]>
Commit | Line | Data |
---|---|---|
db9ecf05 | 1 | /* SPDX-License-Identifier: LGPL-2.1-or-later */ |
cec736d2 | 2 | |
cec736d2 | 3 | #include <errno.h> |
cec736d2 | 4 | #include <fcntl.h> |
11689d2a | 5 | #include <linux/fs.h> |
65ddc2c5 | 6 | #include <linux/magic.h> |
ac2e41f5 | 7 | #include <pthread.h> |
07630cea LP |
8 | #include <stddef.h> |
9 | #include <sys/mman.h> | |
10 | #include <sys/statvfs.h> | |
11 | #include <sys/uio.h> | |
12 | #include <unistd.h> | |
fb0951b0 | 13 | |
a03d4359 ZJS |
14 | #include "sd-event.h" |
15 | ||
b5efdb8a | 16 | #include "alloc-util.h" |
c8b3094d | 17 | #include "chattr-util.h" |
07630cea | 18 | #include "compress.h" |
4ce534f4 | 19 | #include "env-util.h" |
3ffd4af2 | 20 | #include "fd-util.h" |
aa892669 | 21 | #include "format-util.h" |
11b29a96 | 22 | #include "fs-util.h" |
0284adc6 | 23 | #include "journal-authenticate.h" |
cec736d2 LP |
24 | #include "journal-def.h" |
25 | #include "journal-file.h" | |
26 | #include "lookup3.h" | |
0a970718 | 27 | #include "memory-util.h" |
5d1ce257 | 28 | #include "path-util.h" |
3df3e884 | 29 | #include "random-util.h" |
b58c888f | 30 | #include "set.h" |
760877e9 | 31 | #include "sort-util.h" |
3cc44114 | 32 | #include "stat-util.h" |
363b2b9a | 33 | #include "string-table.h" |
07630cea | 34 | #include "string-util.h" |
4761fd0f | 35 | #include "strv.h" |
bf819d3a | 36 | #include "sync-util.h" |
89a5a90c | 37 | #include "xattr-util.h" |
cec736d2 | 38 | |
4a92baf3 LP |
39 | #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) |
40 | #define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) | |
cec736d2 | 41 | |
57850536 AG |
42 | #define DEFAULT_COMPRESS_THRESHOLD (512ULL) |
43 | #define MIN_COMPRESS_THRESHOLD (8ULL) | |
807e17f0 | 44 | |
babfc091 | 45 | /* This is the minimum journal file size */ |
6aae0b1a | 46 | #define JOURNAL_FILE_SIZE_MIN (512 * 1024ULL) /* 512 KiB */ |
babfc091 LP |
47 | |
48 | /* These are the lower and upper bounds if we deduce the max_use value | |
49 | * from the file system size */ | |
6aae0b1a ZJS |
50 | #define MAX_USE_LOWER (1 * 1024 * 1024ULL) /* 1 MiB */ |
51 | #define MAX_USE_UPPER (4 * 1024 * 1024 * 1024ULL) /* 4 GiB */ | |
babfc091 | 52 | |
6aae0b1a ZJS |
53 | /* Those are the lower and upper bounds for the minimal use limit, |
54 | * i.e. how much we'll use even if keep_free suggests otherwise. */ | |
55 | #define MIN_USE_LOW (1 * 1024 * 1024ULL) /* 1 MiB */ | |
56 | #define MIN_USE_HIGH (16 * 1024 * 1024ULL) /* 16 MiB */ | |
8580d1f7 | 57 | |
babfc091 | 58 | /* This is the upper bound if we deduce max_size from max_use */ |
6aae0b1a | 59 | #define MAX_SIZE_UPPER (128 * 1024 * 1024ULL) /* 128 MiB */ |
babfc091 LP |
60 | |
61 | /* This is the upper bound if we deduce the keep_free value from the | |
62 | * file system size */ | |
6aae0b1a | 63 | #define KEEP_FREE_UPPER (4 * 1024 * 1024 * 1024ULL) /* 4 GiB */ |
babfc091 LP |
64 | |
65 | /* This is the keep_free value when we can't determine the system | |
66 | * size */ | |
6aae0b1a | 67 | #define DEFAULT_KEEP_FREE (1024 * 1024ULL) /* 1 MB */ |
babfc091 | 68 | |
8580d1f7 | 69 | /* This is the default maximum number of journal files to keep around. */ |
6aae0b1a | 70 | #define DEFAULT_N_MAX_FILES 100 |
8580d1f7 | 71 | |
dca6219e LP |
72 | /* n_data was the first entry we added after the initial file format design */ |
73 | #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) | |
cec736d2 | 74 | |
a4bcff5b LP |
75 | /* How many entries to keep in the entry array chain cache at max */ |
76 | #define CHAIN_CACHE_MAX 20 | |
77 | ||
a676e665 | 78 | /* How much to increase the journal file size at once each time we allocate something new. */ |
6aae0b1a | 79 | #define FILE_SIZE_INCREASE (8 * 1024 * 1024ULL) /* 8MB */ |
a676e665 | 80 | |
2678031a LP |
81 | /* Reread fstat() of the file for detecting deletions at least this often */ |
82 | #define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC) | |
83 | ||
fa6ac760 LP |
84 | /* The mmap context to use for the header we pick as one above the last defined typed */ |
85 | #define CONTEXT_HEADER _OBJECT_TYPE_MAX | |
86 | ||
0dbe57ee LP |
87 | /* Longest hash chain to rotate after */ |
88 | #define HASH_CHAIN_DEPTH_MAX 100 | |
89 | ||
51804460 ZJS |
90 | #ifdef __clang__ |
91 | # pragma GCC diagnostic ignored "-Waddress-of-packed-member" | |
92 | #endif | |
93 | ||
ac2e41f5 VC |
94 | /* This may be called from a separate thread to prevent blocking the caller for the duration of fsync(). |
95 | * As a result we use atomic operations on f->offline_state for inter-thread communications with | |
96 | * journal_file_set_offline() and journal_file_set_online(). */ | |
97 | static void journal_file_set_offline_internal(JournalFile *f) { | |
26687bf8 | 98 | assert(f); |
ac2e41f5 VC |
99 | assert(f->fd >= 0); |
100 | assert(f->header); | |
101 | ||
102 | for (;;) { | |
103 | switch (f->offline_state) { | |
104 | case OFFLINE_CANCEL: | |
105 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_DONE)) | |
106 | continue; | |
107 | return; | |
108 | ||
109 | case OFFLINE_AGAIN_FROM_SYNCING: | |
110 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_SYNCING)) | |
111 | continue; | |
112 | break; | |
113 | ||
114 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
115 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_SYNCING)) | |
116 | continue; | |
117 | break; | |
118 | ||
119 | case OFFLINE_SYNCING: | |
120 | (void) fsync(f->fd); | |
26687bf8 | 121 | |
ac2e41f5 VC |
122 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_OFFLINING)) |
123 | continue; | |
26687bf8 | 124 | |
8eb85171 | 125 | f->header->state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE; |
ac2e41f5 VC |
126 | (void) fsync(f->fd); |
127 | break; | |
128 | ||
129 | case OFFLINE_OFFLINING: | |
130 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_DONE)) | |
131 | continue; | |
4831981d | 132 | _fallthrough_; |
ac2e41f5 VC |
133 | case OFFLINE_DONE: |
134 | return; | |
135 | ||
136 | case OFFLINE_JOINED: | |
137 | log_debug("OFFLINE_JOINED unexpected offline state for journal_file_set_offline_internal()"); | |
138 | return; | |
139 | } | |
140 | } | |
141 | } | |
142 | ||
143 | static void * journal_file_set_offline_thread(void *arg) { | |
144 | JournalFile *f = arg; | |
145 | ||
fa7ff4cf LP |
146 | (void) pthread_setname_np(pthread_self(), "journal-offline"); |
147 | ||
ac2e41f5 VC |
148 | journal_file_set_offline_internal(f); |
149 | ||
150 | return NULL; | |
151 | } | |
152 | ||
153 | static int journal_file_set_offline_thread_join(JournalFile *f) { | |
154 | int r; | |
155 | ||
156 | assert(f); | |
157 | ||
158 | if (f->offline_state == OFFLINE_JOINED) | |
159 | return 0; | |
160 | ||
161 | r = pthread_join(f->offline_thread, NULL); | |
162 | if (r) | |
163 | return -r; | |
164 | ||
165 | f->offline_state = OFFLINE_JOINED; | |
26687bf8 | 166 | |
c3bd54bf | 167 | if (mmap_cache_fd_got_sigbus(f->cache_fd)) |
fa6ac760 LP |
168 | return -EIO; |
169 | ||
ac2e41f5 VC |
170 | return 0; |
171 | } | |
26687bf8 | 172 | |
ac2e41f5 VC |
173 | /* Trigger a restart if the offline thread is mid-flight in a restartable state. */ |
174 | static bool journal_file_set_offline_try_restart(JournalFile *f) { | |
175 | for (;;) { | |
176 | switch (f->offline_state) { | |
177 | case OFFLINE_AGAIN_FROM_SYNCING: | |
178 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
179 | return true; | |
180 | ||
181 | case OFFLINE_CANCEL: | |
182 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_AGAIN_FROM_SYNCING)) | |
183 | continue; | |
184 | return true; | |
185 | ||
186 | case OFFLINE_SYNCING: | |
187 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_AGAIN_FROM_SYNCING)) | |
188 | continue; | |
189 | return true; | |
190 | ||
191 | case OFFLINE_OFFLINING: | |
192 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_AGAIN_FROM_OFFLINING)) | |
193 | continue; | |
194 | return true; | |
26687bf8 OS |
195 | |
196 | default: | |
ac2e41f5 VC |
197 | return false; |
198 | } | |
26687bf8 OS |
199 | } |
200 | } | |
201 | ||
ac2e41f5 VC |
202 | /* Sets a journal offline. |
203 | * | |
204 | * If wait is false then an offline is dispatched in a separate thread for a | |
205 | * subsequent journal_file_set_offline() or journal_file_set_online() of the | |
206 | * same journal to synchronize with. | |
207 | * | |
208 | * If wait is true, then either an existing offline thread will be restarted | |
209 | * and joined, or if none exists the offline is simply performed in this | |
210 | * context without involving another thread. | |
211 | */ | |
212 | int journal_file_set_offline(JournalFile *f, bool wait) { | |
bb1296b5 | 213 | int target_state; |
ac2e41f5 VC |
214 | bool restarted; |
215 | int r; | |
216 | ||
26687bf8 OS |
217 | assert(f); |
218 | ||
219 | if (!f->writable) | |
220 | return -EPERM; | |
221 | ||
846e5418 | 222 | if (f->fd < 0 || !f->header) |
26687bf8 OS |
223 | return -EINVAL; |
224 | ||
bb1296b5 VC |
225 | target_state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE; |
226 | ||
b8f99e27 | 227 | /* An offlining journal is implicitly online and may modify f->header->state, |
bb1296b5 VC |
228 | * we must also join any potentially lingering offline thread when already in |
229 | * the desired offline state. | |
230 | */ | |
231 | if (!journal_file_is_offlining(f) && f->header->state == target_state) | |
b8f99e27 | 232 | return journal_file_set_offline_thread_join(f); |
26687bf8 | 233 | |
ac2e41f5 VC |
234 | /* Restart an in-flight offline thread and wait if needed, or join a lingering done one. */ |
235 | restarted = journal_file_set_offline_try_restart(f); | |
236 | if ((restarted && wait) || !restarted) { | |
237 | r = journal_file_set_offline_thread_join(f); | |
238 | if (r < 0) | |
239 | return r; | |
240 | } | |
26687bf8 | 241 | |
ac2e41f5 VC |
242 | if (restarted) |
243 | return 0; | |
244 | ||
245 | /* Initiate a new offline. */ | |
246 | f->offline_state = OFFLINE_SYNCING; | |
fa6ac760 | 247 | |
ac2e41f5 VC |
248 | if (wait) /* Without using a thread if waiting. */ |
249 | journal_file_set_offline_internal(f); | |
250 | else { | |
5e9f01e8 LP |
251 | sigset_t ss, saved_ss; |
252 | int k; | |
253 | ||
cd2a429e | 254 | assert_se(sigfillset(&ss) >= 0); |
08f9e80b CM |
255 | /* Don't block SIGBUS since the offlining thread accesses a memory mapped file. |
256 | * Asynchronous SIGBUS signals can safely be handled by either thread. */ | |
257 | assert_se(sigdelset(&ss, SIGBUS) >= 0); | |
5e9f01e8 LP |
258 | |
259 | r = pthread_sigmask(SIG_BLOCK, &ss, &saved_ss); | |
260 | if (r > 0) | |
261 | return -r; | |
262 | ||
ac2e41f5 | 263 | r = pthread_create(&f->offline_thread, NULL, journal_file_set_offline_thread, f); |
5e9f01e8 LP |
264 | |
265 | k = pthread_sigmask(SIG_SETMASK, &saved_ss, NULL); | |
ec9ffa2c VC |
266 | if (r > 0) { |
267 | f->offline_state = OFFLINE_JOINED; | |
ac2e41f5 | 268 | return -r; |
ec9ffa2c | 269 | } |
5e9f01e8 LP |
270 | if (k > 0) |
271 | return -k; | |
ac2e41f5 VC |
272 | } |
273 | ||
274 | return 0; | |
275 | } | |
276 | ||
277 | static int journal_file_set_online(JournalFile *f) { | |
83bf6b67 | 278 | bool wait = true; |
ac2e41f5 VC |
279 | |
280 | assert(f); | |
281 | ||
282 | if (!f->writable) | |
283 | return -EPERM; | |
284 | ||
846e5418 | 285 | if (f->fd < 0 || !f->header) |
ac2e41f5 VC |
286 | return -EINVAL; |
287 | ||
83bf6b67 | 288 | while (wait) { |
ac2e41f5 VC |
289 | switch (f->offline_state) { |
290 | case OFFLINE_JOINED: | |
291 | /* No offline thread, no need to wait. */ | |
83bf6b67 | 292 | wait = false; |
ac2e41f5 VC |
293 | break; |
294 | ||
295 | case OFFLINE_SYNCING: | |
296 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_CANCEL)) | |
297 | continue; | |
298 | /* Canceled syncing prior to offlining, no need to wait. */ | |
83bf6b67 | 299 | wait = false; |
ac2e41f5 VC |
300 | break; |
301 | ||
302 | case OFFLINE_AGAIN_FROM_SYNCING: | |
303 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_CANCEL)) | |
304 | continue; | |
305 | /* Canceled restart from syncing, no need to wait. */ | |
83bf6b67 | 306 | wait = false; |
ac2e41f5 VC |
307 | break; |
308 | ||
309 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
310 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_CANCEL)) | |
311 | continue; | |
312 | /* Canceled restart from offlining, must wait for offlining to complete however. */ | |
4831981d | 313 | _fallthrough_; |
ac2e41f5 VC |
314 | default: { |
315 | int r; | |
316 | ||
317 | r = journal_file_set_offline_thread_join(f); | |
318 | if (r < 0) | |
319 | return r; | |
320 | ||
83bf6b67 | 321 | wait = false; |
ac2e41f5 VC |
322 | break; |
323 | } | |
324 | } | |
325 | } | |
26687bf8 | 326 | |
c3bd54bf | 327 | if (mmap_cache_fd_got_sigbus(f->cache_fd)) |
fa6ac760 LP |
328 | return -EIO; |
329 | ||
ac2e41f5 VC |
330 | switch (f->header->state) { |
331 | case STATE_ONLINE: | |
332 | return 0; | |
26687bf8 | 333 | |
ac2e41f5 VC |
334 | case STATE_OFFLINE: |
335 | f->header->state = STATE_ONLINE; | |
336 | (void) fsync(f->fd); | |
337 | return 0; | |
338 | ||
339 | default: | |
340 | return -EINVAL; | |
341 | } | |
26687bf8 OS |
342 | } |
343 | ||
b58c888f VC |
344 | bool journal_file_is_offlining(JournalFile *f) { |
345 | assert(f); | |
346 | ||
347 | __sync_synchronize(); | |
348 | ||
3742095b | 349 | if (IN_SET(f->offline_state, OFFLINE_DONE, OFFLINE_JOINED)) |
b58c888f VC |
350 | return false; |
351 | ||
352 | return true; | |
353 | } | |
354 | ||
804ae586 | 355 | JournalFile* journal_file_close(JournalFile *f) { |
c377a6f3 YW |
356 | if (!f) |
357 | return NULL; | |
cec736d2 | 358 | |
349cc4a5 | 359 | #if HAVE_GCRYPT |
b0af6f41 | 360 | /* Write the final tag */ |
43cd8794 FB |
361 | if (f->seal && f->writable) { |
362 | int r; | |
363 | ||
364 | r = journal_file_append_tag(f); | |
365 | if (r < 0) | |
366 | log_error_errno(r, "Failed to append tag when closing journal: %m"); | |
367 | } | |
feb12d3e | 368 | #endif |
b0af6f41 | 369 | |
7a24f3bf | 370 | if (f->post_change_timer) { |
b6cdfbe5 ZJS |
371 | if (sd_event_source_get_enabled(f->post_change_timer, NULL) > 0) |
372 | journal_file_post_change(f); | |
7a24f3bf | 373 | |
1d3fe304 | 374 | sd_event_source_disable_unref(f->post_change_timer); |
7a24f3bf VC |
375 | } |
376 | ||
ac2e41f5 | 377 | journal_file_set_offline(f, true); |
cec736d2 | 378 | |
be7cdd8e | 379 | if (f->mmap && f->cache_fd) |
c3bd54bf | 380 | mmap_cache_fd_free(f->cache_fd); |
cec736d2 | 381 | |
11689d2a LP |
382 | if (f->fd >= 0 && f->defrag_on_close) { |
383 | ||
384 | /* Be friendly to btrfs: turn COW back on again now, | |
385 | * and defragment the file. We won't write to the file | |
386 | * ever again, hence remove all fragmentation, and | |
387 | * reenable all the good bits COW usually provides | |
388 | * (such as data checksumming). */ | |
389 | ||
db9a4254 | 390 | (void) chattr_fd(f->fd, 0, FS_NOCOW_FL, NULL); |
11689d2a LP |
391 | (void) btrfs_defrag_fd(f->fd); |
392 | } | |
f27a3864 | 393 | |
5d1ce257 LP |
394 | if (f->close_fd) |
395 | safe_close(f->fd); | |
cec736d2 | 396 | free(f->path); |
807e17f0 | 397 | |
f649045c | 398 | mmap_cache_unref(f->mmap); |
16e9f408 | 399 | |
4743015d | 400 | ordered_hashmap_free_free(f->chain_cache); |
a4bcff5b | 401 | |
d80b051c | 402 | #if HAVE_COMPRESSION |
807e17f0 LP |
403 | free(f->compress_buffer); |
404 | #endif | |
405 | ||
349cc4a5 | 406 | #if HAVE_GCRYPT |
baed47c3 LP |
407 | if (f->fss_file) |
408 | munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size)); | |
dc4ebc07 | 409 | else |
b7c9ae91 LP |
410 | free(f->fsprg_state); |
411 | ||
412 | free(f->fsprg_seed); | |
7560fffc LP |
413 | |
414 | if (f->hmac) | |
415 | gcry_md_close(f->hmac); | |
416 | #endif | |
417 | ||
6b430fdb | 418 | return mfree(f); |
cec736d2 LP |
419 | } |
420 | ||
0ac38b70 | 421 | static int journal_file_init_header(JournalFile *f, JournalFile *template) { |
d89c8fdf | 422 | Header h = {}; |
cec736d2 LP |
423 | ssize_t k; |
424 | int r; | |
425 | ||
426 | assert(f); | |
427 | ||
7560fffc | 428 | memcpy(h.signature, HEADER_SIGNATURE, 8); |
23b0b2b2 | 429 | h.header_size = htole64(ALIGN64(sizeof(h))); |
cec736d2 | 430 | |
d89c8fdf ZJS |
431 | h.incompatible_flags |= htole32( |
432 | f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ | | |
