]>
Commit | Line | Data |
---|---|---|
cec736d2 LP |
1 | /*** |
2 | This file is part of systemd. | |
3 | ||
4 | Copyright 2011 Lennart Poettering | |
5 | ||
6 | systemd is free software; you can redistribute it and/or modify it | |
5430f7f2 LP |
7 | under the terms of the GNU Lesser General Public License as published by |
8 | the Free Software Foundation; either version 2.1 of the License, or | |
cec736d2 LP |
9 | (at your option) any later version. |
10 | ||
11 | systemd is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
5430f7f2 | 14 | Lesser General Public License for more details. |
cec736d2 | 15 | |
5430f7f2 | 16 | You should have received a copy of the GNU Lesser General Public License |
cec736d2 LP |
17 | along with systemd; If not, see <http://www.gnu.org/licenses/>. |
18 | ***/ | |
19 | ||
cec736d2 | 20 | #include <errno.h> |
cec736d2 | 21 | #include <fcntl.h> |
11689d2a | 22 | #include <linux/fs.h> |
ac2e41f5 | 23 | #include <pthread.h> |
07630cea LP |
24 | #include <stddef.h> |
25 | #include <sys/mman.h> | |
26 | #include <sys/statvfs.h> | |
27 | #include <sys/uio.h> | |
28 | #include <unistd.h> | |
fb0951b0 | 29 | |
b5efdb8a | 30 | #include "alloc-util.h" |
f27a3864 | 31 | #include "btrfs-util.h" |
c8b3094d | 32 | #include "chattr-util.h" |
07630cea | 33 | #include "compress.h" |
3ffd4af2 | 34 | #include "fd-util.h" |
0284adc6 | 35 | #include "journal-authenticate.h" |
cec736d2 LP |
36 | #include "journal-def.h" |
37 | #include "journal-file.h" | |
38 | #include "lookup3.h" | |
6bedfcbb | 39 | #include "parse-util.h" |
5d1ce257 | 40 | #include "path-util.h" |
3df3e884 | 41 | #include "random-util.h" |
7a24f3bf | 42 | #include "sd-event.h" |
b58c888f | 43 | #include "set.h" |
07630cea | 44 | #include "string-util.h" |
89a5a90c | 45 | #include "xattr-util.h" |
cec736d2 | 46 | |
4a92baf3 LP |
47 | #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) |
48 | #define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) | |
cec736d2 | 49 | |
be19b7df | 50 | #define COMPRESSION_SIZE_THRESHOLD (512ULL) |
807e17f0 | 51 | |
babfc091 | 52 | /* This is the minimum journal file size */ |
16098e93 | 53 | #define JOURNAL_FILE_SIZE_MIN (512ULL*1024ULL) /* 512 KiB */ |
babfc091 LP |
54 | |
55 | /* These are the lower and upper bounds if we deduce the max_use value | |
56 | * from the file system size */ | |
57 | #define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */ | |
58 | #define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ | |
59 | ||
8580d1f7 LP |
60 | /* This is the default minimal use limit, how much we'll use even if keep_free suggests otherwise. */ |
61 | #define DEFAULT_MIN_USE (1ULL*1024ULL*1024ULL) /* 1 MiB */ | |
62 | ||
babfc091 | 63 | /* This is the upper bound if we deduce max_size from max_use */ |
71100051 | 64 | #define DEFAULT_MAX_SIZE_UPPER (128ULL*1024ULL*1024ULL) /* 128 MiB */ |
babfc091 LP |
65 | |
66 | /* This is the upper bound if we deduce the keep_free value from the | |
67 | * file system size */ | |
68 | #define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ | |
69 | ||
70 | /* This is the keep_free value when we can't determine the system | |
71 | * size */ | |
72 | #define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ | |
73 | ||
8580d1f7 LP |
74 | /* This is the default maximum number of journal files to keep around. */ |
75 | #define DEFAULT_N_MAX_FILES (100) | |
76 | ||
dca6219e LP |
77 | /* n_data was the first entry we added after the initial file format design */ |
78 | #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) | |
cec736d2 | 79 | |
a4bcff5b LP |
80 | /* How many entries to keep in the entry array chain cache at max */ |
81 | #define CHAIN_CACHE_MAX 20 | |
82 | ||
a676e665 LP |
83 | /* How much to increase the journal file size at once each time we allocate something new. */ |
84 | #define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL) /* 8MB */ | |
85 | ||
2678031a LP |
86 | /* Reread fstat() of the file for detecting deletions at least this often */ |
87 | #define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC) | |
88 | ||
fa6ac760 LP |
89 | /* The mmap context to use for the header we pick as one above the last defined typed */ |
90 | #define CONTEXT_HEADER _OBJECT_TYPE_MAX | |
91 | ||
ac2e41f5 VC |
92 | /* This may be called from a separate thread to prevent blocking the caller for the duration of fsync(). |
93 | * As a result we use atomic operations on f->offline_state for inter-thread communications with | |
94 | * journal_file_set_offline() and journal_file_set_online(). */ | |
95 | static void journal_file_set_offline_internal(JournalFile *f) { | |
26687bf8 | 96 | assert(f); |
ac2e41f5 VC |
97 | assert(f->fd >= 0); |
98 | assert(f->header); | |
99 | ||
100 | for (;;) { | |
101 | switch (f->offline_state) { | |
102 | case OFFLINE_CANCEL: | |
103 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_DONE)) | |
104 | continue; | |
105 | return; | |
106 | ||
107 | case OFFLINE_AGAIN_FROM_SYNCING: | |
108 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_SYNCING)) | |
109 | continue; | |
110 | break; | |
111 | ||
112 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
113 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_SYNCING)) | |
114 | continue; | |
115 | break; | |
116 | ||
117 | case OFFLINE_SYNCING: | |
118 | (void) fsync(f->fd); | |
26687bf8 | 119 | |
ac2e41f5 VC |
120 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_OFFLINING)) |
121 | continue; | |
26687bf8 | 122 | |
8eb85171 | 123 | f->header->state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE; |
ac2e41f5 VC |
124 | (void) fsync(f->fd); |
125 | break; | |
126 | ||
127 | case OFFLINE_OFFLINING: | |
128 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_DONE)) | |
129 | continue; | |
130 | /* fall through */ | |
131 | ||
132 | case OFFLINE_DONE: | |
133 | return; | |
134 | ||
135 | case OFFLINE_JOINED: | |
136 | log_debug("OFFLINE_JOINED unexpected offline state for journal_file_set_offline_internal()"); | |
137 | return; | |
138 | } | |
139 | } | |
140 | } | |
141 | ||
142 | static void * journal_file_set_offline_thread(void *arg) { | |
143 | JournalFile *f = arg; | |
144 | ||
145 | journal_file_set_offline_internal(f); | |
146 | ||
147 | return NULL; | |
148 | } | |
149 | ||
150 | static int journal_file_set_offline_thread_join(JournalFile *f) { | |
151 | int r; | |
152 | ||
153 | assert(f); | |
154 | ||
155 | if (f->offline_state == OFFLINE_JOINED) | |
156 | return 0; | |
157 | ||
158 | r = pthread_join(f->offline_thread, NULL); | |
159 | if (r) | |
160 | return -r; | |
161 | ||
162 | f->offline_state = OFFLINE_JOINED; | |
26687bf8 | 163 | |
fa6ac760 LP |
164 | if (mmap_cache_got_sigbus(f->mmap, f->fd)) |
165 | return -EIO; | |
166 | ||
ac2e41f5 VC |
167 | return 0; |
168 | } | |
26687bf8 | 169 | |
ac2e41f5 VC |
170 | /* Trigger a restart if the offline thread is mid-flight in a restartable state. */ |
171 | static bool journal_file_set_offline_try_restart(JournalFile *f) { | |
172 | for (;;) { | |
173 | switch (f->offline_state) { | |
174 | case OFFLINE_AGAIN_FROM_SYNCING: | |
175 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
176 | return true; | |
177 | ||
178 | case OFFLINE_CANCEL: | |
179 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_AGAIN_FROM_SYNCING)) | |
180 | continue; | |
181 | return true; | |
182 | ||
183 | case OFFLINE_SYNCING: | |
184 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_AGAIN_FROM_SYNCING)) | |
185 | continue; | |
186 | return true; | |
187 | ||
188 | case OFFLINE_OFFLINING: | |
189 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_AGAIN_FROM_OFFLINING)) | |
190 | continue; | |
191 | return true; | |
26687bf8 OS |
192 | |
193 | default: | |
ac2e41f5 VC |
194 | return false; |
195 | } | |
26687bf8 OS |
196 | } |
197 | } | |
198 | ||
ac2e41f5 VC |
199 | /* Sets a journal offline. |
200 | * | |
201 | * If wait is false then an offline is dispatched in a separate thread for a | |
202 | * subsequent journal_file_set_offline() or journal_file_set_online() of the | |
203 | * same journal to synchronize with. | |
204 | * | |
205 | * If wait is true, then either an existing offline thread will be restarted | |
206 | * and joined, or if none exists the offline is simply performed in this | |
207 | * context without involving another thread. | |
208 | */ | |
209 | int journal_file_set_offline(JournalFile *f, bool wait) { | |
210 | bool restarted; | |
211 | int r; | |
212 | ||
26687bf8 OS |
213 | assert(f); |
214 | ||
215 | if (!f->writable) | |
216 | return -EPERM; | |
217 | ||
218 | if (!(f->fd >= 0 && f->header)) | |
219 | return -EINVAL; | |
220 | ||
b8f99e27 VC |
221 | /* An offlining journal is implicitly online and may modify f->header->state, |
222 | * we must also join any potentially lingering offline thread when not online. */ | |
223 | if (!journal_file_is_offlining(f) && f->header->state != STATE_ONLINE) | |
224 | return journal_file_set_offline_thread_join(f); | |
26687bf8 | 225 | |
ac2e41f5 VC |
226 | /* Restart an in-flight offline thread and wait if needed, or join a lingering done one. */ |
227 | restarted = journal_file_set_offline_try_restart(f); | |
228 | if ((restarted && wait) || !restarted) { | |
229 | r = journal_file_set_offline_thread_join(f); | |
230 | if (r < 0) | |
231 | return r; | |
232 | } | |
26687bf8 | 233 | |
ac2e41f5 VC |
234 | if (restarted) |
235 | return 0; | |
236 | ||
237 | /* Initiate a new offline. */ | |
238 | f->offline_state = OFFLINE_SYNCING; | |
fa6ac760 | 239 | |
ac2e41f5 VC |
240 | if (wait) /* Without using a thread if waiting. */ |
241 | journal_file_set_offline_internal(f); | |
242 | else { | |
243 | r = pthread_create(&f->offline_thread, NULL, journal_file_set_offline_thread, f); | |
ec9ffa2c VC |
244 | if (r > 0) { |
245 | f->offline_state = OFFLINE_JOINED; | |
ac2e41f5 | 246 | return -r; |
ec9ffa2c | 247 | } |
ac2e41f5 VC |
248 | } |
249 | ||
250 | return 0; | |
251 | } | |
252 | ||
253 | static int journal_file_set_online(JournalFile *f) { | |
254 | bool joined = false; | |
255 | ||
256 | assert(f); | |
257 | ||
258 | if (!f->writable) | |
259 | return -EPERM; | |
260 | ||
261 | if (!(f->fd >= 0 && f->header)) | |
262 | return -EINVAL; | |
263 | ||
264 | while (!joined) { | |
265 | switch (f->offline_state) { | |
266 | case OFFLINE_JOINED: | |
267 | /* No offline thread, no need to wait. */ | |
268 | joined = true; | |
269 | break; | |
270 | ||
271 | case OFFLINE_SYNCING: | |
272 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_CANCEL)) | |
273 | continue; | |
274 | /* Canceled syncing prior to offlining, no need to wait. */ | |
275 | break; | |
276 | ||
277 | case OFFLINE_AGAIN_FROM_SYNCING: | |
278 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_CANCEL)) | |
279 | continue; | |
280 | /* Canceled restart from syncing, no need to wait. */ | |
281 | break; | |
282 | ||
283 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
284 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_CANCEL)) | |
285 | continue; | |
286 | /* Canceled restart from offlining, must wait for offlining to complete however. */ | |
287 | ||
288 | /* fall through to wait */ | |
289 | default: { | |
290 | int r; | |
291 | ||
292 | r = journal_file_set_offline_thread_join(f); | |
293 | if (r < 0) | |
294 | return r; | |
295 | ||
296 | joined = true; | |
297 | break; | |
298 | } | |
299 | } | |
300 | } | |
26687bf8 | 301 | |
fa6ac760 LP |
302 | if (mmap_cache_got_sigbus(f->mmap, f->fd)) |
303 | return -EIO; | |
304 | ||
ac2e41f5 VC |
305 | switch (f->header->state) { |
306 | case STATE_ONLINE: | |
307 | return 0; | |
26687bf8 | 308 | |
ac2e41f5 VC |
309 | case STATE_OFFLINE: |
310 | f->header->state = STATE_ONLINE; | |
311 | (void) fsync(f->fd); | |
312 | return 0; | |
313 | ||
314 | default: | |
315 | return -EINVAL; | |
316 | } | |
26687bf8 OS |
317 | } |
318 | ||
b58c888f VC |
319 | bool journal_file_is_offlining(JournalFile *f) { |
320 | assert(f); | |
321 | ||
322 | __sync_synchronize(); | |
323 | ||
324 | if (f->offline_state == OFFLINE_DONE || | |
325 | f->offline_state == OFFLINE_JOINED) | |
326 | return false; | |
327 | ||
328 | return true; | |
329 | } | |
330 | ||
804ae586 | 331 | JournalFile* journal_file_close(JournalFile *f) { |
de190aef | 332 | assert(f); |
cec736d2 | 333 | |
feb12d3e | 334 | #ifdef HAVE_GCRYPT |
b0af6f41 | 335 | /* Write the final tag */ |
43cd8794 FB |
336 | if (f->seal && f->writable) { |
337 | int r; | |
338 | ||
339 | r = journal_file_append_tag(f); | |
340 | if (r < 0) | |
341 | log_error_errno(r, "Failed to append tag when closing journal: %m"); | |
342 | } | |
feb12d3e | 343 | #endif |
b0af6f41 | 344 | |
7a24f3bf VC |
345 | if (f->post_change_timer) { |
346 | int enabled; | |
347 | ||
348 | if (sd_event_source_get_enabled(f->post_change_timer, &enabled) >= 0) | |
349 | if (enabled == SD_EVENT_ONESHOT) | |
350 | journal_file_post_change(f); | |
351 | ||
e167d7fd | 352 | (void) sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_OFF); |
7a24f3bf VC |
353 | sd_event_source_unref(f->post_change_timer); |
354 | } | |
355 | ||
ac2e41f5 | 356 | journal_file_set_offline(f, true); |
cec736d2 | 357 | |
fa6ac760 LP |
358 | if (f->mmap && f->fd >= 0) |
359 | mmap_cache_close_fd(f->mmap, f->fd); | |
cec736d2 | 360 | |
11689d2a LP |
361 | if (f->fd >= 0 && f->defrag_on_close) { |
362 | ||
363 | /* Be friendly to btrfs: turn COW back on again now, | |
364 | * and defragment the file. We won't write to the file | |
365 | * ever again, hence remove all fragmentation, and | |
366 | * reenable all the good bits COW usually provides | |
367 | * (such as data checksumming). */ | |
368 | ||
1ed8f8c1 | 369 | (void) chattr_fd(f->fd, 0, FS_NOCOW_FL); |
11689d2a LP |
370 | (void) btrfs_defrag_fd(f->fd); |
371 | } | |
f27a3864 | 372 | |
5d1ce257 LP |
373 | if (f->close_fd) |
374 | safe_close(f->fd); | |
cec736d2 | 375 | free(f->path); |
807e17f0 | 376 | |
f649045c | 377 | mmap_cache_unref(f->mmap); |
16e9f408 | 378 | |
4743015d | 379 | ordered_hashmap_free_free(f->chain_cache); |
a4bcff5b | 380 | |
d89c8fdf | 381 | #if defined(HAVE_XZ) || defined(HAVE_LZ4) |
807e17f0 LP |
382 | free(f->compress_buffer); |
383 | #endif | |
384 | ||
7560fffc | 385 | #ifdef HAVE_GCRYPT |
baed47c3 LP |
386 | if (f->fss_file) |
387 | munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size)); | |
dc4ebc07 | 388 | else |
b7c9ae91 LP |
389 | free(f->fsprg_state); |
390 | ||
391 | free(f->fsprg_seed); | |
7560fffc LP |
392 | |
393 | if (f->hmac) | |
394 | gcry_md_close(f->hmac); | |
395 | #endif | |
396 | ||
6b430fdb | 397 | return mfree(f); |
cec736d2 LP |
398 | } |
399 | ||
b58c888f VC |
400 | void journal_file_close_set(Set *s) { |
401 | JournalFile *f; | |
402 | ||
403 | assert(s); | |
404 | ||
405 | while ((f = set_steal_first(s))) | |
406 | (void) journal_file_close(f); | |
407 | } | |
408 | ||
0ac38b70 | 409 | static int journal_file_init_header(JournalFile *f, JournalFile *template) { |
d89c8fdf | 410 | Header h = {}; |
cec736d2 LP |
411 | ssize_t k; |
412 | int r; | |
413 | ||
414 | assert(f); | |
415 | ||
7560fffc | 416 | memcpy(h.signature, HEADER_SIGNATURE, 8); |
23b0b2b2 | 417 | h.header_size = htole64(ALIGN64(sizeof(h))); |
cec736d2 | 418 | |
d89c8fdf ZJS |
419 | h.incompatible_flags |= htole32( |
420 | f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ | | |
421 | f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4); | |
7560fffc | 422 | |
d89c8fdf ZJS |
423 | h.compatible_flags = htole32( |
424 | f->seal * HEADER_COMPATIBLE_SEALED); | |
7560fffc | 425 | |
cec736d2 LP |
426 | r = sd_id128_randomize(&h.file_id); |
427 | if (r < 0) | |
428 | return r; | |
429 | ||
0ac38b70 LP |
430 | if (template) { |
431 | h.seqnum_id = template->header->seqnum_id; | |
beec0085 | 432 | h.tail_entry_seqnum = template->header->tail_entry_seqnum; |
0ac38b70 LP |
433 | } else |
434 | h.seqnum_id = h.file_id; | |
cec736d2 LP |
435 | |
436 | k = pwrite(f->fd, &h, sizeof(h), 0); | |
437 | if (k < 0) | |
