2 This file is part of systemd.
4 Copyright 2011 Lennart Poettering
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.
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.
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/>.
23 #include <linux/magic.h>
26 #include <sys/inotify.h>
30 #include "sd-journal.h"
32 #include "alloc-util.h"
35 #include "dirent-util.h"
38 #include "formats-util.h"
41 #include "hostname-util.h"
43 #include "journal-def.h"
44 #include "journal-file.h"
45 #include "journal-internal.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"
56 #define JOURNAL_FILES_MAX 7168
58 #define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC)
60 #define REPLACE_VAR_MAX 256
62 #define DEFAULT_DATA_THRESHOLD (64*1024)
64 static void remove_file_real(sd_journal
*j
, JournalFile
*f
);
66 static bool journal_pid_changed(sd_journal
*j
) {
69 /* We don't support people creating a journal object and
70 * keeping it around over a fork(). Let's complain. */
72 return j
->original_pid
!= getpid();
75 static int journal_put_error(sd_journal
*j
, int r
, const char *path
) {
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
87 * We return an error here only if we didn't manage to
88 * memorize the real error. */
93 k
= hashmap_ensure_allocated(&j
->errors
, NULL
);
104 k
= hashmap_put(j
->errors
, INT_TO_PTR(r
), copy
);
117 static void detach_location(sd_journal
*j
) {
123 j
->current_file
= NULL
;
124 j
->current_field
= 0;
126 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
)
127 journal_file_reset_location(f
);
130 static void reset_location(sd_journal
*j
) {
134 zero(j
->current_location
);
137 static void init_location(Location
*l
, LocationType type
, JournalFile
*f
, Object
*o
) {
139 assert(type
== LOCATION_DISCRETE
|| type
== LOCATION_SEEK
);
141 assert(o
->object
.type
== OBJECT_ENTRY
);
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
);
151 l
->seqnum_set
= l
->realtime_set
= l
->monotonic_set
= l
->xor_hash_set
= true;
154 static void set_location(sd_journal
*j
, JournalFile
*f
, Object
*o
) {
159 init_location(&j
->current_location
, LOCATION_DISCRETE
, f
, o
);
162 j
->current_field
= 0;
164 /* Let f know its candidate entry was picked. */
165 assert(f
->location_type
== LOCATION_SEEK
);
166 f
->location_type
= LOCATION_DISCRETE
;
169 static int match_is_valid(const void *data
, size_t size
) {
177 if (startswith(data
, "__"))
181 for (p
= b
; p
< b
+ size
; p
++) {
189 if (*p
>= 'A' && *p
<= 'Z')
192 if (*p
>= '0' && *p
<= '9')
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
;
205 for (j
= 0; j
< s
&& j
< t
; j
++) {
214 assert_not_reached("\"=\" not found");
217 static Match
*match_new(Match
*p
, MatchType t
) {
228 LIST_PREPEND(matches
, p
->matches
, m
);
234 static void match_free(Match
*m
) {
238 match_free(m
->matches
);
241 LIST_REMOVE(matches
, m
->parent
->matches
, m
);
247 static void match_free_if_empty(Match
*m
) {
248 if (!m
|| m
->matches
)
254 _public_
int sd_journal_add_match(sd_journal
*j
, const void *data
, size_t size
) {
255 Match
*l3
, *l4
, *add_here
= NULL
, *m
;
258 assert_return(j
, -EINVAL
);
259 assert_return(!journal_pid_changed(j
), -ECHILD
);
260 assert_return(data
, -EINVAL
);
265 assert_return(match_is_valid(data
, size
), -EINVAL
);
271 * level 4: concrete matches */
274 j
->level0
= match_new(NULL
, MATCH_AND_TERM
);
280 j
->level1
= match_new(j
->level0
, MATCH_OR_TERM
);
286 j
->level2
= match_new(j
->level1
, MATCH_AND_TERM
);
291 assert(j
->level0
->type
== MATCH_AND_TERM
);
292 assert(j
->level1
->type
== MATCH_OR_TERM
);
293 assert(j
->level2
->type
== MATCH_AND_TERM
);
295 le_hash
= htole64(hash64(data
, size
));
297 LIST_FOREACH(matches
, l3
, j
->level2
->matches
) {
298 assert(l3
->type
== MATCH_OR_TERM
);
300 LIST_FOREACH(matches
, l4
, l3
->matches
) {
301 assert(l4
->type
== MATCH_DISCRETE
);
303 /* Exactly the same match already? Then ignore
305 if (l4
->le_hash
== le_hash
&&
307 memcmp(l4
->data
, data
, size
) == 0)
310 /* Same field? Then let's add this to this OR term */
311 if (same_field(data
, size
, l4
->data
, l4
->size
)) {
322 add_here
= match_new(j
->level2
, MATCH_OR_TERM
);
327 m
= match_new(add_here
, MATCH_DISCRETE
);
331 m
->le_hash
= le_hash
;
333 m
->data
= memdup(data
, size
);
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
);
350 _public_
int sd_journal_add_conjunction(sd_journal
*j
) {
351 assert_return(j
, -EINVAL
);
352 assert_return(!journal_pid_changed(j
), -ECHILD
);
360 if (!j
->level1
->matches
)
369 _public_
int sd_journal_add_disjunction(sd_journal
*j
) {
370 assert_return(j
, -EINVAL
);
371 assert_return(!journal_pid_changed(j
), -ECHILD
);
382 if (!j
->level2
->matches
)
389 static char *match_make_string(Match
*m
) {
392 bool enclose
= false;
395 return strdup("none");
397 if (m
->type
== MATCH_DISCRETE
)
398 return strndup(m
->data
, m
->size
);
400 LIST_FOREACH(matches
, i
, m
->matches
) {
403 t
= match_make_string(i
);
408 k
= strjoin(p
, m
->type
== MATCH_OR_TERM
? " OR " : " AND ", t
);
423 r
= strjoin("(", p
, ")");
431 char *journal_make_match_string(sd_journal
*j
) {
434 return match_make_string(j
->level0
);
437 _public_
void sd_journal_flush_matches(sd_journal
*j
) {
442 match_free(j
->level0
);
444 j
->level0
= j
->level1
= j
->level2
= NULL
;
449 _pure_
static int compare_with_location(JournalFile
*f
, Location
*l
) {
452 assert(f
->location_type
== LOCATION_SEEK
);
453 assert(l
->type
== LOCATION_DISCRETE
|| l
->type
== LOCATION_SEEK
);
455 if (l
->monotonic_set
&&
456 sd_id128_equal(f
->current_boot_id
, l
->boot_id
) &&
458 f
->current_realtime
== l
->realtime
&&
460 f
->current_xor_hash
== l
->xor_hash
)
464 sd_id128_equal(f
->header
->seqnum_id
, l
->seqnum_id
)) {
466 if (f
->current_seqnum
< l
->seqnum
)
468 if (f
->current_seqnum
> l
->seqnum
)
472 if (l
->monotonic_set
&&
473 sd_id128_equal(f
->current_boot_id
, l
->boot_id
)) {
475 if (f
->current_monotonic
< l
->monotonic
)
477 if (f
->current_monotonic
> l
->monotonic
)
481 if (l
->realtime_set
) {
483 if (f
->current_realtime
< l
->realtime
)
485 if (f
->current_realtime
> l
->realtime
)
489 if (l
->xor_hash_set
) {
491 if (f
->current_xor_hash
< l
->xor_hash
)
493 if (f
->current_xor_hash
> l
->xor_hash
)
500 static int next_for_match(
504 uint64_t after_offset
,
505 direction_t direction
,
517 if (m
->type
== MATCH_DISCRETE
) {
520 r
= journal_file_find_data_object_with_hash(f
, m
->data
, m
->size
, le64toh(m
->le_hash
), NULL
, &dp
);
524 return journal_file_move_to_entry_by_offset_for_data(f
, dp
, after_offset
