1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2011 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 2 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
26 #include <sys/statvfs.h>
30 #include "journal-def.h"
31 #include "journal-file.h"
34 #define DEFAULT_ARENA_MAX_SIZE (16ULL*1024ULL*1024ULL*1024ULL)
35 #define DEFAULT_ARENA_MIN_SIZE (256ULL*1024ULL)
36 #define DEFAULT_ARENA_KEEP_FREE (1ULL*1024ULL*1024ULL)
38 #define DEFAULT_MAX_USE (16ULL*1024ULL*1024ULL*16ULL)
40 #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
41 #define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
43 #define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
45 static const char signature
[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
47 #define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
49 void journal_file_close(JournalFile
*f
) {
54 if (f
->header
&& f
->writable
)
55 f
->header
->state
= STATE_OFFLINE
;
58 for (t
= 0; t
< _WINDOW_MAX
; t
++)
59 if (f
->windows
[t
].ptr
)
60 munmap(f
->windows
[t
].ptr
, f
->windows
[t
].size
);
63 close_nointr_nofail(f
->fd
);
69 static int journal_file_init_header(JournalFile
*f
, JournalFile
*template) {
77 memcpy(h
.signature
, signature
, 8);
78 h
.arena_offset
= htole64(ALIGN64(sizeof(h
)));
79 h
.arena_max_size
= htole64(DEFAULT_ARENA_MAX_SIZE
);
80 h
.arena_min_size
= htole64(DEFAULT_ARENA_MIN_SIZE
);
81 h
.arena_keep_free
= htole64(DEFAULT_ARENA_KEEP_FREE
);
83 r
= sd_id128_randomize(&h
.file_id
);
88 h
.seqnum_id
= template->header
->seqnum_id
;
89 h
.seqnum
= template->header
->seqnum
;
91 h
.seqnum_id
= h
.file_id
;
93 k
= pwrite(f
->fd
, &h
, sizeof(h
), 0);
103 static int journal_file_refresh_header(JournalFile
*f
) {
109 r
= sd_id128_get_machine(&f
->header
->machine_id
);
113 r
= sd_id128_get_boot(&boot_id
);
117 if (sd_id128_equal(boot_id
, f
->header
->boot_id
))
118 f
->tail_entry_monotonic_valid
= true;
120 f
->header
->boot_id
= boot_id
;
122 f
->header
->state
= STATE_ONLINE
;
126 static int journal_file_verify_header(JournalFile
*f
) {
129 if (memcmp(f
->header
, signature
, 8))
132 if (f
->header
->incompatible_flags
!= 0)
133 return -EPROTONOSUPPORT
;
135 if ((uint64_t) f
->last_stat
.st_size
< (le64toh(f
->header
->arena_offset
) + le64toh(f
->header
->arena_size
)))
140 sd_id128_t machine_id
;
143 r
= sd_id128_get_machine(&machine_id
);
147 if (!sd_id128_equal(machine_id
, f
->header
->machine_id
))
150 state
= f
->header
->state
;
152 if (state
== STATE_ONLINE
)
153 log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f
->path
);
154 else if (state
== STATE_ARCHIVED
)
156 else if (state
!= STATE_OFFLINE
)
157 log_debug("Journal file %s has unknown state %u. Ignoring.", f
->path
, state
);
163 static int journal_file_allocate(JournalFile
*f
, uint64_t offset
, uint64_t size
) {
165 uint64_t old_size
, new_size
;
169 if (offset
< le64toh(f
->header
->arena_offset
))
172 new_size
= PAGE_ALIGN(offset
+ size
);
174 /* We assume that this file is not sparse, and we know that
175 * for sure, since we always call posix_fallocate()
179 le64toh(f
->header
->arena_offset
) +
180 le64toh(f
->header
->arena_size
);
182 if (old_size
>= new_size
)
185 asize
= new_size
- le64toh(f
->header
->arena_offset
);
187 if (asize
> le64toh(f
->header
->arena_min_size
)) {
190 if (fstatvfs(f
->fd
, &svfs
) >= 0) {
193 available
= svfs
.f_bfree
* svfs
.f_bsize
;
195 if (available
>= f
->header
->arena_keep_free
)
