1 /* SPDX-License-Identifier: LGPL-2.1+ */
6 #include <linux/magic.h>
9 #include <sys/inotify.h>
13 #include "sd-journal.h"
15 #include "alloc-util.h"
18 #include "dirent-util.h"
21 #include "format-util.h"
24 #include "hostname-util.h"
25 #include "id128-util.h"
27 #include "journal-def.h"
28 #include "journal-file.h"
29 #include "journal-internal.h"
33 #include "path-util.h"
34 #include "process-util.h"
35 #include "replace-var.h"
36 #include "stat-util.h"
37 #include "stat-util.h"
38 #include "stdio-util.h"
39 #include "string-util.h"
42 #define JOURNAL_FILES_MAX 7168
44 #define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC)
46 #define REPLACE_VAR_MAX 256
48 #define DEFAULT_DATA_THRESHOLD (64*1024)
50 static void remove_file_real(sd_journal
*j
, JournalFile
*f
);
52 static bool journal_pid_changed(sd_journal
*j
) {
55 /* We don't support people creating a journal object and
56 * keeping it around over a fork(). Let's complain. */
58 return j
->original_pid
!= getpid_cached();
61 static int journal_put_error(sd_journal
*j
, int r
, const char *path
) {
65 /* Memorize an error we encountered, and store which
66 * file/directory it was generated from. Note that we store
67 * only *one* path per error code, as the error code is the
68 * key into the hashmap, and the path is the value. This means
69 * we keep track only of all error kinds, but not of all error
70 * locations. This has the benefit that the hashmap cannot
73 * We return an error here only if we didn't manage to
74 * memorize the real error. */
79 k
= hashmap_ensure_allocated(&j
->errors
, NULL
);
90 k
= hashmap_put(j
->errors
, INT_TO_PTR(r
), copy
);
103 static void detach_location(sd_journal
*j
) {
109 j
->current_file
= NULL
;
110 j
->current_field
= 0;
112 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
)
113 journal_file_reset_location(f
);
116 static void reset_location(sd_journal
*j
) {
120 zero(j
->current_location
);
123 static void init_location(Location
*l
, LocationType type
, JournalFile
*f
, Object
*o
) {
125 assert(IN_SET(type
, LOCATION_DISCRETE
, LOCATION_SEEK
));
127 assert(o
->object
.type
== OBJECT_ENTRY
);
130 l
->seqnum
= le64toh(o
->entry
.seqnum
);
131 l
->seqnum_id
= f
->header
->seqnum_id
;
132 l
->realtime
= le64toh(o
->entry
.realtime
);
133 l
->monotonic
= le64toh(o
->entry
.monotonic
);
134 l
->boot_id
= o
->entry
.boot_id
;
135 l
->xor_hash
= le64toh(o
->entry
.xor_hash
);
137 l
->seqnum_set
= l
->realtime_set
= l
->monotonic_set
= l
->xor_hash_set
= true;
140 static void set_location(sd_journal
*j
, JournalFile
*f
, Object
*o
) {
145 init_location(&j
->current_location
, LOCATION_DISCRETE
, f
, o
);
148 j
->current_field
= 0;
150 /* Let f know its candidate entry was picked. */
151 assert(f
->location_type
== LOCATION_SEEK
);
152 f
->location_type
= LOCATION_DISCRETE
;
155 static int match_is_valid(const void *data
, size_t size
) {
163 if (startswith(data
, "__"))
167 for (p
= b
; p
< b
+ size
; p
++) {
175 if (*p
>= 'A' && *p
<= 'Z')
178 if (*p
>= '0' && *p
<= '9')
187 static bool same_field(const void *_a
, size_t s
, const void *_b
, size_t t
) {
188 const uint8_t *a
= _a
, *b
= _b
;
191 for (j
= 0; j
< s
&& j
< t
; j
++) {
200 assert_not_reached("\"=\" not found");
203 static Match
*match_new(Match
*p
, MatchType t
) {
214 LIST_PREPEND(matches
, p
->matches
, m
);
220 static void match_free(Match
*m
) {
224 match_free(m
->matches
);
227 LIST_REMOVE(matches
, m
->parent
->matches
, m
);
233 static void match_free_if_empty(Match
*m
) {
234 if (!m
|| m
->matches
)
240 _public_
int sd_journal_add_match(sd_journal
*j
, const void *data
, size_t size
) {
241 Match
*l3
, *l4
, *add_here
= NULL
, *m
;
244 assert_return(j
, -EINVAL
);
245 assert_return(!journal_pid_changed(j
), -ECHILD
);
246 assert_return(data
, -EINVAL
);
251 assert_return(match_is_valid(data
, size
), -EINVAL
);
257 * level 4: concrete matches */
260 j
->level0
= match_new(NULL
, MATCH_AND_TERM
);
266 j
->level1
= match_new(j
->level0
, MATCH_OR_TERM
);
272 j
->level2
= match_new(j
->level1
, MATCH_AND_TERM
);
277 assert(j
->level0
->type
== MATCH_AND_TERM
);
278 assert(j
->level1
->type
== MATCH_OR_TERM
);
279 assert(j
->level2
->type
== MATCH_AND_TERM
);
281 le_hash
= htole64(hash64(data
, size
));
283 LIST_FOREACH(matches
, l3
, j
->level2
->matches
) {
284 assert(l3
->type
== MATCH_OR_TERM
);
286 LIST_FOREACH(matches
, l4
, l3
->matches
) {
287 assert(l4
->type
== MATCH_DISCRETE
);
289 /* Exactly the same match already? Then ignore
291 if (l4
->le_hash
== le_hash
&&
293 memcmp(l4
->data
, data
, size
) == 0)
296 /* Same field? Then let's add this to this OR term */
297 if (same_field(data
, size
, l4
->data
, l4
->size
)) {
308 add_here
= match_new(j
->level2
, MATCH_OR_TERM
);
313 m
= match_new(add_here
, MATCH_DISCRETE
);
317 m
->le_hash
= le_hash
;
319 m
->data
= memdup(data
, size
);
328 match_free_if_empty(add_here
);
329 match_free_if_empty(j
->level2
);
330 match_free_if_empty(j
->level1
);
331 match_free_if_empty(j
->level0
);
336 _public_
int sd_journal_add_conjunction(sd_journal
*j
) {
337 assert_return(j
, -EINVAL
);
338 assert_return(!journal_pid_changed(j
), -ECHILD
);
346 if (!j
->level1
->matches
)
355 _public_
int sd_journal_add_disjunction(sd_journal
*j
) {
356 assert_return(j
, -EINVAL
);
357 assert_return(!journal_pid_changed(j
), -ECHILD
);
368 if (!j
->level2
->matches
)
375 static char *match_make_string(Match
*m
) {
378 bool enclose
= false;
381 return strdup("none");
383 if (m
->type
== MATCH_DISCRETE
)
384 return strndup(m
->data
, m
->size
);
386 LIST_FOREACH(matches
, i
, m
->matches
) {
389 t
= match_make_string(i
);
394 k
= strjoin(p
, m
->type
== MATCH_OR_TERM
? " OR " : " AND ", t
);
409 r
= strjoin("(", p
, ")");
417 char *journal_make_match_string(sd_journal
*j
) {
420 return match_make_string(j
->level0
);
423 _public_
void sd_journal_flush_matches(sd_journal
*j
) {
428 match_free(j
->level0
);
430 j
->level0
= j
->level1
= j
->level2
= NULL
;
435 _pure_
static int compare_with_location(JournalFile
*f
, Location
*l
) {
438 assert(f
->location_type
== LOCATION_SEEK
);
439 assert(IN_SET(l
->type
, LOCATION_DISCRETE
, LOCATION_SEEK
));
441 if (l
->monotonic_set
&&
442 sd_id128_equal(f
->current_boot_id
, l
->boot_id
) &&
444 f
->current_realtime
== l
->realtime
&&
446 f
->current_xor_hash
== l
->xor_hash
)
450 sd_id128_equal(f
->header
->seqnum_id
, l
->seqnum_id
)) {
452 if (f
->current_seqnum
< l
->seqnum
)
454 if (f
->current_seqnum
> l
->seqnum
)
458 if (l
->monotonic_set
&&
459 sd_id128_equal(f
->current_boot_id
, l
->boot_id
)) {
461 if (f
->current_monotonic
< l
->monotonic
)
463 if (f
->current_monotonic
> l
->monotonic
)
467 if (l
->realtime_set
) {
469 if (f
->current_realtime
< l
->realtime
)
471 if (f
->current_realtime
> l
->realtime
)
475 if (l
->xor_hash_set
) {
477 if (f
->current_xor_hash
< l
->xor_hash
)
479 if (f
->current_xor_hash
> l
->xor_hash
)
486 static int next_for_match(
490 uint64_t after_offset
,
491 direction_t direction
,
503 if (m
->type
== MATCH_DISCRETE
) {
506 r
= journal_file_find_data_object_with_hash(f
, m
->data
, m
->size
, le64toh(m