, direction
, ret
, offset
);
526 } else if (m
->type
== MATCH_OR_TERM
) {
529 /* Find the earliest match beyond after_offset */
531 LIST_FOREACH(matches
, i
, m
->matches
) {
534 r
= next_for_match(j
, i
, f
, after_offset
, direction
, NULL
, &cp
);
538 if (np
== 0 || (direction
== DIRECTION_DOWN
? cp
< np
: cp
> np
))
546 } else if (m
->type
== MATCH_AND_TERM
) {
547 Match
*i
, *last_moved
;
549 /* Always jump to the next matching entry and repeat
550 * this until we find an offset that matches for all
556 r
= next_for_match(j
, m
->matches
, f
, after_offset
, direction
, NULL
, &np
);
560 assert(direction
== DIRECTION_DOWN
? np
>= after_offset
: np
<= after_offset
);
561 last_moved
= m
->matches
;
563 LIST_LOOP_BUT_ONE(matches
, i
, m
->matches
, last_moved
) {
566 r
= next_for_match(j
, i
, f
, np
, direction
, NULL
, &cp
);
570 assert(direction
== DIRECTION_DOWN
? cp
>= np
: cp
<= np
);
571 if (direction
== DIRECTION_DOWN
? cp
> np
: cp
< np
) {
580 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, np
, &n
);
592 static int find_location_for_match(
596 direction_t direction
,
606 if (m
->type
== MATCH_DISCRETE
) {
609 r
= journal_file_find_data_object_with_hash(f
, m
->data
, m
->size
, le64toh(m
->le_hash
), NULL
, &dp
);
613 /* FIXME: missing: find by monotonic */
615 if (j
->current_location
.type
== LOCATION_HEAD
)
616 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, DIRECTION_DOWN
, ret
, offset
);
617 if (j
->current_location
.type
== LOCATION_TAIL
)
618 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, DIRECTION_UP
, ret
, offset
);
619 if (j
->current_location
.seqnum_set
&& sd_id128_equal(j
->current_location
.seqnum_id
, f
->header
->seqnum_id
))
620 return journal_file_move_to_entry_by_seqnum_for_data(f
, dp
, j
->current_location
.seqnum
, direction
, ret
, offset
);
621 if (j
->current_location
.monotonic_set
) {
622 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 (j
->current_location
.realtime_set
)
627 return journal_file_move_to_entry_by_realtime_for_data(f
, dp
, j
->current_location
.realtime
, direction
, ret
, offset
);
629 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, direction
, ret
, offset
);
631 } else if (m
->type
== MATCH_OR_TERM
) {
636 /* Find the earliest match */
638 LIST_FOREACH(matches
, i
, m
->matches
) {
641 r
= find_location_for_match(j
, i
, f
, direction
, NULL
, &cp
);
645 if (np
== 0 || (direction
== DIRECTION_DOWN
? np
> cp
: np
< cp
))
653 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, np
, &n
);
668 assert(m
->type
== MATCH_AND_TERM
);
670 /* First jump to the last match, and then find the
671 * next one where all matches match */
676 LIST_FOREACH(matches
, i
, m
->matches
) {
679 r
= find_location_for_match(j
, i
, f
, direction
, NULL
, &cp
);
683 if (np
== 0 || (direction
== DIRECTION_DOWN
? cp
> np
: cp
< np
))
687 return next_for_match(j
, m
, f
, np
, direction
, ret
, offset
);
691 static int find_location_with_matches(
694 direction_t direction
,
706 /* No matches is simple */
708 if (j
->current_location
.type
== LOCATION_HEAD
)
709 return journal_file_next_entry(f
, 0, DIRECTION_DOWN
, ret
, offset
);
710 if (j
->current_location
.type
== LOCATION_TAIL
)
711 return journal_file_next_entry(f
, 0, DIRECTION_UP
, ret
, offset
);
712 if (j
->current_location
.seqnum_set
&& sd_id128_equal(j
->current_location
.seqnum_id
, f
->header
->seqnum_id
))
713 return journal_file_move_to_entry_by_seqnum(f
, j
->current_location
.seqnum
, direction
, ret
, offset
);
714 if (j
->current_location
.monotonic_set
) {
715 r
= journal_file_move_to_entry_by_monotonic(f
, j
->current_location
.boot_id
, j
->current_location
.monotonic
, direction
, ret
, offset
);
719 if (j
->current_location
.realtime_set
)
720 return journal_file_move_to_entry_by_realtime(f
, j
->current_location
.realtime
, direction
, ret
, offset
);
722 return journal_file_next_entry(f
, 0, direction
, ret
, offset
);
724 return find_location_for_match(j
, j
->level0
, f
, direction
, ret
, offset
);
727 static int next_with_matches(
730 direction_t direction
,
739 /* No matches is easy. We simple advance the file
742 return journal_file_next_entry(f
, f
->current_offset
, direction
, ret
, offset
);
744 /* If we have a match then we look for the next matching entry
745 * with an offset at least one step larger */
746 return next_for_match(j
, j
->level0
, f
,
747 direction
== DIRECTION_DOWN
? f
->current_offset
+ 1
748 : f
->current_offset
- 1,
749 direction
, ret
, offset
);
752 static int next_beyond_location(sd_journal
*j
, JournalFile
*f
, direction_t direction
) {
754 uint64_t cp
, n_entries
;
760 n_entries
= le64toh(f
->header
->n_entries
);
762 /* If we hit EOF before, we don't need to look into this file again
763 * unless direction changed or new entries appeared. */
764 if (f
->last_direction
== direction
&& f
->location_type
== LOCATION_TAIL
&&
765 n_entries
== f
->last_n_entries
)
768 f
->last_n_entries
= n_entries
;
770 if (f
->last_direction
== direction
&& f
->current_offset
> 0) {
771 /* LOCATION_SEEK here means we did the work in a previous
772 * iteration and the current location already points to a
773 * candidate entry. */
774 if (f
->location_type
!= LOCATION_SEEK
) {
775 r
= next_with_matches(j
, f
, direction
, &c
, &cp
);
779 journal_file_save_location(f
, c
, cp
);
782 f
->last_direction
= direction
;
784 r
= find_location_with_matches(j
, f
, direction
, &c
, &cp
);
788 journal_file_save_location(f
, c
, cp
);
791 /* OK, we found the spot, now let's advance until an entry
792 * that is actually different from what we were previously
793 * looking at. This is necessary to handle entries which exist
794 * in two (or more) journal files, and which shall all be
795 * suppressed but one. */
800 if (j
->current_location
.type
== LOCATION_DISCRETE
) {
803 k
= compare_with_location(f
, &j
->current_location
);
805 found
= direction
== DIRECTION_DOWN
? k
> 0 : k
< 0;
812 r
= next_with_matches(j
, f
, direction
, &c
, &cp
);
816 journal_file_save_location(f
, c
, cp
);
820 static int real_journal_next(sd_journal
*j
, direction_t direction
) {
821 JournalFile
*f
, *new_file
= NULL
;
826 assert_return(j
, -EINVAL
);
827 assert_return(!journal_pid_changed(j
), -ECHILD
);
829 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
832 r
= next_beyond_location(j
, f
, direction
);
834 log_debug_errno(r
, "Can't iterate through %s, ignoring: %m", f
->path
);
835 remove_file_real(j
, f
);
838 f
->location_type
= LOCATION_TAIL
;
847 k
= journal_file_compare_locations(f