4ce534f4 | 433 | f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4 | |
8653185a | 434 | f->compress_zstd * HEADER_INCOMPATIBLE_COMPRESSED_ZSTD | |
4ce534f4 | 435 | f->keyed_hash * HEADER_INCOMPATIBLE_KEYED_HASH); |
7560fffc | 436 | |
d89c8fdf ZJS |
437 | h.compatible_flags = htole32( |
438 | f->seal * HEADER_COMPATIBLE_SEALED); | |
7560fffc | 439 | |
cec736d2 LP |
440 | r = sd_id128_randomize(&h.file_id); |
441 | if (r < 0) | |
442 | return r; | |
443 | ||
0ac38b70 LP |
444 | if (template) { |
445 | h.seqnum_id = template->header->seqnum_id; | |
beec0085 | 446 | h.tail_entry_seqnum = template->header->tail_entry_seqnum; |
0ac38b70 LP |
447 | } else |
448 | h.seqnum_id = h.file_id; | |
cec736d2 LP |
449 | |
450 | k = pwrite(f->fd, &h, sizeof(h), 0); | |
451 | if (k < 0) | |
452 | return -errno; | |
453 | ||
454 | if (k != sizeof(h)) | |
455 | return -EIO; | |
456 | ||
457 | return 0; | |
458 | } | |
459 | ||
460 | static int journal_file_refresh_header(JournalFile *f) { | |
fa6ac760 | 461 | int r; |
cec736d2 LP |
462 | |
463 | assert(f); | |
c88cc6af | 464 | assert(f->header); |
cec736d2 LP |
465 | |
466 | r = sd_id128_get_machine(&f->header->machine_id); | |
fd4885df ZJS |
467 | if (IN_SET(r, -ENOENT, -ENOMEDIUM)) |
468 | /* We don't have a machine-id, let's continue without */ | |
469 | zero(f->header->machine_id); | |
470 | else if (r < 0) | |
cec736d2 LP |
471 | return r; |
472 | ||
e958c057 | 473 | r = sd_id128_get_boot(&f->header->boot_id); |
cec736d2 LP |
474 | if (r < 0) |
475 | return r; | |
476 | ||
fa6ac760 | 477 | r = journal_file_set_online(f); |
b788cc23 | 478 | |
bf819d3a LP |
479 | /* Sync the online state to disk; likely just created a new file, also sync the directory this file |
480 | * is located in. */ | |
481 | (void) fsync_full(f->fd); | |
a0fe2a2d | 482 | |
fa6ac760 | 483 | return r; |
cec736d2 LP |
484 | } |
485 | ||
4214009f ZJS |
486 | static bool warn_wrong_flags(const JournalFile *f, bool compatible) { |
487 | const uint32_t any = compatible ? HEADER_COMPATIBLE_ANY : HEADER_INCOMPATIBLE_ANY, | |
488 | supported = compatible ? HEADER_COMPATIBLE_SUPPORTED : HEADER_INCOMPATIBLE_SUPPORTED; | |
489 | const char *type = compatible ? "compatible" : "incompatible"; | |
d89c8fdf ZJS |
490 | uint32_t flags; |
491 | ||
4214009f ZJS |
492 | flags = le32toh(compatible ? f->header->compatible_flags : f->header->incompatible_flags); |
493 | ||
494 | if (flags & ~supported) { | |
495 | if (flags & ~any) | |
4761fd0f | 496 | log_debug("Journal file %s has unknown %s flags 0x%"PRIx32, |
4214009f ZJS |
497 | f->path, type, flags & ~any); |
498 | flags = (flags & any) & ~supported; | |
4761fd0f | 499 | if (flags) { |
8653185a | 500 | const char* strv[5]; |
86e68f38 | 501 | size_t n = 0; |
4761fd0f ZJS |
502 | _cleanup_free_ char *t = NULL; |
503 | ||
4ce534f4 LP |
504 | if (compatible) { |
505 | if (flags & HEADER_COMPATIBLE_SEALED) | |
506 | strv[n++] = "sealed"; | |
507 | } else { | |
508 | if (flags & HEADER_INCOMPATIBLE_COMPRESSED_XZ) | |
509 | strv[n++] = "xz-compressed"; | |
510 | if (flags & HEADER_INCOMPATIBLE_COMPRESSED_LZ4) | |
511 | strv[n++] = "lz4-compressed"; | |
8653185a LP |
512 | if (flags & HEADER_INCOMPATIBLE_COMPRESSED_ZSTD) |
513 | strv[n++] = "zstd-compressed"; | |
4ce534f4 LP |
514 | if (flags & HEADER_INCOMPATIBLE_KEYED_HASH) |
515 | strv[n++] = "keyed-hash"; | |
516 | } | |
4761fd0f ZJS |
517 | strv[n] = NULL; |
518 | assert(n < ELEMENTSOF(strv)); | |
519 | ||
520 | t = strv_join((char**) strv, ", "); | |
521 | log_debug("Journal file %s uses %s %s %s disabled at compilation time.", | |
522 | f->path, type, n > 1 ? "flags" : "flag", strnull(t)); | |
523 | } | |
4214009f ZJS |
524 | return true; |
525 | } | |
526 | ||
527 | return false; | |
528 | } | |
529 | ||
530 | static int journal_file_verify_header(JournalFile *f) { | |
6f94e420 TS |
531 | uint64_t arena_size, header_size; |
532 | ||
cec736d2 | 533 | assert(f); |
c88cc6af | 534 | assert(f->header); |
cec736d2 | 535 | |
7560fffc | 536 | if (memcmp(f->header->signature, HEADER_SIGNATURE, 8)) |
cec736d2 LP |
537 | return -EBADMSG; |
538 | ||
4214009f ZJS |
539 | /* In both read and write mode we refuse to open files with incompatible |
540 | * flags we don't know. */ | |
541 | if (warn_wrong_flags(f, false)) | |
cec736d2 LP |
542 | return -EPROTONOSUPPORT; |
543 | ||
4214009f ZJS |
544 | /* When open for writing we refuse to open files with compatible flags, too. */ |
545 | if (f->writable && warn_wrong_flags(f, true)) | |
d89c8fdf | 546 | return -EPROTONOSUPPORT; |
7560fffc | 547 | |
db11ac1a LP |
548 | if (f->header->state >= _STATE_MAX) |
549 | return -EBADMSG; | |
550 | ||
893e0f8f | 551 | header_size = le64toh(READ_NOW(f->header->header_size)); |
6f94e420 | 552 | |
dca6219e | 553 | /* The first addition was n_data, so check that we are at least this large */ |
6f94e420 | 554 | if (header_size < HEADER_SIZE_MIN) |
23b0b2b2 LP |
555 | return -EBADMSG; |
556 | ||
8088cbd3 | 557 | if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
beec0085 LP |
558 | return -EBADMSG; |
559 | ||
893e0f8f | 560 | arena_size = le64toh(READ_NOW(f->header->arena_size)); |
6f94e420 TS |
561 | |
562 | if (UINT64_MAX - header_size < arena_size || header_size + arena_size > (uint64_t) f->last_stat.st_size) | |
db11ac1a LP |
563 | return -ENODATA; |
564 | ||
6f94e420 | 565 | if (le64toh(f->header->tail_object_offset) > header_size + arena_size) |
db11ac1a LP |
566 | return -ENODATA; |
567 | ||
7762e02b LP |
568 | if (!VALID64(le64toh(f->header->data_hash_table_offset)) || |
569 | !VALID64(le64toh(f->header->field_hash_table_offset)) || | |
570 | !VALID64(le64toh(f->header->tail_object_offset)) || | |
571 | !VALID64(le64toh(f->header->entry_array_offset))) | |
572 | return -ENODATA; | |
573 | ||
cec736d2 | 574 | if (f->writable) { |
cec736d2 | 575 | sd_id128_t machine_id; |
ae739cc1 | 576 | uint8_t state; |
cec736d2 LP |
577 | int r; |
578 | ||
579 | r = sd_id128_get_machine(&machine_id); | |
580 | if (r < 0) | |
581 | return r; | |
582 | ||
583 | if (!sd_id128_equal(machine_id, f->header->machine_id)) | |
584 | return -EHOSTDOWN; | |
585 | ||
de190aef | 586 | state = f->header->state; |
cec736d2 | 587 | |
b288cdeb ZJS |
588 | if (state == STATE_ARCHIVED) |
589 | return -ESHUTDOWN; /* Already archived */ | |
baaa35ad ZJS |
590 | else if (state == STATE_ONLINE) |
591 | return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), | |
592 | "Journal file %s is already online. Assuming unclean closing.", | |
593 | f->path); | |
594 | else if (state != STATE_OFFLINE) | |
595 | return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), | |
596 | "Journal file %s has unknown state %i.", | |
597 | f->path, state); | |
ae739cc1 | 598 | |
5b3cc0c8 YN |
599 | if (f->header->field_hash_table_size == 0 || f->header->data_hash_table_size == 0) |
600 | return -EBADMSG; | |
601 | ||
ae739cc1 LP |
602 | /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't |
603 | * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks | |
604 | * bisection. */ | |
baaa35ad ZJS |
605 | if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) |
606 | return log_debug_errno(SYNTHETIC_ERRNO(ETXTBSY), | |
607 | "Journal file %s is from the future, refusing to append new data to it that'd be older.", | |
608 | f->path); | |
cec736d2 LP |
609 | } |
610 | ||
d89c8fdf ZJS |
611 | f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header); |
612 | f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header); | |
8653185a | 613 | f->compress_zstd = JOURNAL_HEADER_COMPRESSED_ZSTD(f->header); |
c586dbf1 | 614 | |
f1889c91 | 615 | f->seal = JOURNAL_HEADER_SEALED(f->header); |
7560fffc | 616 | |
4ce534f4 LP |
617 | f->keyed_hash = JOURNAL_HEADER_KEYED_HASH(f->header); |
618 | ||
cec736d2 LP |
619 | return 0; |
620 | } | |
621 | ||
28ca867a | 622 | int journal_file_fstat(JournalFile *f) { |
3cc44114 LP |
623 | int r; |
624 | ||
2678031a LP |
625 | assert(f); |
626 | assert(f->fd >= 0); | |
627 | ||
628 | if (fstat(f->fd, &f->last_stat) < 0) | |
629 | return -errno; | |
630 | ||
631 | f->last_stat_usec = now(CLOCK_MONOTONIC); | |
632 | ||
e9dd6984 | 633 | /* Refuse dealing with files that aren't regular */ |
3cc44114 LP |
634 | r = stat_verify_regular(&f->last_stat); |
635 | if (r < 0) | |
636 | return r; | |
8d6a4d33 | 637 | |
2678031a LP |
638 | /* Refuse appending to files that are already deleted */ |
639 | if (f->last_stat.st_nlink <= 0) | |
640 | return -EIDRM; | |
641 | ||
642 | return 0; | |
643 | } | |
644 | ||
cec736d2 | 645 | static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { |
893e0f8f | 646 | uint64_t old_size, new_size, old_header_size, old_arena_size; |
fec2aa2f | 647 | int r; |
cec736d2 LP |
648 | |
649 | assert(f); | |
c88cc6af | 650 | assert(f->header); |
cec736d2 | 651 | |
893e0f8f LP |
652 | /* We assume that this file is not sparse, and we know that for sure, since we always call |
653 | * posix_fallocate() ourselves */ | |
654 | ||
655 | if (size > PAGE_ALIGN_DOWN(UINT64_MAX) - offset) | |
656 | return -EINVAL; | |
cec736d2 | 657 | |
c3bd54bf | 658 | if (mmap_cache_fd_got_sigbus(f->cache_fd)) |
fa6ac760 LP |
659 | return -EIO; |
660 | ||
893e0f8f LP |
661 | old_header_size = le64toh(READ_NOW(f->header->header_size)); |
662 | old_arena_size = le64toh(READ_NOW(f->header->arena_size)); | |
663 | if (old_arena_size > PAGE_ALIGN_DOWN(UINT64_MAX) - old_header_size) | |
664 | return -EBADMSG; | |
665 | ||
666 | old_size = old_header_size + old_arena_size; | |
cec736d2 | 667 | |
893e0f8f | 668 | new_size = MAX(PAGE_ALIGN(offset + size), old_header_size); |
bc85bfee | 669 | |
2678031a LP |
670 | if (new_size <= old_size) { |
671 | ||
672 | /* We already pre-allocated enough space, but before | |
673 | * we write to it, let's check with fstat() if the | |
674 | * file got deleted, in order make sure we don't throw | |
675 | * away the data immediately. Don't check fstat() for | |
676 | * all writes though, but only once ever 10s. */ | |
677 | ||
678 | if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC)) | |
679 | return 0; | |
680 | ||
681 | return journal_file_fstat(f); | |
682 | } | |
683 | ||
684 | /* Allocate more space. */ | |
cec736d2 | 685 | |
a676e665 | 686 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) |
bc85bfee | 687 | return -E2BIG; |
cec736d2 | 688 | |
a676e665 | 689 | if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) { |
cec736d2 LP |
690 | struct statvfs svfs; |
691 | ||
692 | if (fstatvfs(f->fd, &svfs) >= 0) { | |
693 | uint64_t available; | |
694 | ||
070052ab | 695 | available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free); |
cec736d2 LP |
696 | |
697 | if (new_size - old_size > available) | |
698 | return -E2BIG; | |
699 | } | |
700 | } | |
701 | ||
eda4b58b | 702 | /* Increase by larger blocks at once */ |
be6b0c21 | 703 | new_size = DIV_ROUND_UP(new_size, FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE; |
eda4b58b LP |
704 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) |
705 | new_size = f->metrics.max_size; | |
706 | ||
bc85bfee LP |
707 | /* Note that the glibc fallocate() fallback is very |
708 | inefficient, hence we try to minimize the allocation area | |
709 | as we can. */ | |
4c54768c IZ |
710 | r = posix_fallocate_loop(f->fd, old_size, new_size - old_size); |
711 | if (r < 0) | |
712 | return r; | |
cec736d2 | 713 | |
893e0f8f | 714 | f->header->arena_size = htole64(new_size - old_header_size); |
cec736d2 | 715 | |
2678031a | 716 | return journal_file_fstat(f); |
cec736d2 LP |
717 | } |
718 | ||
78519831 | 719 | static unsigned type_to_context(ObjectType type) { |
d3d3208f | 720 | /* One context for each type, plus one catch-all for the rest */ |
69adae51 | 721 | assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS); |
fa6ac760 | 722 | assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS); |
d05089d8 | 723 | return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0; |
d3d3208f MS |
724 | } |
725 | ||
71139898 LP |
726 | static int journal_file_move_to( |
727 | JournalFile *f, | |
728 | ObjectType type, | |
729 | bool keep_always, | |
730 | uint64_t offset, | |
731 | uint64_t size, | |
258190a0 | 732 | void **ret) { |
71139898 | 733 | |
2678031a LP |
734 | int r; |
735 | ||
cec736d2 | 736 | assert(f); |
cec736d2 LP |
737 | assert(ret); |
738 | ||
7762e02b LP |
739 | if (size <= 0) |
740 | return -EINVAL; | |
741 | ||
893e0f8f LP |
742 | if (size > UINT64_MAX - offset) |
743 | return -EBADMSG; | |
744 | ||
2a59ea54 | 745 | /* Avoid SIGBUS on invalid accesses */ |
4bbdcdb3 LP |
746 | if (offset + size > (uint64_t) f->last_stat.st_size) { |
747 | /* Hmm, out of range? Let's refresh the fstat() data | |
748 | * first, before we trust that check. */ | |
749 | ||
2678031a LP |
750 | r = journal_file_fstat(f); |
751 | if (r < 0) | |
752 | return r; | |
753 | ||
754 | if (offset + size > (uint64_t) f->last_stat.st_size) | |
4bbdcdb3 LP |
755 | return -EADDRNOTAVAIL; |
756 | } | |
757 | ||
c3bd54bf | 758 | return mmap_cache_fd_get(f->cache_fd, type_to_context(type), keep_always, offset, size, &f->last_stat, ret); |
cec736d2 LP |
759 | } |
760 | ||
16e9f408 LP |
761 | static uint64_t minimum_header_size(Object *o) { |
762 | ||
b8e891e6 | 763 | static const uint64_t table[] = { |
16e9f408 LP |
764 | [OBJECT_DATA] = sizeof(DataObject), |
765 | [OBJECT_FIELD] = sizeof(FieldObject), | |
766 | [OBJECT_ENTRY] = sizeof(EntryObject), | |
767 | [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), | |
768 | [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), | |
769 | [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), | |
770 | [OBJECT_TAG] = sizeof(TagObject), | |
771 | }; | |
772 | ||
773 | if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) | |
774 | return sizeof(ObjectHeader); | |
775 | ||
776 | return table[o->object.type]; | |
777 | } | |
778 | ||
24754f36 TR |
779 | /* Lightweight object checks. We want this to be fast, so that we won't |
780 | * slowdown every journal_file_move_to_object() call too much. */ | |
781 | static int journal_file_check_object(JournalFile *f, uint64_t offset, Object *o) { | |
782 | assert(f); | |
783 | assert(o); | |
784 | ||
785 | switch (o->object.type) { | |
786 | ||
a602d93e | 787 | case OBJECT_DATA: |
baaa35ad ZJS |
788 | if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) |
789 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
790 | "Bad n_entries: %" PRIu64 ": %" PRIu64, | |
791 | le64toh(o->data.n_entries), | |
792 | offset); | |
793 | ||
20ee282b | 794 | if (le64toh(o->object.size) <= offsetof(DataObject, payload)) |
baaa35ad ZJS |
795 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
796 | "Bad object size (<= %zu): %" PRIu64 ": %" PRIu64, | |
797 | offsetof(DataObject, payload), | |
798 | le64toh(o->object.size), | |
799 | offset); | |
24754f36 | 800 | |
10e8445b TR |
801 | if (!VALID64(le64toh(o->data.next_hash_offset)) || |
802 | !VALID64(le64toh(o->data.next_field_offset)) || | |
803 | !VALID64(le64toh(o->data.entry_offset)) || | |
baaa35ad ZJS |
804 | !VALID64(le64toh(o->data.entry_array_offset))) |
805 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
806 | "Invalid offset, next_hash_offset=" OFSfmt ", next_field_offset=" OFSfmt ", entry_offset=" OFSfmt ", entry_array_offset=" OFSfmt ": %" PRIu64, | |
807 | le64toh(o->data.next_hash_offset), | |
808 | le64toh(o->data.next_field_offset), | |
809 | le64toh(o->data.entry_offset), | |
810 | le64toh(o->data.entry_array_offset), | |
811 | offset); | |
24754f36 TR |
812 | |
813 | break; | |
24754f36 TR |
814 | |
815 | case OBJECT_FIELD: | |
20ee282b | 816 | if (le64toh(o->object.size) <= offsetof(FieldObject, payload)) |
baaa35ad ZJS |
817 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
818 | "Bad field size (<= %zu): %" PRIu64 ": %" PRIu64, | |
819 | offsetof(FieldObject, payload), | |
820 | le64toh(o->object.size), | |
821 | offset); | |
24754f36 | 822 | |
10e8445b | 823 | if (!VALID64(le64toh(o->field.next_hash_offset)) || |
baaa35ad ZJS |
824 | !VALID64(le64toh(o->field.head_data_offset))) |
825 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
826 | "Invalid offset, next_hash_offset=" OFSfmt ", head_data_offset=" OFSfmt ": %" PRIu64, | |
827 | le64toh(o->field.next_hash_offset), | |
828 | le64toh(o->field.head_data_offset), | |