196 available
-= f
->header
->arena_keep_free
;
200 if (new_size
- old_size
> available
)
205 if (asize
> le64toh(f
->header
->arena_max_size
))
208 if (posix_fallocate(f
->fd
, old_size
, new_size
- old_size
) < 0)
211 if (fstat(f
->fd
, &f
->last_stat
) < 0)
214 f
->header
->arena_size
= htole64(asize
);
219 static int journal_file_map(
228 uint64_t woffset
, wsize
;
235 woffset
= offset
& ~((uint64_t) page_size() - 1ULL);
236 wsize
= size
+ (offset
- woffset
);
237 wsize
= PAGE_ALIGN(wsize
);
239 window
= mmap(NULL
, wsize
, f
->prot
, MAP_SHARED
, f
->fd
, woffset
);
240 if (window
== MAP_FAILED
)
252 *ret
= (uint8_t*) window
+ (offset
- woffset
);
257 static int journal_file_move_to(JournalFile
*f
, int wt
, uint64_t offset
, uint64_t size
, void **ret
) {
266 assert(wt
< _WINDOW_MAX
);
270 if (_likely_(w
->ptr
&&
271 w
->offset
<= offset
&&
272 w
->offset
+ w
->size
>= offset
+ size
)) {
274 *ret
= (uint8_t*) w
->ptr
+ (offset
- w
->offset
);
279 if (munmap(w
->ptr
, w
->size
) < 0)
283 w
->size
= w
->offset
= 0;
286 if (size
< DEFAULT_WINDOW_SIZE
) {
287 /* If the default window size is larger then what was
288 * asked for extend the mapping a bit in the hope to
289 * minimize needed remappings later on. We add half
290 * the window space before and half behind the
291 * requested mapping */
293 delta
= PAGE_ALIGN((DEFAULT_WINDOW_SIZE
- size
) / 2);
299 size
+= (DEFAULT_WINDOW_SIZE
- delta
);
303 r
= journal_file_map(f
,
305 &w
->ptr
, &w
->offset
, &w
->size
,
311 *ret
= (uint8_t*) p
+ delta
;
315 static bool verify_hash(Object
*o
) {
320 if (o
->object
.type
== OBJECT_DATA
) {
321 h1
= le64toh(o
->data
.hash
);
322 h2
= hash64(o
->data
.payload
, le64toh(o
->object
.size
) - offsetof(Object
, data
.payload
));
323 } else if (o
->object
.type
== OBJECT_FIELD
) {
324 h1
= le64toh(o
->field
.hash
);
325 h2
= hash64(o
->field
.payload
, le64toh(o
->object
.size
) - offsetof(Object
, field
.payload
));
332 int journal_file_move_to_object(JournalFile
*f
, int type
, uint64_t offset
, Object
**ret
) {
340 assert(type
< _OBJECT_TYPE_MAX
);
342 r
= journal_file_move_to(f
, type
>= 0 ? type
: WINDOW_UNKNOWN
, offset
, sizeof(ObjectHeader
), &t
);
347 s
= le64toh(o
->object
.size
);
349 if (s
< sizeof(ObjectHeader
))
352 if (type
>= 0 && o
->object
.type
!= type
)
355 if (s
> sizeof(ObjectHeader
)) {
356 r
= journal_file_move_to(f
, o
->object
.type
, offset
, s
, &t
);
370 static uint64_t journal_file_seqnum(JournalFile
*f
, uint64_t *seqnum
) {
375 r
= le64toh(f
->header
->seqnum
) + 1;
378 /* If an external seqnum counter was passed, we update
379 * both the local and the external one, and set it to
380 * the maximum of both */
388 f
->header
->seqnum
= htole64(r
);
390 if (f
->header
->first_seqnum
== 0)
391 f
->header
->first_seqnum
= htole64(r
);
396 static int journal_file_append_object(JournalFile
*f
, int type
, uint64_t size
, Object
**ret
, uint64_t *offset
) {
403 assert(size
>= sizeof(ObjectHeader
));
407 p
= le64toh(f
->header
->tail_object_offset
);
409 p
= le64toh(f
->header
->arena_offset
);
411 r
= journal_file_move_to_object(f
, -1, p
, &tail
);
415 p
+= ALIGN64(le64toh(tail
->object
.size
));
418 r
= journal_file_allocate(f
, p
, size
);
422 r
= journal_file_move_to(f
, type
, p
, size
, &t
);
429 o
->object
.type
= type
;
430 o
->object
.size
= htole64(size
);
432 f
->header
->tail_object_offset
= htole64(p
);
433 f
->header
->n_objects
= htole64(le64toh(f
->header