, new_file
);
849 found
= direction
== DIRECTION_DOWN
? k
< 0 : k
> 0;
859 r
= journal_file_move_to_object(new_file
, OBJECT_ENTRY
, new_file
->current_offset
, &o
);
863 set_location(j
, new_file
, o
);
868 _public_
int sd_journal_next(sd_journal
*j
) {
869 return real_journal_next(j
, DIRECTION_DOWN
);
872 _public_
int sd_journal_previous(sd_journal
*j
) {
873 return real_journal_next(j
, DIRECTION_UP
);
876 static int real_journal_next_skip(sd_journal
*j
, direction_t direction
, uint64_t skip
) {
879 assert_return(j
, -EINVAL
);
880 assert_return(!journal_pid_changed(j
), -ECHILD
);
883 /* If this is not a discrete skip, then at least
884 * resolve the current location */
885 if (j
->current_location
.type
!= LOCATION_DISCRETE
)
886 return real_journal_next(j
, direction
);
892 r
= real_journal_next(j
, direction
);
906 _public_
int sd_journal_next_skip(sd_journal
*j
, uint64_t skip
) {
907 return real_journal_next_skip(j
, DIRECTION_DOWN
, skip
);
910 _public_
int sd_journal_previous_skip(sd_journal
*j
, uint64_t skip
) {
911 return real_journal_next_skip(j
, DIRECTION_UP
, skip
);
914 _public_
int sd_journal_get_cursor(sd_journal
*j
, char **cursor
) {
917 char bid
[33], sid
[33];
919 assert_return(j
, -EINVAL
);
920 assert_return(!journal_pid_changed(j
), -ECHILD
);
921 assert_return(cursor
, -EINVAL
);
923 if (!j
->current_file
|| j
->current_file
->current_offset
<= 0)
924 return -EADDRNOTAVAIL
;
926 r
= journal_file_move_to_object(j
->current_file
, OBJECT_ENTRY
, j
->current_file
->current_offset
, &o
);
930 sd_id128_to_string(j
->current_file
->header
->seqnum_id
, sid
);
931 sd_id128_to_string(o
->entry
.boot_id
, bid
);
934 "s=%s;i=%"PRIx64
";b=%s;m=%"PRIx64
";t=%"PRIx64
";x=%"PRIx64
,
935 sid
, le64toh(o
->entry
.seqnum
),
936 bid
, le64toh(o
->entry
.monotonic
),
937 le64toh(o
->entry
.realtime
),
938 le64toh(o
->entry
.xor_hash
)) < 0)
944 _public_
int sd_journal_seek_cursor(sd_journal
*j
, const char *cursor
) {
945 const char *word
, *state
;
947 unsigned long long seqnum
, monotonic
, realtime
, xor_hash
;
949 seqnum_id_set
= false,
952 monotonic_set
= false,
953 realtime_set
= false,
954 xor_hash_set
= false;
955 sd_id128_t seqnum_id
, boot_id
;
957 assert_return(j
, -EINVAL
);
958 assert_return(!journal_pid_changed(j
), -ECHILD
);
959 assert_return(!isempty(cursor
), -EINVAL
);
961 FOREACH_WORD_SEPARATOR(word
, l
, cursor
, ";", state
) {
965 if (l
< 2 || word
[1] != '=')
968 item
= strndup(word
, l
);
975 seqnum_id_set
= true;
976 k
= sd_id128_from_string(item
+2, &seqnum_id
);
981 if (sscanf(item
+2, "%llx", &seqnum
) != 1)
987 k
= sd_id128_from_string(item
+2, &boot_id
);
991 monotonic_set
= true;
992 if (sscanf(item
+2, "%llx", &monotonic
) != 1)
998 if (sscanf(item
+2, "%llx", &realtime
) != 1)
1003 xor_hash_set
= true;
1004 if (sscanf(item
+2, "%llx", &xor_hash
) != 1)
1015 if ((!seqnum_set
|| !seqnum_id_set
) &&
1016 (!monotonic_set
|| !boot_id_set
) &&
1022 j
->current_location
.type
= LOCATION_SEEK
;
1025 j
->current_location
.realtime
= (uint64_t) realtime
;
1026 j
->current_location
.realtime_set
= true;
1029 if (seqnum_set
&& seqnum_id_set
) {
1030 j
->current_location
.seqnum
= (uint64_t) seqnum
;
1031 j
->current_location
.seqnum_id
= seqnum_id
;
1032 j
->current_location
.seqnum_set
= true;
1035 if (monotonic_set
&& boot_id_set
) {
1036 j
->current_location
.monotonic
= (uint64_t) monotonic
;
1037 j
->current_location
.boot_id
= boot_id
;
1038 j
->current_location
.monotonic_set
= true;
1042 j
->current_location
.xor_hash
= (uint64_t) xor_hash
;
1043 j
->current_location
.xor_hash_set
= true;
1049 _public_
int sd_journal_test_cursor(sd_journal
*j
, const char *cursor
) {
1053 assert_return(j
, -EINVAL
);
1054 assert_return(!journal_pid_changed(j
), -ECHILD
);
1055 assert_return(!isempty(cursor
), -EINVAL
);
1057 if (!j
->current_file
|| j
->current_file
->current_offset
<= 0)
1058 return -EADDRNOTAVAIL
;
1060 r
= journal_file_move_to_object(j
->current_file
, OBJECT_ENTRY
, j
->current_file
->current_offset
, &o
);
1065 _cleanup_free_
char *item
= NULL
;
1066 unsigned long long ll
;
1070 r
= extract_first_word(&cursor
, &item
, ";", EXTRACT_DONT_COALESCE_SEPARATORS
);
1077 if (strlen(item
) < 2 || item
[1] != '=')
1083 k
= sd_id128_from_string(item
+2, &id
);
1086 if (!sd_id128_equal(id
, j
->current_file
->header
->seqnum_id
))
1091 if (sscanf(item
+2, "%llx", &ll
) != 1)
1093 if (ll
!= le64toh(o
->entry
.seqnum
))
1098 k
= sd_id128_from_string(item
+2, &id
);
1101 if (!sd_id128_equal(id
, o
->entry
.boot_id
))
1106 if (sscanf(item
+2, "%llx", &ll
) != 1)
1108 if (ll
!= le64toh(o
->entry
.monotonic
))
1113 if (sscanf(item
+2, "%llx", &ll
) != 1)
1115 if (ll
!= le64toh(o
->entry
.realtime
))
1120 if (sscanf(item
+2, "%llx", &ll
) != 1)
1122 if (ll
!= le64toh(o
->entry
.xor_hash
))
1132 _public_
int sd_journal_seek_monotonic_usec(sd_journal
*j
, sd_id128_t boot_id
, uint64_t usec
) {
1133 assert_return(j
, -EINVAL
);
1134 assert_return(!journal_pid_changed(j
), -ECHILD
);
1137 j
->current_location
.type
= LOCATION_SEEK
;
1138 j
->current_location
.boot_id
= boot_id
;
1139 j
->current_location
.monotonic
= usec
;
1140 j
->current_location
.monotonic_set
= true;
1145 _public_
int sd_journal_seek_realtime_usec(sd_journal
*j
, uint64_t usec
) {
1146 assert_return(j
, -EINVAL
);
1147 assert_return(!journal_pid_changed(j
), -ECHILD
);
1150 j
->current_location
.type
= LOCATION_SEEK
;
1151 j
->current_location
.realtime
= usec
;
1152 j
->current_location
.realtime_set
= true;
1157 _public_
int sd_journal_seek_head(sd_journal
*j
) {
1158 assert_return(j
, -EINVAL
);
1159 assert_return(!journal_pid_changed(j
), -ECHILD
);
1162 j
->current_location
.type
= LOCATION_HEAD
;
1167 _public_
int sd_journal_seek_tail(sd_journal
*j
) {
1168 assert_return(j
, -EINVAL
);
1169 assert_return(!journal_pid_changed(j
), -ECHILD
);
1172 j
->current_location
.type
= LOCATION_TAIL
;
1177 static void check_network(sd_journal
*j
, int fd
) {
1185 if (fstatfs(fd
, &sfs
) < 0)
1189 F_TYPE_EQUAL(sfs
.f_type
, CIFS_MAGIC_NUMBER
) ||
1190 F_TYPE_EQUAL(sfs
.f_type
, CODA_SUPER_MAGIC
) ||
1191 F_TYPE_EQUAL(sfs
.f_type
, NCP_SUPER_MAGIC
) ||
1192 F_TYPE_EQUAL(sfs
.f_type
, NFS_SUPER_MAGIC
) ||
1193 F_TYPE_EQUAL(sfs
.f_type
, SMB_SUPER_MAGIC
);
1196 static bool file_has_type_prefix(const char *prefix
, const char *filename
) {
1197 const char *full
, *tilded
, *atted
;
1199 full
= strjoina(prefix
, ".journal");
1200 tilded
= strjoina(full
, "~");
1201 atted
= strjoina(prefix
, "@");
1203 return streq(filename
, full
) ||
1204 streq(filename