829 | offset); | |
24754f36 TR |
830 | break; |
831 | ||
893e0f8f LP |
832 | case OBJECT_ENTRY: { |
833 | uint64_t sz; | |
834 | ||
835 | sz = le64toh(READ_NOW(o->object.size)); | |
836 | if (sz < offsetof(EntryObject, items) || | |
837 | (sz - offsetof(EntryObject, items)) % sizeof(EntryItem) != 0) | |
baaa35ad ZJS |
838 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
839 | "Bad entry size (<= %zu): %" PRIu64 ": %" PRIu64, | |
840 | offsetof(EntryObject, items), | |
893e0f8f | 841 | sz, |
baaa35ad ZJS |
842 | offset); |
843 | ||
893e0f8f | 844 | if ((sz - offsetof(EntryObject, items)) / sizeof(EntryItem) <= 0) |
baaa35ad ZJS |
845 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
846 | "Invalid number items in entry: %" PRIu64 ": %" PRIu64, | |
893e0f8f | 847 | (sz - offsetof(EntryObject, items)) / sizeof(EntryItem), |
baaa35ad ZJS |
848 | offset); |
849 | ||
850 | if (le64toh(o->entry.seqnum) <= 0) | |
851 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
852 | "Invalid entry seqnum: %" PRIx64 ": %" PRIu64, | |
853 | le64toh(o->entry.seqnum), | |
854 | offset); | |
855 | ||
856 | if (!VALID_REALTIME(le64toh(o->entry.realtime))) | |
857 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
858 | "Invalid entry realtime timestamp: %" PRIu64 ": %" PRIu64, | |
859 | le64toh(o->entry.realtime), | |
860 | offset); | |
861 | ||
862 | if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) | |
863 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
864 | "Invalid entry monotonic timestamp: %" PRIu64 ": %" PRIu64, | |
865 | le64toh(o->entry.monotonic), | |
866 | offset); | |
24754f36 TR |
867 | |
868 | break; | |
893e0f8f | 869 | } |
24754f36 TR |
870 | |
871 | case OBJECT_DATA_HASH_TABLE: | |
893e0f8f LP |
872 | case OBJECT_FIELD_HASH_TABLE: { |
873 | uint64_t sz; | |
874 | ||
875 | sz = le64toh(READ_NOW(o->object.size)); | |
876 | if (sz < offsetof(HashTableObject, items) || | |
877 | (sz - offsetof(HashTableObject, items)) % sizeof(HashItem) != 0 || | |
878 | (sz - offsetof(HashTableObject, items)) / sizeof(HashItem) <= 0) | |
baaa35ad ZJS |
879 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
880 | "Invalid %s hash table size: %" PRIu64 ": %" PRIu64, | |
881 | o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", | |
893e0f8f | 882 | sz, |
baaa35ad | 883 | offset); |
24754f36 TR |
884 | |
885 | break; | |
893e0f8f | 886 | } |
24754f36 | 887 | |
893e0f8f LP |
888 | case OBJECT_ENTRY_ARRAY: { |
889 | uint64_t sz; | |
890 | ||
891 | sz = le64toh(READ_NOW(o->object.size)); | |
892 | if (sz < offsetof(EntryArrayObject, items) || | |
893 | (sz - offsetof(EntryArrayObject, items)) % sizeof(le64_t) != 0 || | |
894 | (sz - offsetof(EntryArrayObject, items)) / sizeof(le64_t) <= 0) | |
baaa35ad ZJS |
895 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), |
896 | "Invalid object entry array size: %" PRIu64 ": %" PRIu64, | |
893e0f8f | 897 | sz, |
baaa35ad ZJS |
898 | offset); |
899 | ||
900 | if (!VALID64(le64toh(o->entry_array.next_entry_array_offset))) | |
901 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
902 | "Invalid object entry array next_entry_array_offset: " OFSfmt ": %" PRIu64, | |
903 | le64toh(o->entry_array.next_entry_array_offset), | |
904 | offset); | |
24754f36 TR |
905 | |
906 | break; | |
893e0f8f | 907 | } |
24754f36 TR |
908 | |
909 | case OBJECT_TAG: | |
baaa35ad ZJS |
910 | if (le64toh(o->object.size) != sizeof(TagObject)) |
911 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
912 | "Invalid object tag size: %" PRIu64 ": %" PRIu64, | |
913 | le64toh(o->object.size), | |
914 | offset); | |
24754f36 | 915 | |
baaa35ad ZJS |
916 | if (!VALID_EPOCH(le64toh(o->tag.epoch))) |
917 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
918 | "Invalid object tag epoch: %" PRIu64 ": %" PRIu64, | |
919 | le64toh(o->tag.epoch), offset); | |
24754f36 TR |
920 | |
921 | break; | |
922 | } | |
923 | ||
924 | return 0; | |
925 | } | |
926 | ||
78519831 | 927 | int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) { |
cec736d2 LP |
928 | int r; |
929 | void *t; | |
930 | Object *o; | |
931 | uint64_t s; | |
932 | ||
933 | assert(f); | |
934 | assert(ret); | |
935 | ||
db11ac1a | 936 | /* Objects may only be located at multiple of 64 bit */ |
baaa35ad ZJS |
937 | if (!VALID64(offset)) |
938 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
939 | "Attempt to move to object at non-64bit boundary: %" PRIu64, | |
940 | offset); | |
db11ac1a | 941 | |
50809d7a | 942 | /* Object may not be located in the file header */ |
baaa35ad ZJS |
943 | if (offset < le64toh(f->header->header_size)) |
944 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
945 | "Attempt to move to object located in file header: %" PRIu64, | |
946 | offset); | |
50809d7a | 947 | |
258190a0 | 948 | r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t); |
cec736d2 LP |
949 | if (r < 0) |
950 | return r; | |
951 | ||
952 | o = (Object*) t; | |
893e0f8f | 953 | s = le64toh(READ_NOW(o->object.size)); |
cec736d2 | 954 | |
baaa35ad ZJS |
955 | if (s == 0) |
956 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
957 | "Attempt to move to uninitialized object: %" PRIu64, | |
958 | offset); | |
959 | if (s < sizeof(ObjectHeader)) | |
960 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
961 | "Attempt to move to overly short object: %" PRIu64, | |
962 | offset); | |
963 | ||
964 | if (o->object.type <= OBJECT_UNUSED) | |
965 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
966 | "Attempt to move to object with invalid type: %" PRIu64, | |
967 | offset); | |
968 | ||
969 | if (s < minimum_header_size(o)) | |
970 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
971 | "Attempt to move to truncated object: %" PRIu64, | |
972 | offset); | |
973 | ||
974 | if (type > OBJECT_UNUSED && o->object.type != type) | |
975 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
976 | "Attempt to move to object of unexpected type: %" PRIu64, | |
977 | offset); | |
cec736d2 | 978 | |
258190a0 VC |
979 | r = journal_file_move_to(f, type, false, offset, s, &t); |
980 | if (r < 0) | |
981 | return r; | |
cec736d2 | 982 | |
258190a0 | 983 | o = (Object*) t; |
cec736d2 | 984 | |
24754f36 TR |
985 | r = journal_file_check_object(f, offset, o); |
986 | if (r < 0) | |
987 | return r; | |
988 | ||
cec736d2 LP |
989 | *ret = o; |
990 | return 0; | |
991 | } | |
992 | ||
0eaee828 LP |
993 | static uint64_t journal_file_entry_seqnum( |
994 | JournalFile *f, | |
995 | uint64_t *seqnum) { | |
996 | ||
997 | uint64_t ret; | |
cec736d2 LP |
998 | |
999 | assert(f); | |
c88cc6af | 1000 | assert(f->header); |
cec736d2 | 1001 | |
0eaee828 LP |
1002 | /* Picks a new sequence number for the entry we are about to add and returns it. */ |
1003 | ||
1004 | ret = le64toh(f->header->tail_entry_seqnum) + 1; | |
c2373f84 LP |
1005 | |
1006 | if (seqnum) { | |
0eaee828 LP |
1007 | /* If an external seqnum counter was passed, we update both the local and the external one, |
1008 | * and set it to the maximum of both */ | |
c2373f84 | 1009 | |
0eaee828 LP |
1010 | if (*seqnum + 1 > ret) |
1011 | ret = *seqnum + 1; | |
c2373f84 | 1012 | |
0eaee828 | 1013 | *seqnum = ret; |
c2373f84 LP |
1014 | } |
1015 | ||
0eaee828 | 1016 | f->header->tail_entry_seqnum = htole64(ret); |
cec736d2 | 1017 | |
beec0085 | 1018 | if (f->header->head_entry_seqnum == 0) |
0eaee828 | 1019 | f->header->head_entry_seqnum = htole64(ret); |
de190aef | 1020 | |
0eaee828 LP |
1021 | return ret; |
1022 | } | |
1023 | ||
f4474e00 LP |
1024 | int journal_file_append_object( |
1025 | JournalFile *f, | |
1026 | ObjectType type, | |
1027 | uint64_t size, | |
1028 | Object **ret, | |
1029 | uint64_t *ret_offset) { | |
1030 | ||
cec736d2 LP |
1031 | int r; |
1032 | uint64_t p; | |
0aa649b1 | 1033 | Object *tail, *o; |
cec736d2 LP |
1034 | void *t; |
1035 | ||
1036 | assert(f); | |
c88cc6af | 1037 | assert(f->header); |
d05089d8 | 1038 | assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX); |
cec736d2 | 1039 | assert(size >= sizeof(ObjectHeader)); |
cec736d2 | 1040 | |
26687bf8 OS |
1041 | r = journal_file_set_online(f); |
1042 | if (r < 0) | |
1043 | return r; | |
1044 | ||
0aa649b1 YW |
1045 | p = le64toh(f->header->tail_object_offset); |
1046 | if (p == 0) | |
1047 | p = le64toh(f->header->header_size); | |
1048 | else { | |
1049 | uint64_t sz; | |
1050 | ||
1051 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail); | |
1052 | if (r < 0) | |
1053 | return r; | |
1054 | ||
1055 | sz = le64toh(READ_NOW(tail->object.size)); | |
1056 | if (sz > UINT64_MAX - sizeof(uint64_t) + 1) | |
1057 | return -EBADMSG; | |
1058 | ||
1059 | sz = ALIGN64(sz); | |
1060 | if (p > UINT64_MAX - sz) | |
1061 | return -EBADMSG; | |
1062 | ||
1063 | p += sz; | |
1064 | } | |
cec736d2 LP |
1065 | |
1066 | r = journal_file_allocate(f, p, size); | |
1067 | if (r < 0) | |
1068 | return r; | |
1069 | ||
258190a0 | 1070 | r = journal_file_move_to(f, type, false, p, size, &t); |
cec736d2 LP |
1071 | if (r < 0) |
1072 | return r; | |
1073 | ||
1074 | o = (Object*) t; | |
71139898 LP |
1075 | o->object = (ObjectHeader) { |
1076 | .type = type, | |
1077 | .size = htole64(size), | |
1078 | }; | |
cec736d2 LP |
1079 | |
1080 | f->header->tail_object_offset = htole64(p); | |
cec736d2 LP |
1081 | f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); |
1082 | ||
f4474e00 LP |
1083 | if (ret) |
1084 | *ret = o; | |
1085 | ||
1086 | if (ret_offset) | |
1087 | *ret_offset = p; | |
cec736d2 LP |
1088 | |
1089 | return 0; | |
1090 | } | |
1091 | ||
de190aef | 1092 | static int journal_file_setup_data_hash_table(JournalFile *f) { |
cec736d2 LP |
1093 | uint64_t s, p; |
1094 | Object *o; | |
1095 | int r; | |
1096 | ||
1097 | assert(f); | |
c88cc6af | 1098 | assert(f->header); |
cec736d2 | 1099 | |
070052ab LP |
1100 | /* We estimate that we need 1 hash table entry per 768 bytes |
1101 | of journal file and we want to make sure we never get | |
1102 | beyond 75% fill level. Calculate the hash table size for | |
1103 | the maximum file size based on these metrics. */ | |
4a92baf3 | 1104 | |
dfabe643 | 1105 | s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); |
4a92baf3 LP |
1106 | if (s < DEFAULT_DATA_HASH_TABLE_SIZE) |
1107 | s = DEFAULT_DATA_HASH_TABLE_SIZE; | |
1108 | ||
5030c85a | 1109 | log_debug("Reserving %"PRIu64" entries in data hash table.", s / sizeof(HashItem)); |
4a92baf3 | 1110 | |
de190aef LP |
1111 | r = journal_file_append_object(f, |
1112 | OBJECT_DATA_HASH_TABLE, | |
1113 | offsetof(Object, hash_table.items) + s, | |
1114 | &o, &p); | |
cec736d2 LP |
1115 | if (r < 0) |
1116 | return r; | |
1117 | ||
29804cc1 | 1118 | memzero(o->hash_table.items, s); |
cec736d2 | 1119 | |
de190aef LP |
1120 | f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
1121 | f->header->data_hash_table_size = htole64(s); | |
cec736d2 LP |
1122 | |
1123 | return 0; | |
1124 | } | |
1125 | ||
de190aef | 1126 | static int journal_file_setup_field_hash_table(JournalFile *f) { |
cec736d2 LP |
1127 | uint64_t s, p; |
1128 | Object *o; | |
1129 | int r; | |
1130 | ||
1131 | assert(f); | |
c88cc6af | 1132 | assert(f->header); |
cec736d2 | 1133 | |
3c1668da LP |
1134 | /* We use a fixed size hash table for the fields as this |
1135 | * number should grow very slowly only */ | |
1136 | ||
de190aef | 1137 | s = DEFAULT_FIELD_HASH_TABLE_SIZE; |
5030c85a LP |
1138 | log_debug("Reserving %"PRIu64" entries in field hash table.", s / sizeof(HashItem)); |
1139 | ||
de190aef LP |
1140 | r = journal_file_append_object(f, |
1141 | OBJECT_FIELD_HASH_TABLE, | |
1142 | offsetof(Object, hash_table.items) + s, | |
1143 | &o, &p); | |
cec736d2 LP |
1144 | if (r < 0) |
1145 | return r; | |
1146 | ||
29804cc1 | 1147 | memzero(o->hash_table.items, s); |
cec736d2 | 1148 | |
de190aef LP |
1149 | f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
1150 | f->header->field_hash_table_size = htole64(s); | |
cec736d2 LP |
1151 | |
1152 | return 0; | |
1153 | } | |
1154 | ||
dade37d4 | 1155 | int journal_file_map_data_hash_table(JournalFile *f) { |
cec736d2 LP |
1156 | uint64_t s, p; |
1157 | void *t; | |
1158 | int r; | |
1159 | ||
1160 | assert(f); | |
c88cc6af | 1161 | assert(f->header); |
cec736d2 | 1162 | |
dade37d4 LP |
1163 | if (f->data_hash_table) |
1164 | return 0; | |
1165 | ||
de190aef LP |
1166 | p = le64toh(f->header->data_hash_table_offset); |
1167 | s = le64toh(f->header->data_hash_table_size); | |
cec736d2 | 1168 | |
de190aef | 1169 | r = journal_file_move_to(f, |
16e9f408 | 1170 | OBJECT_DATA_HASH_TABLE, |
fcde2389 | 1171 | true, |
de190aef | 1172 | p, s, |
258190a0 | 1173 | &t); |
cec736d2 LP |
1174 | if (r < 0) |
1175 | return r; | |
1176 | ||
de190aef | 1177 | f->data_hash_table = t; |
cec736d2 LP |
1178 | return 0; |
1179 | } | |
1180 | ||
dade37d4 | 1181 | int journal_file_map_field_hash_table(JournalFile *f) { |
cec736d2 LP |
1182 | uint64_t s, p; |
1183 | void *t; | |
1184 | int r; | |
1185 | ||
1186 | assert(f); | |
c88cc6af | 1187 | assert(f->header); |
cec736d2 | 1188 | |
dade37d4 LP |
1189 | if (f->field_hash_table) |
1190 | return 0; | |
1191 | ||
de190aef LP |
1192 | p = le64toh(f->header->field_hash_table_offset); |
1193 | s = le64toh(f->header->field_hash_table_size); | |
cec736d2 | 1194 | |
de190aef | 1195 | r = journal_file_move_to(f, |
16e9f408 | 1196 | OBJECT_FIELD_HASH_TABLE, |
fcde2389 | 1197 | true, |
de190aef | 1198 | p, s, |
258190a0 | 1199 | &t); |
cec736d2 LP |
1200 | if (r < 0) |
1201 | return r; | |
1202 | ||
de190aef | 1203 | f->field_hash_table = t; |
cec736d2 LP |
1204 | return 0; |
1205 | } | |
1206 | ||
3c1668da LP |
1207 | static int journal_file_link_field( |
1208 | JournalFile *f, | |
1209 | Object *o, | |
1210 | uint64_t offset, | |
1211 | uint64_t hash) { | |
1212 | ||
805d1486 | 1213 | uint64_t p, h, m; |
3c1668da LP |
1214 | int r; |
1215 | ||
1216 | assert(f); | |
c88cc6af | 1217 | assert(f->header); |
90d222c1 | 1218 | assert(f->field_hash_table); |
3c1668da LP |
1219 | assert(o); |
1220 | assert(offset > 0); | |
1221 | ||
1222 | if (o->object.type != OBJECT_FIELD) | |
1223 | return -EINVAL; | |
1224 | ||
893e0f8f | 1225 | m = le64toh(READ_NOW(f->header->field_hash_table_size)) / sizeof(HashItem); |
805d1486 LP |
1226 | if (m <= 0) |
1227 | return -EBADMSG; | |
3c1668da | 1228 | |
805d1486 | 1229 | /* This might alter the window we are looking at */ |
3c1668da LP |
1230 | o->field.next_hash_offset = o->field.head_data_offset = 0; |
1231 | ||
805d1486 | 1232 | h = hash % m; |
3c1668da LP |
1233 | p = le64toh(f->field_hash_table[h].tail_hash_offset); |
1234 | if (p == 0) | |
1235 | f->field_hash_table[h].head_hash_offset = htole64(offset); | |
1236 | else { | |
1237 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1238 | if (r < 0) | |
1239 | return r; | |
1240 | ||
1241 | o->field.next_hash_offset = htole64(offset); | |
1242 | } | |
1243 | ||
1244 | f->field_hash_table[h].tail_hash_offset = htole64(offset); | |
1245 | ||
1246 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
1247 | f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1); | |
1248 | ||
1249 | return 0; | |
1250 | } | |
1251 | ||
1252 | static int journal_file_link_data( | |
1253 | JournalFile *f, | |
1254 | Object *o, | |
1255 | uint64_t offset, | |
1256 | uint64_t hash) { | |
1257 | ||
805d1486 | 1258 | uint64_t p, h, m; |
cec736d2 LP |
1259 | int r; |
1260 | ||
1261 | assert(f); | |
c88cc6af | 1262 | assert(f->header); |
90d222c1 | 1263 | assert(f->data_hash_table); |
cec736d2 LP |
1264 | assert(o); |
1265 | assert(offset > 0); | |
b588975f LP |
1266 | |
1267 | if (o->object.type != OBJECT_DATA) | |
1268 | return -EINVAL; | |
cec736d2 | 1269 | |
893e0f8f | 1270 | m = le64toh(READ_NOW(f->header->data_hash_table_size)) / sizeof(HashItem); |
805d1486 LP |
1271 | if (m <= 0) |
1272 | return -EBADMSG; | |
48496df6 | 1273 | |
805d1486 | 1274 | /* This might alter the window we are looking at */ |
de190aef LP |
1275 | o->data.next_hash_offset = o->data.next_field_offset = 0; |
1276 | o->data.entry_offset = o->data.entry_array_offset = 0; | |
1277 | o->data.n_entries = 0; | |
cec736d2 | 1278 | |
805d1486 | 1279 | h = hash % m; |
8db4213e | 1280 | p = le64toh(f->data_hash_table[h].tail_hash_offset); |
3c1668da | 1281 | if (p == 0) |
cec736d2 | 1282 | /* Only entry in the hash table is easy */ |