438 | return -errno; | |
439 | ||
440 | if (k != sizeof(h)) | |
441 | return -EIO; | |
442 | ||
443 | return 0; | |
444 | } | |
445 | ||
a0fe2a2d LP |
446 | static int fsync_directory_of_file(int fd) { |
447 | _cleanup_free_ char *path = NULL, *dn = NULL; | |
448 | _cleanup_close_ int dfd = -1; | |
449 | struct stat st; | |
450 | int r; | |
451 | ||
452 | if (fstat(fd, &st) < 0) | |
453 | return -errno; | |
454 | ||
455 | if (!S_ISREG(st.st_mode)) | |
456 | return -EBADFD; | |
457 | ||
458 | r = fd_get_path(fd, &path); | |
459 | if (r < 0) | |
460 | return r; | |
461 | ||
462 | if (!path_is_absolute(path)) | |
463 | return -EINVAL; | |
464 | ||
465 | dn = dirname_malloc(path); | |
466 | if (!dn) | |
467 | return -ENOMEM; | |
468 | ||
469 | dfd = open(dn, O_RDONLY|O_CLOEXEC|O_DIRECTORY); | |
470 | if (dfd < 0) | |
471 | return -errno; | |
472 | ||
473 | if (fsync(dfd) < 0) | |
474 | return -errno; | |
475 | ||
476 | return 0; | |
477 | } | |
478 | ||
cec736d2 | 479 | static int journal_file_refresh_header(JournalFile *f) { |
de190aef | 480 | sd_id128_t boot_id; |
fa6ac760 | 481 | int r; |
cec736d2 LP |
482 | |
483 | assert(f); | |
c88cc6af | 484 | assert(f->header); |
cec736d2 LP |
485 | |
486 | r = sd_id128_get_machine(&f->header->machine_id); | |
487 | if (r < 0) | |
488 | return r; | |
489 | ||
de190aef | 490 | r = sd_id128_get_boot(&boot_id); |
cec736d2 LP |
491 | if (r < 0) |
492 | return r; | |
493 | ||
de190aef LP |
494 | if (sd_id128_equal(boot_id, f->header->boot_id)) |
495 | f->tail_entry_monotonic_valid = true; | |
496 | ||
497 | f->header->boot_id = boot_id; | |
498 | ||
fa6ac760 | 499 | r = journal_file_set_online(f); |
b788cc23 | 500 | |
7560fffc | 501 | /* Sync the online state to disk */ |
fb426037 | 502 | (void) fsync(f->fd); |
b788cc23 | 503 | |
a0fe2a2d LP |
504 | /* We likely just created a new file, also sync the directory this file is located in. */ |
505 | (void) fsync_directory_of_file(f->fd); | |
506 | ||
fa6ac760 | 507 | return r; |
cec736d2 LP |
508 | } |
509 | ||
510 | static int journal_file_verify_header(JournalFile *f) { | |
d89c8fdf ZJS |
511 | uint32_t flags; |
512 | ||
cec736d2 | 513 | assert(f); |
c88cc6af | 514 | assert(f->header); |
cec736d2 | 515 | |
7560fffc | 516 | if (memcmp(f->header->signature, HEADER_SIGNATURE, 8)) |
cec736d2 LP |
517 | return -EBADMSG; |
518 | ||
7560fffc LP |
519 | /* In both read and write mode we refuse to open files with |
520 | * incompatible flags we don't know */ | |
d89c8fdf ZJS |
521 | flags = le32toh(f->header->incompatible_flags); |
522 | if (flags & ~HEADER_INCOMPATIBLE_SUPPORTED) { | |
523 | if (flags & ~HEADER_INCOMPATIBLE_ANY) | |
524 | log_debug("Journal file %s has unknown incompatible flags %"PRIx32, | |
525 | f->path, flags & ~HEADER_INCOMPATIBLE_ANY); | |
526 | flags = (flags & HEADER_INCOMPATIBLE_ANY) & ~HEADER_INCOMPATIBLE_SUPPORTED; | |
527 | if (flags) | |
528 | log_debug("Journal file %s uses incompatible flags %"PRIx32 | |
529 | " disabled at compilation time.", f->path, flags); | |
cec736d2 | 530 | return -EPROTONOSUPPORT; |
d89c8fdf | 531 | } |
cec736d2 | 532 | |
7560fffc LP |
533 | /* When open for writing we refuse to open files with |
534 | * compatible flags, too */ | |
d89c8fdf ZJS |
535 | flags = le32toh(f->header->compatible_flags); |
536 | if (f->writable && (flags & ~HEADER_COMPATIBLE_SUPPORTED)) { | |
537 | if (flags & ~HEADER_COMPATIBLE_ANY) | |
538 | log_debug("Journal file %s has unknown compatible flags %"PRIx32, | |
539 | f->path, flags & ~HEADER_COMPATIBLE_ANY); | |
540 | flags = (flags & HEADER_COMPATIBLE_ANY) & ~HEADER_COMPATIBLE_SUPPORTED; | |
541 | if (flags) | |
542 | log_debug("Journal file %s uses compatible flags %"PRIx32 | |
543 | " disabled at compilation time.", f->path, flags); | |
544 | return -EPROTONOSUPPORT; | |
7560fffc LP |
545 | } |
546 | ||
db11ac1a LP |
547 | if (f->header->state >= _STATE_MAX) |
548 | return -EBADMSG; | |
549 | ||
dca6219e LP |
550 | /* The first addition was n_data, so check that we are at least this large */ |
551 | if (le64toh(f->header->header_size) < HEADER_SIZE_MIN) | |
23b0b2b2 LP |
552 | return -EBADMSG; |
553 | ||
8088cbd3 | 554 | if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
beec0085 LP |
555 | return -EBADMSG; |
556 | ||
db11ac1a LP |
557 | if ((le64toh(f->header->header_size) + le64toh(f->header->arena_size)) > (uint64_t) f->last_stat.st_size) |
558 | return -ENODATA; | |
559 | ||
560 | if (le64toh(f->header->tail_object_offset) > (le64toh(f->header->header_size) + le64toh(f->header->arena_size))) | |
561 | return -ENODATA; | |
562 | ||
7762e02b LP |
563 | if (!VALID64(le64toh(f->header->data_hash_table_offset)) || |
564 | !VALID64(le64toh(f->header->field_hash_table_offset)) || | |
565 | !VALID64(le64toh(f->header->tail_object_offset)) || | |
566 | !VALID64(le64toh(f->header->entry_array_offset))) | |
567 | return -ENODATA; | |
568 | ||
cec736d2 | 569 | if (f->writable) { |
cec736d2 | 570 | sd_id128_t machine_id; |
ae739cc1 | 571 | uint8_t state; |
cec736d2 LP |
572 | int r; |
573 | ||
574 | r = sd_id128_get_machine(&machine_id); | |
575 | if (r < 0) | |
576 | return r; | |
577 | ||
578 | if (!sd_id128_equal(machine_id, f->header->machine_id)) | |
579 | return -EHOSTDOWN; | |
580 | ||
de190aef | 581 | state = f->header->state; |
cec736d2 | 582 | |
71fa6f00 LP |
583 | if (state == STATE_ONLINE) { |
584 | log_debug("Journal file %s is already online. Assuming unclean closing.", f->path); | |
585 | return -EBUSY; | |
586 | } else if (state == STATE_ARCHIVED) | |
cec736d2 | 587 | return -ESHUTDOWN; |
71fa6f00 | 588 | else if (state != STATE_OFFLINE) { |
8facc349 | 589 | log_debug("Journal file %s has unknown state %i.", f->path, state); |
71fa6f00 LP |
590 | return -EBUSY; |
591 | } | |
ae739cc1 LP |
592 | |
593 | /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't | |
594 | * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks | |
595 | * bisection. */ | |
596 | if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) { | |
597 | log_debug("Journal file %s is from the future, refusing to append new data to it that'd be older.", f->path); | |
598 | return -ETXTBSY; | |
599 | } | |
cec736d2 LP |
600 | } |
601 | ||
d89c8fdf ZJS |
602 | f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header); |
603 | f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header); | |
c586dbf1 | 604 | |
f1889c91 | 605 | f->seal = JOURNAL_HEADER_SEALED(f->header); |
7560fffc | 606 | |
cec736d2 LP |
607 | return 0; |
608 | } | |
609 | ||
2678031a LP |
610 | static int journal_file_fstat(JournalFile *f) { |
611 | assert(f); | |
612 | assert(f->fd >= 0); | |
613 | ||
614 | if (fstat(f->fd, &f->last_stat) < 0) | |
615 | return -errno; | |
616 | ||
617 | f->last_stat_usec = now(CLOCK_MONOTONIC); | |
618 | ||
619 | /* Refuse appending to files that are already deleted */ | |
620 | if (f->last_stat.st_nlink <= 0) | |
621 | return -EIDRM; | |
622 | ||
623 | return 0; | |
624 | } | |
625 | ||
cec736d2 | 626 | static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { |
eda4b58b | 627 | uint64_t old_size, new_size; |
fec2aa2f | 628 | int r; |
cec736d2 LP |
629 | |
630 | assert(f); | |
c88cc6af | 631 | assert(f->header); |
cec736d2 | 632 | |
cec736d2 | 633 | /* We assume that this file is not sparse, and we know that |
38ac38b2 | 634 | * for sure, since we always call posix_fallocate() |
cec736d2 LP |
635 | * ourselves */ |
636 | ||
fa6ac760 LP |
637 | if (mmap_cache_got_sigbus(f->mmap, f->fd)) |
638 | return -EIO; | |
639 | ||
cec736d2 | 640 | old_size = |
23b0b2b2 | 641 | le64toh(f->header->header_size) + |
cec736d2 LP |
642 | le64toh(f->header->arena_size); |
643 | ||
bc85bfee | 644 | new_size = PAGE_ALIGN(offset + size); |
23b0b2b2 LP |
645 | if (new_size < le64toh(f->header->header_size)) |
646 | new_size = le64toh(f->header->header_size); | |
bc85bfee | 647 | |
2678031a LP |
648 | if (new_size <= old_size) { |
649 | ||
650 | /* We already pre-allocated enough space, but before | |
651 | * we write to it, let's check with fstat() if the | |
652 | * file got deleted, in order make sure we don't throw | |
653 | * away the data immediately. Don't check fstat() for | |
654 | * all writes though, but only once ever 10s. */ | |
655 | ||
656 | if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC)) | |
657 | return 0; | |
658 | ||
659 | return journal_file_fstat(f); | |
660 | } | |
661 | ||
662 | /* Allocate more space. */ | |
cec736d2 | 663 | |
a676e665 | 664 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) |
bc85bfee | 665 | return -E2BIG; |
cec736d2 | 666 | |
a676e665 | 667 | if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) { |
cec736d2 LP |
668 | struct statvfs svfs; |
669 | ||
670 | if (fstatvfs(f->fd, &svfs) >= 0) { | |
671 | uint64_t available; | |
672 | ||
070052ab | 673 | available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free); |
cec736d2 LP |
674 | |
675 | if (new_size - old_size > available) | |
676 | return -E2BIG; | |
677 | } | |
678 | } | |
679 | ||
eda4b58b LP |
680 | /* Increase by larger blocks at once */ |
681 | new_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE; | |
682 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) | |
683 | new_size = f->metrics.max_size; | |
684 | ||
bc85bfee LP |
685 | /* Note that the glibc fallocate() fallback is very |
686 | inefficient, hence we try to minimize the allocation area | |
687 | as we can. */ | |
fec2aa2f GV |
688 | r = posix_fallocate(f->fd, old_size, new_size - old_size); |
689 | if (r != 0) | |
690 | return -r; | |
cec736d2 | 691 | |
23b0b2b2 | 692 | f->header->arena_size = htole64(new_size - le64toh(f->header->header_size)); |
cec736d2 | 693 | |
2678031a | 694 | return journal_file_fstat(f); |
cec736d2 LP |
695 | } |
696 | ||
78519831 | 697 | static unsigned type_to_context(ObjectType type) { |
d3d3208f | 698 | /* One context for each type, plus one catch-all for the rest */ |
69adae51 | 699 | assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS); |
fa6ac760 | 700 | assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS); |
d05089d8 | 701 | return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0; |
d3d3208f MS |
702 | } |
703 | ||
7a9dabea | 704 | static int journal_file_move_to(JournalFile *f, ObjectType type, bool keep_always, uint64_t offset, uint64_t size, void **ret) { |
2678031a LP |
705 | int r; |
706 | ||
cec736d2 | 707 | assert(f); |
cec736d2 LP |
708 | assert(ret); |
709 | ||
7762e02b LP |
710 | if (size <= 0) |
711 | return -EINVAL; | |
712 | ||
2a59ea54 | 713 | /* Avoid SIGBUS on invalid accesses */ |
4bbdcdb3 LP |
714 | if (offset + size > (uint64_t) f->last_stat.st_size) { |
715 | /* Hmm, out of range? Let's refresh the fstat() data | |
716 | * first, before we trust that check. */ | |
717 | ||
2678031a LP |
718 | r = journal_file_fstat(f); |
719 | if (r < 0) | |
720 | return r; | |
721 | ||
722 | if (offset + size > (uint64_t) f->last_stat.st_size) | |
4bbdcdb3 LP |
723 | return -EADDRNOTAVAIL; |
724 | } | |
725 | ||
7a9dabea | 726 | return mmap_cache_get(f->mmap, f->fd, f->prot, type_to_context(type), keep_always, offset, size, &f->last_stat, ret); |
cec736d2 LP |
727 | } |
728 | ||
16e9f408 LP |
729 | static uint64_t minimum_header_size(Object *o) { |
730 | ||
b8e891e6 | 731 | static const uint64_t table[] = { |
16e9f408 LP |
732 | [OBJECT_DATA] = sizeof(DataObject), |
733 | [OBJECT_FIELD] = sizeof(FieldObject), | |
734 | [OBJECT_ENTRY] = sizeof(EntryObject), | |
735 | [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), | |
736 | [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), | |
737 | [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), | |
738 | [OBJECT_TAG] = sizeof(TagObject), | |
739 | }; | |
740 | ||
741 | if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) | |
742 | return sizeof(ObjectHeader); | |
743 | ||
744 | return table[o->object.type]; | |
745 | } | |
746 | ||
78519831 | 747 | int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) { |
cec736d2 LP |
748 | int r; |
749 | void *t; | |
750 | Object *o; | |
751 | uint64_t s; | |
752 | ||
753 | assert(f); | |
754 | assert(ret); | |
755 | ||
db11ac1a | 756 | /* Objects may only be located at multiple of 64 bit */ |
202fd896 LP |
757 | if (!VALID64(offset)) { |
758 | log_debug("Attempt to move to object at non-64bit boundary: %" PRIu64, offset); | |
bd30fdf2 | 759 | return -EBADMSG; |
202fd896 | 760 | } |
db11ac1a | 761 | |
50809d7a | 762 | /* Object may not be located in the file header */ |
202fd896 LP |
763 | if (offset < le64toh(f->header->header_size)) { |
764 | log_debug("Attempt to move to object located in file header: %" PRIu64, offset); | |
50809d7a | 765 | return -EBADMSG; |
202fd896 | 766 | } |
50809d7a | 767 | |
7a9dabea | 768 | r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t); |
cec736d2 LP |
769 | if (r < 0) |
770 | return r; | |
771 | ||
772 | o = (Object*) t; | |
773 | s = le64toh(o->object.size); | |
774 | ||
1c69f096 LP |
775 | if (s == 0) { |
776 | log_debug("Attempt to move to uninitialized object: %" PRIu64, offset); | |
777 | return -EBADMSG; | |
778 | } | |
202fd896 LP |
779 | if (s < sizeof(ObjectHeader)) { |
780 | log_debug("Attempt to move to overly short object: %" PRIu64, offset); | |
cec736d2 | 781 | return -EBADMSG; |
202fd896 | 782 | } |
cec736d2 | 783 | |
202fd896 LP |
784 | if (o->object.type <= OBJECT_UNUSED) { |
785 | log_debug("Attempt to move to object with invalid type: %" PRIu64, offset); | |
16e9f408 | 786 | return -EBADMSG; |
202fd896 | 787 | } |
16e9f408 | 788 | |
202fd896 LP |
789 | if (s < minimum_header_size(o)) { |
790 | log_debug("Attempt to move to truncated object: %" PRIu64, offset); | |
16e9f408 | 791 | return -EBADMSG; |
202fd896 | 792 | } |
16e9f408 | 793 | |
202fd896 LP |
794 | if (type > OBJECT_UNUSED && o->object.type != type) { |
795 | log_debug("Attempt to move to object of unexpected type: %" PRIu64, offset); | |
cec736d2 | 796 | return -EBADMSG; |
202fd896 | 797 | } |
cec736d2 LP |
798 | |
799 | if (s > sizeof(ObjectHeader)) { | |
7a9dabea | 800 | r = journal_file_move_to(f, type, false, offset, s, &t); |
cec736d2 LP |
801 | if (r < 0) |
802 | return r; | |
803 | ||
804 | o = (Object*) t; | |
805 | } | |
806 | ||
cec736d2 LP |
807 | *ret = o; |
808 | return 0; | |
809 | } | |
810 | ||
d98cc1f2 | 811 | static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { |
cec736d2 LP |
812 | uint64_t r; |
813 | ||
814 | assert(f); | |
c88cc6af | 815 | assert(f->header); |
cec736d2 | 816 | |
beec0085 | 817 | r = le64toh(f->header->tail_entry_seqnum) + 1; |
c2373f84 LP |
818 | |
819 | if (seqnum) { | |
de190aef | 820 | /* If an external seqnum counter was passed, we update |
c2373f84 LP |
821 | * both the local and the external one, and set it to |
822 | * the maximum of both */ | |
823 | ||
824 | if (*seqnum + 1 > r) | |
825 | r = *seqnum + 1; | |
826 | ||
827 | *seqnum = r; | |
828 | } | |
829 | ||
beec0085 | 830 | f->header->tail_entry_seqnum = htole64(r); |
cec736d2 | 831 | |
beec0085 LP |
832 | if (f->header->head_entry_seqnum == 0) |
833 | f->header->head_entry_seqnum = htole64(r); | |
de190aef | 834 | |
cec736d2 LP |
835 | return r; |
836 | } | |
837 | ||
78519831 | 838 | int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset) { |
cec736d2 LP |
839 | int r; |
840 | uint64_t p; | |
841 | Object *tail, *o; | |
842 | void *t; | |
843 | ||
844 | assert(f); | |
c88cc6af | 845 | assert(f->header); |
d05089d8 | 846 | assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX); |
cec736d2 LP |
847 | assert(size >= sizeof(ObjectHeader)); |
848 | assert(offset); | |
849 | assert(ret); | |
850 | ||
26687bf8 OS |
851 | r = journal_file_set_online(f); |
852 | if (r < 0) | |
853 | return r; | |
854 | ||
cec736d2 | 855 | p = le64toh(f->header->tail_object_offset); |
cec736d2 | 856 | if (p == 0) |
23b0b2b2 | 857 | p = le64toh(f->header->header_size); |
cec736d2 | 858 | else { |
d05089d8 | 859 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail); |