, tilded
) ||
1205 startswith(filename
, atted
);
1208 static bool file_type_wanted(int flags
, const char *filename
) {
1211 if (!endswith(filename
, ".journal") && !endswith(filename
, ".journal~"))
1214 /* no flags set → every type is OK */
1215 if (!(flags
& (SD_JOURNAL_SYSTEM
| SD_JOURNAL_CURRENT_USER
)))
1218 if (flags
& SD_JOURNAL_SYSTEM
&& file_has_type_prefix("system", filename
))
1221 if (flags
& SD_JOURNAL_CURRENT_USER
) {
1222 char prefix
[5 + DECIMAL_STR_MAX(uid_t
) + 1];
1224 xsprintf(prefix
, "user-"UID_FMT
, getuid());
1226 if (file_has_type_prefix(prefix
, filename
))
1233 static bool path_has_prefix(sd_journal
*j
, const char *path
, const char *prefix
) {
1238 if (j
->toplevel_fd
>= 0)
1241 return path_startswith(path
, prefix
);
1244 static const char *skip_slash(const char *p
) {
1255 static int add_any_file(sd_journal
*j
, int fd
, const char *path
) {
1256 JournalFile
*f
= NULL
;
1257 bool close_fd
= false;
1261 assert(fd
>= 0 || path
);
1263 if (path
&& ordered_hashmap_get(j
->files
, path
))
1266 if (ordered_hashmap_size(j
->files
) >= JOURNAL_FILES_MAX
) {
1267 log_debug("Too many open journal files, not adding %s.", path
);
1272 if (fd
< 0 && j
->toplevel_fd
>= 0) {
1274 /* If there's a top-level fd defined, open the file relative to this now. (Make the path relative,
1275 * explicitly, since otherwise openat() ignores the first argument.) */
1277 fd
= openat(j
->toplevel_fd
, skip_slash(path
), O_RDONLY
|O_CLOEXEC
);
1279 r
= log_debug_errno(errno
, "Failed to open journal file %s: %m", path
);
1286 r
= journal_file_open(fd
, path
, O_RDONLY
, 0, false, false, NULL
, j
->mmap
, NULL
, NULL
, &f
);
1290 log_debug_errno(r
, "Failed to open journal file %s: %m", path
);
1294 /* journal_file_dump(f); */
1296 r
= ordered_hashmap_put(j
->files
, f
->path
, f
);
1298 f
->close_fd
= close_fd
;
1299 (void) journal_file_close(f
);
1303 if (!j
->has_runtime_files
&& path_has_prefix(j
, f
->path
, "/run"))
1304 j
->has_runtime_files
= true;
1305 else if (!j
->has_persistent_files
&& path_has_prefix(j
, f
->path
, "/var"))
1306 j
->has_persistent_files
= true;
1308 log_debug("File %s added.", f
->path
);
1310 check_network(j
, f
->fd
);
1312 j
->current_invalidate_counter
++;
1317 k
= journal_put_error(j
, r
, path
);
1324 static int add_file(sd_journal
*j
, const char *prefix
, const char *filename
) {
1331 if (j
->no_new_files
)
1334 if (!file_type_wanted(j
->flags
, filename
))
1337 path
= strjoina(prefix
, "/", filename
);
1338 return add_any_file(j
, -1, path
);
1341 static void remove_file(sd_journal
*j
, const char *prefix
, const char *filename
) {
1349 path
= strjoina(prefix
, "/", filename
);
1350 f
= ordered_hashmap_get(j
->files
, path
);
1354 remove_file_real(j
, f
);
1357 static void remove_file_real(sd_journal
*j
, JournalFile
*f
) {
1361 ordered_hashmap_remove(j
->files
, f
->path
);
1363 log_debug("File %s removed.", f
->path
);
1365 if (j
->current_file
== f
) {
1366 j
->current_file
= NULL
;
1367 j
->current_field
= 0;
1370 if (j
->unique_file
== f
) {
1371 /* Jump to the next unique_file or NULL if that one was last */
1372 j
->unique_file
= ordered_hashmap_next(j
->files
, j
->unique_file
->path
);
1373 j
->unique_offset
= 0;
1374 if (!j
->unique_file
)
1375 j
->unique_file_lost
= true;
1378 if (j
->fields_file
== f
) {
1379 j
->fields_file
= ordered_hashmap_next(j
->files
, j
->fields_file
->path
);
1380 j
->fields_offset
= 0;
1381 if (!j
->fields_file
)
1382 j
->fields_file_lost
= true;
1385 (void) journal_file_close(f
);
1387 j
->current_invalidate_counter
++;
1390 static int dirname_is_machine_id(const char *fn
) {
1391 sd_id128_t id
, machine
;
1394 r
= sd_id128_get_machine(&machine
);
1398 r
= sd_id128_from_string(fn
, &id
);
1402 return sd_id128_equal(id
, machine
);
1405 static int add_directory(sd_journal
*j
, const char *prefix
, const char *dirname
) {
1406 _cleanup_free_
char *path
= NULL
;
1407 _cleanup_closedir_
DIR *d
= NULL
;
1408 struct dirent
*de
= NULL
;
1415 /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch
1416 * and reenumerates directory contents */
1419 path
= strjoin(prefix
, "/", dirname
);
1421 path
= strdup(prefix
);
1427 log_debug("Considering directory %s.", path
);
1429 /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */
1430 if ((j
->flags
& SD_JOURNAL_LOCAL_ONLY
) &&
1431 !((dirname
&& dirname_is_machine_id(dirname
) > 0) || path_has_prefix(j
, path
, "/run")))
1435 if (j
->toplevel_fd
< 0)
1438 /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is
1439 * relative, by dropping the initial slash */
1440 d
= xopendirat(j
->toplevel_fd
, skip_slash(path
), 0);
1442 r
= log_debug_errno(errno
, "Failed to open directory %s: %m", path
);
1446 m
= hashmap_get(j
->directories_by_path
, path
);
1448 m
= new0(Directory
, 1);
1457 if (hashmap_put(j
->directories_by_path
, m
->path
, m
) < 0) {
1463 path
= NULL
; /* avoid freeing in cleanup */
1464 j
->current_invalidate_counter
++;
1466 log_debug("Directory %s added.", m
->path
);
1468 } else if (m
->is_root
)
1471 if (m
->wd
<= 0 && j
->inotify_fd
>= 0) {
1472 /* Watch this directory, if it not being watched yet. */
1474 m
->wd
= inotify_add_watch_fd(j
->inotify_fd
, dirfd(d
),
1475 IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
|IN_DELETE
|
1476 IN_DELETE_SELF
|IN_MOVE_SELF
|IN_UNMOUNT
|IN_MOVED_FROM
|
1479 if (m
->wd
> 0 && hashmap_put(j
->directories_by_wd
, INT_TO_PTR(m
->wd
), m
) < 0)
1480 inotify_rm_watch(j
->inotify_fd
, m
->wd
);
1483 FOREACH_DIRENT_ALL(de
, d
, r
= log_debug_errno(errno
, "Failed to read directory %s: %m", m
->path
); goto fail
) {
1485 if (dirent_is_file_with_suffix(de
, ".journal") ||
1486 dirent_is_file_with_suffix(de
, ".journal~"))
1487 (void) add_file(j
, m
->path
, de
->d_name
);
1490 check_network(j
, dirfd(d
));
1495 k
= journal_put_error(j
, r
, path
?: prefix
);
1502 static int add_root_directory(sd_journal
*j
, const char *p
, bool missing_ok
) {
1504 _cleanup_closedir_
DIR *d
= NULL
;
1511 /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we
1512 * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially
1513 * populate the set, as well as to update it later. */
1516 /* If there's a path specified, use it. */
1518 if ((j
->flags
& SD_JOURNAL_RUNTIME_ONLY
) &&
1519 !path_has_prefix(j
, p
, "/run"))
1523 p
= strjoina(j
->prefix
, p
);
1525 if (j
->toplevel_fd
< 0)
1528 d
= xopendirat(j
->toplevel_fd
, skip_slash(p
), 0);
1531 if (errno
== ENOENT
&& missing_ok
)
1534 r
= log_debug_errno(errno
, "Failed to open root directory %s: %m", p
);
