]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/journal/journal-file.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
journal: fix hash table lookup logic
[thirdparty/systemd.git] / src / journal / journal-file.c
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 #include "compress.h"
34
35 #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
36 #define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
37
38 #define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
39
40 #define COMPRESSION_SIZE_THRESHOLD (64ULL)
41
42 static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
43
44 #define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
45
46 void journal_file_close(JournalFile *f) {
47 int t;
48
49 assert(f);
50
51 if (f->header && f->writable)
52 f->header->state = STATE_OFFLINE;
53
54
55 for (t = 0; t < _WINDOW_MAX; t++)
56 if (f->windows[t].ptr)
57 munmap(f->windows[t].ptr, f->windows[t].size);
58
59 if (f->fd >= 0)
60 close_nointr_nofail(f->fd);
61
62 free(f->path);
63
64 #ifdef HAVE_XZ
65 free(f->compress_buffer);
66 #endif
67
68 free(f);
69 }
70
71 static int journal_file_init_header(JournalFile *f, JournalFile *template) {
72 Header h;
73 ssize_t k;
74 int r;
75
76 assert(f);
77
78 zero(h);
79 memcpy(h.signature, signature, 8);
80 h.arena_offset = htole64(ALIGN64(sizeof(h)));
81
82 r = sd_id128_randomize(&h.file_id);
83 if (r < 0)
84 return r;
85
86 if (template) {
87 h.seqnum_id = template->header->seqnum_id;
88 h.seqnum = template->header->seqnum;
89 } else
90 h.seqnum_id = h.file_id;
91
92 k = pwrite(f->fd, &h, sizeof(h), 0);
93 if (k < 0)
94 return -errno;
95
96 if (k != sizeof(h))
97 return -EIO;
98
99 return 0;
100 }
101
102 static int journal_file_refresh_header(JournalFile *f) {
103 int r;
104 sd_id128_t boot_id;
105
106 assert(f);
107
108 r = sd_id128_get_machine(&f->header->machine_id);
109 if (r < 0)
110 return r;
111
112 r = sd_id128_get_boot(&boot_id);
113 if (r < 0)
114 return r;
115
116 if (sd_id128_equal(boot_id, f->header->boot_id))
117 f->tail_entry_monotonic_valid = true;
118
119 f->header->boot_id = boot_id;
120
121 f->header->state = STATE_ONLINE;
122 return 0;
123 }
124
125 static int journal_file_verify_header(JournalFile *f) {
126 assert(f);
127
128 if (memcmp(f->header, signature, 8))
129 return -EBADMSG;
130
131 #ifdef HAVE_XZ
132 if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
133 return -EPROTONOSUPPORT;
134 #else
135 if (f->header->incompatible_flags != 0)
136 return -EPROTONOSUPPORT;
137 #endif
138
139 if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->arena_offset) + le64toh(f->header->arena_size)))
140 return -ENODATA;
141
142 if (f->writable) {
143 uint32_t state;
144 sd_id128_t machine_id;
145 int r;
146
147 r = sd_id128_get_machine(&machine_id);
148 if (r < 0)
149 return r;
150
151 if (!sd_id128_equal(machine_id, f->header->machine_id))
152 return -EHOSTDOWN;
153
154 state = f->header->state;
155
156 if (state == STATE_ONLINE)
157 log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
158 else if (state == STATE_ARCHIVED)
159 return -ESHUTDOWN;
160 else if (state != STATE_OFFLINE)
161 log_debug("Journal file %s has unknown state %u. Ignoring.", f->path, state);
162 }
163
164 return 0;
165 }
166
167 static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
168 uint64_t old_size, new_size;
169
170 assert(f);
171
172 /* We assume that this file is not sparse, and we know that
173 * for sure, since we always call posix_fallocate()
174 * ourselves */
175
176 old_size =
177 le64toh(f->header->arena_offset) +
178 le64toh(f->header->arena_size);
179
180 new_size = PAGE_ALIGN(offset + size);
181 if (new_size < le64toh(f->header->arena_offset))
182 new_size = le64toh(f->header->arena_offset);
183
184 if (new_size <= old_size)
185 return 0;
186
187 if (f->metrics.max_size > 0 &&
188 new_size > f->metrics.max_size)
189 return -E2BIG;
190
191 if (new_size > f->metrics.min_size &&
192 f->metrics.keep_free > 0) {
193 struct statvfs svfs;
194
195 if (fstatvfs(f->fd, &svfs) >= 0) {
196 uint64_t available;
197
198 available = svfs.f_bfree * svfs.f_bsize;
199
200 if (available >= f->metrics.keep_free)
201 available -= f->metrics.keep_free;
202 else
203 available = 0;
204
205 if (new_size - old_size > available)
206 return -E2BIG;
207 }
208 }
209
210 /* Note that the glibc fallocate() fallback is very
211 inefficient, hence we try to minimize the allocation area
212 as we can. */
213 if (posix_fallocate(f->fd, old_size, new_size - old_size) < 0)
214 return -errno;
215
216 if (fstat(f->fd, &f->last_stat) < 0)
217 return -errno;
218
219 f->header->arena_size = new_size - htole64(f->header->arena_offset);
220
221 return 0;
222 }
223
224 static int journal_file_map(
225 JournalFile *f,
226 uint64_t offset,
227 uint64_t size,
228 void **_window,
229 uint64_t *_woffset,
230 uint64_t *_wsize,
231 void **ret) {
232
233 uint64_t woffset, wsize;
234 void *window;
235
236 assert(f);
237 assert(size > 0);
238 assert(ret);
239
240 woffset = offset & ~((uint64_t) page_size() - 1ULL);
241 wsize = size + (offset - woffset);
242 wsize = PAGE_ALIGN(wsize);
243
244 window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset);
245 if (window == MAP_FAILED)
246 return -errno;
247
248 if (_window)
249 *_window = window;
250
251 if (_woffset)
252 *_woffset = woffset;
253
254 if (_wsize)