->le_hash
), NULL
, &dp
);
510 return journal_file_move_to_entry_by_offset_for_data(f
, dp
, after_offset
, direction
, ret
, offset
);
512 } else if (m
->type
== MATCH_OR_TERM
) {
515 /* Find the earliest match beyond after_offset */
517 LIST_FOREACH(matches
, i
, m
->matches
) {
520 r
= next_for_match(j
, i
, f
, after_offset
, direction
, NULL
, &cp
);
524 if (np
== 0 || (direction
== DIRECTION_DOWN
? cp
< np
: cp
> np
))
532 } else if (m
->type
== MATCH_AND_TERM
) {
533 Match
*i
, *last_moved
;
535 /* Always jump to the next matching entry and repeat
536 * this until we find an offset that matches for all
542 r
= next_for_match(j
, m
->matches
, f
, after_offset
, direction
, NULL
, &np
);
546 assert(direction
== DIRECTION_DOWN
? np
>= after_offset
: np
<= after_offset
);
547 last_moved
= m
->matches
;
549 LIST_LOOP_BUT_ONE(matches
, i
, m
->matches
, last_moved
) {
552 r
= next_for_match(j
, i
, f
, np
, direction
, NULL
, &cp
);
556 assert(direction
== DIRECTION_DOWN
? cp
>= np
: cp
<= np
);
557 if (direction
== DIRECTION_DOWN
? cp
> np
: cp
< np
) {
566 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, np
, &n
);
578 static int find_location_for_match(
582 direction_t direction
,
592 if (m
->type
== MATCH_DISCRETE
) {
595 r
= journal_file_find_data_object_with_hash(f
, m
->data
, m
->size
, le64toh(m
->le_hash
), NULL
, &dp
);
599 /* FIXME: missing: find by monotonic */
601 if (j
->current_location
.type
== LOCATION_HEAD
)
602 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, DIRECTION_DOWN
, ret
, offset
);
603 if (j
->current_location
.type
== LOCATION_TAIL
)
604 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, DIRECTION_UP
, ret
, offset
);
605 if (j
->current_location
.seqnum_set
&& sd_id128_equal(j
->current_location
.seqnum_id
, f
->header
->seqnum_id
))
606 return journal_file_move_to_entry_by_seqnum_for_data(f
, dp
, j
->current_location
.seqnum
, direction
, ret
, offset
);
607 if (j
->current_location
.monotonic_set
) {
608 r
= journal_file_move_to_entry_by_monotonic_for_data(f
, dp
, j
->current_location
.boot_id
, j
->current_location
.monotonic
, direction
, ret
, offset
);
612 if (j
->current_location
.realtime_set
)
613 return journal_file_move_to_entry_by_realtime_for_data(f
, dp
, j
->current_location
.realtime
, direction
, ret
, offset
);
615 return journal_file_next_entry_for_data(f
, NULL
, 0, dp
, direction
, ret
, offset
);
617 } else if (m
->type
== MATCH_OR_TERM
) {
622 /* Find the earliest match */
624 LIST_FOREACH(matches
, i
, m
->matches
) {
627 r
= find_location_for_match(j
, i
, f
, direction
, NULL
, &cp
);
631 if (np
== 0 || (direction
== DIRECTION_DOWN
? np
> cp
: np
< cp
))
639 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, np
, &n
);
654 assert(m
->type
== MATCH_AND_TERM
);
656 /* First jump to the last match, and then find the
657 * next one where all matches match */
662 LIST_FOREACH(matches
, i
, m
->matches
) {
665 r
= find_location_for_match(j
, i
, f
, direction
, NULL
, &cp
);
669 if (np
== 0 || (direction
== DIRECTION_DOWN
? cp
> np
: cp
< np
))
673 return next_for_match(j
, m
, f
, np
, direction
, ret
, offset
);
677 static int find_location_with_matches(
680 direction_t direction
,
692 /* No matches is simple */
694 if (j
->current_location
.type
== LOCATION_HEAD
)
695 return journal_file_next_entry(f
, 0, DIRECTION_DOWN
, ret
, offset
);
696 if (j
->current_location
.type
== LOCATION_TAIL
)
697 return journal_file_next_entry(f
, 0, DIRECTION_UP
, ret
, offset
);
698 if (j
->current_location
.seqnum_set
&& sd_id128_equal(j
->current_location
.seqnum_id
, f
->header
->seqnum_id
))
699 return journal_file_move_to_entry_by_seqnum(f
, j
->current_location
.seqnum
, direction
, ret
, offset
);
700 if (j
->current_location
.monotonic_set
) {
701 r
= journal_file_move_to_entry_by_monotonic(f
, j
->current_location
.boot_id
, j
->current_location
.monotonic
, direction
, ret
, offset
);
705 if (j
->current_location
.realtime_set
)
706 return journal_file_move_to_entry_by_realtime(f
, j
->current_location
.realtime
, direction
, ret
, offset
);
708 return journal_file_next_entry(f
, 0, direction
, ret
, offset
);
710 return find_location_for_match(j
, j
->level0
, f
, direction
, ret
, offset
);
713 static int next_with_matches(
716 direction_t direction
,
725 /* No matches is easy. We simple advance the file
728 return journal_file_next_entry(f
, f
->current_offset
, direction
, ret
, offset
);
730 /* If we have a match then we look for the next matching entry
731 * with an offset at least one step larger */
732 return next_for_match(j
, j
->level0
, f
,
733 direction
== DIRECTION_DOWN
? f
->current_offset
+ 1
734 : f
->current_offset
- 1,
735 direction
, ret
, offset
);
738 static int next_beyond_location(sd_journal
*j
, JournalFile
*f
, direction_t direction
) {
740 uint64_t cp
, n_entries
;
746 n_entries
= le64toh(f
->header
->n_entries
);
748 /* If we hit EOF before, we don't need to look into this file again
749 * unless direction changed or new entries appeared. */
750 if (f
->last_direction
== direction
&& f
->location_type
== LOCATION_TAIL
&&
751 n_entries
== f
->last_n_entries
)
754 f
->last_n_entries
= n_entries
;
756 if (f
->last_direction
== direction
&& f
->current_offset
> 0) {
757 /* LOCATION_SEEK here means we did the work in a previous
758 * iteration and the current location already points to a
759 * candidate entry. */
760 if (f
->location_type
!= LOCATION_SEEK
) {
761 r
= next_with_matches(j
, f
, direction
, &c
, &cp
);
765 journal_file_save_location(f
, c
, cp
);
768 f
->last_direction
= direction
;
770 r
= find_location_with_matches(j
, f
, direction
, &c
, &cp
);
774 journal_file_save_location(f
, c
, cp
);
777 /* OK, we found the spot, now let's advance until an entry
778 * that is actually different from what we were previously
779 * looking at. This is necessary to handle entries which exist
780 * in two (or more) journal files, and which shall all be
781 * suppressed but one. */
786 if (j
->current_location
.type
== LOCATION_DISCRETE
) {
789 k
= compare_with_location(f
, &j
->current_location
);
791 found
= direction
== DIRECTION_DOWN
? k
> 0 : k
< 0;
798 r
= next_with_matches(j
, f
, direction
, &c
, &cp
);
802 journal_file_save_location(f
, c
, cp
);
806 static int real_journal_next(sd_journal
*j
, direction_t direction
) {
807 JournalFile
*new_file
= NULL
;
813 assert_return(j
, -EINVAL
);
814 assert_return(!journal_pid_changed(j
), -ECHILD
);
816 r
= iterated_cache_get(j
->files_cache
, NULL
, &files
, &n_files
);
820 for (i
= 0; i
< n_files
; i
++) {
821 JournalFile
*f
= (JournalFile
*)files
[i
];
824 r
= next_beyond_location(j
, f
, direction
);
826 log_debug_errno(r
, "Can't iterate through %s, ignoring: %m", f
->path
);
827 remove_file_real(j
, f
);
830 f
->location_type
= LOCATION_TAIL
;
839 k
= journal_file_compare_locations(f
, new_file
);
841 found
= direction
== DIRECTION_DOWN
? k
< 0 : k
> 0;
851 r
= journal_file_move_to_object(new_file
, OBJECT_ENTRY
, new_file
->current_offset
, &o
);
855 set_location(j
, new_file
, o
);
860 _public_
int sd_journal_next(sd_journal