cec736d2 LP |
860 | if (r < 0) |
861 | return r; | |
862 | ||
863 | p += ALIGN64(le64toh(tail->object.size)); | |
864 | } | |
865 | ||
866 | r = journal_file_allocate(f, p, size); | |
867 | if (r < 0) | |
868 | return r; | |
869 | ||
fcde2389 | 870 | r = journal_file_move_to(f, type, false, p, size, &t); |
cec736d2 LP |
871 | if (r < 0) |
872 | return r; | |
873 | ||
874 | o = (Object*) t; | |
875 | ||
876 | zero(o->object); | |
de190aef | 877 | o->object.type = type; |
cec736d2 LP |
878 | o->object.size = htole64(size); |
879 | ||
880 | f->header->tail_object_offset = htole64(p); | |
cec736d2 LP |
881 | f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); |
882 | ||
883 | *ret = o; | |
884 | *offset = p; | |
885 | ||
886 | return 0; | |
887 | } | |
888 | ||
de190aef | 889 | static int journal_file_setup_data_hash_table(JournalFile *f) { |
cec736d2 LP |
890 | uint64_t s, p; |
891 | Object *o; | |
892 | int r; | |
893 | ||
894 | assert(f); | |
c88cc6af | 895 | assert(f->header); |
cec736d2 | 896 | |
070052ab LP |
897 | /* We estimate that we need 1 hash table entry per 768 bytes |
898 | of journal file and we want to make sure we never get | |
899 | beyond 75% fill level. Calculate the hash table size for | |
900 | the maximum file size based on these metrics. */ | |
4a92baf3 | 901 | |
dfabe643 | 902 | s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); |
4a92baf3 LP |
903 | if (s < DEFAULT_DATA_HASH_TABLE_SIZE) |
904 | s = DEFAULT_DATA_HASH_TABLE_SIZE; | |
905 | ||
507f22bd | 906 | log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem)); |
4a92baf3 | 907 | |
de190aef LP |
908 | r = journal_file_append_object(f, |
909 | OBJECT_DATA_HASH_TABLE, | |
910 | offsetof(Object, hash_table.items) + s, | |
911 | &o, &p); | |
cec736d2 LP |
912 | if (r < 0) |
913 | return r; | |
914 | ||
29804cc1 | 915 | memzero(o->hash_table.items, s); |
cec736d2 | 916 | |
de190aef LP |
917 | f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
918 | f->header->data_hash_table_size = htole64(s); | |
cec736d2 LP |
919 | |
920 | return 0; | |
921 | } | |
922 | ||
de190aef | 923 | static int journal_file_setup_field_hash_table(JournalFile *f) { |
cec736d2 LP |
924 | uint64_t s, p; |
925 | Object *o; | |
926 | int r; | |
927 | ||
928 | assert(f); | |
c88cc6af | 929 | assert(f->header); |
cec736d2 | 930 | |
3c1668da LP |
931 | /* We use a fixed size hash table for the fields as this |
932 | * number should grow very slowly only */ | |
933 | ||
de190aef LP |
934 | s = DEFAULT_FIELD_HASH_TABLE_SIZE; |
935 | r = journal_file_append_object(f, | |
936 | OBJECT_FIELD_HASH_TABLE, | |
937 | offsetof(Object, hash_table.items) + s, | |
938 | &o, &p); | |
cec736d2 LP |
939 | if (r < 0) |
940 | return r; | |
941 | ||
29804cc1 | 942 | memzero(o->hash_table.items, s); |
cec736d2 | 943 | |
de190aef LP |
944 | f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
945 | f->header->field_hash_table_size = htole64(s); | |
cec736d2 LP |
946 | |
947 | return 0; | |
948 | } | |
949 | ||
dade37d4 | 950 | int journal_file_map_data_hash_table(JournalFile *f) { |
cec736d2 LP |
951 | uint64_t s, p; |
952 | void *t; | |
953 | int r; | |
954 | ||
955 | assert(f); | |
c88cc6af | 956 | assert(f->header); |
cec736d2 | 957 | |
dade37d4 LP |
958 | if (f->data_hash_table) |
959 | return 0; | |
960 | ||
de190aef LP |
961 | p = le64toh(f->header->data_hash_table_offset); |
962 | s = le64toh(f->header->data_hash_table_size); | |
cec736d2 | 963 | |
de190aef | 964 | r = journal_file_move_to(f, |
16e9f408 | 965 | OBJECT_DATA_HASH_TABLE, |
fcde2389 | 966 | true, |
de190aef LP |
967 | p, s, |
968 | &t); | |
cec736d2 LP |
969 | if (r < 0) |
970 | return r; | |
971 | ||
de190aef | 972 | f->data_hash_table = t; |
cec736d2 LP |
973 | return 0; |
974 | } | |
975 | ||
dade37d4 | 976 | int journal_file_map_field_hash_table(JournalFile *f) { |
cec736d2 LP |
977 | uint64_t s, p; |
978 | void *t; | |
979 | int r; | |
980 | ||
981 | assert(f); | |
c88cc6af | 982 | assert(f->header); |
cec736d2 | 983 | |
dade37d4 LP |
984 | if (f->field_hash_table) |
985 | return 0; | |
986 | ||
de190aef LP |
987 | p = le64toh(f->header->field_hash_table_offset); |
988 | s = le64toh(f->header->field_hash_table_size); | |
cec736d2 | 989 | |
de190aef | 990 | r = journal_file_move_to(f, |
16e9f408 | 991 | OBJECT_FIELD_HASH_TABLE, |
fcde2389 | 992 | true, |
de190aef LP |
993 | p, s, |
994 | &t); | |
cec736d2 LP |
995 | if (r < 0) |
996 | return r; | |
997 | ||
de190aef | 998 | f->field_hash_table = t; |
cec736d2 LP |
999 | return 0; |
1000 | } | |
1001 | ||
3c1668da LP |
1002 | static int journal_file_link_field( |
1003 | JournalFile *f, | |
1004 | Object *o, | |
1005 | uint64_t offset, | |
1006 | uint64_t hash) { | |
1007 | ||
805d1486 | 1008 | uint64_t p, h, m; |
3c1668da LP |
1009 | int r; |
1010 | ||
1011 | assert(f); | |
c88cc6af | 1012 | assert(f->header); |
90d222c1 | 1013 | assert(f->field_hash_table); |
3c1668da LP |
1014 | assert(o); |
1015 | assert(offset > 0); | |
1016 | ||
1017 | if (o->object.type != OBJECT_FIELD) | |
1018 | return -EINVAL; | |
1019 | ||
805d1486 LP |
1020 | m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); |
1021 | if (m <= 0) | |
1022 | return -EBADMSG; | |
3c1668da | 1023 | |
805d1486 | 1024 | /* This might alter the window we are looking at */ |
3c1668da LP |
1025 | o->field.next_hash_offset = o->field.head_data_offset = 0; |
1026 | ||
805d1486 | 1027 | h = hash % m; |
3c1668da LP |
1028 | p = le64toh(f->field_hash_table[h].tail_hash_offset); |
1029 | if (p == 0) | |
1030 | f->field_hash_table[h].head_hash_offset = htole64(offset); | |
1031 | else { | |
1032 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1033 | if (r < 0) | |
1034 | return r; | |
1035 | ||
1036 | o->field.next_hash_offset = htole64(offset); | |
1037 | } | |
1038 | ||
1039 | f->field_hash_table[h].tail_hash_offset = htole64(offset); | |
1040 | ||
1041 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
1042 | f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1); | |
1043 | ||
1044 | return 0; | |
1045 | } | |
1046 | ||
1047 | static int journal_file_link_data( | |
1048 | JournalFile *f, | |
1049 | Object *o, | |
1050 | uint64_t offset, | |
1051 | uint64_t hash) { | |
1052 | ||
805d1486 | 1053 | uint64_t p, h, m; |
cec736d2 LP |
1054 | int r; |
1055 | ||
1056 | assert(f); | |
c88cc6af | 1057 | assert(f->header); |
90d222c1 | 1058 | assert(f->data_hash_table); |
cec736d2 LP |
1059 | assert(o); |
1060 | assert(offset > 0); | |
b588975f LP |
1061 | |
1062 | if (o->object.type != OBJECT_DATA) | |
1063 | return -EINVAL; | |
cec736d2 | 1064 | |
805d1486 LP |
1065 | m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); |
1066 | if (m <= 0) | |
1067 | return -EBADMSG; | |
48496df6 | 1068 | |
805d1486 | 1069 | /* This might alter the window we are looking at */ |
de190aef LP |
1070 | o->data.next_hash_offset = o->data.next_field_offset = 0; |
1071 | o->data.entry_offset = o->data.entry_array_offset = 0; | |
1072 | o->data.n_entries = 0; | |
cec736d2 | 1073 | |
805d1486 | 1074 | h = hash % m; |
8db4213e | 1075 | p = le64toh(f->data_hash_table[h].tail_hash_offset); |
3c1668da | 1076 | if (p == 0) |
cec736d2 | 1077 | /* Only entry in the hash table is easy */ |
de190aef | 1078 | f->data_hash_table[h].head_hash_offset = htole64(offset); |
3c1668da | 1079 | else { |
48496df6 LP |
1080 | /* Move back to the previous data object, to patch in |
1081 | * pointer */ | |
cec736d2 | 1082 | |
de190aef | 1083 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1084 | if (r < 0) |
1085 | return r; | |
1086 | ||
de190aef | 1087 | o->data.next_hash_offset = htole64(offset); |
cec736d2 LP |
1088 | } |
1089 | ||
de190aef | 1090 | f->data_hash_table[h].tail_hash_offset = htole64(offset); |
cec736d2 | 1091 | |
dca6219e LP |
1092 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
1093 | f->header->n_data = htole64(le64toh(f->header->n_data) + 1); | |
1094 | ||
cec736d2 LP |
1095 | return 0; |
1096 | } | |
1097 | ||
3c1668da LP |
1098 | int journal_file_find_field_object_with_hash( |
1099 | JournalFile *f, | |
1100 | const void *field, uint64_t size, uint64_t hash, | |
1101 | Object **ret, uint64_t *offset) { | |
1102 | ||
805d1486 | 1103 | uint64_t p, osize, h, m; |
3c1668da LP |
1104 | int r; |
1105 | ||
1106 | assert(f); | |
c88cc6af | 1107 | assert(f->header); |
3c1668da LP |
1108 | assert(field && size > 0); |
1109 | ||
dade37d4 LP |
1110 | /* If the field hash table is empty, we can't find anything */ |
1111 | if (le64toh(f->header->field_hash_table_size) <= 0) | |
1112 | return 0; | |
1113 | ||
1114 | /* Map the field hash table, if it isn't mapped yet. */ | |
1115 | r = journal_file_map_field_hash_table(f); | |
1116 | if (r < 0) | |
1117 | return r; | |
1118 | ||
3c1668da LP |
1119 | osize = offsetof(Object, field.payload) + size; |
1120 | ||
805d1486 | 1121 | m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); |
805d1486 | 1122 | if (m <= 0) |
3c1668da LP |
1123 | return -EBADMSG; |
1124 | ||
805d1486 | 1125 | h = hash % m; |
3c1668da LP |
1126 | p = le64toh(f->field_hash_table[h].head_hash_offset); |
1127 | ||
1128 | while (p > 0) { | |
1129 | Object *o; | |
1130 | ||
1131 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1132 | if (r < 0) | |
1133 | return r; | |
1134 | ||
1135 | if (le64toh(o->field.hash) == hash && | |
1136 | le64toh(o->object.size) == osize && | |
1137 | memcmp(o->field.payload, field, size) == 0) { | |
1138 | ||
1139 | if (ret) | |
1140 | *ret = o; | |
1141 | if (offset) | |
1142 | *offset = p; | |
1143 | ||
1144 | return 1; | |
1145 | } | |
1146 | ||
1147 | p = le64toh(o->field.next_hash_offset); | |
1148 | } | |
1149 | ||
1150 | return 0; | |
1151 | } | |
1152 | ||
1153 | int journal_file_find_field_object( | |
1154 | JournalFile *f, | |
1155 | const void *field, uint64_t size, | |
1156 | Object **ret, uint64_t *offset) { | |
1157 | ||
1158 | uint64_t hash; | |
1159 | ||
1160 | assert(f); | |
1161 | assert(field && size > 0); | |
1162 | ||
1163 | hash = hash64(field, size); | |
1164 | ||
1165 | return journal_file_find_field_object_with_hash(f, | |
1166 | field, size, hash, | |
1167 | ret, offset); | |
1168 | } | |
1169 | ||
de190aef LP |
1170 | int journal_file_find_data_object_with_hash( |
1171 | JournalFile *f, | |
1172 | const void *data, uint64_t size, uint64_t hash, | |
1173 | Object **ret, uint64_t *offset) { | |
48496df6 | 1174 | |
805d1486 | 1175 | uint64_t p, osize, h, m; |
cec736d2 LP |
1176 | int r; |
1177 | ||
1178 | assert(f); | |
c88cc6af | 1179 | assert(f->header); |
cec736d2 LP |
1180 | assert(data || size == 0); |
1181 | ||
dade37d4 LP |
1182 | /* If there's no data hash table, then there's no entry. */ |
1183 | if (le64toh(f->header->data_hash_table_size) <= 0) | |
1184 | return 0; | |
1185 | ||
1186 | /* Map the data hash table, if it isn't mapped yet. */ | |
1187 | r = journal_file_map_data_hash_table(f); | |
1188 | if (r < 0) | |
1189 | return r; | |
1190 | ||
cec736d2 LP |
1191 | osize = offsetof(Object, data.payload) + size; |
1192 | ||
805d1486 LP |
1193 | m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); |
1194 | if (m <= 0) | |
bc85bfee LP |
1195 | return -EBADMSG; |
1196 | ||
805d1486 | 1197 | h = hash % m; |
de190aef | 1198 | p = le64toh(f->data_hash_table[h].head_hash_offset); |
cec736d2 | 1199 | |
de190aef LP |
1200 | while (p > 0) { |
1201 | Object *o; | |
cec736d2 | 1202 | |
de190aef | 1203 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1204 | if (r < 0) |
1205 | return r; | |
1206 | ||
807e17f0 | 1207 | if (le64toh(o->data.hash) != hash) |
85a131e8 | 1208 | goto next; |
807e17f0 | 1209 | |
d89c8fdf | 1210 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
3b1a55e1 | 1211 | #if defined(HAVE_XZ) || defined(HAVE_LZ4) |
fa1c4b51 | 1212 | uint64_t l; |
a7f7d1bd | 1213 | size_t rsize = 0; |
cec736d2 | 1214 | |
807e17f0 LP |
1215 | l = le64toh(o->object.size); |
1216 | if (l <= offsetof(Object, data.payload)) | |
cec736d2 LP |
1217 | return -EBADMSG; |
1218 | ||
807e17f0 LP |
1219 | l -= offsetof(Object, data.payload); |
1220 | ||
d89c8fdf ZJS |
1221 | r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, |
1222 | o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0); | |
1223 | if (r < 0) | |
1224 | return r; | |
807e17f0 | 1225 | |
b785c858 | 1226 | if (rsize == size && |
807e17f0 LP |
1227 | memcmp(f->compress_buffer, data, size) == 0) { |
1228 | ||
1229 | if (ret) | |
1230 | *ret = o; | |
1231 | ||
1232 | if (offset) | |
1233 | *offset = p; | |
1234 | ||
1235 | return 1; | |
1236 | } | |
3b1a55e1 ZJS |
1237 | #else |
1238 | return -EPROTONOSUPPORT; | |
1239 | #endif | |
807e17f0 LP |
1240 | } else if (le64toh(o->object.size) == osize && |
1241 | memcmp(o->data.payload, data, size) == 0) { | |
1242 | ||
cec736d2 LP |
1243 | if (ret) |
1244 | *ret = o; | |
1245 | ||
1246 | if (offset) | |
1247 | *offset = p; | |
1248 | ||
de190aef | 1249 | return 1; |
cec736d2 LP |
1250 | } |
1251 | ||
85a131e8 | 1252 | next: |
cec736d2 LP |
1253 | p = le64toh(o->data.next_hash_offset); |
1254 | } | |
1255 | ||
de190aef LP |
1256 | return 0; |
1257 | } | |
1258 | ||
1259 | int journal_file_find_data_object( | |
1260 | JournalFile *f, | |
1261 | const void *data, uint64_t size, | |
1262 | Object **ret, uint64_t *offset) { | |
1263 | ||
1264 | uint64_t hash; | |
1265 | ||
1266 | assert(f); | |
1267 | assert(data || size == 0); | |
1268 | ||
1269 | hash = hash64(data, size); | |
1270 | ||
1271 | return journal_file_find_data_object_with_hash(f, | |
1272 | data, size, hash, | |
1273 | ret, offset); | |
1274 | } | |
1275 | ||
3c1668da LP |
1276 | static int journal_file_append_field( |
1277 | JournalFile *f, | |
1278 | const void *field, uint64_t size, | |
1279 | Object **ret, uint64_t *offset) { | |
1280 | ||
1281 | uint64_t hash, p; | |
1282 | uint64_t osize; | |
1283 | Object *o; | |
1284 | int r; | |
1285 | ||
1286 | assert(f); | |
1287 | assert(field && size > 0); | |
1288 | ||
1289 | hash = hash64(field, size); | |
1290 | ||
1291 | r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p); | |
1292 | if (r < 0) | |
1293 | return r; | |
1294 | else if (r > 0) { | |
1295 | ||
1296 | if (ret) | |
1297 | *ret = o; | |
1298 | ||
1299 | if (offset) | |
1300 | *offset = p; | |
1301 | ||
1302 | return 0; | |
1303 | } | |
1304 | ||
1305 | osize = offsetof(Object, field.payload) + size; | |
1306 | r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p); | |
8c92d4bb LP |
1307 | if (r < 0) |
1308 | return r; | |
3c1668da LP |
1309 | |
1310 | o->field.hash = htole64(hash); | |
1311 | memcpy(o->field.payload, field, size); | |
1312 | ||
1313 | r = journal_file_link_field(f, o, p, hash); | |
1314 | if (r < 0) | |
1315 | return r; | |
1316 | ||
1317 | /* The linking might have altered the window, so let's | |
1318 | * refresh our pointer */ | |
1319 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1320 | if (r < 0) | |
1321 | return r; | |
1322 | ||
1323 | #ifdef HAVE_GCRYPT | |
1324 | r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p); | |
1325 | if (r < 0) | |
1326 | return r; | |
1327 | #endif | |
1328 | ||
1329 | if (ret) | |
1330 | *ret = o; | |
1331 | ||
1332 | if (offset) | |
1333 | *offset = p; | |
1334 | ||
1335 | return 0; | |
1336 | } | |
1337 | ||
48496df6 LP |
1338 | static int journal_file_append_data( |
1339 | JournalFile *f, | |
1340 | const void *data, uint64_t size, | |
1341 | Object **ret, uint64_t *offset) { | |
1342 | ||
de190aef LP |
1343 | uint64_t hash, p; |
1344 | uint64_t osize; | |
1345 | Object *o; | |
d89c8fdf | 1346 | int r, compression = 0; |
3c1668da | 1347 | const void *eq; |
de190aef LP |
1348 | |
1349 | assert(f); | |
1350 | assert(data || size == 0); | |
1351 | ||
1352 | hash = hash64(data, size); | |
1353 | ||
1354 | r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p); | |
1355 | if (r < 0) | |
1356 | return r; | |
0240c603 | 1357 | if (r > 0) { |
de190aef LP |
1358 | |
1359 | if (ret) | |
1360 | *ret = o; | |
1361 | ||
1362 | if (offset) | |
