]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/journal/journal-file.c
projects / thirdparty / systemd.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
journalctl: find journal files in right path
[thirdparty/systemd.git] / src / journal / journal-file.c
CommitLineData
cec736d2
LP		 1/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
2
3/***
4 This file is part of systemd.
5
6 Copyright 2011 Lennart Poettering
7
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.
12
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.
17
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/>.
20***/
21
22#include <sys/mman.h>
23#include <errno.h>
24#include <sys/uio.h>
25#include <unistd.h>
26#include <sys/statvfs.h>
27#include <fcntl.h>
28#include <stddef.h>
29
30#include "journal-def.h"
31#include "journal-file.h"
32#include "lookup3.h"
33
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)
37
38#define DEFAULT_HASH_TABLE_SIZE (2047ULL*16ULL)
39#define DEFAULT_BISECT_TABLE_SIZE ((DEFAULT_ARENA_MAX_SIZE/(64ULL*1024ULL))*8ULL)
40
41#define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
42
43static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
44
45#define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
46
47void journal_file_close(JournalFile *f) {
48 assert(f);
49
50 if (f->fd >= 0)
51 close_nointr_nofail(f->fd);
52
53 if (f->header)
54 munmap(f->header, PAGE_ALIGN(sizeof(Header)));
55
56 if (f->hash_table_window)
57 munmap(f->hash_table_window, f->hash_table_window_size);
58
59 if (f->bisect_table_window)
60 munmap(f->bisect_table_window, f->bisect_table_window_size);
61
62 if (f->window)
63 munmap(f->window, f->window_size);
64
65 free(f->path);
66 free(f);
67}
68
69static int journal_file_init_header(JournalFile *f) {
70 Header h;
71 ssize_t k;
72 int r;
73
74 assert(f);
75
76 zero(h);
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);
82
83 r = sd_id128_randomize(&h.file_id);
84 if (r < 0)
85 return r;
86
87 h.seqnum_id = h.file_id;
88
89 k = pwrite(f->fd, &h, sizeof(h), 0);
90 if (k < 0)
91 return -errno;
92
93 if (k != sizeof(h))
94 return -EIO;
95
96 return 0;
97}
98
99static int journal_file_refresh_header(JournalFile *f) {
100 int r;
101
102 assert(f);
103
104 r = sd_id128_get_machine(&f->header->machine_id);
105 if (r < 0)
106 return r;
107
108 r = sd_id128_get_boot(&f->header->boot_id);
109 if (r < 0)
110 return r;
111
112 f->header->state = htole32(STATE_ONLINE);
113 return 0;
114}
115
116static int journal_file_verify_header(JournalFile *f) {
117 assert(f);
118
119 if (memcmp(f->header, signature, 8))
120 return -EBADMSG;
121
122 if (f->header->incompatible_flags != 0)
123 return -EPROTONOSUPPORT;
124
125 if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->arena_offset) + le64toh(f->header->arena_size)))
126 return -ENODATA;
127
128 if (f->writable) {
129 uint32_t state;
130 sd_id128_t machine_id;
131 int r;
132
133 r = sd_id128_get_machine(&machine_id);
134 if (r < 0)
135 return r;
136
137 if (!sd_id128_equal(machine_id, f->header->machine_id))
138 return -EHOSTDOWN;
139
140 state = le32toh(f->header->state);
141
142 if (state == STATE_ONLINE)
143 log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
144 else if (state == STATE_ARCHIVED)
145 return -ESHUTDOWN;
146 else if (state != STATE_OFFLINE)
147 log_debug("Journal file %s has unknown state %u. Ignoring.", f->path, state);
148 }
149
150 return 0;
151}
152
153static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
154 uint64_t asize;
155 uint64_t old_size, new_size;
156
157 assert(f);
158
159 if (offset < le64toh(f->header->arena_offset))
160 return -EINVAL;
161
162 new_size = PAGE_ALIGN(offset + size);
163
164 /* We assume that this file is not sparse, and we know that