->n_objects
) + 1);
441 static int journal_file_setup_data_hash_table(JournalFile
*f
) {
448 s
= DEFAULT_DATA_HASH_TABLE_SIZE
;
449 r
= journal_file_append_object(f
,
450 OBJECT_DATA_HASH_TABLE
,
451 offsetof(Object
, hash_table
.items
) + s
,
456 memset(o
->hash_table
.items
, 0, s
);
458 f
->header
->data_hash_table_offset
= htole64(p
+ offsetof(Object
, hash_table
.items
));
459 f
->header
->data_hash_table_size
= htole64(s
);
464 static int journal_file_setup_field_hash_table(JournalFile
*f
) {
471 s
= DEFAULT_FIELD_HASH_TABLE_SIZE
;
472 r
= journal_file_append_object(f
,
473 OBJECT_FIELD_HASH_TABLE
,
474 offsetof(Object
, hash_table
.items
) + s
,
479 memset(o
->hash_table
.items
, 0, s
);
481 f
->header
->field_hash_table_offset
= htole64(p
+ offsetof(Object
, hash_table
.items
));
482 f
->header
->field_hash_table_size
= htole64(s
);
487 static int journal_file_map_data_hash_table(JournalFile
*f
) {
494 p
= le64toh(f
->header
->data_hash_table_offset
);
495 s
= le64toh(f
->header
->data_hash_table_size
);
497 r
= journal_file_move_to(f
,
498 WINDOW_DATA_HASH_TABLE
,
504 f
->data_hash_table
= t
;
508 static int journal_file_map_field_hash_table(JournalFile
*f
) {
515 p
= le64toh(f
->header
->field_hash_table_offset
);
516 s
= le64toh(f
->header
->field_hash_table_size
);
518 r
= journal_file_move_to(f
,
519 WINDOW_FIELD_HASH_TABLE
,
525 f
->field_hash_table
= t
;
529 static int journal_file_link_data(JournalFile
*f
, Object
*o
, uint64_t offset
, uint64_t hash
) {
536 assert(o
->object
.type
== OBJECT_DATA
);
538 o
->data
.next_hash_offset
= o
->data
.next_field_offset
= 0;
539 o
->data
.entry_offset
= o
->data
.entry_array_offset
= 0;
540 o
->data
.n_entries
= 0;
542 h
= hash
% (le64toh(f
->header
->data_hash_table_size
) / sizeof(HashItem
));
543 p
= le64toh(f
->data_hash_table
[h
].head_hash_offset
);
545 /* Only entry in the hash table is easy */
546 f
->data_hash_table
[h
].head_hash_offset
= htole64(offset
);
548 /* Temporarily move back to the previous data object,
549 * to patch in pointer */
551 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
555 o
->data
.next_hash_offset
= htole64(offset
);
557 r
= journal_file_move_to_object(f
, OBJECT_DATA
, offset
, &o
);
562 f
->data_hash_table
[h
].tail_hash_offset
= htole64(offset
);
567 int journal_file_find_data_object_with_hash(
569 const void *data
, uint64_t size
, uint64_t hash
,
570 Object
**ret
, uint64_t *offset
) {
571 uint64_t p
, osize
, h
;
575 assert(data
|| size
== 0);
577 osize
= offsetof(Object
, data
.payload
) + size
;
579 h
= hash
% (le64toh(f
->header
->data_hash_table_size
) / sizeof(HashItem
));
580 p
= le64toh(f
->data_hash_table
[h
].head_hash_offset
);
585 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
589 if (le64toh(o
->object
.size
) == osize
&&
590 memcmp(o
->data
.payload
, data
, size
) == 0) {
592 if (le64toh(o
->data
.hash
) != hash
)
604 p
= le64toh(o
->data
.next_hash_offset
);
610 int journal_file_find_data_object(
612 const void *data
, uint64_t size
,
613 Object
**ret
, uint64_t *offset
) {
618 assert(data
|| size
== 0);
620 hash
= hash64(data
, size
);
622 return journal_file_find_data_object_with_hash(f
,
627 static int journal_file_append_data(JournalFile
*f
, const void *data
, uint64_t size
, Object
**ret
, uint64_t *offset
) {
634 assert(data
|| size
== 0);
636 hash
= hash64(data
, size
);
638 r
= journal_file_find_data_object_with_hash(f
, data
, size
, hash
, &o
, &p
);