de190aef | 1283 | f->data_hash_table[h].head_hash_offset = htole64(offset); |
3c1668da | 1284 | else { |
48496df6 LP |
1285 | /* Move back to the previous data object, to patch in |
1286 | * pointer */ | |
cec736d2 | 1287 | |
de190aef | 1288 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1289 | if (r < 0) |
1290 | return r; | |
1291 | ||
de190aef | 1292 | o->data.next_hash_offset = htole64(offset); |
cec736d2 LP |
1293 | } |
1294 | ||
de190aef | 1295 | f->data_hash_table[h].tail_hash_offset = htole64(offset); |
cec736d2 | 1296 | |
dca6219e LP |
1297 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
1298 | f->header->n_data = htole64(le64toh(f->header->n_data) + 1); | |
1299 | ||
cec736d2 LP |
1300 | return 0; |
1301 | } | |
1302 | ||
0dbe57ee LP |
1303 | static int next_hash_offset( |
1304 | JournalFile *f, | |
1305 | uint64_t *p, | |
1306 | le64_t *next_hash_offset, | |
1307 | uint64_t *depth, | |
1308 | le64_t *header_max_depth) { | |
1309 | ||
1310 | uint64_t nextp; | |
1311 | ||
1312 | nextp = le64toh(READ_NOW(*next_hash_offset)); | |
1313 | if (nextp > 0) { | |
1314 | if (nextp <= *p) /* Refuse going in loops */ | |
1315 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
1316 | "Detected hash item loop in %s, refusing.", f->path); | |
1317 | ||
1318 | (*depth)++; | |
1319 | ||
1320 | /* If the depth of this hash chain is larger than all others we have seen so far, record it */ | |
1321 | if (header_max_depth && f->writable) | |
1322 | *header_max_depth = htole64(MAX(*depth, le64toh(*header_max_depth))); | |
1323 | } | |
1324 | ||
1325 | *p = nextp; | |
1326 | return 0; | |
1327 | } | |
1328 | ||
3c1668da LP |
1329 | int journal_file_find_field_object_with_hash( |
1330 | JournalFile *f, | |
1331 | const void *field, uint64_t size, uint64_t hash, | |
f4474e00 | 1332 | Object **ret, uint64_t *ret_offset) { |
3c1668da | 1333 | |
0dbe57ee | 1334 | uint64_t p, osize, h, m, depth = 0; |
3c1668da LP |
1335 | int r; |
1336 | ||
1337 | assert(f); | |
c88cc6af | 1338 | assert(f->header); |
3c1668da LP |
1339 | assert(field && size > 0); |
1340 | ||
dade37d4 LP |
1341 | /* If the field hash table is empty, we can't find anything */ |
1342 | if (le64toh(f->header->field_hash_table_size) <= 0) | |
1343 | return 0; | |
1344 | ||
1345 | /* Map the field hash table, if it isn't mapped yet. */ | |
1346 | r = journal_file_map_field_hash_table(f); | |
1347 | if (r < 0) | |
1348 | return r; | |
1349 | ||
3c1668da LP |
1350 | osize = offsetof(Object, field.payload) + size; |
1351 | ||
893e0f8f | 1352 | m = le64toh(READ_NOW(f->header->field_hash_table_size)) / sizeof(HashItem); |
805d1486 | 1353 | if (m <= 0) |
3c1668da LP |
1354 | return -EBADMSG; |
1355 | ||
805d1486 | 1356 | h = hash % m; |
3c1668da | 1357 | p = le64toh(f->field_hash_table[h].head_hash_offset); |
3c1668da LP |
1358 | while (p > 0) { |
1359 | Object *o; | |
1360 | ||
1361 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1362 | if (r < 0) | |
1363 | return r; | |
1364 | ||
1365 | if (le64toh(o->field.hash) == hash && | |
1366 | le64toh(o->object.size) == osize && | |
1367 | memcmp(o->field.payload, field, size) == 0) { | |
1368 | ||
1369 | if (ret) | |
1370 | *ret = o; | |
f4474e00 LP |
1371 | if (ret_offset) |
1372 | *ret_offset = p; | |
3c1668da LP |
1373 | |
1374 | return 1; | |
1375 | } | |
1376 | ||
0dbe57ee LP |
1377 | r = next_hash_offset( |
1378 | f, | |
1379 | &p, | |
1380 | &o->field.next_hash_offset, | |
1381 | &depth, | |
1382 | JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth) ? &f->header->field_hash_chain_depth : NULL); | |
1383 | if (r < 0) | |
1384 | return r; | |
3c1668da LP |
1385 | } |
1386 | ||
1387 | return 0; | |
1388 | } | |
1389 | ||
4ce534f4 LP |
1390 | uint64_t journal_file_hash_data( |
1391 | JournalFile *f, | |
1392 | const void *data, | |
1393 | size_t sz) { | |
1394 | ||
1395 | assert(f); | |
1396 | assert(data || sz == 0); | |
1397 | ||
1398 | /* We try to unify our codebase on siphash, hence new-styled journal files utilizing the keyed hash | |
1399 | * function use siphash. Old journal files use the Jenkins hash. */ | |
1400 | ||
1401 | if (JOURNAL_HEADER_KEYED_HASH(f->header)) | |
1402 | return siphash24(data, sz, f->header->file_id.bytes); | |
1403 | ||
1404 | return jenkins_hash64(data, sz); | |
1405 | } | |
1406 | ||
3c1668da LP |
1407 | int journal_file_find_field_object( |
1408 | JournalFile *f, | |
1409 | const void *field, uint64_t size, | |
f4474e00 | 1410 | Object **ret, uint64_t *ret_offset) { |
3c1668da | 1411 | |
3c1668da LP |
1412 | assert(f); |
1413 | assert(field && size > 0); | |
1414 | ||
f4474e00 LP |
1415 | return journal_file_find_field_object_with_hash( |
1416 | f, | |
4ce534f4 LP |
1417 | field, size, |
1418 | journal_file_hash_data(f, field, size), | |
f4474e00 | 1419 | ret, ret_offset); |
3c1668da LP |
1420 | } |
1421 | ||
de190aef LP |
1422 | int journal_file_find_data_object_with_hash( |
1423 | JournalFile *f, | |
1424 | const void *data, uint64_t size, uint64_t hash, | |
f4474e00 | 1425 | Object **ret, uint64_t *ret_offset) { |
48496df6 | 1426 | |
0dbe57ee | 1427 | uint64_t p, osize, h, m, depth = 0; |
cec736d2 LP |
1428 | int r; |
1429 | ||
1430 | assert(f); | |
c88cc6af | 1431 | assert(f->header); |
cec736d2 LP |
1432 | assert(data || size == 0); |
1433 | ||
dade37d4 LP |
1434 | /* If there's no data hash table, then there's no entry. */ |
1435 | if (le64toh(f->header->data_hash_table_size) <= 0) | |
1436 | return 0; | |
1437 | ||
1438 | /* Map the data hash table, if it isn't mapped yet. */ | |
1439 | r = journal_file_map_data_hash_table(f); | |
1440 | if (r < 0) | |
1441 | return r; | |
1442 | ||
cec736d2 LP |
1443 | osize = offsetof(Object, data.payload) + size; |
1444 | ||
893e0f8f | 1445 | m = le64toh(READ_NOW(f->header->data_hash_table_size)) / sizeof(HashItem); |
805d1486 | 1446 | if (m <= 0) |
bc85bfee LP |
1447 | return -EBADMSG; |
1448 | ||
805d1486 | 1449 | h = hash % m; |
de190aef | 1450 | p = le64toh(f->data_hash_table[h].head_hash_offset); |
cec736d2 | 1451 | |
de190aef LP |
1452 | while (p > 0) { |
1453 | Object *o; | |
cec736d2 | 1454 | |
de190aef | 1455 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1456 | if (r < 0) |
1457 | return r; | |
1458 | ||
807e17f0 | 1459 | if (le64toh(o->data.hash) != hash) |
85a131e8 | 1460 | goto next; |
807e17f0 | 1461 | |
d89c8fdf | 1462 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
d80b051c | 1463 | #if HAVE_COMPRESSION |
fa1c4b51 | 1464 | uint64_t l; |
a7f7d1bd | 1465 | size_t rsize = 0; |
cec736d2 | 1466 | |
893e0f8f | 1467 | l = le64toh(READ_NOW(o->object.size)); |
807e17f0 | 1468 | if (l <= offsetof(Object, data.payload)) |
cec736d2 LP |
1469 | return -EBADMSG; |
1470 | ||
807e17f0 LP |
1471 | l -= offsetof(Object, data.payload); |
1472 | ||
d89c8fdf | 1473 | r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, |
319a4f4b | 1474 | o->data.payload, l, &f->compress_buffer, &rsize, 0); |
d89c8fdf ZJS |
1475 | if (r < 0) |
1476 | return r; | |
807e17f0 | 1477 | |
b785c858 | 1478 | if (rsize == size && |
807e17f0 LP |
1479 | memcmp(f->compress_buffer, data, size) == 0) { |
1480 | ||
1481 | if (ret) | |
1482 | *ret = o; | |
1483 | ||
f4474e00 LP |
1484 | if (ret_offset) |
1485 | *ret_offset = p; | |
807e17f0 LP |
1486 | |
1487 | return 1; | |
1488 | } | |
3b1a55e1 ZJS |
1489 | #else |
1490 | return -EPROTONOSUPPORT; | |
1491 | #endif | |
807e17f0 LP |
1492 | } else if (le64toh(o->object.size) == osize && |
1493 | memcmp(o->data.payload, data, size) == 0) { | |
1494 | ||
cec736d2 LP |
1495 | if (ret) |
1496 | *ret = o; | |
1497 | ||
f4474e00 LP |
1498 | if (ret_offset) |
1499 | *ret_offset = p; | |
cec736d2 | 1500 | |
de190aef | 1501 | return 1; |
cec736d2 LP |
1502 | } |
1503 | ||
85a131e8 | 1504 | next: |
0dbe57ee LP |
1505 | r = next_hash_offset( |
1506 | f, | |
1507 | &p, | |
1508 | &o->data.next_hash_offset, | |
1509 | &depth, | |
1510 | JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth) ? &f->header->data_hash_chain_depth : NULL); | |
1511 | if (r < 0) | |
1512 | return r; | |
cec736d2 LP |
1513 | } |
1514 | ||
de190aef LP |
1515 | return 0; |
1516 | } | |
1517 | ||
1518 | int journal_file_find_data_object( | |
1519 | JournalFile *f, | |
1520 | const void *data, uint64_t size, | |
f4474e00 | 1521 | Object **ret, uint64_t *ret_offset) { |
de190aef | 1522 | |
de190aef LP |
1523 | assert(f); |
1524 | assert(data || size == 0); | |
1525 | ||
f4474e00 LP |
1526 | return journal_file_find_data_object_with_hash( |
1527 | f, | |
4ce534f4 LP |
1528 | data, size, |
1529 | journal_file_hash_data(f, data, size), | |
f4474e00 | 1530 | ret, ret_offset); |
de190aef LP |
1531 | } |
1532 | ||
adce225a | 1533 | bool journal_field_valid(const char *p, size_t l, bool allow_protected) { |
adce225a YW |
1534 | /* We kinda enforce POSIX syntax recommendations for |
1535 | environment variables here, but make a couple of additional | |
1536 | requirements. | |
1537 | ||
1538 | http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */ | |
1539 | ||
f5fbe71d | 1540 | if (l == SIZE_MAX) |
adce225a YW |
1541 | l = strlen(p); |
1542 | ||
1543 | /* No empty field names */ | |
1544 | if (l <= 0) | |
1545 | return false; | |
1546 | ||
1547 | /* Don't allow names longer than 64 chars */ | |
1548 | if (l > 64) | |
1549 | return false; | |
1550 | ||
1551 | /* Variables starting with an underscore are protected */ | |
1552 | if (!allow_protected && p[0] == '_') | |
1553 | return false; | |
1554 | ||
1555 | /* Don't allow digits as first character */ | |
1556 | if (p[0] >= '0' && p[0] <= '9') | |
1557 | return false; | |
1558 | ||
1559 | /* Only allow A-Z0-9 and '_' */ | |
f6a0cfa5 | 1560 | for (const char *a = p; a < p + l; a++) |
adce225a YW |
1561 | if ((*a < 'A' || *a > 'Z') && |
1562 | (*a < '0' || *a > '9') && | |
1563 | *a != '_') | |
1564 | return false; | |
1565 | ||
1566 | return true; | |
1567 | } | |
1568 | ||
3c1668da LP |
1569 | static int journal_file_append_field( |
1570 | JournalFile *f, | |
1571 | const void *field, uint64_t size, | |
f4474e00 | 1572 | Object **ret, uint64_t *ret_offset) { |
3c1668da LP |
1573 | |
1574 | uint64_t hash, p; | |
1575 | uint64_t osize; | |
1576 | Object *o; | |
1577 | int r; | |
1578 | ||
1579 | assert(f); | |
1580 | assert(field && size > 0); | |
1581 | ||
f2bd0320 YW |
1582 | if (!journal_field_valid(field, size, true)) |
1583 | return -EBADMSG; | |
1584 | ||
4ce534f4 | 1585 | hash = journal_file_hash_data(f, field, size); |
3c1668da LP |
1586 | |
1587 | r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p); | |
1588 | if (r < 0) | |
1589 | return r; | |
8d5a1082 | 1590 | if (r > 0) { |
3c1668da LP |
1591 | |
1592 | if (ret) | |
1593 | *ret = o; | |
1594 | ||
f4474e00 LP |
1595 | if (ret_offset) |
1596 | *ret_offset = p; | |
3c1668da LP |
1597 | |
1598 | return 0; | |
1599 | } | |
1600 | ||
1601 | osize = offsetof(Object, field.payload) + size; | |
1602 | r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p); | |
8c92d4bb LP |
1603 | if (r < 0) |
1604 | return r; | |
3c1668da LP |
1605 | |
1606 | o->field.hash = htole64(hash); | |
1607 | memcpy(o->field.payload, field, size); | |
1608 | ||
1609 | r = journal_file_link_field(f, o, p, hash); | |
1610 | if (r < 0) | |
1611 | return r; | |
1612 | ||
1613 | /* The linking might have altered the window, so let's | |
1614 | * refresh our pointer */ | |
1615 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1616 | if (r < 0) | |
1617 | return r; | |
1618 | ||
349cc4a5 | 1619 | #if HAVE_GCRYPT |
3c1668da LP |
1620 | r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p); |
1621 | if (r < 0) | |
1622 | return r; | |
1623 | #endif | |
1624 | ||
1625 | if (ret) | |
1626 | *ret = o; | |
1627 | ||
f4474e00 LP |
1628 | if (ret_offset) |
1629 | *ret_offset = p; | |
3c1668da LP |
1630 | |
1631 | return 0; | |
1632 | } | |
1633 | ||
48496df6 LP |
1634 | static int journal_file_append_data( |
1635 | JournalFile *f, | |
1636 | const void *data, uint64_t size, | |
f4474e00 | 1637 | Object **ret, uint64_t *ret_offset) { |
48496df6 | 1638 | |
bc6b326d DDM |
1639 | uint64_t hash, p, fp, osize; |
1640 | Object *o, *fo; | |
d89c8fdf | 1641 | int r, compression = 0; |
3c1668da | 1642 | const void *eq; |
de190aef LP |
1643 | |
1644 | assert(f); | |
bc6b326d DDM |
1645 | |
1646 | if (!data || size == 0) | |
1647 | return -EINVAL; | |
de190aef | 1648 | |
4ce534f4 | 1649 | hash = journal_file_hash_data(f, data, size); |
de190aef LP |
1650 | |
1651 | r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p); | |
1652 | if (r < 0) | |
1653 | return r; | |
0240c603 | 1654 | if (r > 0) { |
de190aef LP |
1655 | |
1656 | if (ret) | |
1657 | *ret = o; | |
1658 | ||
f4474e00 LP |
1659 | if (ret_offset) |
1660 | *ret_offset = p; | |
de190aef LP |
1661 | |
1662 | return 0; | |
1663 | } | |
1664 | ||
bc6b326d DDM |
1665 | eq = memchr(data, '=', size); |
1666 | if (!eq) | |
1667 | return -EINVAL; | |
1668 | ||
de190aef LP |
1669 | osize = offsetof(Object, data.payload) + size; |
1670 | r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); | |
cec736d2 LP |
1671 | if (r < 0) |
1672 | return r; | |
1673 | ||
cec736d2 | 1674 | o->data.hash = htole64(hash); |
807e17f0 | 1675 | |
d80b051c | 1676 | #if HAVE_COMPRESSION |
57850536 | 1677 | if (JOURNAL_FILE_COMPRESS(f) && size >= f->compress_threshold_bytes) { |
a7f7d1bd | 1678 | size_t rsize = 0; |
807e17f0 | 1679 | |
5d6f46b6 | 1680 | compression = compress_blob(data, size, o->data.payload, size - 1, &rsize); |
807e17f0 | 1681 | |
d1afbcd2 | 1682 | if (compression >= 0) { |
807e17f0 | 1683 | o->object.size = htole64(offsetof(Object, data.payload) + rsize); |
d89c8fdf | 1684 | o->object.flags |= compression; |
807e17f0 | 1685 | |
fa1c4b51 | 1686 | log_debug("Compressed data object %"PRIu64" -> %zu using %s", |
d89c8fdf | 1687 | size, rsize, object_compressed_to_string(compression)); |
d1afbcd2 LP |
1688 | } else |
1689 | /* Compression didn't work, we don't really care why, let's continue without compression */ | |
1690 | compression = 0; | |
807e17f0 LP |
1691 | } |
1692 | #endif | |
1693 | ||
75f32f04 ZJS |
1694 | if (compression == 0) |
1695 | memcpy_safe(o->data.payload, data, size); | |
cec736d2 | 1696 | |
de190aef | 1697 | r = journal_file_link_data(f, o, p, hash); |
cec736d2 LP |
1698 | if (r < 0) |
1699 | return r; | |
1700 | ||
349cc4a5 | 1701 | #if HAVE_GCRYPT |
33685a5a FB |
1702 | r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p); |
1703 | if (r < 0) | |
1704 | return r; | |
1705 | #endif | |
1706 | ||
48496df6 LP |
1707 | /* The linking might have altered the window, so let's |
1708 | * refresh our pointer */ | |
1709 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
1710 | if (r < 0) | |
1711 | return r; | |
1712 | ||
bc6b326d DDM |
1713 | /* Create field object ... */ |
1714 | r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp); | |
1715 | if (r < 0) | |
1716 | return r; | |
3c1668da | 1717 | |
bc6b326d DDM |
1718 | /* ... and link it in. */ |
1719 | o->data.next_field_offset = fo->field.head_data_offset; | |
1720 | fo->field.head_data_offset = le64toh(p); | |
3c1668da | 1721 | |
cec736d2 LP |
1722 | if (ret) |
1723 | *ret = o; | |
1724 | ||
f4474e00 LP |
1725 | if (ret_offset) |
1726 | *ret_offset = p; | |
cec736d2 LP |
1727 | |
1728 | return 0; | |
1729 | } | |
1730 | ||
1731 | uint64_t journal_file_entry_n_items(Object *o) { | |
893e0f8f | 1732 | uint64_t sz; |
cec736d2 | 1733 | assert(o); |
b588975f LP |
1734 | |
1735 | if (o->object.type != OBJECT_ENTRY) | |
1736 | return 0; | |
cec736d2 | 1737 | |
893e0f8f LP |
1738 | sz = le64toh(READ_NOW(o->object.size)); |
1739 | if (sz < offsetof(Object, entry.items)) | |
1740 | return 0; | |
1741 | ||
1742 | return (sz - offsetof(Object, entry.items)) / sizeof(EntryItem); | |
cec736d2 LP |
1743 | } |
1744 | ||
0284adc6 | 1745 | uint64_t journal_file_entry_array_n_items(Object *o) { |
893e0f8f LP |
1746 | uint64_t sz; |
1747 | ||
de190aef | 1748 | assert(o); |
b588975f LP |
1749 | |
1750 | if (o->object.type != OBJECT_ENTRY_ARRAY) | |
1751 | return 0; | |
de190aef | 1752 | |
893e0f8f LP |
1753 | sz = le64toh(READ_NOW(o->object.size)); |
1754 | if (sz < offsetof(Object, entry_array.items)) | |
1755 | return 0; | |
1756 | ||
1757 | return (sz - offsetof(Object, entry_array.items)) / sizeof(uint64_t); | |
