.xor_hash = (uint64_t) xor_hash;
1092 j->current_location.xor_hash_set = true;
1098 _public_ int sd_journal_test_cursor(sd_journal *j, const char *cursor) {
1106 if (isempty(cursor))
1109 if (!j->current_file || j->current_file->current_offset <= 0)
1110 return -EADDRNOTAVAIL;
1112 r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
1116 FOREACH_WORD_SEPARATOR(w, l, cursor, ";", state) {
1117 _cleanup_free_ char *item = NULL;
1119 unsigned long long ll;
1122 if (l < 2 || w[1] != '=')
1125 item = strndup(w, l);
1132 k = sd_id128_from_string(item+2, &id);
1135 if (!sd_id128_equal(id, j->current_file->header->seqnum_id))
1140 if (sscanf(item+2, "%llx", &ll) != 1)
1142 if (ll != le64toh(o->entry.seqnum))
1147 k = sd_id128_from_string(item+2, &id);
1150 if (!sd_id128_equal(id, o->entry.boot_id))
1155 if (sscanf(item+2, "%llx", &ll) != 1)
1157 if (ll != le64toh(o->entry.monotonic))
1162 if (sscanf(item+2, "%llx", &ll) != 1)
1164 if (ll != le64toh(o->entry.realtime))
1169 if (sscanf(item+2, "%llx", &ll) != 1)
1171 if (ll != le64toh(o->entry.xor_hash))
1181 _public_ int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) {
1186 j->current_location.type = LOCATION_SEEK;
1187 j->current_location.boot_id = boot_id;
1188 j->current_location.monotonic = usec;
1189 j->current_location.monotonic_set = true;
1194 _public_ int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) {
1199 j->current_location.type = LOCATION_SEEK;
1200 j->current_location.realtime = usec;
1201 j->current_location.realtime_set = true;
1206 _public_ int sd_journal_seek_head(sd_journal *j) {
1211 j->current_location.type = LOCATION_HEAD;
1216 _public_ int sd_journal_seek_tail(sd_journal *j) {
1221 j->current_location.type = LOCATION_TAIL;
1226 static void check_network(sd_journal *j, int fd) {
1234 if (fstatfs(fd, &sfs) < 0)
1238 F_TYPE_CMP(sfs.f_type, CIFS_MAGIC_NUMBER) ||
1239 F_TYPE_CMP(sfs.f_type, CODA_SUPER_MAGIC) ||
1240 F_TYPE_CMP(sfs.f_type, NCP_SUPER_MAGIC) ||
1241 F_TYPE_CMP(sfs.f_type, NFS_SUPER_MAGIC) ||
1242 F_TYPE_CMP(sfs.f_type, SMB_SUPER_MAGIC);
1245 static bool file_has_type_prefix(const char *prefix, const char *filename) {
1246 const char *full, *tilded, *atted;
1248 full = strappend(prefix, ".journal");
1249 tilded = strappenda(full, "~");
1250 atted = strappenda(prefix, "@");
1252 return streq(filename, full) ||
1253 streq(filename, tilded) ||
1254 startswith(filename, atted);
1257 static bool file_type_wanted(int flags, const char *filename) {
1258 if (!endswith(filename, ".journal") && !endswith(filename, ".journal~"))
1261 /* no flags set → every type is OK */
1262 if (!(flags & (SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)))
1265 if (flags & SD_JOURNAL_SYSTEM && file_has_type_prefix("system", filename))
1268 if (flags & SD_JOURNAL_CURRENT_USER) {
1269 char prefix[5 + DECIMAL_STR_MAX(uid_t) + 1];
1271 assert_se(snprintf(prefix, sizeof(prefix), "user-%lu", (unsigned long) getuid())
1272 < (int) sizeof(prefix));
1274 if (file_has_type_prefix(prefix, filename))
1281 static int add_file(sd_journal *j, const char *prefix, const char *filename) {
1282 _cleanup_free_ char *path = NULL;
1290 if (!file_type_wanted(j->flags, filename))
1293 path = strjoin(prefix, "/", filename, NULL);
1297 if (hashmap_get(j->files, path))
1300 if (hashmap_size(j->files) >= JOURNAL_FILES_MAX) {
1301 log_debug("Too many open journal files, not adding %s, ignoring.", path);
1302 return set_put_error(j, -ETOOMANYREFS);
1305 r = journal_file_open(path, O_RDONLY, 0, false, false, NULL, j->mmap, NULL, &f);