652 osize
= offsetof(Object
, data
.payload
) + size
;
653 r
= journal_file_append_object(f
, OBJECT_DATA
, osize
, &o
, &p
);
657 o
->data
.hash
= htole64(hash
);
658 memcpy(o
->data
.payload
, data
, size
);
660 r
= journal_file_link_data(f
, o
, p
, hash
);
673 uint64_t journal_file_entry_n_items(Object
*o
) {
675 assert(o
->object
.type
== htole64(OBJECT_ENTRY
));
677 return (le64toh(o
->object
.size
) - offsetof(Object
, entry
.items
)) / sizeof(EntryItem
);
680 static uint64_t journal_file_entry_array_n_items(Object
*o
) {
682 assert(o
->object
.type
== htole64(OBJECT_ENTRY_ARRAY
));
684 return (le64toh(o
->object
.size
) - offsetof(Object
, entry_array
.items
)) / sizeof(uint64_t);
687 static int link_entry_into_array(JournalFile
*f
,
692 uint64_t n
= 0, ap
= 0, q
, i
, a
, hidx
;
701 i
= hidx
= le64toh(*idx
);
704 r
= journal_file_move_to_object(f
, OBJECT_ENTRY_ARRAY
, a
, &o
);
708 n
= journal_file_entry_array_n_items(o
);
710 o
->entry_array
.items
[i
] = htole64(p
);
711 *idx
= htole64(hidx
+ 1);
717 a
= le64toh(o
->entry_array
.next_entry_array_offset
);
728 r
= journal_file_append_object(f
, OBJECT_ENTRY_ARRAY
,
729 offsetof(Object
, entry_array
.items
) + n
* sizeof(uint64_t),
734 o
->entry_array
.items
[i
] = htole64(p
);
739 r
= journal_file_move_to_object(f
, OBJECT_ENTRY_ARRAY
, ap
, &o
);
743 o
->entry_array
.next_entry_array_offset
= htole64(q
);
746 *idx
= htole64(hidx
+ 1);
751 static int link_entry_into_array_plus_one(JournalFile
*f
,
770 i
= le64toh(*idx
) - 1;
771 r
= link_entry_into_array(f
, first
, &i
, p
);
776 *idx
= htole64(le64toh(*idx
) + 1);
780 static int journal_file_link_entry_item(JournalFile
*f
, Object
*o
, uint64_t offset
, uint64_t i
) {
787 p
= le64toh(o
->entry
.items
[i
].object_offset
);
791 r
= journal_file_move_to_object(f
, OBJECT_DATA
, p
, &o
);
795 return link_entry_into_array_plus_one(f
,
796 &o
->data
.entry_offset
,
797 &o
->data
.entry_array_offset
,
802 static int journal_file_link_entry(JournalFile
*f
, Object
*o
, uint64_t offset
) {
809 assert(o
->object
.type
== OBJECT_ENTRY
);
811 /* Link up the entry itself */
812 r
= link_entry_into_array(f
,
813 &f
->header
->entry_array_offset
,
814 &f
->header
->n_entries
,
819 log_error("%s %lu", f
->path
, (unsigned long) f
->header
->n_entries
);
821 if (f
->header
->head_entry_realtime
== 0)
822 f
->header
->head_entry_realtime
= o
->entry
.realtime
;
824 f
->header
->tail_entry_realtime
= o
->entry
.realtime
;
825 f
->header
->tail_entry_monotonic
= o
->entry
.monotonic
;
827 f
->tail_entry_monotonic_valid
= true;
829 /* Link up the items */
830 n
= journal_file_entry_n_items(o
);
831 for (i
= 0; i
< n
; i
++) {
832 r
= journal_file_link_entry_item(f
, o
, offset
, i
);
840 static int journal_file_append_entry_internal(
842 const dual_timestamp
*ts
,
844 const EntryItem items
[], unsigned n_items
,
846 Object
**ret
, uint64_t *offset
) {
853 assert(items
|| n_items
== 0);
856 osize
= offsetof(Object
, entry
.items
) + (n_items
* sizeof(EntryItem
));
858 r
= journal_file_append_object(f
, OBJECT_ENTRY
, osize
, &o
, &np
);
862 o
->entry
.seqnum
= htole64(journal_file_seqnum(f
, seqnum
));
863 memcpy(o
->entry
.items
, items
, n_items
* sizeof(EntryItem
));
864 o
->entry
.realtime
= htole64(ts
->realtime
);
865 o
->entry
.monotonic
= htole64(ts
->monotonic
);
866 o
->entry
.xor_hash
= htole64(xor_hash
);
867 o
->entry
.boot_id
= f
->header
->boot_id
;
869 r
= journal_file_link_entry(f