1363 | *offset = p; | |
1364 | ||
1365 | return 0; | |
1366 | } | |
1367 | ||
1368 | osize = offsetof(Object, data.payload) + size; | |
1369 | r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); | |
cec736d2 LP |
1370 | if (r < 0) |
1371 | return r; | |
1372 | ||
cec736d2 | 1373 | o->data.hash = htole64(hash); |
807e17f0 | 1374 | |
d89c8fdf | 1375 | #if defined(HAVE_XZ) || defined(HAVE_LZ4) |
d1afbcd2 | 1376 | if (JOURNAL_FILE_COMPRESS(f) && size >= COMPRESSION_SIZE_THRESHOLD) { |
a7f7d1bd | 1377 | size_t rsize = 0; |
807e17f0 | 1378 | |
5d6f46b6 | 1379 | compression = compress_blob(data, size, o->data.payload, size - 1, &rsize); |
807e17f0 | 1380 | |
d1afbcd2 | 1381 | if (compression >= 0) { |
807e17f0 | 1382 | o->object.size = htole64(offsetof(Object, data.payload) + rsize); |
d89c8fdf | 1383 | o->object.flags |= compression; |
807e17f0 | 1384 | |
fa1c4b51 | 1385 | log_debug("Compressed data object %"PRIu64" -> %zu using %s", |
d89c8fdf | 1386 | size, rsize, object_compressed_to_string(compression)); |
d1afbcd2 LP |
1387 | } else |
1388 | /* Compression didn't work, we don't really care why, let's continue without compression */ | |
1389 | compression = 0; | |
807e17f0 LP |
1390 | } |
1391 | #endif | |
1392 | ||
75f32f04 ZJS |
1393 | if (compression == 0) |
1394 | memcpy_safe(o->data.payload, data, size); | |
cec736d2 | 1395 | |
de190aef | 1396 | r = journal_file_link_data(f, o, p, hash); |
cec736d2 LP |
1397 | if (r < 0) |
1398 | return r; | |
1399 | ||
33685a5a FB |
1400 | #ifdef HAVE_GCRYPT |
1401 | r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p); | |
1402 | if (r < 0) | |
1403 | return r; | |
1404 | #endif | |
1405 | ||
48496df6 LP |
1406 | /* The linking might have altered the window, so let's |
1407 | * refresh our pointer */ | |
1408 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
1409 | if (r < 0) | |
1410 | return r; | |
1411 | ||
08c6f819 SL |
1412 | if (!data) |
1413 | eq = NULL; | |
1414 | else | |
1415 | eq = memchr(data, '=', size); | |
3c1668da | 1416 | if (eq && eq > data) { |
748db592 | 1417 | Object *fo = NULL; |
3c1668da | 1418 | uint64_t fp; |
3c1668da LP |
1419 | |
1420 | /* Create field object ... */ | |
1421 | r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp); | |
1422 | if (r < 0) | |
1423 | return r; | |
1424 | ||
1425 | /* ... and link it in. */ | |
1426 | o->data.next_field_offset = fo->field.head_data_offset; | |
1427 | fo->field.head_data_offset = le64toh(p); | |
1428 | } | |
1429 | ||
cec736d2 LP |
1430 | if (ret) |
1431 | *ret = o; | |
1432 | ||
1433 | if (offset) | |
de190aef | 1434 | *offset = p; |
cec736d2 LP |
1435 | |
1436 | return 0; | |
1437 | } | |
1438 | ||
1439 | uint64_t journal_file_entry_n_items(Object *o) { | |
1440 | assert(o); | |
b588975f LP |
1441 | |
1442 | if (o->object.type != OBJECT_ENTRY) | |
1443 | return 0; | |
cec736d2 LP |
1444 | |
1445 | return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem); | |
1446 | } | |
1447 | ||
0284adc6 | 1448 | uint64_t journal_file_entry_array_n_items(Object *o) { |
de190aef | 1449 | assert(o); |
b588975f LP |
1450 | |
1451 | if (o->object.type != OBJECT_ENTRY_ARRAY) | |
1452 | return 0; | |
de190aef LP |
1453 | |
1454 | return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t); | |
1455 | } | |
1456 | ||
fb9a24b6 LP |
1457 | uint64_t journal_file_hash_table_n_items(Object *o) { |
1458 | assert(o); | |
b588975f LP |
1459 | |
1460 | if (o->object.type != OBJECT_DATA_HASH_TABLE && | |
1461 | o->object.type != OBJECT_FIELD_HASH_TABLE) | |
1462 | return 0; | |
fb9a24b6 LP |
1463 | |
1464 | return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem); | |
1465 | } | |
1466 | ||
de190aef | 1467 | static int link_entry_into_array(JournalFile *f, |
4fd052ae FC |
1468 | le64_t *first, |
1469 | le64_t *idx, | |
de190aef | 1470 | uint64_t p) { |
cec736d2 | 1471 | int r; |
de190aef LP |
1472 | uint64_t n = 0, ap = 0, q, i, a, hidx; |
1473 | Object *o; | |
1474 | ||
cec736d2 | 1475 | assert(f); |
c88cc6af | 1476 | assert(f->header); |
de190aef LP |
1477 | assert(first); |
1478 | assert(idx); | |
1479 | assert(p > 0); | |
cec736d2 | 1480 | |
de190aef LP |
1481 | a = le64toh(*first); |
1482 | i = hidx = le64toh(*idx); | |
1483 | while (a > 0) { | |
1484 | ||
1485 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); | |
1486 | if (r < 0) | |
1487 | return r; | |
cec736d2 | 1488 | |
de190aef LP |
1489 | n = journal_file_entry_array_n_items(o); |
1490 | if (i < n) { | |
1491 | o->entry_array.items[i] = htole64(p); | |
1492 | *idx = htole64(hidx + 1); | |
1493 | return 0; | |
1494 | } | |
cec736d2 | 1495 | |
de190aef LP |
1496 | i -= n; |
1497 | ap = a; | |
1498 | a = le64toh(o->entry_array.next_entry_array_offset); | |
1499 | } | |
1500 | ||
1501 | if (hidx > n) | |
1502 | n = (hidx+1) * 2; | |
1503 | else | |
1504 | n = n * 2; | |
1505 | ||
1506 | if (n < 4) | |
1507 | n = 4; | |
1508 | ||
1509 | r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, | |
1510 | offsetof(Object, entry_array.items) + n * sizeof(uint64_t), | |
1511 | &o, &q); | |
cec736d2 LP |
1512 | if (r < 0) |
1513 | return r; | |
1514 | ||
feb12d3e | 1515 | #ifdef HAVE_GCRYPT |
5996c7c2 | 1516 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q); |
b0af6f41 LP |
1517 | if (r < 0) |
1518 | return r; | |
feb12d3e | 1519 | #endif |
b0af6f41 | 1520 | |
de190aef | 1521 | o->entry_array.items[i] = htole64(p); |
cec736d2 | 1522 | |
de190aef | 1523 | if (ap == 0) |
7be3aa17 | 1524 | *first = htole64(q); |
cec736d2 | 1525 | else { |
de190aef | 1526 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); |
cec736d2 LP |
1527 | if (r < 0) |
1528 | return r; | |
1529 | ||
de190aef LP |
1530 | o->entry_array.next_entry_array_offset = htole64(q); |
1531 | } | |
cec736d2 | 1532 | |
2dee23eb LP |
1533 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
1534 | f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); | |
1535 | ||
de190aef LP |
1536 | *idx = htole64(hidx + 1); |
1537 | ||
1538 | return 0; | |
1539 | } | |
cec736d2 | 1540 | |
de190aef | 1541 | static int link_entry_into_array_plus_one(JournalFile *f, |
4fd052ae FC |
1542 | le64_t *extra, |
1543 | le64_t *first, | |
1544 | le64_t *idx, | |
de190aef LP |
1545 | uint64_t p) { |
1546 | ||
1547 | int r; | |
1548 | ||
1549 | assert(f); | |
1550 | assert(extra); | |
1551 | assert(first); | |
1552 | assert(idx); | |
1553 | assert(p > 0); | |
1554 | ||
1555 | if (*idx == 0) | |
1556 | *extra = htole64(p); | |
1557 | else { | |
4fd052ae | 1558 | le64_t i; |
de190aef | 1559 | |
7be3aa17 | 1560 | i = htole64(le64toh(*idx) - 1); |
de190aef LP |
1561 | r = link_entry_into_array(f, first, &i, p); |
1562 | if (r < 0) | |
1563 | return r; | |
cec736d2 LP |
1564 | } |
1565 | ||
de190aef LP |
1566 | *idx = htole64(le64toh(*idx) + 1); |
1567 | return 0; | |
1568 | } | |
1569 | ||
1570 | static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { | |
1571 | uint64_t p; | |
1572 | int r; | |
1573 | assert(f); | |
1574 | assert(o); | |
1575 | assert(offset > 0); | |
1576 | ||
1577 | p = le64toh(o->entry.items[i].object_offset); | |
1578 | if (p == 0) | |
1579 | return -EINVAL; | |
1580 | ||
1581 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
cec736d2 LP |
1582 | if (r < 0) |
1583 | return r; | |
1584 | ||
de190aef LP |
1585 | return link_entry_into_array_plus_one(f, |
1586 | &o->data.entry_offset, | |
1587 | &o->data.entry_array_offset, | |
1588 | &o->data.n_entries, | |
1589 | offset); | |
cec736d2 LP |
1590 | } |
1591 | ||
1592 | static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { | |
de190aef | 1593 | uint64_t n, i; |
cec736d2 LP |
1594 | int r; |
1595 | ||
1596 | assert(f); | |
c88cc6af | 1597 | assert(f->header); |
cec736d2 LP |
1598 | assert(o); |
1599 | assert(offset > 0); | |
b588975f LP |
1600 | |
1601 | if (o->object.type != OBJECT_ENTRY) | |
1602 | return -EINVAL; | |
cec736d2 | 1603 | |
b788cc23 LP |
1604 | __sync_synchronize(); |
1605 | ||
cec736d2 | 1606 | /* Link up the entry itself */ |
de190aef LP |
1607 | r = link_entry_into_array(f, |
1608 | &f->header->entry_array_offset, | |
1609 | &f->header->n_entries, | |
1610 | offset); | |
1611 | if (r < 0) | |
1612 | return r; | |
cec736d2 | 1613 | |
507f22bd | 1614 | /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */ |
cec736d2 | 1615 | |
de190aef | 1616 | if (f->header->head_entry_realtime == 0) |
0ac38b70 | 1617 | f->header->head_entry_realtime = o->entry.realtime; |
cec736d2 | 1618 | |
0ac38b70 | 1619 | f->header->tail_entry_realtime = o->entry.realtime; |
de190aef LP |
1620 | f->header->tail_entry_monotonic = o->entry.monotonic; |
1621 | ||
1622 | f->tail_entry_monotonic_valid = true; | |
cec736d2 LP |
1623 | |
1624 | /* Link up the items */ | |
1625 | n = journal_file_entry_n_items(o); | |
1626 | for (i = 0; i < n; i++) { | |
1627 | r = journal_file_link_entry_item(f, o, offset, i); | |
1628 | if (r < 0) | |
1629 | return r; | |
1630 | } | |
1631 | ||
cec736d2 LP |
1632 | return 0; |
1633 | } | |
1634 | ||
1635 | static int journal_file_append_entry_internal( | |
1636 | JournalFile *f, | |
1637 | const dual_timestamp *ts, | |
1638 | uint64_t xor_hash, | |
1639 | const EntryItem items[], unsigned n_items, | |
de190aef | 1640 | uint64_t *seqnum, |
cec736d2 LP |
1641 | Object **ret, uint64_t *offset) { |
1642 | uint64_t np; | |
1643 | uint64_t osize; | |
1644 | Object *o; | |
1645 | int r; | |
1646 | ||
1647 | assert(f); | |
c88cc6af | 1648 | assert(f->header); |
cec736d2 | 1649 | assert(items || n_items == 0); |
de190aef | 1650 | assert(ts); |
cec736d2 LP |
1651 | |
1652 | osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem)); | |
1653 | ||
de190aef | 1654 | r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); |
cec736d2 LP |
1655 | if (r < 0) |
1656 | return r; | |
1657 | ||
d98cc1f2 | 1658 | o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); |
75f32f04 | 1659 | memcpy_safe(o->entry.items, items, n_items * sizeof(EntryItem)); |
de190aef LP |
1660 | o->entry.realtime = htole64(ts->realtime); |
1661 | o->entry.monotonic = htole64(ts->monotonic); | |
cec736d2 LP |
1662 | o->entry.xor_hash = htole64(xor_hash); |
1663 | o->entry.boot_id = f->header->boot_id; | |
1664 | ||
feb12d3e | 1665 | #ifdef HAVE_GCRYPT |
5996c7c2 | 1666 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np); |
b0af6f41 LP |
1667 | if (r < 0) |
1668 | return r; | |
feb12d3e | 1669 | #endif |
b0af6f41 | 1670 | |
cec736d2 LP |
1671 | r = journal_file_link_entry(f, o, np); |
1672 | if (r < 0) | |
1673 | return r; | |
1674 | ||
1675 | if (ret) | |
1676 | *ret = o; | |
1677 | ||
1678 | if (offset) | |
1679 | *offset = np; | |
1680 | ||
1681 | return 0; | |
1682 | } | |
1683 | ||
cf244689 | 1684 | void journal_file_post_change(JournalFile *f) { |
50f20cfd LP |
1685 | assert(f); |
1686 | ||
1687 | /* inotify() does not receive IN_MODIFY events from file | |
1688 | * accesses done via mmap(). After each access we hence | |
1689 | * trigger IN_MODIFY by truncating the journal file to its | |
1690 | * current size which triggers IN_MODIFY. */ | |
1691 | ||
bc85bfee LP |
1692 | __sync_synchronize(); |
1693 | ||
50f20cfd | 1694 | if (ftruncate(f->fd, f->last_stat.st_size) < 0) |
e167d7fd | 1695 | log_debug_errno(errno, "Failed to truncate file to its own size: %m"); |
50f20cfd LP |
1696 | } |
1697 | ||
7a24f3bf VC |
1698 | static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) { |
1699 | assert(userdata); | |
1700 | ||
1701 | journal_file_post_change(userdata); | |
1702 | ||
1703 | return 1; | |
1704 | } | |
1705 | ||
1706 | static void schedule_post_change(JournalFile *f) { | |
1707 | sd_event_source *timer; | |
1708 | int enabled, r; | |
1709 | uint64_t now; | |
1710 | ||
1711 | assert(f); | |
1712 | assert(f->post_change_timer); | |
1713 | ||
1714 | timer = f->post_change_timer; | |
1715 | ||
1716 | r = sd_event_source_get_enabled(timer, &enabled); | |
1717 | if (r < 0) { | |
e167d7fd LP |
1718 | log_debug_errno(r, "Failed to get ftruncate timer state: %m"); |
1719 | goto fail; | |
7a24f3bf VC |
1720 | } |
1721 | ||
1722 | if (enabled == SD_EVENT_ONESHOT) | |
1723 | return; | |
1724 | ||
1725 | r = sd_event_now(sd_event_source_get_event(timer), CLOCK_MONOTONIC, &now); | |
1726 | if (r < 0) { | |
e167d7fd LP |
1727 | log_debug_errno(r, "Failed to get clock's now for scheduling ftruncate: %m"); |
1728 | goto fail; | |
7a24f3bf VC |
1729 | } |
1730 | ||
1731 | r = sd_event_source_set_time(timer, now+f->post_change_timer_period); | |
1732 | if (r < 0) { | |
e167d7fd LP |
1733 | log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m"); |
1734 | goto fail; | |
7a24f3bf VC |
1735 | } |
1736 | ||
1737 | r = sd_event_source_set_enabled(timer, SD_EVENT_ONESHOT); | |
1738 | if (r < 0) { | |
e167d7fd LP |
1739 | log_debug_errno(r, "Failed to enable scheduled ftruncate: %m"); |
1740 | goto fail; | |
7a24f3bf | 1741 | } |
e167d7fd LP |
1742 | |
1743 | return; | |
1744 | ||
1745 | fail: | |
1746 | /* On failure, let's simply post the change immediately. */ | |
1747 | journal_file_post_change(f); | |
7a24f3bf VC |
1748 | } |
1749 | ||
1750 | /* Enable coalesced change posting in a timer on the provided sd_event instance */ | |
1751 | int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) { | |
1752 | _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL; | |
1753 | int r; | |
1754 | ||
1755 | assert(f); | |
1756 | assert_return(!f->post_change_timer, -EINVAL); | |
1757 | assert(e); | |
1758 | assert(t); | |
1759 | ||
1760 | r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f); | |
1761 | if (r < 0) | |
1762 | return r; | |
1763 | ||
1764 | r = sd_event_source_set_enabled(timer, SD_EVENT_OFF); | |
1765 | if (r < 0) | |
1766 | return r; | |
1767 | ||
1768 | f->post_change_timer = timer; | |
1769 | timer = NULL; | |
1770 | f->post_change_timer_period = t; | |
1771 | ||
1772 | return r; | |
1773 | } | |
1774 | ||
1f2da9ec LP |
1775 | static int entry_item_cmp(const void *_a, const void *_b) { |
1776 | const EntryItem *a = _a, *b = _b; | |
1777 | ||
1778 | if (le64toh(a->object_offset) < le64toh(b->object_offset)) | |
1779 | return -1; | |
1780 | if (le64toh(a->object_offset) > le64toh(b->object_offset)) | |
1781 | return 1; | |
1782 | return 0; | |
1783 | } | |
1784 | ||
de190aef | 1785 | int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqnum, Object **ret, uint64_t *offset) { |
cec736d2 LP |
1786 | unsigned i; |
1787 | EntryItem *items; | |
1788 | int r; | |
1789 | uint64_t xor_hash = 0; | |
de190aef | 1790 | struct dual_timestamp _ts; |
cec736d2 LP |
1791 | |
1792 | assert(f); | |
c88cc6af | 1793 | assert(f->header); |
cec736d2 LP |
1794 | assert(iovec || n_iovec == 0); |
1795 | ||
de190aef LP |
1796 | if (!ts) { |
1797 | dual_timestamp_get(&_ts); | |
1798 | ts = &_ts; | |
1799 | } | |
1800 | ||
feb12d3e | 1801 | #ifdef HAVE_GCRYPT |
7560fffc LP |
1802 | r = journal_file_maybe_append_tag(f, ts->realtime); |
1803 | if (r < 0) | |
1804 | return r; | |
feb12d3e | 1805 | #endif |
7560fffc | 1806 | |
64825d3c | 1807 | /* alloca() can't take 0, hence let's allocate at least one */ |
9607d947 | 1808 | items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec)); |
cec736d2 LP |
1809 | |
1810 | for (i = 0; i < n_iovec; i++) { | |
1811 | uint64_t p; | |
1812 | Object *o; | |
1813 | ||
1814 | r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); | |
1815 | if (r < 0) | |
cf244689 | 1816 | return r; |
cec736d2 LP |
1817 | |
1818 | xor_hash ^= le64toh(o->data.hash); | |
1819 | items[i].object_offset = htole64(p); | |
de7b95cd | 1820 | items[i].hash = o->data.hash; |
cec736d2 LP |
1821 | } |
1822 | ||
1f2da9ec LP |
1823 | /* Order by the position on disk, in order to improve seek |
1824 | * times for rotating media. */ | |
7ff7394d | 1825 | qsort_safe(items, n_iovec, sizeof(EntryItem), entry_item_cmp); |
1f2da9ec | 1826 | |
de190aef | 1827 | r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset); |
cec736d2 | 1828 | |
fa6ac760 LP |
1829 | /* If the memory mapping triggered a SIGBUS then we return an |