1307 if (errno == ENOENT)
1313 /* journal_file_dump(f); */
1315 r = hashmap_put(j->files, f->path, f);
1317 journal_file_close(f);
1321 log_debug("File %s added.", f->path);
1323 check_network(j, f->fd);
1325 j->current_invalidate_counter ++;
1330 static int remove_file(sd_journal *j, const char *prefix, const char *filename) {
1338 path = strjoin(prefix, "/", filename, NULL);
1342 f = hashmap_get(j->files, path);
1347 hashmap_remove(j->files, f->path);
1349 log_debug("File %s removed.", f->path);
1351 if (j->current_file == f) {
1352 j->current_file = NULL;
1353 j->current_field = 0;
1356 if (j->unique_file == f) {
1357 j->unique_file = NULL;
1358 j->unique_offset = 0;
1361 journal_file_close(f);
1363 j->current_invalidate_counter ++;
1368 static int add_directory(sd_journal *j, const char *prefix, const char *dirname) {
1369 _cleanup_free_ char *path = NULL;
1371 _cleanup_closedir_ DIR *d = NULL;
1379 log_debug("Considering %s/%s.", prefix, dirname);
1381 if ((j->flags & SD_JOURNAL_LOCAL_ONLY) &&
1382 (sd_id128_from_string(dirname, &id) < 0 ||
1383 sd_id128_get_machine(&mid) < 0 ||
1384 !(sd_id128_equal(id, mid) || path_startswith(prefix, "/run"))))
1387 path = strjoin(prefix, "/", dirname, NULL);
1393 log_debug("Failed to open %s: %m", path);
1394 if (errno == ENOENT)
1399 m = hashmap_get(j->directories_by_path, path);
1401 m = new0(Directory, 1);
1408 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1413 path = NULL; /* avoid freeing in cleanup */
1414 j->current_invalidate_counter ++;
1416 log_debug("Directory %s added.", m->path);
1418 } else if (m->is_root)
1421 if (m->wd <= 0 && j->inotify_fd >= 0) {
1423 m->wd = inotify_add_watch(j->inotify_fd, m->path,
1424 IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1425 IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT|IN_MOVED_FROM|
1428 if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0)
1429 inotify_rm_watch(j->inotify_fd, m->wd);
1434 union dirent_storage buf;
1436 r = readdir_r(d, &buf.de, &de);
1440 if (dirent_is_file_with_suffix(de, ".journal") ||
1441 dirent_is_file_with_suffix(de, ".journal~")) {
1442 r = add_file(j, m->path, de->d_name);
1444 log_debug("Failed to add file %s/%s: %s",
1445 m->path, de->d_name, strerror(-r));
1446 r = set_put_error(j, r);
1453 check_network(j, dirfd(d));
1458 static int add_root_directory(sd_journal *j, const char *p) {
1459 _cleanup_closedir_ DIR *d = NULL;
1466 if ((j->flags & SD_JOURNAL_RUNTIME_ONLY) &&
1467 !path_startswith(p, "/run"))
1474 m = hashmap_get(j->directories_by_path, p);
1476 m = new0(Directory, 1);
1481 m->path = strdup(p);
1487 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1493 j->current_invalidate_counter ++;
1495 log_debug("Root directory %s added.", m->path);
1497 } else if (!m->is_root)
1500 if (m->wd <= 0 && j->inotify_fd >= 0) {
1502 m->wd = inotify_add_watch(j->inotify_fd, m->path,
1503 IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1506 if (m->wd > 0 && hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m) < 0)
1507 inotify_rm_watch(j->inotify_fd, m->wd);
1512 union dirent_storage buf;
1515 r = readdir_r(d, &buf.de, &de);
1519 if (dirent_is_file_with_suffix(de, ".journal") ||
1520 dirent_is_file_with_suffix(de, ".journal~")) {
1521 r = add_file(j, m->path, de->d_name);
1523 log_debug("Failed to add file %s/%s: %s",
1524 m->path, de->d_name, strerror(-r));
1525 r = set_put_error(j, r);
1529 } else if ((de->d_type == DT_DIR || de->d_type == DT_LNK || de->d_type == DT_UNKNOWN) &&