1540 /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since
1541 * opendir() will take possession of the fd, and close it, which we don't want. */
1543 p
= "."; /* store this as "." in the directories hashmap */
1545 dfd
= fcntl(j
->toplevel_fd
, F_DUPFD_CLOEXEC
, 3);
1561 m
= hashmap_get(j
->directories_by_path
, p
);
1563 m
= new0(Directory
, 1);
1571 m
->path
= strdup(p
);
1578 if (hashmap_put(j
->directories_by_path
, m
->path
, m
) < 0) {
1585 j
->current_invalidate_counter
++;
1587 log_debug("Root directory %s added.", m
->path
);
1589 } else if (!m
->is_root
)
1592 if (m
->wd
<= 0 && j
->inotify_fd
>= 0) {
1594 m
->wd
= inotify_add_watch_fd(j
->inotify_fd
, dirfd(d
),
1595 IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
|IN_DELETE
|
1598 if (m
->wd
> 0 && hashmap_put(j
->directories_by_wd
, INT_TO_PTR(m
->wd
), m
) < 0)
1599 inotify_rm_watch(j
->inotify_fd
, m
->wd
);
1602 if (j
->no_new_files
)
1605 FOREACH_DIRENT_ALL(de
, d
, r
= log_debug_errno(errno
, "Failed to read directory %s: %m", m
->path
); goto fail
) {
1608 if (dirent_is_file_with_suffix(de
, ".journal") ||
1609 dirent_is_file_with_suffix(de
, ".journal~"))
1610 (void) add_file(j
, m
->path
, de
->d_name
);
1611 else if (IN_SET(de
->d_type
, DT_DIR
, DT_LNK
, DT_UNKNOWN
) &&
1612 sd_id128_from_string(de
->d_name
, &id
) >= 0)
1613 (void) add_directory(j
, m
->path
, de
->d_name
);
1616 check_network(j
, dirfd(d
));
1621 k
= journal_put_error(j
, r
, p
);
1628 static void remove_directory(sd_journal
*j
, Directory
*d
) {
1632 hashmap_remove(j
->directories_by_wd
, INT_TO_PTR(d
->wd
));
1634 if (j
->inotify_fd
>= 0)
1635 inotify_rm_watch(j
->inotify_fd
, d
->wd
);
1638 hashmap_remove(j
->directories_by_path
, d
->path
);
1641 log_debug("Root directory %s removed.", d
->path
);
1643 log_debug("Directory %s removed.", d
->path
);
1649 static int add_search_paths(sd_journal
*j
) {
1651 static const char search_paths
[] =
1652 "/run/log/journal\0"
1653 "/var/log/journal\0";
1658 /* We ignore most errors here, since the idea is to only open
1659 * what's actually accessible, and ignore the rest. */
1661 NULSTR_FOREACH(p
, search_paths
)
1662 (void) add_root_directory(j
, p
, true);
1667 static int add_current_paths(sd_journal
*j
) {
1672 assert(j
->no_new_files
);
1674 /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we
1675 * treat them as fatal. */
1677 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
1678 _cleanup_free_
char *dir
;
1681 dir
= dirname_malloc(f
->path
);
1685 r
= add_directory(j
, dir
, NULL
);
1693 static int allocate_inotify(sd_journal
*j
) {
1696 if (j
->inotify_fd
< 0) {
1697 j
->inotify_fd
= inotify_init1(IN_NONBLOCK
|IN_CLOEXEC
);
1698 if (j
->inotify_fd
< 0)
1702 return hashmap_ensure_allocated(&j
->directories_by_wd
, NULL
);
1705 static sd_journal
*journal_new(int flags
, const char *path
) {
1708 j
= new0(sd_journal
, 1);
1712 j
->original_pid
= getpid();
1713 j
->toplevel_fd
= -1;
1716 j
->data_threshold
= DEFAULT_DATA_THRESHOLD
;
1725 if (flags
& SD_JOURNAL_OS_ROOT
)
1731 j
->files
= ordered_hashmap_new(&string_hash_ops
);
1732 j
->directories_by_path
= hashmap_new(&string_hash_ops
);
1733 j
->mmap
= mmap_cache_new();
1734 if (!j
->files
|| !j
->directories_by_path
|| !j
->mmap
)
1740 sd_journal_close(j
);
1744 #define OPEN_ALLOWED_FLAGS \
1745 (SD_JOURNAL_LOCAL_ONLY | \
1746 SD_JOURNAL_RUNTIME_ONLY | \
1747 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)
1749 _public_
int sd_journal_open(sd_journal
**ret
, int flags
) {
1753 assert_return(ret
, -EINVAL
);
1754 assert_return((flags
& ~OPEN_ALLOWED_FLAGS
) == 0, -EINVAL
);
1756 j
= journal_new(flags
, NULL
);
1760 r
= add_search_paths(j
);
1768 sd_journal_close(j
);
1773 #define OPEN_CONTAINER_ALLOWED_FLAGS \
1774 (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM)
1776 _public_
int sd_journal_open_container(sd_journal
**ret
, const char *machine
, int flags
) {
1777 _cleanup_free_
char *root
= NULL
, *class = NULL
;
1782 /* This is pretty much deprecated, people should use machined's OpenMachineRootDirectory() call instead in
1783 * combination with sd_journal_open_directory_fd(). */
1785 assert_return(machine
, -EINVAL
);
1786 assert_return(ret
, -EINVAL
);
1787 assert_return((flags
& ~OPEN_CONTAINER_ALLOWED_FLAGS
) == 0, -EINVAL
);
1788 assert_return(machine_name_is_valid(machine
), -EINVAL
);
1790 p
= strjoina("/run/systemd/machines/", machine
);
1791 r
= parse_env_file(p
, NEWLINE
, "ROOT", &root
, "CLASS", &class, NULL
);
1799 if (!streq_ptr(class, "container"))
1802 j
= journal_new(flags
, root
);
1806 r
= add_search_paths(j
);
1814 sd_journal_close(j
);
1818 #define OPEN_DIRECTORY_ALLOWED_FLAGS \
1819 (SD_JOURNAL_OS_ROOT | \
1820 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1822 _public_
int sd_journal_open_directory(sd_journal
**ret
, const char *path
, int flags
) {
1826 assert_return(ret
, -EINVAL
);
1827 assert_return(path
, -EINVAL
);
1828 assert_return((flags
& ~OPEN_DIRECTORY_ALLOWED_FLAGS
) == 0, -EINVAL
);
1830 j
= journal_new(flags
, path
);
1834 if (flags
& SD_JOURNAL_OS_ROOT
)
1835 r
= add_search_paths(j
);
1837 r
= add_root_directory(j
, path
, false);
1845 sd_journal_close(j
);
1849 _public_
int sd_journal_open_files(sd_journal
**ret
, const char **paths
, int flags
) {
1854 assert_return(ret
, -EINVAL
);
1855 assert_return(flags
== 0, -EINVAL
);
1857 j
= journal_new(flags
, NULL
);
1861 STRV_FOREACH(path
, paths
) {
1862 r
= add_any_file(j
, -1, *path
);
1867 j
->no_new_files
= true;
1873 sd_journal_close(j
);
1877 #define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \
1878 (SD_JOURNAL_OS_ROOT | \
1879 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1881 _public_
int sd_journal_open_directory_fd(sd_journal
**ret
, int fd
, int flags
) {
1886 assert_return(ret
, -EINVAL
);
1887 assert_return(fd
>= 0, -EBADF
);
1888 assert_return((flags
& ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS
) == 0, -EINVAL
);
1890 if (fstat(fd
, &st
) < 0)
1893 if (!S_ISDIR(st
.st_mode
))
1896 j
= journal_new(flags
, NULL
);
1900 j
->toplevel_fd
= fd
;
1902 if (flags
& SD_JOURNAL_OS_ROOT
)
1903 r
= add_search_paths(j
);
1905 r
= add_root_directory(j
, NULL
, false);
1913 sd_journal_close(j
);
1917 _public_
int sd_journal_open_files_fd(sd_journal
**ret
, int fds
[], unsigned n_fds
, int flags
) {
1924 assert_return(ret
, -EINVAL
);
1925 assert_return(n_fds
> 0, -EBADF
);
1926 assert_return(flags
== 0, -EINVAL
);
1928 j
= journal_new(flags
, NULL
);
1932 for (i
= 0; i
< n_fds
; i
++) {
1940 if (fstat(fds
[i
], &st
) < 0) {
1945 if (!S_ISREG(st