, o
, np
);
882 int journal_file_append_entry(JournalFile
*f
, const dual_timestamp
*ts
, const struct iovec iovec
[], unsigned n_iovec
, uint64_t *seqnum
, Object
**ret
, uint64_t *offset
) {
886 uint64_t xor_hash
= 0;
887 struct dual_timestamp _ts
;
890 assert(iovec
|| n_iovec
== 0);
896 dual_timestamp_get(&_ts
);
900 if (f
->tail_entry_monotonic_valid
&&
901 ts
->monotonic
< le64toh(f
->header
->tail_entry_monotonic
))
904 if (ts
->realtime
< le64toh(f
->header
->tail_entry_realtime
))
907 items
= new(EntryItem
, n_iovec
);
911 for (i
= 0; i
< n_iovec
; i
++) {
915 r
= journal_file_append_data(f
, iovec
[i
].iov_base
, iovec
[i
].iov_len
, &o
, &p
);
919 xor_hash
^= le64toh(o
->data
.hash
);
920 items
[i
].object_offset
= htole64(p
);
921 items
[i
].hash
= o
->data
.hash
;
924 r
= journal_file_append_entry_internal(f
, ts
, xor_hash
, items
, n_iovec
, seqnum
, ret
, offset
);
932 static int generic_array_get(JournalFile
*f
,
935 Object
**ret
, uint64_t *offset
) {
947 r
= journal_file_move_to_object(f
, OBJECT_ENTRY_ARRAY
, a
, &o
);
951 n
= journal_file_entry_array_n_items(o
);
953 p
= le64toh(o
->entry_array
.items
[i
]);
958 a
= le64toh(o
->entry_array
.next_entry_array_offset
);
961 if (a
<= 0 || p
<= 0)
964 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, p
, &o
);
977 static int generic_array_get_plus_one(JournalFile
*f
,
981 Object
**ret
, uint64_t *offset
) {
990 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, extra
, &o
);
1003 return generic_array_get(f
, first
, i
-1, ret
, offset
);
1012 static int generic_array_bisect(JournalFile
*f
,
1016 int (*test_object
)(JournalFile
*f
, uint64_t p
, uint64_t needle
),
1017 direction_t direction
,
1022 uint64_t a
, p
, t
= 0, i
= 0, last_p
= 0;
1023 bool subtract_one
= false;
1024 Object
*o
, *array
= NULL
;
1028 assert(test_object
);
1032 uint64_t left
, right
, k
, lp
;
1034 r
= journal_file_move_to_object(f
, OBJECT_ENTRY_ARRAY
, a
, &array
);
1038 k
= journal_file_entry_array_n_items(array
);
1044 lp
= p
= le64toh(array
->entry_array
.items
[i
]);
1048 r
= test_object(f
, p
, needle
);
1052 if (r
== TEST_FOUND
)
1053 r
= direction
== DIRECTION_DOWN
? TEST_RIGHT
: TEST_LEFT
;
1055 if (r
== TEST_RIGHT
) {
1059 if (left
== right
) {
1060 if (direction
== DIRECTION_UP
)
1061 subtract_one
= true;
1067 assert(left
< right
);
1069 i
= (left
+ right
) / 2;
1070 p
= le64toh(array
->entry_array
.items
[i
]);
1074 r
= test_object(f
, p
, needle
);
1078 if (r
== TEST_FOUND
)
1079 r
= direction
== DIRECTION_DOWN
? TEST_RIGHT
: TEST_LEFT
;
1081 if (r
== TEST_RIGHT
)
1095 a
= le64toh(array
->entry_array
.next_entry_array_offset
);
1101 if (subtract_one
&& t
== 0 && i
== 0)
1104 if (subtract_one
&& i
== 0)
1106 else if (subtract_one
)
1107 p
= le64toh(array
->entry_array
.items
[i
-1]);
1109 p
= le64toh(array
->entry_array
.items
[i
]);
1111 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, p
, &o
);
1122 *idx
= t
+ i
- (subtract_one
? 1 : 0);
1127 static int generic_array_bisect_plus_one(JournalFile
*f
,
1132 int (*test_object
)(JournalFile
*f
, uint64_t p
, uint64_t needle
),
1133 direction_t direction
,
1141 assert(test_object
);
1146 /* This bisects the array in object 'first', but first checks
1149 r
= test_object(f
, extra
, needle
);
1152 else if (r
== TEST_FOUND
) {
1155 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, extra
, &o
);
1164 } else if (r
== TEST_RIGHT
)
1167 r
= generic_array_bisect(f
, first