1530 sd_id128_from_string(de->d_name, &id) >= 0) {
1532 r = add_directory(j, m->path, de->d_name);
1534 log_debug("Failed to add directory %s/%s: %s", m->path, de->d_name, strerror(-r));
1538 check_network(j, dirfd(d));
1543 static int remove_directory(sd_journal *j, Directory *d) {
1547 hashmap_remove(j->directories_by_wd, INT_TO_PTR(d->wd));
1549 if (j->inotify_fd >= 0)
1550 inotify_rm_watch(j->inotify_fd, d->wd);
1553 hashmap_remove(j->directories_by_path, d->path);
1556 log_debug("Root directory %s removed.", d->path);
1558 log_debug("Directory %s removed.", d->path);
1566 static int add_search_paths(sd_journal *j) {
1568 const char search_paths[] =
1569 "/run/log/journal\0"
1570 "/var/log/journal\0";
1575 /* We ignore most errors here, since the idea is to only open
1576 * what's actually accessible, and ignore the rest. */
1578 NULSTR_FOREACH(p, search_paths) {
1579 r = add_root_directory(j, p);
1580 if (r < 0 && r != -ENOENT) {
1581 r = set_put_error(j, r);
1590 static int allocate_inotify(sd_journal *j) {
1593 if (j->inotify_fd < 0) {
1594 j->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC);
1595 if (j->inotify_fd < 0)
1599 if (!j->directories_by_wd) {
1600 j->directories_by_wd = hashmap_new(trivial_hash_func, trivial_compare_func);
1601 if (!j->directories_by_wd)
1608 static sd_journal *journal_new(int flags, const char *path) {
1611 j = new0(sd_journal, 1);
1617 j->data_threshold = DEFAULT_DATA_THRESHOLD;
1620 j->path = strdup(path);
1625 j->files = hashmap_new(string_hash_func, string_compare_func);
1626 j->directories_by_path = hashmap_new(string_hash_func, string_compare_func);
1627 j->mmap = mmap_cache_new();
1628 if (!j->files || !j->directories_by_path || !j->mmap)
1634 sd_journal_close(j);
1638 _public_ int sd_journal_open(sd_journal **ret, int flags) {
1645 if (flags & ~(SD_JOURNAL_LOCAL_ONLY|
1646 SD_JOURNAL_RUNTIME_ONLY|
1648 SD_JOURNAL_CURRENT_USER))
1651 j = journal_new(flags, NULL);
1655 r = add_search_paths(j);
1663 sd_journal_close(j);
1668 _public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) {
1681 j = journal_new(flags, path);
1685 r = add_root_directory(j, path);
1687 set_put_error(j, r);
1695 sd_journal_close(j);
1700 _public_ void sd_journal_close(sd_journal *j) {
1707 sd_journal_flush_matches(j);
1709 while ((f = hashmap_steal_first(j->files)))
1710 journal_file_close(f);
1712 hashmap_free(j->files);
1714 while ((d = hashmap_first(j->directories_by_path)))
1715 remove_directory(j, d);
1717 while ((d = hashmap_first(j->directories_by_wd)))
1718 remove_directory(j, d);
1720 hashmap_free(j->directories_by_path);
1721 hashmap_free(j->directories_by_wd);
1723 if (j->inotify_fd >= 0)
1724 close_nointr_nofail(j->inotify_fd);
1727 mmap_cache_unref(j->mmap);
1730 free(j->unique_field);
1731 set_free(j->errors);
1735 _public_ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) {
1745 f = j->current_file;
1747 return -EADDRNOTAVAIL;
1749 if (f->current_offset <= 0)
1750 return -EADDRNOTAVAIL;
1752 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1756 *ret = le64toh(o->entry.realtime);
1760 _public_ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) {
1769 f = j->current_file;
1771 return -EADDRNOTAVAIL;
1773 if (f->current_offset <= 0)
1774 return -EADDRNOTAVAIL;
1776 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1781 *ret_boot_id = o->entry.boot_id;
1783 r = sd_id128_get_boot(&id);
1787 if (!sd_id128_equal(id, o->entry.boot_id))
1792 *ret = le64toh(o->entry.monotonic);
1797 static bool field_is_valid(const char *field) {