1830 | * IO error and ignore the error code passed down to us, since | |
1831 | * it is very likely just an effect of a nullified replacement | |
1832 | * mapping page */ | |
1833 | ||
1834 | if (mmap_cache_got_sigbus(f->mmap, f->fd)) | |
1835 | r = -EIO; | |
1836 | ||
7a24f3bf VC |
1837 | if (f->post_change_timer) |
1838 | schedule_post_change(f); | |
1839 | else | |
1840 | journal_file_post_change(f); | |
50f20cfd | 1841 | |
cec736d2 LP |
1842 | return r; |
1843 | } | |
1844 | ||
a4bcff5b | 1845 | typedef struct ChainCacheItem { |
fb099c8d | 1846 | uint64_t first; /* the array at the beginning of the chain */ |
a4bcff5b LP |
1847 | uint64_t array; /* the cached array */ |
1848 | uint64_t begin; /* the first item in the cached array */ | |
1849 | uint64_t total; /* the total number of items in all arrays before this one in the chain */ | |
f268980d | 1850 | uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */ |
a4bcff5b LP |
1851 | } ChainCacheItem; |
1852 | ||
1853 | static void chain_cache_put( | |
4743015d | 1854 | OrderedHashmap *h, |
a4bcff5b LP |
1855 | ChainCacheItem *ci, |
1856 | uint64_t first, | |
1857 | uint64_t array, | |
1858 | uint64_t begin, | |
f268980d LP |
1859 | uint64_t total, |
1860 | uint64_t last_index) { | |
a4bcff5b LP |
1861 | |
1862 | if (!ci) { | |
34741aa3 LP |
1863 | /* If the chain item to cache for this chain is the |
1864 | * first one it's not worth caching anything */ | |
1865 | if (array == first) | |
1866 | return; | |
1867 | ||
29433089 | 1868 | if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) { |
4743015d | 1869 | ci = ordered_hashmap_steal_first(h); |
29433089 LP |
1870 | assert(ci); |
1871 | } else { | |
a4bcff5b LP |
1872 | ci = new(ChainCacheItem, 1); |
1873 | if (!ci) | |
1874 | return; | |
1875 | } | |
1876 | ||
1877 | ci->first = first; | |
1878 | ||
4743015d | 1879 | if (ordered_hashmap_put(h, &ci->first, ci) < 0) { |
a4bcff5b LP |
1880 | free(ci); |
1881 | return; | |
1882 | } | |
1883 | } else | |
1884 | assert(ci->first == first); | |
1885 | ||
1886 | ci->array = array; | |
1887 | ci->begin = begin; | |
1888 | ci->total = total; | |
f268980d | 1889 | ci->last_index = last_index; |
a4bcff5b LP |
1890 | } |
1891 | ||
f268980d LP |
1892 | static int generic_array_get( |
1893 | JournalFile *f, | |
1894 | uint64_t first, | |
1895 | uint64_t i, | |
1896 | Object **ret, uint64_t *offset) { | |
de190aef | 1897 | |
cec736d2 | 1898 | Object *o; |
a4bcff5b | 1899 | uint64_t p = 0, a, t = 0; |
cec736d2 | 1900 | int r; |
a4bcff5b | 1901 | ChainCacheItem *ci; |
cec736d2 LP |
1902 | |
1903 | assert(f); | |
1904 | ||
de190aef | 1905 | a = first; |
a4bcff5b LP |
1906 | |
1907 | /* Try the chain cache first */ | |
4743015d | 1908 | ci = ordered_hashmap_get(f->chain_cache, &first); |
a4bcff5b LP |
1909 | if (ci && i > ci->total) { |
1910 | a = ci->array; | |
1911 | i -= ci->total; | |
1912 | t = ci->total; | |
1913 | } | |
1914 | ||
de190aef | 1915 | while (a > 0) { |
a4bcff5b | 1916 | uint64_t k; |
cec736d2 | 1917 | |
de190aef LP |
1918 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); |
1919 | if (r < 0) | |
1920 | return r; | |
cec736d2 | 1921 | |
a4bcff5b LP |
1922 | k = journal_file_entry_array_n_items(o); |
1923 | if (i < k) { | |
de190aef | 1924 | p = le64toh(o->entry_array.items[i]); |
a4bcff5b | 1925 | goto found; |
cec736d2 LP |
1926 | } |
1927 | ||
a4bcff5b LP |
1928 | i -= k; |
1929 | t += k; | |
de190aef LP |
1930 | a = le64toh(o->entry_array.next_entry_array_offset); |
1931 | } | |
1932 | ||
a4bcff5b LP |
1933 | return 0; |
1934 | ||
1935 | found: | |
1936 | /* Let's cache this item for the next invocation */ | |
af13a6b0 | 1937 | chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i); |
de190aef LP |
1938 | |
1939 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
1940 | if (r < 0) | |
1941 | return r; | |
1942 | ||
1943 | if (ret) | |
1944 | *ret = o; | |
1945 | ||
1946 | if (offset) | |
1947 | *offset = p; | |
1948 | ||
1949 | return 1; | |
1950 | } | |
1951 | ||
f268980d LP |
1952 | static int generic_array_get_plus_one( |
1953 | JournalFile *f, | |
1954 | uint64_t extra, | |
1955 | uint64_t first, | |
1956 | uint64_t i, | |
1957 | Object **ret, uint64_t *offset) { | |
de190aef LP |
1958 | |
1959 | Object *o; | |
1960 | ||
1961 | assert(f); | |
1962 | ||
1963 | if (i == 0) { | |
1964 | int r; | |
1965 | ||
1966 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
cec736d2 LP |
1967 | if (r < 0) |
1968 | return r; | |
1969 | ||
de190aef LP |
1970 | if (ret) |
1971 | *ret = o; | |
cec736d2 | 1972 | |
de190aef LP |
1973 | if (offset) |
1974 | *offset = extra; | |
cec736d2 | 1975 | |
de190aef | 1976 | return 1; |
cec736d2 LP |
1977 | } |
1978 | ||
de190aef LP |
1979 | return generic_array_get(f, first, i-1, ret, offset); |
1980 | } | |
cec736d2 | 1981 | |
de190aef LP |
1982 | enum { |
1983 | TEST_FOUND, | |
1984 | TEST_LEFT, | |
1985 | TEST_RIGHT | |
1986 | }; | |
cec736d2 | 1987 | |
f268980d LP |
1988 | static int generic_array_bisect( |
1989 | JournalFile *f, | |
1990 | uint64_t first, | |
1991 | uint64_t n, | |
1992 | uint64_t needle, | |
1993 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
1994 | direction_t direction, | |
1995 | Object **ret, | |
1996 | uint64_t *offset, | |
1997 | uint64_t *idx) { | |
1998 | ||
1999 | uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1; | |
de190aef LP |
2000 | bool subtract_one = false; |
2001 | Object *o, *array = NULL; | |
2002 | int r; | |
a4bcff5b | 2003 | ChainCacheItem *ci; |
cec736d2 | 2004 | |
de190aef LP |
2005 | assert(f); |
2006 | assert(test_object); | |
cec736d2 | 2007 | |
a4bcff5b | 2008 | /* Start with the first array in the chain */ |
de190aef | 2009 | a = first; |
a4bcff5b | 2010 | |
4743015d | 2011 | ci = ordered_hashmap_get(f->chain_cache, &first); |
a4bcff5b LP |
2012 | if (ci && n > ci->total) { |
2013 | /* Ah, we have iterated this bisection array chain | |
2014 | * previously! Let's see if we can skip ahead in the | |
2015 | * chain, as far as the last time. But we can't jump | |
2016 | * backwards in the chain, so let's check that | |
2017 | * first. */ | |
2018 | ||
2019 | r = test_object(f, ci->begin, needle); | |
2020 | if (r < 0) | |
2021 | return r; | |
2022 | ||
2023 | if (r == TEST_LEFT) { | |
f268980d | 2024 | /* OK, what we are looking for is right of the |
a4bcff5b LP |
2025 | * begin of this EntryArray, so let's jump |
2026 | * straight to previously cached array in the | |
2027 | * chain */ | |
2028 | ||
2029 | a = ci->array; | |
2030 | n -= ci->total; | |
2031 | t = ci->total; | |
f268980d | 2032 | last_index = ci->last_index; |
a4bcff5b LP |
2033 | } |
2034 | } | |
2035 | ||
de190aef LP |
2036 | while (a > 0) { |
2037 | uint64_t left, right, k, lp; | |
2038 | ||
2039 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); | |
cec736d2 LP |
2040 | if (r < 0) |
2041 | return r; | |
2042 | ||
de190aef LP |
2043 | k = journal_file_entry_array_n_items(array); |
2044 | right = MIN(k, n); | |
2045 | if (right <= 0) | |
2046 | return 0; | |
cec736d2 | 2047 | |
de190aef LP |
2048 | i = right - 1; |
2049 | lp = p = le64toh(array->entry_array.items[i]); | |
2050 | if (p <= 0) | |
bee6a291 LP |
2051 | r = -EBADMSG; |
2052 | else | |
2053 | r = test_object(f, p, needle); | |
2054 | if (r == -EBADMSG) { | |
2055 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)"); | |
2056 | n = i; | |
2057 | continue; | |
2058 | } | |
de190aef LP |
2059 | if (r < 0) |
2060 | return r; | |
cec736d2 | 2061 | |
de190aef LP |
2062 | if (r == TEST_FOUND) |
2063 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2064 | ||
2065 | if (r == TEST_RIGHT) { | |
2066 | left = 0; | |
2067 | right -= 1; | |
f268980d LP |
2068 | |
2069 | if (last_index != (uint64_t) -1) { | |
2070 | assert(last_index <= right); | |
2071 | ||
2072 | /* If we cached the last index we | |
2073 | * looked at, let's try to not to jump | |
2074 | * too wildly around and see if we can | |
2075 | * limit the range to look at early to | |
2076 | * the immediate neighbors of the last | |
2077 | * index we looked at. */ | |
2078 | ||
2079 | if (last_index > 0) { | |
2080 | uint64_t x = last_index - 1; | |
2081 | ||
2082 | p = le64toh(array->entry_array.items[x]); | |
2083 | if (p <= 0) | |
2084 | return -EBADMSG; | |
2085 | ||
2086 | r = test_object(f, p, needle); | |
2087 | if (r < 0) | |
2088 | return r; | |
2089 | ||
2090 | if (r == TEST_FOUND) | |
2091 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2092 | ||
2093 | if (r == TEST_RIGHT) | |
2094 | right = x; | |
2095 | else | |
2096 | left = x + 1; | |
2097 | } | |
2098 | ||
2099 | if (last_index < right) { | |
2100 | uint64_t y = last_index + 1; | |
2101 | ||
2102 | p = le64toh(array->entry_array.items[y]); | |
2103 | if (p <= 0) | |
2104 | return -EBADMSG; | |
2105 | ||
2106 | r = test_object(f, p, needle); | |
2107 | if (r < 0) | |
2108 | return r; | |
2109 | ||
2110 | if (r == TEST_FOUND) | |
2111 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2112 | ||
2113 | if (r == TEST_RIGHT) | |
2114 | right = y; | |
2115 | else | |
2116 | left = y + 1; | |
2117 | } | |
f268980d LP |
2118 | } |
2119 | ||
de190aef LP |
2120 | for (;;) { |
2121 | if (left == right) { | |
2122 | if (direction == DIRECTION_UP) | |
2123 | subtract_one = true; | |
2124 | ||
2125 | i = left; | |
2126 | goto found; | |
2127 | } | |
2128 | ||
2129 | assert(left < right); | |
de190aef | 2130 | i = (left + right) / 2; |
f268980d | 2131 | |
de190aef LP |
2132 | p = le64toh(array->entry_array.items[i]); |
2133 | if (p <= 0) | |
bee6a291 LP |
2134 | r = -EBADMSG; |
2135 | else | |
2136 | r = test_object(f, p, needle); | |
2137 | if (r == -EBADMSG) { | |
2138 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)"); | |
2139 | right = n = i; | |
2140 | continue; | |
2141 | } | |
de190aef LP |
2142 | if (r < 0) |
2143 | return r; | |
cec736d2 | 2144 | |
de190aef LP |
2145 | if (r == TEST_FOUND) |
2146 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2147 | ||
2148 | if (r == TEST_RIGHT) | |
2149 | right = i; | |
2150 | else | |
2151 | left = i + 1; | |
2152 | } | |
2153 | } | |
2154 | ||
2173cbf8 | 2155 | if (k >= n) { |
cbdca852 LP |
2156 | if (direction == DIRECTION_UP) { |
2157 | i = n; | |
2158 | subtract_one = true; | |
2159 | goto found; | |
2160 | } | |
2161 | ||
cec736d2 | 2162 | return 0; |
cbdca852 | 2163 | } |
cec736d2 | 2164 | |
de190aef LP |
2165 | last_p = lp; |
2166 | ||
2167 | n -= k; | |
2168 | t += k; | |
f268980d | 2169 | last_index = (uint64_t) -1; |
de190aef | 2170 | a = le64toh(array->entry_array.next_entry_array_offset); |
cec736d2 LP |
2171 | } |
2172 | ||
2173 | return 0; | |
de190aef LP |
2174 | |
2175 | found: | |
2176 | if (subtract_one && t == 0 && i == 0) | |
2177 | return 0; | |
2178 | ||
a4bcff5b | 2179 | /* Let's cache this item for the next invocation */ |
af13a6b0 | 2180 | 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 | 2181 | |
de190aef LP |
2182 | if (subtract_one && i == 0) |
2183 | p = last_p; | |
2184 | else if (subtract_one) | |
2185 | p = le64toh(array->entry_array.items[i-1]); | |
2186 | else | |
2187 | p = le64toh(array->entry_array.items[i]); | |
2188 | ||
2189 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2190 | if (r < 0) | |
2191 | return r; | |
2192 | ||
2193 | if (ret) | |
2194 | *ret = o; | |
2195 | ||
2196 | if (offset) | |
2197 | *offset = p; | |
2198 | ||
2199 | if (idx) | |
cbdca852 | 2200 | *idx = t + i + (subtract_one ? -1 : 0); |
de190aef LP |
2201 | |
2202 | return 1; | |
cec736d2 LP |
2203 | } |
2204 | ||
f268980d LP |
2205 | static int generic_array_bisect_plus_one( |
2206 | JournalFile *f, | |
2207 | uint64_t extra, | |
2208 | uint64_t first, | |
2209 | uint64_t n, | |
2210 | uint64_t needle, | |
2211 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
2212 | direction_t direction, | |
2213 | Object **ret, | |
2214 | uint64_t *offset, | |
2215 | uint64_t *idx) { | |
de190aef | 2216 | |
cec736d2 | 2217 | int r; |
cbdca852 LP |
2218 | bool step_back = false; |
2219 | Object *o; | |
cec736d2 LP |
2220 | |
2221 | assert(f); | |
de190aef | 2222 | assert(test_object); |
cec736d2 | 2223 | |
de190aef LP |
2224 | if (n <= 0) |
2225 | return 0; | |
cec736d2 | 2226 | |
de190aef LP |
2227 | /* This bisects the array in object 'first', but first checks |
2228 | * an extra */ | |
de190aef LP |
2229 | r = test_object(f, extra, needle); |
2230 | if (r < 0) | |
2231 | return r; | |
a536e261 LP |
2232 | |
2233 | if (r == TEST_FOUND) | |
2234 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2235 | ||
cbdca852 LP |
2236 | /* if we are looking with DIRECTION_UP then we need to first |
2237 | see if in the actual array there is a matching entry, and | |
2238 | return the last one of that. But if there isn't any we need | |
2239 | to return this one. Hence remember this, and return it | |
2240 | below. */ | |
2241 | if (r == TEST_LEFT) | |
2242 | step_back = direction == DIRECTION_UP; | |
de190aef | 2243 | |
cbdca852 LP |
2244 | if (r == TEST_RIGHT) { |
2245 | if (direction == DIRECTION_DOWN) | |
2246 | goto found; | |
2247 | else | |
2248 | return 0; | |
a536e261 | 2249 | } |
cec736d2 | 2250 | |
de190aef LP |
2251 | r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx); |
2252 | ||
cbdca852 LP |
2253 | if (r == 0 && step_back) |
2254 | goto found; | |
2255 | ||
ecf68b1d | 2256 | if (r > 0 && idx) |
313cefa1 | 2257 | (*idx)++; |
de190aef LP |
2258 | |
2259 | return r; | |
cbdca852 LP |
2260 | |
2261 | found: | |
2262 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
2263 | if (r < 0) | |
2264 | return r; | |
2265 | ||
2266 | if (ret) | |
2267 | *ret = o; | |
2268 | ||
2269 | if (offset) | |
2270 | *offset = extra; | |
2271 | ||
2272 | if (idx) | |
2273 | *idx = 0; | |
2274 | ||
2275 | return 1; | |
2276 | } | |
2277 | ||
44a6b1b6 | 2278 | _pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { |
cbdca852 LP |
2279 | assert(f); |
2280 | assert(p > 0); | |
2281 | ||
2282 | if (p == needle) | |
2283 | return TEST_FOUND; | |
2284 | else if (p < needle) | |
2285 | return TEST_LEFT; | |
2286 | else | |
2287 | return TEST_RIGHT; | |
2288 | } | |
2289 | ||
de190aef LP |
2290 | static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { |
2291 | Object *o; | |
2292 | int r; | |
2293 | ||
2294 | assert(f); | |
2295 | assert(p > 0); | |
2296 | ||
2297 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
cec736d2 LP |
2298 | if (r < 0) |
2299 | return r; | |
2300 | ||
de190aef LP |
2301 | if (le64toh(o->entry.seqnum) == needle) |
2302 | return TEST_FOUND; | |
2303 | else if (le64toh(o->entry.seqnum) < needle) | |
2304 | return TEST_LEFT; | |
2305 | else | |
2306 | return TEST_RIGHT; | |
2307 | } | |
cec736d2 | 2308 | |
de190aef LP |
2309 | int journal_file_move_to_entry_by_seqnum( |
2310 | JournalFile *f, | |
2311 | uint64_t seqnum, | |
2312 | direction_t direction, | |
2313 | Object **ret, | |
2314 | uint64_t *offset) { | |
c88cc6af VC |
2315 | assert(f); |
2316 | assert(f->header); | |
de190aef LP |
2317 | |
2318 | return generic_array_bisect(f, | |
2319 | le64toh(f->header->entry_array_offset), | |
2320 | le64toh(f->header->n_entries), | |
2321 | seqnum, | |
2322 | test_object_seqnum, | |
2323 | direction, | |
2324 | ret, offset, NULL); | |
2325 | } | |
cec736d2 | 2326 | |
de190aef LP |
2327 | static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { |
2328 | Object *o; | |
2329 | int r; | |
2330 | ||
2331 | assert(f); | |
2332 | assert(p > 0); | |
2333 | ||
2334 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2335 | if (r < 0) | |
2336 | return r; | |
2337 | ||
2338 | if (le64toh(o->entry.realtime) == needle) | |
2339 | return TEST_FOUND; | |
2340 | else if (le64toh(o->entry.realtime) < needle) | |
2341 | return TEST_LEFT; | |
2342 | else | |
2343 | return TEST_RIGHT; | |
cec736d2 LP |
2344 | } |
2345 | ||
de190aef LP |
2346 | int journal_file_move_to_entry_by_realtime( |
2347 | JournalFile *f, | |
2348 | uint64_t realtime, | |
2349 | direction_t direction, | |