38ac38b2 165 * for sure, since we always call posix_fallocate()
cec736d2
LP		 166 * ourselves */
167
168 old_size =
169 le64toh(f->header->arena_offset) +
170 le64toh(f->header->arena_size);
171
172 if (old_size >= new_size)
173 return 0;
174
175 asize = new_size - le64toh(f->header->arena_offset);
176
177 if (asize > le64toh(f->header->arena_min_size)) {
178 struct statvfs svfs;
179
180 if (fstatvfs(f->fd, &svfs) >= 0) {
181 uint64_t available;
182
183 available = svfs.f_bfree * svfs.f_bsize;
184
185 if (available >= f->header->arena_keep_free)
186 available -= f->header->arena_keep_free;
187 else
188 available = 0;
189
190 if (new_size - old_size > available)
191 return -E2BIG;
192 }
193 }
194
195 if (asize > le64toh(f->header->arena_max_size))
196 return -E2BIG;
197
38ac38b2 198 if (posix_fallocate(f->fd, old_size, new_size - old_size) < 0)
cec736d2
LP		 199 return -errno;
200
201 if (fstat(f->fd, &f->last_stat) < 0)
202 return -errno;
203
204 f->header->arena_size = htole64(asize);
205
206 return 0;
207}
208
209static int journal_file_map(
210 JournalFile *f,
211 uint64_t offset,
212 uint64_t size,
213 void **_window,
214 uint64_t *_woffset,
215 uint64_t *_wsize,
216 void **ret) {
217
218 uint64_t woffset, wsize;
219 void *window;
220
221 assert(f);
222 assert(size > 0);
223 assert(ret);
224
225 woffset = offset & ~((uint64_t) page_size() - 1ULL);
226 wsize = size + (offset - woffset);
227 wsize = PAGE_ALIGN(wsize);
228
229 window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset);
230 if (window == MAP_FAILED)
231 return -errno;
232
233 if (_window)
234 *_window = window;
235
236 if (_woffset)
237 *_woffset = woffset;
238
239 if (_wsize)
240 *_wsize = wsize;
241
242 *ret = (uint8_t*) window + (offset - woffset);
243
244 return 0;
245}
246
247static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size, void **ret) {
248 void *p;
249 uint64_t delta;
250 int r;
251
252 assert(f);
253 assert(ret);
254
255 if (_likely_(f->window &&
256 f->window_offset <= offset &&
257 f->window_offset+f->window_size >= offset + size)) {
258
259 *ret = (uint8_t*) f->window + (offset - f->window_offset);
260 return 0;
261 }
262
263 if (f->window) {
264 if (munmap(f->window, f->window_size) < 0)
265 return -errno;
266
267 f->window = NULL;
268 f->window_size = f->window_offset = 0;
269 }
270
271 if (size < DEFAULT_WINDOW_SIZE) {
272 /* If the default window size is larger then what was
273 * asked for extend the mapping a bit in the hope to
274 * minimize needed remappings later on. We add half
275 * the window space before and half behind the
276 * requested mapping */
277
278 delta = PAGE_ALIGN((DEFAULT_WINDOW_SIZE - size) / 2);
279
280 if (offset < delta)
281 delta = offset;
282
283 offset -= delta;
284 size += (DEFAULT_WINDOW_SIZE - delta);
285 } else
286 delta = 0;
287
288 r = journal_file_map(f,
289 offset, size,
290 &f->window, &f->window_offset, &f->window_size,
291 & p);
292
293 if (r < 0)
294 return r;
295
296 *ret = (uint8_t*) p + delta;
297 return 0;
298}
299
300static bool verify_hash(Object *o) {
301 uint64_t t;
302
303 assert(o);
304
305 t = le64toh(o->object.type);
306 if (t == OBJECT_DATA) {
307 uint64_t s, h1, h2;
308
309 s = le64toh(o->object.size);
310
311 h1 = le64toh(o->data.hash);
312 h2 = hash64(o->data.payload, s - offsetof(Object, data.payload));
313
314 return h1 == h2;
315 }
316
317 return true;
318}
319
320int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Object **ret) {
321 int r;
322 void *t;
323 Object *o;
324 uint64_t s;
325
326 assert(f);
327 assert(ret);
328
329 r = journal_file_move_to(f, offset, sizeof(ObjectHeader), &t);
330 if (r < 0)
331 return r;
332
333 o = (Object*) t;
334 s = le64toh(o->object.size);