*j
) {
861 return real_journal_next(j
, DIRECTION_DOWN
);
864 _public_
int sd_journal_previous(sd_journal
*j
) {
865 return real_journal_next(j
, DIRECTION_UP
);
868 static int real_journal_next_skip(sd_journal
*j
, direction_t direction
, uint64_t skip
) {
871 assert_return(j
, -EINVAL
);
872 assert_return(!journal_pid_changed(j
), -ECHILD
);
875 /* If this is not a discrete skip, then at least
876 * resolve the current location */
877 if (j
->current_location
.type
!= LOCATION_DISCRETE
) {
878 r
= real_journal_next(j
, direction
);
887 r
= real_journal_next(j
, direction
);
901 _public_
int sd_journal_next_skip(sd_journal
*j
, uint64_t skip
) {
902 return real_journal_next_skip(j
, DIRECTION_DOWN
, skip
);
905 _public_
int sd_journal_previous_skip(sd_journal
*j
, uint64_t skip
) {
906 return real_journal_next_skip(j
, DIRECTION_UP
, skip
);
909 _public_
int sd_journal_get_cursor(sd_journal
*j
, char **cursor
) {
912 char bid
[33], sid
[33];
914 assert_return(j
, -EINVAL
);
915 assert_return(!journal_pid_changed(j
), -ECHILD
);
916 assert_return(cursor
, -EINVAL
);
918 if (!j
->current_file
|| j
->current_file
->current_offset
<= 0)
919 return -EADDRNOTAVAIL
;
921 r
= journal_file_move_to_object(j
->current_file
, OBJECT_ENTRY
, j
->current_file
->current_offset
, &o
);
925 sd_id128_to_string(j
->current_file
->header
->seqnum_id
, sid
);
926 sd_id128_to_string(o
->entry
.boot_id
, bid
);
929 "s=%s;i=%"PRIx64
";b=%s;m=%"PRIx64
";t=%"PRIx64
";x=%"PRIx64
,
930 sid
, le64toh(o
->entry
.seqnum
),
931 bid
, le64toh(o
->entry
.monotonic
),
932 le64toh(o
->entry
.realtime
),
933 le64toh(o
->entry
.xor_hash
)) < 0)
939 _public_
int sd_journal_seek_cursor(sd_journal
*j
, const char *cursor
) {
940 const char *word
, *state
;
942 unsigned long long seqnum
, monotonic
, realtime
, xor_hash
;
944 seqnum_id_set
= false,
947 monotonic_set
= false,
948 realtime_set
= false,
949 xor_hash_set
= false;
950 sd_id128_t seqnum_id
, boot_id
;
952 assert_return(j
, -EINVAL
);
953 assert_return(!journal_pid_changed(j
), -ECHILD
);
954 assert_return(!isempty(cursor
), -EINVAL
);
956 FOREACH_WORD_SEPARATOR(word
, l
, cursor
, ";", state
) {
960 if (l
< 2 || word
[1] != '=')
963 item
= strndup(word
, l
);
970 seqnum_id_set
= true;
971 k
= sd_id128_from_string(item
+2, &seqnum_id
);
976 if (sscanf(item
+2, "%llx", &seqnum
) != 1)
982 k
= sd_id128_from_string(item
+2, &boot_id
);
986 monotonic_set
= true;
987 if (sscanf(item
+2, "%llx", &monotonic
) != 1)
993 if (sscanf(item
+2, "%llx", &realtime
) != 1)
999 if (sscanf(item
+2, "%llx", &xor_hash
) != 1)
1010 if ((!seqnum_set
|| !seqnum_id_set
) &&
1011 (!monotonic_set
|| !boot_id_set
) &&
1017 j
->current_location
.type
= LOCATION_SEEK
;
1020 j
->current_location
.realtime
= (uint64_t) realtime
;
1021 j
->current_location
.realtime_set
= true;
1024 if (seqnum_set
&& seqnum_id_set
) {
1025 j
->current_location
.seqnum
= (uint64_t) seqnum
;
1026 j
->current_location
.seqnum_id
= seqnum_id
;
1027 j
->current_location
.seqnum_set
= true;
1030 if (monotonic_set
&& boot_id_set
) {
1031 j
->current_location
.monotonic
= (uint64_t) monotonic
;
1032 j
->current_location
.boot_id
= boot_id
;
1033 j
->current_location
.monotonic_set
= true;
1037 j
->current_location
.xor_hash
= (uint64_t) xor_hash
;
1038 j
->current_location
.xor_hash_set
= true;
1044 _public_
int sd_journal_test_cursor(sd_journal
*j
, const char *cursor
) {
1048 assert_return(j
, -EINVAL
);
1049 assert_return(!journal_pid_changed(j
), -ECHILD
);
1050 assert_return(!isempty(cursor
), -EINVAL
);
1052 if (!j
->current_file
|| j
->current_file
->current_offset
<= 0)
1053 return -EADDRNOTAVAIL
;
1055 r
= journal_file_move_to_object(j
->current_file
, OBJECT_ENTRY
, j
->current_file
->current_offset
, &o
);
1060 _cleanup_free_
char *item
= NULL
;
1061 unsigned long long ll
;
1065 r
= extract_first_word(&cursor
, &item
, ";", EXTRACT_DONT_COALESCE_SEPARATORS
);
1072 if (strlen(item
) < 2 || item
[1] != '=')
1078 k
= sd_id128_from_string(item
+2, &id
);
1081 if (!sd_id128_equal(id
, j
->current_file
->header
->seqnum_id
))
1086 if (sscanf(item
+2, "%llx", &ll
) != 1)
1088 if (ll
!= le64toh(o
->entry
.seqnum
))
1093 k
= sd_id128_from_string(item
+2, &id
);
1096 if (!sd_id128_equal(id
, o
->entry
.boot_id
))
1101 if (sscanf(item
+2, "%llx", &ll
) != 1)
1103 if (ll
!= le64toh(o
->entry
.monotonic
))
1108 if (sscanf(item
+2, "%llx", &ll
) != 1)
1110 if (ll
!= le64toh(o
->entry
.realtime
))
1115 if (sscanf(item
+2, "%llx", &ll
) != 1)
1117 if (ll
!= le64toh(o
->entry
.xor_hash
))
1126 _public_
int sd_journal_seek_monotonic_usec(sd_journal
*j
, sd_id128_t boot_id
, uint64_t usec
) {
1127 assert_return(j
, -EINVAL
);
1128 assert_return(!journal_pid_changed(j
), -ECHILD
);
1131 j
->current_location
.type
= LOCATION_SEEK
;
1132 j
->current_location
.boot_id
= boot_id
;
1133 j
->current_location
.monotonic
= usec
;
1134 j
->current_location
.monotonic_set
= true;
1139 _public_
int sd_journal_seek_realtime_usec(sd_journal
*j
, uint64_t usec
) {
1140 assert_return(j
, -EINVAL
);
1141 assert_return(!journal_pid_changed(j
), -ECHILD
);
1144 j
->current_location
.type
= LOCATION_SEEK
;
1145 j
->current_location
.realtime
= usec
;
1146 j
->current_location
.realtime_set
= true;
1151 _public_
int sd_journal_seek_head(sd_journal
*j
) {
1152 assert_return(j
, -EINVAL
);
1153 assert_return(!journal_pid_changed(j
), -ECHILD
);
1156 j
->current_location
.type
= LOCATION_HEAD
;
1161 _public_
int sd_journal_seek_tail(sd_journal
*j
) {
1162 assert_return(j
, -EINVAL
);
1163 assert_return(!journal_pid_changed(j
), -ECHILD
);
1166 j
->current_location
.type
= LOCATION_TAIL
;
1171 static void check_network(sd_journal
*j
, int fd
) {
1177 j
->on_network
= fd_is_network_fs(fd
);
1180 static bool file_has_type_prefix(const char *prefix
, const char *filename
) {
1181 const char *full
, *tilded
, *atted
;
1183 full
= strjoina(prefix
, ".journal");
1184 tilded
= strjoina(full
, "~");
1185 atted
= strjoina(prefix
, "@");
1187 return streq(filename
, full
) ||
1188 streq(filename
, tilded
) ||
1189 startswith(filename
, atted
);
1192 static bool file_type_wanted(int flags
, const char *filename
) {
1195 if (!endswith(filename
, ".journal") && !endswith(filename
, ".journal~"))
1198 /* no flags set → every type is OK */
1199 if (!(flags
& (SD_JOURNAL_SYSTEM
| SD_JOURNAL_CURRENT_USER
)))
1202 if (flags
& SD_JOURNAL_SYSTEM
&& file_has_type_prefix("system", filename
))
1205 if (flags
& SD_JOURNAL_CURRENT_USER
) {
1206 char prefix
[5 + DECIMAL_STR_MAX(uid_t
) + 1];
1208 xsprintf(prefix
, "user-"UID_FMT
, getuid());
1210 if (file_has_type_prefix(prefix
, filename
))
1217 static bool path_has_prefix(sd_journal
*j
, const char *path
, const char *prefix
) {
1222 if (j
->toplevel_fd
>= 0)
1225 return path_startswith(path
, prefix
);
1228 static void track_file_disposition(sd_journal
*j
, JournalFile
*f
) {
1232 if (!j
->has_runtime_files
&& path_has_prefix(j
, f
->path
, "/run"))
1233 j
->has_runtime_files
= true;
1234 else if (!j
->has_persistent_files
&& path_has_prefix(j
, f
->path
, "/var"))
1235 j
->has_persistent_files
= true;