255 *_wsize = wsize;
256
257 *ret = (uint8_t*) window + (offset - woffset);
258
259 return 0;
260 }
261
262 static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) {
263 void *p;
264 uint64_t delta;
265 int r;
266 Window *w;
267
268 assert(f);
269 assert(ret);
270 assert(wt >= 0);
271 assert(wt < _WINDOW_MAX);
272
273 w = f->windows + wt;
274
275 if (_likely_(w->ptr &&
276 w->offset <= offset &&
277 w->offset + w->size >= offset + size)) {
278
279 *ret = (uint8_t*) w->ptr + (offset - w->offset);
280 return 0;
281 }
282
283 if (w->ptr) {
284 if (munmap(w->ptr, w->size) < 0)
285 return -errno;
286
287 w->ptr = NULL;
288 w->size = w->offset = 0;
289 }
290
291 if (size < DEFAULT_WINDOW_SIZE) {
292 /* If the default window size is larger then what was
293 * asked for extend the mapping a bit in the hope to
294 * minimize needed remappings later on. We add half
295 * the window space before and half behind the
296 * requested mapping */
297
298 delta = PAGE_ALIGN((DEFAULT_WINDOW_SIZE - size) / 2);
299
300 if (offset < delta)
301 delta = offset;
302
303 offset -= delta;
304 size += (DEFAULT_WINDOW_SIZE - delta);
305 } else
306 delta = 0;
307
308 r = journal_file_map(f,
309 offset, size,
310 &w->ptr, &w->offset, &w->size,
311 &p);
312
313 if (r < 0)
314 return r;
315
316 *ret = (uint8_t*) p + delta;
317 return 0;
318 }
319
320 static bool verify_hash(Object *o) {
321 uint64_t h1, h2;
322
323 assert(o);
324
325 if (o->object.type == OBJECT_DATA && !(o->object.flags & OBJECT_COMPRESSED)) {
326 h1 = le64toh(o->data.hash);
327 h2 = hash64(o->data.payload, le64toh(o->object.size) - offsetof(Object, data.payload));
328 } else if (o->object.type == OBJECT_FIELD) {
329 h1 = le64toh(o->field.hash);
330 h2 = hash64(o->field.payload, le64toh(o->object.size) - offsetof(Object, field.payload));
331 } else
332 return true;
333
334 return h1 == h2;
335 }
336
337 int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
338 int r;
339 void *t;
340 Object *o;
341 uint64_t s;
342
343 assert(f);
344 assert(ret);
345 assert(type < _OBJECT_TYPE_MAX);
346
347 r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t);
348 if (r < 0)
349 return r;
350
351 o = (Object*) t;
352 s = le64toh(o->object.size);
353
354 if (s < sizeof(ObjectHeader))
355 return -EBADMSG;
356
357 if (type >= 0 && o->object.type != type)
358 return -EBADMSG;
359
360 if (s > sizeof(ObjectHeader)) {
361 r = journal_file_move_to(f, o->object.type, offset, s, &t);
362 if (r < 0)
363 return r;
364
365 o = (Object*) t;
366 }
367
368 if (!verify_hash(o))
369 return -EBADMSG;
370
371 *ret = o;
372 return 0;
373 }
374
375 static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
376 uint64_t r;
377
378 assert(f);
379
380 r = le64toh(f->header->seqnum) + 1;
381
382 if (seqnum) {
383 /* If an external seqnum counter was passed, we update
384 * both the local and the external one, and set it to
385 * the maximum of both */
386
387 if (*seqnum + 1 > r)
388 r = *seqnum + 1;
389
390 *seqnum = r;
391 }
392
393 f->header->seqnum = htole64(r);
394
395 if (f->header->first_seqnum == 0)
396 f->header->first_seqnum = htole64(r);
397
398 return r;
399 }
400
401 static int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object **ret, uint64_t *offset) {
402 int r;
403 uint64_t p;
404 Object *tail, *o;
405 void *t;
406
407 assert(f);
408 assert(size >= sizeof(ObjectHeader));
409 assert(offset);
410 assert(ret);
411
412 p = le64toh(f->header->tail_object_offset);
413 if (p == 0)
414 p = le64toh(f->header->arena_offset);
415 else {
416 r = journal_file_move_to_object(f, -1, p, &tail);
417 if (r < 0)
418 return r;
419
420 p += ALIGN64(le64toh(tail->object.size));
421 }
422
423 r = journal_file_allocate(f, p, size);
424 if (r < 0)
425 return r;
426
427 r = journal_file_move_to(f, type, p, size, &t);
428 if (r < 0)
429 return r;
430
431 o = (Object*) t;
432
433 zero(o->object);
434 o->object.type = type;
435 o->object.size = htole64(size);
436
437 f->header->tail_object_offset = htole64(p);
438 f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1);
439
440 *ret = o;
441 *offset = p;
442
443 return 0;
444 }
445
446 static int journal_file_setup_data_hash_table(JournalFile *f) {
447 uint64_t s, p;
448 Object *o;
449 int r;
450
451 assert(f);
452
453 s = DEFAULT_DATA_HASH_TABLE_SIZE;
454 r = journal_file_append_object(f,
455 OBJECT_DATA_HASH_TABLE,
456 offsetof(Object, hash_table.items) + s,
457 &o, &p);
458 if (r < 0)
459 return r;
460
461 memset(o->hash_table.items, 0, s);
462
463 f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
464 f->header->data_hash_table_size = htole64(s);
465
466 return 0;
467 }
468
469 static int journal_file_setup_field_hash_table(JournalFile *f) {
470 uint64_t s, p;
471 Object *o;
472 int r;
473
474 assert(f);
475
476 s = DEFAULT_FIELD_HASH_TABLE_SIZE;
477 r = journal_file_append_object(f,
478 OBJECT_FIELD_HASH_TABLE,
479 offsetof(Object, hash_table.items) + s,
480 &o, &p);
481 if (r < 0)
482 return r;
483
484 memset(o->hash_table.items, 0, s);
485
486 f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
487 f->header->field_hash_table_size = htole64(s);
488
489 return 0;
490 }
491
492 static int journal_file_map_data_hash_table(JournalFile *f) {
493 uint64_t s, p;
494 void *t;
495 int r;
496
497 assert(f);
498
499 p = le64toh(f->header->data_hash_table_offset);
500 s = le64toh(f->header->data_hash_table_size);