, n
-1, needle
, test_object
, direction
, ret
, offset
, idx
);
1175 static int test_object_seqnum(JournalFile
*f
, uint64_t p
, uint64_t needle
) {
1182 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, p
, &o
);
1186 if (le64toh(o
->entry
.seqnum
) == needle
)
1188 else if (le64toh(o
->entry
.seqnum
) < needle
)
1194 int journal_file_move_to_entry_by_seqnum(
1197 direction_t direction
,
1201 return generic_array_bisect(f
,
1202 le64toh(f
->header
->entry_array_offset
),
1203 le64toh(f
->header
->n_entries
),
1210 static int test_object_realtime(JournalFile
*f
, uint64_t p
, uint64_t needle
) {
1217 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, p
, &o
);
1221 if (le64toh(o
->entry
.realtime
) == needle
)
1223 else if (le64toh(o
->entry
.realtime
) < needle
)
1229 int journal_file_move_to_entry_by_realtime(
1232 direction_t direction
,
1236 return generic_array_bisect(f
,
1237 le64toh(f
->header
->entry_array_offset
),
1238 le64toh(f
->header
->n_entries
),
1240 test_object_realtime
,
1245 static int test_object_monotonic(JournalFile
*f
, uint64_t p
, uint64_t needle
) {
1252 r
= journal_file_move_to_object(f
, OBJECT_ENTRY
, p
, &o
);
1256 if (le64toh(o
->entry
.monotonic
) == needle
)
1258 else if (le64toh(o
->entry
.monotonic
) < needle
)
1264 int journal_file_move_to_entry_by_monotonic(
1268 direction_t direction
,
1272 char t
[8+32+1] = "_BOOT_ID=";
1276 sd_id128_to_string(boot_id
, t
+ 8);
1278 r
= journal_file_find_data_object(f
, t
, strlen(t
), &o
, NULL
);
1284 return generic_array_bisect_plus_one(f
,
1285 le64toh(o
->data
.entry_offset
),
1286 le64toh(o
->data
.entry_array_offset
),
1287 le64toh(o
->data
.n_entries
),
1289 test_object_monotonic
,
1294 static int test_object_offset(JournalFile
*f
, uint64_t p
, uint64_t needle
) {
1300 else if (p
< needle
)
1306 int journal_file_next_entry(
1308 Object
*o
, uint64_t p
,
1309 direction_t direction
,
1310 Object
**ret
, uint64_t *offset
) {
1316 assert(p
> 0 || !o
);
1318 n
= le64toh(f
->header
->n_entries
);
1323 i
= direction
== DIRECTION_DOWN
? 0 : n
- 1;
1325 if (o
->object
.type
!= OBJECT_ENTRY
)
1328 r
= generic_array_bisect(f
,
1329 le64toh(f
->header
->entry_array_offset
),
1330 le64toh(f
->header
->n_entries
),
1339 if (direction
== DIRECTION_DOWN
) {
1352 /* And jump to it */
1353 return generic_array_get(f
,
1354 le64toh(f
->header
->entry_array_offset
),
1359 int journal_file_skip_entry(
1361 Object
*o
, uint64_t p
,
1363 Object
**ret
, uint64_t *offset
) {
1372 if (o
->object
.type
!= OBJECT_ENTRY
)
1375 r
= generic_array_bisect(f
,
1376 le64toh(f
->header
->entry_array_offset
),
1377 le64toh(f
->header
->n_entries
),
1386 /* Calculate new index */
1388 if ((uint64_t) -skip
>= i
)
1391 i
= i
- (uint64_t) -skip
;
1393 i
+= (uint64_t) skip
;
1395 n
= le64toh(f
->header
->n_entries
);
1402 return generic_array_get(f
,
1403 le64toh(f
->header
->entry_array_offset
),
1408 int journal_file_next_entry_for_data(
1410 Object
*o
, uint64_t p
,
1411 uint64_t data_offset
,
1412 direction_t direction
,
1413 Object
**ret
, uint64_t *offset
) {
1420 assert(p
> 0 || !o
);
1422 r
= journal_file_move_to_object(f
, OBJECT_DATA
, data_offset
, &d
);
1426 n
= le64toh(d
->data
.n_entries
);
1431 i
= direction
== DIRECTION_DOWN
? 0 : n
- 1;
1433 if (o
->object
.type
!= OBJECT_ENTRY
)
1436 r
= generic_array_bisect_plus_one(f
,
1437 le64toh(d
->data
.entry_offset
),
1438 le64toh(d
->data
.entry_array_offset
),
1439 le64toh(d