de190aef LP |
1758 | } |
1759 | ||
fb9a24b6 | 1760 | uint64_t journal_file_hash_table_n_items(Object *o) { |
893e0f8f LP |
1761 | uint64_t sz; |
1762 | ||
fb9a24b6 | 1763 | assert(o); |
b588975f | 1764 | |
ec2ce0c5 | 1765 | if (!IN_SET(o->object.type, OBJECT_DATA_HASH_TABLE, OBJECT_FIELD_HASH_TABLE)) |
b588975f | 1766 | return 0; |
fb9a24b6 | 1767 | |
893e0f8f LP |
1768 | sz = le64toh(READ_NOW(o->object.size)); |
1769 | if (sz < offsetof(Object, hash_table.items)) | |
1770 | return 0; | |
1771 | ||
1772 | return (sz - offsetof(Object, hash_table.items)) / sizeof(HashItem); | |
fb9a24b6 LP |
1773 | } |
1774 | ||
de190aef | 1775 | static int link_entry_into_array(JournalFile *f, |
4fd052ae FC |
1776 | le64_t *first, |
1777 | le64_t *idx, | |
de190aef | 1778 | uint64_t p) { |
cec736d2 | 1779 | int r; |
de190aef LP |
1780 | uint64_t n = 0, ap = 0, q, i, a, hidx; |
1781 | Object *o; | |
1782 | ||
cec736d2 | 1783 | assert(f); |
c88cc6af | 1784 | assert(f->header); |
de190aef LP |
1785 | assert(first); |
1786 | assert(idx); | |
1787 | assert(p > 0); | |
cec736d2 | 1788 | |
de190aef | 1789 | a = le64toh(*first); |
893e0f8f | 1790 | i = hidx = le64toh(READ_NOW(*idx)); |
de190aef LP |
1791 | while (a > 0) { |
1792 | ||
1793 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); | |
1794 | if (r < 0) | |
1795 | return r; | |
cec736d2 | 1796 | |
de190aef LP |
1797 | n = journal_file_entry_array_n_items(o); |
1798 | if (i < n) { | |
1799 | o->entry_array.items[i] = htole64(p); | |
1800 | *idx = htole64(hidx + 1); | |
1801 | return 0; | |
1802 | } | |
cec736d2 | 1803 | |
de190aef LP |
1804 | i -= n; |
1805 | ap = a; | |
1806 | a = le64toh(o->entry_array.next_entry_array_offset); | |
1807 | } | |
1808 | ||
1809 | if (hidx > n) | |
1810 | n = (hidx+1) * 2; | |
1811 | else | |
1812 | n = n * 2; | |
1813 | ||
1814 | if (n < 4) | |
1815 | n = 4; | |
1816 | ||
1817 | r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, | |
1818 | offsetof(Object, entry_array.items) + n * sizeof(uint64_t), | |
1819 | &o, &q); | |
cec736d2 LP |
1820 | if (r < 0) |
1821 | return r; | |
1822 | ||
349cc4a5 | 1823 | #if HAVE_GCRYPT |
5996c7c2 | 1824 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q); |
b0af6f41 LP |
1825 | if (r < 0) |
1826 | return r; | |
feb12d3e | 1827 | #endif |
b0af6f41 | 1828 | |
de190aef | 1829 | o->entry_array.items[i] = htole64(p); |
cec736d2 | 1830 | |
de190aef | 1831 | if (ap == 0) |
7be3aa17 | 1832 | *first = htole64(q); |
cec736d2 | 1833 | else { |
de190aef | 1834 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); |
cec736d2 LP |
1835 | if (r < 0) |
1836 | return r; | |
1837 | ||
de190aef LP |
1838 | o->entry_array.next_entry_array_offset = htole64(q); |
1839 | } | |
cec736d2 | 1840 | |
2dee23eb LP |
1841 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
1842 | f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); | |
1843 | ||
de190aef LP |
1844 | *idx = htole64(hidx + 1); |
1845 | ||
1846 | return 0; | |
1847 | } | |
cec736d2 | 1848 | |
de190aef | 1849 | static int link_entry_into_array_plus_one(JournalFile *f, |
4fd052ae FC |
1850 | le64_t *extra, |
1851 | le64_t *first, | |
1852 | le64_t *idx, | |
de190aef LP |
1853 | uint64_t p) { |
1854 | ||
893e0f8f | 1855 | uint64_t hidx; |
de190aef LP |
1856 | int r; |
1857 | ||
1858 | assert(f); | |
1859 | assert(extra); | |
1860 | assert(first); | |
1861 | assert(idx); | |
1862 | assert(p > 0); | |
1863 | ||
893e0f8f LP |
1864 | hidx = le64toh(READ_NOW(*idx)); |
1865 | if (hidx == UINT64_MAX) | |
1866 | return -EBADMSG; | |
1867 | if (hidx == 0) | |
de190aef LP |
1868 | *extra = htole64(p); |
1869 | else { | |
4fd052ae | 1870 | le64_t i; |
de190aef | 1871 | |
893e0f8f | 1872 | i = htole64(hidx - 1); |
de190aef LP |
1873 | r = link_entry_into_array(f, first, &i, p); |
1874 | if (r < 0) | |
1875 | return r; | |
cec736d2 LP |
1876 | } |
1877 | ||
893e0f8f | 1878 | *idx = htole64(hidx + 1); |
de190aef LP |
1879 | return 0; |
1880 | } | |
1881 | ||
1882 | static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { | |
1883 | uint64_t p; | |
1884 | int r; | |
bfbd5be0 | 1885 | |
de190aef LP |
1886 | assert(f); |
1887 | assert(o); | |
1888 | assert(offset > 0); | |
1889 | ||
1890 | p = le64toh(o->entry.items[i].object_offset); | |
de190aef | 1891 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1892 | if (r < 0) |
1893 | return r; | |
1894 | ||
de190aef LP |
1895 | return link_entry_into_array_plus_one(f, |
1896 | &o->data.entry_offset, | |
1897 | &o->data.entry_array_offset, | |
1898 | &o->data.n_entries, | |
1899 | offset); | |
cec736d2 LP |
1900 | } |
1901 | ||
1902 | static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { | |
f6a0cfa5 | 1903 | uint64_t n; |
cec736d2 LP |
1904 | int r; |
1905 | ||
1906 | assert(f); | |
c88cc6af | 1907 | assert(f->header); |
cec736d2 LP |
1908 | assert(o); |
1909 | assert(offset > 0); | |
b588975f LP |
1910 | |
1911 | if (o->object.type != OBJECT_ENTRY) | |
1912 | return -EINVAL; | |
cec736d2 | 1913 | |
b788cc23 LP |
1914 | __sync_synchronize(); |
1915 | ||
cec736d2 | 1916 | /* Link up the entry itself */ |
de190aef LP |
1917 | r = link_entry_into_array(f, |
1918 | &f->header->entry_array_offset, | |
1919 | &f->header->n_entries, | |
1920 | offset); | |
1921 | if (r < 0) | |
1922 | return r; | |
cec736d2 | 1923 | |
507f22bd | 1924 | /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */ |
cec736d2 | 1925 | |
de190aef | 1926 | if (f->header->head_entry_realtime == 0) |
0ac38b70 | 1927 | f->header->head_entry_realtime = o->entry.realtime; |
cec736d2 | 1928 | |
0ac38b70 | 1929 | f->header->tail_entry_realtime = o->entry.realtime; |
de190aef LP |
1930 | f->header->tail_entry_monotonic = o->entry.monotonic; |
1931 | ||
cec736d2 LP |
1932 | /* Link up the items */ |
1933 | n = journal_file_entry_n_items(o); | |
f6a0cfa5 | 1934 | for (uint64_t i = 0; i < n; i++) { |
cec736d2 LP |
1935 | r = journal_file_link_entry_item(f, o, offset, i); |
1936 | if (r < 0) | |
1937 | return r; | |
1938 | } | |
1939 | ||
cec736d2 LP |
1940 | return 0; |
1941 | } | |
1942 | ||
1943 | static int journal_file_append_entry_internal( | |
1944 | JournalFile *f, | |
1945 | const dual_timestamp *ts, | |
d180c349 | 1946 | const sd_id128_t *boot_id, |
cec736d2 LP |
1947 | uint64_t xor_hash, |
1948 | const EntryItem items[], unsigned n_items, | |
de190aef | 1949 | uint64_t *seqnum, |
f4474e00 | 1950 | Object **ret, uint64_t *ret_offset) { |
cec736d2 LP |
1951 | uint64_t np; |
1952 | uint64_t osize; | |
1953 | Object *o; | |
1954 | int r; | |
1955 | ||
1956 | assert(f); | |
c88cc6af | 1957 | assert(f->header); |
cec736d2 | 1958 | assert(items || n_items == 0); |
de190aef | 1959 | assert(ts); |
cec736d2 LP |
1960 | |
1961 | osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem)); | |
1962 | ||
de190aef | 1963 | r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); |
cec736d2 LP |
1964 | if (r < 0) |
1965 | return r; | |
1966 | ||
d98cc1f2 | 1967 | o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); |
75f32f04 | 1968 | memcpy_safe(o->entry.items, items, n_items * sizeof(EntryItem)); |
de190aef LP |
1969 | o->entry.realtime = htole64(ts->realtime); |
1970 | o->entry.monotonic = htole64(ts->monotonic); | |
cec736d2 | 1971 | o->entry.xor_hash = htole64(xor_hash); |
924426a7 CM |
1972 | if (boot_id) |
1973 | f->header->boot_id = *boot_id; | |
1974 | o->entry.boot_id = f->header->boot_id; | |
cec736d2 | 1975 | |
349cc4a5 | 1976 | #if HAVE_GCRYPT |
5996c7c2 | 1977 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np); |
b0af6f41 | 1978 | if (r < 0) |
b41b682b | 1979 | return r; |
feb12d3e | 1980 | #endif |
b0af6f41 | 1981 | |
cec736d2 LP |
1982 | r = journal_file_link_entry(f, o, np); |
1983 | if (r < 0) | |
b41b682b | 1984 | return r; |
cec736d2 LP |
1985 | |
1986 | if (ret) | |
1987 | *ret = o; | |
1988 | ||
f4474e00 LP |
1989 | if (ret_offset) |
1990 | *ret_offset = np; | |
cec736d2 | 1991 | |
0eaee828 | 1992 | return r; |
cec736d2 LP |
1993 | } |
1994 | ||
cf244689 | 1995 | void journal_file_post_change(JournalFile *f) { |
50f20cfd LP |
1996 | assert(f); |
1997 | ||
c5236850 DT |
1998 | if (f->fd < 0) |
1999 | return; | |
2000 | ||
50f20cfd LP |
2001 | /* inotify() does not receive IN_MODIFY events from file |
2002 | * accesses done via mmap(). After each access we hence | |
2003 | * trigger IN_MODIFY by truncating the journal file to its | |
2004 | * current size which triggers IN_MODIFY. */ | |
2005 | ||
bc85bfee LP |
2006 | __sync_synchronize(); |
2007 | ||
50f20cfd | 2008 | if (ftruncate(f->fd, f->last_stat.st_size) < 0) |
e167d7fd | 2009 | log_debug_errno(errno, "Failed to truncate file to its own size: %m"); |
50f20cfd LP |
2010 | } |
2011 | ||
7a24f3bf VC |
2012 | static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) { |
2013 | assert(userdata); | |
2014 | ||
2015 | journal_file_post_change(userdata); | |
2016 | ||
2017 | return 1; | |
2018 | } | |
2019 | ||
2020 | static void schedule_post_change(JournalFile *f) { | |
b6cdfbe5 | 2021 | int r; |
7a24f3bf VC |
2022 | |
2023 | assert(f); | |
2024 | assert(f->post_change_timer); | |
2025 | ||
b6cdfbe5 | 2026 | r = sd_event_source_get_enabled(f->post_change_timer, NULL); |
7a24f3bf | 2027 | if (r < 0) { |
e167d7fd LP |
2028 | log_debug_errno(r, "Failed to get ftruncate timer state: %m"); |
2029 | goto fail; | |
7a24f3bf | 2030 | } |
b6cdfbe5 | 2031 | if (r > 0) |
7a24f3bf VC |
2032 | return; |
2033 | ||
39cf0351 | 2034 | r = sd_event_source_set_time_relative(f->post_change_timer, f->post_change_timer_period); |
7a24f3bf | 2035 | if (r < 0) { |
e167d7fd LP |
2036 | log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m"); |
2037 | goto fail; | |
7a24f3bf VC |
2038 | } |
2039 | ||
ca5d90d4 | 2040 | r = sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_ONESHOT); |
7a24f3bf | 2041 | if (r < 0) { |
e167d7fd LP |
2042 | log_debug_errno(r, "Failed to enable scheduled ftruncate: %m"); |
2043 | goto fail; | |
7a24f3bf | 2044 | } |
e167d7fd LP |
2045 | |
2046 | return; | |
2047 | ||
2048 | fail: | |
2049 | /* On failure, let's simply post the change immediately. */ | |
2050 | journal_file_post_change(f); | |
7a24f3bf VC |
2051 | } |
2052 | ||
2053 | /* Enable coalesced change posting in a timer on the provided sd_event instance */ | |
2054 | int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) { | |
2055 | _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL; | |
2056 | int r; | |
2057 | ||
2058 | assert(f); | |
2059 | assert_return(!f->post_change_timer, -EINVAL); | |
2060 | assert(e); | |
2061 | assert(t); | |
2062 | ||
2063 | r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f); | |
2064 | if (r < 0) | |
2065 | return r; | |
2066 | ||
2067 | r = sd_event_source_set_enabled(timer, SD_EVENT_OFF); | |
2068 | if (r < 0) | |
2069 | return r; | |
2070 | ||
1cc6c93a | 2071 | f->post_change_timer = TAKE_PTR(timer); |
7a24f3bf VC |
2072 | f->post_change_timer_period = t; |
2073 | ||
2074 | return r; | |
2075 | } | |
2076 | ||
93bab288 YW |
2077 | static int entry_item_cmp(const EntryItem *a, const EntryItem *b) { |
2078 | return CMP(le64toh(a->object_offset), le64toh(b->object_offset)); | |
1f2da9ec LP |
2079 | } |
2080 | ||
d180c349 ZJS |
2081 | int journal_file_append_entry( |
2082 | JournalFile *f, | |
2083 | const dual_timestamp *ts, | |
2084 | const sd_id128_t *boot_id, | |
2085 | const struct iovec iovec[], unsigned n_iovec, | |
2086 | uint64_t *seqnum, | |
f4474e00 | 2087 | Object **ret, uint64_t *ret_offset) { |
d180c349 | 2088 | |
cec736d2 LP |
2089 | EntryItem *items; |
2090 | int r; | |
2091 | uint64_t xor_hash = 0; | |
de190aef | 2092 | struct dual_timestamp _ts; |
cec736d2 LP |
2093 | |
2094 | assert(f); | |
c88cc6af | 2095 | assert(f->header); |
bc6b326d | 2096 | assert(iovec && n_iovec > 0); |
cec736d2 | 2097 | |
c6273953 | 2098 | if (ts) { |
baaa35ad ZJS |
2099 | if (!VALID_REALTIME(ts->realtime)) |
2100 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2101 | "Invalid realtime timestamp %" PRIu64 ", refusing entry.", | |
2102 | ts->realtime); | |
2103 | if (!VALID_MONOTONIC(ts->monotonic)) | |
2104 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2105 | "Invalid monotomic timestamp %" PRIu64 ", refusing entry.", | |
2106 | ts->monotonic); | |
c6273953 | 2107 | } else { |
de190aef LP |
2108 | dual_timestamp_get(&_ts); |
2109 | ts = &_ts; | |
2110 | } | |
2111 | ||
349cc4a5 | 2112 | #if HAVE_GCRYPT |
7560fffc LP |
2113 | r = journal_file_maybe_append_tag(f, ts->realtime); |
2114 | if (r < 0) | |
2115 | return r; | |
feb12d3e | 2116 | #endif |
7560fffc | 2117 | |
5222651e | 2118 | items = newa(EntryItem, n_iovec); |
cec736d2 | 2119 | |
86e68f38 | 2120 | for (size_t i = 0; i < n_iovec; i++) { |
cec736d2 LP |
2121 | uint64_t p; |
2122 | Object *o; | |
2123 | ||
2124 | r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); | |
2125 | if (r < 0) | |
cf244689 | 2126 | return r; |
cec736d2 | 2127 | |
4ce534f4 LP |
2128 | /* When calculating the XOR hash field, we need to take special care if the "keyed-hash" |
2129 | * journal file flag is on. We use the XOR hash field to quickly determine the identity of a | |
2130 | * specific record, and give records with otherwise identical position (i.e. match in seqno, | |
2131 | * timestamp, …) a stable ordering. But for that we can't have it that the hash of the | |
2132 | * objects in each file is different since they are keyed. Hence let's calculate the Jenkins | |
2133 | * hash here for that. This also has the benefit that cursors for old and new journal files | |
2134 | * are completely identical (they include the XOR hash after all). For classic Jenkins-hash | |
2135 | * files things are easier, we can just take the value from the stored record directly. */ | |
2136 | ||
2137 | if (JOURNAL_HEADER_KEYED_HASH(f->header)) | |
2138 | xor_hash ^= jenkins_hash64(iovec[i].iov_base, iovec[i].iov_len); | |
2139 | else | |
2140 | xor_hash ^= le64toh(o->data.hash); | |
2141 | ||
d164ac77 DDM |
2142 | items[i] = (EntryItem) { |
2143 | .object_offset = htole64(p), | |
2144 | .hash = o->data.hash, | |
2145 | }; | |
cec736d2 LP |
2146 | } |
2147 | ||
1f2da9ec LP |
2148 | /* Order by the position on disk, in order to improve seek |
2149 | * times for rotating media. */ | |
93bab288 | 2150 | typesafe_qsort(items, n_iovec, entry_item_cmp); |
1f2da9ec | 2151 | |
f4474e00 | 2152 | r = journal_file_append_entry_internal(f, ts, boot_id, xor_hash, items, n_iovec, seqnum, ret, ret_offset); |
cec736d2 | 2153 | |
fa6ac760 LP |
2154 | /* If the memory mapping triggered a SIGBUS then we return an |
2155 | * IO error and ignore the error code passed down to us, since | |
2156 | * it is very likely just an effect of a nullified replacement | |
2157 | * mapping page */ | |
2158 | ||
c3bd54bf | 2159 | if (mmap_cache_fd_got_sigbus(f->cache_fd)) |
fa6ac760 LP |
2160 | r = -EIO; |
2161 | ||
7a24f3bf VC |
2162 | if (f->post_change_timer) |
2163 | schedule_post_change(f); | |
2164 | else | |
2165 | journal_file_post_change(f); | |
50f20cfd | 2166 | |
cec736d2 LP |
2167 | return r; |
2168 | } | |
2169 | ||
a4bcff5b | 2170 | typedef struct ChainCacheItem { |
fb099c8d | 2171 | uint64_t first; /* the array at the beginning of the chain */ |
a4bcff5b LP |
2172 | uint64_t array; /* the cached array */ |
2173 | uint64_t begin; /* the first item in the cached array */ | |
2174 | uint64_t total; /* the total number of items in all arrays before this one in the chain */ | |
f268980d | 2175 | uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */ |
a4bcff5b LP |
2176 | } ChainCacheItem; |
2177 | ||
2178 | static void chain_cache_put( | |
4743015d | 2179 | OrderedHashmap *h, |
a4bcff5b LP |
2180 | ChainCacheItem *ci, |
2181 | uint64_t first, | |
2182 | uint64_t array, | |
2183 | uint64_t begin, | |
f268980d LP |
2184 | uint64_t total, |
2185 | uint64_t last_index) { | |
a4bcff5b LP |
2186 | |
2187 | if (!ci) { | |
34741aa3 LP |
2188 | /* If the chain item to cache for this chain is the |
2189 | * first one it's not worth caching anything */ | |
2190 | if (array == first) | |
2191 | return; | |
2192 | ||
29433089 | 2193 | if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) { |
4743015d | 2194 | ci = ordered_hashmap_steal_first(h); |
29433089 LP |
2195 | assert(ci); |
2196 | } else { | |
a4bcff5b LP |
2197 | ci = new(ChainCacheItem, 1); |