1238 static const char *skip_slash(const char *p
) {
1249 static int add_any_file(
1254 bool close_fd
= false;
1260 assert(fd
>= 0 || path
);
1263 if (j
->toplevel_fd
>= 0)
1264 /* If there's a top-level fd defined make the path relative, explicitly, since otherwise
1265 * openat() ignores the first argument. */
1267 fd
= openat(j
->toplevel_fd
, skip_slash(path
), O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
);
1269 fd
= open(path
, O_RDONLY
|O_CLOEXEC
|O_NONBLOCK
);
1271 r
= log_debug_errno(errno
, "Failed to open journal file %s: %m", path
);
1277 r
= fd_nonblock(fd
, false);
1279 r
= log_debug_errno(errno
, "Failed to turn off O_NONBLOCK for %s: %m", path
);
1284 if (fstat(fd
, &st
) < 0) {
1285 r
= log_debug_errno(errno
, "Failed to fstat file '%s': %m", path
);
1289 r
= stat_verify_regular(&st
);
1291 log_debug_errno(r
, "Refusing to open '%s', as it is not a regular file.", path
);
1295 f
= ordered_hashmap_get(j
->files
, path
);
1297 if (f
->last_stat
.st_dev
== st
.st_dev
&&
1298 f
->last_stat
.st_ino
== st
.st_ino
) {
1300 /* We already track this file, under the same path and with the same device/inode numbers, it's
1301 * hence really the same. Mark this file as seen in this generation. This is used to GC old
1302 * files in process_q_overflow() to detect journal files that are still there and discern them
1303 * from those which are gone. */
1305 f
->last_seen_generation
= j
->generation
;
1310 /* So we tracked a file under this name, but it has a different inode/device. In that case, it got
1311 * replaced (probably due to rotation?), let's drop it hence from our list. */
1312 remove_file_real(j
, f
);
1316 if (ordered_hashmap_size(j
->files
) >= JOURNAL_FILES_MAX
) {
1317 log_debug("Too many open journal files, not adding %s.", path
);
1322 r
= journal_file_open(fd
, path
, O_RDONLY
, 0, false, 0, false, NULL
, j
->mmap
, NULL
, NULL
, &f
);
1324 log_debug_errno(r
, "Failed to open journal file %s: %m", path
);
1328 /* journal_file_dump(f); */
1330 r
= ordered_hashmap_put(j
->files
, f
->path
, f
);
1332 f
->close_fd
= false; /* make sure journal_file_close() doesn't close the caller's fd (or our own). We'll let the caller do that, or ourselves */
1333 (void) journal_file_close(f
);
1337 close_fd
= false; /* the fd is now owned by the JournalFile object */
1339 f
->last_seen_generation
= j
->generation
;
1341 track_file_disposition(j
, f
);
1342 check_network(j
, f
->fd
);
1344 j
->current_invalidate_counter
++;
1346 log_debug("File %s added.", f
->path
);
1355 k
= journal_put_error(j
, r
, path
);
1363 static int add_file_by_name(
1366 const char *filename
) {
1374 if (j
->no_new_files
)
1377 if (!file_type_wanted(j
->flags
, filename
))
1380 path
= strjoina(prefix
, "/", filename
);
1381 return add_any_file(j
, -1, path
);
1384 static void remove_file_by_name(
1387 const char *filename
) {
1396 path
= strjoina(prefix
, "/", filename
);
1397 f
= ordered_hashmap_get(j
->files
, path
);
1401 remove_file_real(j
, f
);
1404 static void remove_file_real(sd_journal
*j
, JournalFile
*f
) {
1408 (void) ordered_hashmap_remove(j
->files
, f
->path
);
1410 log_debug("File %s removed.", f
->path
);
1412 if (j
->current_file
== f
) {
1413 j
->current_file
= NULL
;
1414 j
->current_field
= 0;
1417 if (j
->unique_file
== f
) {
1418 /* Jump to the next unique_file or NULL if that one was last */
1419 j
->unique_file
= ordered_hashmap_next(j
->files
, j
->unique_file
->path
);
1420 j
->unique_offset
= 0;
1421 if (!j
->unique_file
)
1422 j
->unique_file_lost
= true;
1425 if (j
->fields_file
== f
) {
1426 j
->fields_file
= ordered_hashmap_next(j
->files
, j
->fields_file
->path
);
1427 j
->fields_offset
= 0;
1428 if (!j
->fields_file
)
1429 j
->fields_file_lost
= true;
1432 (void) journal_file_close(f
);
1434 j
->current_invalidate_counter
++;
1437 static int dirname_is_machine_id(const char *fn
) {
1438 sd_id128_t id
, machine
;
1441 r
= sd_id128_get_machine(&machine
);
1445 r
= sd_id128_from_string(fn
, &id
);
1449 return sd_id128_equal(id
, machine
);
1452 static bool dirent_is_journal_file(const struct dirent
*de
) {
1455 if (!IN_SET(de
->d_type
, DT_REG
, DT_LNK
, DT_UNKNOWN
))
1458 return endswith(de
->d_name
, ".journal") ||
1459 endswith(de
->d_name
, ".journal~");
1462 static bool dirent_is_id128_subdir(const struct dirent
*de
) {
1465 if (!IN_SET(de
->d_type
, DT_DIR
, DT_LNK
, DT_UNKNOWN
))
1468 return id128_is_valid(de
->d_name
);
1471 static int directory_open(sd_journal
*j
, const char *path
, DIR **ret
) {
1478 if (j
->toplevel_fd
< 0)
1481 /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is
1482 * relative, by dropping the initial slash */
1483 d
= xopendirat(j
->toplevel_fd
, skip_slash(path
), 0);
1491 static int add_directory(sd_journal
*j
, const char *prefix
, const char *dirname
);
1493 static void directory_enumerate(sd_journal
*j
, Directory
*m
, DIR *d
) {
1500 FOREACH_DIRENT_ALL(de
, d
, goto fail
) {
1502 if (dirent_is_journal_file(de
))
1503 (void) add_file_by_name(j
, m
->path
, de
->d_name
);
1505 if (m
->is_root
&& dirent_is_id128_subdir(de
))
1506 (void) add_directory(j
, m
->path
, de
->d_name
);
1512 log_debug_errno(errno
, "Failed to enumerate directory %s, ignoring: %m", m
->path
);
1515 static void directory_watch(sd_journal
*j
, Directory
*m
, int fd
, uint32_t mask
) {
1522 /* Watch this directory if that's enabled and if it not being watched yet. */
1524 if (m
->wd
> 0) /* Already have a watch? */
1526 if (j
->inotify_fd
< 0) /* Not watching at all? */
1529 m
->wd
= inotify_add_watch_fd(j
->inotify_fd
, fd
, mask
);
1531 log_debug_errno(errno
, "Failed to watch journal directory '%s', ignoring: %m", m
->path
);
1535 r
= hashmap_put(j
->directories_by_wd
, INT_TO_PTR(m
->wd
), m
);
1537 log_debug_errno(r
, "Directory '%s' already being watched under a different path, ignoring: %m", m
->path
);
1539 log_debug_errno(r
, "Failed to add watch for journal directory '%s' to hashmap, ignoring: %m", m
->path
);
1540 (void) inotify_rm_watch(j
->inotify_fd
, m
->wd
);
1545 static int add_directory(sd_journal
*j
, const char *prefix
, const char *dirname
) {
1546 _cleanup_free_
char *path
= NULL
;
1547 _cleanup_closedir_
DIR *d
= NULL
;
1554 /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch
1555 * and reenumerates directory contents */
1558 path
= strjoin(prefix
, "/", dirname
);
1560 path
= strdup(prefix
);
1566 log_debug("Considering directory '%s'.", path
);
1568 /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */
1569 if ((j
->flags
& SD_JOURNAL_LOCAL_ONLY
) &&
1570 !((dirname
&& dirname_is_machine_id(dirname
) > 0) || path_has_prefix(j
, path
, "/run")))
1573 r
= directory_open(j
, path
, &d
);
1575 log_debug_errno(r
, "Failed to open directory '%s': %m", path
);
1579 m
= hashmap_get(j
->directories_by_path
, path
);
1581 m
= new0(Directory
, 1);
1590 if (hashmap_put(j
->directories_by_path
, m
->path
, m
) < 0) {
1596 path
= NULL
; /* avoid freeing in cleanup */
1597 j
->current_invalidate_counter
++;
1599 log_debug("Directory %s added.", m
->path
);
1601 } else if (m
->is_root
)
1602 return 0; /* Don't 'downgrade' from root directory */