->data
.n_entries
),
1449 if (direction
== DIRECTION_DOWN
) {
1463 return generic_array_get_plus_one(f
,
1464 le64toh(d
->data
.entry_offset
),
1465 le64toh(d
->data
.entry_array_offset
),
1470 int journal_file_move_to_entry_by_seqnum_for_data(
1472 uint64_t data_offset
,
1474 direction_t direction
,
1475 Object
**ret
, uint64_t *offset
) {
1480 r
= journal_file_move_to_object(f
, OBJECT_DATA
, data_offset
, &d
);
1484 return generic_array_bisect_plus_one(f
,
1485 le64toh(d
->data
.entry_offset
),
1486 le64toh(d
->data
.entry_array_offset
),
1487 le64toh(d
->data
.n_entries
),
1494 int journal_file_move_to_entry_by_realtime_for_data(
1496 uint64_t data_offset
,
1498 direction_t direction
,
1499 Object
**ret
, uint64_t *offset
) {
1504 r
= journal_file_move_to_object(f
, OBJECT_DATA
, data_offset
, &d
);
1508 return generic_array_bisect_plus_one(f
,
1509 le64toh(d
->data
.entry_offset
),
1510 le64toh(d
->data
.entry_array_offset
),
1511 le64toh(d
->data
.n_entries
),
1513 test_object_realtime
,
1518 void journal_file_dump(JournalFile
*f
) {
1519 char a
[33], b
[33], c
[33];
1526 printf("File Path: %s\n"
1530 "Arena size: %llu\n"
1534 sd_id128_to_string(f
->header
->file_id
, a
),
1535 sd_id128_to_string(f
->header
->machine_id
, b
),
1536 sd_id128_to_string(f
->header
->boot_id
, c
),
1537 (unsigned long long) le64toh(f
->header
->arena_size
),
1538 (unsigned long) le64toh(f
->header
->n_objects
),
1539 (unsigned long) le64toh(f
->header
->n_entries
));
1541 p
= le64toh(f
->header
->arena_offset
);
1543 r
= journal_file_move_to_object(f
, -1, p
, &o
);
1547 switch (o
->object
.type
) {
1550 printf("Type: OBJECT_UNUSED\n");
1554 printf("Type: OBJECT_DATA\n");
1558 printf("Type: OBJECT_ENTRY %llu %llu %llu\n",
1559 (unsigned long long) le64toh(o
->entry
.seqnum
),
1560 (unsigned long long) le64toh(o
->entry
.monotonic
),
1561 (unsigned long long) le64toh(o
->entry
.realtime
));
1564 case OBJECT_FIELD_HASH_TABLE
:
1565 printf("Type: OBJECT_FIELD_HASH_TABLE\n");
1568 case OBJECT_DATA_HASH_TABLE
:
1569 printf("Type: OBJECT_DATA_HASH_TABLE\n");
1572 case OBJECT_ENTRY_ARRAY
:
1573 printf("Type: OBJECT_ENTRY_ARRAY\n");
1577 if (p
== le64toh(f
->header
->tail_object_offset
))
1580 p
= p
+ ALIGN64(le64toh(o
->object
.size
));
1585 log_error("File corrupt");
1588 int journal_file_open(
1592 JournalFile
*template,
1593 JournalFile
**ret
) {
1597 bool newly_created
= false;
1601 if ((flags
& O_ACCMODE
) != O_RDONLY
&&
1602 (flags
& O_ACCMODE
) != O_RDWR
)
1605 f
= new0(JournalFile
, 1);
1612 f
->writable
= (flags
& O_ACCMODE
) != O_RDONLY
;
1613 f
->prot
= prot_from_flags(flags
);
1615 f
->path
= strdup(fname
);
1621 f
->fd
= open(f
->path
, f
->flags
|O_CLOEXEC
, f
->mode
);
1627 if (fstat(f
->fd
, &f
->last_stat
) < 0) {
1632 if (f
->last_stat
.st_size
== 0 && f
->writable
) {
1633 newly_created
= true;
1635 r
= journal_file_init_header(f
, template);
1639 if (fstat(f
->fd
, &f
->last_stat
) < 0) {
1645 if (f
->last_stat
.st_size
< (off_t
) sizeof(Header
)) {
1650 f
->header
= mmap(NULL
, PAGE_ALIGN(sizeof(Header
)), prot_from_flags(flags
), MAP_SHARED
, f
->fd
, 0);
1651 if (f
->header
== MAP_FAILED
) {
1657 if (!newly_created
) {
1658 r
= journal_file_verify_header(f
);
1664 r
= journal_file_refresh_header(f
);
1669 if (newly_created
) {
1671 r
= journal_file_setup_field_hash_table(f
);
1675 r
= journal_file_setup_data_hash_table(f
);
1680 r
= journal_file_map_field_hash_table(f
);
1684 r