335
336 if (s < sizeof(ObjectHeader))
337 return -EBADMSG;
338
339 if (type >= 0 && le64toh(o->object.type) != type)
340 return -EBADMSG;
341
342 if (s > sizeof(ObjectHeader)) {
343 r = journal_file_move_to(f, offset, s, &t);
344 if (r < 0)
345 return r;
346
347 o = (Object*) t;
348 }
349
350 if (!verify_hash(o))
351 return -EBADMSG;
352
353 *ret = o;
354 return 0;
355}
356
357static uint64_t journal_file_seqnum(JournalFile *f) {
358 uint64_t r;
359
360 assert(f);
361
362 r = le64toh(f->header->seqnum) + 1;
363 f->header->seqnum = htole64(r);
364
365 return r;
366}
367
368static int journal_file_append_object(JournalFile *f, uint64_t size, Object **ret, uint64_t *offset) {
369 int r;
370 uint64_t p;
371 Object *tail, *o;
372 void *t;
373
374 assert(f);
375 assert(size >= sizeof(ObjectHeader));
376 assert(offset);
377 assert(ret);
378
379 p = le64toh(f->header->tail_object_offset);
380
381 if (p == 0)
382 p = le64toh(f->header->arena_offset);
383 else {
384 r = journal_file_move_to_object(f, p, -1, &tail);
385 if (r < 0)
386 return r;
387
388 p += ALIGN64(le64toh(tail->object.size));
389 }
390
391 r = journal_file_allocate(f, p, size);
392 if (r < 0)
393 return r;
394
395 r = journal_file_move_to(f, p, size, &t);
396 if (r < 0)
397 return r;
398
399 o = (Object*) t;
400
401 zero(o->object);
402 o->object.type = htole64(OBJECT_UNUSED);
403 zero(o->object.reserved);
404 o->object.size = htole64(size);
405
406 f->header->tail_object_offset = htole64(p);
407 if (f->header->head_object_offset == 0)
408 f->header->head_object_offset = htole64(p);
409
410 f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1);
411
412 *ret = o;
413 *offset = p;
414
415 return 0;
416}
417
418static int journal_file_setup_hash_table(JournalFile *f) {
419 uint64_t s, p;
420 Object *o;
421 int r;
422
423 assert(f);
424
425 s = DEFAULT_HASH_TABLE_SIZE;
426 r = journal_file_append_object(f, offsetof(Object, hash_table.table) + s, &o, &p);
427 if (r < 0)
428 return r;
429
430 o->object.type = htole64(OBJECT_HASH_TABLE);
431 memset(o->hash_table.table, 0, s);
432
433 f->header->hash_table_offset = htole64(p + offsetof(Object, hash_table.table));
434 f->header->hash_table_size = htole64(s);
435
436 return 0;
437}
438
439static int journal_file_setup_bisect_table(JournalFile *f) {
440 uint64_t s, p;
441 Object *o;
442 int r;
443
444 assert(f);
445
446 s = DEFAULT_BISECT_TABLE_SIZE;
447 r = journal_file_append_object(f, offsetof(Object, bisect_table.table) + s, &o, &p);
448 if (r < 0)
449 return r;
450
451 o->object.type = htole64(OBJECT_BISECT_TABLE);
452 memset(o->bisect_table.table, 0, s);
453
454 f->header->bisect_table_offset = htole64(p + offsetof(Object, bisect_table.table));
455 f->header->bisect_table_size = htole64(s);
456
457 return 0;
458}
459
460static int journal_file_map_hash_table(JournalFile *f) {
461 uint64_t s, p;
462 void *t;
463 int r;
464
465 assert(f);
466
467 p = le64toh(f->header->hash_table_offset);
468 s = le64toh(f->header->hash_table_size);
469
470 r = journal_file_map(f,
471 p, s,
472 &f->hash_table_window, NULL, &f->hash_table_window_size,
473 &t);
474 if (r < 0)
475 return r;
476
477 f->hash_table = t;
478 return 0;
479}
480
481static int journal_file_map_bisect_table(JournalFile *f) {
482 uint64_t s, p;
483 void *t;
484 int r;
485
486 assert(f);
487
488 p = le64toh(f->header->bisect_table_offset);
489 s = le64toh(f->header->bisect_table_size);
490
491 r = journal_file_map(f,
492 p, s,
493 &f->bisect_table_window, NULL, &f->bisect_table_window_size,
494 &t);
495
496 if (r < 0)
497 return r;
498
499 f->bisect_table = t;
500 return 0;
501}
502
503static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash_index) {
504 uint64_t p;
505 int r;
506
507 assert(f);