1604 m
->last_seen_generation
= j
->generation
;
1606 directory_watch(j
, m
, dirfd(d
),
1607 IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
|IN_DELETE
|
1608 IN_DELETE_SELF
|IN_MOVE_SELF
|IN_UNMOUNT
|IN_MOVED_FROM
|
1611 if (!j
->no_new_files
)
1612 directory_enumerate(j
, m
, d
);
1614 check_network(j
, dirfd(d
));
1619 k
= journal_put_error(j
, r
, path
?: prefix
);
1626 static int add_root_directory(sd_journal
*j
, const char *p
, bool missing_ok
) {
1628 _cleanup_closedir_
DIR *d
= NULL
;
1634 /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we
1635 * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially
1636 * populate the set, as well as to update it later. */
1639 /* If there's a path specified, use it. */
1641 log_debug("Considering root directory '%s'.", p
);
1643 if ((j
->flags
& SD_JOURNAL_RUNTIME_ONLY
) &&
1644 !path_has_prefix(j
, p
, "/run"))
1648 p
= strjoina(j
->prefix
, p
);
1650 r
= directory_open(j
, p
, &d
);
1651 if (r
== -ENOENT
&& missing_ok
)
1654 log_debug_errno(r
, "Failed to open root directory %s: %m", p
);
1660 /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since
1661 * opendir() will take possession of the fd, and close it, which we don't want. */
1663 p
= "."; /* store this as "." in the directories hashmap */
1665 dfd
= fcntl(j
->toplevel_fd
, F_DUPFD_CLOEXEC
, 3);
1681 m
= hashmap_get(j
->directories_by_path
, p
);
1683 m
= new0(Directory
, 1);
1691 m
->path
= strdup(p
);
1698 if (hashmap_put(j
->directories_by_path
, m
->path
, m
) < 0) {
1705 j
->current_invalidate_counter
++;
1707 log_debug("Root directory %s added.", m
->path
);
1709 } else if (!m
->is_root
)
1712 directory_watch(j
, m
, dirfd(d
),
1713 IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
|IN_DELETE
|
1716 if (!j
->no_new_files
)
1717 directory_enumerate(j
, m
, d
);
1719 check_network(j
, dirfd(d
));
1724 k
= journal_put_error(j
, r
, p
);
1731 static void remove_directory(sd_journal
*j
, Directory
*d
) {
1735 hashmap_remove(j
->directories_by_wd
, INT_TO_PTR(d
->wd
));
1737 if (j
->inotify_fd
>= 0)
1738 inotify_rm_watch(j
->inotify_fd
, d
->wd
);
1741 hashmap_remove(j
->directories_by_path
, d
->path
);
1744 log_debug("Root directory %s removed.", d
->path
);
1746 log_debug("Directory %s removed.", d
->path
);
1752 static int add_search_paths(sd_journal
*j
) {
1754 static const char search_paths
[] =
1755 "/run/log/journal\0"
1756 "/var/log/journal\0";
1761 /* We ignore most errors here, since the idea is to only open
1762 * what's actually accessible, and ignore the rest. */
1764 NULSTR_FOREACH(p
, search_paths
)
1765 (void) add_root_directory(j
, p
, true);
1767 if (!(j
->flags
& SD_JOURNAL_LOCAL_ONLY
))
1768 (void) add_root_directory(j
, "/var/log/journal/remote", true);
1773 static int add_current_paths(sd_journal
*j
) {
1778 assert(j
->no_new_files
);
1780 /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we
1781 * treat them as fatal. */
1783 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
1784 _cleanup_free_
char *dir
;
1787 dir
= dirname_malloc(f
->path
);
1791 r
= add_directory(j
, dir
, NULL
);
1799 static int allocate_inotify(sd_journal
*j
) {
1802 if (j
->inotify_fd
< 0) {
1803 j
->inotify_fd
= inotify_init1(IN_NONBLOCK
|IN_CLOEXEC
);
1804 if (j
->inotify_fd
< 0)
1808 return hashmap_ensure_allocated(&j
->directories_by_wd
, NULL
);
1811 static sd_journal
*journal_new(int flags
, const char *path
) {
1812 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1814 j
= new0(sd_journal
, 1);
1818 j
->original_pid
= getpid_cached();
1819 j
->toplevel_fd
= -1;
1822 j
->data_threshold
= DEFAULT_DATA_THRESHOLD
;
1831 if (flags
& SD_JOURNAL_OS_ROOT
)
1837 j
->files
= ordered_hashmap_new(&path_hash_ops
);
1841 j
->files_cache
= ordered_hashmap_iterated_cache_new(j
->files
);
1842 j
->directories_by_path
= hashmap_new(&path_hash_ops
);
1843 j
->mmap
= mmap_cache_new();
1844 if (!j
->files_cache
|| !j
->directories_by_path
|| !j
->mmap
)
1850 #define OPEN_ALLOWED_FLAGS \
1851 (SD_JOURNAL_LOCAL_ONLY | \
1852 SD_JOURNAL_RUNTIME_ONLY | \
1853 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)
1855 _public_
int sd_journal_open(sd_journal
**ret
, int flags
) {
1856 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1859 assert_return(ret
, -EINVAL
);
1860 assert_return((flags
& ~OPEN_ALLOWED_FLAGS
) == 0, -EINVAL
);
1862 j
= journal_new(flags
, NULL
);
1866 r
= add_search_paths(j
);
1874 #define OPEN_CONTAINER_ALLOWED_FLAGS \
1875 (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM)
1877 _public_
int sd_journal_open_container(sd_journal
**ret
, const char *machine
, int flags
) {
1878 _cleanup_free_
char *root
= NULL
, *class = NULL
;
1879 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1883 /* This is pretty much deprecated, people should use machined's OpenMachineRootDirectory() call instead in
1884 * combination with sd_journal_open_directory_fd(). */
1886 assert_return(machine
, -EINVAL
);
1887 assert_return(ret
, -EINVAL
);
1888 assert_return((flags
& ~OPEN_CONTAINER_ALLOWED_FLAGS
) == 0, -EINVAL
);
1889 assert_return(machine_name_is_valid(machine
), -EINVAL
);
1891 p
= strjoina("/run/systemd/machines/", machine
);
1892 r
= parse_env_file(NULL
, p
, NEWLINE
, "ROOT", &root
, "CLASS", &class, NULL
);
1900 if (!streq_ptr(class, "container"))
1903 j
= journal_new(flags
, root
);
1907 r
= add_search_paths(j
);
1915 #define OPEN_DIRECTORY_ALLOWED_FLAGS \
1916 (SD_JOURNAL_OS_ROOT | \
1917 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1919 _public_
int sd_journal_open_directory(sd_journal
**ret
, const char *path
, int flags
) {
1920 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1923 assert_return(ret
, -EINVAL
);
1924 assert_return(path
, -EINVAL
);
1925 assert_return((flags
& ~OPEN_DIRECTORY_ALLOWED_FLAGS
) == 0, -EINVAL
);
1927 j
= journal_new(flags
, path
);
1931 if (flags
& SD_JOURNAL_OS_ROOT
)
1932 r
= add_search_paths(j
);
1934 r
= add_root_directory(j
, path
, false);
1942 _public_
int sd_journal_open_files(sd_journal
**ret
, const char **paths
, int flags
) {
1943 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1947 assert_return(ret
, -EINVAL
);
1948 assert_return(flags
== 0, -EINVAL
);
1950 j
= journal_new(flags
, NULL
);
1954 STRV_FOREACH(path
, paths
) {
1955 r
= add_any_file(j
, -1, *path
);
1960 j
->no_new_files
= true;
1966 #define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \
1967 (SD_JOURNAL_OS_ROOT | \
1968 SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1970 _public_
int sd_journal_open_directory_fd(sd_journal
**ret
, int fd
, int flags
) {
1971 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
1975 assert_return(ret
, -EINVAL
);
1976 assert_return(fd
>= 0, -EBADF
);
1977 assert_return((flags
& ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS
) == 0, -EINVAL
);
1979 if (fstat(fd
, &st
) < 0)
1982 if (!S_ISDIR(st
.st_mode
))
1985 j
= journal_new(flags
, NULL
);
1989 j
->toplevel_fd
= fd
;
1991 if (flags
& SD_JOURNAL_OS_ROOT
)
1992 r
= add_search_paths(j
);
1994 r
= add_root_directory(j
, NULL
, false);
2002 _public_
int sd_journal_open_files_fd(sd_journal