501
502 r = journal_file_move_to(f,
503 WINDOW_DATA_HASH_TABLE,
504 p, s,
505 &t);
506 if (r < 0)
507 return r;
508
509 f->data_hash_table = t;
510 return 0;
511 }
512
513 static int journal_file_map_field_hash_table(JournalFile *f) {
514 uint64_t s, p;
515 void *t;
516 int r;
517
518 assert(f);
519
520 p = le64toh(f->header->field_hash_table_offset);
521 s = le64toh(f->header->field_hash_table_size);
522
523 r = journal_file_move_to(f,
524 WINDOW_FIELD_HASH_TABLE,
525 p, s,
526 &t);
527 if (r < 0)
528 return r;
529
530 f->field_hash_table = t;
531 return 0;
532 }
533
534 static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash) {
535 uint64_t p, h;
536 int r;
537
538 assert(f);
539 assert(o);
540 assert(offset > 0);
541 assert(o->object.type == OBJECT_DATA);
542
543 o->data.next_hash_offset = o->data.next_field_offset = 0;
544 o->data.entry_offset = o->data.entry_array_offset = 0;
545 o->data.n_entries = 0;
546
547 h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
548 p = le64toh(f->data_hash_table[h].head_hash_offset);
549 if (p == 0) {
550 /* Only entry in the hash table is easy */
551 f->data_hash_table[h].head_hash_offset = htole64(offset);
552 } else {
553 /* Temporarily move back to the previous data object,
554 * to patch in pointer */
555
556 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
557 if (r < 0)
558 return r;
559
560 o->data.next_hash_offset = htole64(offset);
561
562 r = journal_file_move_to_object(f, OBJECT_DATA, offset, &o);
563 if (r < 0)
564 return r;
565 }
566
567 f->data_hash_table[h].tail_hash_offset = htole64(offset);
568
569 return 0;
570 }
571
572 int journal_file_find_data_object_with_hash(
573 JournalFile *f,
574 const void *data, uint64_t size, uint64_t hash,
575 Object **ret, uint64_t *offset) {
576 uint64_t p, osize, h;
577 int r;
578
579 assert(f);
580 assert(data || size == 0);
581
582 osize = offsetof(Object, data.payload) + size;
583
584 if (f->header->data_hash_table_size == 0)
585 return -EBADMSG;
586
587 h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
588 p = le64toh(f->data_hash_table[h].head_hash_offset);
589
590 while (p > 0) {
591 Object *o;
592
593 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
594 if (r < 0)
595 return r;
596
597 if (le64toh(o->data.hash) != hash)
598 goto next;
599
600 if (o->object.flags & OBJECT_COMPRESSED) {
601 #ifdef HAVE_XZ
602 uint64_t l, rsize;
603
604 l = le64toh(o->object.size);
605 if (l <= offsetof(Object, data.payload))
606 return -EBADMSG;
607
608 l -= offsetof(Object, data.payload);
609
610 if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize))
611 return -EBADMSG;
612
613 if (rsize == size &&
614 memcmp(f->compress_buffer, data, size) == 0) {
615
616 if (ret)
617 *ret = o;
618
619 if (offset)
620 *offset = p;
621
622 return 1;
623 }
624 #else
625 return -EPROTONOSUPPORT;
626 #endif
627
628 } else if (le64toh(o->object.size) == osize &&
629 memcmp(o->data.payload, data, size) == 0) {
630
631 if (ret)
632 *ret = o;
633
634 if (offset)
635 *offset = p;
636
637 return 1;
638 }
639
640 next:
641 p = le64toh(o->data.next_hash_offset);
642 }
643
644 return 0;
645 }
646
647 int journal_file_find_data_object(
648 JournalFile *f,
649 const void *data, uint64_t size,
650 Object **ret, uint64_t *offset) {
651
652 uint64_t hash;
653
654 assert(f);
655 assert(data || size == 0);
656
657 hash = hash64(data, size);
658
659 return journal_file_find_data_object_with_hash(f,
660 data, size, hash,
661 ret, offset);
662 }
663
664 static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
665 uint64_t hash, p;
666 uint64_t osize;
667 Object *o;
668 int r;
669 bool compressed = false;
670
671 assert(f);
672 assert(data || size == 0);
673
674 hash = hash64(data, size);
675
676 r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p);
677 if (r < 0)
678 return r;
679 else if (r > 0) {
680
681 if (ret)
682 *ret = o;
683
684 if (offset)
685 *offset = p;
686
687 return 0;
688 }
689
690 osize = offsetof(Object, data.payload) + size;
691 r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p);
692 if (r < 0)
693 return r;
694
695 o->data.hash = htole64(hash);
696
697 #ifdef HAVE_XZ
698 if (f->compress &&
699 size >= COMPRESSION_SIZE_THRESHOLD) {
700 uint64_t rsize;
701
702 compressed = compress_blob(data, size, o->data.payload, &rsize);
703
704 if (compressed) {
705 o->object.size = htole64(offsetof(Object, data.payload) + rsize);
706 o->object.flags |= OBJECT_COMPRESSED;
707
708 f->header->incompatible_flags = htole32(le32toh(f->header->incompatible_flags) | HEADER_INCOMPATIBLE_COMPRESSED);
709
710 log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize);
711 }
712 }
713 #endif
714
715 if (!compressed)
716 memcpy(o->data.payload, data, size);
717
718 r = journal_file_link_data(f, o, p, hash);
719 if (r < 0)
720 return r;
721
722 if (ret)
723 *ret = o;
724
725 if (offset)
726 *offset = p;
727
728 return 0;
729 }
730
731 uint64_t journal_file_entry_n_items(Object *o) {
732 assert(o);
733 assert(o->object.type == htole64(OBJECT_ENTRY));
734
735 return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
736 }
737
738 static uint64_t journal_file_entry_array_n_items(Object *o) {
739 assert(o);
740 assert(o->object.type == htole64(OBJECT_ENTRY_ARRAY));
741
742 return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
743 }
744