508 assert(o);
509 assert(offset > 0);
510 assert(o->object.type == htole64(OBJECT_DATA));
511
512 o->data.head_entry_offset = o->data.tail_entry_offset = 0;
513 o->data.next_hash_offset = 0;
514
515 p = le64toh(f->hash_table[hash_index].tail_hash_offset);
516 if (p == 0) {
517 /* Only entry in the hash table is easy */
518
519 o->data.prev_hash_offset = 0;
520 f->hash_table[hash_index].head_hash_offset = htole64(offset);
521 } else {
522 o->data.prev_hash_offset = htole64(p);
523
524 /* Temporarily move back to the previous data object,
525 * to patch in pointer */
526
527 r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
528 if (r < 0)
529 return r;
530
531 o->data.next_hash_offset = offset;
532
533 r = journal_file_move_to_object(f, offset, OBJECT_DATA, &o);
534 if (r < 0)
535 return r;
536 }
537
538 f->hash_table[hash_index].tail_hash_offset = htole64(offset);
539
540 return 0;
541}
542
543static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
544 uint64_t hash, h, p, np;
545 uint64_t osize;
546 Object *o;
547 int r;
548
549 assert(f);
550 assert(data || size == 0);
551
552 osize = offsetof(Object, data.payload) + size;
553
554 hash = hash64(data, size);
555 h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
556 p = le64toh(f->hash_table[h].head_hash_offset);
557
558 while (p != 0) {
559 /* Look for this data object in the hash table */
560
561 r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
562 if (r < 0)
563 return r;
564
565 if (le64toh(o->object.size) == osize &&
566 memcmp(o->data.payload, data, size) == 0) {
567
568 if (le64toh(o->data.hash) != hash)
569 return -EBADMSG;
570
571 if (ret)
572 *ret = o;
573
574 if (offset)
575 *offset = p;
576
577 return 0;
578 }
579
580 p = le64toh(o->data.next_hash_offset);
581 }
582
583 r = journal_file_append_object(f, osize, &o, &np);
584 if (r < 0)
585 return r;
586
587 o->object.type = htole64(OBJECT_DATA);
588 o->data.hash = htole64(hash);
589 memcpy(o->data.payload, data, size);
590
591 r = journal_file_link_data(f, o, np, h);
592 if (r < 0)
593 return r;
594
595 if (ret)
596 *ret = o;
597
598 if (offset)
599 *offset = np;
600
601 return 0;
602}
603
604uint64_t journal_file_entry_n_items(Object *o) {
605 assert(o);
606 assert(o->object.type == htole64(OBJECT_ENTRY));
607
608 return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
609}
610
611static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
612 uint64_t p, q;
613 int r;
614 assert(f);
615 assert(o);
616 assert(offset > 0);
617
618 p = le64toh(o->entry.items[i].object_offset);
619 if (p == 0)
620 return -EINVAL;
621
622 o->entry.items[i].next_entry_offset = 0;
623
624 /* Move to the data object */
625 r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
626 if (r < 0)
627 return r;
628
629 q = le64toh(o->data.tail_entry_offset);
630 o->data.tail_entry_offset = htole64(offset);
631
632 if (q == 0)
633 o->data.head_entry_offset = htole64(offset);
634 else {
635 uint64_t n, j;
636
637 /* Move to previous entry */
638 r = journal_file_move_to_object(f, q, OBJECT_ENTRY, &o);
639 if (r < 0)
640 return r;
641
642 n = journal_file_entry_n_items(o);
643 for (j = 0; j < n; j++)
644 if (le64toh(o->entry.items[j].object_offset) == p)
645 break;
646
647 if (j >= n)
648 return -EBADMSG;
649
650 o->entry.items[j].next_entry_offset = offset;
651 }
652
653 /* Move back to original entry */
654 r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
655 if (r < 0)
656 return r;
657
658 o->entry.items[i].prev_entry_offset = q;
659 return 0;
660}
661
662static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
663 uint64_t p, i, n, k, a, b;
664 int r;
665
666 assert(f);
667 assert(o);
668 assert(offset > 0);
669 assert(o->object.type == htole64(OBJECT_ENTRY));