**ret
, int fds
[], unsigned n_fds
, int flags
) {
2005 _cleanup_(sd_journal_closep
) sd_journal
*j
= NULL
;
2009 assert_return(ret
, -EINVAL
);
2010 assert_return(n_fds
> 0, -EBADF
);
2011 assert_return(flags
== 0, -EINVAL
);
2013 j
= journal_new(flags
, NULL
);
2017 for (i
= 0; i
< n_fds
; i
++) {
2025 if (fstat(fds
[i
], &st
) < 0) {
2030 r
= stat_verify_regular(&st
);
2034 r
= add_any_file(j
, fds
[i
], NULL
);
2039 j
->no_new_files
= true;
2040 j
->no_inotify
= true;
2046 /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they
2048 ORDERED_HASHMAP_FOREACH(f
, j
->files
, iterator
)
2049 f
->close_fd
= false;
2054 _public_
void sd_journal_close(sd_journal
*j
) {
2060 sd_journal_flush_matches(j
);
2062 ordered_hashmap_free_with_destructor(j
->files
, journal_file_close
);
2063 iterated_cache_free(j
->files_cache
);
2065 while ((d
= hashmap_first(j
->directories_by_path
)))
2066 remove_directory(j
, d
);
2068 while ((d
= hashmap_first(j
->directories_by_wd
)))
2069 remove_directory(j
, d
);
2071 hashmap_free(j
->directories_by_path
);
2072 hashmap_free(j
->directories_by_wd
);
2074 safe_close(j
->inotify_fd
);
2077 log_debug("mmap cache statistics: %u hit, %u miss", mmap_cache_get_hit(j
->mmap
), mmap_cache_get_missed(j
->mmap
));
2078 mmap_cache_unref(j
->mmap
);
2081 hashmap_free_free(j
->errors
);
2085 free(j
->unique_field
);
2086 free(j
->fields_buffer
);
2090 _public_
int sd_journal_get_realtime_usec(sd_journal
*j
, uint64_t *ret
) {
2095 assert_return(j
, -EINVAL
);
2096 assert_return(!journal_pid_changed(j
), -ECHILD
);
2097 assert_return(ret
, -EINVAL
);
2099 f
= j
->current_file
;
2101 return -EADDRNOTAVAIL
;
2103 if (f
->current_offset
<= 0)
2104 return -EADDRNOTAVAIL
;
2106 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2110 *ret
= le64toh(o
->entry
.realtime
);
2114 _public_
int sd_journal_get_monotonic_usec(sd_journal
*j
, uint64_t *ret
, sd_id128_t
*ret_boot_id
) {
2120 assert_return(j
, -EINVAL
);
2121 assert_return(!journal_pid_changed(j
), -ECHILD
);
2123 f
= j
->current_file
;
2125 return -EADDRNOTAVAIL
;
2127 if (f
->current_offset
<= 0)
2128 return -EADDRNOTAVAIL
;
2130 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2135 *ret_boot_id
= o
->entry
.boot_id
;
2137 r
= sd_id128_get_boot(&id
);
2141 if (!sd_id128_equal(id
, o
->entry
.boot_id
))
2146 *ret
= le64toh(o
->entry
.monotonic
);
2151 static bool field_is_valid(const char *field
) {
2159 if (startswith(field
, "__"))
2162 for (p
= field
; *p
; p
++) {
2167 if (*p
>= 'A' && *p
<= 'Z')
2170 if (*p
>= '0' && *p
<= '9')
2179 _public_
int sd_journal_get_data(sd_journal
*j
, const char *field
, const void **data
, size_t *size
) {
2182 size_t field_length
;
2186 assert_return(j
, -EINVAL
);
2187 assert_return(!journal_pid_changed(j
), -ECHILD
);
2188 assert_return(field
, -EINVAL
);
2189 assert_return(data
, -EINVAL
);
2190 assert_return(size
, -EINVAL
);
2191 assert_return(field_is_valid(field
), -EINVAL
);
2193 f
= j
->current_file
;
2195 return -EADDRNOTAVAIL
;
2197 if (f
->current_offset
<= 0)
2198 return -EADDRNOTAVAIL
;
2200 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2204 field_length
= strlen(field
);
2206 n
= journal_file_entry_n_items(o
);
2207 for (i
= 0; i
< n
; i
++) {
2213 p
= le64toh(o
->entry
.items
[i
].object_offset
);
2214 le_hash
= o
->entry
.items
[i
].hash
;
2215 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
2219 if (le_hash
!= o
->data
.hash
)
2222 l
= le64toh(o
->object
.size
) - offsetof(Object
, data
.payload
);
2224 compression
= o
->object
.flags
& OBJECT_COMPRESSION_MASK
;
2226 #if HAVE_XZ || HAVE_LZ4
2227 r
= decompress_startswith(compression
,
2229 &f
->compress_buffer
, &f
->compress_buffer_size
,
2230 field
, field_length
, '=');
2232 log_debug_errno(r
, "Cannot decompress %s object of length %"PRIu64
" at offset "OFSfmt
": %m",
2233 object_compressed_to_string(compression
), l
, p
);
2238 r
= decompress_blob(compression
,
2240 &f
->compress_buffer
, &f
->compress_buffer_size
, &rsize
,
2245 *data
= f
->compress_buffer
;
2246 *size
= (size_t) rsize
;
2251 return -EPROTONOSUPPORT
;
2253 } else if (l
>= field_length
+1 &&
2254 memcmp(o
->data
.payload
, field
, field_length
) == 0 &&
2255 o
->data
.payload
[field_length
] == '=') {
2259 if ((uint64_t) t
!= l
)
2262 *data
= o
->data
.payload
;
2268 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2276 static int return_data(sd_journal
*j
, JournalFile
*f
, Object
*o
, const void **data
, size_t *size
) {
2281 l
= le64toh(o
->object
.size
) - offsetof(Object
, data
.payload
);
2284 /* We can't read objects larger than 4G on a 32bit machine */
2285 if ((uint64_t) t
!= l
)
2288 compression
= o
->object
.flags
& OBJECT_COMPRESSION_MASK
;
2290 #if HAVE_XZ || HAVE_LZ4
2294 r
= decompress_blob(compression
,
2295 o
->data
.payload
, l
, &f
->compress_buffer
,
2296 &f
->compress_buffer_size
, &rsize
, j
->data_threshold
);
2300 *data
= f
->compress_buffer
;
2301 *size
= (size_t) rsize
;
2303 return -EPROTONOSUPPORT
;
2306 *data
= o
->data
.payload
;
2313 _public_
int sd_journal_enumerate_data(sd_journal
*j
, const void **data
, size_t *size
) {
2320 assert_return(j
, -EINVAL
);
2321 assert_return(!journal_pid_changed(j
), -ECHILD
);
2322 assert_return(data
, -EINVAL
);
2323 assert_return(size
, -EINVAL
);
2325 f
= j
->current_file
;
2327 return -EADDRNOTAVAIL
;
2329 if (f
->current_offset
<= 0)
2330 return -EADDRNOTAVAIL
;
2332 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, f
->current_offset
, &o
);
2336 n
= journal_file_entry_n_items(o
);
2337 if (j
->current_field
>= n
)
2340 p
= le64toh(o
->entry
.items
[j
->current_field
].object_offset
);
2341 le_hash
= o
->entry
.items
[j
->current_field
].hash
;
2342 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
2346 if (le_hash
!= o
->data
.hash
)
2349 r
= return_data(j
, f
, o
, data
, size
);
2358 _public_
void sd_journal_restart_data(sd_journal
*j
) {
2362 j
->current_field
= 0;
2365 static int reiterate_all_paths(sd_journal
*j
) {
2368 if (j
->no_new_files
)
2369 return add_current_paths(j
);
2371 if (j
->flags
& SD_JOURNAL_OS_ROOT
)
2372 return add_search_paths(j
);
2374 if (j
->toplevel_fd
>= 0)
2375 return add_root_directory(j
, NULL
, false);
2378 return add_root_directory(j
, j
->path
, true);
2380 return add_search_paths(j
);
2383 _public_
int sd_journal_get_fd(sd_journal
*j
) {
2386 assert_return(j
, -EINVAL
);
2387 assert_return(!journal_pid_changed(j
), -ECHILD
);
2390 return -EMEDIUMTYPE
;
2392 if (j
->inotify_fd
>= 0)
2393 return j
->inotify_fd
;
2395 r
= allocate_inotify(j
);
2399 log_debug("Reiterating files to get inotify watches established.");
2401 /* Iterate through all dirs again, to add them to the inotify */
2402 r
= reiterate_all_paths(j
);
2406 return j
->inotify_fd
;
2409 _public_
int sd_journal_get_events(sd_journal
*j
) {
2412 assert_return(j
, -EINVAL
);
2413 assert_return(!journal_pid_changed(j
), -ECHILD
);
2415 fd
= sd_journal_get_fd(j
);
2422 _public_
int sd_journal_get_timeout(sd_journal
*j
, uint64_t *timeout_usec
) {
2425 assert_return(j
, -EINVAL
);
2426 assert_return(!journal_pid_changed(j
), -ECHILD
);
