]> 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
hwdb: add Aiptek Hyperpen 12000U (#10424)
[thirdparty/systemd.git] / src / journal / journal-file.c
CommitLineData
53e1b683 1/* SPDX-License-Identifier: LGPL-2.1+ */
cec736d2 2
cec736d2 3#include <errno.h>
cec736d2 4#include <fcntl.h>
11689d2a 5#include <linux/fs.h>
ac2e41f5 6#include <pthread.h>
07630cea
LP		 7#include <stddef.h>
8#include <sys/mman.h>
9#include <sys/statvfs.h>
10#include <sys/uio.h>
11#include <unistd.h>
fb0951b0 12
b5efdb8a 13#include "alloc-util.h"
f27a3864 14#include "btrfs-util.h"
c8b3094d 15#include "chattr-util.h"
07630cea 16#include "compress.h"
3ffd4af2 17#include "fd-util.h"
11b29a96 18#include "fs-util.h"
0284adc6 19#include "journal-authenticate.h"
cec736d2
LP		 20#include "journal-def.h"
21#include "journal-file.h"
22#include "lookup3.h"
6bedfcbb 23#include "parse-util.h"
5d1ce257 24#include "path-util.h"
3df3e884 25#include "random-util.h"
7a24f3bf 26#include "sd-event.h"
b58c888f 27#include "set.h"
3cc44114 28#include "stat-util.h"
07630cea 29#include "string-util.h"
4761fd0f 30#include "strv.h"
89a5a90c 31#include "xattr-util.h"
cec736d2 32
4a92baf3
LP		 33#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem))
34#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem))
cec736d2 35
57850536
AG		 36#define DEFAULT_COMPRESS_THRESHOLD (512ULL)
37#define MIN_COMPRESS_THRESHOLD (8ULL)
807e17f0 38
babfc091 39/* This is the minimum journal file size */
16098e93 40#define JOURNAL_FILE_SIZE_MIN (512ULL*1024ULL) /* 512 KiB */
babfc091
LP		 41
42/* These are the lower and upper bounds if we deduce the max_use value
43 * from the file system size */
44#define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */
45#define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */
46
8580d1f7
LP		 47/* This is the default minimal use limit, how much we'll use even if keep_free suggests otherwise. */
48#define DEFAULT_MIN_USE (1ULL*1024ULL*1024ULL) /* 1 MiB */
49
babfc091 50/* This is the upper bound if we deduce max_size from max_use */
71100051 51#define DEFAULT_MAX_SIZE_UPPER (128ULL*1024ULL*1024ULL) /* 128 MiB */
babfc091
LP		 52
53/* This is the upper bound if we deduce the keep_free value from the
54 * file system size */
55#define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */
56
57/* This is the keep_free value when we can't determine the system
58 * size */
59#define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */
60
8580d1f7
LP		 61/* This is the default maximum number of journal files to keep around. */
62#define DEFAULT_N_MAX_FILES (100)
63
dca6219e
LP		 64/* n_data was the first entry we added after the initial file format design */
65#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
cec736d2 66
a4bcff5b
LP		 67/* How many entries to keep in the entry array chain cache at max */
68#define CHAIN_CACHE_MAX 20
69
a676e665
LP		 70/* How much to increase the journal file size at once each time we allocate something new. */
71#define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL) /* 8MB */
72
2678031a
LP		 73/* Reread fstat() of the file for detecting deletions at least this often */
74#define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC)
75
fa6ac760
LP		 76/* The mmap context to use for the header we pick as one above the last defined typed */
77#define CONTEXT_HEADER _OBJECT_TYPE_MAX
78
51804460
ZJS		 79#ifdef __clang__
80# pragma GCC diagnostic ignored "-Waddress-of-packed-member"
81#endif
82
ac2e41f5
VC		 83/* This may be called from a separate thread to prevent blocking the caller for the duration of fsync().
84 * As a result we use atomic operations on f->offline_state for inter-thread communications with
85 * journal_file_set_offline() and journal_file_set_online(). */
86static void journal_file_set_offline_internal(JournalFile *f) {
26687bf8 87 assert(f);
ac2e41f5
VC		 88 assert(f->fd >= 0);
89 assert(f->header);
90
91 for (;;) {
92 switch (f->offline_state) {
93 case OFFLINE_CANCEL:
94 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_DONE))
95 continue;
96 return;
97
98 case OFFLINE_AGAIN_FROM_SYNCING:
99 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_SYNCING))
100 continue;
101 break;
102
103 case OFFLINE_AGAIN_FROM_OFFLINING:
104 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_SYNCING))
105 continue;
106 break;
107
108 case OFFLINE_SYNCING:
109 (void) fsync(f->fd);
26687bf8 110
ac2e41f5
VC		 111 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_OFFLINING))
112 continue;
26687bf8 113
8eb85171 114 f->header->state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE;
ac2e41f5
VC		 115 (void) fsync(f->fd);
116 break;
117
118 case OFFLINE_OFFLINING:
119 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_DONE))
120 continue;
4831981d 121 _fallthrough_;
ac2e41f5
VC		 122 case OFFLINE_DONE:
123 return;
124
125 case OFFLINE_JOINED:
126 log_debug("OFFLINE_JOINED unexpected offline state for journal_file_set_offline_internal()");
127 return;
128 }
129 }
130}
131
132static void * journal_file_set_offline_thread(void *arg) {
133 JournalFile *f = arg;
134
fa7ff4cf
LP		 135 (void) pthread_setname_np(pthread_self(), "journal-offline");
136
ac2e41f5
VC		 137 journal_file_set_offline_internal(f);
138
139 return NULL;
140}
141
142static int journal_file_set_offline_thread_join(JournalFile *f) {
143 int r;
144
145 assert(f);
146
147 if (f->offline_state == OFFLINE_JOINED)
148 return 0;
149
150 r = pthread_join(f->offline_thread, NULL);
151 if (r)
152 return -r;
153
154 f->offline_state = OFFLINE_JOINED;
26687bf8 155
be7cdd8e 156 if (mmap_cache_got_sigbus(f->mmap, f->cache_fd))
fa6ac760
LP		 157 return -EIO;
158
ac2e41f5
VC		 159 return 0;
160}
26687bf8 161
ac2e41f5
VC		 162/* Trigger a restart if the offline thread is mid-flight in a restartable state. */
163static bool journal_file_set_offline_try_restart(JournalFile *f) {
164 for (;;) {
165 switch (f->offline_state) {
166 case OFFLINE_AGAIN_FROM_SYNCING:
167 case OFFLINE_AGAIN_FROM_OFFLINING:
168 return true;
169
170 case OFFLINE_CANCEL:
171 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_AGAIN_FROM_SYNCING))
172 continue;
173 return true;
174
175 case OFFLINE_SYNCING:
176 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_AGAIN_FROM_SYNCING))
177 continue;
178 return true;
179
180 case OFFLINE_OFFLINING:
181 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_AGAIN_FROM_OFFLINING))
182 continue;
183 return true;
26687bf8
OS		 184
185 default:
ac2e41f5
VC		 186 return false;
187 }
26687bf8
OS		 188 }
189}
190
ac2e41f5
VC		 191/* Sets a journal offline.
192 *
193 * If wait is false then an offline is dispatched in a separate thread for a
194 * subsequent journal_file_set_offline() or journal_file_set_online() of the
195 * same journal to synchronize with.
196 *
197 * If wait is true, then either an existing offline thread will be restarted
198 * and joined, or if none exists the offline is simply performed in this
199 * context without involving another thread.
200 */
201int journal_file_set_offline(JournalFile *f, bool wait) {
202 bool restarted;
203 int r;
204
26687bf8
OS		 205 assert(f);
206
207 if (!f->writable)
208 return -EPERM;
209
846e5418 210 if (f->fd < 0 || !f->header)
26687bf8
OS		 211 return -EINVAL;
212
b8f99e27
VC		 213 /* An offlining journal is implicitly online and may modify f->header->state,
214 * we must also join any potentially lingering offline thread when not online. */
215 if (!journal_file_is_offlining(f) && f->header->state != STATE_ONLINE)
216 return journal_file_set_offline_thread_join(f);
26687bf8 217
ac2e41f5
VC		 218 /* Restart an in-flight offline thread and wait if needed, or join a lingering done one. */
219 restarted = journal_file_set_offline_try_restart(f);
220 if ((restarted && wait) || !restarted) {
221 r = journal_file_set_offline_thread_join(f);
222 if (r < 0)
223 return r;
224 }
26687bf8 225
ac2e41f5
VC		 226 if (restarted)
227 return 0;
228
229 /* Initiate a new offline. */
230 f->offline_state = OFFLINE_SYNCING;
fa6ac760 231
ac2e41f5
VC		 232 if (wait) /* Without using a thread if waiting. */
233 journal_file_set_offline_internal(f);
234 else {
5e9f01e8
LP		 235 sigset_t ss, saved_ss;
236 int k;
237
238 if (sigfillset(&ss) < 0)
239 return -errno;
240
241 r = pthread_sigmask(SIG_BLOCK, &ss, &saved_ss);
242 if (r > 0)
243 return -r;
244
ac2e41f5 245 r = pthread_create(&f->offline_thread, NULL, journal_file_set_offline_thread, f);
5e9f01e8
LP		 246
247 k = pthread_sigmask(SIG_SETMASK, &saved_ss, NULL);
ec9ffa2c
VC		 248 if (r > 0) {
249 f->offline_state = OFFLINE_JOINED;
ac2e41f5 250 return -r;
ec9ffa2c 251 }
5e9f01e8
LP		 252 if (k > 0)
253 return -k;
ac2e41f5
VC		 254 }
255
256 return 0;
257}
258
259static int journal_file_set_online(JournalFile *f) {
83bf6b67 260 bool wait = true;
ac2e41f5
VC		 261
262 assert(f);
263
264 if (!f->writable)
265 return -EPERM;
266
846e5418 267 if (f->fd < 0 || !f->header)
ac2e41f5
VC		 268 return -EINVAL;
269
83bf6b67 270 while (wait) {
ac2e41f5
VC		 271 switch (f->offline_state) {
272 case OFFLINE_JOINED:
273 /* No offline thread, no need to wait. */
83bf6b67 274 wait = false;
ac2e41f5
VC		 275 break;
276
277 case OFFLINE_SYNCING:
278 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_CANCEL))
279 continue;
280 /* Canceled syncing prior to offlining, no need to wait. */
83bf6b67 281 wait = false;
ac2e41f5
VC		 282 break;
283
284 case OFFLINE_AGAIN_FROM_SYNCING:
285 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_CANCEL))
286 continue;
287 /* Canceled restart from syncing, no need to wait. */
83bf6b67 288 wait = false;
ac2e41f5
VC		 289 break;
290
291 case OFFLINE_AGAIN_FROM_OFFLINING:
292 if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_CANCEL))
293 continue;
294 /* Canceled restart from offlining, must wait for offlining to complete however. */
4831981d 295 _fallthrough_;
ac2e41f5
VC		 296 default: {
297 int r;
298
299 r = journal_file_set_offline_thread_join(f);
300 if (r < 0)
301 return r;
302
83bf6b67 303 wait = false;
ac2e41f5
VC		 304 break;
305 }
306 }
307 }
26687bf8 308
be7cdd8e 309 if (mmap_cache_got_sigbus(f->mmap, f->cache_fd))
fa6ac760
LP		 310 return -EIO;
311
ac2e41f5
VC		 312 switch (f->header->state) {
313 case STATE_ONLINE:
314 return 0;
26687bf8 315
ac2e41f5
VC		 316 case STATE_OFFLINE:
317 f->header->state = STATE_ONLINE;
318 (void) fsync(f->fd);
319 return 0;
320
321 default:
322 return -EINVAL;
323 }
26687bf8
OS		 324}
325
b58c888f
VC		 326bool journal_file_is_offlining(JournalFile *f) {
327 assert(f);
328
329 __sync_synchronize();
330
3742095b 331 if (IN_SET(f->offline_state, OFFLINE_DONE, OFFLINE_JOINED))
b58c888f
VC		 332 return false;
333
334 return true;
335}
336
804ae586 337JournalFile* journal_file_close(JournalFile *f) {
de190aef 338 assert(f);
cec736d2 339
349cc4a5 340#if HAVE_GCRYPT
b0af6f41 341 /* Write the final tag */
43cd8794
FB		 342 if (f->seal && f->writable) {
343 int r;
344
345 r = journal_file_append_tag(f);
346 if (r < 0)
347 log_error_errno(r, "Failed to append tag when closing journal: %m");
348 }
feb12d3e 349#endif
b0af6f41 350
7a24f3bf
VC		 351 if (f->post_change_timer) {
352 int enabled;
353
354 if (sd_event_source_get_enabled(f->post_change_timer, &enabled) >= 0)
355 if (enabled == SD_EVENT_ONESHOT)
356 journal_file_post_change(f);
357
e167d7fd 358 (void) sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_OFF);
7a24f3bf
VC		 359 sd_event_source_unref(f->post_change_timer);
360 }
361
ac2e41f5 362 journal_file_set_offline(f, true);
cec736d2 363
be7cdd8e
VC		 364 if (f->mmap && f->cache_fd)
365 mmap_cache_free_fd(f->mmap, f->cache_fd);
cec736d2 366
11689d2a
LP		 367 if (f->fd >= 0 && f->defrag_on_close) {
368
369 /* Be friendly to btrfs: turn COW back on again now,
370 * and defragment the file. We won't write to the file
371 * ever again, hence remove all fragmentation, and
372 * reenable all the good bits COW usually provides
373 * (such as data checksumming). */
374
db9a4254 375 (void) chattr_fd(f->fd, 0, FS_NOCOW_FL, NULL);
11689d2a
LP		 376 (void) btrfs_defrag_fd(f->fd);
377 }
f27a3864 378
5d1ce257
LP		 379 if (f->close_fd)
380 safe_close(f->fd);
cec736d2 381 free(f->path);
807e17f0 382
f649045c 383 mmap_cache_unref(f->mmap);
16e9f408 384
4743015d 385 ordered_hashmap_free_free(f->chain_cache);
a4bcff5b 386
349cc4a5 387#if HAVE_XZ || HAVE_LZ4
807e17f0
LP		 388 free(f->compress_buffer);
389#endif
390
349cc4a5 391#if HAVE_GCRYPT
baed47c3
LP		 392 if (f->fss_file)
393 munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size));
dc4ebc07 394 else
b7c9ae91
LP		 395 free(f->fsprg_state);
396
397 free(f->fsprg_seed);
7560fffc
LP		 398
399 if (f->hmac)
400 gcry_md_close(f->hmac);
401#endif
402
6b430fdb 403 return mfree(f);
cec736d2
LP		 404}
405
0ac38b70 406static int journal_file_init_header(JournalFile *f, JournalFile *template) {
d89c8fdf 407 Header h = {};
cec736d2
LP		 408 ssize_t k;
409 int r;
410
411 assert(f);
412
7560fffc 413 memcpy(h.signature, HEADER_SIGNATURE, 8);
23b0b2b2 414 h.header_size = htole64(ALIGN64(sizeof(h)));
cec736d2 415
d89c8fdf
ZJS		 416 h.incompatible_flags |= htole32(
417 f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ |
418 f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4);
7560fffc 419
d89c8fdf
ZJS		 420 h.compatible_flags = htole32(
421 f->seal * HEADER_COMPATIBLE_SEALED);
7560fffc 422
cec736d2
LP		 423 r = sd_id128_randomize(&h.file_id);
424 if (r < 0)
425 return r;
426
0ac38b70
LP		 427 if (template) {
428 h.seqnum_id = template->header->seqnum_id;
beec0085 429 h.tail_entry_seqnum = template->header->tail_entry_seqnum;
0ac38b70
LP		 430 } else
431 h.seqnum_id = h.file_id;
cec736d2
LP		 432
433 k = pwrite(f->fd, &h, sizeof(h), 0);
434 if (k < 0)
435 return -errno;
436
437 if (k != sizeof(h))
438 return -EIO;
439
440 return 0;
441}
442
443static int journal_file_refresh_header(JournalFile *f) {
de190aef 444 sd_id128_t boot_id;
fa6ac760 445 int r;
cec736d2
LP		 446
447 assert(f);
c88cc6af 448 assert(f->header);
cec736d2
LP		 449
450 r = sd_id128_get_machine(&f->header->machine_id);
fd4885df
ZJS		 451 if (IN_SET(r, -ENOENT, -ENOMEDIUM))
452 /* We don't have a machine-id, let's continue without */
453 zero(f->header->machine_id);
454 else if (r < 0)
cec736d2
LP		 455 return r;
456
de190aef 457 r = sd_id128_get_boot(&boot_id);
cec736d2
LP		 458 if (r < 0)
459 return r;
460
de190aef
LP		 461 f->header->boot_id = boot_id;
462
fa6ac760 463 r = journal_file_set_online(f);
b788cc23 464
7560fffc 465 /* Sync the online state to disk */
fb426037 466 (void) fsync(f->fd);
b788cc23 467
a0fe2a2d
LP		 468 /* We likely just created a new file, also sync the directory this file is located in. */