745 static int link_entry_into_array(JournalFile *f,
746 uint64_t *first,
747 uint64_t *idx,
748 uint64_t p) {
749 int r;
750 uint64_t n = 0, ap = 0, q, i, a, hidx;
751 Object *o;
752
753 assert(f);
754 assert(first);
755 assert(idx);
756 assert(p > 0);
757
758 a = le64toh(*first);
759 i = hidx = le64toh(*idx);
760 while (a > 0) {
761
762 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
763 if (r < 0)
764 return r;
765
766 n = journal_file_entry_array_n_items(o);
767 if (i < n) {
768 o->entry_array.items[i] = htole64(p);
769 *idx = htole64(hidx + 1);
770 return 0;
771 }
772
773 i -= n;
774 ap = a;
775 a = le64toh(o->entry_array.next_entry_array_offset);
776 }
777
778 if (hidx > n)
779 n = (hidx+1) * 2;
780 else
781 n = n * 2;
782
783 if (n < 4)
784 n = 4;
785
786 r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY,
787 offsetof(Object, entry_array.items) + n * sizeof(uint64_t),
788 &o, &q);
789 if (r < 0)
790 return r;
791
792 o->entry_array.items[i] = htole64(p);
793
794 if (ap == 0)
795 *first = q;
796 else {
797 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o);
798 if (r < 0)
799 return r;
800
801 o->entry_array.next_entry_array_offset = htole64(q);
802 }
803
804 *idx = htole64(hidx + 1);
805
806 return 0;
807 }
808
809 static int link_entry_into_array_plus_one(JournalFile *f,
810 uint64_t *extra,
811 uint64_t *first,
812 uint64_t *idx,
813 uint64_t p) {
814
815 int r;
816
817 assert(f);
818 assert(extra);
819 assert(first);
820 assert(idx);
821 assert(p > 0);
822
823 if (*idx == 0)
824 *extra = htole64(p);
825 else {
826 uint64_t i;
827
828 i = le64toh(*idx) - 1;
829 r = link_entry_into_array(f, first, &i, p);
830 if (r < 0)
831 return r;
832 }
833
834 *idx = htole64(le64toh(*idx) + 1);
835 return 0;
836 }
837
838 static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
839 uint64_t p;
840 int r;
841 assert(f);
842 assert(o);
843 assert(offset > 0);
844
845 p = le64toh(o->entry.items[i].object_offset);
846 if (p == 0)
847 return -EINVAL;
848
849 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
850 if (r < 0)
851 return r;
852
853 return link_entry_into_array_plus_one(f,
854 &o->data.entry_offset,
855 &o->data.entry_array_offset,
856 &o->data.n_entries,
857 offset);
858 }
859
860 static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
861 uint64_t n, i;
862 int r;
863
864 assert(f);
865 assert(o);
866 assert(offset > 0);
867 assert(o->object.type == OBJECT_ENTRY);
868
869 /* Link up the entry itself */
870 r = link_entry_into_array(f,
871 &f->header->entry_array_offset,
872 &f->header->n_entries,
873 offset);
874 if (r < 0)
875 return r;
876
877 log_error("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries);
878
879 if (f->header->head_entry_realtime == 0)
880 f->header->head_entry_realtime = o->entry.realtime;
881
882 f->header->tail_entry_realtime = o->entry.realtime;
883 f->header->tail_entry_monotonic = o->entry.monotonic;
884
885 f->tail_entry_monotonic_valid = true;
886
887 /* Link up the items */
888 n = journal_file_entry_n_items(o);
889 for (i = 0; i < n; i++) {
890 r = journal_file_link_entry_item(f, o, offset, i);
891 if (r < 0)
892 return r;
893 }
894
895 return 0;
896 }
897
898 static int journal_file_append_entry_internal(
899 JournalFile *f,
900 const dual_timestamp *ts,
901 uint64_t xor_hash,
902 const EntryItem items[], unsigned n_items,
903 uint64_t *seqnum,
904 Object **ret, uint64_t *offset) {
905 uint64_t np;
906 uint64_t osize;
907 Object *o;
908 int r;
909
910 assert(f);
911 assert(items || n_items == 0);
912 assert(ts);
913
914 osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem));
915
916 r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np);
917 if (r < 0)
918 return r;
919
920 o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum));
921 memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
922 o->entry.realtime = htole64(ts->realtime);
923 o->entry.monotonic = htole64(ts->monotonic);
924 o->entry.xor_hash = htole64(xor_hash);
925 o->entry.boot_id = f->header->boot_id;
926
927 r = journal_file_link_entry(f, o, np);
928 if (r < 0)
929 return r;
930
931 if (ret)
932 *ret = o;
933
934 if (offset)
935 *offset = np;
936
937 return 0;
938 }
939
940 static void journal_file_post_change(JournalFile *f) {
941 assert(f);
942
943 /* inotify() does not receive IN_MODIFY events from file
944 * accesses done via mmap(). After each access we hence
945 * trigger IN_MODIFY by truncating the journal file to its
946 * current size which triggers IN_MODIFY. */
947
948 __sync_synchronize();
949
950 if (ftruncate(f->fd, f->last_stat.st_size) < 0)
951 log_error("Failed to to truncate file to its own size: %m");
952 }
953
954 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) {
955 unsigned i;
956 EntryItem *items;
957 int r;
958 uint64_t xor_hash = 0;
959 struct dual_timestamp _ts;
960
961 assert(f);
962 assert(iovec || n_iovec == 0);
963
964 if (!f->writable)
965 return -EPERM;
966
967 if (!ts) {
968 dual_timestamp_get(&_ts);
969 ts = &_ts;
970 }
971
972 if (f->tail_entry_monotonic_valid &&
973 ts->monotonic < le64toh(f->header->tail_entry_monotonic))
974 return -EINVAL;
975
976 if (ts->realtime < le64toh(f->header->tail_entry_realtime))
977 return -EINVAL;
978
979 items = new(EntryItem, n_iovec);
980 if (!items)
981 return -ENOMEM;
982
983 for (i = 0; i < n_iovec; i++) {
984 uint64_t p;
985 Object *o;