2198 | if (!ci) | |
2199 | return; | |
2200 | } | |
2201 | ||
2202 | ci->first = first; | |
2203 | ||
4743015d | 2204 | if (ordered_hashmap_put(h, &ci->first, ci) < 0) { |
a4bcff5b LP |
2205 | free(ci); |
2206 | return; | |
2207 | } | |
2208 | } else | |
2209 | assert(ci->first == first); | |
2210 | ||
2211 | ci->array = array; | |
2212 | ci->begin = begin; | |
2213 | ci->total = total; | |
f268980d | 2214 | ci->last_index = last_index; |
a4bcff5b LP |
2215 | } |
2216 | ||
f268980d LP |
2217 | static int generic_array_get( |
2218 | JournalFile *f, | |
2219 | uint64_t first, | |
2220 | uint64_t i, | |
f4474e00 | 2221 | Object **ret, uint64_t *ret_offset) { |
de190aef | 2222 | |
cec736d2 | 2223 | Object *o; |
a4bcff5b | 2224 | uint64_t p = 0, a, t = 0; |
cec736d2 | 2225 | int r; |
a4bcff5b | 2226 | ChainCacheItem *ci; |
cec736d2 LP |
2227 | |
2228 | assert(f); | |
2229 | ||
de190aef | 2230 | a = first; |
a4bcff5b LP |
2231 | |
2232 | /* Try the chain cache first */ | |
4743015d | 2233 | ci = ordered_hashmap_get(f->chain_cache, &first); |
a4bcff5b LP |
2234 | if (ci && i > ci->total) { |
2235 | a = ci->array; | |
2236 | i -= ci->total; | |
2237 | t = ci->total; | |
2238 | } | |
2239 | ||
de190aef | 2240 | while (a > 0) { |
a4bcff5b | 2241 | uint64_t k; |
cec736d2 | 2242 | |
de190aef LP |
2243 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); |
2244 | if (r < 0) | |
2245 | return r; | |
cec736d2 | 2246 | |
a4bcff5b LP |
2247 | k = journal_file_entry_array_n_items(o); |
2248 | if (i < k) { | |
de190aef | 2249 | p = le64toh(o->entry_array.items[i]); |
a4bcff5b | 2250 | goto found; |
cec736d2 LP |
2251 | } |
2252 | ||
a4bcff5b LP |
2253 | i -= k; |
2254 | t += k; | |
de190aef LP |
2255 | a = le64toh(o->entry_array.next_entry_array_offset); |
2256 | } | |
2257 | ||
a4bcff5b LP |
2258 | return 0; |
2259 | ||
2260 | found: | |
2261 | /* Let's cache this item for the next invocation */ | |
af13a6b0 | 2262 | chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i); |
de190aef LP |
2263 | |
2264 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2265 | if (r < 0) | |
2266 | return r; | |
2267 | ||
2268 | if (ret) | |
2269 | *ret = o; | |
2270 | ||
f4474e00 LP |
2271 | if (ret_offset) |
2272 | *ret_offset = p; | |
de190aef LP |
2273 | |
2274 | return 1; | |
2275 | } | |
2276 | ||
f268980d LP |
2277 | static int generic_array_get_plus_one( |
2278 | JournalFile *f, | |
2279 | uint64_t extra, | |
2280 | uint64_t first, | |
2281 | uint64_t i, | |
f4474e00 | 2282 | Object **ret, uint64_t *ret_offset) { |
de190aef LP |
2283 | |
2284 | Object *o; | |
2285 | ||
2286 | assert(f); | |
2287 | ||
2288 | if (i == 0) { | |
2289 | int r; | |
2290 | ||
2291 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
cec736d2 LP |
2292 | if (r < 0) |
2293 | return r; | |
2294 | ||
de190aef LP |
2295 | if (ret) |
2296 | *ret = o; | |
cec736d2 | 2297 | |
f4474e00 LP |
2298 | if (ret_offset) |
2299 | *ret_offset = extra; | |
cec736d2 | 2300 | |
de190aef | 2301 | return 1; |
cec736d2 LP |
2302 | } |
2303 | ||
f4474e00 | 2304 | return generic_array_get(f, first, i-1, ret, ret_offset); |
de190aef | 2305 | } |
cec736d2 | 2306 | |
de190aef LP |
2307 | enum { |
2308 | TEST_FOUND, | |
2309 | TEST_LEFT, | |
2310 | TEST_RIGHT | |
2311 | }; | |
cec736d2 | 2312 | |
f268980d LP |
2313 | static int generic_array_bisect( |
2314 | JournalFile *f, | |
2315 | uint64_t first, | |
2316 | uint64_t n, | |
2317 | uint64_t needle, | |
2318 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
2319 | direction_t direction, | |
2320 | Object **ret, | |
f4474e00 LP |
2321 | uint64_t *ret_offset, |
2322 | uint64_t *ret_idx) { | |
f268980d | 2323 | |
f5fbe71d | 2324 | uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = UINT64_MAX; |
de190aef LP |
2325 | bool subtract_one = false; |
2326 | Object *o, *array = NULL; | |
2327 | int r; | |
a4bcff5b | 2328 | ChainCacheItem *ci; |
cec736d2 | 2329 | |
de190aef LP |
2330 | assert(f); |
2331 | assert(test_object); | |
cec736d2 | 2332 | |
a4bcff5b | 2333 | /* Start with the first array in the chain */ |
de190aef | 2334 | a = first; |
a4bcff5b | 2335 | |
4743015d | 2336 | ci = ordered_hashmap_get(f->chain_cache, &first); |
96d4d024 | 2337 | if (ci && n > ci->total && ci->begin != 0) { |
a4bcff5b LP |
2338 | /* Ah, we have iterated this bisection array chain |
2339 | * previously! Let's see if we can skip ahead in the | |
2340 | * chain, as far as the last time. But we can't jump | |
2341 | * backwards in the chain, so let's check that | |
2342 | * first. */ | |
2343 | ||
2344 | r = test_object(f, ci->begin, needle); | |
2345 | if (r < 0) | |
2346 | return r; | |
2347 | ||
2348 | if (r == TEST_LEFT) { | |
f268980d | 2349 | /* OK, what we are looking for is right of the |
a4bcff5b LP |
2350 | * begin of this EntryArray, so let's jump |
2351 | * straight to previously cached array in the | |
2352 | * chain */ | |
2353 | ||
2354 | a = ci->array; | |
2355 | n -= ci->total; | |
2356 | t = ci->total; | |
f268980d | 2357 | last_index = ci->last_index; |
a4bcff5b LP |
2358 | } |
2359 | } | |
2360 | ||
de190aef LP |
2361 | while (a > 0) { |
2362 | uint64_t left, right, k, lp; | |
2363 | ||
2364 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); | |
cec736d2 LP |
2365 | if (r < 0) |
2366 | return r; | |
2367 | ||
de190aef LP |
2368 | k = journal_file_entry_array_n_items(array); |
2369 | right = MIN(k, n); | |
2370 | if (right <= 0) | |
2371 | return 0; | |
cec736d2 | 2372 | |
de190aef LP |
2373 | i = right - 1; |
2374 | lp = p = le64toh(array->entry_array.items[i]); | |
2375 | if (p <= 0) | |
bee6a291 LP |
2376 | r = -EBADMSG; |
2377 | else | |
2378 | r = test_object(f, p, needle); | |
2379 | if (r == -EBADMSG) { | |
2380 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)"); | |
2381 | n = i; | |
2382 | continue; | |
2383 | } | |
de190aef LP |
2384 | if (r < 0) |
2385 | return r; | |
cec736d2 | 2386 | |
de190aef LP |
2387 | if (r == TEST_FOUND) |
2388 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2389 | ||
2390 | if (r == TEST_RIGHT) { | |
2391 | left = 0; | |
2392 | right -= 1; | |
f268980d | 2393 | |
f5fbe71d | 2394 | if (last_index != UINT64_MAX) { |
f268980d LP |
2395 | assert(last_index <= right); |
2396 | ||
2397 | /* If we cached the last index we | |
2398 | * looked at, let's try to not to jump | |
2399 | * too wildly around and see if we can | |
2400 | * limit the range to look at early to | |
2401 | * the immediate neighbors of the last | |
2402 | * index we looked at. */ | |
2403 | ||
2404 | if (last_index > 0) { | |
2405 | uint64_t x = last_index - 1; | |
2406 | ||
2407 | p = le64toh(array->entry_array.items[x]); | |
2408 | if (p <= 0) | |
2409 | return -EBADMSG; | |
2410 | ||
2411 | r = test_object(f, p, needle); | |
2412 | if (r < 0) | |
2413 | return r; | |
2414 | ||
2415 | if (r == TEST_FOUND) | |
2416 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2417 | ||
2418 | if (r == TEST_RIGHT) | |
2419 | right = x; | |
2420 | else | |
2421 | left = x + 1; | |
2422 | } | |
2423 | ||
2424 | if (last_index < right) { | |
2425 | uint64_t y = last_index + 1; | |
2426 | ||
2427 | p = le64toh(array->entry_array.items[y]); | |
2428 | if (p <= 0) | |
2429 | return -EBADMSG; | |
2430 | ||
2431 | r = test_object(f, p, needle); | |
2432 | if (r < 0) | |
2433 | return r; | |
2434 | ||
2435 | if (r == TEST_FOUND) | |
2436 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2437 | ||
2438 | if (r == TEST_RIGHT) | |
2439 | right = y; | |
2440 | else | |
2441 | left = y + 1; | |
2442 | } | |
f268980d LP |
2443 | } |
2444 | ||
de190aef LP |
2445 | for (;;) { |
2446 | if (left == right) { | |
2447 | if (direction == DIRECTION_UP) | |
2448 | subtract_one = true; | |
2449 | ||
2450 | i = left; | |
2451 | goto found; | |
2452 | } | |
2453 | ||
2454 | assert(left < right); | |
de190aef | 2455 | i = (left + right) / 2; |
f268980d | 2456 | |
de190aef LP |
2457 | p = le64toh(array->entry_array.items[i]); |
2458 | if (p <= 0) | |
bee6a291 LP |
2459 | r = -EBADMSG; |
2460 | else | |
2461 | r = test_object(f, p, needle); | |
2462 | if (r == -EBADMSG) { | |
2463 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)"); | |
2464 | right = n = i; | |
2465 | continue; | |
2466 | } | |
de190aef LP |
2467 | if (r < 0) |
2468 | return r; | |
cec736d2 | 2469 | |
de190aef LP |
2470 | if (r == TEST_FOUND) |
2471 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2472 | ||
2473 | if (r == TEST_RIGHT) | |
2474 | right = i; | |
2475 | else | |
2476 | left = i + 1; | |
2477 | } | |
2478 | } | |
2479 | ||
2173cbf8 | 2480 | if (k >= n) { |
cbdca852 LP |
2481 | if (direction == DIRECTION_UP) { |
2482 | i = n; | |
2483 | subtract_one = true; | |
2484 | goto found; | |
2485 | } | |
2486 | ||
cec736d2 | 2487 | return 0; |
cbdca852 | 2488 | } |
cec736d2 | 2489 | |
de190aef LP |
2490 | last_p = lp; |
2491 | ||
2492 | n -= k; | |
2493 | t += k; | |
f5fbe71d | 2494 | last_index = UINT64_MAX; |
de190aef | 2495 | a = le64toh(array->entry_array.next_entry_array_offset); |
cec736d2 LP |
2496 | } |
2497 | ||
2498 | return 0; | |
de190aef LP |
2499 | |
2500 | found: | |
2501 | if (subtract_one && t == 0 && i == 0) | |
2502 | return 0; | |
2503 | ||
a4bcff5b | 2504 | /* Let's cache this item for the next invocation */ |
f5fbe71d | 2505 | chain_cache_put(f->chain_cache, ci, first, a, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : UINT64_MAX) : i); |
a4bcff5b | 2506 | |
de190aef LP |
2507 | if (subtract_one && i == 0) |
2508 | p = last_p; | |
2509 | else if (subtract_one) | |
2510 | p = le64toh(array->entry_array.items[i-1]); | |
2511 | else | |
2512 | p = le64toh(array->entry_array.items[i]); | |
2513 | ||
2514 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2515 | if (r < 0) | |
2516 | return r; | |
2517 | ||
2518 | if (ret) | |
2519 | *ret = o; | |
2520 | ||
f4474e00 LP |
2521 | if (ret_offset) |
2522 | *ret_offset = p; | |
de190aef | 2523 | |
f4474e00 LP |
2524 | if (ret_idx) |
2525 | *ret_idx = t + i + (subtract_one ? -1 : 0); | |
de190aef LP |
2526 | |
2527 | return 1; | |
cec736d2 LP |
2528 | } |
2529 | ||
f268980d LP |
2530 | static int generic_array_bisect_plus_one( |
2531 | JournalFile *f, | |
2532 | uint64_t extra, | |
2533 | uint64_t first, | |
2534 | uint64_t n, | |
2535 | uint64_t needle, | |
2536 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
2537 | direction_t direction, | |
2538 | Object **ret, | |
f4474e00 LP |
2539 | uint64_t *ret_offset, |
2540 | uint64_t *ret_idx) { | |
de190aef | 2541 | |
cec736d2 | 2542 | int r; |
cbdca852 LP |
2543 | bool step_back = false; |
2544 | Object *o; | |
cec736d2 LP |
2545 | |
2546 | assert(f); | |
de190aef | 2547 | assert(test_object); |
cec736d2 | 2548 | |
de190aef LP |
2549 | if (n <= 0) |
2550 | return 0; | |
cec736d2 | 2551 | |
de190aef LP |
2552 | /* This bisects the array in object 'first', but first checks |
2553 | * an extra */ | |
de190aef LP |
2554 | r = test_object(f, extra, needle); |
2555 | if (r < 0) | |
2556 | return r; | |
a536e261 LP |
2557 | |
2558 | if (r == TEST_FOUND) | |
2559 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2560 | ||
cbdca852 LP |
2561 | /* if we are looking with DIRECTION_UP then we need to first |
2562 | see if in the actual array there is a matching entry, and | |
2563 | return the last one of that. But if there isn't any we need | |
2564 | to return this one. Hence remember this, and return it | |
2565 | below. */ | |
2566 | if (r == TEST_LEFT) | |
2567 | step_back = direction == DIRECTION_UP; | |
de190aef | 2568 | |
cbdca852 LP |
2569 | if (r == TEST_RIGHT) { |
2570 | if (direction == DIRECTION_DOWN) | |
2571 | goto found; | |
2572 | else | |
2573 | return 0; | |
a536e261 | 2574 | } |
cec736d2 | 2575 | |
f4474e00 | 2576 | r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, ret_offset, ret_idx); |
de190aef | 2577 | |
cbdca852 LP |
2578 | if (r == 0 && step_back) |
2579 | goto found; | |
2580 | ||
f4474e00 LP |
2581 | if (r > 0 && ret_idx) |
2582 | (*ret_idx)++; | |
de190aef LP |
2583 | |
2584 | return r; | |
cbdca852 LP |
2585 | |
2586 | found: | |
2587 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
2588 | if (r < 0) | |
2589 | return r; | |
2590 | ||
2591 | if (ret) | |
2592 | *ret = o; | |
2593 | ||
f4474e00 LP |
2594 | if (ret_offset) |
2595 | *ret_offset = extra; | |
cbdca852 | 2596 | |
f4474e00 LP |
2597 | if (ret_idx) |
2598 | *ret_idx = 0; | |
cbdca852 LP |
2599 | |
2600 | return 1; | |
2601 | } | |
2602 | ||
44a6b1b6 | 2603 | _pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { |
cbdca852 LP |
2604 | assert(f); |
2605 | assert(p > 0); | |
2606 | ||
2607 | if (p == needle) | |
2608 | return TEST_FOUND; | |
2609 | else if (p < needle) | |
2610 | return TEST_LEFT; | |
2611 | else | |
2612 | return TEST_RIGHT; | |
2613 | } | |
2614 | ||
de190aef | 2615 | static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { |
893e0f8f | 2616 | uint64_t sq; |
de190aef LP |
2617 | Object *o; |
2618 | int r; | |
2619 | ||
2620 | assert(f); | |
2621 | assert(p > 0); | |
2622 | ||
2623 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
cec736d2 LP |
2624 | if (r < 0) |
2625 | return r; | |
2626 | ||
893e0f8f LP |
2627 | sq = le64toh(READ_NOW(o->entry.seqnum)); |
2628 | if (sq == needle) | |
de190aef | 2629 | return TEST_FOUND; |
893e0f8f | 2630 | else if (sq < needle) |
de190aef LP |
2631 | return TEST_LEFT; |
2632 | else | |
2633 | return TEST_RIGHT; | |
2634 | } | |
cec736d2 | 2635 | |
de190aef LP |
2636 | int journal_file_move_to_entry_by_seqnum( |
2637 | JournalFile *f, | |
2638 | uint64_t seqnum, | |
2639 | direction_t direction, | |
2640 | Object **ret, | |
f4474e00 | 2641 | uint64_t *ret_offset) { |
c88cc6af VC |
2642 | assert(f); |
2643 | assert(f->header); | |
de190aef | 2644 | |
f4474e00 LP |
2645 | return generic_array_bisect( |
2646 | f, | |
2647 | le64toh(f->header->entry_array_offset), | |
2648 | le64toh(f->header->n_entries), | |
2649 | seqnum, | |
2650 | test_object_seqnum, | |
2651 | direction, | |
2652 | ret, ret_offset, NULL); | |
de190aef | 2653 | } |
cec736d2 | 2654 | |
de190aef LP |
2655 | static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { |
2656 | Object *o; | |
893e0f8f | 2657 | uint64_t rt; |
de190aef LP |
2658 | int r; |
2659 | ||
2660 | assert(f); | |
2661 | assert(p > 0); | |
2662 | ||
2663 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2664 | if (r < 0) | |
2665 | return r; | |
2666 | ||
893e0f8f LP |
2667 | rt = le64toh(READ_NOW(o->entry.realtime)); |
2668 | if (rt == needle) | |
de190aef | 2669 | return TEST_FOUND; |
893e0f8f | 2670 | else if (rt < needle) |
de190aef LP |
2671 | return TEST_LEFT; |
2672 | else | |
2673 | return TEST_RIGHT; | |
cec736d2 LP |
2674 | } |
2675 | ||
de190aef LP |
2676 | int journal_file_move_to_entry_by_realtime( |
2677 | JournalFile *f, | |
2678 | uint64_t realtime, | |
2679 | direction_t direction, | |
2680 | Object **ret, | |
f4474e00 | 2681 | uint64_t *ret_offset) { |
c88cc6af VC |
2682 | assert(f); |
2683 | assert(f->header); | |
de190aef | 2684 | |
f4474e00 LP |
2685 | return generic_array_bisect( |
2686 | f, | |
2687 | le64toh(f->header->entry_array_offset), | |
2688 | le64toh(f->header->n_entries), | |
2689 | realtime, | |
2690 | test_object_realtime, | |
2691 | direction, | |
2692 | ret, ret_offset, NULL); | |
de190aef LP |
2693 | } |
2694 | ||
2695 | static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { | |
2696 | Object *o; | |
893e0f8f | 2697 | uint64_t m; |
de190aef LP |
2698 | int r; |
2699 | ||
2700 | assert(f); | |
2701 | assert(p > 0); | |
2702 | ||
2703 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2704 | if (r < 0) | |
2705 | return r; | |
2706 | ||
893e0f8f LP |
2707 | m = le64toh(READ_NOW(o->entry.monotonic)); |
2708 | if (m == needle) | |
de190aef | 2709 | return TEST_FOUND; |
893e0f8f | 2710 | else if (m < needle) |
de190aef LP |
2711 | return TEST_LEFT; |
2712 | else | |
2713 | return TEST_RIGHT; | |
2714 | } | |
2715 | ||
2a560338 | 2716 | static int find_data_object_by_boot_id( |
47838ab3 ZJS |
2717 | JournalFile *f, |
2718 | sd_id128_t boot_id, | |
2719 | Object **o, | |
2720 | uint64_t *b) { | |
2a560338 | 2721 | |
fbd0b64f | 2722 | char t[STRLEN("_BOOT_ID=") + 32 + 1] = "_BOOT_ID="; |
47838ab3 ZJS |
2723 | |
2724 | sd_id128_to_string(boot_id, t + 9); | |