1805 if (startswith(field, "__"))
1808 for (p = field; *p; p++) {
1813 if (*p >= 'A' && *p <= 'Z')
1816 if (*p >= '0' && *p <= '9')
1825 _public_ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) {
1828 size_t field_length;
1841 if (!field_is_valid(field))
1844 f = j->current_file;
1846 return -EADDRNOTAVAIL;
1848 if (f->current_offset <= 0)
1849 return -EADDRNOTAVAIL;
1851 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1855 field_length = strlen(field);
1857 n = journal_file_entry_n_items(o);
1858 for (i = 0; i < n; i++) {
1863 p = le64toh(o->entry.items[i].object_offset);
1864 le_hash = o->entry.items[i].hash;
1865 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
1869 if (le_hash != o->data.hash)
1872 l = le64toh(o->object.size) - offsetof(Object, data.payload);
1874 if (o->object.flags & OBJECT_COMPRESSED) {
1877 if (uncompress_startswith(o->data.payload, l,
1878 &f->compress_buffer, &f->compress_buffer_size,
1879 field, field_length, '=')) {
1883 if (!uncompress_blob(o->data.payload, l,
1884 &f->compress_buffer, &f->compress_buffer_size, &rsize,
1888 *data = f->compress_buffer;
1889 *size = (size_t) rsize;
1894 return -EPROTONOSUPPORT;
1897 } else if (l >= field_length+1 &&
1898 memcmp(o->data.payload, field, field_length) == 0 &&
1899 o->data.payload[field_length] == '=') {
1903 if ((uint64_t) t != l)
1906 *data = o->data.payload;
1912 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1920 static int return_data(sd_journal *j, JournalFile *f, Object *o, const void **data, size_t *size) {
1924 l = le64toh(o->object.size) - offsetof(Object, data.payload);
1927 /* We can't read objects larger than 4G on a 32bit machine */
1928 if ((uint64_t) t != l)
1931 if (o->object.flags & OBJECT_COMPRESSED) {
1935 if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, j->data_threshold))
1938 *data = f->compress_buffer;
1939 *size = (size_t) rsize;
1941 return -EPROTONOSUPPORT;
1944 *data = o->data.payload;
1951 _public_ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
1965 f = j->current_file;
1967 return -EADDRNOTAVAIL;
1969 if (f->current_offset <= 0)
1970 return -EADDRNOTAVAIL;
1972 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1976 n = journal_file_entry_n_items(o);
1977 if (j->current_field >= n)
1980 p = le64toh(o->entry.items[j->current_field].object_offset);
1981 le_hash = o->entry.items[j->current_field].hash;
1982 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
1986 if (le_hash != o->data.hash)
1989 r = return_data(j, f, o, data, size);
1993 j->current_field ++;
1998 _public_ void sd_journal_restart_data(sd_journal *j) {
2002 j->current_field = 0;
2005 _public_ int sd_journal_get_fd(sd_journal *j) {
2011 if (j->inotify_fd >= 0)
2012 return j->inotify_fd;
2014 r = allocate_inotify(j);
2018 /* Iterate through all dirs again, to add them to the
2021 r = add_root_directory(j, j->path);
2023 r = add_search_paths(j);
2027 return j->inotify_fd;
2030 _public_ int sd_journal_get_events(sd_journal *j) {
2036 fd = sd_journal_get_fd(j);
2043 _public_ int sd_journal_get_timeout(sd_journal *j, uint64_t *timeout_usec) {
2051 fd = sd_journal_get_fd(j);
2055 if (!j->on_network) {
2056 *timeout_usec = (uint64_t) -1;
2060 /* If we are on the network we need to regularly check for
2061 * changes manually */
2063 *timeout_usec = j->last_process_usec + JOURNAL_FILES_RECHECK_USEC;
2067 static void process_inotify_event(sd_journal *j, struct inotify_event *e) {
2074 /* Is this a subdirectory we watch? */