.st_mode
)) {
1950 r
= add_any_file(j
, fds
[i
], NULL
);
1955 j
->no_new_files
= true;
1956 j
->no_inotify
= true;
1962 /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they
1964 ORDERED_HASHMAP_FOREACH(f
, j
->files
, iterator
)
1965 f
->close_fd
= false;
1967 sd_journal_close(j
);
1971 _public_
void sd_journal_close(sd_journal
*j
) {
1979 sd_journal_flush_matches(j
);
1981 while ((f
= ordered_hashmap_steal_first(j
->files
)))
1982 (void) journal_file_close(f
);
1984 ordered_hashmap_free(j
->files
);
1986 while ((d
= hashmap_first(j
->directories_by_path
)))
1987 remove_directory(j
, d
);
1989 while ((d
= hashmap_first(j
->directories_by_wd
)))
1990 remove_directory(j
, d
);
1992 hashmap_free(j
->directories_by_path
);
1993 hashmap_free(j
->directories_by_wd
);
1995 safe_close(j
->inotify_fd
);
1998 log_debug("mmap cache statistics: %u hit, %u miss", mmap_cache_get_hit(j
->mmap
), mmap_cache_get_missed(j
->mmap
));
1999 mmap_cache_unref(j
->mmap
);
2002 while ((p
= hashmap_steal_first(j
->errors
)))
2004 hashmap_free(j
->errors
);
2008 free(j
->unique_field
);
2009 free(j
->fields_buffer
);
2013 _public_
int sd_journal_get_realtime_usec(sd_journal
*j
, uint64_t *ret
) {
2018 assert_return(j
, -EINVAL
);
2019 assert_return(!journal_pid_changed(j
), -ECHILD
);
2020 assert_return(ret
, -EINVAL
);
2022 f
= j
->current_file
;
2024 return -EADDRNOTAVAIL
;
2026 if (f
->current_offset
<= 0)
2027 return -EADDRNOTAVAIL
;
2029 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2033 *ret
= le64toh(o
->entry
.realtime
);
2037 _public_
int sd_journal_get_monotonic_usec(sd_journal
*j
, uint64_t *ret
, sd_id128_t
*ret_boot_id
) {
2043 assert_return(j
, -EINVAL
);
2044 assert_return(!journal_pid_changed(j
), -ECHILD
);
2046 f
= j
->current_file
;
2048 return -EADDRNOTAVAIL
;
2050 if (f
->current_offset
<= 0)
2051 return -EADDRNOTAVAIL
;
2053 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2058 *ret_boot_id
= o
->entry
.boot_id
;
2060 r
= sd_id128_get_boot(&id
);
2064 if (!sd_id128_equal(id
, o
->entry
.boot_id
))
2069 *ret
= le64toh(o
->entry
.monotonic
);
2074 static bool field_is_valid(const char *field
) {
2082 if (startswith(field
, "__"))
2085 for (p
= field
; *p
; p
++) {
2090 if (*p
>= 'A' && *p
<= 'Z')
2093 if (*p
>= '0' && *p
<= '9')
2102 _public_
int sd_journal_get_data(sd_journal
*j
, const char *field
, const void **data
, size_t *size
) {
2105 size_t field_length
;
2109 assert_return(j
, -EINVAL
);
2110 assert_return(!journal_pid_changed(j
), -ECHILD
);
2111 assert_return(field
, -EINVAL
);
2112 assert_return(data
, -EINVAL
);
2113 assert_return(size
, -EINVAL
);
2114 assert_return(field_is_valid(field
), -EINVAL
);
2116 f
= j
->current_file
;
2118 return -EADDRNOTAVAIL
;
2120 if (f
->current_offset
<= 0)
2121 return -EADDRNOTAVAIL
;
2123 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2127 field_length
= strlen(field
);
2129 n
= journal_file_entry_n_items(o
);
2130 for (i
= 0; i
< n
; i
++) {
2136 p
= le64toh(o
->entry
.items
[i
].object_offset
);
2137 le_hash
= o
->entry
.items
[i
].hash
;
2138 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
2142 if (le_hash
!= o
->data
.hash
)
2145 l
= le64toh(o
->object
.size
) - offsetof(Object
, data
.payload
);
2147 compression
= o
->object
.flags
& OBJECT_COMPRESSION_MASK
;
2149 #if defined(HAVE_XZ) || defined(HAVE_LZ4)
2150 r
= decompress_startswith(compression
,
2152 &f
->compress_buffer
, &f
->compress_buffer_size
,
2153 field
, field_length
, '=');
2155 log_debug_errno(r
, "Cannot decompress %s object of length %"PRIu64
" at offset "OFSfmt
": %m",
2156 object_compressed_to_string(compression
), l
, p
);
2161 r
= decompress_blob(compression
,
2163 &f
->compress_buffer
, &f
->compress_buffer_size
, &rsize
,
2168 *data
= f
->compress_buffer
;
2169 *size
= (size_t) rsize
;
2174 return -EPROTONOSUPPORT
;
2176 } else if (l
>= field_length
+1 &&
2177 memcmp(o
->data
.payload
, field
, field_length
) == 0 &&
2178 o
->data
.payload
[field_length
] == '=') {
2182 if ((uint64_t) t
!= l
)
2185 *data
= o
->data
.payload
;
2191 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2199 static int return_data(sd_journal
*j
, JournalFile
*f
, Object
*o
, const void **data
, size_t *size
) {
2204 l
= le64toh(o
->object
.size
) - offsetof(Object
, data
.payload
);
2207 /* We can't read objects larger than 4G on a 32bit machine */
2208 if ((uint64_t) t
!= l
)
2211 compression
= o
->object
.flags
& OBJECT_COMPRESSION_MASK
;
2213 #if defined(HAVE_XZ) || defined(HAVE_LZ4)
2217 r
= decompress_blob(compression
,
2218 o
->data
.payload
, l
, &f
->compress_buffer
,
2219 &f
->compress_buffer_size
, &rsize
, j
->data_threshold
);
2223 *data
= f
->compress_buffer
;
2224 *size
= (size_t) rsize
;
2226 return -EPROTONOSUPPORT
;
2229 *data
= o
->data
.payload
;
2236 _public_
int sd_journal_enumerate_data(sd_journal
*j
, const void **data
, size_t *size
) {
2243 assert_return(j
, -EINVAL
);
2244 assert_return(!journal_pid_changed(j
), -ECHILD
);
2245 assert_return(data
, -EINVAL
);
2246 assert_return(size
, -EINVAL
);
2248 f
= j
->current_file
;
2250 return -EADDRNOTAVAIL
;
2252 if (f
->current_offset
<= 0)
2253 return -EADDRNOTAVAIL
;
2255 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2259 n
= journal_file_entry_n_items(o
);
2260 if (j
->current_field
>= n
)
2263 p
= le64toh(o
->entry
.items
[j
->current_field
].object_offset
);
2264 le_hash
= o
->entry
.items
[j
->current_field
].hash
;
2265 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
2269 if (le_hash
!= o
->data
.hash
)
2272 r
= return_data(j
, f
, o
, data
, size
);
2281 _public_
void sd_journal_restart_data(sd_journal
*j
) {
2285 j
->current_field
= 0;
2288 _public_
int sd_journal_get_fd(sd_journal
*j
) {
2291 assert_return(j
, -EINVAL
);
2292 assert_return(!journal_pid_changed(j
), -ECHILD
);
2295 return -EMEDIUMTYPE
;
2297 if (j
->inotify_fd
>= 0)
2298 return j
->inotify_fd
;
2300 r
= allocate_inotify(j
);
2304 log_debug("Reiterating files to get inotify watches established");
2306 /* Iterate through all dirs again, to add them to the
2308 if (j
->no_new_files
)
2309 r
= add_current_paths(j
);
2310 else if (j
->flags
& SD_JOURNAL_OS_ROOT
)
2311 r
= add_search_paths(j
);
2312 else if (j
->toplevel_fd
>= 0)
2313 r
= add_root_directory(j
, NULL
, false);
2315 r
= add_root_directory(j
, j
->path
, true);
2317 r
= add_search_paths(j
);
2321 return j
->inotify_fd
;
2324 _public_
int sd_journal_get_events(sd_journal
*j
) {
2327 assert_return(j
, -EINVAL
);
2328 assert_return(!journal_pid_changed(j
), -ECHILD
);
2330 fd
= sd_journal_get_fd(j