2725 | return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b); | |
2726 | } | |
2727 | ||
de190aef LP |
2728 | int journal_file_move_to_entry_by_monotonic( |
2729 | JournalFile *f, | |
2730 | sd_id128_t boot_id, | |
2731 | uint64_t monotonic, | |
2732 | direction_t direction, | |
2733 | Object **ret, | |
f4474e00 | 2734 | uint64_t *ret_offset) { |
de190aef | 2735 | |
de190aef LP |
2736 | Object *o; |
2737 | int r; | |
2738 | ||
cbdca852 | 2739 | assert(f); |
de190aef | 2740 | |
47838ab3 | 2741 | r = find_data_object_by_boot_id(f, boot_id, &o, NULL); |
de190aef LP |
2742 | if (r < 0) |
2743 | return r; | |
cbdca852 | 2744 | if (r == 0) |
de190aef LP |
2745 | return -ENOENT; |
2746 | ||
f4474e00 LP |
2747 | return generic_array_bisect_plus_one( |
2748 | f, | |
2749 | le64toh(o->data.entry_offset), | |
2750 | le64toh(o->data.entry_array_offset), | |
2751 | le64toh(o->data.n_entries), | |
2752 | monotonic, | |
2753 | test_object_monotonic, | |
2754 | direction, | |
2755 | ret, ret_offset, NULL); | |
de190aef LP |
2756 | } |
2757 | ||
1fc605b0 | 2758 | void journal_file_reset_location(JournalFile *f) { |
6573ef05 | 2759 | f->location_type = LOCATION_HEAD; |
1fc605b0 | 2760 | f->current_offset = 0; |
6573ef05 MS |
2761 | f->current_seqnum = 0; |
2762 | f->current_realtime = 0; | |
2763 | f->current_monotonic = 0; | |
2764 | zero(f->current_boot_id); | |
2765 | f->current_xor_hash = 0; | |
2766 | } | |
2767 | ||
950c07d4 | 2768 | void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) { |
6573ef05 MS |
2769 | f->location_type = LOCATION_SEEK; |
2770 | f->current_offset = offset; | |
2771 | f->current_seqnum = le64toh(o->entry.seqnum); | |
2772 | f->current_realtime = le64toh(o->entry.realtime); | |
2773 | f->current_monotonic = le64toh(o->entry.monotonic); | |
2774 | f->current_boot_id = o->entry.boot_id; | |
2775 | f->current_xor_hash = le64toh(o->entry.xor_hash); | |
1fc605b0 MS |
2776 | } |
2777 | ||
d8ae66d7 | 2778 | int journal_file_compare_locations(JournalFile *af, JournalFile *bf) { |
90c88092 YW |
2779 | int r; |
2780 | ||
d8ae66d7 | 2781 | assert(af); |
c88cc6af | 2782 | assert(af->header); |
d8ae66d7 | 2783 | assert(bf); |
c88cc6af | 2784 | assert(bf->header); |
d8ae66d7 MS |
2785 | assert(af->location_type == LOCATION_SEEK); |
2786 | assert(bf->location_type == LOCATION_SEEK); | |
2787 | ||
b17f651a | 2788 | /* If contents, timestamps and seqnum match, these entries are |
7802194a | 2789 | * identical. */ |
d8ae66d7 MS |
2790 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) && |
2791 | af->current_monotonic == bf->current_monotonic && | |
2792 | af->current_realtime == bf->current_realtime && | |
b17f651a | 2793 | af->current_xor_hash == bf->current_xor_hash && |
2794 | sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id) && | |
2795 | af->current_seqnum == bf->current_seqnum) | |
d8ae66d7 MS |
2796 | return 0; |
2797 | ||
2798 | if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) { | |
2799 | ||
2800 | /* If this is from the same seqnum source, compare | |
2801 | * seqnums */ | |
90c88092 YW |
2802 | r = CMP(af->current_seqnum, bf->current_seqnum); |
2803 | if (r != 0) | |
2804 | return r; | |
d8ae66d7 MS |
2805 | |
2806 | /* Wow! This is weird, different data but the same | |
2807 | * seqnums? Something is borked, but let's make the | |
2808 | * best of it and compare by time. */ | |
2809 | } | |
2810 | ||
2811 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) { | |
2812 | ||
2813 | /* If the boot id matches, compare monotonic time */ | |
90c88092 YW |
2814 | r = CMP(af->current_monotonic, bf->current_monotonic); |
2815 | if (r != 0) | |
2816 | return r; | |
d8ae66d7 MS |
2817 | } |
2818 | ||
2819 | /* Otherwise, compare UTC time */ | |
90c88092 YW |
2820 | r = CMP(af->current_realtime, bf->current_realtime); |
2821 | if (r != 0) | |
2822 | return r; | |
d8ae66d7 MS |
2823 | |
2824 | /* Finally, compare by contents */ | |
6dd91b36 | 2825 | return CMP(af->current_xor_hash, bf->current_xor_hash); |
d8ae66d7 MS |
2826 | } |
2827 | ||
aa598ba5 LP |
2828 | static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) { |
2829 | ||
2830 | /* Increase or decrease the specified index, in the right direction. */ | |
2831 | ||
2832 | if (direction == DIRECTION_DOWN) { | |
2833 | if (*i >= n - 1) | |
2834 | return 0; | |
2835 | ||
2836 | (*i) ++; | |
2837 | } else { | |
2838 | if (*i <= 0) | |
2839 | return 0; | |
2840 | ||
2841 | (*i) --; | |
2842 | } | |
2843 | ||
2844 | return 1; | |
2845 | } | |
2846 | ||
b6da4ed0 LP |
2847 | static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) { |
2848 | ||
2849 | /* Consider it an error if any of the two offsets is uninitialized */ | |
2850 | if (old_offset == 0 || new_offset == 0) | |
2851 | return false; | |
2852 | ||
2853 | /* If we go down, the new offset must be larger than the old one. */ | |
2854 | return direction == DIRECTION_DOWN ? | |
2855 | new_offset > old_offset : | |
2856 | new_offset < old_offset; | |
2857 | } | |
2858 | ||
de190aef LP |
2859 | int journal_file_next_entry( |
2860 | JournalFile *f, | |
f534928a | 2861 | uint64_t p, |
de190aef | 2862 | direction_t direction, |
f4474e00 | 2863 | Object **ret, uint64_t *ret_offset) { |
de190aef | 2864 | |
fb099c8d | 2865 | uint64_t i, n, ofs; |
cec736d2 LP |
2866 | int r; |
2867 | ||
2868 | assert(f); | |
c88cc6af | 2869 | assert(f->header); |
de190aef | 2870 | |
893e0f8f | 2871 | n = le64toh(READ_NOW(f->header->n_entries)); |
de190aef LP |
2872 | if (n <= 0) |
2873 | return 0; | |
cec736d2 | 2874 | |
f534928a | 2875 | if (p == 0) |
de190aef | 2876 | i = direction == DIRECTION_DOWN ? 0 : n - 1; |
cec736d2 | 2877 | else { |
de190aef LP |
2878 | r = generic_array_bisect(f, |
2879 | le64toh(f->header->entry_array_offset), | |
2880 | le64toh(f->header->n_entries), | |
2881 | p, | |
2882 | test_object_offset, | |
2883 | DIRECTION_DOWN, | |
2884 | NULL, NULL, | |
2885 | &i); | |
2886 | if (r <= 0) | |
2887 | return r; | |
2888 | ||
aa598ba5 LP |
2889 | r = bump_array_index(&i, direction, n); |
2890 | if (r <= 0) | |
2891 | return r; | |
cec736d2 LP |
2892 | } |
2893 | ||
de190aef | 2894 | /* And jump to it */ |
989793d3 LP |
2895 | for (;;) { |
2896 | r = generic_array_get(f, | |
2897 | le64toh(f->header->entry_array_offset), | |
2898 | i, | |
2899 | ret, &ofs); | |
2900 | if (r > 0) | |
2901 | break; | |
2902 | if (r != -EBADMSG) | |
2903 | return r; | |
2904 | ||
2905 | /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if | |
2906 | * the next one might work for us instead. */ | |
2907 | log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i); | |
2908 | ||
2909 | r = bump_array_index(&i, direction, n); | |
2910 | if (r <= 0) | |
2911 | return r; | |
caeab8f6 | 2912 | } |
fb099c8d | 2913 | |
b6da4ed0 | 2914 | /* Ensure our array is properly ordered. */ |
baaa35ad ZJS |
2915 | if (p > 0 && !check_properly_ordered(ofs, p, direction)) |
2916 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2917 | "%s: entry array not properly ordered at entry %" PRIu64, | |
2918 | f->path, i); | |
fb099c8d | 2919 | |
f4474e00 LP |
2920 | if (ret_offset) |
2921 | *ret_offset = ofs; | |
fb099c8d ZJS |
2922 | |
2923 | return 1; | |
de190aef | 2924 | } |
cec736d2 | 2925 | |
de190aef LP |
2926 | int journal_file_next_entry_for_data( |
2927 | JournalFile *f, | |
2928 | Object *o, uint64_t p, | |
2929 | uint64_t data_offset, | |
2930 | direction_t direction, | |
f4474e00 | 2931 | Object **ret, uint64_t *ret_offset) { |
de190aef | 2932 | |
ded5034e | 2933 | uint64_t i, n, ofs; |
de190aef | 2934 | Object *d; |
989793d3 | 2935 | int r; |
cec736d2 LP |
2936 | |
2937 | assert(f); | |
de190aef | 2938 | assert(p > 0 || !o); |
cec736d2 | 2939 | |
de190aef | 2940 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
466ccd92 | 2941 | if (r < 0) |
de190aef | 2942 | return r; |
cec736d2 | 2943 | |
893e0f8f | 2944 | n = le64toh(READ_NOW(d->data.n_entries)); |
de190aef LP |
2945 | if (n <= 0) |
2946 | return n; | |
cec736d2 | 2947 | |
de190aef LP |
2948 | if (!o) |
2949 | i = direction == DIRECTION_DOWN ? 0 : n - 1; | |
2950 | else { | |
2951 | if (o->object.type != OBJECT_ENTRY) | |
2952 | return -EINVAL; | |
cec736d2 | 2953 | |
de190aef LP |
2954 | r = generic_array_bisect_plus_one(f, |
2955 | le64toh(d->data.entry_offset), | |
2956 | le64toh(d->data.entry_array_offset), | |
2957 | le64toh(d->data.n_entries), | |
2958 | p, | |
2959 | test_object_offset, | |
2960 | DIRECTION_DOWN, | |
2961 | NULL, NULL, | |
2962 | &i); | |
2963 | ||
2964 | if (r <= 0) | |
cec736d2 LP |
2965 | return r; |
2966 | ||
aa598ba5 LP |
2967 | r = bump_array_index(&i, direction, n); |
2968 | if (r <= 0) | |
2969 | return r; | |
de190aef | 2970 | } |
cec736d2 | 2971 | |
989793d3 LP |
2972 | for (;;) { |
2973 | r = generic_array_get_plus_one(f, | |
2974 | le64toh(d->data.entry_offset), | |
2975 | le64toh(d->data.entry_array_offset), | |
2976 | i, | |
2977 | ret, &ofs); | |
2978 | if (r > 0) | |
2979 | break; | |
2980 | if (r != -EBADMSG) | |
2981 | return r; | |
2982 | ||
2983 | log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i); | |
2984 | ||
2985 | r = bump_array_index(&i, direction, n); | |
2986 | if (r <= 0) | |
2987 | return r; | |
2988 | } | |
ded5034e LP |
2989 | |
2990 | /* Ensure our array is properly ordered. */ | |
baaa35ad ZJS |
2991 | if (p > 0 && check_properly_ordered(ofs, p, direction)) |
2992 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2993 | "%s data entry array not properly ordered at entry %" PRIu64, | |
2994 | f->path, i); | |
ded5034e | 2995 | |
f4474e00 LP |
2996 | if (ret_offset) |
2997 | *ret_offset = ofs; | |
ded5034e LP |
2998 | |
2999 | return 1; | |
de190aef | 3000 | } |
cec736d2 | 3001 | |
cbdca852 LP |
3002 | int journal_file_move_to_entry_by_offset_for_data( |
3003 | JournalFile *f, | |
3004 | uint64_t data_offset, | |
3005 | uint64_t p, | |
3006 | direction_t direction, | |
f4474e00 | 3007 | Object **ret, uint64_t *ret_offset) { |
cbdca852 LP |
3008 | |
3009 | int r; | |
3010 | Object *d; | |
3011 | ||
3012 | assert(f); | |
3013 | ||
3014 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
3015 | if (r < 0) | |
3016 | return r; | |
3017 | ||
f4474e00 LP |
3018 | return generic_array_bisect_plus_one( |
3019 | f, | |
3020 | le64toh(d->data.entry_offset), | |
3021 | le64toh(d->data.entry_array_offset), | |
3022 | le64toh(d->data.n_entries), | |
3023 | p, | |
3024 | test_object_offset, | |
3025 | direction, | |
3026 | ret, ret_offset, NULL); | |
cbdca852 LP |
3027 | } |
3028 | ||
3029 | int journal_file_move_to_entry_by_monotonic_for_data( | |
3030 | JournalFile *f, | |
3031 | uint64_t data_offset, | |
3032 | sd_id128_t boot_id, | |
3033 | uint64_t monotonic, | |
3034 | direction_t direction, | |
f4474e00 | 3035 | Object **ret, uint64_t *ret_offset) { |
cbdca852 | 3036 | |
cbdca852 LP |
3037 | Object *o, *d; |
3038 | int r; | |
3039 | uint64_t b, z; | |
3040 | ||
3041 | assert(f); | |
3042 | ||
3043 | /* First, seek by time */ | |
47838ab3 | 3044 | r = find_data_object_by_boot_id(f, boot_id, &o, &b); |
cbdca852 LP |
3045 | if (r < 0) |
3046 | return r; | |
3047 | if (r == 0) | |
3048 | return -ENOENT; | |
3049 | ||
3050 | r = generic_array_bisect_plus_one(f, | |
3051 | le64toh(o->data.entry_offset), | |
3052 | le64toh(o->data.entry_array_offset), | |
3053 | le64toh(o->data.n_entries), | |
3054 | monotonic, | |
3055 | test_object_monotonic, | |
3056 | direction, | |
3057 | NULL, &z, NULL); | |
3058 | if (r <= 0) | |
3059 | return r; | |
3060 | ||
3061 | /* And now, continue seeking until we find an entry that | |
3062 | * exists in both bisection arrays */ | |
3063 | ||
3064 | for (;;) { | |
3065 | Object *qo; | |
3066 | uint64_t p, q; | |
3067 | ||
3068 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
3069 | if (r < 0) | |
3070 | return r; | |
3071 | ||
3072 | r = generic_array_bisect_plus_one(f, | |
3073 | le64toh(d->data.entry_offset), | |
3074 | le64toh(d->data.entry_array_offset), | |
3075 | le64toh(d->data.n_entries), | |
3076 | z, | |
3077 | test_object_offset, | |
3078 | direction, | |
3079 | NULL, &p, NULL); | |
3080 | if (r <= 0) | |
3081 | return r; | |
3082 | ||
3083 | r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); | |
3084 | if (r < 0) | |
3085 | return r; | |
3086 | ||
3087 | r = generic_array_bisect_plus_one(f, | |
3088 | le64toh(o->data.entry_offset), | |
3089 | le64toh(o->data.entry_array_offset), | |
3090 | le64toh(o->data.n_entries), | |
3091 | p, | |
3092 | test_object_offset, | |
3093 | direction, | |
3094 | &qo, &q, NULL); | |
3095 | ||
3096 | if (r <= 0) | |
3097 | return r; | |
3098 | ||
3099 | if (p == q) { | |
3100 | if (ret) | |
3101 | *ret = qo; | |
f4474e00 LP |
3102 | if (ret_offset) |
3103 | *ret_offset = q; | |
cbdca852 LP |
3104 | |
3105 | return 1; | |
3106 | } | |
3107 | ||
3108 | z = q; | |
3109 | } | |
cbdca852 LP |
3110 | } |
3111 | ||
de190aef LP |
3112 | int journal_file_move_to_entry_by_seqnum_for_data( |
3113 | JournalFile *f, | |
3114 | uint64_t data_offset, | |
3115 | uint64_t seqnum, | |
3116 | direction_t direction, | |
f4474e00 | 3117 | Object **ret, uint64_t *ret_offset) { |
cec736d2 | 3118 | |
de190aef LP |
3119 | Object *d; |
3120 | int r; | |
cec736d2 | 3121 | |
91a31dde LP |
3122 | assert(f); |
3123 | ||
de190aef | 3124 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 3125 | if (r < 0) |
de190aef | 3126 | return r; |
cec736d2 | 3127 | |
f4474e00 LP |
3128 | return generic_array_bisect_plus_one( |
3129 | f, | |
3130 | le64toh(d->data.entry_offset), | |
3131 | le64toh(d->data.entry_array_offset), | |
3132 | le64toh(d->data.n_entries), | |
3133 | seqnum, | |
3134 | test_object_seqnum, | |
3135 | direction, | |
3136 | ret, ret_offset, NULL); | |
de190aef | 3137 | } |
cec736d2 | 3138 | |
de190aef LP |
3139 | int journal_file_move_to_entry_by_realtime_for_data( |
3140 | JournalFile *f, | |
3141 | uint64_t data_offset, | |
3142 | uint64_t realtime, | |
3143 | direction_t direction, | |
f4474e00 | 3144 | Object **ret, uint64_t *ret_offset) { |
de190aef LP |
3145 | |
3146 | Object *d; | |
3147 | int r; | |
3148 | ||
91a31dde LP |
3149 | assert(f); |
3150 | ||
de190aef | 3151 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 3152 | if (r < 0) |
de190aef LP |
3153 | return r; |
3154 | ||
f4474e00 LP |
3155 | return generic_array_bisect_plus_one( |
3156 | f, | |
3157 | le64toh(d->data.entry_offset), | |
3158 | le64toh(d->data.entry_array_offset), | |
3159 | le64toh(d->data.n_entries), | |
3160 | realtime, | |
3161 | test_object_realtime, | |
3162 | direction, | |
3163 | ret, ret_offset, NULL); | |
cec736d2 LP |
3164 | } |
3165 | ||
0284adc6 | 3166 | void journal_file_dump(JournalFile *f) { |
7560fffc | 3167 | Object *o; |
7560fffc | 3168 | int r; |
0284adc6 | 3169 | uint64_t p; |
7560fffc LP |
3170 | |
3171 | assert(f); | |
c88cc6af | 3172 | assert(f->header); |
7560fffc | 3173 | |
0284adc6 | 3174 | journal_file_print_header(f); |
7560fffc | 3175 | |
893e0f8f | 3176 | p = le64toh(READ_NOW(f->header->header_size)); |
0284adc6 | 3177 | while (p != 0) { |
363b2b9a DDM |
3178 | const char *s; |
3179 | ||
d05089d8 | 3180 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); |
0284adc6 LP |
3181 | if (r < 0) |
3182 | goto fail; | |
7560fffc | 3183 | |
363b2b9a | 3184 | s = journal_object_type_to_string(o->object.type); |
7560fffc | 3185 | |
363b2b9a | 3186 | switch (o->object.type) { |
3c1668da | 3187 | |
0284adc6 | 3188 | case OBJECT_ENTRY: |
363b2b9a DDM |
3189 | assert(s); |
3190 | ||
3191 | printf("Type: %s seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n", | |
3192 | s, | |
507f22bd ZJS |
3193 | le64toh(o->entry.seqnum), |
3194 | le64toh(o->entry.monotonic), | |
3195 | le64toh(o->entry.realtime)); | |
0284adc6 | 3196 | break; |
7560fffc | 3197 | |
0284adc6 | 3198 | case OBJECT_TAG: |
363b2b9a DDM |
3199 | assert(s); |
3200 | ||
3201 | printf("Type: %s seqnum=%"PRIu64" epoch=%"PRIu64"\n", | |
3202 | s, | |
507f22bd ZJS |
3203 | le64toh(o->tag.seqnum), |
3204 | le64toh(o->tag.epoch)); | |
0284adc6 | 3205 | break; |
3c1668da LP |
3206 | |
3207 | default: | |
363b2b9a DDM |
3208 | if (s) |
3209 | printf("Type: %s \n", s); | |
3210 | else | |
3211 | printf("Type: unknown (%i)", o->object.type); | |
3212 | ||
3c1668da | 3213 | break; |
0284adc6 | 3214 | } |
7560fffc | 3215 | |
d89c8fdf ZJS |