2350 | Object **ret, | |
2351 | uint64_t *offset) { | |
c88cc6af VC |
2352 | assert(f); |
2353 | assert(f->header); | |
de190aef LP |
2354 | |
2355 | return generic_array_bisect(f, | |
2356 | le64toh(f->header->entry_array_offset), | |
2357 | le64toh(f->header->n_entries), | |
2358 | realtime, | |
2359 | test_object_realtime, | |
2360 | direction, | |
2361 | ret, offset, NULL); | |
2362 | } | |
2363 | ||
2364 | static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { | |
2365 | Object *o; | |
2366 | int r; | |
2367 | ||
2368 | assert(f); | |
2369 | assert(p > 0); | |
2370 | ||
2371 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2372 | if (r < 0) | |
2373 | return r; | |
2374 | ||
2375 | if (le64toh(o->entry.monotonic) == needle) | |
2376 | return TEST_FOUND; | |
2377 | else if (le64toh(o->entry.monotonic) < needle) | |
2378 | return TEST_LEFT; | |
2379 | else | |
2380 | return TEST_RIGHT; | |
2381 | } | |
2382 | ||
2a560338 | 2383 | static int find_data_object_by_boot_id( |
47838ab3 ZJS |
2384 | JournalFile *f, |
2385 | sd_id128_t boot_id, | |
2386 | Object **o, | |
2387 | uint64_t *b) { | |
2a560338 | 2388 | |
47838ab3 ZJS |
2389 | char t[sizeof("_BOOT_ID=")-1 + 32 + 1] = "_BOOT_ID="; |
2390 | ||
2391 | sd_id128_to_string(boot_id, t + 9); | |
2392 | return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b); | |
2393 | } | |
2394 | ||
de190aef LP |
2395 | int journal_file_move_to_entry_by_monotonic( |
2396 | JournalFile *f, | |
2397 | sd_id128_t boot_id, | |
2398 | uint64_t monotonic, | |
2399 | direction_t direction, | |
2400 | Object **ret, | |
2401 | uint64_t *offset) { | |
2402 | ||
de190aef LP |
2403 | Object *o; |
2404 | int r; | |
2405 | ||
cbdca852 | 2406 | assert(f); |
de190aef | 2407 | |
47838ab3 | 2408 | r = find_data_object_by_boot_id(f, boot_id, &o, NULL); |
de190aef LP |
2409 | if (r < 0) |
2410 | return r; | |
cbdca852 | 2411 | if (r == 0) |
de190aef LP |
2412 | return -ENOENT; |
2413 | ||
2414 | return generic_array_bisect_plus_one(f, | |
2415 | le64toh(o->data.entry_offset), | |
2416 | le64toh(o->data.entry_array_offset), | |
2417 | le64toh(o->data.n_entries), | |
2418 | monotonic, | |
2419 | test_object_monotonic, | |
2420 | direction, | |
2421 | ret, offset, NULL); | |
2422 | } | |
2423 | ||
1fc605b0 | 2424 | void journal_file_reset_location(JournalFile *f) { |
6573ef05 | 2425 | f->location_type = LOCATION_HEAD; |
1fc605b0 | 2426 | f->current_offset = 0; |
6573ef05 MS |
2427 | f->current_seqnum = 0; |
2428 | f->current_realtime = 0; | |
2429 | f->current_monotonic = 0; | |
2430 | zero(f->current_boot_id); | |
2431 | f->current_xor_hash = 0; | |
2432 | } | |
2433 | ||
950c07d4 | 2434 | void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) { |
6573ef05 MS |
2435 | f->location_type = LOCATION_SEEK; |
2436 | f->current_offset = offset; | |
2437 | f->current_seqnum = le64toh(o->entry.seqnum); | |
2438 | f->current_realtime = le64toh(o->entry.realtime); | |
2439 | f->current_monotonic = le64toh(o->entry.monotonic); | |
2440 | f->current_boot_id = o->entry.boot_id; | |
2441 | f->current_xor_hash = le64toh(o->entry.xor_hash); | |
1fc605b0 MS |
2442 | } |
2443 | ||
d8ae66d7 MS |
2444 | int journal_file_compare_locations(JournalFile *af, JournalFile *bf) { |
2445 | assert(af); | |
c88cc6af | 2446 | assert(af->header); |
d8ae66d7 | 2447 | assert(bf); |
c88cc6af | 2448 | assert(bf->header); |
d8ae66d7 MS |
2449 | assert(af->location_type == LOCATION_SEEK); |
2450 | assert(bf->location_type == LOCATION_SEEK); | |
2451 | ||
2452 | /* If contents and timestamps match, these entries are | |
2453 | * identical, even if the seqnum does not match */ | |
2454 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) && | |
2455 | af->current_monotonic == bf->current_monotonic && | |
2456 | af->current_realtime == bf->current_realtime && | |
2457 | af->current_xor_hash == bf->current_xor_hash) | |
2458 | return 0; | |
2459 | ||
2460 | if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) { | |
2461 | ||
2462 | /* If this is from the same seqnum source, compare | |
2463 | * seqnums */ | |
2464 | if (af->current_seqnum < bf->current_seqnum) | |
2465 | return -1; | |
2466 | if (af->current_seqnum > bf->current_seqnum) | |
2467 | return 1; | |
2468 | ||
2469 | /* Wow! This is weird, different data but the same | |
2470 | * seqnums? Something is borked, but let's make the | |
2471 | * best of it and compare by time. */ | |
2472 | } | |
2473 | ||
2474 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) { | |
2475 | ||
2476 | /* If the boot id matches, compare monotonic time */ | |
2477 | if (af->current_monotonic < bf->current_monotonic) | |
2478 | return -1; | |
2479 | if (af->current_monotonic > bf->current_monotonic) | |
2480 | return 1; | |
2481 | } | |
2482 | ||
2483 | /* Otherwise, compare UTC time */ | |
2484 | if (af->current_realtime < bf->current_realtime) | |
2485 | return -1; | |
2486 | if (af->current_realtime > bf->current_realtime) | |
2487 | return 1; | |
2488 | ||
2489 | /* Finally, compare by contents */ | |
2490 | if (af->current_xor_hash < bf->current_xor_hash) | |
2491 | return -1; | |
2492 | if (af->current_xor_hash > bf->current_xor_hash) | |
2493 | return 1; | |
2494 | ||
2495 | return 0; | |
2496 | } | |
2497 | ||
aa598ba5 LP |
2498 | static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) { |
2499 | ||
2500 | /* Increase or decrease the specified index, in the right direction. */ | |
2501 | ||
2502 | if (direction == DIRECTION_DOWN) { | |
2503 | if (*i >= n - 1) | |
2504 | return 0; | |
2505 | ||
2506 | (*i) ++; | |
2507 | } else { | |
2508 | if (*i <= 0) | |
2509 | return 0; | |
2510 | ||
2511 | (*i) --; | |
2512 | } | |
2513 | ||
2514 | return 1; | |
2515 | } | |
2516 | ||
b6da4ed0 LP |
2517 | static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) { |
2518 | ||
2519 | /* Consider it an error if any of the two offsets is uninitialized */ | |
2520 | if (old_offset == 0 || new_offset == 0) | |
2521 | return false; | |
2522 | ||
2523 | /* If we go down, the new offset must be larger than the old one. */ | |
2524 | return direction == DIRECTION_DOWN ? | |
2525 | new_offset > old_offset : | |
2526 | new_offset < old_offset; | |
2527 | } | |
2528 | ||
de190aef LP |
2529 | int journal_file_next_entry( |
2530 | JournalFile *f, | |
f534928a | 2531 | uint64_t p, |
de190aef LP |
2532 | direction_t direction, |
2533 | Object **ret, uint64_t *offset) { | |
2534 | ||
fb099c8d | 2535 | uint64_t i, n, ofs; |
cec736d2 LP |
2536 | int r; |
2537 | ||
2538 | assert(f); | |
c88cc6af | 2539 | assert(f->header); |
de190aef LP |
2540 | |
2541 | n = le64toh(f->header->n_entries); | |
2542 | if (n <= 0) | |
2543 | return 0; | |
cec736d2 | 2544 | |
f534928a | 2545 | if (p == 0) |
de190aef | 2546 | i = direction == DIRECTION_DOWN ? 0 : n - 1; |
cec736d2 | 2547 | else { |
de190aef LP |
2548 | r = generic_array_bisect(f, |
2549 | le64toh(f->header->entry_array_offset), | |
2550 | le64toh(f->header->n_entries), | |
2551 | p, | |
2552 | test_object_offset, | |
2553 | DIRECTION_DOWN, | |
2554 | NULL, NULL, | |
2555 | &i); | |
2556 | if (r <= 0) | |
2557 | return r; | |
2558 | ||
aa598ba5 LP |
2559 | r = bump_array_index(&i, direction, n); |
2560 | if (r <= 0) | |
2561 | return r; | |
cec736d2 LP |
2562 | } |
2563 | ||
de190aef | 2564 | /* And jump to it */ |
989793d3 LP |
2565 | for (;;) { |
2566 | r = generic_array_get(f, | |
2567 | le64toh(f->header->entry_array_offset), | |
2568 | i, | |
2569 | ret, &ofs); | |
2570 | if (r > 0) | |
2571 | break; | |
2572 | if (r != -EBADMSG) | |
2573 | return r; | |
2574 | ||
2575 | /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if | |
2576 | * the next one might work for us instead. */ | |
2577 | log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i); | |
2578 | ||
2579 | r = bump_array_index(&i, direction, n); | |
2580 | if (r <= 0) | |
2581 | return r; | |
caeab8f6 | 2582 | } |
fb099c8d | 2583 | |
b6da4ed0 LP |
2584 | /* Ensure our array is properly ordered. */ |
2585 | if (p > 0 && !check_properly_ordered(ofs, p, direction)) { | |
2586 | log_debug("%s: entry array not properly ordered at entry %" PRIu64, f->path, i); | |
fb099c8d ZJS |
2587 | return -EBADMSG; |
2588 | } | |
2589 | ||
2590 | if (offset) | |
2591 | *offset = ofs; | |
2592 | ||
2593 | return 1; | |
de190aef | 2594 | } |
cec736d2 | 2595 | |
de190aef LP |
2596 | int journal_file_next_entry_for_data( |
2597 | JournalFile *f, | |
2598 | Object *o, uint64_t p, | |
2599 | uint64_t data_offset, | |
2600 | direction_t direction, | |
2601 | Object **ret, uint64_t *offset) { | |
2602 | ||
ded5034e | 2603 | uint64_t i, n, ofs; |
de190aef | 2604 | Object *d; |
989793d3 | 2605 | int r; |
cec736d2 LP |
2606 | |
2607 | assert(f); | |
de190aef | 2608 | assert(p > 0 || !o); |
cec736d2 | 2609 | |
de190aef | 2610 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
466ccd92 | 2611 | if (r < 0) |
de190aef | 2612 | return r; |
cec736d2 | 2613 | |
de190aef LP |
2614 | n = le64toh(d->data.n_entries); |
2615 | if (n <= 0) | |
2616 | return n; | |
cec736d2 | 2617 | |
de190aef LP |
2618 | if (!o) |
2619 | i = direction == DIRECTION_DOWN ? 0 : n - 1; | |
2620 | else { | |
2621 | if (o->object.type != OBJECT_ENTRY) | |
2622 | return -EINVAL; | |
cec736d2 | 2623 | |
de190aef LP |
2624 | r = generic_array_bisect_plus_one(f, |
2625 | le64toh(d->data.entry_offset), | |
2626 | le64toh(d->data.entry_array_offset), | |
2627 | le64toh(d->data.n_entries), | |
2628 | p, | |
2629 | test_object_offset, | |
2630 | DIRECTION_DOWN, | |
2631 | NULL, NULL, | |
2632 | &i); | |
2633 | ||
2634 | if (r <= 0) | |
cec736d2 LP |
2635 | return r; |
2636 | ||
aa598ba5 LP |
2637 | r = bump_array_index(&i, direction, n); |
2638 | if (r <= 0) | |
2639 | return r; | |
de190aef | 2640 | } |
cec736d2 | 2641 | |
989793d3 LP |
2642 | for (;;) { |
2643 | r = generic_array_get_plus_one(f, | |
2644 | le64toh(d->data.entry_offset), | |
2645 | le64toh(d->data.entry_array_offset), | |
2646 | i, | |
2647 | ret, &ofs); | |
2648 | if (r > 0) | |
2649 | break; | |
2650 | if (r != -EBADMSG) | |
2651 | return r; | |
2652 | ||
2653 | log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i); | |
2654 | ||
2655 | r = bump_array_index(&i, direction, n); | |
2656 | if (r <= 0) | |
2657 | return r; | |
2658 | } | |
ded5034e LP |
2659 | |
2660 | /* Ensure our array is properly ordered. */ | |
2661 | if (p > 0 && check_properly_ordered(ofs, p, direction)) { | |
2662 | log_debug("%s data entry array not properly ordered at entry %" PRIu64, f->path, i); | |
2663 | return -EBADMSG; | |
2664 | } | |
2665 | ||
2666 | if (offset) | |
2667 | *offset = ofs; | |
2668 | ||
2669 | return 1; | |
de190aef | 2670 | } |
cec736d2 | 2671 | |
cbdca852 LP |
2672 | int journal_file_move_to_entry_by_offset_for_data( |
2673 | JournalFile *f, | |
2674 | uint64_t data_offset, | |
2675 | uint64_t p, | |
2676 | direction_t direction, | |
2677 | Object **ret, uint64_t *offset) { | |
2678 | ||
2679 | int r; | |
2680 | Object *d; | |
2681 | ||
2682 | assert(f); | |
2683 | ||
2684 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
2685 | if (r < 0) | |
2686 | return r; | |
2687 | ||
2688 | return generic_array_bisect_plus_one(f, | |
2689 | le64toh(d->data.entry_offset), | |
2690 | le64toh(d->data.entry_array_offset), | |
2691 | le64toh(d->data.n_entries), | |
2692 | p, | |
2693 | test_object_offset, | |
2694 | direction, | |
2695 | ret, offset, NULL); | |
2696 | } | |
2697 | ||
2698 | int journal_file_move_to_entry_by_monotonic_for_data( | |
2699 | JournalFile *f, | |
2700 | uint64_t data_offset, | |
2701 | sd_id128_t boot_id, | |
2702 | uint64_t monotonic, | |
2703 | direction_t direction, | |
2704 | Object **ret, uint64_t *offset) { | |
2705 | ||
cbdca852 LP |
2706 | Object *o, *d; |
2707 | int r; | |
2708 | uint64_t b, z; | |
2709 | ||
2710 | assert(f); | |
2711 | ||
2712 | /* First, seek by time */ | |
47838ab3 | 2713 | r = find_data_object_by_boot_id(f, boot_id, &o, &b); |
cbdca852 LP |
2714 | if (r < 0) |
2715 | return r; | |
2716 | if (r == 0) | |
2717 | return -ENOENT; | |
2718 | ||
2719 | r = generic_array_bisect_plus_one(f, | |
2720 | le64toh(o->data.entry_offset), | |
2721 | le64toh(o->data.entry_array_offset), | |
2722 | le64toh(o->data.n_entries), | |
2723 | monotonic, | |
2724 | test_object_monotonic, | |
2725 | direction, | |
2726 | NULL, &z, NULL); | |
2727 | if (r <= 0) | |
2728 | return r; | |
2729 | ||
2730 | /* And now, continue seeking until we find an entry that | |
2731 | * exists in both bisection arrays */ | |
2732 | ||
2733 | for (;;) { | |
2734 | Object *qo; | |
2735 | uint64_t p, q; | |
2736 | ||
2737 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
2738 | if (r < 0) | |
2739 | return r; | |
2740 | ||
2741 | r = generic_array_bisect_plus_one(f, | |
2742 | le64toh(d->data.entry_offset), | |
2743 | le64toh(d->data.entry_array_offset), | |
2744 | le64toh(d->data.n_entries), | |
2745 | z, | |
2746 | test_object_offset, | |
2747 | direction, | |
2748 | NULL, &p, NULL); | |
2749 | if (r <= 0) | |
2750 | return r; | |
2751 | ||
2752 | r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); | |
2753 | if (r < 0) | |
2754 | return r; | |
2755 | ||
2756 | r = generic_array_bisect_plus_one(f, | |
2757 | le64toh(o->data.entry_offset), | |
2758 | le64toh(o->data.entry_array_offset), | |
2759 | le64toh(o->data.n_entries), | |
2760 | p, | |
2761 | test_object_offset, | |
2762 | direction, | |
2763 | &qo, &q, NULL); | |
2764 | ||
2765 | if (r <= 0) | |
2766 | return r; | |
2767 | ||
2768 | if (p == q) { | |
2769 | if (ret) | |
2770 | *ret = qo; | |
2771 | if (offset) | |
2772 | *offset = q; | |
2773 | ||
2774 | return 1; | |
2775 | } | |
2776 | ||
2777 | z = q; | |
2778 | } | |
cbdca852 LP |
2779 | } |
2780 | ||
de190aef LP |
2781 | int journal_file_move_to_entry_by_seqnum_for_data( |
2782 | JournalFile *f, | |
2783 | uint64_t data_offset, | |
2784 | uint64_t seqnum, | |
2785 | direction_t direction, | |
2786 | Object **ret, uint64_t *offset) { | |
cec736d2 | 2787 | |
de190aef LP |
2788 | Object *d; |
2789 | int r; | |
cec736d2 | 2790 | |
91a31dde LP |
2791 | assert(f); |
2792 | ||
de190aef | 2793 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 2794 | if (r < 0) |
de190aef | 2795 | return r; |
cec736d2 | 2796 | |
de190aef LP |
2797 | return generic_array_bisect_plus_one(f, |
2798 | le64toh(d->data.entry_offset), | |
2799 | le64toh(d->data.entry_array_offset), | |
2800 | le64toh(d->data.n_entries), | |
2801 | seqnum, | |
2802 | test_object_seqnum, | |
2803 | direction, | |
2804 | ret, offset, NULL); | |
2805 | } | |
cec736d2 | 2806 | |
de190aef LP |
2807 | int journal_file_move_to_entry_by_realtime_for_data( |
2808 | JournalFile *f, | |
2809 | uint64_t data_offset, | |
2810 | uint64_t realtime, | |
2811 | direction_t direction, | |
2812 | Object **ret, uint64_t *offset) { | |
2813 | ||
2814 | Object *d; | |
2815 | int r; | |
2816 | ||
91a31dde LP |
2817 | assert(f); |
2818 | ||
de190aef | 2819 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 2820 | if (r < 0) |
de190aef LP |
2821 | return r; |
2822 | ||
2823 | return generic_array_bisect_plus_one(f, | |
2824 | le64toh(d->data.entry_offset), | |
2825 | le64toh(d->data.entry_array_offset), | |
2826 | le64toh(d->data.n_entries), | |
2827 | realtime, | |
2828 | test_object_realtime, | |
2829 | direction, | |
2830 | ret, offset, NULL); | |
cec736d2 LP |
2831 | } |
2832 | ||
0284adc6 | 2833 | void journal_file_dump(JournalFile *f) { |
7560fffc | 2834 | Object *o; |
7560fffc | 2835 | int r; |
0284adc6 | 2836 | uint64_t p; |
7560fffc LP |
2837 | |
2838 | assert(f); | |
c88cc6af | 2839 | assert(f->header); |
7560fffc | 2840 | |
0284adc6 | 2841 | journal_file_print_header(f); |
7560fffc | 2842 | |
0284adc6 LP |
2843 | p = le64toh(f->header->header_size); |
2844 | while (p != 0) { | |
d05089d8 | 2845 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); |
0284adc6 LP |
2846 | if (r < 0) |
2847 | goto fail; | |
7560fffc | 2848 | |
0284adc6 | 2849 | switch (o->object.type) { |