2427 assert_return(timeout_usec
, -EINVAL
);
2429 fd
= sd_journal_get_fd(j
);
2433 if (!j
->on_network
) {
2434 *timeout_usec
= (uint64_t) -1;
2438 /* If we are on the network we need to regularly check for
2439 * changes manually */
2441 *timeout_usec
= j
->last_process_usec
+ JOURNAL_FILES_RECHECK_USEC
;
2445 static void process_q_overflow(sd_journal
*j
) {
2452 /* When the inotify queue overruns we need to enumerate and re-validate all journal files to bring our list
2453 * back in sync with what's on disk. For this we pick a new generation counter value. It'll be assigned to all
2454 * journal files we encounter. All journal files and all directories that don't carry it after reenumeration
2455 * are subject for unloading. */
2457 log_debug("Inotify queue overrun, reiterating everything.");
2460 (void) reiterate_all_paths(j
);
2462 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2464 if (f
->last_seen_generation
== j
->generation
)
2467 log_debug("File '%s' hasn't been seen in this enumeration, removing.", f
->path
);
2468 remove_file_real(j
, f
);
2471 HASHMAP_FOREACH(m
, j
->directories_by_path
, i
) {
2473 if (m
->last_seen_generation
== j
->generation
)
2476 if (m
->is_root
) /* Never GC root directories */
2479 log_debug("Directory '%s' hasn't been seen in this enumeration, removing.", f
->path
);
2480 remove_directory(j
, m
);
2483 log_debug("Reiteration complete.");
2486 static void process_inotify_event(sd_journal
*j
, struct inotify_event
*e
) {
2492 if (e
->mask
& IN_Q_OVERFLOW
) {
2493 process_q_overflow(j
);
2497 /* Is this a subdirectory we watch? */
2498 d
= hashmap_get(j
->directories_by_wd
, INT_TO_PTR(e
->wd
));
2500 if (!(e
->mask
& IN_ISDIR
) && e
->len
> 0 &&
2501 (endswith(e
->name
, ".journal") ||
2502 endswith(e
->name
, ".journal~"))) {
2504 /* Event for a journal file */
2506 if (e
->mask
& (IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
))
2507 (void) add_file_by_name(j
, d
->path
, e
->name
);
2508 else if (e
->mask
& (IN_DELETE
|IN_MOVED_FROM
|IN_UNMOUNT
))
2509 remove_file_by_name(j
, d
->path
, e
->name
);
2511 } else if (!d
->is_root
&& e
->len
== 0) {
2513 /* Event for a subdirectory */
2515 if (e
->mask
& (IN_DELETE_SELF
|IN_MOVE_SELF
|IN_UNMOUNT
))
2516 remove_directory(j
, d
);
2518 } else if (d
->is_root
&& (e
->mask
& IN_ISDIR
) && e
->len
> 0 && id128_is_valid(e
->name
)) {
2520 /* Event for root directory */
2522 if (e
->mask
& (IN_CREATE
|IN_MOVED_TO
|IN_MODIFY
|IN_ATTRIB
))
2523 (void) add_directory(j
, d
->path
, e
->name
);
2529 if (e
->mask
& IN_IGNORED
)
2532 log_debug("Unexpected inotify event.");
2535 static int determine_change(sd_journal
*j
) {
2540 b
= j
->current_invalidate_counter
!= j
->last_invalidate_counter
;
2541 j
->last_invalidate_counter
= j
->current_invalidate_counter
;
2543 return b
? SD_JOURNAL_INVALIDATE
: SD_JOURNAL_APPEND
;
2546 _public_
int sd_journal_process(sd_journal
*j
) {
2547 bool got_something
= false;
2549 assert_return(j
, -EINVAL
);
2550 assert_return(!journal_pid_changed(j
), -ECHILD
);
2552 if (j
->inotify_fd
< 0) /* We have no inotify fd yet? Then there's noting to process. */
2555 j
->last_process_usec
= now(CLOCK_MONOTONIC
);
2556 j
->last_invalidate_counter
= j
->current_invalidate_counter
;
2559 union inotify_event_buffer buffer
;
2560 struct inotify_event
*e
;
2563 l
= read(j
->inotify_fd
, &buffer
, sizeof(buffer
));
2565 if (IN_SET(errno
, EAGAIN
, EINTR
))
2566 return got_something
? determine_change(j
) : SD_JOURNAL_NOP
;
2571 got_something
= true;
2573 FOREACH_INOTIFY_EVENT(e
, buffer
, l
)
2574 process_inotify_event(j
, e
);
2578 _public_
int sd_journal_wait(sd_journal
*j
, uint64_t timeout_usec
) {
2582 assert_return(j
, -EINVAL
);
2583 assert_return(!journal_pid_changed(j
), -ECHILD
);
2585 if (j
->inotify_fd
< 0) {
2587 /* This is the first invocation, hence create the
2589 r
= sd_journal_get_fd(j
);
2593 /* The journal might have changed since the context
2594 * object was created and we weren't watching before,
2595 * hence don't wait for anything, and return
2597 return determine_change(j
);
2600 r
= sd_journal_get_timeout(j
, &t
);
2604 if (t
!= (uint64_t) -1) {
2607 n
= now(CLOCK_MONOTONIC
);
2608 t
= t
> n
? t
- n
: 0;
2610 if (timeout_usec
== (uint64_t) -1 || timeout_usec
> t
)
2615 r
= fd_wait_for_event(j
->inotify_fd
, POLLIN
, timeout_usec
);
2616 } while (r
== -EINTR
);
2621 return sd_journal_process(j
);
2624 _public_
int sd_journal_get_cutoff_realtime_usec(sd_journal
*j
, uint64_t *from
, uint64_t *to
) {
2628 uint64_t fmin
= 0, tmax
= 0;
2631 assert_return(j
, -EINVAL
);
2632 assert_return(!journal_pid_changed(j
), -ECHILD
);
2633 assert_return(from
|| to
, -EINVAL
);
2634 assert_return(from
!= to
, -EINVAL
);
2636 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2639 r
= journal_file_get_cutoff_realtime_usec(f
, &fr
, &t
);
2652 fmin
= MIN(fr
, fmin
);
2653 tmax
= MAX(t
, tmax
);
2662 return first
? 0 : 1;
2665 _public_
int sd_journal_get_cutoff_monotonic_usec(sd_journal
*j
, sd_id128_t boot_id
, uint64_t *from
, uint64_t *to
) {
2671 assert_return(j
, -EINVAL
);
2672 assert_return(!journal_pid_changed(j
), -ECHILD
);
2673 assert_return(from
|| to
, -EINVAL
);
2674 assert_return(from
!= to
, -EINVAL
);
2676 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2679 r
= journal_file_get_cutoff_monotonic_usec(f
, boot_id
, &fr
, &t
);
2689 *from
= MIN(fr
, *from
);
2704 void journal_print_header(sd_journal
*j
) {
2707 bool newline
= false;
2711 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2717 journal_file_print_header(f
);
2721 _public_
int sd_journal_get_usage(sd_journal
*j
, uint64_t *bytes
) {
2726 assert_return(j
, -EINVAL
);
2727 assert_return(!journal_pid_changed(j
), -ECHILD
);
2728 assert_return(bytes
, -EINVAL
);
2730 ORDERED_HASHMAP_FOREACH(f
, j
->files
, i
) {
2733 if (fstat(f
->fd
, &st
) < 0)
2736 sum
+= (uint64_t) st
.st_blocks
* 512ULL;
2743 _public_
int sd_journal_query_unique(sd_journal
*j
, const char *field
) {
2746 assert_return(j
, -EINVAL
);
2747 assert_return(!journal_pid_changed(j
), -ECHILD
);
2748 assert_return(!isempty(field
), -EINVAL
);
2749 assert_return(field_is_valid(field
), -EINVAL
);
2755 free(j
->unique_field
);
2756 j
->unique_field
= f
;
2757 j
->unique_file
= NULL
;
2758 j
->unique_offset
= 0;
2759 j
->unique_file_lost
= false;
2764 _public_
int sd_journal_enumerate_unique(sd_journal
*j
, const void **data
, size_t *l
) {
2767 assert_return(j
, -EINVAL
);
2768 assert_return(!journal_pid_changed(j
), -ECHILD
);
2769 assert_return(data
, -EINVAL
);
2770 assert_return(l
, -EINVAL
);
2771 assert_return(j
->unique_field
, -EINVAL
);
2773 k
= strlen(j
->unique_field
);
2775 if (!j
->unique_file
) {
2776 if (j
->unique_file_lost
)
2779 j
->unique_file
= ordered_hashmap_first(j
->files
);
2780 if (!j
->unique_file
)
2783 j
->unique_offset
= 0;
2795 /* Proceed to next data object in the field's linked list */
2796 if (j
->unique_offset
== 0) {
2797 r
= journal_file_find_field_object(j
->unique_file
, j
->unique_field
, k
, &o
, NULL
);
2801 j
->unique_offset
= r
> 0 ? le64toh(o
->field
.head_data_offset
) : 0;
2803 r