469 (void) fsync_directory_of_file(f->fd);
470
fa6ac760 471 return r;
cec736d2
LP		 472}
473
4214009f
ZJS		 474static bool warn_wrong_flags(const JournalFile *f, bool compatible) {
475 const uint32_t any = compatible ? HEADER_COMPATIBLE_ANY : HEADER_INCOMPATIBLE_ANY,
476 supported = compatible ? HEADER_COMPATIBLE_SUPPORTED : HEADER_INCOMPATIBLE_SUPPORTED;
477 const char *type = compatible ? "compatible" : "incompatible";
d89c8fdf
ZJS		 478 uint32_t flags;
479
4214009f
ZJS		 480 flags = le32toh(compatible ? f->header->compatible_flags : f->header->incompatible_flags);
481
482 if (flags & ~supported) {
483 if (flags & ~any)
4761fd0f 484 log_debug("Journal file %s has unknown %s flags 0x%"PRIx32,
4214009f
ZJS		 485 f->path, type, flags & ~any);
486 flags = (flags & any) & ~supported;
4761fd0f
ZJS		 487 if (flags) {
488 const char* strv[3];
489 unsigned n = 0;
490 _cleanup_free_ char *t = NULL;
491
492 if (compatible && (flags & HEADER_COMPATIBLE_SEALED))
493 strv[n++] = "sealed";
494 if (!compatible && (flags & HEADER_INCOMPATIBLE_COMPRESSED_XZ))
495 strv[n++] = "xz-compressed";
496 if (!compatible && (flags & HEADER_INCOMPATIBLE_COMPRESSED_LZ4))
497 strv[n++] = "lz4-compressed";
498 strv[n] = NULL;
499 assert(n < ELEMENTSOF(strv));
500
501 t = strv_join((char**) strv, ", ");
502 log_debug("Journal file %s uses %s %s %s disabled at compilation time.",
503 f->path, type, n > 1 ? "flags" : "flag", strnull(t));
504 }
4214009f
ZJS		 505 return true;
506 }
507
508 return false;
509}
510
511static int journal_file_verify_header(JournalFile *f) {
6f94e420
TS		 512 uint64_t arena_size, header_size;
513
cec736d2 514 assert(f);
c88cc6af 515 assert(f->header);
cec736d2 516
7560fffc 517 if (memcmp(f->header->signature, HEADER_SIGNATURE, 8))
cec736d2
LP		 518 return -EBADMSG;
519
4214009f
ZJS		 520 /* In both read and write mode we refuse to open files with incompatible
521 * flags we don't know. */
522 if (warn_wrong_flags(f, false))
cec736d2
LP		 523 return -EPROTONOSUPPORT;
524
4214009f
ZJS		 525 /* When open for writing we refuse to open files with compatible flags, too. */
526 if (f->writable && warn_wrong_flags(f, true))
d89c8fdf 527 return -EPROTONOSUPPORT;
7560fffc 528
db11ac1a
LP		 529 if (f->header->state >= _STATE_MAX)
530 return -EBADMSG;
531
6f94e420
TS		 532 header_size = le64toh(f->header->header_size);
533
dca6219e 534 /* The first addition was n_data, so check that we are at least this large */
6f94e420 535 if (header_size < HEADER_SIZE_MIN)
23b0b2b2
LP		 536 return -EBADMSG;
537
8088cbd3 538 if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
beec0085
LP		 539 return -EBADMSG;
540
6f94e420
TS		 541 arena_size = le64toh(f->header->arena_size);
542
543 if (UINT64_MAX - header_size < arena_size || header_size + arena_size > (uint64_t) f->last_stat.st_size)
db11ac1a
LP		 544 return -ENODATA;
545
6f94e420 546 if (le64toh(f->header->tail_object_offset) > header_size + arena_size)
db11ac1a
LP		 547 return -ENODATA;
548
7762e02b
LP		 549 if (!VALID64(le64toh(f->header->data_hash_table_offset)) ||
550 !VALID64(le64toh(f->header->field_hash_table_offset)) ||
551 !VALID64(le64toh(f->header->tail_object_offset)) ||
552 !VALID64(le64toh(f->header->entry_array_offset)))
553 return -ENODATA;
554
cec736d2 555 if (f->writable) {
cec736d2 556 sd_id128_t machine_id;
ae739cc1 557 uint8_t state;
cec736d2
LP		 558 int r;
559
560 r = sd_id128_get_machine(&machine_id);
561 if (r < 0)
562 return r;
563
564 if (!sd_id128_equal(machine_id, f->header->machine_id))
565 return -EHOSTDOWN;
566
de190aef 567 state = f->header->state;
cec736d2 568
b288cdeb
ZJS		 569 if (state == STATE_ARCHIVED)
570 return -ESHUTDOWN; /* Already archived */
571 else if (state == STATE_ONLINE) {
71fa6f00
LP		 572 log_debug("Journal file %s is already online. Assuming unclean closing.", f->path);
573 return -EBUSY;
b288cdeb 574 } else if (state != STATE_OFFLINE) {
8facc349 575 log_debug("Journal file %s has unknown state %i.", f->path, state);
71fa6f00
LP		 576 return -EBUSY;
577 }
ae739cc1 578
5b3cc0c8
YN		 579 if (f->header->field_hash_table_size == 0 || f->header->data_hash_table_size == 0)
580 return -EBADMSG;
581
ae739cc1
LP		 582 /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't
583 * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks
584 * bisection. */
585 if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) {
586 log_debug("Journal file %s is from the future, refusing to append new data to it that'd be older.", f->path);
587 return -ETXTBSY;
588 }
cec736d2
LP		 589 }
590
d89c8fdf
ZJS		 591 f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header);
592 f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header);
c586dbf1 593
f1889c91 594 f->seal = JOURNAL_HEADER_SEALED(f->header);
7560fffc 595
cec736d2
LP		 596 return 0;
597}
598
2678031a 599static int journal_file_fstat(JournalFile *f) {
3cc44114
LP		 600 int r;
601
2678031a
LP		 602 assert(f);
603 assert(f->fd >= 0);
604
605 if (fstat(f->fd, &f->last_stat) < 0)
606 return -errno;
607
608 f->last_stat_usec = now(CLOCK_MONOTONIC);
609
8d6a4d33 610 /* Refuse dealing with with files that aren't regular */
3cc44114
LP		 611 r = stat_verify_regular(&f->last_stat);
612 if (r < 0)
613 return r;
8d6a4d33 614
2678031a
LP		 615 /* Refuse appending to files that are already deleted */
616 if (f->last_stat.st_nlink <= 0)
617 return -EIDRM;
618
619 return 0;
620}
621
cec736d2 622static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
eda4b58b 623 uint64_t old_size, new_size;
fec2aa2f 624 int r;
cec736d2
LP		 625
626 assert(f);
c88cc6af 627 assert(f->header);
cec736d2 628
cec736d2 629 /* We assume that this file is not sparse, and we know that
38ac38b2 630 * for sure, since we always call posix_fallocate()
cec736d2
LP		 631 * ourselves */
632
be7cdd8e 633 if (mmap_cache_got_sigbus(f->mmap, f->cache_fd))
fa6ac760
LP		 634 return -EIO;
635
cec736d2 636 old_size =
23b0b2b2 637 le64toh(f->header->header_size) +
cec736d2
LP		 638 le64toh(f->header->arena_size);
639
bc85bfee 640 new_size = PAGE_ALIGN(offset + size);
23b0b2b2
LP		 641 if (new_size < le64toh(f->header->header_size))
642 new_size = le64toh(f->header->header_size);
bc85bfee 643
2678031a
LP		 644 if (new_size <= old_size) {
645
646 /* We already pre-allocated enough space, but before
647 * we write to it, let's check with fstat() if the
648 * file got deleted, in order make sure we don't throw
649 * away the data immediately. Don't check fstat() for
650 * all writes though, but only once ever 10s. */
651
652 if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC))
653 return 0;
654
655 return journal_file_fstat(f);
656 }
657
658 /* Allocate more space. */
cec736d2 659
a676e665 660 if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
bc85bfee 661 return -E2BIG;
cec736d2 662
a676e665 663 if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) {
cec736d2
LP		 664 struct statvfs svfs;
665
666 if (fstatvfs(f->fd, &svfs) >= 0) {
667 uint64_t available;
668
070052ab 669 available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free);
cec736d2
LP		 670
671 if (new_size - old_size > available)
672 return -E2BIG;
673 }
674 }
675
eda4b58b 676 /* Increase by larger blocks at once */
be6b0c21 677 new_size = DIV_ROUND_UP(new_size, FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE;
eda4b58b
LP		 678 if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
679 new_size = f->metrics.max_size;
680
bc85bfee
LP		 681 /* Note that the glibc fallocate() fallback is very
682 inefficient, hence we try to minimize the allocation area
683 as we can. */
fec2aa2f
GV		 684 r = posix_fallocate(f->fd, old_size, new_size - old_size);
685 if (r != 0)
686 return -r;
cec736d2 687
23b0b2b2 688 f->header->arena_size = htole64(new_size - le64toh(f->header->header_size));
cec736d2 689
2678031a 690 return journal_file_fstat(f);
cec736d2
LP		 691}
692
78519831 693static unsigned type_to_context(ObjectType type) {
d3d3208f 694 /* One context for each type, plus one catch-all for the rest */
69adae51 695 assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS);
fa6ac760 696 assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS);
d05089d8 697 return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0;
d3d3208f
MS		 698}
699
b439282e 700static int journal_file_move_to(JournalFile *f, ObjectType type, bool keep_always, uint64_t offset, uint64_t size, void **ret, size_t *ret_size) {
2678031a
LP		 701 int r;
702
cec736d2 703 assert(f);
cec736d2
LP		 704 assert(ret);
705
7762e02b
LP		 706 if (size <= 0)
707 return -EINVAL;
708
2a59ea54 709 /* Avoid SIGBUS on invalid accesses */
4bbdcdb3
LP		 710 if (offset + size > (uint64_t) f->last_stat.st_size) {
711 /* Hmm, out of range? Let's refresh the fstat() data
712 * first, before we trust that check. */
713
2678031a
LP		 714 r = journal_file_fstat(f);
715 if (r < 0)
716 return r;
717
718 if (offset + size > (uint64_t) f->last_stat.st_size)
4bbdcdb3
LP		 719 return -EADDRNOTAVAIL;
720 }
721
b439282e 722 return mmap_cache_get(f->mmap, f->cache_fd, f->prot, type_to_context(type), keep_always, offset, size, &f->last_stat, ret, ret_size);
cec736d2
LP		 723}
724
16e9f408
LP		 725static uint64_t minimum_header_size(Object *o) {
726
b8e891e6 727 static const uint64_t table[] = {
16e9f408
LP		 728 [OBJECT_DATA] = sizeof(DataObject),
729 [OBJECT_FIELD] = sizeof(FieldObject),
730 [OBJECT_ENTRY] = sizeof(EntryObject),
731 [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject),
732 [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject),
733 [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject),
734 [OBJECT_TAG] = sizeof(TagObject),
735 };
736
737 if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0)
738 return sizeof(ObjectHeader);
739
740 return table[o->object.type];
741}
742
24754f36
TR		 743/* Lightweight object checks. We want this to be fast, so that we won't
744 * slowdown every journal_file_move_to_object() call too much. */
745static int journal_file_check_object(JournalFile *f, uint64_t offset, Object *o) {
746 assert(f);
747 assert(o);
748
749 switch (o->object.type) {
750
751 case OBJECT_DATA: {
752 if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) {
753 log_debug("Bad n_entries: %"PRIu64": %"PRIu64,
10e8445b 754 le64toh(o->data.n_entries), offset);
24754f36
TR		 755 return -EBADMSG;
756 }
757
758 if (le64toh(o->object.size) - offsetof(DataObject, payload) <= 0) {
759 log_debug("Bad object size (<= %zu): %"PRIu64": %"PRIu64,
760 offsetof(DataObject, payload),
761 le64toh(o->object.size),
762 offset);
763 return -EBADMSG;
764 }
765
10e8445b
TR		 766 if (!VALID64(le64toh(o->data.next_hash_offset)) ||
767 !VALID64(le64toh(o->data.next_field_offset)) ||
768 !VALID64(le64toh(o->data.entry_offset)) ||
769 !VALID64(le64toh(o->data.entry_array_offset))) {
24754f36
TR		 770 log_debug("Invalid offset, next_hash_offset="OFSfmt", next_field_offset="OFSfmt
771 ", entry_offset="OFSfmt", entry_array_offset="OFSfmt": %"PRIu64,
10e8445b
TR		 772 le64toh(o->data.next_hash_offset),
773 le64toh(o->data.next_field_offset),
774 le64toh(o->data.entry_offset),
775 le64toh(o->data.entry_array_offset),
24754f36
TR		 776 offset);
777 return -EBADMSG;
778 }
779
780 break;
781 }
782
783 case OBJECT_FIELD:
784 if (le64toh(o->object.size) - offsetof(FieldObject, payload) <= 0) {
785 log_debug(
786 "Bad field size (<= %zu): %"PRIu64": %"PRIu64,
787 offsetof(FieldObject, payload),
788 le64toh(o->object.size),
789 offset);
790 return -EBADMSG;
791 }
792
10e8445b
TR		 793 if (!VALID64(le64toh(o->field.next_hash_offset)) ||
794 !VALID64(le64toh(o->field.head_data_offset))) {
24754f36
TR		 795 log_debug(
796 "Invalid offset, next_hash_offset="OFSfmt
797 ", head_data_offset="OFSfmt": %"PRIu64,
10e8445b
TR		 798 le64toh(o->field.next_hash_offset),
799 le64toh(o->field.head_data_offset),
24754f36
TR		 800 offset);
801 return -EBADMSG;
802 }
803 break;
804
805 case OBJECT_ENTRY:
806 if ((le64toh(o->object.size) - offsetof(EntryObject, items)) % sizeof(EntryItem) != 0) {
807 log_debug(
808 "Bad entry size (<= %zu): %"PRIu64": %"PRIu64,
809 offsetof(EntryObject, items),
810 le64toh(o->object.size),
811 offset);
812 return -EBADMSG;
813 }
814
815 if ((le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem) <= 0) {
816 log_debug(
817 "Invalid number items in entry: %"PRIu64": %"PRIu64,
818 (le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem),
819 offset);
820 return -EBADMSG;
821 }
822
823 if (le64toh(o->entry.seqnum) <= 0) {
824 log_debug(
825 "Invalid entry seqnum: %"PRIx64": %"PRIu64,
826 le64toh(o->entry.seqnum),
827 offset);
828 return -EBADMSG;
829 }
830
831 if (!VALID_REALTIME(le64toh(o->entry.realtime))) {
832 log_debug(
833 "Invalid entry realtime timestamp: %"PRIu64": %"PRIu64,
834 le64toh(o->entry.realtime),
835 offset);
836 return -EBADMSG;
837 }
838
839 if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) {
840 log_debug(
841 "Invalid entry monotonic timestamp: %"PRIu64": %"PRIu64,
842 le64toh(o->entry.monotonic),
843 offset);
844 return -EBADMSG;
845 }
846
847 break;
848
849 case OBJECT_DATA_HASH_TABLE:
850 case OBJECT_FIELD_HASH_TABLE:
851 if ((le64toh(o->object.size) - offsetof(HashTableObject, items)) % sizeof(HashItem) != 0 ||
852 (le64toh(o->object.size) - offsetof(HashTableObject, items)) / sizeof(HashItem) <= 0) {
853 log_debug(
854 "Invalid %s hash table size: %"PRIu64": %"PRIu64,
855 o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field",
856 le64toh(o->object.size),
857 offset);
858 return -EBADMSG;
859 }
860
861 break;
862
863 case OBJECT_ENTRY_ARRAY:
864 if ((le64toh(o->object.size) - offsetof(EntryArrayObject, items)) % sizeof(le64_t) != 0 ||
865 (le64toh(o->object.size) - offsetof(EntryArrayObject, items)) / sizeof(le64_t) <= 0) {
866 log_debug(
867 "Invalid object entry array size: %"PRIu64": %"PRIu64,
868 le64toh(o->object.size),
869 offset);
870 return -EBADMSG;
871 }
872
10e8445b 873 if (!VALID64(le64toh(o->entry_array.next_entry_array_offset))) {
24754f36
TR		 874 log_debug(
875 "Invalid object entry array next_entry_array_offset: "OFSfmt": %"PRIu64,
10e8445b 876 le64toh(o->entry_array.next_entry_array_offset),
24754f36
TR		 877 offset);
878 return -EBADMSG;
879 }
880
881 break;
882
883 case OBJECT_TAG:
884 if (le64toh(o->object.size) != sizeof(TagObject)) {
885 log_debug(
886 "Invalid object tag size: %"PRIu64": %"PRIu64,
887 le64toh(o->object.size),
888 offset);
889 return -EBADMSG;
890 }
891
10e8445b 892 if (!VALID_EPOCH(le64toh(o->tag.epoch))) {
24754f36
TR		 893 log_debug(
894 "Invalid object tag epoch: %"PRIu64": %"PRIu64,
10e8445b 895 le64toh(o->tag.epoch),
24754f36
TR		 896 offset);
897 return -EBADMSG;
898 }
899
900 break;