2075 d = hashmap_get(j->directories_by_wd, INT_TO_PTR(e->wd));
2079 if (!(e->mask & IN_ISDIR) && e->len > 0 &&
2080 (endswith(e->name, ".journal") ||
2081 endswith(e->name, ".journal~"))) {
2083 /* Event for a journal file */
2085 if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) {
2086 r = add_file(j, d->path, e->name);
2088 log_debug("Failed to add file %s/%s: %s",
2089 d->path, e->name, strerror(-r));
2090 set_put_error(j, r);
2093 } else if (e->mask & (IN_DELETE|IN_MOVED_FROM|IN_UNMOUNT)) {
2095 r = remove_file(j, d->path, e->name);
2097 log_debug("Failed to remove file %s/%s: %s", d->path, e->name, strerror(-r));
2100 } else if (!d->is_root && e->len == 0) {
2102 /* Event for a subdirectory */
2104 if (e->mask & (IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT)) {
2105 r = remove_directory(j, d);
2107 log_debug("Failed to remove directory %s: %s", d->path, strerror(-r));
2111 } else if (d->is_root && (e->mask & IN_ISDIR) && e->len > 0 && sd_id128_from_string(e->name, &id) >= 0) {
2113 /* Event for root directory */
2115 if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) {
2116 r = add_directory(j, d->path, e->name);
2118 log_debug("Failed to add directory %s/%s: %s", d->path, e->name, strerror(-r));
2125 if (e->mask & IN_IGNORED)
2128 log_warning("Unknown inotify event.");
2131 static int determine_change(sd_journal *j) {
2136 b = j->current_invalidate_counter != j->last_invalidate_counter;
2137 j->last_invalidate_counter = j->current_invalidate_counter;
2139 return b ? SD_JOURNAL_INVALIDATE : SD_JOURNAL_APPEND;
2142 _public_ int sd_journal_process(sd_journal *j) {
2143 uint8_t buffer[sizeof(struct inotify_event) + FILENAME_MAX] _alignas_(struct inotify_event);
2144 bool got_something = false;
2149 j->last_process_usec = now(CLOCK_MONOTONIC);
2152 struct inotify_event *e;
2155 l = read(j->inotify_fd, buffer, sizeof(buffer));
2157 if (errno == EAGAIN || errno == EINTR)
2158 return got_something ? determine_change(j) : SD_JOURNAL_NOP;
2163 got_something = true;
2165 e = (struct inotify_event*) buffer;
2169 process_inotify_event(j, e);
2171 step = sizeof(struct inotify_event) + e->len;
2172 assert(step <= (size_t) l);
2174 e = (struct inotify_event*) ((uint8_t*) e + step);
2179 return determine_change(j);
2182 _public_ int sd_journal_wait(sd_journal *j, uint64_t timeout_usec) {
2188 if (j->inotify_fd < 0) {
2190 /* This is the first invocation, hence create the
2192 r = sd_journal_get_fd(j);
2196 /* The journal might have changed since the context
2197 * object was created and we weren't watching before,
2198 * hence don't wait for anything, and return
2200 return determine_change(j);
2203 r = sd_journal_get_timeout(j, &t);
2207 if (t != (uint64_t) -1) {
2210 n = now(CLOCK_MONOTONIC);
2211 t = t > n ? t - n : 0;
2213 if (timeout_usec == (uint64_t) -1 || timeout_usec > t)
2218 r = fd_wait_for_event(j->inotify_fd, POLLIN, timeout_usec);
2219 } while (r == -EINTR);
2224 return sd_journal_process(j);
2227 _public_ int sd_journal_get_cutoff_realtime_usec(sd_journal *j, uint64_t *from, uint64_t *to) {
2240 HASHMAP_FOREACH(f, j->files, i) {
2243 r = journal_file_get_cutoff_realtime_usec(f, &fr, &t);
2259 *from = MIN(fr, *from);
2265 return first ? 0 : 1;
2268 _public_ int sd_journal_get_cutoff_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t *from, uint64_t *to) {
2281 HASHMAP_FOREACH(f, j->files, i) {
2284 r = journal_file_get_cutoff_monotonic_usec(f, boot_id, &fr, &t);
2300 *from = MIN(fr, *from);
2306 return first ? 0 : 1;
2309 void journal_print_header(sd_journal *j) {