);
2337 _public_
int sd_journal_get_timeout(sd_journal
*j
, uint64_t *timeout_usec
) {
2340 assert_return(j
, -EINVAL
);
2341 assert_return(!journal_pid_changed(j
), -ECHILD
);
2342 assert_return(timeout_usec
, -EINVAL
);
2344 fd
= sd_journal_get_fd(j
);
2348 if (!j
->on_network
) {
2349 *timeout_usec
= (uint64_t) -1;
2353 /* If we are on the network we need to regularly check for
2354 * changes manually */
2356 *timeout_usec
= j
->last_process_usec
+ JOURNAL_FILES_RECHECK_USEC
;
2360 static void process_inotify_event(sd_journal
*j
, struct inotify_event
*e
) {
2366 /* Is this a subdirectory we watch? */
2367 d
= hashmap_get(j
->directories_by_wd
, INT_TO_PTR(e
->wd
));
2371 if (!(e
->mask
& IN_ISDIR
) && e
->len
> 0 &&
2372 (endswith(e
->name
, ".journal") ||
2373 endswith(e
->name
, ".journal~"))) {
2375 /* Event for a journal file */
2377 if (e
->mask
& (IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
))
2378 (void) add_file(j
, d
->path
, e
->name
);
2379 else if (e
->mask
& (IN_DELETE
|IN_MOVED_FROM
|IN_UNMOUNT
))
2380 remove_file(j
, d
->path
, e
->name
);
2382 } else if (!d
->is_root
&& e
->len
== 0) {
2384 /* Event for a subdirectory */
2386 if (e
->mask
& (IN_DELETE_SELF
|IN_MOVE_SELF
|IN_UNMOUNT
))
2387 remove_directory(j
, d
);
2389 } else if (d
->is_root
&& (e
->mask
& IN_ISDIR
) && e
->len
> 0 && sd_id128_from_string(e
->name
, &id
) >= 0) {
2391 /* Event for root directory */
2393 if (e
->mask
& (IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
))
2394 (void) add_directory(j
, d
->path
, e
->name
);
2400 if (e
->mask
& IN_IGNORED
)
2403 log_debug("Unknown inotify event.");
2406 static int determine_change(sd_journal
*j
) {
2411 b
= j
->current_invalidate_counter
!= j
->last_invalidate_counter
;
2412 j
->last_invalidate_counter
= j
->current_invalidate_counter
;
2414 return b
? SD_JOURNAL_INVALIDATE
: SD_JOURNAL_APPEND
;
2417 _public_
int sd_journal_process(sd_journal
*j
) {
2418 bool got_something
= false;
2420 assert_return(j
, -EINVAL
);
2421 assert_return(!journal_pid_changed(j
), -ECHILD
);
2423 j
->last_process_usec
= now(CLOCK_MONOTONIC
);
2426 union inotify_event_buffer buffer
;
2427 struct inotify_event
*e
;
2430 l
= read(j
->inotify_fd
, &buffer
, sizeof(buffer
));
2432 if (errno
== EAGAIN
|| errno
== EINTR
)
2433 return got_something
? determine_change(j
) : SD_JOURNAL_NOP
;
2438 got_something
= true;
2440 FOREACH_INOTIFY_EVENT(e
, buffer
, l
)
2441 process_inotify_event(j
, e
);
2445 _public_
int sd_journal_wait(sd_journal
*j
, uint64_t timeout_usec
) {
2449 assert_return(j
, -EINVAL
);
2450 assert_return(!journal_pid_changed(j
), -ECHILD
);
2452 if (j
->inotify_fd
< 0) {
2454 /* This is the first invocation, hence create the
2456 r
= sd_journal_get_fd(j
);
2460 /* The journal might have changed since the context
2461 * object was created and we weren't watching before,
2462 * hence don't wait for anything, and return
2464 return determine_change(j
);
2467 r
= sd_journal_get_timeout(j
, &t
);
2471 if (t
!= (uint64_t) -1) {
2474 n
= now(CLOCK_MONOTONIC
);
2475 t
= t
> n
? t
- n
: 0;
2477 if (timeout_usec
== (uint64_t) -1 || timeout_usec
> t
)
2482 r
= fd_wait_for_event(j
->inotify_fd
, POLLIN
, timeout_usec
);
2483 } while (r
== -EINTR
);
2488 return sd_journal_process(j
);
2491 _public_
int sd_journal_get_cutoff_realtime_usec(sd_journal
*j
, uint64_t *from
, uint64_t *to
) {
2495 uint64_t fmin
= 0, tmax
= 0;
2498 assert_return(j
, -EINVAL
);
2499 assert_return(!journal_pid_changed(j
), -ECHILD
);
2500 assert_return(from
|| to
, -EINVAL
);
2501 assert_return(from
!= to
, -EINVAL
);
2503 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2506 r
= journal_file_get_cutoff_realtime_usec(f
, &fr
, &t
);
2519 fmin
= MIN(fr
, fmin
);
2520 tmax
= MAX(t
, tmax
);
2529 return first
? 0 : 1;
2532 _public_
int sd_journal_get_cutoff_monotonic_usec(sd_journal
*j
, sd_id128_t boot_id
, uint64_t *from
, uint64_t *to
) {
2538 assert_return(j
, -EINVAL
);
2539 assert_return(!journal_pid_changed(j
), -ECHILD
);
2540 assert_return(from
|| to
, -EINVAL
);
2541 assert_return(from
!= to
, -EINVAL
);
2543 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2546 r
= journal_file_get_cutoff_monotonic_usec(f
, boot_id
, &fr
, &t
);
2556 *from
= MIN(fr
, *from
);
2571 void journal_print_header(sd_journal
*j
) {
2574 bool newline
= false;
2578 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2584 journal_file_print_header(f
);
2588 _public_
int sd_journal_get_usage(sd_journal
*j
, uint64_t *bytes
) {
2593 assert_return(j
, -EINVAL
);
2594 assert_return(!journal_pid_changed(j
), -ECHILD
);
2595 assert_return(bytes
, -EINVAL
);
2597 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2600 if (fstat(f
->fd
, &st
) < 0)
2603 sum
+= (uint64_t) st
.st_blocks
* 512ULL;
2610 _public_
int sd_journal_query_unique(sd_journal
*j
, const char *field
) {
2613 assert_return(j
, -EINVAL
);
2614 assert_return(!journal_pid_changed(j
), -ECHILD
);
2615 assert_return(!isempty(field
), -EINVAL
);
2616 assert_return(field_is_valid(field
), -EINVAL
);
2622 free(j
->unique_field
);
2623 j
->unique_field
= f
;
2624 j
->unique_file
= NULL
;
2625 j
->unique_offset
= 0;
2626 j
->unique_file_lost
= false;
2631 _public_
int sd_journal_enumerate_unique(sd_journal
*j
, const void **data
, size_t *l
) {
2634 assert_return(j
, -EINVAL
);
2635 assert_return(!journal_pid_changed(j
), -ECHILD
);
2636 assert_return(data
, -EINVAL
);
2637 assert_return(l
, -EINVAL
);
2638 assert_return(j
->unique_field
, -EINVAL
);
2640 k
= strlen(j
->unique_field
);
2642 if (!j
->unique_file
) {
2643 if (j
->unique_file_lost
)
2646 j
->unique_file
= ordered_hashmap_first(j
->files
);
2647 if (!j
->unique_file
)
2650 j
->unique_offset
= 0;
2662 /* Proceed to next data object in the field's linked list */
2663 if (j
->unique_offset
== 0) {
2664 r
= journal_file_find_field_object(j
->unique_file
, j
->unique_field
, k
, &o
, NULL
);
2668 j
->unique_offset
= r
> 0 ? le64toh(o
->field
.head_data_offset
) : 0;
2670 r
= journal_file_move_to_object(j
->unique_file
, OBJECT_DATA
, j
->unique_offset
, &o
);
2674 j
->unique_offset
= le64toh(o
->data
.next_field_offset
);
2677 /* We reached the end of the list? Then start again, with the next file */
2678 if (j
->unique_offset
== 0) {
2679 j
->unique_file
= ordered_hashmap_next(j
->files
, j
->unique_file
->path
);
2680 if (!j
->unique_file
)
2686 /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED
2687 * instead, so that we can look at this data object at the same
2688 * time as one on another file */
2689 r