986
987 r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p);
988 if (r < 0)
989 goto finish;
990
991 xor_hash ^= le64toh(o->data.hash);
992 items[i].object_offset = htole64(p);
993 items[i].hash = o->data.hash;
994 }
995
996 r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
997
998 journal_file_post_change(f);
999
1000 finish:
1001 free(items);
1002
1003 return r;
1004 }
1005
1006 static int generic_array_get(JournalFile *f,
1007 uint64_t first,
1008 uint64_t i,
1009 Object **ret, uint64_t *offset) {
1010
1011 Object *o;
1012 uint64_t p, a;
1013 int r;
1014
1015 assert(f);
1016
1017 a = first;
1018 while (a > 0) {
1019 uint64_t n;
1020
1021 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
1022 if (r < 0)
1023 return r;
1024
1025 n = journal_file_entry_array_n_items(o);
1026 if (i < n) {
1027 p = le64toh(o->entry_array.items[i]);
1028 break;
1029 }
1030
1031 i -= n;
1032 a = le64toh(o->entry_array.next_entry_array_offset);
1033 }
1034
1035 if (a <= 0 || p <= 0)
1036 return 0;
1037
1038 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
1039 if (r < 0)
1040 return r;
1041
1042 if (ret)
1043 *ret = o;
1044
1045 if (offset)
1046 *offset = p;
1047
1048 return 1;
1049 }
1050
1051 static int generic_array_get_plus_one(JournalFile *f,
1052 uint64_t extra,
1053 uint64_t first,
1054 uint64_t i,
1055 Object **ret, uint64_t *offset) {
1056
1057 Object *o;
1058
1059 assert(f);
1060
1061 if (i == 0) {
1062 int r;
1063
1064 r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
1065 if (r < 0)
1066 return r;
1067
1068 if (ret)
1069 *ret = o;
1070
1071 if (offset)
1072 *offset = extra;
1073
1074 return 1;
1075 }
1076
1077 return generic_array_get(f, first, i-1, ret, offset);
1078 }
1079
1080 enum {
1081 TEST_FOUND,
1082 TEST_LEFT,
1083 TEST_RIGHT
1084 };
1085
1086 static int generic_array_bisect(JournalFile *f,
1087 uint64_t first,
1088 uint64_t n,
1089 uint64_t needle,
1090 int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
1091 direction_t direction,
1092 Object **ret,
1093 uint64_t *offset,
1094 uint64_t *idx) {
1095
1096 uint64_t a, p, t = 0, i = 0, last_p = 0;
1097 bool subtract_one = false;
1098 Object *o, *array = NULL;
1099 int r;
1100
1101 assert(f);
1102 assert(test_object);
1103
1104 a = first;
1105 while (a > 0) {
1106 uint64_t left, right, k, lp;
1107
1108 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array);
1109 if (r < 0)
1110 return r;
1111
1112 k = journal_file_entry_array_n_items(array);
1113 right = MIN(k, n);
1114 if (right <= 0)
1115 return 0;
1116
1117 i = right - 1;
1118 lp = p = le64toh(array->entry_array.items[i]);
1119 if (p <= 0)
1120 return -EBADMSG;
1121
1122 r = test_object(f, p, needle);
1123 if (r < 0)
1124 return r;
1125
1126 if (r == TEST_FOUND)
1127 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
1128
1129 if (r == TEST_RIGHT) {
1130 left = 0;
1131 right -= 1;
1132 for (;;) {
1133 if (left == right) {
1134 if (direction == DIRECTION_UP)
1135 subtract_one = true;
1136
1137 i = left;
1138 goto found;
1139 }
1140
1141 assert(left < right);
1142
1143 i = (left + right) / 2;
1144 p = le64toh(array->entry_array.items[i]);
1145 if (p <= 0)
1146 return -EBADMSG;
1147
1148 r = test_object(f, p, needle);
1149 if (r < 0)
1150 return r;
1151
1152 if (r == TEST_FOUND)
1153 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
1154
1155 if (r == TEST_RIGHT)
1156 right = i;
1157 else
1158 left = i + 1;
1159 }
1160 }
1161
1162 if (k > n)
1163 return 0;
1164
1165 last_p = lp;
1166
1167 n -= k;
1168 t += k;
1169 a = le64toh(array->entry_array.next_entry_array_offset);
1170 }
1171
1172 return 0;
1173
1174 found:
1175 if (subtract_one && t == 0 && i == 0)
1176 return 0;
1177
1178 if (subtract_one && i == 0)
1179 p = last_p;
1180 else if (subtract_one)
1181 p = le64toh(array->entry_array.items[i-1]);
1182 else
1183 p = le64toh(array->entry_array.items[i]);
1184
1185 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
1186 if (r < 0)
1187 return r;
1188
1189 if (ret)
1190 *ret = o;
1191
1192 if (offset)
1193 *offset = p;
1194
1195 if (idx)
1196 *idx = t + i - (subtract_one ? 1 : 0);
1197
1198 return 1;
1199 }
1200
1201 static int generic_array_bisect_plus_one(JournalFile *f,
1202 uint64_t extra,
1203 uint64_t first,
1204 uint64_t n,
1205 uint64_t needle,
1206 int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
1207 direction_t direction,
1208 Object **ret,
1209 uint64_t *offset,
1210 uint64_t *idx) {
1211
1212 int r;
1213
1214 assert(f);
1215 assert(test_object);
1216
1217 if (n <= 0)
1218 return 0;
1219
1220 /* This bisects the array in object 'first', but first checks
1221 * an extra */
1222 r = test_object(f, extra, needle);
1223 if (r < 0)
1224 return r;
1225 else if (r == TEST_FOUND) {
1226 Object *o;
1227
1228 r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
1229 if (r < 0)
1230 return r;
1231
1232 if (ret)
1233 *ret = o;
1234
1235 if (offset)
1236 *offset = extra;
1237
1238 if (idx)
1239 *idx = 0;
1240
1241 return 1;
1242 } else if (r == TEST_RIGHT)
1243 return 0;
1244
1245 r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx);
1246
1247 if (r > 0)
1248 (*idx) ++;
1249
1250 return r;
1251 }
1252
1253 static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
1254 Object *o;
1255 int r;
1256
1257 assert(f);
1258 assert(p > 0);
1259
1260 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