901 }
902
903 return 0;
904}
905
78519831 906int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) {
cec736d2
LP		 907 int r;
908 void *t;
b439282e 909 size_t tsize;
cec736d2
LP		 910 Object *o;
911 uint64_t s;
912
913 assert(f);
914 assert(ret);
915
db11ac1a 916 /* Objects may only be located at multiple of 64 bit */
202fd896
LP		 917 if (!VALID64(offset)) {
918 log_debug("Attempt to move to object at non-64bit boundary: %" PRIu64, offset);
bd30fdf2 919 return -EBADMSG;
202fd896 920 }
db11ac1a 921
50809d7a 922 /* Object may not be located in the file header */
202fd896
LP		 923 if (offset < le64toh(f->header->header_size)) {
924 log_debug("Attempt to move to object located in file header: %" PRIu64, offset);
50809d7a 925 return -EBADMSG;
202fd896 926 }
50809d7a 927
b439282e 928 r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t, &tsize);
cec736d2
LP		 929 if (r < 0)
930 return r;
931
932 o = (Object*) t;
933 s = le64toh(o->object.size);
934
1c69f096
LP		 935 if (s == 0) {
936 log_debug("Attempt to move to uninitialized object: %" PRIu64, offset);
937 return -EBADMSG;
938 }
202fd896
LP		 939 if (s < sizeof(ObjectHeader)) {
940 log_debug("Attempt to move to overly short object: %" PRIu64, offset);
cec736d2 941 return -EBADMSG;
202fd896 942 }
cec736d2 943
202fd896
LP		 944 if (o->object.type <= OBJECT_UNUSED) {
945 log_debug("Attempt to move to object with invalid type: %" PRIu64, offset);
16e9f408 946 return -EBADMSG;
202fd896 947 }
16e9f408 948
202fd896
LP		 949 if (s < minimum_header_size(o)) {
950 log_debug("Attempt to move to truncated object: %" PRIu64, offset);
16e9f408 951 return -EBADMSG;
202fd896 952 }
16e9f408 953
202fd896
LP		 954 if (type > OBJECT_UNUSED && o->object.type != type) {
955 log_debug("Attempt to move to object of unexpected type: %" PRIu64, offset);
cec736d2 956 return -EBADMSG;
202fd896 957 }
cec736d2 958
b439282e
VC		 959 if (s > tsize) {
960 r = journal_file_move_to(f, type, false, offset, s, &t, NULL);
cec736d2
LP		 961 if (r < 0)
962 return r;
963
964 o = (Object*) t;
965 }
966
24754f36
TR		 967 r = journal_file_check_object(f, offset, o);
968 if (r < 0)
969 return r;
970
cec736d2
LP		 971 *ret = o;
972 return 0;
973}
974
d98cc1f2 975static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) {
cec736d2
LP		 976 uint64_t r;
977
978 assert(f);
c88cc6af 979 assert(f->header);
cec736d2 980
beec0085 981 r = le64toh(f->header->tail_entry_seqnum) + 1;
c2373f84
LP		 982
983 if (seqnum) {
de190aef 984 /* If an external seqnum counter was passed, we update
c2373f84
LP		 985 * both the local and the external one, and set it to
986 * the maximum of both */
987
988 if (*seqnum + 1 > r)
989 r = *seqnum + 1;
990
991 *seqnum = r;
992 }
993
beec0085 994 f->header->tail_entry_seqnum = htole64(r);
cec736d2 995
beec0085
LP		 996 if (f->header->head_entry_seqnum == 0)
997 f->header->head_entry_seqnum = htole64(r);
de190aef 998
cec736d2
LP		 999 return r;
1000}
1001
78519831 1002int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset) {
cec736d2
LP		 1003 int r;
1004 uint64_t p;
1005 Object *tail, *o;
1006 void *t;
1007
1008 assert(f);
c88cc6af 1009 assert(f->header);
d05089d8 1010 assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX);
cec736d2
LP		 1011 assert(size >= sizeof(ObjectHeader));
1012 assert(offset);
1013 assert(ret);
1014
26687bf8
OS		 1015 r = journal_file_set_online(f);
1016 if (r < 0)
1017 return r;
1018
cec736d2 1019 p = le64toh(f->header->tail_object_offset);
cec736d2 1020 if (p == 0)
23b0b2b2 1021 p = le64toh(f->header->header_size);
cec736d2 1022 else {
d05089d8 1023 r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail);
cec736d2
LP		 1024 if (r < 0)
1025 return r;
1026
1027 p += ALIGN64(le64toh(tail->object.size));
1028 }
1029
1030 r = journal_file_allocate(f, p, size);
1031 if (r < 0)
1032 return r;
1033
b439282e 1034 r = journal_file_move_to(f, type, false, p, size, &t, NULL);
cec736d2
LP		 1035 if (r < 0)
1036 return r;
1037
1038 o = (Object*) t;
1039
1040 zero(o->object);
de190aef 1041 o->object.type = type;
cec736d2
LP		 1042 o->object.size = htole64(size);
1043
1044 f->header->tail_object_offset = htole64(p);
cec736d2
LP		 1045 f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1);
1046
1047 *ret = o;
1048 *offset = p;
1049
1050 return 0;
1051}
1052
de190aef 1053static int journal_file_setup_data_hash_table(JournalFile *f) {
cec736d2
LP		 1054 uint64_t s, p;
1055 Object *o;
1056 int r;
1057
1058 assert(f);
c88cc6af 1059 assert(f->header);
cec736d2 1060
070052ab
LP		 1061 /* We estimate that we need 1 hash table entry per 768 bytes
1062 of journal file and we want to make sure we never get
1063 beyond 75% fill level. Calculate the hash table size for
1064 the maximum file size based on these metrics. */
4a92baf3 1065
dfabe643 1066 s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem);
4a92baf3
LP		 1067 if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
1068 s = DEFAULT_DATA_HASH_TABLE_SIZE;
1069
507f22bd 1070 log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem));
4a92baf3 1071
de190aef
LP		 1072 r = journal_file_append_object(f,
1073 OBJECT_DATA_HASH_TABLE,
1074 offsetof(Object, hash_table.items) + s,
1075 &o, &p);
cec736d2
LP		 1076 if (r < 0)
1077 return r;
1078
29804cc1 1079 memzero(o->hash_table.items, s);
cec736d2 1080
de190aef
LP		 1081 f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
1082 f->header->data_hash_table_size = htole64(s);
cec736d2
LP		 1083
1084 return 0;
1085}
1086
de190aef 1087static int journal_file_setup_field_hash_table(JournalFile *f) {
cec736d2
LP		 1088 uint64_t s, p;
1089 Object *o;
1090 int r;
1091
1092 assert(f);
c88cc6af 1093 assert(f->header);
cec736d2 1094
3c1668da
LP		 1095 /* We use a fixed size hash table for the fields as this
1096 * number should grow very slowly only */
1097
de190aef
LP		 1098 s = DEFAULT_FIELD_HASH_TABLE_SIZE;
1099 r = journal_file_append_object(f,
1100 OBJECT_FIELD_HASH_TABLE,
1101 offsetof(Object, hash_table.items) + s,
1102 &o, &p);
cec736d2
LP		 1103 if (r < 0)
1104 return r;
1105
29804cc1 1106 memzero(o->hash_table.items, s);
cec736d2 1107
de190aef
LP		 1108 f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
1109 f->header->field_hash_table_size = htole64(s);
cec736d2
LP		 1110
1111 return 0;
1112}
1113
dade37d4 1114int journal_file_map_data_hash_table(JournalFile *f) {
cec736d2
LP		 1115 uint64_t s, p;
1116 void *t;
1117 int r;
1118
1119 assert(f);
c88cc6af 1120 assert(f->header);
cec736d2 1121
dade37d4
LP		 1122 if (f->data_hash_table)
1123 return 0;
1124
de190aef
LP		 1125 p = le64toh(f->header->data_hash_table_offset);
1126 s = le64toh(f->header->data_hash_table_size);
cec736d2 1127
de190aef 1128 r = journal_file_move_to(f,
16e9f408 1129 OBJECT_DATA_HASH_TABLE,
fcde2389 1130 true,
de190aef 1131 p, s,
b42549ad 1132 &t, NULL);
cec736d2
LP		 1133 if (r < 0)
1134 return r;
1135
de190aef 1136 f->data_hash_table = t;
cec736d2
LP		 1137 return 0;
1138}
1139
dade37d4 1140int journal_file_map_field_hash_table(JournalFile *f) {
cec736d2
LP		 1141 uint64_t s, p;
1142 void *t;
1143 int r;
1144
1145 assert(f);
c88cc6af 1146 assert(f->header);
cec736d2 1147
dade37d4
LP		 1148 if (f->field_hash_table)
1149 return 0;
1150
de190aef
LP		 1151 p = le64toh(f->header->field_hash_table_offset);
1152 s = le64toh(f->header->field_hash_table_size);
cec736d2 1153
de190aef 1154 r = journal_file_move_to(f,
16e9f408 1155 OBJECT_FIELD_HASH_TABLE,
fcde2389 1156 true,
de190aef 1157 p, s,
b42549ad 1158 &t, NULL);
cec736d2
LP		 1159 if (r < 0)
1160 return r;
1161
de190aef 1162 f->field_hash_table = t;
cec736d2
LP		 1163 return 0;
1164}
1165
3c1668da
LP		 1166static int journal_file_link_field(
1167 JournalFile *f,
1168 Object *o,
1169 uint64_t offset,
1170 uint64_t hash) {
1171
805d1486 1172 uint64_t p, h, m;
3c1668da
LP		 1173 int r;
1174
1175 assert(f);
c88cc6af 1176 assert(f->header);
90d222c1 1177 assert(f->field_hash_table);
3c1668da
LP		 1178 assert(o);
1179 assert(offset > 0);
1180
1181 if (o->object.type != OBJECT_FIELD)
1182 return -EINVAL;
1183
805d1486
LP		 1184 m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
1185 if (m <= 0)
1186 return -EBADMSG;
3c1668da 1187
805d1486 1188 /* This might alter the window we are looking at */
3c1668da
LP		 1189 o->field.next_hash_offset = o->field.head_data_offset = 0;
1190
805d1486 1191 h = hash % m;
3c1668da
LP		 1192 p = le64toh(f->field_hash_table[h].tail_hash_offset);
1193 if (p == 0)
1194 f->field_hash_table[h].head_hash_offset = htole64(offset);
1195 else {
1196 r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
1197 if (r < 0)
1198 return r;
1199
1200 o->field.next_hash_offset = htole64(offset);
1201 }
1202
1203 f->field_hash_table[h].tail_hash_offset = htole64(offset);
1204
1205 if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
1206 f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1);
1207
1208 return 0;
1209}
1210
1211static int journal_file_link_data(
1212 JournalFile *f,
1213 Object *o,
1214 uint64_t offset,
1215 uint64_t hash) {
1216
805d1486 1217 uint64_t p, h, m;
cec736d2
LP		 1218 int r;
1219
1220 assert(f);
c88cc6af 1221 assert(f->header);
90d222c1 1222 assert(f->data_hash_table);
cec736d2
LP		 1223 assert(o);
1224 assert(offset > 0);
b588975f
LP		 1225
1226 if (o->object.type != OBJECT_DATA)
1227 return -EINVAL;
cec736d2 1228
805d1486
LP		 1229 m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem);
1230 if (m <= 0)
1231 return -EBADMSG;
48496df6 1232
805d1486 1233 /* This might alter the window we are looking at */
de190aef
LP		 1234 o->data.next_hash_offset = o->data.next_field_offset = 0;
1235 o->data.entry_offset = o->data.entry_array_offset = 0;
1236 o->data.n_entries = 0;
cec736d2 1237
805d1486 1238 h = hash % m;
8db4213e 1239 p = le64toh(f->data_hash_table[h].tail_hash_offset);
3c1668da 1240 if (p == 0)
cec736d2 1241 /* Only entry in the hash table is easy */
de190aef 1242 f->data_hash_table[h].head_hash_offset = htole64(offset);
3c1668da 1243 else {
48496df6
LP		 1244 /* Move back to the previous data object, to patch in
1245 * pointer */
cec736d2 1246
de190aef 1247 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
cec736d2
LP		 1248 if (r < 0)
1249 return r;
1250
de190aef 1251 o->data.next_hash_offset = htole64(offset);
cec736d2
LP		 1252 }
1253
de190aef 1254 f->data_hash_table[h].tail_hash_offset = htole64(offset);
cec736d2 1255
dca6219e
LP		 1256 if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
1257 f->header->n_data = htole64(le64toh(f->header->n_data) + 1);
1258
cec736d2
LP		 1259 return 0;
1260}
1261
3c1668da
LP		 1262int journal_file_find_field_object_with_hash(
1263 JournalFile *f,
1264 const void *field, uint64_t size, uint64_t hash,
1265 Object **ret, uint64_t *offset) {
1266
805d1486 1267 uint64_t p, osize, h, m;
3c1668da
LP		 1268 int r;
1269
1270 assert(f);
c88cc6af 1271 assert(f->header);
3c1668da
LP		 1272 assert(field && size > 0);
1273
dade37d4
LP		 1274 /* If the field hash table is empty, we can't find anything */
1275 if (le64toh(f->header->field_hash_table_size) <= 0)
1276 return 0;
1277
1278 /* Map the field hash table, if it isn't mapped yet. */
1279 r = journal_file_map_field_hash_table(f);
1280 if (r < 0)
1281 return r;
1282
3c1668da
LP		 1283 osize = offsetof(Object, field.payload) + size;
1284
805d1486 1285 m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
805d1486 1286 if (m <= 0)
3c1668da
LP		 1287 return -EBADMSG;
1288
805d1486 1289 h = hash % m;
3c1668da
LP		 1290 p = le64toh(f->field_hash_table[h].head_hash_offset);
1291
1292 while (p > 0) {
1293 Object *o;
1294
1295 r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
1296 if (r < 0)
1297 return r;
1298
1299 if (le64toh(o->field.hash) == hash &&
1300 le64toh(o->object.size) == osize &&
1301 memcmp(o->field.payload, field, size) == 0) {
1302
1303 if (ret)
1304 *ret = o;
1305 if (offset)
1306 *offset = p;
1307
1308 return 1;
1309 }
1310
1311 p = le64toh(o->field.next_hash_offset);
1312 }
1313
1314 return 0;
1315}
1316
1317int journal_file_find_field_object(
1318 JournalFile *f,
1319 const void *field, uint64_t size,
1320 Object **ret, uint64_t *offset) {
1321
1322 uint64_t hash;
1323
1324 assert(f);
1325 assert(field && size > 0);
1326
1327 hash = hash64(field, size);
1328
1329 return journal_file_find_field_object_with_hash(f,
1330 field, size, hash,
1331 ret, offset);
1332}
1333
de190aef
LP		 1334int journal_file_find_data_object_with_hash(
1335 JournalFile *f,
1336 const void *data, uint64_t size, uint64_t hash,
1337 Object **ret, uint64_t *offset) {
48496df6 1338
805d1486 1339 uint64_t p, osize, h, m;
cec736d2
LP		 1340 int r;
1341
1342 assert(f);
c88cc6af 1343 assert(f->header);
cec736d2
LP		 1344 assert(data || size == 0);
1345
dade37d4
LP		 1346 /* If there's no data hash table, then there's no entry. */
1347 if (le64toh(f->header->data_hash_table_size) <= 0)
1348 return 0;
1349
1350 /* Map the data hash table, if it isn't mapped yet. */
1351 r = journal_file_map_data_hash_table(f);
1352 if (r < 0)
1353 return r;
1354
cec736d2
LP		 1355 osize = offsetof(Object, data.payload) + size;
1356
805d1486
LP		 1357 m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem);
1358 if (m <= 0)
bc85bfee
LP		 1359 return -EBADMSG;
1360
805d1486 1361 h = hash % m;
de190aef 1362 p = le64toh(f->data_hash_table[h].head_hash_offset);
cec736d2 1363
de190aef
LP		 1364 while (p > 0) {
1365 Object *o;
cec736d2 1366
de190aef 1367 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
cec736d2
LP		 1368 if (r < 0)
1369 return r;
1370
807e17f0 1371 if (le64toh(o->data.hash) != hash)
85a131e8 1372 goto next;
807e17f0 1373
d89c8fdf 1374 if (o->object.flags & OBJECT_COMPRESSION_MASK) {
349cc4a5 1375#if HAVE_XZ || HAVE_LZ4
fa1c4b51 1376 uint64_t l;
a7f7d1bd 1377 size_t rsize = 0;
cec736d2 1378
807e17f0
LP		 1379 l = le64toh(o->object.size);
1380 if (l <= offsetof(Object, data.payload))
cec736d2
LP		 1381 return -EBADMSG;
1382
807e17f0
LP		 1383 l -= offsetof(Object, data.payload);
1384
d89c8fdf
ZJS		 1385 r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
1386 o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0);
1387 if (r < 0)
1388 return r;
807e17f0 1389
b785c858 1390 if (rsize == size &&
807e17f0
LP		 1391 memcmp(f->compress_buffer, data, size) == 0) {
1392
1393 if (ret)
1394 *ret = o;
1395
1396 if (offset)
1397 *offset = p;
1398
1399 return 1;
1400 }
3b1a55e1
ZJS		 1401#else
1402 return -EPROTONOSUPPORT;
1403#endif
807e17f0
LP		 1404 } else if (le64toh(o->object.size) == osize &&
1405 memcmp(o->data.payload, data, size) == 0) {
1406
cec736d2
LP		 1407 if (ret)
1408 *ret = o;
1409
1410 if (offset)
1411 *offset = p;
1412
de190aef 1413 return 1;
cec736d2