670
671 /* Link up the entry itself */
672 p = le64toh(f->header->tail_entry_offset);
673
674 o->entry.prev_entry_offset = f->header->tail_entry_offset;
675 o->entry.next_entry_offset = 0;
676
677 if (p == 0)
678 f->header->head_entry_offset = htole64(offset);
679 else {
680 /* Temporarily move back to the previous entry, to
681 * patch in pointer */
682
683 r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
684 if (r < 0)
685 return r;
686
687 o->entry.next_entry_offset = htole64(offset);
688
689 r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
690 if (r < 0)
691 return r;
692 }
693
694 f->header->tail_entry_offset = htole64(offset);
695
696 /* Link up the items */
697 n = journal_file_entry_n_items(o);
698 for (i = 0; i < n; i++) {
699 r = journal_file_link_entry_item(f, o, offset, i);
700 if (r < 0)
701 return r;
702 }
703
704 /* Link up the entry in the bisect table */
705 n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
706 k = le64toh(f->header->arena_max_size) / n;
707
708 a = (le64toh(f->header->last_bisect_offset) + k - 1) / k;
709 b = offset / k;
710
711 for (; a <= b; a++)
712 f->bisect_table[a] = htole64(offset);
713
714 f->header->last_bisect_offset = htole64(offset + le64toh(o->object.size));
715
716 return 0;
717}
718
719static int journal_file_append_entry_internal(
720 JournalFile *f,
721 const dual_timestamp *ts,
722 uint64_t xor_hash,
723 const EntryItem items[], unsigned n_items,
724 Object **ret, uint64_t *offset) {
725 uint64_t np;
726 uint64_t osize;
727 Object *o;
728 int r;
729
730 assert(f);
731 assert(items || n_items == 0);
732
733 osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem));
734
735 r = journal_file_append_object(f, osize, &o, &np);
736 if (r < 0)
737 return r;
738
739 o->object.type = htole64(OBJECT_ENTRY);
740 o->entry.seqnum = htole64(journal_file_seqnum(f));
741 memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
742 o->entry.realtime = ts ? htole64(ts->realtime) : 0;
743 o->entry.monotonic = ts ? htole64(ts->monotonic) : 0;
744 o->entry.xor_hash = htole64(xor_hash);
745 o->entry.boot_id = f->header->boot_id;
746
747 r = journal_file_link_entry(f, o, np);
748 if (r < 0)
749 return r;
750
751 if (ret)
752 *ret = o;
753
754 if (offset)
755 *offset = np;
756
757 return 0;
758}
759
760int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, Object **ret, uint64_t *offset) {
761 unsigned i;
762 EntryItem *items;
763 int r;
764 uint64_t xor_hash = 0;
765
766 assert(f);
767 assert(iovec || n_iovec == 0);
768
769 items = new(EntryItem, n_iovec);
770 if (!items)
771 return -ENOMEM;
772
773 for (i = 0; i < n_iovec; i++) {
774 uint64_t p;
775 Object *o;
776
777 r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p);
778 if (r < 0)
779 goto finish;
780
781 xor_hash ^= le64toh(o->data.hash);
782 items[i].object_offset = htole64(p);
783 }
784
785 r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, ret, offset);
786
787finish:
788 free(items);
789
790 return r;
791}
792
793int journal_file_move_to_entry(JournalFile *f, uint64_t seqnum, Object **ret, uint64_t *offset) {
794 Object *o;
795 uint64_t lower, upper, p, n, k;
796 int r;
797
798 assert(f);
799
800 n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
801 k = le64toh(f->header->arena_max_size) / n;
802
803 lower = 0;
804 upper = le64toh(f->header->last_bisect_offset)/k+1;
805
806 while (lower < upper) {
807 k = (upper + lower) / 2;
808 p = le64toh(f->bisect_table[k]);
809
810 if (p == 0) {
811 upper = k;
812 continue;
813 }
814
815 r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
816 if (r < 0)
817 return r;
818
819 if (o->entry.seqnum == seqnum) {
820 if (ret)
821 *ret = o;
822
823 if (offset)
824 *offset = p;