d98cc1f2 | 2850 | |
0284adc6 LP |
2851 | case OBJECT_UNUSED: |
2852 | printf("Type: OBJECT_UNUSED\n"); | |
2853 | break; | |
d98cc1f2 | 2854 | |
0284adc6 LP |
2855 | case OBJECT_DATA: |
2856 | printf("Type: OBJECT_DATA\n"); | |
2857 | break; | |
7560fffc | 2858 | |
3c1668da LP |
2859 | case OBJECT_FIELD: |
2860 | printf("Type: OBJECT_FIELD\n"); | |
2861 | break; | |
2862 | ||
0284adc6 | 2863 | case OBJECT_ENTRY: |
507f22bd ZJS |
2864 | printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n", |
2865 | le64toh(o->entry.seqnum), | |
2866 | le64toh(o->entry.monotonic), | |
2867 | le64toh(o->entry.realtime)); | |
0284adc6 | 2868 | break; |
7560fffc | 2869 | |
0284adc6 LP |
2870 | case OBJECT_FIELD_HASH_TABLE: |
2871 | printf("Type: OBJECT_FIELD_HASH_TABLE\n"); | |
2872 | break; | |
7560fffc | 2873 | |
0284adc6 LP |
2874 | case OBJECT_DATA_HASH_TABLE: |
2875 | printf("Type: OBJECT_DATA_HASH_TABLE\n"); | |
2876 | break; | |
7560fffc | 2877 | |
0284adc6 LP |
2878 | case OBJECT_ENTRY_ARRAY: |
2879 | printf("Type: OBJECT_ENTRY_ARRAY\n"); | |
2880 | break; | |
7560fffc | 2881 | |
0284adc6 | 2882 | case OBJECT_TAG: |
507f22bd ZJS |
2883 | printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n", |
2884 | le64toh(o->tag.seqnum), | |
2885 | le64toh(o->tag.epoch)); | |
0284adc6 | 2886 | break; |
3c1668da LP |
2887 | |
2888 | default: | |
8facc349 | 2889 | printf("Type: unknown (%i)\n", o->object.type); |
3c1668da | 2890 | break; |
0284adc6 | 2891 | } |
7560fffc | 2892 | |
d89c8fdf ZJS |
2893 | if (o->object.flags & OBJECT_COMPRESSION_MASK) |
2894 | printf("Flags: %s\n", | |
2895 | object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK)); | |
7560fffc | 2896 | |
0284adc6 LP |
2897 | if (p == le64toh(f->header->tail_object_offset)) |
2898 | p = 0; | |
2899 | else | |
2900 | p = p + ALIGN64(le64toh(o->object.size)); | |
2901 | } | |
7560fffc | 2902 | |
0284adc6 LP |
2903 | return; |
2904 | fail: | |
2905 | log_error("File corrupt"); | |
7560fffc LP |
2906 | } |
2907 | ||
718fe4b1 ZJS |
2908 | static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) { |
2909 | const char *x; | |
2910 | ||
2911 | x = format_timestamp(buf, l, t); | |
2912 | if (x) | |
2913 | return x; | |
2914 | return " --- "; | |
2915 | } | |
2916 | ||
0284adc6 | 2917 | void journal_file_print_header(JournalFile *f) { |
2765b7bb | 2918 | char a[33], b[33], c[33], d[33]; |
ed375beb | 2919 | char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX]; |
a1a03e30 LP |
2920 | struct stat st; |
2921 | char bytes[FORMAT_BYTES_MAX]; | |
7560fffc LP |
2922 | |
2923 | assert(f); | |
c88cc6af | 2924 | assert(f->header); |
7560fffc | 2925 | |
0284adc6 LP |
2926 | printf("File Path: %s\n" |
2927 | "File ID: %s\n" | |
2928 | "Machine ID: %s\n" | |
2929 | "Boot ID: %s\n" | |
2930 | "Sequential Number ID: %s\n" | |
2931 | "State: %s\n" | |
2932 | "Compatible Flags:%s%s\n" | |
d89c8fdf | 2933 | "Incompatible Flags:%s%s%s\n" |
507f22bd ZJS |
2934 | "Header size: %"PRIu64"\n" |
2935 | "Arena size: %"PRIu64"\n" | |
2936 | "Data Hash Table Size: %"PRIu64"\n" | |
2937 | "Field Hash Table Size: %"PRIu64"\n" | |
0284adc6 | 2938 | "Rotate Suggested: %s\n" |
0808b92f LP |
2939 | "Head Sequential Number: %"PRIu64" (%"PRIx64")\n" |
2940 | "Tail Sequential Number: %"PRIu64" (%"PRIx64")\n" | |
2941 | "Head Realtime Timestamp: %s (%"PRIx64")\n" | |
2942 | "Tail Realtime Timestamp: %s (%"PRIx64")\n" | |
2943 | "Tail Monotonic Timestamp: %s (%"PRIx64")\n" | |
507f22bd ZJS |
2944 | "Objects: %"PRIu64"\n" |
2945 | "Entry Objects: %"PRIu64"\n", | |
0284adc6 LP |
2946 | f->path, |
2947 | sd_id128_to_string(f->header->file_id, a), | |
2948 | sd_id128_to_string(f->header->machine_id, b), | |
2949 | sd_id128_to_string(f->header->boot_id, c), | |
2765b7bb | 2950 | sd_id128_to_string(f->header->seqnum_id, d), |
3223f44f LP |
2951 | f->header->state == STATE_OFFLINE ? "OFFLINE" : |
2952 | f->header->state == STATE_ONLINE ? "ONLINE" : | |
2953 | f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN", | |
8088cbd3 | 2954 | JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "", |
d89c8fdf ZJS |
2955 | (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "", |
2956 | JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "", | |
2957 | JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "", | |
2958 | (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "", | |
507f22bd ZJS |
2959 | le64toh(f->header->header_size), |
2960 | le64toh(f->header->arena_size), | |
2961 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
2962 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem), | |
fb0951b0 | 2963 | yes_no(journal_file_rotate_suggested(f, 0)), |
0808b92f LP |
2964 | le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum), |
2965 | le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum), | |
2966 | format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime), | |
2967 | format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime), | |
2968 | format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic), | |
507f22bd ZJS |
2969 | le64toh(f->header->n_objects), |
2970 | le64toh(f->header->n_entries)); | |
7560fffc | 2971 | |
0284adc6 | 2972 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
507f22bd | 2973 | printf("Data Objects: %"PRIu64"\n" |
0284adc6 | 2974 | "Data Hash Table Fill: %.1f%%\n", |
507f22bd | 2975 | le64toh(f->header->n_data), |
0284adc6 | 2976 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)))); |
7560fffc | 2977 | |
0284adc6 | 2978 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) |
507f22bd | 2979 | printf("Field Objects: %"PRIu64"\n" |
0284adc6 | 2980 | "Field Hash Table Fill: %.1f%%\n", |
507f22bd | 2981 | le64toh(f->header->n_fields), |
0284adc6 | 2982 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); |
3223f44f LP |
2983 | |
2984 | if (JOURNAL_HEADER_CONTAINS(f->header, n_tags)) | |
507f22bd ZJS |
2985 | printf("Tag Objects: %"PRIu64"\n", |
2986 | le64toh(f->header->n_tags)); | |
3223f44f | 2987 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
507f22bd ZJS |
2988 | printf("Entry Array Objects: %"PRIu64"\n", |
2989 | le64toh(f->header->n_entry_arrays)); | |
a1a03e30 LP |
2990 | |
2991 | if (fstat(f->fd, &st) >= 0) | |
59f448cf | 2992 | printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (uint64_t) st.st_blocks * 512ULL)); |
7560fffc LP |
2993 | } |
2994 | ||
fc68c929 LP |
2995 | static int journal_file_warn_btrfs(JournalFile *f) { |
2996 | unsigned attrs; | |
2997 | int r; | |
2998 | ||
2999 | assert(f); | |
3000 | ||
3001 | /* Before we write anything, check if the COW logic is turned | |
3002 | * off on btrfs. Given our write pattern that is quite | |
3003 | * unfriendly to COW file systems this should greatly improve | |
3004 | * performance on COW file systems, such as btrfs, at the | |
3005 | * expense of data integrity features (which shouldn't be too | |
3006 | * bad, given that we do our own checksumming). */ | |
3007 | ||
3008 | r = btrfs_is_filesystem(f->fd); | |
3009 | if (r < 0) | |
3010 | return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m"); | |
3011 | if (!r) | |
3012 | return 0; | |
3013 | ||
3014 | r = read_attr_fd(f->fd, &attrs); | |
3015 | if (r < 0) | |
3016 | return log_warning_errno(r, "Failed to read file attributes: %m"); | |
3017 | ||
3018 | if (attrs & FS_NOCOW_FL) { | |
3019 | log_debug("Detected btrfs file system with copy-on-write disabled, all is good."); | |
3020 | return 0; | |
3021 | } | |
3022 | ||
3023 | log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. " | |
3024 | "This is likely to slow down journal access substantially, please consider turning " | |
3025 | "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path); | |
3026 | ||
3027 | return 1; | |
3028 | } | |
3029 | ||
0284adc6 | 3030 | int journal_file_open( |
5d1ce257 | 3031 | int fd, |
0284adc6 LP |
3032 | const char *fname, |
3033 | int flags, | |
3034 | mode_t mode, | |
3035 | bool compress, | |
baed47c3 | 3036 | bool seal, |
0284adc6 LP |
3037 | JournalMetrics *metrics, |
3038 | MMapCache *mmap_cache, | |
b58c888f | 3039 | Set *deferred_closes, |
0284adc6 LP |
3040 | JournalFile *template, |
3041 | JournalFile **ret) { | |
7560fffc | 3042 | |
fa6ac760 | 3043 | bool newly_created = false; |
0284adc6 | 3044 | JournalFile *f; |
fa6ac760 | 3045 | void *h; |
0284adc6 | 3046 | int r; |
7560fffc | 3047 | |
0559d3a5 | 3048 | assert(ret); |
5d1ce257 | 3049 | assert(fd >= 0 || fname); |
7560fffc | 3050 | |
0284adc6 LP |
3051 | if ((flags & O_ACCMODE) != O_RDONLY && |
3052 | (flags & O_ACCMODE) != O_RDWR) | |
3053 | return -EINVAL; | |
7560fffc | 3054 | |
5d1ce257 LP |
3055 | if (fname) { |
3056 | if (!endswith(fname, ".journal") && | |
3057 | !endswith(fname, ".journal~")) | |
3058 | return -EINVAL; | |
3059 | } | |
7560fffc | 3060 | |
0284adc6 LP |
3061 | f = new0(JournalFile, 1); |
3062 | if (!f) | |
3063 | return -ENOMEM; | |
7560fffc | 3064 | |
5d1ce257 | 3065 | f->fd = fd; |
0284adc6 | 3066 | f->mode = mode; |
7560fffc | 3067 | |
0284adc6 LP |
3068 | f->flags = flags; |
3069 | f->prot = prot_from_flags(flags); | |
3070 | f->writable = (flags & O_ACCMODE) != O_RDONLY; | |
92261977 | 3071 | #if defined(HAVE_LZ4) |
d89c8fdf ZJS |
3072 | f->compress_lz4 = compress; |
3073 | #elif defined(HAVE_XZ) | |
3074 | f->compress_xz = compress; | |
48b61739 | 3075 | #endif |
49a32d43 | 3076 | #ifdef HAVE_GCRYPT |
baed47c3 | 3077 | f->seal = seal; |
49a32d43 | 3078 | #endif |
7560fffc | 3079 | |
0284adc6 LP |
3080 | if (mmap_cache) |
3081 | f->mmap = mmap_cache_ref(mmap_cache); | |
3082 | else { | |
84168d80 | 3083 | f->mmap = mmap_cache_new(); |
0284adc6 LP |
3084 | if (!f->mmap) { |
3085 | r = -ENOMEM; | |
3086 | goto fail; | |
3087 | } | |
3088 | } | |
7560fffc | 3089 | |
5d1ce257 LP |
3090 | if (fname) |
3091 | f->path = strdup(fname); | |
3092 | else /* If we don't know the path, fill in something explanatory and vaguely useful */ | |
3093 | asprintf(&f->path, "/proc/self/%i", fd); | |
0284adc6 LP |
3094 | if (!f->path) { |
3095 | r = -ENOMEM; | |
3096 | goto fail; | |
3097 | } | |
7560fffc | 3098 | |
4743015d | 3099 | f->chain_cache = ordered_hashmap_new(&uint64_hash_ops); |
a4bcff5b LP |
3100 | if (!f->chain_cache) { |
3101 | r = -ENOMEM; | |
3102 | goto fail; | |
3103 | } | |
3104 | ||
0284adc6 | 3105 | if (f->fd < 0) { |
5d1ce257 LP |
3106 | f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode); |
3107 | if (f->fd < 0) { | |
3108 | r = -errno; | |
3109 | goto fail; | |
3110 | } | |
3111 | ||
3112 | /* fds we opened here by us should also be closed by us. */ | |
3113 | f->close_fd = true; | |
7560fffc | 3114 | } |
7560fffc | 3115 | |
2678031a LP |
3116 | r = journal_file_fstat(f); |
3117 | if (r < 0) | |
0284adc6 | 3118 | goto fail; |
7560fffc | 3119 | |
0284adc6 | 3120 | if (f->last_stat.st_size == 0 && f->writable) { |
11689d2a | 3121 | |
fc68c929 | 3122 | (void) journal_file_warn_btrfs(f); |
11689d2a | 3123 | |
fb0951b0 LP |
3124 | /* Let's attach the creation time to the journal file, |
3125 | * so that the vacuuming code knows the age of this | |
3126 | * file even if the file might end up corrupted one | |
3127 | * day... Ideally we'd just use the creation time many | |
3128 | * file systems maintain for each file, but there is | |
3129 | * currently no usable API to query this, hence let's | |
3130 | * emulate this via extended attributes. If extended | |
3131 | * attributes are not supported we'll just skip this, | |
7517e174 | 3132 | * and rely solely on mtime/atime/ctime of the file. */ |
fb0951b0 | 3133 | |
d61b600d | 3134 | fd_setcrtime(f->fd, 0); |
7560fffc | 3135 | |
feb12d3e | 3136 | #ifdef HAVE_GCRYPT |
0284adc6 | 3137 | /* Try to load the FSPRG state, and if we can't, then |
baed47c3 | 3138 | * just don't do sealing */ |
49a32d43 LP |
3139 | if (f->seal) { |
3140 | r = journal_file_fss_load(f); | |
3141 | if (r < 0) | |
3142 | f->seal = false; | |
3143 | } | |
feb12d3e | 3144 | #endif |
7560fffc | 3145 | |
0284adc6 LP |
3146 | r = journal_file_init_header(f, template); |
3147 | if (r < 0) | |
3148 | goto fail; | |
7560fffc | 3149 | |
2678031a LP |
3150 | r = journal_file_fstat(f); |
3151 | if (r < 0) | |
0284adc6 | 3152 | goto fail; |
fb0951b0 LP |
3153 | |
3154 | newly_created = true; | |
0284adc6 | 3155 | } |
7560fffc | 3156 | |
0284adc6 | 3157 | if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { |
cfb571f3 | 3158 | r = -ENODATA; |
0284adc6 LP |
3159 | goto fail; |
3160 | } | |
7560fffc | 3161 | |
fa6ac760 | 3162 | r = mmap_cache_get(f->mmap, f->fd, f->prot, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h); |
977eaa1e | 3163 | if (r < 0) |
0284adc6 | 3164 | goto fail; |
7560fffc | 3165 | |
fa6ac760 LP |
3166 | f->header = h; |
3167 | ||
0284adc6 | 3168 | if (!newly_created) { |
b58c888f VC |
3169 | if (deferred_closes) |
3170 | journal_file_close_set(deferred_closes); | |
3171 | ||
0284adc6 LP |
3172 | r = journal_file_verify_header(f); |
3173 | if (r < 0) | |
3174 | goto fail; | |
3175 | } | |
7560fffc | 3176 | |
feb12d3e | 3177 | #ifdef HAVE_GCRYPT |
0284adc6 | 3178 | if (!newly_created && f->writable) { |
baed47c3 | 3179 | r = journal_file_fss_load(f); |
0284adc6 LP |
3180 | if (r < 0) |
3181 | goto fail; | |
3182 | } | |
feb12d3e | 3183 | #endif |
cec736d2 LP |
3184 | |
3185 | if (f->writable) { | |
4a92baf3 LP |
3186 | if (metrics) { |
3187 | journal_default_metrics(metrics, f->fd); | |
3188 | f->metrics = *metrics; | |
3189 | } else if (template) | |
3190 | f->metrics = template->metrics; | |
3191 | ||
cec736d2 LP |
3192 | r = journal_file_refresh_header(f); |
3193 | if (r < 0) | |
3194 | goto fail; | |
3195 | } | |
3196 | ||
feb12d3e | 3197 | #ifdef HAVE_GCRYPT |
baed47c3 | 3198 | r = journal_file_hmac_setup(f); |
14d10188 LP |
3199 | if (r < 0) |
3200 | goto fail; | |
feb12d3e | 3201 | #endif |
14d10188 | 3202 | |
cec736d2 | 3203 | if (newly_created) { |
de190aef | 3204 | r = journal_file_setup_field_hash_table(f); |
cec736d2 LP |
3205 | if (r < 0) |
3206 | goto fail; | |
3207 | ||
de190aef | 3208 | r = journal_file_setup_data_hash_table(f); |
cec736d2 LP |
3209 | if (r < 0) |
3210 | goto fail; | |
7560fffc | 3211 | |
feb12d3e | 3212 | #ifdef HAVE_GCRYPT |
7560fffc LP |
3213 | r = journal_file_append_first_tag(f); |
3214 | if (r < 0) | |
3215 | goto fail; | |
feb12d3e | 3216 | #endif |
cec736d2 LP |
3217 | } |
3218 | ||
fa6ac760 LP |
3219 | if (mmap_cache_got_sigbus(f->mmap, f->fd)) { |
3220 | r = -EIO; | |
3221 | goto fail; | |
3222 | } | |
3223 | ||
7a24f3bf | 3224 | if (template && template->post_change_timer) { |
e167d7fd LP |
3225 | r = journal_file_enable_post_change_timer( |
3226 | f, | |
3227 | sd_event_source_get_event(template->post_change_timer), | |
3228 | template->post_change_timer_period); | |
7a24f3bf | 3229 | |
7a24f3bf VC |
3230 | if (r < 0) |
3231 | goto fail; | |
3232 | } | |
3233 | ||
f8e2f4d6 | 3234 | /* The file is opened now successfully, thus we take possession of any passed in fd. */ |
5d1ce257 LP |
3235 | f->close_fd = true; |
3236 | ||
0559d3a5 | 3237 | *ret = f; |
cec736d2 LP |
3238 | return 0; |
3239 | ||
3240 | fail: | |
fa6ac760 LP |
3241 | if (f->fd >= 0 && mmap_cache_got_sigbus(f->mmap, f->fd)) |
3242 | r = -EIO; | |
3243 | ||
69a3a6fd | 3244 | (void) journal_file_close(f); |
cec736d2 LP |
3245 | |
3246 | return r; | |
3247 | } | |
0ac38b70 | 3248 | |
b58c888f | 3249 | int journal_file_rotate(JournalFile **f, bool compress, bool seal, Set *deferred_closes) { |
57535f47 | 3250 | _cleanup_free_ char *p = NULL; |
0ac38b70 LP |
3251 | size_t l; |
3252 | JournalFile *old_file, *new_file = NULL; | |
3253 | int r; | |
3254 | ||
3255 | assert(f); | |
3256 | assert(*f); | |
3257 | ||
3258 | old_file = *f; | |
3259 | ||
3260 | if (!old_file->writable) | |
3261 | return -EINVAL; | |
3262 | ||
5d1ce257 LP |
3263 | /* Is this a journal file that was passed to us as fd? If so, we synthesized a path name for it, and we refuse |
3264 | * rotation, since we don't know the actual path, and couldn't rename the file hence.*/ | |