LP		 1414 }
1415
85a131e8 1416 next:
cec736d2
LP		 1417 p = le64toh(o->data.next_hash_offset);
1418 }
1419
de190aef
LP		 1420 return 0;
1421}
1422
1423int journal_file_find_data_object(
1424 JournalFile *f,
1425 const void *data, uint64_t size,
1426 Object **ret, uint64_t *offset) {
1427
1428 uint64_t hash;
1429
1430 assert(f);
1431 assert(data || size == 0);
1432
1433 hash = hash64(data, size);
1434
1435 return journal_file_find_data_object_with_hash(f,
1436 data, size, hash,
1437 ret, offset);
1438}
1439
3c1668da
LP		 1440static int journal_file_append_field(
1441 JournalFile *f,
1442 const void *field, uint64_t size,
1443 Object **ret, uint64_t *offset) {
1444
1445 uint64_t hash, p;
1446 uint64_t osize;
1447 Object *o;
1448 int r;
1449
1450 assert(f);
1451 assert(field && size > 0);
1452
1453 hash = hash64(field, size);
1454
1455 r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p);
1456 if (r < 0)
1457 return r;
1458 else if (r > 0) {
1459
1460 if (ret)
1461 *ret = o;
1462
1463 if (offset)
1464 *offset = p;
1465
1466 return 0;
1467 }
1468
1469 osize = offsetof(Object, field.payload) + size;
1470 r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p);
8c92d4bb
LP		 1471 if (r < 0)
1472 return r;
3c1668da
LP		 1473
1474 o->field.hash = htole64(hash);
1475 memcpy(o->field.payload, field, size);
1476
1477 r = journal_file_link_field(f, o, p, hash);
1478 if (r < 0)
1479 return r;
1480
1481 /* The linking might have altered the window, so let's
1482 * refresh our pointer */
1483 r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
1484 if (r < 0)
1485 return r;
1486
349cc4a5 1487#if HAVE_GCRYPT
3c1668da
LP		 1488 r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p);
1489 if (r < 0)
1490 return r;
1491#endif
1492
1493 if (ret)
1494 *ret = o;
1495
1496 if (offset)
1497 *offset = p;
1498
1499 return 0;
1500}
1501
48496df6
LP		 1502static int journal_file_append_data(
1503 JournalFile *f,
1504 const void *data, uint64_t size,
1505 Object **ret, uint64_t *offset) {
1506
de190aef
LP		 1507 uint64_t hash, p;
1508 uint64_t osize;
1509 Object *o;
d89c8fdf 1510 int r, compression = 0;
3c1668da 1511 const void *eq;
de190aef
LP		 1512
1513 assert(f);
1514 assert(data || size == 0);
1515
1516 hash = hash64(data, size);
1517
1518 r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p);
1519 if (r < 0)
1520 return r;
0240c603 1521 if (r > 0) {
de190aef
LP		 1522
1523 if (ret)
1524 *ret = o;
1525
1526 if (offset)
1527 *offset = p;
1528
1529 return 0;
1530 }
1531
1532 osize = offsetof(Object, data.payload) + size;
1533 r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p);
cec736d2
LP		 1534 if (r < 0)
1535 return r;
1536
cec736d2 1537 o->data.hash = htole64(hash);
807e17f0 1538
349cc4a5 1539#if HAVE_XZ || HAVE_LZ4
57850536 1540 if (JOURNAL_FILE_COMPRESS(f) && size >= f->compress_threshold_bytes) {
a7f7d1bd 1541 size_t rsize = 0;
807e17f0 1542
5d6f46b6 1543 compression = compress_blob(data, size, o->data.payload, size - 1, &rsize);
807e17f0 1544
d1afbcd2 1545 if (compression >= 0) {
807e17f0 1546 o->object.size = htole64(offsetof(Object, data.payload) + rsize);
d89c8fdf 1547 o->object.flags |= compression;
807e17f0 1548
fa1c4b51 1549 log_debug("Compressed data object %"PRIu64" -> %zu using %s",
d89c8fdf 1550 size, rsize, object_compressed_to_string(compression));
d1afbcd2
LP		 1551 } else
1552 /* Compression didn't work, we don't really care why, let's continue without compression */
1553 compression = 0;
807e17f0
LP		 1554 }
1555#endif
1556
75f32f04
ZJS		 1557 if (compression == 0)
1558 memcpy_safe(o->data.payload, data, size);
cec736d2 1559
de190aef 1560 r = journal_file_link_data(f, o, p, hash);
cec736d2
LP		 1561 if (r < 0)
1562 return r;
1563
349cc4a5 1564#if HAVE_GCRYPT
33685a5a
FB		 1565 r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p);
1566 if (r < 0)
1567 return r;
1568#endif
1569
48496df6
LP		 1570 /* The linking might have altered the window, so let's
1571 * refresh our pointer */
1572 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
1573 if (r < 0)
1574 return r;
1575
08c6f819
SL		 1576 if (!data)
1577 eq = NULL;
1578 else
1579 eq = memchr(data, '=', size);
3c1668da 1580 if (eq && eq > data) {
748db592 1581 Object *fo = NULL;
3c1668da 1582 uint64_t fp;
3c1668da
LP		 1583
1584 /* Create field object ... */
1585 r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp);
1586 if (r < 0)
1587 return r;
1588
1589 /* ... and link it in. */
1590 o->data.next_field_offset = fo->field.head_data_offset;
1591 fo->field.head_data_offset = le64toh(p);
1592 }
1593
cec736d2
LP		 1594 if (ret)
1595 *ret = o;
1596
1597 if (offset)
de190aef 1598 *offset = p;
cec736d2
LP		 1599
1600 return 0;
1601}
1602
1603uint64_t journal_file_entry_n_items(Object *o) {
1604 assert(o);
b588975f
LP		 1605
1606 if (o->object.type != OBJECT_ENTRY)
1607 return 0;
cec736d2
LP		 1608
1609 return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
1610}
1611
0284adc6 1612uint64_t journal_file_entry_array_n_items(Object *o) {
de190aef 1613 assert(o);
b588975f
LP		 1614
1615 if (o->object.type != OBJECT_ENTRY_ARRAY)
1616 return 0;
de190aef
LP		 1617
1618 return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
1619}
1620
fb9a24b6
LP		 1621uint64_t journal_file_hash_table_n_items(Object *o) {
1622 assert(o);
b588975f 1623
ec2ce0c5 1624 if (!IN_SET(o->object.type, OBJECT_DATA_HASH_TABLE, OBJECT_FIELD_HASH_TABLE))
b588975f 1625 return 0;
fb9a24b6
LP		 1626
1627 return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem);
1628}
1629
de190aef 1630static int link_entry_into_array(JournalFile *f,
4fd052ae
FC		 1631 le64_t *first,
1632 le64_t *idx,
de190aef 1633 uint64_t p) {
cec736d2 1634 int r;
de190aef
LP		 1635 uint64_t n = 0, ap = 0, q, i, a, hidx;
1636 Object *o;
1637
cec736d2 1638 assert(f);
c88cc6af 1639 assert(f->header);
de190aef
LP		 1640 assert(first);
1641 assert(idx);
1642 assert(p > 0);
cec736d2 1643
de190aef
LP		 1644 a = le64toh(*first);
1645 i = hidx = le64toh(*idx);
1646 while (a > 0) {
1647
1648 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
1649 if (r < 0)
1650 return r;
cec736d2 1651
de190aef
LP		 1652 n = journal_file_entry_array_n_items(o);
1653 if (i < n) {
1654 o->entry_array.items[i] = htole64(p);
1655 *idx = htole64(hidx + 1);
1656 return 0;
1657 }
cec736d2 1658
de190aef
LP		 1659 i -= n;
1660 ap = a;
1661 a = le64toh(o->entry_array.next_entry_array_offset);
1662 }
1663
1664 if (hidx > n)
1665 n = (hidx+1) * 2;
1666 else
1667 n = n * 2;
1668
1669 if (n < 4)
1670 n = 4;
1671
1672 r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY,
1673 offsetof(Object, entry_array.items) + n * sizeof(uint64_t),
1674 &o, &q);
cec736d2
LP		 1675 if (r < 0)
1676 return r;
1677
349cc4a5 1678#if HAVE_GCRYPT
5996c7c2 1679 r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q);
b0af6f41
LP		 1680 if (r < 0)
1681 return r;
feb12d3e 1682#endif
b0af6f41 1683
de190aef 1684 o->entry_array.items[i] = htole64(p);
cec736d2 1685
de190aef 1686 if (ap == 0)
7be3aa17 1687 *first = htole64(q);
cec736d2 1688 else {
de190aef 1689 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o);
cec736d2
LP		 1690 if (r < 0)
1691 return r;
1692
de190aef
LP		 1693 o->entry_array.next_entry_array_offset = htole64(q);
1694 }
cec736d2 1695
2dee23eb
LP		 1696 if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
1697 f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1);
1698
de190aef
LP		 1699 *idx = htole64(hidx + 1);
1700
1701 return 0;
1702}
cec736d2 1703
de190aef 1704static int link_entry_into_array_plus_one(JournalFile *f,
4fd052ae
FC		 1705 le64_t *extra,
1706 le64_t *first,
1707 le64_t *idx,
de190aef
LP		 1708 uint64_t p) {
1709
1710 int r;
1711
1712 assert(f);
1713 assert(extra);
1714 assert(first);
1715 assert(idx);
1716 assert(p > 0);
1717
1718 if (*idx == 0)
1719 *extra = htole64(p);
1720 else {
4fd052ae 1721 le64_t i;
de190aef 1722
7be3aa17 1723 i = htole64(le64toh(*idx) - 1);
de190aef
LP		 1724 r = link_entry_into_array(f, first, &i, p);
1725 if (r < 0)
1726 return r;
cec736d2
LP		 1727 }
1728
de190aef
LP		 1729 *idx = htole64(le64toh(*idx) + 1);
1730 return 0;
1731}
1732
1733static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
1734 uint64_t p;
1735 int r;
1736 assert(f);
1737 assert(o);
1738 assert(offset > 0);
1739
1740 p = le64toh(o->entry.items[i].object_offset);
1741 if (p == 0)
1742 return -EINVAL;
1743
1744 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
cec736d2
LP		 1745 if (r < 0)
1746 return r;
1747
de190aef
LP		 1748 return link_entry_into_array_plus_one(f,
1749 &o->data.entry_offset,
1750 &o->data.entry_array_offset,
1751 &o->data.n_entries,
1752 offset);
cec736d2
LP		 1753}
1754
1755static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
de190aef 1756 uint64_t n, i;
cec736d2
LP		 1757 int r;
1758
1759 assert(f);
c88cc6af 1760 assert(f->header);
cec736d2
LP		 1761 assert(o);
1762 assert(offset > 0);
b588975f
LP		 1763
1764 if (o->object.type != OBJECT_ENTRY)
1765 return -EINVAL;
cec736d2 1766
b788cc23
LP		 1767 __sync_synchronize();
1768
cec736d2 1769 /* Link up the entry itself */
de190aef
LP		 1770 r = link_entry_into_array(f,
1771 &f->header->entry_array_offset,
1772 &f->header->n_entries,
1773 offset);
1774 if (r < 0)
1775 return r;
cec736d2 1776
507f22bd 1777 /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */
cec736d2 1778
de190aef 1779 if (f->header->head_entry_realtime == 0)
0ac38b70 1780 f->header->head_entry_realtime = o->entry.realtime;
cec736d2 1781
0ac38b70 1782 f->header->tail_entry_realtime = o->entry.realtime;
de190aef
LP		 1783 f->header->tail_entry_monotonic = o->entry.monotonic;
1784
cec736d2
LP		 1785 /* Link up the items */
1786 n = journal_file_entry_n_items(o);
1787 for (i = 0; i < n; i++) {
1788 r = journal_file_link_entry_item(f, o, offset, i);
1789 if (r < 0)
1790 return r;
1791 }
1792
cec736d2
LP		 1793 return 0;
1794}
1795
1796static int journal_file_append_entry_internal(
1797 JournalFile *f,
1798 const dual_timestamp *ts,
d180c349 1799 const sd_id128_t *boot_id,
cec736d2
LP		 1800 uint64_t xor_hash,
1801 const EntryItem items[], unsigned n_items,
de190aef 1802 uint64_t *seqnum,
cec736d2
LP		 1803 Object **ret, uint64_t *offset) {
1804 uint64_t np;
1805 uint64_t osize;
1806 Object *o;
1807 int r;
1808
1809 assert(f);
c88cc6af 1810 assert(f->header);
cec736d2 1811 assert(items || n_items == 0);
de190aef 1812 assert(ts);
cec736d2
LP		 1813
1814 osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem));
1815
de190aef 1816 r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np);
cec736d2
LP		 1817 if (r < 0)
1818 return r;
1819
d98cc1f2 1820 o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum));
75f32f04 1821 memcpy_safe(o->entry.items, items, n_items * sizeof(EntryItem));
de190aef
LP		 1822 o->entry.realtime = htole64(ts->realtime);
1823 o->entry.monotonic = htole64(ts->monotonic);
cec736d2 1824 o->entry.xor_hash = htole64(xor_hash);
d180c349 1825 o->entry.boot_id = boot_id ? *boot_id : f->header->boot_id;
cec736d2 1826
349cc4a5 1827#if HAVE_GCRYPT
5996c7c2 1828 r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np);
b0af6f41
LP		 1829 if (r < 0)
1830 return r;
feb12d3e 1831#endif
b0af6f41 1832
cec736d2
LP		 1833 r = journal_file_link_entry(f, o, np);
1834 if (r < 0)
1835 return r;
1836
1837 if (ret)
1838 *ret = o;
1839
1840 if (offset)
1841 *offset = np;
1842
1843 return 0;
1844}
1845
cf244689 1846void journal_file_post_change(JournalFile *f) {
50f20cfd
LP		 1847 assert(f);
1848
c5236850
DT		 1849 if (f->fd < 0)
1850 return;
1851
50f20cfd
LP		 1852 /* inotify() does not receive IN_MODIFY events from file
1853 * accesses done via mmap(). After each access we hence
1854 * trigger IN_MODIFY by truncating the journal file to its
1855 * current size which triggers IN_MODIFY. */
1856
bc85bfee
LP		 1857 __sync_synchronize();
1858
50f20cfd 1859 if (ftruncate(f->fd, f->last_stat.st_size) < 0)
e167d7fd 1860 log_debug_errno(errno, "Failed to truncate file to its own size: %m");
50f20cfd
LP		 1861}
1862
7a24f3bf
VC		 1863static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) {
1864 assert(userdata);
1865
1866 journal_file_post_change(userdata);
1867
1868 return 1;
1869}
1870
1871static void schedule_post_change(JournalFile *f) {
1872 sd_event_source *timer;
1873 int enabled, r;
1874 uint64_t now;
1875
1876 assert(f);
1877 assert(f->post_change_timer);
1878
1879 timer = f->post_change_timer;
1880
1881 r = sd_event_source_get_enabled(timer, &enabled);
1882 if (r < 0) {
e167d7fd
LP		 1883 log_debug_errno(r, "Failed to get ftruncate timer state: %m");
1884 goto fail;
7a24f3bf
VC		 1885 }
1886
1887 if (enabled == SD_EVENT_ONESHOT)
1888 return;
1889
1890 r = sd_event_now(sd_event_source_get_event(timer), CLOCK_MONOTONIC, &now);
1891 if (r < 0) {
e167d7fd
LP		 1892 log_debug_errno(r, "Failed to get clock's now for scheduling ftruncate: %m");
1893 goto fail;
7a24f3bf
VC		 1894 }
1895
1896 r = sd_event_source_set_time(timer, now+f->post_change_timer_period);
1897 if (r < 0) {
e167d7fd
LP		 1898 log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m");
1899 goto fail;
7a24f3bf
VC		 1900 }
1901
1902 r = sd_event_source_set_enabled(timer, SD_EVENT_ONESHOT);
1903 if (r < 0) {
e167d7fd
LP		 1904 log_debug_errno(r, "Failed to enable scheduled ftruncate: %m");
1905 goto fail;
7a24f3bf 1906 }
e167d7fd
LP		 1907
1908 return;
1909
1910fail:
1911 /* On failure, let's simply post the change immediately. */
1912 journal_file_post_change(f);
7a24f3bf
VC		 1913}
1914
1915/* Enable coalesced change posting in a timer on the provided sd_event instance */
1916int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) {
1917 _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL;
1918 int r;
1919
1920 assert(f);
1921 assert_return(!f->post_change_timer, -EINVAL);
1922 assert(e);
1923 assert(t);
1924
1925 r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f);
1926 if (r < 0)
1927 return r;
1928
1929 r = sd_event_source_set_enabled(timer, SD_EVENT_OFF);
1930 if (r < 0)
1931 return r;
1932
1cc6c93a 1933 f->post_change_timer = TAKE_PTR(timer);
7a24f3bf
VC		 1934 f->post_change_timer_period = t;
1935
1936 return r;
1937}
1938
93bab288
YW		 1939static int entry_item_cmp(const EntryItem *a, const EntryItem *b) {
1940 return CMP(le64toh(a->object_offset), le64toh(b->object_offset));
1f2da9ec
LP		 1941}
1942
d180c349
ZJS		 1943int journal_file_append_entry(
1944 JournalFile *f,
1945 const dual_timestamp *ts,
1946 const sd_id128_t *boot_id,
1947 const struct iovec iovec[], unsigned n_iovec,
1948 uint64_t *seqnum,
1949 Object **ret, uint64_t *offset) {
1950
cec736d2
LP		 1951 unsigned i;
1952 EntryItem *items;
1953 int r;
1954 uint64_t xor_hash = 0;