825
826 return 1;
827 } else if (seqnum < o->entry.seqnum)
828 upper = k;
829 else if (seqnum > o->entry.seqnum)
830 lower = k+1;
831 }
832
833 assert(lower == upper);
834
835 if (lower <= 0)
836 return 0;
837
838 /* The object we are looking for is between
839 * bisect_table[lower-1] and bisect_table[lower] */
840
841 p = le64toh(f->bisect_table[lower-1]);
842
843 for (;;) {
844 r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
845 if (r < 0)
846 return r;
847
848 if (o->entry.seqnum == seqnum) {
849 if (ret)
850 *ret = o;
851
852 if (offset)
853 *offset = p;
854
855 return 1;
856
857 } if (seqnum < o->entry.seqnum)
858 return 0;
859
860 if (o->entry.next_entry_offset == 0)
861 return 0;
862
863 p = le64toh(o->entry.next_entry_offset);
864 }
865
866 return 0;
867}
868
869int journal_file_next_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) {
870 uint64_t np;
871 int r;
872
873 assert(f);
874
875 if (!o)
876 np = le64toh(f->header->head_entry_offset);
877 else {
878 if (le64toh(o->object.type) != OBJECT_ENTRY)
879 return -EINVAL;
880
881 np = le64toh(o->entry.next_entry_offset);
882 }
883
884 if (np == 0)
885 return 0;
886
887 r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
888 if (r < 0)
889 return r;
890
891 if (ret)
892 *ret = o;
893
894 if (offset)
895 *offset = np;
896
897 return 1;
898}
899
900int journal_file_prev_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) {
901 uint64_t np;
902 int r;
903
904 assert(f);
905
906 if (!o)
907 np = le64toh(f->header->tail_entry_offset);
908 else {
909 if (le64toh(o->object.type) != OBJECT_ENTRY)
910 return -EINVAL;
911
912 np = le64toh(o->entry.prev_entry_offset);
913 }
914
915 if (np == 0)
916 return 0;
917
918 r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
919 if (r < 0)
920 return r;
921
922 if (ret)
923 *ret = o;
924
925 if (offset)
926 *offset = np;
927
928 return 1;
929}
930
931int journal_file_find_first_entry(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
932 uint64_t p, osize, hash, h;
933 int r;
934
935 assert(f);
936 assert(data || size == 0);
937
938 osize = offsetof(Object, data.payload) + size;
939
940 hash = hash64(data, size);
941 h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
942 p = le64toh(f->hash_table[h].head_hash_offset);
943
944 while (p != 0) {
945 Object *o;
946
947 r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
948 if (r < 0)
949 return r;
950
951 if (le64toh(o->object.size) == osize &&
952 memcmp(o->data.payload, data, size) == 0) {
953
954 if (le64toh(o->data.hash) != hash)
955 return -EBADMSG;
956
957 if (o->data.head_entry_offset == 0)
958 return 0;
959
960 p = le64toh(o->data.head_entry_offset);
961 r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
962 if (r < 0)
963 return r;
964
965 if (ret)
966 *ret = o;
967
968 if (offset)
969 *offset = p;
970
971 return 1;
972 }
973
974 p = le64toh(o->data.next_hash_offset);
975 }
976
977 return 0;
978}
979
980int journal_file_find_last_entry(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
981 uint64_t p, osize, hash, h;
982 int r;
983
984 assert(f);
985 assert(data || size == 0);
986
987 osize = offsetof(Object, data.payload) + size;
988
989 hash = hash64(data, size);
990 h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
991 p = le64toh(f->hash_table[h].tail_hash_offset);
992
993 while (p != 0) {
994 Object *o;
995
996 r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
997 if (r < 0)
998 return r;
999
1000 if (le64toh(o->object.size) == osize &&
1001 memcmp(o->data.payload, data, size) == 0) {
1002
1003 if (le64toh(o->data.hash) != hash)
1004 return -EBADMSG;
1005
1006 if (o->data.tail_entry_offset == 0)
1007 return 0;
1008
1009 p = le64toh(o->data.tail_entry_offset);