2312 bool newline = false;
2316 HASHMAP_FOREACH(f, j->files, i) {
2322 journal_file_print_header(f);
2326 _public_ int sd_journal_get_usage(sd_journal *j, uint64_t *bytes) {
2336 HASHMAP_FOREACH(f, j->files, i) {
2339 if (fstat(f->fd, &st) < 0)
2342 sum += (uint64_t) st.st_blocks * 512ULL;
2349 _public_ int sd_journal_query_unique(sd_journal *j, const char *field) {
2356 if (!field_is_valid(field))
2363 free(j->unique_field);
2364 j->unique_field = f;
2365 j->unique_file = NULL;
2366 j->unique_offset = 0;
2371 _public_ int sd_journal_enumerate_unique(sd_journal *j, const void **data, size_t *l) {
2382 if (!j->unique_field)
2385 k = strlen(j->unique_field);
2387 if (!j->unique_file) {
2388 j->unique_file = hashmap_first(j->files);
2389 if (!j->unique_file)
2391 j->unique_offset = 0;
2401 /* Proceed to next data object in the field's linked list */
2402 if (j->unique_offset == 0) {
2403 r = journal_file_find_field_object(j->unique_file, j->unique_field, k, &o, NULL);
2407 j->unique_offset = r > 0 ? le64toh(o->field.head_data_offset) : 0;
2409 r = journal_file_move_to_object(j->unique_file, OBJECT_DATA, j->unique_offset, &o);
2413 j->unique_offset = le64toh(o->data.next_field_offset);
2416 /* We reached the end of the list? Then start again, with the next file */
2417 if (j->unique_offset == 0) {
2420 n = hashmap_next(j->files, j->unique_file->path);
2428 /* We do not use the type context here, but 0 instead,
2429 * so that we can look at this data object at the same
2430 * time as one on another file */
2431 r = journal_file_move_to_object(j->unique_file, 0, j->unique_offset, &o);
2435 /* Let's do the type check by hand, since we used 0 context above. */
2436 if (o->object.type != OBJECT_DATA)
2439 r = return_data(j, j->unique_file, o, &odata, &ol);
2443 /* OK, now let's see if we already returned this data
2444 * object by checking if it exists in the earlier
2445 * traversed files. */
2447 HASHMAP_FOREACH(of, j->files, i) {
2451 if (of == j->unique_file)
2454 /* Skip this file it didn't have any fields
2456 if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) &&
2457 le64toh(of->header->n_fields) <= 0)
2460 r = journal_file_find_data_object_with_hash(of, odata, ol, le64toh(o->data.hash), &oo, &op);
2471 r = return_data(j, j->unique_file, o, data, l);
2479 _public_ void sd_journal_restart_unique(sd_journal *j) {
2483 j->unique_file = NULL;
2484 j->unique_offset = 0;
2487 _public_ int sd_journal_reliable_fd(sd_journal *j) {
2491 return !j->on_network;
2494 static char *lookup_field(const char *field, void *userdata) {
2495 sd_journal *j = userdata;
2503 r = sd_journal_get_data(j, field, &data, &size);
2505 size > REPLACE_VAR_MAX)
2506 return strdup(field);
2508 d = strlen(field) + 1;
2510 return strndup((const char*) data + d, size - d);
2513 _public_ int sd_journal_get_catalog(sd_journal *j, char **ret) {
2517 _cleanup_free_ char *text = NULL, *cid = NULL;
2526 r = sd_journal_get_data(j, "MESSAGE_ID", &data, &size);
2530 cid = strndup((const char*) data + 11, size - 11);
2534 r = sd_id128_from_string(cid, &id);
2538 r = catalog_get(CATALOG_DATABASE, id, &text);
2542 t = replace_var(text, lookup_field, j);
2550 _public_ int sd_journal_get_catalog_for_message_id(sd_id128_t id, char **ret) {
2554 return catalog_get(CATALOG_DATABASE, id, ret);
2557 _public_ int sd_journal_set_data_threshold(sd_journal *j, size_t sz) {
2561 j->data_threshold = sz;
2565 _public_ int sd_journal_get_data_threshold(sd_journal *j, size_t *sz) {
2571 *sz = j->data_threshold;