1261 if (r < 0)
1262 return r;
1263
1264 if (le64toh(o->entry.seqnum) == needle)
1265 return TEST_FOUND;
1266 else if (le64toh(o->entry.seqnum) < needle)
1267 return TEST_LEFT;
1268 else
1269 return TEST_RIGHT;
1270 }
1271
1272 int journal_file_move_to_entry_by_seqnum(
1273 JournalFile *f,
1274 uint64_t seqnum,
1275 direction_t direction,
1276 Object **ret,
1277 uint64_t *offset) {
1278
1279 return generic_array_bisect(f,
1280 le64toh(f->header->entry_array_offset),
1281 le64toh(f->header->n_entries),
1282 seqnum,
1283 test_object_seqnum,
1284 direction,
1285 ret, offset, NULL);
1286 }
1287
1288 static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) {
1289 Object *o;
1290 int r;
1291
1292 assert(f);
1293 assert(p > 0);
1294
1295 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
1296 if (r < 0)
1297 return r;
1298
1299 if (le64toh(o->entry.realtime) == needle)
1300 return TEST_FOUND;
1301 else if (le64toh(o->entry.realtime) < needle)
1302 return TEST_LEFT;
1303 else
1304 return TEST_RIGHT;
1305 }
1306
1307 int journal_file_move_to_entry_by_realtime(
1308 JournalFile *f,
1309 uint64_t realtime,
1310 direction_t direction,
1311 Object **ret,
1312 uint64_t *offset) {
1313
1314 return generic_array_bisect(f,
1315 le64toh(f->header->entry_array_offset),
1316 le64toh(f->header->n_entries),
1317 realtime,
1318 test_object_realtime,
1319 direction,
1320 ret, offset, NULL);
1321 }
1322
1323 static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
1324 Object *o;
1325 int r;
1326
1327 assert(f);
1328 assert(p > 0);
1329
1330 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
1331 if (r < 0)
1332 return r;
1333
1334 if (le64toh(o->entry.monotonic) == needle)
1335 return TEST_FOUND;
1336 else if (le64toh(o->entry.monotonic) < needle)
1337 return TEST_LEFT;
1338 else
1339 return TEST_RIGHT;
1340 }
1341
1342 int journal_file_move_to_entry_by_monotonic(
1343 JournalFile *f,
1344 sd_id128_t boot_id,
1345 uint64_t monotonic,
1346 direction_t direction,
1347 Object **ret,
1348 uint64_t *offset) {
1349
1350 char t[8+32+1] = "_BOOT_ID=";
1351 Object *o;
1352 int r;
1353
1354 sd_id128_to_string(boot_id, t + 8);
1355
1356 r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
1357 if (r < 0)
1358 return r;
1359 else if (r == 0)
1360 return -ENOENT;
1361
1362 return generic_array_bisect_plus_one(f,
1363 le64toh(o->data.entry_offset),
1364 le64toh(o->data.entry_array_offset),
1365 le64toh(o->data.n_entries),
1366 monotonic,
1367 test_object_monotonic,
1368 direction,
1369 ret, offset, NULL);
1370 }
1371
1372 static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
1373 assert(f);
1374 assert(p > 0);
1375
1376 if (p == needle)
1377 return TEST_FOUND;
1378 else if (p < needle)
1379 return TEST_LEFT;
1380 else
1381 return TEST_RIGHT;
1382 }
1383
1384 int journal_file_next_entry(
1385 JournalFile *f,
1386 Object *o, uint64_t p,
1387 direction_t direction,
1388 Object **ret, uint64_t *offset) {
1389
1390 uint64_t i, n;
1391 int r;
1392
1393 assert(f);
1394 assert(p > 0 || !o);
1395
1396 n = le64toh(f->header->n_entries);
1397 if (n <= 0)
1398 return 0;
1399
1400 if (!o)
1401 i = direction == DIRECTION_DOWN ? 0 : n - 1;
1402 else {
1403 if (o->object.type != OBJECT_ENTRY)
1404 return -EINVAL;
1405
1406 r = generic_array_bisect(f,
1407 le64toh(f->header->entry_array_offset),
1408 le64toh(f->header->n_entries),
1409 p,
1410 test_object_offset,
1411 DIRECTION_DOWN,
1412 NULL, NULL,
1413 &i);
1414 if (r <= 0)
1415 return r;
1416
1417 if (direction == DIRECTION_DOWN) {
1418 if (i >= n - 1)
1419 return 0;
1420
1421 i++;
1422 } else {
1423 if (i <= 0)
1424 return 0;
1425
1426 i--;
1427 }
1428 }
1429
1430 /* And jump to it */
1431 return generic_array_get(f,
1432 le64toh(f->header->entry_array_offset),
1433 i,
1434 ret, offset);
1435 }
1436
1437 int journal_file_skip_entry(
1438 JournalFile *f,
1439 Object *o, uint64_t p,
1440 int64_t skip,
1441 Object **ret, uint64_t *offset) {
1442
1443 uint64_t i, n;
1444 int r;
1445
1446 assert(f);
1447 assert(o);
1448 assert(p > 0);
1449
1450 if (o->object.type != OBJECT_ENTRY)
1451 return -EINVAL;
1452
1453 r = generic_array_bisect(f,
1454 le64toh(f->header->entry_array_offset),
1455 le64toh(f->header->n_entries),
1456 p,
1457 test_object_offset,
1458 DIRECTION_DOWN,
1459 NULL, NULL,
1460 &i);
1461 if (r <= 0)
1462 return r;
1463
1464 /* Calculate new index */
1465 if (skip < 0) {
1466 if ((uint64_t) -skip >= i)
1467 i = 0;
1468 else
1469 i = i - (uint64_t) -skip;
1470 } else
1471 i += (uint64_t) skip;
1472
1473 n = le64toh(f->header->n_entries);
1474 if (n <= 0)
1475 return -EBADMSG;
1476
1477 if (i >= n)
1478 i = n-1;
1479
1480 return generic_array_get(f,
1481 le64toh(f->header->entry_array_offset),
1482 i,
1483 ret, offset);
1484 }
1485
1486 int journal_file_next_entry_for_data(
1487 JournalFile *f,
1488 Object *o, uint64_t p,
1489 uint64_t data_offset,
1490 direction_t direction,
1491 Object **ret, uint64_t *offset) {
1492
1493 uint64_t n, i;
1494 int r;
1495 Object *d;
1496
1497 assert(f);
1498 assert(p > 0 || !o);
1499
1500 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
1501 if (r < 0)
1502 return r;
1503
1504 n = le64toh(d->data.n_entries);
1505 if (n <= 0)
1506 return n;
1507
1508 if (!o)
1509 i = direction == DIRECTION_DOWN ? 0 : n - 1;
1510 else {
1511 if (o->object.type != OBJECT_ENTRY)
1512 return -EINVAL;