= journal_file_map_data_hash_table(f
);
1694 journal_file_close(f
);
1699 int journal_file_rotate(JournalFile
**f
) {
1702 JournalFile
*old_file
, *new_file
= NULL
;
1710 if (!old_file
->writable
)
1713 if (!endswith(old_file
->path
, ".journal"))
1716 l
= strlen(old_file
->path
);
1718 p
= new(char, l
+ 1 + 16 + 1 + 32 + 1 + 16 + 1);
1722 memcpy(p
, old_file
->path
, l
- 8);
1724 sd_id128_to_string(old_file
->header
->seqnum_id
, p
+ l
- 8 + 1);
1725 snprintf(p
+ l
- 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1,
1726 "-%016llx-%016llx.journal",
1727 (unsigned long long) le64toh((*f
)->header
->seqnum
),
1728 (unsigned long long) le64toh((*f
)->header
->tail_entry_realtime
));
1730 r
= rename(old_file
->path
, p
);
1736 old_file
->header
->state
= le32toh(STATE_ARCHIVED
);
1738 r
= journal_file_open(old_file
->path
, old_file
->flags
, old_file
->mode
, old_file
, &new_file
);
1739 journal_file_close(old_file
);
1745 struct vacuum_info
{
1750 sd_id128_t seqnum_id
;
1754 static int vacuum_compare(const void *_a
, const void *_b
) {
1755 const struct vacuum_info
*a
, *b
;
1760 if (sd_id128_equal(a
->seqnum_id
, b
->seqnum_id
)) {
1761 if (a
->seqnum
< b
->seqnum
)
1763 else if (a
->seqnum
> b
->seqnum
)
1769 if (a
->realtime
< b
->realtime
)
1771 else if (a
->realtime
> b
->realtime
)
1774 return memcmp(&a
->seqnum_id
, &b
->seqnum_id
, 16);
1777 int journal_directory_vacuum(const char *directory
, uint64_t max_use
, uint64_t min_free
) {
1780 struct vacuum_info
*list
= NULL
;
1781 unsigned n_list
= 0, n_allocated
= 0, i
;
1787 max_use
= DEFAULT_MAX_USE
;
1789 d
= opendir(directory
);
1795 struct dirent buf
, *de
;
1799 unsigned long long seqnum
, realtime
;
1800 sd_id128_t seqnum_id
;
1802 k
= readdir_r(d
, &buf
, &de
);
1811 if (!dirent_is_file_with_suffix(de
, ".journal"))
1814 q
= strlen(de
->d_name
);
1816 if (q
< 1 + 32 + 1 + 16 + 1 + 16 + 8)
1819 if (de
->d_name
[q
-8-16-1] != '-' ||
1820 de
->d_name
[q
-8-16-1-16-1] != '-' ||
1821 de
->d_name
[q
-8-16-1-16-1-32-1] != '@')
1824 if (fstatat(dirfd(d
), de
->d_name
, &st
, AT_SYMLINK_NOFOLLOW
) < 0)
1827 if (!S_ISREG(st
.st_mode
))
1830 p
= strdup(de
->d_name
);
1836 de
->d_name
[q
-8-16-1-16-1] = 0;
1837 if (sd_id128_from_string(de
->d_name
+ q
-8-16-1-16-1-32, &seqnum_id
) < 0) {
1842 if (sscanf(de
->d_name
+ q
-8-16-1-16, "%16llx-%16llx.journal", &seqnum
, &realtime
) != 2) {
1847 if (n_list
>= n_allocated
) {
1848 struct vacuum_info
*j
;
1850 n_allocated
= MAX(n_allocated
* 2U, 8U);
1851 j
= realloc(list
, n_allocated
* sizeof(struct vacuum_info
));
1861 list
[n_list
].filename
= p
;
1862 list
[n_list
].usage
= (uint64_t) st
.st_blksize
* (uint64_t) st
.st_blocks
;
1863 list
[n_list
].seqnum
= seqnum
;
1864 list
[n_list
].realtime
= realtime
;
1865 list
[n_list
].seqnum_id
= seqnum_id
;
1867 sum
+= list
[n_list
].usage
;
1872 qsort(list
, n_list
, sizeof(struct vacuum_info
), vacuum_compare
);
1874 for(i
= 0; i
< n_list
; i
++) {
1877 if (fstatvfs(dirfd(d
), &ss
) < 0) {
1882 if (sum
<= max_use
&&
1883 (uint64_t) ss
.f_bavail
* (uint64_t) ss
.f_bsize
>= min_free
)
1886 if (unlinkat(dirfd(d
), list
[i
].filename
, 0) >= 0) {
1887 log_debug("Deleted archived journal %s/%s.", directory
, list
[i
].filename
);
1888 sum
-= list
[i
].usage
;
1889 } else if (errno
!= ENOENT
)
1890 log_warning("Failed to delete %s/%s: %m", directory
, list
[i
].filename
);
1894 for (i
= 0; i
< n_list
; i
++)
1895 free(list
[i
].filename
);