de190aef 1955 struct dual_timestamp _ts;
cec736d2
LP		 1956
1957 assert(f);
c88cc6af 1958 assert(f->header);
cec736d2
LP		 1959 assert(iovec || n_iovec == 0);
1960
c6273953
ZJS		 1961 if (ts) {
1962 if (!VALID_REALTIME(ts->realtime)) {
1963 log_debug("Invalid realtime timestamp %"PRIu64", refusing entry.", ts->realtime);
1964 return -EBADMSG;
1965 }
1966 if (!VALID_MONOTONIC(ts->monotonic)) {
1967 log_debug("Invalid monotomic timestamp %"PRIu64", refusing entry.", ts->monotonic);
1968 return -EBADMSG;
1969 }
1970 } else {
de190aef
LP		 1971 dual_timestamp_get(&_ts);
1972 ts = &_ts;
1973 }
1974
349cc4a5 1975#if HAVE_GCRYPT
7560fffc
LP		 1976 r = journal_file_maybe_append_tag(f, ts->realtime);
1977 if (r < 0)
1978 return r;
feb12d3e 1979#endif
7560fffc 1980
64825d3c 1981 /* alloca() can't take 0, hence let's allocate at least one */
cf409d15 1982 items = newa(EntryItem, MAX(1u, n_iovec));
cec736d2
LP		 1983
1984 for (i = 0; i < n_iovec; i++) {
1985 uint64_t p;
1986 Object *o;
1987
1988 r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p);
1989 if (r < 0)
cf244689 1990 return r;
cec736d2
LP		 1991
1992 xor_hash ^= le64toh(o->data.hash);
1993 items[i].object_offset = htole64(p);
de7b95cd 1994 items[i].hash = o->data.hash;
cec736d2
LP		 1995 }
1996
1f2da9ec
LP		 1997 /* Order by the position on disk, in order to improve seek
1998 * times for rotating media. */
93bab288 1999 typesafe_qsort(items, n_iovec, entry_item_cmp);
1f2da9ec 2000
d180c349 2001 r = journal_file_append_entry_internal(f, ts, boot_id, xor_hash, items, n_iovec, seqnum, ret, offset);
cec736d2 2002
fa6ac760
LP		 2003 /* If the memory mapping triggered a SIGBUS then we return an
2004 * IO error and ignore the error code passed down to us, since
2005 * it is very likely just an effect of a nullified replacement
2006 * mapping page */
2007
be7cdd8e 2008 if (mmap_cache_got_sigbus(f->mmap, f->cache_fd))
fa6ac760
LP		 2009 r = -EIO;
2010
7a24f3bf
VC		 2011 if (f->post_change_timer)
2012 schedule_post_change(f);
2013 else
2014 journal_file_post_change(f);
50f20cfd 2015
cec736d2
LP		 2016 return r;
2017}
2018
a4bcff5b 2019typedef struct ChainCacheItem {
fb099c8d 2020 uint64_t first; /* the array at the beginning of the chain */
a4bcff5b
LP		 2021 uint64_t array; /* the cached array */
2022 uint64_t begin; /* the first item in the cached array */
2023 uint64_t total; /* the total number of items in all arrays before this one in the chain */
f268980d 2024 uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */
a4bcff5b
LP		 2025} ChainCacheItem;
2026
2027static void chain_cache_put(
4743015d 2028 OrderedHashmap *h,
a4bcff5b
LP		 2029 ChainCacheItem *ci,
2030 uint64_t first,
2031 uint64_t array,
2032 uint64_t begin,
f268980d
LP		 2033 uint64_t total,
2034 uint64_t last_index) {
a4bcff5b
LP		 2035
2036 if (!ci) {
34741aa3
LP		 2037 /* If the chain item to cache for this chain is the
2038 * first one it's not worth caching anything */
2039 if (array == first)
2040 return;
2041
29433089 2042 if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) {
4743015d 2043 ci = ordered_hashmap_steal_first(h);
29433089
LP		 2044 assert(ci);
2045 } else {
a4bcff5b
LP		 2046 ci = new(ChainCacheItem, 1);
2047 if (!ci)
2048 return;
2049 }
2050
2051 ci->first = first;
2052
4743015d 2053 if (ordered_hashmap_put(h, &ci->first, ci) < 0) {
a4bcff5b
LP		 2054 free(ci);
2055 return;
2056 }
2057 } else
2058 assert(ci->first == first);
2059
2060 ci->array = array;
2061 ci->begin = begin;
2062 ci->total = total;
f268980d 2063 ci->last_index = last_index;
a4bcff5b
LP		 2064}
2065
f268980d
LP		 2066static int generic_array_get(
2067 JournalFile *f,
2068 uint64_t first,
2069 uint64_t i,
2070 Object **ret, uint64_t *offset) {
de190aef 2071
cec736d2 2072 Object *o;
a4bcff5b 2073 uint64_t p = 0, a, t = 0;
cec736d2 2074 int r;
a4bcff5b 2075 ChainCacheItem *ci;
cec736d2
LP		 2076
2077 assert(f);
2078
de190aef 2079 a = first;
a4bcff5b
LP		 2080
2081 /* Try the chain cache first */
4743015d 2082 ci = ordered_hashmap_get(f->chain_cache, &first);
a4bcff5b
LP		 2083 if (ci && i > ci->total) {
2084 a = ci->array;
2085 i -= ci->total;
2086 t = ci->total;
2087 }
2088
de190aef 2089 while (a > 0) {
a4bcff5b 2090 uint64_t k;
cec736d2 2091
de190aef
LP		 2092 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
2093 if (r < 0)
2094 return r;
cec736d2 2095
a4bcff5b
LP		 2096 k = journal_file_entry_array_n_items(o);
2097 if (i < k) {
de190aef 2098 p = le64toh(o->entry_array.items[i]);
a4bcff5b 2099 goto found;
cec736d2
LP		 2100 }
2101
a4bcff5b
LP		 2102 i -= k;
2103 t += k;
de190aef
LP		 2104 a = le64toh(o->entry_array.next_entry_array_offset);
2105 }
2106
a4bcff5b
LP		 2107 return 0;
2108
2109found:
2110 /* Let's cache this item for the next invocation */
af13a6b0 2111 chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i);
de190aef
LP		 2112
2113 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
2114 if (r < 0)
2115 return r;
2116
2117 if (ret)
2118 *ret = o;
2119
2120 if (offset)
2121 *offset = p;
2122
2123 return 1;
2124}
2125
f268980d
LP		 2126static int generic_array_get_plus_one(
2127 JournalFile *f,
2128 uint64_t extra,
2129 uint64_t first,
2130 uint64_t i,
2131 Object **ret, uint64_t *offset) {
de190aef
LP		 2132
2133 Object *o;
2134
2135 assert(f);
2136
2137 if (i == 0) {
2138 int r;
2139
2140 r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
cec736d2
LP		 2141 if (r < 0)
2142 return r;
2143
de190aef
LP		 2144 if (ret)
2145 *ret = o;
cec736d2 2146
de190aef
LP		 2147 if (offset)
2148 *offset = extra;
cec736d2 2149
de190aef 2150 return 1;
cec736d2
LP		 2151 }
2152
de190aef
LP		 2153 return generic_array_get(f, first, i-1, ret, offset);
2154}
cec736d2 2155
de190aef
LP		 2156enum {
2157 TEST_FOUND,
2158 TEST_LEFT,
2159 TEST_RIGHT
2160};
cec736d2 2161
f268980d
LP		 2162static int generic_array_bisect(
2163 JournalFile *f,
2164 uint64_t first,
2165 uint64_t n,
2166 uint64_t needle,
2167 int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
2168 direction_t direction,
2169 Object **ret,
2170 uint64_t *offset,
2171 uint64_t *idx) {
2172
2173 uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1;
de190aef
LP		 2174 bool subtract_one = false;
2175 Object *o, *array = NULL;
2176 int r;
a4bcff5b 2177 ChainCacheItem *ci;
cec736d2 2178
de190aef
LP		 2179 assert(f);
2180 assert(test_object);
cec736d2 2181
a4bcff5b 2182 /* Start with the first array in the chain */
de190aef 2183 a = first;
a4bcff5b 2184
4743015d 2185 ci = ordered_hashmap_get(f->chain_cache, &first);
96d4d024 2186 if (ci && n > ci->total && ci->begin != 0) {
a4bcff5b
LP		 2187 /* Ah, we have iterated this bisection array chain
2188 * previously! Let's see if we can skip ahead in the
2189 * chain, as far as the last time. But we can't jump
2190 * backwards in the chain, so let's check that
2191 * first. */
2192
2193 r = test_object(f, ci->begin, needle);
2194 if (r < 0)
2195 return r;
2196
2197 if (r == TEST_LEFT) {
f268980d 2198 /* OK, what we are looking for is right of the
a4bcff5b
LP		 2199 * begin of this EntryArray, so let's jump
2200 * straight to previously cached array in the
2201 * chain */
2202
2203 a = ci->array;
2204 n -= ci->total;
2205 t = ci->total;
f268980d 2206 last_index = ci->last_index;
a4bcff5b
LP		 2207 }
2208 }
2209
de190aef
LP		 2210 while (a > 0) {
2211 uint64_t left, right, k, lp;
2212
2213 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array);
cec736d2
LP		 2214 if (r < 0)
2215 return r;
2216
de190aef
LP		 2217 k = journal_file_entry_array_n_items(array);
2218 right = MIN(k, n);
2219 if (right <= 0)
2220 return 0;
cec736d2 2221
de190aef
LP		 2222 i = right - 1;
2223 lp = p = le64toh(array->entry_array.items[i]);
2224 if (p <= 0)
bee6a291
LP		 2225 r = -EBADMSG;
2226 else
2227 r = test_object(f, p, needle);
2228 if (r == -EBADMSG) {
2229 log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)");
2230 n = i;
2231 continue;
2232 }
de190aef
LP		 2233 if (r < 0)
2234 return r;
cec736d2 2235
de190aef
LP		 2236 if (r == TEST_FOUND)
2237 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
2238
2239 if (r == TEST_RIGHT) {
2240 left = 0;
2241 right -= 1;
f268980d
LP		 2242
2243 if (last_index != (uint64_t) -1) {
2244 assert(last_index <= right);
2245
2246 /* If we cached the last index we
2247 * looked at, let's try to not to jump
2248 * too wildly around and see if we can
2249 * limit the range to look at early to
2250 * the immediate neighbors of the last
2251 * index we looked at. */
2252
2253 if (last_index > 0) {
2254 uint64_t x = last_index - 1;
2255
2256 p = le64toh(array->entry_array.items[x]);
2257 if (p <= 0)
2258 return -EBADMSG;
2259
2260 r = test_object(f, p, needle);
2261 if (r < 0)
2262 return r;
2263
2264 if (r == TEST_FOUND)
2265 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
2266
2267 if (r == TEST_RIGHT)
2268 right = x;
2269 else
2270 left = x + 1;
2271 }
2272
2273 if (last_index < right) {
2274 uint64_t y = last_index + 1;
2275
2276 p = le64toh(array->entry_array.items[y]);
2277 if (p <= 0)
2278 return -EBADMSG;
2279
2280 r = test_object(f, p, needle);
2281 if (r < 0)
2282 return r;
2283
2284 if (r == TEST_FOUND)
2285 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
2286
2287 if (r == TEST_RIGHT)
2288 right = y;
2289 else
2290 left = y + 1;
2291 }
f268980d
LP		 2292 }
2293
de190aef
LP		 2294 for (;;) {
2295 if (left == right) {
2296 if (direction == DIRECTION_UP)
2297 subtract_one = true;
2298
2299 i = left;
2300 goto found;
2301 }
2302
2303 assert(left < right);
de190aef 2304 i = (left + right) / 2;
f268980d 2305
de190aef
LP		 2306 p = le64toh(array->entry_array.items[i]);
2307 if (p <= 0)
bee6a291
LP		 2308 r = -EBADMSG;
2309 else
2310 r = test_object(f, p, needle);
2311 if (r == -EBADMSG) {
2312 log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)");
2313 right = n = i;
2314 continue;
2315 }
de190aef
LP		 2316 if (r < 0)
2317 return r;
cec736d2 2318
de190aef
LP		 2319 if (r == TEST_FOUND)
2320 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
2321
2322 if (r == TEST_RIGHT)
2323 right = i;
2324 else
2325 left = i + 1;
2326 }
2327 }
2328
2173cbf8 2329 if (k >= n) {
cbdca852
LP		 2330 if (direction == DIRECTION_UP) {
2331 i = n;
2332 subtract_one = true;
2333 goto found;
2334 }
2335
cec736d2 2336 return 0;
cbdca852 2337 }
cec736d2 2338
de190aef
LP		 2339 last_p = lp;
2340
2341 n -= k;
2342 t += k;
f268980d 2343 last_index = (uint64_t) -1;
de190aef 2344 a = le64toh(array->entry_array.next_entry_array_offset);
cec736d2
LP		 2345 }
2346
2347 return 0;
de190aef
LP		 2348
2349found:
2350 if (subtract_one && t == 0 && i == 0)
2351 return 0;
2352
a4bcff5b 2353 /* Let's cache this item for the next invocation */
af13a6b0 2354 chain_cache_put(f->chain_cache, ci, first, a, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i);
a4bcff5b 2355
de190aef
LP		 2356 if (subtract_one && i == 0)
2357 p = last_p;
2358 else if (subtract_one)
2359 p = le64toh(array->entry_array.items[i-1]);
2360 else
2361 p = le64toh(array->entry_array.items[i]);
2362
2363 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
2364 if (r < 0)
2365 return r;
2366
2367 if (ret)
2368 *ret = o;
2369
2370 if (offset)
2371 *offset = p;
2372
2373 if (idx)
cbdca852 2374 *idx = t + i + (subtract_one ? -1 : 0);
de190aef
LP		 2375
2376 return 1;
cec736d2
LP		 2377}
2378
f268980d
LP		 2379static int generic_array_bisect_plus_one(
2380 JournalFile *f,
2381 uint64_t extra,
2382 uint64_t first,
2383 uint64_t n,
2384 uint64_t needle,
2385 int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
2386 direction_t direction,
2387 Object **ret,
2388 uint64_t *offset,
2389 uint64_t *idx) {
de190aef 2390
cec736d2 2391 int r;
cbdca852
LP		 2392 bool step_back = false;
2393 Object *o;
cec736d2
LP		 2394
2395 assert(f);
de190aef 2396 assert(test_object);
cec736d2 2397
de190aef
LP		 2398 if (n <= 0)
2399 return 0;
cec736d2 2400
de190aef
LP		 2401 /* This bisects the array in object 'first', but first checks
2402 * an extra */
de190aef
LP		 2403 r = test_object(f, extra, needle);
2404 if (r < 0)
2405 return r;
a536e261
LP		 2406
2407 if (r == TEST_FOUND)
2408 r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
2409
cbdca852
LP		 2410 /* if we are looking with DIRECTION_UP then we need to first
2411 see if in the actual array there is a matching entry, and
2412 return the last one of that. But if there isn't any we need
2413 to return this one. Hence remember this, and return it
2414 below. */
2415 if (r == TEST_LEFT)
2416 step_back = direction == DIRECTION_UP;
de190aef 2417
cbdca852
LP		 2418 if (r == TEST_RIGHT) {
2419 if (direction == DIRECTION_DOWN)
2420 goto found;
2421 else
2422 return 0;
a536e261 2423 }
cec736d2 2424
de190aef
LP		 2425 r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx);
2426
cbdca852
LP		 2427 if (r == 0 && step_back)
2428 goto found;
2429
ecf68b1d 2430 if (r > 0 && idx)
313cefa1 2431 (*idx)++;
de190aef
LP		 2432
2433 return r;
cbdca852
LP		 2434
2435found:
2436 r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
2437 if (r < 0)
2438 return r;
2439
2440 if (ret)
2441 *ret = o;
2442
2443 if (offset)
2444 *offset = extra;
2445
2446 if (idx)
2447 *idx = 0;
2448
2449 return 1;
2450}
2451
44a6b1b6 2452_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
cbdca852
LP		 2453 assert(f);
2454 assert(p > 0);
2455
2456 if (p == needle)
2457 return TEST_FOUND;
2458 else if (p < needle)
2459 return TEST_LEFT;
2460 else
2461 return TEST_RIGHT;
2462}
2463
de190aef
LP		 2464static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
2465 Object *o;
2466 int r;
2467
2468 assert(f);
2469 assert(p > 0);
2470
2471 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
cec736d2
LP		 2472 if (r < 0)
2473 return r;
2474
de190aef
LP		 2475 if (le64toh(o->entry.seqnum) == needle)
2476 return TEST_FOUND;
2477 else if (le64toh(o->entry.seqnum) < needle)
2478 return TEST_LEFT;
2479 else
2480 return TEST_RIGHT;
2481}
cec736d2 2482
de190aef
LP		 2483int journal_file_move_to_entry_by_seqnum(
2484 JournalFile *f,
2485 uint64_t seqnum,
2486 direction_t direction,
2487 Object **ret,
2488 uint64_t *offset) {
c88cc6af
VC		 2489 assert(f);
2490 assert(f->header);
de190aef
LP		 2491
2492 return generic_array_bisect(f,
2493 le64toh(f->header->entry_array_offset),
2494 le64toh(f->header->n_entries),
2495 seqnum,
2496 test_object_seqnum,
2497 direction,
2498 ret, offset, NULL);
2499}
cec736d2 2500
de190aef
LP		 2501static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) {
2502 Object *o;
2503 int r;
2504
2505 assert(f);
2506 assert(p > 0);
2507
2508 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
2509 if (r < 0)
2510 return r;
2511
2512 if (le64toh(o->entry.realtime) == needle)