1513
1514 r = generic_array_bisect_plus_one(f,
1515 le64toh(d->data.entry_offset),
1516 le64toh(d->data.entry_array_offset),
1517 le64toh(d->data.n_entries),
1518 p,
1519 test_object_offset,
1520 DIRECTION_DOWN,
1521 NULL, NULL,
1522 &i);
1523
1524 if (r <= 0)
1525 return r;
1526
1527 if (direction == DIRECTION_DOWN) {
1528 if (i >= n - 1)
1529 return 0;
1530
1531 i++;
1532 } else {
1533 if (i <= 0)
1534 return 0;
1535
1536 i--;
1537 }
1538
1539 }
1540
1541 return generic_array_get_plus_one(f,
1542 le64toh(d->data.entry_offset),
1543 le64toh(d->data.entry_array_offset),
1544 i,
1545 ret, offset);
1546 }
1547
1548 int journal_file_move_to_entry_by_seqnum_for_data(
1549 JournalFile *f,
1550 uint64_t data_offset,
1551 uint64_t seqnum,
1552 direction_t direction,
1553 Object **ret, uint64_t *offset) {
1554
1555 Object *d;
1556 int r;
1557
1558 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
1559 if (r <= 0)
1560 return r;
1561
1562 return generic_array_bisect_plus_one(f,
1563 le64toh(d->data.entry_offset),
1564 le64toh(d->data.entry_array_offset),
1565 le64toh(d->data.n_entries),
1566 seqnum,
1567 test_object_seqnum,
1568 direction,
1569 ret, offset, NULL);
1570 }
1571
1572 int journal_file_move_to_entry_by_realtime_for_data(
1573 JournalFile *f,
1574 uint64_t data_offset,
1575 uint64_t realtime,
1576 direction_t direction,
1577 Object **ret, uint64_t *offset) {
1578
1579 Object *d;
1580 int r;
1581
1582 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
1583 if (r <= 0)
1584 return r;
1585
1586 return generic_array_bisect_plus_one(f,
1587 le64toh(d->data.entry_offset),
1588 le64toh(d->data.entry_array_offset),
1589 le64toh(d->data.n_entries),
1590 realtime,
1591 test_object_realtime,
1592 direction,
1593 ret, offset, NULL);
1594 }
1595
1596 void journal_file_dump(JournalFile *f) {
1597 char a[33], b[33], c[33];
1598 Object *o;
1599 int r;
1600 uint64_t p;
1601
1602 assert(f);
1603
1604 printf("File Path: %s\n"
1605 "File ID: %s\n"
1606 "Machine ID: %s\n"
1607 "Boot ID: %s\n"
1608 "Arena size: %llu\n"
1609 "Objects: %lu\n"
1610 "Entries: %lu\n",
1611 f->path,
1612 sd_id128_to_string(f->header->file_id, a),
1613 sd_id128_to_string(f->header->machine_id, b),
1614 sd_id128_to_string(f->header->boot_id, c),
1615 (unsigned long long) le64toh(f->header->arena_size),
1616 (unsigned long) le64toh(f->header->n_objects),
1617 (unsigned long) le64toh(f->header->n_entries));
1618
1619 p = le64toh(f->header->arena_offset);
1620 while (p != 0) {
1621 r = journal_file_move_to_object(f, -1, p, &o);
1622 if (r < 0)
1623 goto fail;
1624
1625 switch (o->object.type) {
1626
1627 case OBJECT_UNUSED:
1628 printf("Type: OBJECT_UNUSED\n");
1629 break;
1630
1631 case OBJECT_DATA:
1632 printf("Type: OBJECT_DATA\n");
1633 break;
1634
1635 case OBJECT_ENTRY:
1636 printf("Type: OBJECT_ENTRY %llu %llu %llu\n",
1637 (unsigned long long) le64toh(o->entry.seqnum),
1638 (unsigned long long) le64toh(o->entry.monotonic),
1639 (unsigned long long) le64toh(o->entry.realtime));
1640 break;
1641
1642 case OBJECT_FIELD_HASH_TABLE:
1643 printf("Type: OBJECT_FIELD_HASH_TABLE\n");
1644 break;
1645
1646 case OBJECT_DATA_HASH_TABLE:
1647 printf("Type: OBJECT_DATA_HASH_TABLE\n");
1648 break;
1649
1650 case OBJECT_ENTRY_ARRAY:
1651 printf("Type: OBJECT_ENTRY_ARRAY\n");
1652 break;
1653 }
1654
1655 if (o->object.flags & OBJECT_COMPRESSED)
1656 printf("Flags: COMPRESSED\n");
1657
1658 if (p == le64toh(f->header->tail_object_offset))
1659 p = 0;
1660 else
1661 p = p + ALIGN64(le64toh(o->object.size));
1662 }
1663
1664 return;
1665 fail:
1666 log_error("File corrupt");
1667 }
1668
1669 int journal_file_open(
1670 const char *fname,
1671 int flags,
1672 mode_t mode,
1673 JournalFile *template,
1674 JournalFile **ret) {
1675
1676 JournalFile *f;
1677 int r;
1678 bool newly_created = false;
1679
1680 assert(fname);
1681
1682 if ((flags & O_ACCMODE) != O_RDONLY &&
1683 (flags & O_ACCMODE) != O_RDWR)
1684 return -EINVAL;
1685
1686 f = new0(JournalFile, 1);
1687 if (!f)
1688 return -ENOMEM;
1689
1690 f->fd = -1;
1691 f->flags = flags;
1692 f->mode = mode;
1693 f->writable = (flags & O_ACCMODE) != O_RDONLY;
1694 f->prot = prot_from_flags(flags);
1695
1696 f->metrics.max_size = DEFAULT_MAX_SIZE;
1697 f->metrics.min_size = DEFAULT_MIN_SIZE;
1698 f->metrics.keep_free = DEFAULT_KEEP_FREE;
1699
1700 f->path = strdup(fname);
1701 if (!f->path) {
1702 r = -ENOMEM;
1703 goto fail;
1704 }
1705
1706 f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
1707 if (f->fd < 0) {
1708 r = -errno;
1709 goto fail;
1710 }
1711
1712 if (fstat(f->fd, &f->last_stat) < 0) {
1713 r = -errno;
1714 goto fail;
1715 }
1716
1717 if (f->last_stat.st_size == 0 && f->writable) {
1718 newly_created = true;
1719
1720 r = journal_file_init_header(f, template);
1721 if (r < 0)
1722 goto fail;
1723
1724 if (fstat(f->fd, &f->last_stat) < 0) {
1725 r = -errno;
1726 goto fail;
1727 }
1728 }
1729
1730 if (f->last_stat.st_size < (off_t) sizeof(Header)) {
1731 r = -EIO;
1732 goto fail;
1733 }
1734
1735 f->header = mmap(NULL, PAGE_ALIGN(sizeof(Header)), prot_from_flags(flags), MAP_SHARED, f->fd, 0);
1736 if (f->header == MAP_FAILED) {
1737 f->header = NULL;
1738 r = -errno;
1739 goto fail;
1740 }
1741
1742 if (!newly_created) {
1743 r = journal_file_verify_header(f);
1744 if (r < 0)
1745 goto fail;
1746 }
1747
1748 if (f->writable) {