= journal_file_move_to_object(j
->unique_file
, OBJECT_DATA
, j
->unique_offset
, &o
);
2807 j
->unique_offset
= le64toh(o
->data
.next_field_offset
);
2810 /* We reached the end of the list? Then start again, with the next file */
2811 if (j
->unique_offset
== 0) {
2812 j
->unique_file
= ordered_hashmap_next(j
->files
, j
->unique_file
->path
);
2813 if (!j
->unique_file
)
2819 /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED
2820 * instead, so that we can look at this data object at the same
2821 * time as one on another file */
2822 r
= journal_file_move_to_object(j
->unique_file
, OBJECT_UNUSED
, j
->unique_offset
, &o
);
2826 /* Let's do the type check by hand, since we used 0 context above. */
2827 if (o
->object
.type
!= OBJECT_DATA
) {
2828 log_debug("%s:offset " OFSfmt
": object has type %d, expected %d",
2829 j
->unique_file
->path
, j
->unique_offset
,
2830 o
->object
.type
, OBJECT_DATA
);
2834 r
= return_data(j
, j
->unique_file
, o
, &odata
, &ol
);
2838 /* Check if we have at least the field name and "=". */
2840 log_debug("%s:offset " OFSfmt
": object has size %zu, expected at least %zu",
2841 j
->unique_file
->path
, j
->unique_offset
,
2846 if (memcmp(odata
, j
->unique_field
, k
) || ((const char*) odata
)[k
] != '=') {
2847 log_debug("%s:offset " OFSfmt
": object does not start with \"%s=\"",
2848 j
->unique_file
->path
, j
->unique_offset
,
2853 /* OK, now let's see if we already returned this data
2854 * object by checking if it exists in the earlier
2855 * traversed files. */
2857 ORDERED_HASHMAP_FOREACH(of
, j
->files
, i
) {
2858 if (of
== j
->unique_file
)
2861 /* Skip this file it didn't have any fields indexed */
2862 if (JOURNAL_HEADER_CONTAINS(of
->header
, n_fields
) && le64toh(of
->header
->n_fields
) <= 0)
2865 r
= journal_file_find_data_object_with_hash(of
, odata
, ol
, le64toh(o
->data
.hash
), NULL
, NULL
);
2877 r
= return_data(j
, j
->unique_file
, o
, data
, l
);
2885 _public_
void sd_journal_restart_unique(sd_journal
*j
) {
2889 j
->unique_file
= NULL
;
2890 j
->unique_offset
= 0;
2891 j
->unique_file_lost
= false;
2894 _public_
int sd_journal_enumerate_fields(sd_journal
*j
, const char **field
) {
2897 assert_return(j
, -EINVAL
);
2898 assert_return(!journal_pid_changed(j
), -ECHILD
);
2899 assert_return(field
, -EINVAL
);
2901 if (!j
->fields_file
) {
2902 if (j
->fields_file_lost
)
2905 j
->fields_file
= ordered_hashmap_first(j
->files
);
2906 if (!j
->fields_file
)
2909 j
->fields_hash_table_index
= 0;
2910 j
->fields_offset
= 0;
2914 JournalFile
*f
, *of
;
2923 if (j
->fields_offset
== 0) {
2926 /* We are not yet positioned at any field. Let's pick the first one */
2927 r
= journal_file_map_field_hash_table(f
);
2931 m
= le64toh(f
->header
->field_hash_table_size
) / sizeof(HashItem
);
2933 if (j
->fields_hash_table_index
>= m
) {
2934 /* Reached the end of the hash table, go to the next file. */
2939 j
->fields_offset
= le64toh(f
->field_hash_table
[j
->fields_hash_table_index
].head_hash_offset
);
2941 if (j
->fields_offset
!= 0)
2944 /* Empty hash table bucket, go to next one */
2945 j
->fields_hash_table_index
++;
2949 /* Proceed with next file */
2950 j
->fields_file
= ordered_hashmap_next(j
->files
, f
->path
);
2951 if (!j
->fields_file
) {
2956 j
->fields_offset
= 0;
2957 j
->fields_hash_table_index
= 0;
2962 /* We are already positioned at a field. If so, let's figure out the next field from it */
2964 r
= journal_file_move_to_object(f
, OBJECT_FIELD
, j
->fields_offset
, &o
);
2968 j
->fields_offset
= le64toh(o
->field
.next_hash_offset
);
2969 if (j
->fields_offset
== 0) {
2970 /* Reached the end of the hash table chain */
2971 j
->fields_hash_table_index
++;
2976 /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */
2977 r
= journal_file_move_to_object(f
, OBJECT_UNUSED
, j
->fields_offset
, &o
);
2981 /* Because we used OBJECT_UNUSED above, we need to do our type check manually */
2982 if (o
->object
.type
!= OBJECT_FIELD
) {
2983 log_debug("%s:offset " OFSfmt
": object has type %i, expected %i", f
->path
, j
->fields_offset
, o
->object
.type
, OBJECT_FIELD
);
2987 sz
= le64toh(o
->object
.size
) - offsetof(Object
, field
.payload
);
2989 /* Let's see if we already returned this field name before. */
2991 ORDERED_HASHMAP_FOREACH(of
, j
->files
, i
) {
2995 /* Skip this file it didn't have any fields indexed */
2996 if (JOURNAL_HEADER_CONTAINS(of
->header
, n_fields
) && le64toh(of
->header
->n_fields
) <= 0)
2999 r
= journal_file_find_field_object_with_hash(of
, o
->field
.payload
, sz
, le64toh(o
->field
.hash
), NULL
, NULL
);
3011 /* Check if this is really a valid string containing no NUL byte */
3012 if (memchr(o
->field
.payload
, 0, sz
))
3015 if (sz
> j
->data_threshold
)
3016 sz
= j
->data_threshold
;
3018 if (!GREEDY_REALLOC(j
->fields_buffer
, j
->fields_buffer_allocated
, sz
+ 1))
3021 memcpy(j
->fields_buffer
, o
->field
.payload
, sz
);
3022 j
->fields_buffer
[sz
] = 0;
3024 if (!field_is_valid(j
->fields_buffer
))
3027 *field
= j
->fields_buffer
;
3032 _public_
void sd_journal_restart_fields(sd_journal
*j
) {
3036 j
->fields_file
= NULL
;
3037 j
->fields_hash_table_index
= 0;
3038 j
->fields_offset
= 0;
3039 j
->fields_file_lost
= false;
3042 _public_
int sd_journal_reliable_fd(sd_journal
*j
) {
3043 assert_return(j
, -EINVAL
);
3044 assert_return(!journal_pid_changed(j
), -ECHILD
);
3046 return !j
->on_network
;
3049 static char *lookup_field(const char *field
, void *userdata
) {
3050 sd_journal
*j
= userdata
;
3058 r
= sd_journal_get_data(j
, field
, &data
, &size
);
3060 size
> REPLACE_VAR_MAX
)
3061 return strdup(field
);
3063 d
= strlen(field
) + 1;
3065 return strndup((const char*) data
+ d
, size
- d
);
3068 _public_
int sd_journal_get_catalog(sd_journal
*j
, char **ret
) {
3072 _cleanup_free_
char *text
= NULL
, *cid
= NULL
;
3076 assert_return(j
, -EINVAL
);
3077 assert_return(!journal_pid_changed(j
), -ECHILD
);
3078 assert_return(ret
, -EINVAL
);
3080 r
= sd_journal_get_data(j
, "MESSAGE_ID", &data
, &size
);
3084 cid
= strndup((const char*) data
+ 11, size
- 11);
3088 r
= sd_id128_from_string(cid
, &id
);
3092 r
= catalog_get(CATALOG_DATABASE
, id
, &text
);
3096 t
= replace_var(text
, lookup_field
, j
);
3104 _public_
int sd_journal_get_catalog_for_message_id(sd_id128_t id
, char **ret
) {
3105 assert_return(ret
, -EINVAL
);
3107 return catalog_get(CATALOG_DATABASE
, id
, ret
);
3110 _public_
int sd_journal_set_data_threshold(sd_journal
*j
, size_t sz
) {
3111 assert_return(j
, -EINVAL
);
3112 assert_return(!journal_pid_changed(j
), -ECHILD
);
3114 j
->data_threshold
= sz
;
3118 _public_
int sd_journal_get_data_threshold(sd_journal
*j
, size_t *sz
) {
3119 assert_return(j
, -EINVAL
);
3120 assert_return(!journal_pid_changed(j
), -ECHILD
);
3121 assert_return(sz
, -EINVAL
);
3123 *sz
= j
->data_threshold
;
3127 _public_
int sd_journal_has_runtime_files(sd_journal
*j
) {
3128 assert_return(j
, -EINVAL
);
3130 return j
->has_runtime_files
;
3133 _public_
int sd_journal_has_persistent_files(sd_journal
*j
) {
3134 assert_return(j
, -EINVAL
);
3136 return j
->has_persistent_files
;