1010 r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
1011 if (r < 0)
1012 return r;
1013
1014 if (ret)
1015 *ret = o;
1016
1017 if (offset)
1018 *offset = p;
1019
1020 return 1;
1021 }
1022
1023 p = le64toh(o->data.prev_hash_offset);
1024 }
1025
1026 return 0;
1027}
1028
1029void journal_file_dump(JournalFile *f) {
1030 char a[33], b[33], c[33];
1031 Object *o;
1032 int r;
1033 uint64_t p;
1034
1035 assert(f);
1036
1037 printf("File ID: %s\n"
1038 "Machine ID: %s\n"
1039 "Boot ID: %s\n"
1040 "Arena size: %llu\n",
1041 sd_id128_to_string(f->header->file_id, a),
1042 sd_id128_to_string(f->header->machine_id, b),
1043 sd_id128_to_string(f->header->boot_id, c),
1044 (unsigned long long) le64toh(f->header->arena_size));
1045
1046 p = le64toh(f->header->head_object_offset);
1047 while (p != 0) {
1048 r = journal_file_move_to_object(f, p, -1, &o);
1049 if (r < 0)
1050 goto fail;
1051
1052 switch (o->object.type) {
1053
1054 case OBJECT_UNUSED:
1055 printf("Type: OBJECT_UNUSED\n");
1056 break;
1057
1058 case OBJECT_DATA:
1059 printf("Type: OBJECT_DATA\n");
1060 break;
1061
1062 case OBJECT_ENTRY:
1063 printf("Type: OBJECT_ENTRY %llu\n", (unsigned long long) le64toh(o->entry.seqnum));
1064 break;
1065
1066 case OBJECT_HASH_TABLE:
1067 printf("Type: OBJECT_HASH_TABLE\n");
1068 break;
1069
1070 case OBJECT_BISECT_TABLE:
1071 printf("Type: OBJECT_BISECT_TABLE\n");
1072 break;
1073 }
1074
1075 if (p == le64toh(f->header->tail_object_offset))
1076 p = 0;
1077 else
1078 p = p + ALIGN64(le64toh(o->object.size));
1079 }
1080
1081 return;
1082fail:
1083 log_error("File corrupt");
1084}
1085
1086int journal_file_open(
1087 const char *fname,
1088 int flags,
1089 mode_t mode,
1090 JournalFile **ret) {
1091
1092 JournalFile *f;
1093 int r;
1094 bool newly_created = false;
1095
1096 assert(fname);
1097
1098 if ((flags & O_ACCMODE) != O_RDONLY &&
1099 (flags & O_ACCMODE) != O_RDWR)
1100 return -EINVAL;
1101
1102 f = new0(JournalFile, 1);
1103 if (!f)
1104 return -ENOMEM;
1105
1106 f->writable = (flags & O_ACCMODE) != O_RDONLY;
1107 f->prot = prot_from_flags(flags);
1108
1109 f->fd = open(fname, flags|O_CLOEXEC, mode);
1110 if (f->fd < 0) {
1111 r = -errno;
1112 goto fail;
1113 }
1114
1115 f->path = strdup(fname);
1116 if (!f->path) {
1117 r = -ENOMEM;
1118 goto fail;
1119 }
1120
1121 if (fstat(f->fd, &f->last_stat) < 0) {
1122 r = -errno;
1123 goto fail;
1124 }
1125
1126 if (f->last_stat.st_size == 0 && f->writable) {
1127 newly_created = true;
1128
1129 r = journal_file_init_header(f);
1130 if (r < 0)
1131 goto fail;
1132
1133 if (fstat(f->fd, &f->last_stat) < 0) {
1134 r = -errno;
1135 goto fail;
1136 }
1137 }
1138
1139 if (f->last_stat.st_size < (off_t) sizeof(Header)) {
1140 r = -EIO;
1141 goto fail;
1142 }
1143
1144 f->header = mmap(NULL, PAGE_ALIGN(sizeof(Header)), prot_from_flags(flags), MAP_SHARED, f->fd, 0);
1145 if (f->header == MAP_FAILED) {
1146 f->header = NULL;
1147 r = -errno;
1148 goto fail;
1149 }
1150
1151 if (!newly_created) {
1152 r = journal_file_verify_header(f);
1153 if (r < 0)
1154 goto fail;
1155 }
1156
1157 if (f->writable) {
1158 r = journal_file_refresh_header(f);
1159 if (r < 0)
1160 goto fail;
1161 }
1162
1163 if (newly_created) {
1164
1165 r = journal_file_setup_hash_table(f);
1166 if (r < 0)
1167 goto fail;
1168
1169 r = journal_file_setup_bisect_table(f);
1170 if (r < 0)
1171 goto fail;
1172 }
1173
1174 r = journal_file_map_hash_table(f);
1175 if (r < 0)
1176 goto fail;
1177
1178 r = journal_file_map_bisect_table(f);
1179 if (r < 0)
1180 goto fail;
1181
1182 if (ret)
1183 *ret = f;
1184
1185 return 0;
1186
1187fail:
1188 journal_file_close(f);
1189
1190 return r;
1191}
Unnamed repository; edit this file 'description' to name the repository.