1749 r = journal_file_refresh_header(f);
1750 if (r < 0)
1751 goto fail;
1752 }
1753
1754 if (newly_created) {
1755
1756 r = journal_file_setup_field_hash_table(f);
1757 if (r < 0)
1758 goto fail;
1759
1760 r = journal_file_setup_data_hash_table(f);
1761 if (r < 0)
1762 goto fail;
1763 }
1764
1765 r = journal_file_map_field_hash_table(f);
1766 if (r < 0)
1767 goto fail;
1768
1769 r = journal_file_map_data_hash_table(f);
1770 if (r < 0)
1771 goto fail;
1772
1773 if (ret)
1774 *ret = f;
1775
1776 return 0;
1777
1778 fail:
1779 journal_file_close(f);
1780
1781 return r;
1782 }
1783
1784 int journal_file_rotate(JournalFile **f) {
1785 char *p;
1786 size_t l;
1787 JournalFile *old_file, *new_file = NULL;
1788 int r;
1789
1790 assert(f);
1791 assert(*f);
1792
1793 old_file = *f;
1794
1795 if (!old_file->writable)
1796 return -EINVAL;
1797
1798 if (!endswith(old_file->path, ".journal"))
1799 return -EINVAL;
1800
1801 l = strlen(old_file->path);
1802
1803 p = new(char, l + 1 + 16 + 1 + 32 + 1 + 16 + 1);
1804 if (!p)
1805 return -ENOMEM;
1806
1807 memcpy(p, old_file->path, l - 8);
1808 p[l-8] = '@';
1809 sd_id128_to_string(old_file->header->seqnum_id, p + l - 8 + 1);
1810 snprintf(p + l - 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1,
1811 "-%016llx-%016llx.journal",
1812 (unsigned long long) le64toh((*f)->header->seqnum),
1813 (unsigned long long) le64toh((*f)->header->tail_entry_realtime));
1814
1815 r = rename(old_file->path, p);
1816 free(p);
1817
1818 if (r < 0)
1819 return -errno;
1820
1821 old_file->header->state = le32toh(STATE_ARCHIVED);
1822
1823 r = journal_file_open(old_file->path, old_file->flags, old_file->mode, old_file, &new_file);
1824 journal_file_close(old_file);
1825
1826 *f = new_file;
1827 return r;
1828 }
1829
1830 struct vacuum_info {
1831 off_t usage;
1832 char *filename;
1833
1834 uint64_t realtime;
1835 sd_id128_t seqnum_id;
1836 uint64_t seqnum;
1837 };
1838
1839 static int vacuum_compare(const void *_a, const void *_b) {
1840 const struct vacuum_info *a, *b;
1841
1842 a = _a;
1843 b = _b;
1844
1845 if (sd_id128_equal(a->seqnum_id, b->seqnum_id)) {
1846 if (a->seqnum < b->seqnum)
1847 return -1;
1848 else if (a->seqnum > b->seqnum)
1849 return 1;
1850 else
1851 return 0;
1852 }
1853
1854 if (a->realtime < b->realtime)
1855 return -1;
1856 else if (a->realtime > b->realtime)
1857 return 1;
1858 else
1859 return memcmp(&a->seqnum_id, &b->seqnum_id, 16);
1860 }
1861
1862 int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t min_free) {
1863 DIR *d;
1864 int r = 0;
1865 struct vacuum_info *list = NULL;
1866 unsigned n_list = 0, n_allocated = 0, i;
1867 uint64_t sum = 0;
1868
1869 assert(directory);
1870
1871 if (max_use <= 0)
1872 max_use = DEFAULT_MAX_USE;
1873
1874 d = opendir(directory);
1875 if (!d)
1876 return -errno;
1877
1878 for (;;) {
1879 int k;
1880 struct dirent buf, *de;
1881 size_t q;
1882 struct stat st;
1883 char *p;
1884 unsigned long long seqnum, realtime;
1885 sd_id128_t seqnum_id;
1886
1887 k = readdir_r(d, &buf, &de);
1888 if (k != 0) {
1889 r = -k;
1890 goto finish;
1891 }
1892
1893 if (!de)
1894 break;
1895
1896 if (!dirent_is_file_with_suffix(de, ".journal"))
1897 continue;
1898
1899 q = strlen(de->d_name);
1900
1901 if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8)
1902 continue;
1903
1904 if (de->d_name[q-8-16-1] != '-' ||
1905 de->d_name[q-8-16-1-16-1] != '-' ||
1906 de->d_name[q-8-16-1-16-1-32-1] != '@')
1907 continue;
1908
1909 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
1910 continue;
1911
1912 if (!S_ISREG(st.st_mode))
1913 continue;
1914
1915 p = strdup(de->d_name);
1916 if (!p) {
1917 r = -ENOMEM;
1918 goto finish;
1919 }
1920
1921 de->d_name[q-8-16-1-16-1] = 0;
1922 if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) {
1923 free(p);
1924 continue;
1925 }
1926
1927 if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) {
1928 free(p);
1929 continue;
1930 }
1931
1932 if (n_list >= n_allocated) {
1933 struct vacuum_info *j;
1934
1935 n_allocated = MAX(n_allocated * 2U, 8U);
1936 j = realloc(list, n_allocated * sizeof(struct vacuum_info));
1937 if (!j) {
1938 free(p);
1939 r = -ENOMEM;
1940 goto finish;
1941 }
1942
1943 list = j;
1944 }
1945
1946 list[n_list].filename = p;
1947 list[n_list].usage = (uint64_t) st.st_blksize * (uint64_t) st.st_blocks;
1948 list[n_list].seqnum = seqnum;
1949 list[n_list].realtime = realtime;
1950 list[n_list].seqnum_id = seqnum_id;
1951
1952 sum += list[n_list].usage;
1953
1954 n_list ++;
1955 }
1956
1957 qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare);
1958
1959 for(i = 0; i < n_list; i++) {
1960 struct statvfs ss;
1961
1962 if (fstatvfs(dirfd(d), &ss) < 0) {
1963 r = -errno;
1964 goto finish;
1965 }
1966
1967 if (sum <= max_use &&
1968 (uint64_t) ss.f_bavail * (uint64_t) ss.f_bsize >= min_free)
1969 break;
1970
1971 if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) {
1972 log_debug("Deleted archived journal %s/%s.", directory, list[i].filename);
1973 sum -= list[i].usage;
1974 } else if (errno != ENOENT)
1975 log_warning("Failed to delete %s/%s: %m", directory, list[i].filename);
1976 }
1977
1978 finish:
1979 for (i = 0; i < n_list; i++)
1980 free(list[i].filename);
1981
1982 free(list);
1983
1984 if (d)
1985 closedir(d);
1986
1987 return r;
1988 }
Unnamed repository; edit this file 'description' to name the repository.