3265 | if (path_startswith(old_file->path, "/proc/self/fd")) | |
3266 | return -EINVAL; | |
3267 | ||
0ac38b70 LP |
3268 | if (!endswith(old_file->path, ".journal")) |
3269 | return -EINVAL; | |
3270 | ||
3271 | l = strlen(old_file->path); | |
57535f47 ZJS |
3272 | r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal", |
3273 | (int) l - 8, old_file->path, | |
3274 | SD_ID128_FORMAT_VAL(old_file->header->seqnum_id), | |
3275 | le64toh((*f)->header->head_entry_seqnum), | |
3276 | le64toh((*f)->header->head_entry_realtime)); | |
3277 | if (r < 0) | |
0ac38b70 LP |
3278 | return -ENOMEM; |
3279 | ||
2678031a LP |
3280 | /* Try to rename the file to the archived version. If the file |
3281 | * already was deleted, we'll get ENOENT, let's ignore that | |
3282 | * case. */ | |
0ac38b70 | 3283 | r = rename(old_file->path, p); |
2678031a | 3284 | if (r < 0 && errno != ENOENT) |
0ac38b70 LP |
3285 | return -errno; |
3286 | ||
1fcefd88 LP |
3287 | /* Sync the rename to disk */ |
3288 | (void) fsync_directory_of_file(old_file->fd); | |
3289 | ||
8eb85171 VC |
3290 | /* Set as archive so offlining commits w/state=STATE_ARCHIVED. |
3291 | * Previously we would set old_file->header->state to STATE_ARCHIVED directly here, | |
3292 | * but journal_file_set_offline() short-circuits when state != STATE_ONLINE, which | |
3293 | * would result in the rotated journal never getting fsync() called before closing. | |
3294 | * Now we simply queue the archive state by setting an archive bit, leaving the state | |
3295 | * as STATE_ONLINE so proper offlining occurs. */ | |
3296 | old_file->archive = true; | |
0ac38b70 | 3297 | |
f27a3864 LP |
3298 | /* Currently, btrfs is not very good with out write patterns |
3299 | * and fragments heavily. Let's defrag our journal files when | |
3300 | * we archive them */ | |
3301 | old_file->defrag_on_close = true; | |
3302 | ||
5d1ce257 | 3303 | r = journal_file_open(-1, old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, deferred_closes, old_file, &new_file); |
b58c888f VC |
3304 | |
3305 | if (deferred_closes && | |
3306 | set_put(deferred_closes, old_file) >= 0) | |
3307 | (void) journal_file_set_offline(old_file, false); | |
3308 | else | |
3309 | (void) journal_file_close(old_file); | |
0ac38b70 LP |
3310 | |
3311 | *f = new_file; | |
3312 | return r; | |
3313 | } | |
3314 | ||
9447a7f1 LP |
3315 | int journal_file_open_reliably( |
3316 | const char *fname, | |
3317 | int flags, | |
3318 | mode_t mode, | |
7560fffc | 3319 | bool compress, |
baed47c3 | 3320 | bool seal, |
4a92baf3 | 3321 | JournalMetrics *metrics, |
27370278 | 3322 | MMapCache *mmap_cache, |
b58c888f | 3323 | Set *deferred_closes, |
9447a7f1 LP |
3324 | JournalFile *template, |
3325 | JournalFile **ret) { | |
3326 | ||
3327 | int r; | |
3328 | size_t l; | |
ed375beb | 3329 | _cleanup_free_ char *p = NULL; |
9447a7f1 | 3330 | |
5d1ce257 | 3331 | r = journal_file_open(-1, fname, flags, mode, compress, seal, metrics, mmap_cache, deferred_closes, template, ret); |
288359db ZJS |
3332 | if (!IN_SET(r, |
3333 | -EBADMSG, /* corrupted */ | |
3334 | -ENODATA, /* truncated */ | |
3335 | -EHOSTDOWN, /* other machine */ | |
3336 | -EPROTONOSUPPORT, /* incompatible feature */ | |
3337 | -EBUSY, /* unclean shutdown */ | |
3338 | -ESHUTDOWN, /* already archived */ | |
3339 | -EIO, /* IO error, including SIGBUS on mmap */ | |
ae739cc1 LP |
3340 | -EIDRM, /* File has been deleted */ |
3341 | -ETXTBSY)) /* File is from the future */ | |
9447a7f1 LP |
3342 | return r; |
3343 | ||
3344 | if ((flags & O_ACCMODE) == O_RDONLY) | |
3345 | return r; | |
3346 | ||
3347 | if (!(flags & O_CREAT)) | |
3348 | return r; | |
3349 | ||
7560fffc LP |
3350 | if (!endswith(fname, ".journal")) |
3351 | return r; | |
3352 | ||
5c70eab4 LP |
3353 | /* The file is corrupted. Rotate it away and try it again (but only once) */ |
3354 | ||
9447a7f1 | 3355 | l = strlen(fname); |
d587eca5 | 3356 | if (asprintf(&p, "%.*s@%016"PRIx64 "-%016"PRIx64 ".journal~", |
57535f47 | 3357 | (int) l - 8, fname, |
d587eca5 | 3358 | now(CLOCK_REALTIME), |
9bf3b535 | 3359 | random_u64()) < 0) |
9447a7f1 LP |
3360 | return -ENOMEM; |
3361 | ||
65089b82 | 3362 | if (rename(fname, p) < 0) |
9447a7f1 LP |
3363 | return -errno; |
3364 | ||
f27a3864 LP |
3365 | /* btrfs doesn't cope well with our write pattern and |
3366 | * fragments heavily. Let's defrag all files we rotate */ | |
11689d2a | 3367 | |
a67d68b8 | 3368 | (void) chattr_path(p, 0, FS_NOCOW_FL); |
f27a3864 LP |
3369 | (void) btrfs_defrag(p); |
3370 | ||
65089b82 | 3371 | log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname); |
9447a7f1 | 3372 | |
5d1ce257 | 3373 | return journal_file_open(-1, fname, flags, mode, compress, seal, metrics, mmap_cache, deferred_closes, template, ret); |
9447a7f1 LP |
3374 | } |
3375 | ||
cf244689 LP |
3376 | int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset) { |
3377 | uint64_t i, n; | |
3378 | uint64_t q, xor_hash = 0; | |
3379 | int r; | |
3380 | EntryItem *items; | |
3381 | dual_timestamp ts; | |
3382 | ||
3383 | assert(from); | |
3384 | assert(to); | |
3385 | assert(o); | |
3386 | assert(p); | |
3387 | ||
3388 | if (!to->writable) | |
3389 | return -EPERM; | |
3390 | ||
3391 | ts.monotonic = le64toh(o->entry.monotonic); | |
3392 | ts.realtime = le64toh(o->entry.realtime); | |
3393 | ||
cf244689 | 3394 | n = journal_file_entry_n_items(o); |
4faa7004 TA |
3395 | /* alloca() can't take 0, hence let's allocate at least one */ |
3396 | items = alloca(sizeof(EntryItem) * MAX(1u, n)); | |
cf244689 LP |
3397 | |
3398 | for (i = 0; i < n; i++) { | |
4fd052ae FC |
3399 | uint64_t l, h; |
3400 | le64_t le_hash; | |
cf244689 LP |
3401 | size_t t; |
3402 | void *data; | |
3403 | Object *u; | |
3404 | ||
3405 | q = le64toh(o->entry.items[i].object_offset); | |
3406 | le_hash = o->entry.items[i].hash; | |
3407 | ||
3408 | r = journal_file_move_to_object(from, OBJECT_DATA, q, &o); | |
3409 | if (r < 0) | |
3410 | return r; | |
3411 | ||
3412 | if (le_hash != o->data.hash) | |
3413 | return -EBADMSG; | |
3414 | ||
3415 | l = le64toh(o->object.size) - offsetof(Object, data.payload); | |
3416 | t = (size_t) l; | |
3417 | ||
3418 | /* We hit the limit on 32bit machines */ | |
3419 | if ((uint64_t) t != l) | |
3420 | return -E2BIG; | |
3421 | ||
d89c8fdf | 3422 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
3b1a55e1 | 3423 | #if defined(HAVE_XZ) || defined(HAVE_LZ4) |
a7f7d1bd | 3424 | size_t rsize = 0; |
cf244689 | 3425 | |
d89c8fdf ZJS |
3426 | r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, |
3427 | o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0); | |
3428 | if (r < 0) | |
3429 | return r; | |
cf244689 LP |
3430 | |
3431 | data = from->compress_buffer; | |
3432 | l = rsize; | |
3b1a55e1 ZJS |
3433 | #else |
3434 | return -EPROTONOSUPPORT; | |
3435 | #endif | |
cf244689 LP |
3436 | } else |
3437 | data = o->data.payload; | |
3438 | ||
3439 | r = journal_file_append_data(to, data, l, &u, &h); | |
3440 | if (r < 0) | |
3441 | return r; | |
3442 | ||
3443 | xor_hash ^= le64toh(u->data.hash); | |
3444 | items[i].object_offset = htole64(h); | |
3445 | items[i].hash = u->data.hash; | |
3446 | ||
3447 | r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); | |
3448 | if (r < 0) | |
3449 | return r; | |
3450 | } | |
3451 | ||
fa6ac760 LP |
3452 | r = journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset); |
3453 | ||
3454 | if (mmap_cache_got_sigbus(to->mmap, to->fd)) | |
3455 | return -EIO; | |
3456 | ||
3457 | return r; | |
cf244689 | 3458 | } |
babfc091 | 3459 | |
8580d1f7 LP |
3460 | void journal_reset_metrics(JournalMetrics *m) { |
3461 | assert(m); | |
3462 | ||
3463 | /* Set everything to "pick automatic values". */ | |
3464 | ||
3465 | *m = (JournalMetrics) { | |
3466 | .min_use = (uint64_t) -1, | |
3467 | .max_use = (uint64_t) -1, | |
3468 | .min_size = (uint64_t) -1, | |
3469 | .max_size = (uint64_t) -1, | |
3470 | .keep_free = (uint64_t) -1, | |
3471 | .n_max_files = (uint64_t) -1, | |
3472 | }; | |
3473 | } | |
3474 | ||
babfc091 | 3475 | void journal_default_metrics(JournalMetrics *m, int fd) { |
8580d1f7 | 3476 | char a[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX], e[FORMAT_BYTES_MAX]; |
babfc091 | 3477 | struct statvfs ss; |
8580d1f7 | 3478 | uint64_t fs_size; |
babfc091 LP |
3479 | |
3480 | assert(m); | |
3481 | assert(fd >= 0); | |
3482 | ||
3483 | if (fstatvfs(fd, &ss) >= 0) | |
3484 | fs_size = ss.f_frsize * ss.f_blocks; | |
8580d1f7 LP |
3485 | else { |
3486 | log_debug_errno(errno, "Failed to detremine disk size: %m"); | |
3487 | fs_size = 0; | |
3488 | } | |
babfc091 LP |
3489 | |
3490 | if (m->max_use == (uint64_t) -1) { | |
3491 | ||
3492 | if (fs_size > 0) { | |
3493 | m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */ | |
3494 | ||
3495 | if (m->max_use > DEFAULT_MAX_USE_UPPER) | |
3496 | m->max_use = DEFAULT_MAX_USE_UPPER; | |
3497 | ||
3498 | if (m->max_use < DEFAULT_MAX_USE_LOWER) | |
3499 | m->max_use = DEFAULT_MAX_USE_LOWER; | |
3500 | } else | |
3501 | m->max_use = DEFAULT_MAX_USE_LOWER; | |
3502 | } else { | |
3503 | m->max_use = PAGE_ALIGN(m->max_use); | |
3504 | ||
8580d1f7 | 3505 | if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2) |
babfc091 LP |
3506 | m->max_use = JOURNAL_FILE_SIZE_MIN*2; |
3507 | } | |
3508 | ||
8580d1f7 LP |
3509 | if (m->min_use == (uint64_t) -1) |
3510 | m->min_use = DEFAULT_MIN_USE; | |
3511 | ||
3512 | if (m->min_use > m->max_use) | |
3513 | m->min_use = m->max_use; | |
3514 | ||
babfc091 LP |
3515 | if (m->max_size == (uint64_t) -1) { |
3516 | m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */ | |
3517 | ||
3518 | if (m->max_size > DEFAULT_MAX_SIZE_UPPER) | |
3519 | m->max_size = DEFAULT_MAX_SIZE_UPPER; | |
3520 | } else | |
3521 | m->max_size = PAGE_ALIGN(m->max_size); | |
3522 | ||
8580d1f7 LP |
3523 | if (m->max_size != 0) { |
3524 | if (m->max_size < JOURNAL_FILE_SIZE_MIN) | |
3525 | m->max_size = JOURNAL_FILE_SIZE_MIN; | |
babfc091 | 3526 | |
8580d1f7 LP |
3527 | if (m->max_use != 0 && m->max_size*2 > m->max_use) |
3528 | m->max_use = m->max_size*2; | |
3529 | } | |
babfc091 LP |
3530 | |
3531 | if (m->min_size == (uint64_t) -1) | |
3532 | m->min_size = JOURNAL_FILE_SIZE_MIN; | |
3533 | else { | |
3534 | m->min_size = PAGE_ALIGN(m->min_size); | |
3535 | ||
3536 | if (m->min_size < JOURNAL_FILE_SIZE_MIN) | |
3537 | m->min_size = JOURNAL_FILE_SIZE_MIN; | |
3538 | ||
8580d1f7 | 3539 | if (m->max_size != 0 && m->min_size > m->max_size) |
babfc091 LP |
3540 | m->max_size = m->min_size; |
3541 | } | |
3542 | ||
3543 | if (m->keep_free == (uint64_t) -1) { | |
3544 | ||
3545 | if (fs_size > 0) { | |
8621b110 | 3546 | m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */ |
babfc091 LP |
3547 | |
3548 | if (m->keep_free > DEFAULT_KEEP_FREE_UPPER) | |
3549 | m->keep_free = DEFAULT_KEEP_FREE_UPPER; | |
3550 | ||
3551 | } else | |
3552 | m->keep_free = DEFAULT_KEEP_FREE; | |
3553 | } | |
3554 | ||
8580d1f7 LP |
3555 | if (m->n_max_files == (uint64_t) -1) |
3556 | m->n_max_files = DEFAULT_N_MAX_FILES; | |
3557 | ||
3558 | log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64, | |
3559 | format_bytes(a, sizeof(a), m->min_use), | |
3560 | format_bytes(b, sizeof(b), m->max_use), | |
3561 | format_bytes(c, sizeof(c), m->max_size), | |
3562 | format_bytes(d, sizeof(d), m->min_size), | |
3563 | format_bytes(e, sizeof(e), m->keep_free), | |
3564 | m->n_max_files); | |
babfc091 | 3565 | } |
08984293 LP |
3566 | |
3567 | int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { | |
08984293 | 3568 | assert(f); |
c88cc6af | 3569 | assert(f->header); |
08984293 LP |
3570 | assert(from || to); |
3571 | ||
3572 | if (from) { | |
162566a4 LP |
3573 | if (f->header->head_entry_realtime == 0) |
3574 | return -ENOENT; | |
08984293 | 3575 | |
162566a4 | 3576 | *from = le64toh(f->header->head_entry_realtime); |
08984293 LP |
3577 | } |
3578 | ||
3579 | if (to) { | |
162566a4 LP |
3580 | if (f->header->tail_entry_realtime == 0) |
3581 | return -ENOENT; | |
08984293 | 3582 | |
162566a4 | 3583 | *to = le64toh(f->header->tail_entry_realtime); |
08984293 LP |
3584 | } |
3585 | ||
3586 | return 1; | |
3587 | } | |
3588 | ||
3589 | int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) { | |
08984293 LP |
3590 | Object *o; |
3591 | uint64_t p; | |
3592 | int r; | |
3593 | ||
3594 | assert(f); | |
3595 | assert(from || to); | |
3596 | ||
47838ab3 | 3597 | r = find_data_object_by_boot_id(f, boot_id, &o, &p); |
08984293 LP |
3598 | if (r <= 0) |
3599 | return r; | |
3600 | ||
3601 | if (le64toh(o->data.n_entries) <= 0) | |
3602 | return 0; | |
3603 | ||
3604 | if (from) { | |
3605 | r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); | |
3606 | if (r < 0) | |
3607 | return r; | |
3608 | ||
3609 | *from = le64toh(o->entry.monotonic); | |
3610 | } | |
3611 | ||
3612 | if (to) { | |
3613 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
3614 | if (r < 0) | |
3615 | return r; | |
3616 | ||
3617 | r = generic_array_get_plus_one(f, | |
3618 | le64toh(o->data.entry_offset), | |
3619 | le64toh(o->data.entry_array_offset), | |
3620 | le64toh(o->data.n_entries)-1, | |
3621 | &o, NULL); | |
3622 | if (r <= 0) | |
3623 | return r; | |
3624 | ||
3625 | *to = le64toh(o->entry.monotonic); | |
3626 | } | |
3627 | ||
3628 | return 1; | |
3629 | } | |
dca6219e | 3630 | |
fb0951b0 | 3631 | bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) { |
dca6219e | 3632 | assert(f); |
c88cc6af | 3633 | assert(f->header); |
dca6219e LP |
3634 | |
3635 | /* If we gained new header fields we gained new features, | |
3636 | * hence suggest a rotation */ | |
361f9cbc LP |
3637 | if (le64toh(f->header->header_size) < sizeof(Header)) { |
3638 | log_debug("%s uses an outdated header, suggesting rotation.", f->path); | |
dca6219e | 3639 | return true; |
361f9cbc | 3640 | } |
dca6219e LP |
3641 | |
3642 | /* Let's check if the hash tables grew over a certain fill | |
3643 | * level (75%, borrowing this value from Java's hash table | |
3644 | * implementation), and if so suggest a rotation. To calculate | |
3645 | * the fill level we need the n_data field, which only exists | |
3646 | * in newer versions. */ | |
3647 | ||
3648 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) | |
361f9cbc | 3649 | if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
507f22bd | 3650 | 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 |
3651 | f->path, |
3652 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))), | |
507f22bd ZJS |
3653 | le64toh(f->header->n_data), |
3654 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
3655 | (unsigned long long) f->last_stat.st_size, | |
3656 | f->last_stat.st_size / le64toh(f->header->n_data)); | |
dca6219e | 3657 | return true; |
361f9cbc | 3658 | } |
dca6219e LP |
3659 | |
3660 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
361f9cbc | 3661 | if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
507f22bd | 3662 | log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.", |
361f9cbc LP |
3663 | f->path, |
3664 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), | |
507f22bd ZJS |
3665 | le64toh(f->header->n_fields), |
3666 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem)); | |
dca6219e | 3667 | return true; |
361f9cbc | 3668 | } |
dca6219e | 3669 | |
0598fd4a LP |
3670 | /* Are the data objects properly indexed by field objects? */ |
3671 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && | |
3672 | JOURNAL_HEADER_CONTAINS(f->header, n_fields) && | |
3673 | le64toh(f->header->n_data) > 0 && | |
3674 | le64toh(f->header->n_fields) == 0) | |
3675 | return true; | |
3676 | ||
fb0951b0 LP |
3677 | if (max_file_usec > 0) { |
3678 | usec_t t, h; | |
3679 | ||
3680 | h = le64toh(f->header->head_entry_realtime); | |
3681 | t = now(CLOCK_REALTIME); | |
3682 | ||
3683 | if (h > 0 && t > h + max_file_usec) | |
3684 | return true; | |
3685 | } | |
3686 | ||
dca6219e LP |
3687 | return false; |
3688 | } |