2513 return TEST_FOUND;
2514 else if (le64toh(o->entry.realtime) < needle)
2515 return TEST_LEFT;
2516 else
2517 return TEST_RIGHT;
cec736d2
LP		 2518}
2519
de190aef
LP		 2520int journal_file_move_to_entry_by_realtime(
2521 JournalFile *f,
2522 uint64_t realtime,
2523 direction_t direction,
2524 Object **ret,
2525 uint64_t *offset) {
c88cc6af
VC		 2526 assert(f);
2527 assert(f->header);
de190aef
LP		 2528
2529 return generic_array_bisect(f,
2530 le64toh(f->header->entry_array_offset),
2531 le64toh(f->header->n_entries),
2532 realtime,
2533 test_object_realtime,
2534 direction,
2535 ret, offset, NULL);
2536}
2537
2538static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
2539 Object *o;
2540 int r;
2541
2542 assert(f);
2543 assert(p > 0);
2544
2545 r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
2546 if (r < 0)
2547 return r;
2548
2549 if (le64toh(o->entry.monotonic) == needle)
2550 return TEST_FOUND;
2551 else if (le64toh(o->entry.monotonic) < needle)
2552 return TEST_LEFT;
2553 else
2554 return TEST_RIGHT;
2555}
2556
2a560338 2557static int find_data_object_by_boot_id(
47838ab3
ZJS		 2558 JournalFile *f,
2559 sd_id128_t boot_id,
2560 Object **o,
2561 uint64_t *b) {
2a560338 2562
fbd0b64f 2563 char t[STRLEN("_BOOT_ID=") + 32 + 1] = "_BOOT_ID=";
47838ab3
ZJS		 2564
2565 sd_id128_to_string(boot_id, t + 9);
2566 return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b);
2567}
2568
de190aef
LP		 2569int journal_file_move_to_entry_by_monotonic(
2570 JournalFile *f,
2571 sd_id128_t boot_id,
2572 uint64_t monotonic,
2573 direction_t direction,
2574 Object **ret,
2575 uint64_t *offset) {
2576
de190aef
LP		 2577 Object *o;
2578 int r;
2579
cbdca852 2580 assert(f);
de190aef 2581
47838ab3 2582 r = find_data_object_by_boot_id(f, boot_id, &o, NULL);
de190aef
LP		 2583 if (r < 0)
2584 return r;
cbdca852 2585 if (r == 0)
de190aef
LP		 2586 return -ENOENT;
2587
2588 return generic_array_bisect_plus_one(f,
2589 le64toh(o->data.entry_offset),
2590 le64toh(o->data.entry_array_offset),
2591 le64toh(o->data.n_entries),
2592 monotonic,
2593 test_object_monotonic,
2594 direction,
2595 ret, offset, NULL);
2596}
2597
1fc605b0 2598void journal_file_reset_location(JournalFile *f) {
6573ef05 2599 f->location_type = LOCATION_HEAD;
1fc605b0 2600 f->current_offset = 0;
6573ef05
MS		 2601 f->current_seqnum = 0;
2602 f->current_realtime = 0;
2603 f->current_monotonic = 0;
2604 zero(f->current_boot_id);
2605 f->current_xor_hash = 0;
2606}
2607
950c07d4 2608void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) {
6573ef05
MS		 2609 f->location_type = LOCATION_SEEK;
2610 f->current_offset = offset;
2611 f->current_seqnum = le64toh(o->entry.seqnum);
2612 f->current_realtime = le64toh(o->entry.realtime);
2613 f->current_monotonic = le64toh(o->entry.monotonic);
2614 f->current_boot_id = o->entry.boot_id;
2615 f->current_xor_hash = le64toh(o->entry.xor_hash);
1fc605b0
MS		 2616}
2617
d8ae66d7
MS		 2618int journal_file_compare_locations(JournalFile *af, JournalFile *bf) {
2619 assert(af);
c88cc6af 2620 assert(af->header);
d8ae66d7 2621 assert(bf);
c88cc6af 2622 assert(bf->header);
d8ae66d7
MS		 2623 assert(af->location_type == LOCATION_SEEK);
2624 assert(bf->location_type == LOCATION_SEEK);
2625
2626 /* If contents and timestamps match, these entries are
2627 * identical, even if the seqnum does not match */
2628 if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) &&
2629 af->current_monotonic == bf->current_monotonic &&
2630 af->current_realtime == bf->current_realtime &&
2631 af->current_xor_hash == bf->current_xor_hash)
2632 return 0;
2633
2634 if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) {
2635
2636 /* If this is from the same seqnum source, compare
2637 * seqnums */
2638 if (af->current_seqnum < bf->current_seqnum)
2639 return -1;
2640 if (af->current_seqnum > bf->current_seqnum)
2641 return 1;
2642
2643 /* Wow! This is weird, different data but the same
2644 * seqnums? Something is borked, but let's make the
2645 * best of it and compare by time. */
2646 }
2647
2648 if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) {
2649
2650 /* If the boot id matches, compare monotonic time */
2651 if (af->current_monotonic < bf->current_monotonic)
2652 return -1;
2653 if (af->current_monotonic > bf->current_monotonic)
2654 return 1;
2655 }
2656
2657 /* Otherwise, compare UTC time */
2658 if (af->current_realtime < bf->current_realtime)
2659 return -1;
2660 if (af->current_realtime > bf->current_realtime)
2661 return 1;
2662
2663 /* Finally, compare by contents */
6dd91b36 2664 return CMP(af->current_xor_hash, bf->current_xor_hash);
d8ae66d7
MS		 2665}
2666
aa598ba5
LP		 2667static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) {
2668
2669 /* Increase or decrease the specified index, in the right direction. */
2670
2671 if (direction == DIRECTION_DOWN) {
2672 if (*i >= n - 1)
2673 return 0;
2674
2675 (*i) ++;
2676 } else {
2677 if (*i <= 0)
2678 return 0;
2679
2680 (*i) --;
2681 }
2682
2683 return 1;
2684}
2685
b6da4ed0
LP		 2686static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) {
2687
2688 /* Consider it an error if any of the two offsets is uninitialized */
2689 if (old_offset == 0 || new_offset == 0)
2690 return false;
2691
2692 /* If we go down, the new offset must be larger than the old one. */
2693 return direction == DIRECTION_DOWN ?
2694 new_offset > old_offset :
2695 new_offset < old_offset;
2696}
2697
de190aef
LP		 2698int journal_file_next_entry(
2699 JournalFile *f,
f534928a 2700 uint64_t p,
de190aef
LP		 2701 direction_t direction,
2702 Object **ret, uint64_t *offset) {
2703
fb099c8d 2704 uint64_t i, n, ofs;
cec736d2
LP		 2705 int r;
2706
2707 assert(f);
c88cc6af 2708 assert(f->header);
de190aef
LP		 2709
2710 n = le64toh(f->header->n_entries);
2711 if (n <= 0)
2712 return 0;
cec736d2 2713
f534928a 2714 if (p == 0)
de190aef 2715 i = direction == DIRECTION_DOWN ? 0 : n - 1;
cec736d2 2716 else {
de190aef
LP		 2717 r = generic_array_bisect(f,
2718 le64toh(f->header->entry_array_offset),
2719 le64toh(f->header->n_entries),
2720 p,
2721 test_object_offset,
2722 DIRECTION_DOWN,
2723 NULL, NULL,
2724 &i);
2725 if (r <= 0)
2726 return r;
2727
aa598ba5
LP		 2728 r = bump_array_index(&i, direction, n);
2729 if (r <= 0)
2730 return r;
cec736d2
LP		 2731 }
2732
de190aef 2733 /* And jump to it */
989793d3
LP		 2734 for (;;) {
2735 r = generic_array_get(f,
2736 le64toh(f->header->entry_array_offset),
2737 i,
2738 ret, &ofs);
2739 if (r > 0)
2740 break;
2741 if (r != -EBADMSG)
2742 return r;
2743
2744 /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if
2745 * the next one might work for us instead. */
2746 log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i);
2747
2748 r = bump_array_index(&i, direction, n);
2749 if (r <= 0)
2750 return r;
caeab8f6 2751 }
fb099c8d 2752
b6da4ed0
LP		 2753 /* Ensure our array is properly ordered. */
2754 if (p > 0 && !check_properly_ordered(ofs, p, direction)) {
2755 log_debug("%s: entry array not properly ordered at entry %" PRIu64, f->path, i);
fb099c8d
ZJS		 2756 return -EBADMSG;
2757 }
2758
2759 if (offset)
2760 *offset = ofs;
2761
2762 return 1;
de190aef 2763}
cec736d2 2764
de190aef
LP		 2765int journal_file_next_entry_for_data(
2766 JournalFile *f,
2767 Object *o, uint64_t p,
2768 uint64_t data_offset,
2769 direction_t direction,
2770 Object **ret, uint64_t *offset) {
2771
ded5034e 2772 uint64_t i, n, ofs;
de190aef 2773 Object *d;
989793d3 2774 int r;
cec736d2
LP		 2775
2776 assert(f);
de190aef 2777 assert(p > 0 || !o);
cec736d2 2778
de190aef 2779 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
466ccd92 2780 if (r < 0)
de190aef 2781 return r;
cec736d2 2782
de190aef
LP		 2783 n = le64toh(d->data.n_entries);
2784 if (n <= 0)
2785 return n;
cec736d2 2786
de190aef
LP		 2787 if (!o)
2788 i = direction == DIRECTION_DOWN ? 0 : n - 1;
2789 else {
2790 if (o->object.type != OBJECT_ENTRY)
2791 return -EINVAL;
cec736d2 2792
de190aef
LP		 2793 r = generic_array_bisect_plus_one(f,
2794 le64toh(d->data.entry_offset),
2795 le64toh(d->data.entry_array_offset),
2796 le64toh(d->data.n_entries),
2797 p,
2798 test_object_offset,
2799 DIRECTION_DOWN,
2800 NULL, NULL,
2801 &i);
2802
2803 if (r <= 0)
cec736d2
LP		 2804 return r;
2805
aa598ba5
LP		 2806 r = bump_array_index(&i, direction, n);
2807 if (r <= 0)
2808 return r;
de190aef 2809 }
cec736d2 2810
989793d3
LP		 2811 for (;;) {
2812 r = generic_array_get_plus_one(f,
2813 le64toh(d->data.entry_offset),
2814 le64toh(d->data.entry_array_offset),
2815 i,
2816 ret, &ofs);
2817 if (r > 0)
2818 break;
2819 if (r != -EBADMSG)
2820 return r;
2821
2822 log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i);
2823
2824 r = bump_array_index(&i, direction, n);
2825 if (r <= 0)
2826 return r;
2827 }
ded5034e
LP		 2828
2829 /* Ensure our array is properly ordered. */
2830 if (p > 0 && check_properly_ordered(ofs, p, direction)) {
2831 log_debug("%s data entry array not properly ordered at entry %" PRIu64, f->path, i);
2832 return -EBADMSG;
2833 }
2834
2835 if (offset)
2836 *offset = ofs;
2837
2838 return 1;
de190aef 2839}
cec736d2 2840
cbdca852
LP		 2841int journal_file_move_to_entry_by_offset_for_data(
2842 JournalFile *f,
2843 uint64_t data_offset,
2844 uint64_t p,
2845 direction_t direction,
2846 Object **ret, uint64_t *offset) {
2847
2848 int r;
2849 Object *d;
2850
2851 assert(f);
2852
2853 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
2854 if (r < 0)
2855 return r;
2856
2857 return generic_array_bisect_plus_one(f,
2858 le64toh(d->data.entry_offset),
2859 le64toh(d->data.entry_array_offset),
2860 le64toh(d->data.n_entries),
2861 p,
2862 test_object_offset,
2863 direction,
2864 ret, offset, NULL);
2865}
2866
2867int journal_file_move_to_entry_by_monotonic_for_data(
2868 JournalFile *f,
2869 uint64_t data_offset,
2870 sd_id128_t boot_id,
2871 uint64_t monotonic,
2872 direction_t direction,
2873 Object **ret, uint64_t *offset) {
2874
cbdca852
LP		 2875 Object *o, *d;
2876 int r;
2877 uint64_t b, z;
2878
2879 assert(f);
2880
2881 /* First, seek by time */
47838ab3 2882 r = find_data_object_by_boot_id(f, boot_id, &o, &b);
cbdca852
LP		 2883 if (r < 0)
2884 return r;
2885 if (r == 0)
2886 return -ENOENT;
2887
2888 r = generic_array_bisect_plus_one(f,
2889 le64toh(o->data.entry_offset),
2890 le64toh(o->data.entry_array_offset),
2891 le64toh(o->data.n_entries),
2892 monotonic,
2893 test_object_monotonic,
2894 direction,
2895 NULL, &z, NULL);
2896 if (r <= 0)
2897 return r;
2898
2899 /* And now, continue seeking until we find an entry that
2900 * exists in both bisection arrays */
2901
2902 for (;;) {
2903 Object *qo;
2904 uint64_t p, q;
2905
2906 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
2907 if (r < 0)
2908 return r;
2909
2910 r = generic_array_bisect_plus_one(f,
2911 le64toh(d->data.entry_offset),
2912 le64toh(d->data.entry_array_offset),
2913 le64toh(d->data.n_entries),
2914 z,
2915 test_object_offset,
2916 direction,
2917 NULL, &p, NULL);
2918 if (r <= 0)
2919 return r;
2920
2921 r = journal_file_move_to_object(f, OBJECT_DATA, b, &o);
2922 if (r < 0)
2923 return r;
2924
2925 r = generic_array_bisect_plus_one(f,
2926 le64toh(o->data.entry_offset),
2927 le64toh(o->data.entry_array_offset),
2928 le64toh(o->data.n_entries),
2929 p,
2930 test_object_offset,
2931 direction,
2932 &qo, &q, NULL);
2933
2934 if (r <= 0)
2935 return r;
2936
2937 if (p == q) {
2938 if (ret)
2939 *ret = qo;
2940 if (offset)
2941 *offset = q;
2942
2943 return 1;
2944 }
2945
2946 z = q;
2947 }
cbdca852
LP		 2948}
2949
de190aef
LP		 2950int journal_file_move_to_entry_by_seqnum_for_data(
2951 JournalFile *f,
2952 uint64_t data_offset,
2953 uint64_t seqnum,
2954 direction_t direction,
2955 Object **ret, uint64_t *offset) {
cec736d2 2956
de190aef
LP		 2957 Object *d;
2958 int r;
cec736d2 2959
91a31dde
LP		 2960 assert(f);
2961
de190aef 2962 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
91a31dde 2963 if (r < 0)
de190aef 2964 return r;
cec736d2 2965
de190aef
LP		 2966 return generic_array_bisect_plus_one(f,
2967 le64toh(d->data.entry_offset),
2968 le64toh(d->data.entry_array_offset),
2969 le64toh(d->data.n_entries),
2970 seqnum,
2971 test_object_seqnum,
2972 direction,
2973 ret, offset, NULL);
2974}
cec736d2 2975
de190aef
LP		 2976int journal_file_move_to_entry_by_realtime_for_data(
2977 JournalFile *f,
2978 uint64_t data_offset,
2979 uint64_t realtime,
2980 direction_t direction,
2981 Object **ret, uint64_t *offset) {
2982
2983 Object *d;
2984 int r;
2985
91a31dde
LP		 2986 assert(f);
2987
de190aef 2988 r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
91a31dde 2989 if (r < 0)
de190aef
LP		 2990 return r;
2991
2992 return generic_array_bisect_plus_one(f,
2993 le64toh(d->data.entry_offset),
2994 le64toh(d->data.entry_array_offset),
2995 le64toh(d->data.n_entries),
2996 realtime,
2997 test_object_realtime,
2998 direction,
2999 ret, offset, NULL);
cec736d2
LP		 3000}
3001
0284adc6 3002void journal_file_dump(JournalFile *f) {
7560fffc 3003 Object *o;
7560fffc 3004 int r;
0284adc6 3005 uint64_t p;
7560fffc
LP		 3006
3007 assert(f);
c88cc6af 3008 assert(f->header);
7560fffc 3009
0284adc6 3010 journal_file_print_header(f);
7560fffc 3011
0284adc6
LP		 3012 p = le64toh(f->header->header_size);
3013 while (p != 0) {
d05089d8 3014 r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o);
0284adc6
LP		 3015 if (r < 0)
3016 goto fail;
7560fffc 3017
0284adc6 3018 switch (o->object.type) {
d98cc1f2 3019
0284adc6
LP		 3020 case OBJECT_UNUSED:
3021 printf("Type: OBJECT_UNUSED\n");
3022 break;
d98cc1f2 3023
0284adc6
LP		 3024 case OBJECT_DATA:
3025 printf("Type: OBJECT_DATA\n");
3026 break;
7560fffc 3027
3c1668da
LP		 3028 case OBJECT_FIELD:
3029 printf("Type: OBJECT_FIELD\n");
3030 break;
3031
0284adc6 3032 case OBJECT_ENTRY:
507f22bd
ZJS		 3033 printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n",
3034 le64toh(o->entry.seqnum),
3035 le64toh(o->entry.monotonic),
3036 le64toh(o->entry.realtime));
0284adc6 3037 break;
7560fffc 3038
0284adc6
LP		 3039 case OBJECT_FIELD_HASH_TABLE:
3040 printf("Type: OBJECT_FIELD_HASH_TABLE\n");
3041 break;
7560fffc 3042
0284adc6
LP		 3043 case OBJECT_DATA_HASH_TABLE:
3044 printf("Type: OBJECT_DATA_HASH_TABLE\n");
3045 break;
7560fffc 3046
0284adc6
LP		 3047 case OBJECT_ENTRY_ARRAY:
3048 printf("Type: OBJECT_ENTRY_ARRAY\n");
3049 break;
7560fffc 3050
0284adc6 3051 case OBJECT_TAG:
507f22bd
ZJS		 3052 printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n",
3053 le64toh(o->tag.seqnum),
3054 le64toh(o->tag.epoch));
0284adc6 3055 break;
3c1668da
LP		 3056
3057 default:
8facc349 3058 printf("Type: unknown (%i)\n", o->object.type);
3c1668da 3059 break;