= journal_file_move_to_object(j
->unique_file
, OBJECT_UNUSED
, j
->unique_offset
, &o
);
2693 /* Let's do the type check by hand, since we used 0 context above. */
2694 if (o
->object
.type
!= OBJECT_DATA
) {
2695 log_debug("%s:offset " OFSfmt
": object has type %d, expected %d",
2696 j
->unique_file
->path
, j
->unique_offset
,
2697 o
->object
.type
, OBJECT_DATA
);
2701 r
= return_data(j
, j
->unique_file
, o
, &odata
, &ol
);
2705 /* Check if we have at least the field name and "=". */
2707 log_debug("%s:offset " OFSfmt
": object has size %zu, expected at least %zu",
2708 j
->unique_file
->path
, j
->unique_offset
,
2713 if (memcmp(odata
, j
->unique_field
, k
) || ((const char*) odata
)[k
] != '=') {
2714 log_debug("%s:offset " OFSfmt
": object does not start with \"%s=\"",
2715 j
->unique_file
->path
, j
->unique_offset
,
2720 /* OK, now let's see if we already returned this data
2721 * object by checking if it exists in the earlier
2722 * traversed files. */
2724 ORDERED_HASHMAP_FOREACH(of
, j
->files
, i
) {
2725 if (of
== j
->unique_file
)
2728 /* Skip this file it didn't have any fields indexed */
2729 if (JOURNAL_HEADER_CONTAINS(of
->header
, n_fields
) && le64toh(of
->header
->n_fields
) <= 0)
2732 r
= journal_file_find_data_object_with_hash(of
, odata
, ol
, le64toh(o
->data
.hash
), NULL
, NULL
);
2744 r
= return_data(j
, j
->unique_file
, o
, data
, l
);
2752 _public_
void sd_journal_restart_unique(sd_journal
*j
) {
2756 j
->unique_file
= NULL
;
2757 j
->unique_offset
= 0;
2758 j
->unique_file_lost
= false;
2761 _public_
int sd_journal_enumerate_fields(sd_journal
*j
, const char **field
) {
2764 assert_return(j
, -EINVAL
);
2765 assert_return(!journal_pid_changed(j
), -ECHILD
);
2766 assert_return(field
, -EINVAL
);
2768 if (!j
->fields_file
) {
2769 if (j
->fields_file_lost
)
2772 j
->fields_file
= ordered_hashmap_first(j
->files
);
2773 if (!j
->fields_file
)
2776 j
->fields_hash_table_index
= 0;
2777 j
->fields_offset
= 0;
2781 JournalFile
*f
, *of
;
2790 if (j
->fields_offset
== 0) {
2793 /* We are not yet positioned at any field. Let's pick the first one */
2794 r
= journal_file_map_field_hash_table(f
);
2798 m
= le64toh(f
->header
->field_hash_table_size
) / sizeof(HashItem
);
2800 if (j
->fields_hash_table_index
>= m
) {
2801 /* Reached the end of the hash table, go to the next file. */
2806 j
->fields_offset
= le64toh(f
->field_hash_table
[j
->fields_hash_table_index
].head_hash_offset
);
2808 if (j
->fields_offset
!= 0)
2811 /* Empty hash table bucket, go to next one */
2812 j
->fields_hash_table_index
++;
2816 /* Proceed with next file */
2817 j
->fields_file
= ordered_hashmap_next(j
->files
, f
->path
);
2818 if (!j
->fields_file
) {
2823 j
->fields_offset
= 0;
2824 j
->fields_hash_table_index
= 0;
2829 /* We are already positioned at a field. If so, let's figure out the next field from it */
2831 r
= journal_file_move_to_object(f
, OBJECT_FIELD
, j
->fields_offset
, &o
);
2835 j
->fields_offset
= le64toh(o
->field
.next_hash_offset
);
2836 if (j
->fields_offset
== 0) {
2837 /* Reached the end of the hash table chain */
2838 j
->fields_hash_table_index
++;
2843 /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */
2844 r
= journal_file_move_to_object(f
, OBJECT_UNUSED
, j
->fields_offset
, &o
);
2848 /* Because we used OBJECT_UNUSED above, we need to do our type check manually */
2849 if (o
->object
.type
!= OBJECT_FIELD
) {
2850 log_debug("%s:offset " OFSfmt
": object has type %i, expected %i", f
->path
, j
->fields_offset
, o
->object
.type
, OBJECT_FIELD
);
2854 sz
= le64toh(o
->object
.size
) - offsetof(Object
, field
.payload
);
2856 /* Let's see if we already returned this field name before. */
2858 ORDERED_HASHMAP_FOREACH(of
, j
->files
, i
) {
2862 /* Skip this file it didn't have any fields indexed */
2863 if (JOURNAL_HEADER_CONTAINS(of
->header
, n_fields
) && le64toh(of
->header
->n_fields
) <= 0)
2866 r
= journal_file_find_field_object_with_hash(of
, o
->field
.payload
, sz
, le64toh(o
->field
.hash
), NULL
, NULL
);
2878 /* Check if this is really a valid string containing no NUL byte */
2879 if (memchr(o
->field
.payload
, 0, sz
))
2882 if (sz
> j
->data_threshold
)
2883 sz
= j
->data_threshold
;
2885 if (!GREEDY_REALLOC(j
->fields_buffer
, j
->fields_buffer_allocated
, sz
+ 1))
2888 memcpy(j
->fields_buffer
, o
->field
.payload
, sz
);
2889 j
->fields_buffer
[sz
] = 0;
2891 if (!field_is_valid(j
->fields_buffer
))
2894 *field
= j
->fields_buffer
;
2899 _public_
void sd_journal_restart_fields(sd_journal
*j
) {
2903 j
->fields_file
= NULL
;
2904 j
->fields_hash_table_index
= 0;
2905 j
->fields_offset
= 0;
2906 j
->fields_file_lost
= false;
2909 _public_
int sd_journal_reliable_fd(sd_journal
*j
) {
2910 assert_return(j
, -EINVAL
);
2911 assert_return(!journal_pid_changed(j
), -ECHILD
);
2913 return !j
->on_network
;
2916 static char *lookup_field(const char *field
, void *userdata
) {
2917 sd_journal
*j
= userdata
;
2925 r
= sd_journal_get_data(j
, field
, &data
, &size
);
2927 size
> REPLACE_VAR_MAX
)
2928 return strdup(field
);
2930 d
= strlen(field
) + 1;
2932 return strndup((const char*) data
+ d
, size
- d
);
2935 _public_
int sd_journal_get_catalog(sd_journal
*j
, char **ret
) {
2939 _cleanup_free_
char *text
= NULL
, *cid
= NULL
;
2943 assert_return(j
, -EINVAL
);
2944 assert_return(!journal_pid_changed(j
), -ECHILD
);
2945 assert_return(ret
, -EINVAL
);
2947 r
= sd_journal_get_data(j
, "MESSAGE_ID", &data
, &size
);
2951 cid
= strndup((const char*) data
+ 11, size
- 11);
2955 r
= sd_id128_from_string(cid
, &id
);
2959 r
= catalog_get(CATALOG_DATABASE
, id
, &text
);
2963 t
= replace_var(text
, lookup_field
, j
);
2971 _public_
int sd_journal_get_catalog_for_message_id(sd_id128_t id
, char **ret
) {
2972 assert_return(ret
, -EINVAL
);
2974 return catalog_get(CATALOG_DATABASE
, id
, ret
);
2977 _public_
int sd_journal_set_data_threshold(sd_journal
*j
, size_t sz
) {
2978 assert_return(j
, -EINVAL
);
2979 assert_return(!journal_pid_changed(j
), -ECHILD
);
2981 j
->data_threshold
= sz
;
2985 _public_
int sd_journal_get_data_threshold(sd_journal
*j
, size_t *sz
) {
2986 assert_return(j
, -EINVAL
);
2987 assert_return(!journal_pid_changed(j
), -ECHILD
);
2988 assert_return(sz
, -EINVAL
);
2990 *sz
= j
->data_threshold
;
2994 _public_
int sd_journal_has_runtime_files(sd_journal
*j
) {
2995 assert_return(j
, -EINVAL
);
2997 return j
->has_runtime_files
;
3000 _public_
int sd_journal_has_persistent_files(sd_journal
*j
) {
3001 assert_return(j
, -EINVAL
);
3003 return j
->has_persistent_files
;