0284adc6 3060 }
7560fffc 3061
d89c8fdf
ZJS		 3062 if (o->object.flags & OBJECT_COMPRESSION_MASK)
3063 printf("Flags: %s\n",
3064 object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK));
7560fffc 3065
0284adc6
LP		 3066 if (p == le64toh(f->header->tail_object_offset))
3067 p = 0;
3068 else
3069 p = p + ALIGN64(le64toh(o->object.size));
3070 }
7560fffc 3071
0284adc6
LP		 3072 return;
3073fail:
3074 log_error("File corrupt");
7560fffc
LP		 3075}
3076
718fe4b1
ZJS		 3077static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) {
3078 const char *x;
3079
3080 x = format_timestamp(buf, l, t);
3081 if (x)
3082 return x;
3083 return " --- ";
3084}
3085
0284adc6 3086void journal_file_print_header(JournalFile *f) {
2765b7bb 3087 char a[33], b[33], c[33], d[33];
ed375beb 3088 char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX];
a1a03e30
LP		 3089 struct stat st;
3090 char bytes[FORMAT_BYTES_MAX];
7560fffc
LP		 3091
3092 assert(f);
c88cc6af 3093 assert(f->header);
7560fffc 3094
0284adc6
LP		 3095 printf("File Path: %s\n"
3096 "File ID: %s\n"
3097 "Machine ID: %s\n"
3098 "Boot ID: %s\n"
3099 "Sequential Number ID: %s\n"
3100 "State: %s\n"
3101 "Compatible Flags:%s%s\n"
d89c8fdf 3102 "Incompatible Flags:%s%s%s\n"
507f22bd
ZJS		 3103 "Header size: %"PRIu64"\n"
3104 "Arena size: %"PRIu64"\n"
3105 "Data Hash Table Size: %"PRIu64"\n"
3106 "Field Hash Table Size: %"PRIu64"\n"
0284adc6 3107 "Rotate Suggested: %s\n"
0808b92f
LP		 3108 "Head Sequential Number: %"PRIu64" (%"PRIx64")\n"
3109 "Tail Sequential Number: %"PRIu64" (%"PRIx64")\n"
3110 "Head Realtime Timestamp: %s (%"PRIx64")\n"
3111 "Tail Realtime Timestamp: %s (%"PRIx64")\n"
3112 "Tail Monotonic Timestamp: %s (%"PRIx64")\n"
507f22bd
ZJS		 3113 "Objects: %"PRIu64"\n"
3114 "Entry Objects: %"PRIu64"\n",
0284adc6
LP		 3115 f->path,
3116 sd_id128_to_string(f->header->file_id, a),
3117 sd_id128_to_string(f->header->machine_id, b),
3118 sd_id128_to_string(f->header->boot_id, c),
2765b7bb 3119 sd_id128_to_string(f->header->seqnum_id, d),
3223f44f
LP		 3120 f->header->state == STATE_OFFLINE ? "OFFLINE" :
3121 f->header->state == STATE_ONLINE ? "ONLINE" :
3122 f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN",
8088cbd3 3123 JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "",
d89c8fdf
ZJS		 3124 (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "",
3125 JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "",
3126 JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "",
3127 (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "",
507f22bd
ZJS		 3128 le64toh(f->header->header_size),
3129 le64toh(f->header->arena_size),
3130 le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
3131 le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
fb0951b0 3132 yes_no(journal_file_rotate_suggested(f, 0)),
0808b92f
LP		 3133 le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum),
3134 le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum),
3135 format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime),
3136 format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime),
3137 format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic),
507f22bd
ZJS		 3138 le64toh(f->header->n_objects),
3139 le64toh(f->header->n_entries));
7560fffc 3140
0284adc6 3141 if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
507f22bd 3142 printf("Data Objects: %"PRIu64"\n"
0284adc6 3143 "Data Hash Table Fill: %.1f%%\n",
507f22bd 3144 le64toh(f->header->n_data),
0284adc6 3145 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))));
7560fffc 3146
0284adc6 3147 if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
507f22bd 3148 printf("Field Objects: %"PRIu64"\n"
0284adc6 3149 "Field Hash Table Fill: %.1f%%\n",
507f22bd 3150 le64toh(f->header->n_fields),
0284adc6 3151 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
3223f44f
LP		 3152
3153 if (JOURNAL_HEADER_CONTAINS(f->header, n_tags))
507f22bd
ZJS		 3154 printf("Tag Objects: %"PRIu64"\n",
3155 le64toh(f->header->n_tags));
3223f44f 3156 if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
507f22bd
ZJS		 3157 printf("Entry Array Objects: %"PRIu64"\n",
3158 le64toh(f->header->n_entry_arrays));
a1a03e30
LP		 3159
3160 if (fstat(f->fd, &st) >= 0)
59f448cf 3161 printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (uint64_t) st.st_blocks * 512ULL));
7560fffc
LP		 3162}
3163
fc68c929
LP		 3164static int journal_file_warn_btrfs(JournalFile *f) {
3165 unsigned attrs;
3166 int r;
3167
3168 assert(f);
3169
3170 /* Before we write anything, check if the COW logic is turned
3171 * off on btrfs. Given our write pattern that is quite
3172 * unfriendly to COW file systems this should greatly improve
3173 * performance on COW file systems, such as btrfs, at the
3174 * expense of data integrity features (which shouldn't be too
3175 * bad, given that we do our own checksumming). */
3176
3177 r = btrfs_is_filesystem(f->fd);
3178 if (r < 0)
3179 return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m");
3180 if (!r)
3181 return 0;
3182
3183 r = read_attr_fd(f->fd, &attrs);
3184 if (r < 0)
3185 return log_warning_errno(r, "Failed to read file attributes: %m");
3186
3187 if (attrs & FS_NOCOW_FL) {
3188 log_debug("Detected btrfs file system with copy-on-write disabled, all is good.");
3189 return 0;
3190 }
3191
3192 log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. "
3193 "This is likely to slow down journal access substantially, please consider turning "
3194 "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path);
3195
3196 return 1;
3197}
3198
0284adc6 3199int journal_file_open(
5d1ce257 3200 int fd,
0284adc6
LP		 3201 const char *fname,
3202 int flags,
3203 mode_t mode,
3204 bool compress,
57850536 3205 uint64_t compress_threshold_bytes,
baed47c3 3206 bool seal,
0284adc6
LP		 3207 JournalMetrics *metrics,
3208 MMapCache *mmap_cache,
b58c888f 3209 Set *deferred_closes,
0284adc6
LP		 3210 JournalFile *template,
3211 JournalFile **ret) {
7560fffc 3212
fa6ac760 3213 bool newly_created = false;
0284adc6 3214 JournalFile *f;
fa6ac760 3215 void *h;
0284adc6 3216 int r;
57850536 3217 char bytes[FORMAT_BYTES_MAX];
7560fffc 3218
0559d3a5 3219 assert(ret);
5d1ce257 3220 assert(fd >= 0 || fname);
7560fffc 3221
ec2ce0c5 3222 if (!IN_SET((flags & O_ACCMODE), O_RDONLY, O_RDWR))
0284adc6 3223 return -EINVAL;
7560fffc 3224
6eda13d3
LP		 3225 if (fname && (flags & O_CREAT) && !endswith(fname, ".journal"))
3226 return -EINVAL;
7560fffc 3227
0284adc6
LP		 3228 f = new0(JournalFile, 1);
3229 if (!f)
3230 return -ENOMEM;
7560fffc 3231
5d1ce257 3232 f->fd = fd;
0284adc6 3233 f->mode = mode;
7560fffc 3234
0284adc6
LP		 3235 f->flags = flags;
3236 f->prot = prot_from_flags(flags);
3237 f->writable = (flags & O_ACCMODE) != O_RDONLY;
349cc4a5 3238#if HAVE_LZ4
d89c8fdf 3239 f->compress_lz4 = compress;
349cc4a5 3240#elif HAVE_XZ
d89c8fdf 3241 f->compress_xz = compress;
48b61739 3242#endif
57850536
AG		 3243
3244 if (compress_threshold_bytes == (uint64_t) -1)
3245 f->compress_threshold_bytes = DEFAULT_COMPRESS_THRESHOLD;
3246 else
3247 f->compress_threshold_bytes = MAX(MIN_COMPRESS_THRESHOLD, compress_threshold_bytes);
3248
349cc4a5 3249#if HAVE_GCRYPT
baed47c3 3250 f->seal = seal;
49a32d43 3251#endif
7560fffc 3252
57850536
AG		 3253 log_debug("Journal effective settings seal=%s compress=%s compress_threshold_bytes=%s",
3254 yes_no(f->seal), yes_no(JOURNAL_FILE_COMPRESS(f)),
3255 format_bytes(bytes, sizeof(bytes), f->compress_threshold_bytes));
3256
0284adc6
LP		 3257 if (mmap_cache)
3258 f->mmap = mmap_cache_ref(mmap_cache);
3259 else {
84168d80 3260 f->mmap = mmap_cache_new();
0284adc6
LP		 3261 if (!f->mmap) {
3262 r = -ENOMEM;
3263 goto fail;
3264 }
3265 }
7560fffc 3266
7645c77b 3267 if (fname) {
5d1ce257 3268 f->path = strdup(fname);
7645c77b
ZJS		 3269 if (!f->path) {
3270 r = -ENOMEM;
3271 goto fail;
3272 }
3273 } else {
817b1c5b
LP		 3274 assert(fd >= 0);
3275
7645c77b
ZJS		 3276 /* If we don't know the path, fill in something explanatory and vaguely useful */
3277 if (asprintf(&f->path, "/proc/self/%i", fd) < 0) {
3278 r = -ENOMEM;
3279 goto fail;
3280 }
0284adc6 3281 }
7560fffc 3282
4743015d 3283 f->chain_cache = ordered_hashmap_new(&uint64_hash_ops);
a4bcff5b
LP		 3284 if (!f->chain_cache) {
3285 r = -ENOMEM;
3286 goto fail;
3287 }
3288
0284adc6 3289 if (f->fd < 0) {
817b1c5b
LP		 3290 /* We pass O_NONBLOCK here, so that in case somebody pointed us to some character device node or FIFO
3291 * or so, we likely fail quickly than block for long. For regular files O_NONBLOCK has no effect, hence
3292 * it doesn't hurt in that case. */
3293
3294 f->fd = open(f->path, f->flags|O_CLOEXEC|O_NONBLOCK, f->mode);
5d1ce257
LP		 3295 if (f->fd < 0) {
3296 r = -errno;
3297 goto fail;
3298 }
3299
3300 /* fds we opened here by us should also be closed by us. */
3301 f->close_fd = true;
817b1c5b
LP		 3302
3303 r = fd_nonblock(f->fd, false);
3304 if (r < 0)
3305 goto fail;
7560fffc 3306 }
7560fffc 3307
be7cdd8e
VC		 3308 f->cache_fd = mmap_cache_add_fd(f->mmap, f->fd);
3309 if (!f->cache_fd) {
3310 r = -ENOMEM;
3311 goto fail;
3312 }
3313
2678031a
LP		 3314 r = journal_file_fstat(f);
3315 if (r < 0)
0284adc6 3316 goto fail;
7560fffc 3317
0284adc6 3318 if (f->last_stat.st_size == 0 && f->writable) {
11689d2a 3319
fc68c929 3320 (void) journal_file_warn_btrfs(f);
11689d2a 3321
4c2e1b39
LP		 3322 /* Let's attach the creation time to the journal file, so that the vacuuming code knows the age of this
3323 * file even if the file might end up corrupted one day... Ideally we'd just use the creation time many
3324 * file systems maintain for each file, but the API to query this is very new, hence let's emulate this
3325 * via extended attributes. If extended attributes are not supported we'll just skip this, and rely
3326 * solely on mtime/atime/ctime of the file. */
3327 (void) fd_setcrtime(f->fd, 0);
7560fffc 3328
349cc4a5 3329#if HAVE_GCRYPT
0284adc6 3330 /* Try to load the FSPRG state, and if we can't, then
baed47c3 3331 * just don't do sealing */
49a32d43
LP		 3332 if (f->seal) {
3333 r = journal_file_fss_load(f);
3334 if (r < 0)
3335 f->seal = false;
3336 }
feb12d3e 3337#endif
7560fffc 3338
0284adc6
LP		 3339 r = journal_file_init_header(f, template);
3340 if (r < 0)
3341 goto fail;
7560fffc 3342
2678031a
LP		 3343 r = journal_file_fstat(f);
3344 if (r < 0)
0284adc6 3345 goto fail;
fb0951b0
LP		 3346
3347 newly_created = true;
0284adc6 3348 }
7560fffc 3349
0284adc6 3350 if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) {
cfb571f3 3351 r = -ENODATA;
0284adc6
LP		 3352 goto fail;
3353 }
7560fffc 3354
b42549ad 3355 r = mmap_cache_get(f->mmap, f->cache_fd, f->prot, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h, NULL);
977eaa1e 3356 if (r < 0)
0284adc6 3357 goto fail;
7560fffc 3358
fa6ac760
LP		 3359 f->header = h;
3360
0284adc6 3361 if (!newly_created) {
f9168190 3362 set_clear_with_destructor(deferred_closes, journal_file_close);
b58c888f 3363
0284adc6
LP		 3364 r = journal_file_verify_header(f);
3365 if (r < 0)
3366 goto fail;
3367 }
7560fffc 3368
349cc4a5 3369#if HAVE_GCRYPT
0284adc6 3370 if (!newly_created && f->writable) {
baed47c3 3371 r = journal_file_fss_load(f);
0284adc6
LP		 3372 if (r < 0)
3373 goto fail;
3374 }
feb12d3e 3375#endif
cec736d2
LP		 3376
3377 if (f->writable) {
4a92baf3
LP		 3378 if (metrics) {
3379 journal_default_metrics(metrics, f->fd);
3380 f->metrics = *metrics;
3381 } else if (template)
3382 f->metrics = template->metrics;
3383
cec736d2
LP		 3384 r = journal_file_refresh_header(f);
3385 if (r < 0)
3386 goto fail;
3387 }
3388
349cc4a5 3389#if HAVE_GCRYPT
baed47c3 3390 r = journal_file_hmac_setup(f);
14d10188
LP		 3391 if (r < 0)
3392 goto fail;
feb12d3e 3393#endif
14d10188 3394
cec736d2 3395 if (newly_created) {
de190aef 3396 r = journal_file_setup_field_hash_table(f);
cec736d2
LP		 3397 if (r < 0)
3398 goto fail;
3399
de190aef 3400 r = journal_file_setup_data_hash_table(f);
cec736d2
LP		 3401 if (r < 0)
3402 goto fail;
7560fffc 3403
349cc4a5 3404#if HAVE_GCRYPT
7560fffc
LP		 3405 r = journal_file_append_first_tag(f);
3406 if (r < 0)
3407 goto fail;
feb12d3e 3408#endif
cec736d2
LP		 3409 }
3410
be7cdd8e 3411 if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) {
fa6ac760
LP		 3412 r = -EIO;
3413 goto fail;
3414 }
3415
7a24f3bf 3416 if (template && template->post_change_timer) {
e167d7fd
LP		 3417 r = journal_file_enable_post_change_timer(
3418 f,
3419 sd_event_source_get_event(template->post_change_timer),
3420 template->post_change_timer_period);
7a24f3bf 3421
7a24f3bf
VC		 3422 if (r < 0)
3423 goto fail;
3424 }
3425
f8e2f4d6 3426 /* The file is opened now successfully, thus we take possession of any passed in fd. */
5d1ce257
LP		 3427 f->close_fd = true;
3428
0559d3a5 3429 *ret = f;
cec736d2
LP		 3430 return 0;
3431
3432fail:
be7cdd8e 3433 if (f->cache_fd && mmap_cache_got_sigbus(f->mmap, f->cache_fd))
fa6ac760
LP		 3434 r = -EIO;
3435
69a3a6fd 3436 (void) journal_file_close(f);
cec736d2
LP		 3437
3438 return r;
3439}
0ac38b70 3440
57850536 3441int journal_file_rotate(JournalFile **f, bool compress, uint64_t compress_threshold_bytes, bool seal, Set *deferred_closes) {
57535f47 3442 _cleanup_free_ char *p = NULL;
0ac38b70
LP		 3443 size_t l;
3444 JournalFile *old_file, *new_file = NULL;
3445 int r;
3446
3447 assert(f);
3448 assert(*f);
3449
3450 old_file = *f;
3451
3452 if (!old_file->writable)
3453 return -EINVAL;
3454
5d1ce257 3455 /* Is this a journal file that was passed to us as fd? If so, we synthesized a path name for it, and we refuse
13e785f7 3456 * rotation, since we don't know the actual path, and couldn't rename the file hence. */
5d1ce257
LP		 3457 if (path_startswith(old_file->path, "/proc/self/fd"))
3458 return -EINVAL;
3459
0ac38b70
LP		 3460 if (!endswith(old_file->path, ".journal"))
3461 return -EINVAL;
3462
3463 l = strlen(old_file->path);
57535f47
ZJS		 3464 r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal",
3465 (int) l - 8, old_file->path,
3466 SD_ID128_FORMAT_VAL(old_file->header->seqnum_id),
3467 le64toh((*f)->header->head_entry_seqnum),
3468 le64toh((*f)->header->head_entry_realtime));
3469 if (r < 0)
0ac38b70
LP		 3470 return -ENOMEM;
3471
2678031a
LP		 3472 /* Try to rename the file to the archived version. If the file
3473 * already was deleted, we'll get ENOENT, let's ignore that
3474 * case. */
0ac38b70 3475 r = rename(old_file->path, p);
2678031a 3476 if (r < 0 && errno != ENOENT)
0ac38b70
LP		 3477 return -errno;
3478
1fcefd88
LP		 3479 /* Sync the rename to disk */
3480 (void) fsync_directory_of_file(old_file->fd);
3481
8eb85171
VC		 3482 /* Set as archive so offlining commits w/state=STATE_ARCHIVED.
3483 * Previously we would set old_file->header->state to STATE_ARCHIVED directly here,
3484 * but journal_file_set_offline() short-circuits when state != STATE_ONLINE, which
3485 * would result in the rotated journal never getting fsync() called before closing.
3486 * Now we simply queue the archive state by setting an archive bit, leaving the state
3487 * as STATE_ONLINE so proper offlining occurs. */
3488 old_file->archive = true;
0ac38b70 3489
f27a3864
LP		 3490 /* Currently, btrfs is not very good with out write patterns
3491 * and fragments heavily. Let's defrag our journal files when
3492 * we archive them */
3493 old_file->defrag_on_close = true;
3494
57850536
AG		 3495 r = journal_file_open(-1, old_file->path, old_file->flags, old_file->mode, compress,
3496 compress_threshold_bytes, seal, NULL, old_file->mmap, deferred_closes,
3497 old_file, &new_file);
b58c888f
VC		 3498
3499 if (deferred_closes &&
3500 set_put(deferred_closes, old_file) >= 0)
3501 (void) journal_file_set_offline(old_file, false);
3502 else
3503 (void) journal_file_close(old_file);
0ac38b70
LP		 3504
3505 *f = new_file;
3506 return r;
3507}
3508
9447a7f1
LP		 3509int journal_file_open_reliably(
3510 const char *fname,
3511 int flags,
3512 mode_t mode,
7560fffc 3513 bool compress,
57850536 3514 uint64_t compress_threshold_bytes,
baed47c3 3515 bool seal,
4a92baf3 3516 JournalMetrics *metrics,
27370278 3517 MMapCache *mmap_cache,
b58c888f 3518 Set *deferred_closes,
9447a7f1
LP		 3519 JournalFile *template,
3520 JournalFile **ret) {
3521
3522 int r;
3523 size_t l;
ed375beb 3524 _cleanup_free_ char *p = NULL;
9447a7f1 3525
57850536
AG		 3526 r = journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache,
3527 deferred_closes, template, ret);
288359db 3528 if (!IN_SET(r,
b288cdeb
ZJS		 3529 -EBADMSG, /* Corrupted */
3530 -ENODATA, /* Truncated */
3531 -EHOSTDOWN, /* Other machine */
3532 -EPROTONOSUPPORT, /* Incompatible feature */
3533 -EBUSY, /* Unclean shutdown */
3534 -ESHUTDOWN, /* Already archived */
288359db 3535 -EIO, /* IO error, including SIGBUS on mmap */
ae739cc1
LP		 3536 -EIDRM, /* File has been deleted */
3537 -ETXTBSY)) /* File is from the future */
9447a7f1
LP		 3538 return r;
3539
3540 if ((flags & O_ACCMODE) == O_RDONLY)
3541 return r;
3542
3543 if (!(flags & O_CREAT))
3544 return r;
3545
7560fffc
LP		 3546 if (!endswith(fname, ".journal"))
3547 return r;
3548
5c70eab4
LP		 3549 /* The file is corrupted. Rotate it away and try it again (but only once) */
3550
9447a7f1 3551 l = strlen(fname);
d587eca5 3552 if (asprintf(&p, "%.*s@%016"PRIx64 "-%016"PRIx64 ".journal~",
57535f47 3553 (int) l - 8, fname,
d587eca5 3554 now(CLOCK_REALTIME),
9bf3b535 3555 random_u64()) < 0)
9447a7f1
LP		 3556 return -ENOMEM;
3557
65089b82 3558 if (rename(fname, p) < 0)
9447a7f1
LP		 3559 return -errno;
3560
f27a3864
LP		 3561 /* btrfs doesn't cope well with our write pattern and
3562 * fragments heavily. Let's defrag all files we rotate */
11689d2a 3563
db9a4254 3564 (void) chattr_path(p, 0, FS_NOCOW_FL, NULL);
f27a3864
LP		 3565 (void) btrfs_defrag(p);
3566
65089b82 3567 log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
9447a7f1 3568
57850536
AG		 3569 return journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache,
3570 deferred_closes, template, ret);
9447a7f1
LP		 3571}
3572
5a271b08 3573int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p) {
cf244689
LP		 3574 uint64_t i, n;
3575 uint64_t q, xor_hash = 0;
3576 int r;
3577 EntryItem *items;
3578 dual_timestamp ts;
d180c349 3579 const sd_id128_t *boot_id;
cf244689
LP		 3580
3581 assert(from);
3582 assert(to);
3583 assert(o);
3584 assert(p);
3585
3586 if (!to->writable)
3587 return -EPERM;
3588
3589 ts.monotonic = le64toh(o->entry.monotonic);
3590 ts.realtime = le64toh(o->entry.realtime);
d180c349 3591 boot_id = &o->entry.boot_id;
cf244689 3592
cf244689 3593 n = journal_file_entry_n_items(o);
4faa7004 3594 /* alloca() can't take 0, hence let's allocate at least one */
cf409d15 3595 items = newa(EntryItem, MAX(1u, n));
cf244689
LP		 3596
3597 for (i = 0; i < n; i++) {
4fd052ae
FC		 3598 uint64_t l, h;
3599 le64_t le_hash;
cf244689
LP		 3600 size_t t;
3601 void *data;
3602 Object *u;
3603
3604 q = le64toh(o->entry.items[i].object_offset);
3605 le_hash = o->entry.items[i].hash;
3606
3607 r = journal_file_move_to_object(from, OBJECT_DATA, q, &o);
3608 if (r < 0)
3609 return r;
3610
3611 if (le_hash != o->data.hash)
3612 return -EBADMSG;
3613
3614 l = le64toh(o->object.size) - offsetof(Object, data.payload);
3615 t = (size_t) l;
3616
3617 /* We hit the limit on 32bit machines */
3618 if ((uint64_t) t != l)
3619 return -E2BIG;
3620
d89c8fdf 3621 if (o->object.flags & OBJECT_COMPRESSION_MASK) {
349cc4a5 3622#if HAVE_XZ || HAVE_LZ4
a7f7d1bd 3623 size_t rsize = 0;
cf244689 3624
d89c8fdf
ZJS		 3625 r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
3626 o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0);
3627 if (r < 0)
3628 return r;
cf244689
LP		 3629
3630 data = from->compress_buffer;
3631 l = rsize;
3b1a55e1
ZJS		 3632#else
3633 return -EPROTONOSUPPORT;
3634#endif
cf244689
LP		 3635 } else
3636 data = o->data.payload;
3637
3638 r = journal_file_append_data(to, data, l, &u, &h);
3639 if (r < 0)
3640 return r;
3641
3642 xor_hash ^= le64toh(u->data.hash);
3643 items[i].object_offset = htole64(h);
3644 items[i].hash = u->data.hash;
3645
3646 r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o);
3647 if (r < 0)
3648 return r;
3649 }
3650
d180c349
ZJS		 3651 r = journal_file_append_entry_internal(to, &ts, boot_id, xor_hash, items, n,
3652 NULL, NULL, NULL);
fa6ac760 3653
be7cdd8e 3654 if (mmap_cache_got_sigbus(to->mmap, to->cache_fd))
fa6ac760
LP		 3655 return -EIO;
3656
3657 return r;
cf244689 3658}
babfc091 3659
8580d1f7
LP		 3660void journal_reset_metrics(JournalMetrics *m) {
3661 assert(m);
3662
3663 /* Set everything to "pick automatic values". */
3664
3665 *m = (JournalMetrics) {
3666 .min_use = (uint64_t) -1,
3667 .max_use = (uint64_t) -1,
3668 .min_size = (uint64_t) -1,
3669 .max_size = (uint64_t) -1,
3670 .keep_free = (uint64_t) -1,
3671 .n_max_files = (uint64_t) -1,
3672 };
3673}
3674
babfc091 3675void journal_default_metrics(JournalMetrics *m, int fd) {
8580d1f7 3676 char a[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX], e[FORMAT_BYTES_MAX];
babfc091 3677 struct statvfs ss;
8580d1f7 3678 uint64_t fs_size;
babfc091
LP		 3679
3680 assert(m);
3681 assert(fd >= 0);
3682
3683 if (fstatvfs(fd, &ss) >= 0)
3684 fs_size = ss.f_frsize * ss.f_blocks;
8580d1f7 3685 else {
8fc58f1a 3686 log_debug_errno(errno, "Failed to determine disk size: %m");
8580d1f7
LP		 3687 fs_size = 0;
3688 }
babfc091
LP		 3689
3690 if (m->max_use == (uint64_t) -1) {
3691
3692 if (fs_size > 0) {
3693 m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */
3694
3695 if (m->max_use > DEFAULT_MAX_USE_UPPER)
3696 m->max_use = DEFAULT_MAX_USE_UPPER;
3697
3698 if (m->max_use < DEFAULT_MAX_USE_LOWER)
3699 m->max_use = DEFAULT_MAX_USE_LOWER;
3700 } else
3701 m->max_use = DEFAULT_MAX_USE_LOWER;
3702 } else {
3703 m->max_use = PAGE_ALIGN(m->max_use);
3704
8580d1f7 3705 if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2)
babfc091
LP		 3706 m->max_use = JOURNAL_FILE_SIZE_MIN*2;
3707 }
3708
8580d1f7
LP		 3709 if (m->min_use == (uint64_t) -1)
3710 m->min_use = DEFAULT_MIN_USE;
3711
3712 if (m->min_use > m->max_use)
3713 m->min_use = m->max_use;
3714
babfc091
LP		 3715 if (m->max_size == (uint64_t) -1) {
3716 m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */
3717
3718 if (m->max_size > DEFAULT_MAX_SIZE_UPPER)
3719 m->max_size = DEFAULT_MAX_SIZE_UPPER;
3720 } else
3721 m->max_size = PAGE_ALIGN(m->max_size);
3722
8580d1f7
LP		 3723 if (m->max_size != 0) {
3724 if (m->max_size < JOURNAL_FILE_SIZE_MIN)
3725 m->max_size = JOURNAL_FILE_SIZE_MIN;
babfc091 3726
8580d1f7
LP		 3727 if (m->max_use != 0 && m->max_size*2 > m->max_use)
3728 m->max_use = m->max_size*2;
3729 }
babfc091
LP		 3730
3731 if (m->min_size == (uint64_t) -1)
3732 m->min_size = JOURNAL_FILE_SIZE_MIN;
3733 else {
3734 m->min_size = PAGE_ALIGN(m->min_size);
3735
3736 if (m->min_size < JOURNAL_FILE_SIZE_MIN)
3737 m->min_size = JOURNAL_FILE_SIZE_MIN;
3738
8580d1f7 3739 if (m->max_size != 0 && m->min_size > m->max_size)
babfc091
LP		 3740 m->max_size = m->min_size;
3741 }
3742
3743 if (m->keep_free == (uint64_t) -1) {
3744
3745 if (fs_size > 0) {
8621b110 3746 m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */
babfc091
LP		 3747
3748 if (m->keep_free > DEFAULT_KEEP_FREE_UPPER)
3749 m->keep_free = DEFAULT_KEEP_FREE_UPPER;
3750
3751 } else
3752 m->keep_free = DEFAULT_KEEP_FREE;
3753 }
3754
8580d1f7
LP		 3755 if (m->n_max_files == (uint64_t) -1)
3756 m->n_max_files = DEFAULT_N_MAX_FILES;
3757
3758 log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64,
3759 format_bytes(a, sizeof(a), m->min_use),
3760 format_bytes(b, sizeof(b), m->max_use),
3761 format_bytes(c, sizeof(c), m->max_size),
3762 format_bytes(d, sizeof(d), m->min_size),
3763 format_bytes(e, sizeof(e), m->keep_free),
3764 m->n_max_files);
babfc091 3765}
08984293
LP		 3766
3767int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) {
08984293 3768 assert(f);
c88cc6af 3769 assert(f->header);
08984293
LP		 3770 assert(from || to);
3771
3772 if (from) {
162566a4
LP		 3773 if (f->header->head_entry_realtime == 0)
3774 return -ENOENT;
08984293 3775
162566a4 3776 *from = le64toh(f->header->head_entry_realtime);
08984293
LP		 3777 }
3778
3779 if (to) {
162566a4
LP		 3780 if (f->header->tail_entry_realtime == 0)
3781 return -ENOENT;
08984293 3782
162566a4 3783 *to = le64toh(f->header->tail_entry_realtime);
08984293
LP		 3784 }
3785
3786 return 1;
3787}
3788
3789int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) {
08984293
LP		 3790 Object *o;
3791 uint64_t p;
3792 int r;
3793
3794 assert(f);
3795 assert(from || to);
3796
47838ab3 3797 r = find_data_object_by_boot_id(f, boot_id, &o, &p);
08984293
LP		 3798 if (r <= 0)
3799 return r;
3800
3801 if (le64toh(o->data.n_entries) <= 0)
3802 return 0;
3803
3804 if (from) {
3805 r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o);
3806 if (r < 0)
3807 return r;
3808
3809 *from = le64toh(o->entry.monotonic);
3810 }
3811
3812 if (to) {
3813 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
3814 if (r < 0)
3815 return r;
3816
3817 r = generic_array_get_plus_one(f,
3818 le64toh(o->data.entry_offset),
3819 le64toh(o->data.entry_array_offset),
3820 le64toh(o->data.n_entries)-1,
3821 &o, NULL);
3822 if (r <= 0)
3823 return r;
3824
3825 *to = le64toh(o->entry.monotonic);
3826 }
3827
3828 return 1;
3829}
dca6219e 3830
fb0951b0 3831bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) {
dca6219e 3832 assert(f);
c88cc6af 3833 assert(f->header);
dca6219e
LP		 3834
3835 /* If we gained new header fields we gained new features,
3836 * hence suggest a rotation */
361f9cbc
LP		 3837 if (le64toh(f->header->header_size) < sizeof(Header)) {
3838 log_debug("%s uses an outdated header, suggesting rotation.", f->path);
dca6219e 3839 return true;
361f9cbc 3840 }
dca6219e
LP		 3841
3842 /* Let's check if the hash tables grew over a certain fill
3843 * level (75%, borrowing this value from Java's hash table
3844 * implementation), and if so suggest a rotation. To calculate
3845 * the fill level we need the n_data field, which only exists
3846 * in newer versions. */
3847
3848 if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
361f9cbc 3849 if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) {
507f22bd 3850 log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.",
361f9cbc
LP		 3851 f->path,
3852 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))),
507f22bd
ZJS		 3853 le64toh(f->header->n_data),
3854 le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
3855 (unsigned long long) f->last_stat.st_size,
3856 f->last_stat.st_size / le64toh(f->header->n_data));
dca6219e 3857 return true;
361f9cbc 3858 }
dca6219e
LP		 3859
3860 if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
361f9cbc 3861 if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) {
507f22bd 3862 log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.",
361f9cbc
LP		 3863 f->path,
3864 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))),
507f22bd
ZJS		 3865 le64toh(f->header->n_fields),
3866 le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
dca6219e 3867 return true;
361f9cbc 3868 }
dca6219e 3869
0598fd4a
LP		 3870 /* Are the data objects properly indexed by field objects? */
3871 if (JOURNAL_HEADER_CONTAINS(f->header, n_data) &&
3872 JOURNAL_HEADER_CONTAINS(f->header, n_fields) &&
3873 le64toh(f->header->n_data) > 0 &&
3874 le64toh(f->header->n_fields) == 0)
3875 return true;
3876
fb0951b0
LP		 3877 if (max_file_usec > 0) {
3878 usec_t t, h;
3879
3880 h = le64toh(f->header->head_entry_realtime);
3881 t = now(CLOCK_REALTIME);
3882
3883 if (h > 0 && t > h + max_file_usec)
3884 return true;
3885 }
3886
dca6219e
LP		 3887 return false;
3888}
Unnamed repository; edit this file 'description' to name the repository.