3216 | if (o->object.flags & OBJECT_COMPRESSION_MASK) |
3217 | printf("Flags: %s\n", | |
3218 | object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK)); | |
7560fffc | 3219 | |
0284adc6 LP |
3220 | if (p == le64toh(f->header->tail_object_offset)) |
3221 | p = 0; | |
3222 | else | |
71139898 | 3223 | p += ALIGN64(le64toh(o->object.size)); |
0284adc6 | 3224 | } |
7560fffc | 3225 | |
0284adc6 LP |
3226 | return; |
3227 | fail: | |
3228 | log_error("File corrupt"); | |
7560fffc LP |
3229 | } |
3230 | ||
5e62ac8b ZJS |
3231 | /* Note: the lifetime of the compound literal is the immediately surrounding block. */ |
3232 | #define FORMAT_TIMESTAMP_SAFE(t) (FORMAT_TIMESTAMP(t) ?: " --- ") | |
718fe4b1 | 3233 | |
0284adc6 | 3234 | void journal_file_print_header(JournalFile *f) { |
a1a03e30 | 3235 | struct stat st; |
7560fffc LP |
3236 | |
3237 | assert(f); | |
c88cc6af | 3238 | assert(f->header); |
7560fffc | 3239 | |
2c54acb1 | 3240 | printf("File path: %s\n" |
0284adc6 LP |
3241 | "File ID: %s\n" |
3242 | "Machine ID: %s\n" | |
3243 | "Boot ID: %s\n" | |
2c54acb1 | 3244 | "Sequential number ID: %s\n" |
0284adc6 | 3245 | "State: %s\n" |
2c54acb1 | 3246 | "Compatible flags:%s%s\n" |
8653185a | 3247 | "Incompatible flags:%s%s%s%s%s\n" |
507f22bd ZJS |
3248 | "Header size: %"PRIu64"\n" |
3249 | "Arena size: %"PRIu64"\n" | |
2c54acb1 TN |
3250 | "Data hash table size: %"PRIu64"\n" |
3251 | "Field hash table size: %"PRIu64"\n" | |
3252 | "Rotate suggested: %s\n" | |
3253 | "Head sequential number: %"PRIu64" (%"PRIx64")\n" | |
3254 | "Tail sequential number: %"PRIu64" (%"PRIx64")\n" | |
3255 | "Head realtime timestamp: %s (%"PRIx64")\n" | |
3256 | "Tail realtime timestamp: %s (%"PRIx64")\n" | |
3257 | "Tail monotonic timestamp: %s (%"PRIx64")\n" | |
507f22bd | 3258 | "Objects: %"PRIu64"\n" |
2c54acb1 | 3259 | "Entry objects: %"PRIu64"\n", |
0284adc6 | 3260 | f->path, |
85b55869 LP |
3261 | SD_ID128_TO_STRING(f->header->file_id), |
3262 | SD_ID128_TO_STRING(f->header->machine_id), | |
3263 | SD_ID128_TO_STRING(f->header->boot_id), | |
3264 | SD_ID128_TO_STRING(f->header->seqnum_id), | |
3223f44f LP |
3265 | f->header->state == STATE_OFFLINE ? "OFFLINE" : |
3266 | f->header->state == STATE_ONLINE ? "ONLINE" : | |
3267 | f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN", | |
8088cbd3 | 3268 | JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "", |
d89c8fdf ZJS |
3269 | (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "", |
3270 | JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "", | |
3271 | JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "", | |
8653185a | 3272 | JOURNAL_HEADER_COMPRESSED_ZSTD(f->header) ? " COMPRESSED-ZSTD" : "", |
4ce534f4 | 3273 | JOURNAL_HEADER_KEYED_HASH(f->header) ? " KEYED-HASH" : "", |
d89c8fdf | 3274 | (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "", |
507f22bd ZJS |
3275 | le64toh(f->header->header_size), |
3276 | le64toh(f->header->arena_size), | |
3277 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
3278 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem), | |
c8e6e1f1 | 3279 | yes_no(journal_file_rotate_suggested(f, 0, LOG_DEBUG)), |
0808b92f LP |
3280 | le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum), |
3281 | le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum), | |
5e62ac8b ZJS |
3282 | FORMAT_TIMESTAMP_SAFE(le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime), |
3283 | FORMAT_TIMESTAMP_SAFE(le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime), | |
5291f26d | 3284 | FORMAT_TIMESPAN(le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic), |
507f22bd ZJS |
3285 | le64toh(f->header->n_objects), |
3286 | le64toh(f->header->n_entries)); | |
7560fffc | 3287 | |
0284adc6 | 3288 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
2c54acb1 TN |
3289 | printf("Data objects: %"PRIu64"\n" |
3290 | "Data hash table fill: %.1f%%\n", | |
507f22bd | 3291 | le64toh(f->header->n_data), |
0284adc6 | 3292 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)))); |
7560fffc | 3293 | |
0284adc6 | 3294 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) |
2c54acb1 TN |
3295 | printf("Field objects: %"PRIu64"\n" |
3296 | "Field hash table fill: %.1f%%\n", | |
507f22bd | 3297 | le64toh(f->header->n_fields), |
0284adc6 | 3298 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); |
3223f44f LP |
3299 | |
3300 | if (JOURNAL_HEADER_CONTAINS(f->header, n_tags)) | |
2c54acb1 | 3301 | printf("Tag objects: %"PRIu64"\n", |
507f22bd | 3302 | le64toh(f->header->n_tags)); |
3223f44f | 3303 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
2c54acb1 | 3304 | printf("Entry array objects: %"PRIu64"\n", |
507f22bd | 3305 | le64toh(f->header->n_entry_arrays)); |
a1a03e30 | 3306 | |
0dbe57ee LP |
3307 | if (JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth)) |
3308 | printf("Deepest field hash chain: %" PRIu64"\n", | |
3309 | f->header->field_hash_chain_depth); | |
3310 | ||
3311 | if (JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth)) | |
3312 | printf("Deepest data hash chain: %" PRIu64"\n", | |
3313 | f->header->data_hash_chain_depth); | |
3314 | ||
a1a03e30 | 3315 | if (fstat(f->fd, &st) >= 0) |
2b59bf51 | 3316 | printf("Disk usage: %s\n", FORMAT_BYTES((uint64_t) st.st_blocks * 512ULL)); |
7560fffc LP |
3317 | } |
3318 | ||
fc68c929 LP |
3319 | static int journal_file_warn_btrfs(JournalFile *f) { |
3320 | unsigned attrs; | |
3321 | int r; | |
3322 | ||
3323 | assert(f); | |
3324 | ||
3325 | /* Before we write anything, check if the COW logic is turned | |
3326 | * off on btrfs. Given our write pattern that is quite | |
3327 | * unfriendly to COW file systems this should greatly improve | |
3328 | * performance on COW file systems, such as btrfs, at the | |
3329 | * expense of data integrity features (which shouldn't be too | |
3330 | * bad, given that we do our own checksumming). */ | |
3331 | ||
65ddc2c5 | 3332 | r = fd_is_fs_type(f->fd, BTRFS_SUPER_MAGIC); |
fc68c929 LP |
3333 | if (r < 0) |
3334 | return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m"); | |
3335 | if (!r) | |
3336 | return 0; | |
3337 | ||
3338 | r = read_attr_fd(f->fd, &attrs); | |
3339 | if (r < 0) | |
3340 | return log_warning_errno(r, "Failed to read file attributes: %m"); | |
3341 | ||
3342 | if (attrs & FS_NOCOW_FL) { | |
3343 | log_debug("Detected btrfs file system with copy-on-write disabled, all is good."); | |
3344 | return 0; | |
3345 | } | |
3346 | ||
3347 | log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. " | |
3348 | "This is likely to slow down journal access substantially, please consider turning " | |
3349 | "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path); | |
3350 | ||
3351 | return 1; | |
3352 | } | |
3353 | ||
0284adc6 | 3354 | int journal_file_open( |
5d1ce257 | 3355 | int fd, |
0284adc6 LP |
3356 | const char *fname, |
3357 | int flags, | |
3358 | mode_t mode, | |
3359 | bool compress, | |
57850536 | 3360 | uint64_t compress_threshold_bytes, |
baed47c3 | 3361 | bool seal, |
0284adc6 LP |
3362 | JournalMetrics *metrics, |
3363 | MMapCache *mmap_cache, | |
b58c888f | 3364 | Set *deferred_closes, |
0284adc6 LP |
3365 | JournalFile *template, |
3366 | JournalFile **ret) { | |
7560fffc | 3367 | |
fa6ac760 | 3368 | bool newly_created = false; |
0284adc6 | 3369 | JournalFile *f; |
fa6ac760 | 3370 | void *h; |
0284adc6 | 3371 | int r; |
7560fffc | 3372 | |
0559d3a5 | 3373 | assert(ret); |
5d1ce257 | 3374 | assert(fd >= 0 || fname); |
7560fffc | 3375 | |
ec2ce0c5 | 3376 | if (!IN_SET((flags & O_ACCMODE), O_RDONLY, O_RDWR)) |
0284adc6 | 3377 | return -EINVAL; |
7560fffc | 3378 | |
6eda13d3 LP |
3379 | if (fname && (flags & O_CREAT) && !endswith(fname, ".journal")) |
3380 | return -EINVAL; | |
7560fffc | 3381 | |
971b52c4 | 3382 | f = new(JournalFile, 1); |
0284adc6 LP |
3383 | if (!f) |
3384 | return -ENOMEM; | |
7560fffc | 3385 | |
971b52c4 LP |
3386 | *f = (JournalFile) { |
3387 | .fd = fd, | |
3388 | .mode = mode, | |
3389 | ||
3390 | .flags = flags, | |
971b52c4 | 3391 | .writable = (flags & O_ACCMODE) != O_RDONLY, |
7560fffc | 3392 | |
8653185a LP |
3393 | #if HAVE_ZSTD |
3394 | .compress_zstd = compress, | |
3395 | #elif HAVE_LZ4 | |
971b52c4 | 3396 | .compress_lz4 = compress, |
349cc4a5 | 3397 | #elif HAVE_XZ |
971b52c4 | 3398 | .compress_xz = compress, |
48b61739 | 3399 | #endif |
f5fbe71d | 3400 | .compress_threshold_bytes = compress_threshold_bytes == UINT64_MAX ? |
971b52c4 LP |
3401 | DEFAULT_COMPRESS_THRESHOLD : |
3402 | MAX(MIN_COMPRESS_THRESHOLD, compress_threshold_bytes), | |
349cc4a5 | 3403 | #if HAVE_GCRYPT |
971b52c4 | 3404 | .seal = seal, |
49a32d43 | 3405 | #endif |
971b52c4 | 3406 | }; |
7560fffc | 3407 | |
4ce534f4 LP |
3408 | /* We turn on keyed hashes by default, but provide an environment variable to turn them off, if |
3409 | * people really want that */ | |
3410 | r = getenv_bool("SYSTEMD_JOURNAL_KEYED_HASH"); | |
3411 | if (r < 0) { | |
3412 | if (r != -ENXIO) | |
3413 | log_debug_errno(r, "Failed to parse $SYSTEMD_JOURNAL_KEYED_HASH environment variable, ignoring."); | |
3414 | f->keyed_hash = true; | |
3415 | } else | |
3416 | f->keyed_hash = r; | |
3417 | ||
170a434c | 3418 | if (DEBUG_LOGGING) { |
4ce534f4 | 3419 | static int last_seal = -1, last_compress = -1, last_keyed_hash = -1; |
170a434c | 3420 | static uint64_t last_bytes = UINT64_MAX; |
170a434c ZJS |
3421 | |
3422 | if (last_seal != f->seal || | |
4ce534f4 | 3423 | last_keyed_hash != f->keyed_hash || |
170a434c ZJS |
3424 | last_compress != JOURNAL_FILE_COMPRESS(f) || |
3425 | last_bytes != f->compress_threshold_bytes) { | |
3426 | ||
4ce534f4 LP |
3427 | log_debug("Journal effective settings seal=%s keyed_hash=%s compress=%s compress_threshold_bytes=%s", |
3428 | yes_no(f->seal), yes_no(f->keyed_hash), yes_no(JOURNAL_FILE_COMPRESS(f)), | |
2b59bf51 | 3429 | FORMAT_BYTES(f->compress_threshold_bytes)); |
170a434c | 3430 | last_seal = f->seal; |
4ce534f4 | 3431 | last_keyed_hash = f->keyed_hash; |
170a434c ZJS |
3432 | last_compress = JOURNAL_FILE_COMPRESS(f); |
3433 | last_bytes = f->compress_threshold_bytes; | |
3434 | } | |
3435 | } | |
57850536 | 3436 | |
0284adc6 LP |
3437 | if (mmap_cache) |
3438 | f->mmap = mmap_cache_ref(mmap_cache); | |
3439 | else { | |
84168d80 | 3440 | f->mmap = mmap_cache_new(); |
0284adc6 LP |
3441 | if (!f->mmap) { |
3442 | r = -ENOMEM; | |
3443 | goto fail; | |
3444 | } | |
3445 | } | |
7560fffc | 3446 | |
7645c77b | 3447 | if (fname) { |
5d1ce257 | 3448 | f->path = strdup(fname); |
7645c77b ZJS |
3449 | if (!f->path) { |
3450 | r = -ENOMEM; | |
3451 | goto fail; | |
3452 | } | |
3453 | } else { | |
817b1c5b LP |
3454 | assert(fd >= 0); |
3455 | ||
7645c77b ZJS |
3456 | /* If we don't know the path, fill in something explanatory and vaguely useful */ |
3457 | if (asprintf(&f->path, "/proc/self/%i", fd) < 0) { | |
3458 | r = -ENOMEM; | |
3459 | goto fail; | |
3460 | } | |
0284adc6 | 3461 | } |
7560fffc | 3462 | |
4743015d | 3463 | f->chain_cache = ordered_hashmap_new(&uint64_hash_ops); |
a4bcff5b LP |
3464 | if (!f->chain_cache) { |
3465 | r = -ENOMEM; | |
3466 | goto fail; | |
3467 | } | |
3468 | ||
0284adc6 | 3469 | if (f->fd < 0) { |
817b1c5b LP |
3470 | /* We pass O_NONBLOCK here, so that in case somebody pointed us to some character device node or FIFO |
3471 | * or so, we likely fail quickly than block for long. For regular files O_NONBLOCK has no effect, hence | |
3472 | * it doesn't hurt in that case. */ | |
3473 | ||
3474 | f->fd = open(f->path, f->flags|O_CLOEXEC|O_NONBLOCK, f->mode); | |
5d1ce257 LP |
3475 | if (f->fd < 0) { |
3476 | r = -errno; | |
3477 | goto fail; | |
3478 | } | |
3479 | ||
3480 | /* fds we opened here by us should also be closed by us. */ | |
3481 | f->close_fd = true; | |
817b1c5b LP |
3482 | |
3483 | r = fd_nonblock(f->fd, false); | |
3484 | if (r < 0) | |
3485 | goto fail; | |
7560fffc | 3486 | } |
7560fffc | 3487 | |
104fc4be | 3488 | f->cache_fd = mmap_cache_add_fd(f->mmap, f->fd, prot_from_flags(flags)); |
be7cdd8e VC |
3489 | if (!f->cache_fd) { |
3490 | r = -ENOMEM; | |
3491 | goto fail; | |
3492 | } | |
3493 | ||
2678031a LP |
3494 | r = journal_file_fstat(f); |
3495 | if (r < 0) | |
0284adc6 | 3496 | goto fail; |
7560fffc | 3497 | |
0284adc6 | 3498 | if (f->last_stat.st_size == 0 && f->writable) { |
11689d2a | 3499 | |
fc68c929 | 3500 | (void) journal_file_warn_btrfs(f); |
11689d2a | 3501 | |
4c2e1b39 LP |
3502 | /* Let's attach the creation time to the journal file, so that the vacuuming code knows the age of this |
3503 | * file even if the file might end up corrupted one day... Ideally we'd just use the creation time many | |
3504 | * file systems maintain for each file, but the API to query this is very new, hence let's emulate this | |
3505 | * via extended attributes. If extended attributes are not supported we'll just skip this, and rely | |
3506 | * solely on mtime/atime/ctime of the file. */ | |
3507 | (void) fd_setcrtime(f->fd, 0); | |
7560fffc | 3508 | |
349cc4a5 | 3509 | #if HAVE_GCRYPT |
0284adc6 | 3510 | /* Try to load the FSPRG state, and if we can't, then |
baed47c3 | 3511 | * just don't do sealing */ |
49a32d43 LP |
3512 | if (f->seal) { |
3513 | r = journal_file_fss_load(f); | |
3514 | if (r < 0) | |
3515 | f->seal = false; | |
3516 | } | |
feb12d3e | 3517 | #endif |
7560fffc | 3518 | |
0284adc6 LP |
3519 | r = journal_file_init_header(f, template); |
3520 | if (r < 0) | |
3521 | goto fail; | |
7560fffc | 3522 | |
2678031a LP |
3523 | r = journal_file_fstat(f); |
3524 | if (r < 0) | |
0284adc6 | 3525 | goto fail; |
fb0951b0 LP |
3526 | |
3527 | newly_created = true; | |
0284adc6 | 3528 | } |
7560fffc | 3529 | |
0284adc6 | 3530 | if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { |
cfb571f3 | 3531 | r = -ENODATA; |
0284adc6 LP |
3532 | goto fail; |
3533 | } | |
7560fffc | 3534 | |
c3bd54bf | 3535 | r = mmap_cache_fd_get(f->cache_fd, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h); |
5087825e LP |
3536 | if (r == -EINVAL) { |
3537 | /* Some file systems (jffs2 or p9fs) don't support mmap() properly (or only read-only | |
3538 | * mmap()), and return EINVAL in that case. Let's propagate that as a more recognizable error | |
3539 | * code. */ | |
3540 | r = -EAFNOSUPPORT; | |
3541 | goto fail; | |
3542 | } | |
977eaa1e | 3543 | if (r < 0) |
0284adc6 | 3544 | goto fail; |
7560fffc | 3545 | |
fa6ac760 LP |
3546 | f->header = h; |
3547 | ||
0284adc6 | 3548 | if (!newly_created) { |
f9168190 | 3549 | set_clear_with_destructor(deferred_closes, journal_file_close); |
b58c888f | 3550 | |
0284adc6 LP |
3551 | r = journal_file_verify_header(f); |
3552 | if (r < 0) | |
3553 | goto fail; | |
3554 | } | |
7560fffc | 3555 | |
349cc4a5 | 3556 | #if HAVE_GCRYPT |
0284adc6 | 3557 | if (!newly_created && f->writable) { |
baed47c3 | 3558 | r = journal_file_fss_load(f); |
0284adc6 LP |
3559 | if (r < 0) |
3560 | goto fail; | |
3561 | } | |
feb12d3e | 3562 | #endif |
cec736d2 LP |
3563 | |
3564 | if (f->writable) { | |
4a92baf3 LP |
3565 | if (metrics) { |
3566 | journal_default_metrics(metrics, f->fd); | |
3567 | f->metrics = *metrics; | |
3568 | } else if (template) | |
3569 | f->metrics = template->metrics; | |
3570 | ||
cec736d2 LP |
3571 | r = journal_file_refresh_header(f); |
3572 | if (r < 0) | |
3573 | goto fail; | |
3574 | } | |
3575 | ||
349cc4a5 | 3576 | #if HAVE_GCRYPT |
baed47c3 | 3577 | r = journal_file_hmac_setup(f); |
14d10188 LP |
3578 | if (r < 0) |
3579 | goto fail; | |
feb12d3e | 3580 | #endif |
14d10188 | 3581 | |
cec736d2 | 3582 | if (newly_created) { |
de190aef | 3583 | r = journal_file_setup_field_hash_table(f); |
cec736d2 LP |
3584 | if (r < 0) |
3585 | goto fail; | |
3586 | ||
de190aef | 3587 | r = journal_file_setup_data_hash_table(f); |
cec736d2 LP |
3588 | if (r < 0) |
3589 | goto fail; | |
7560fffc | 3590 | |
349cc4a5 | 3591 | #if HAVE_GCRYPT |
7560fffc LP |
3592 | r = journal_file_append_first_tag(f); |
3593 | if (r < 0) | |
3594 | goto fail; | |
feb12d3e | 3595 | #endif |
cec736d2 LP |
3596 | } |
3597 | ||
c3bd54bf | 3598 | if (mmap_cache_fd_got_sigbus(f->cache_fd)) { |
fa6ac760 LP |
3599 | r = -EIO; |
3600 | goto fail; | |
3601 | } | |
3602 | ||
7a24f3bf | 3603 | if (template && template->post_change_timer) { |
e167d7fd LP |
3604 | r = journal_file_enable_post_change_timer( |
3605 | f, | |
3606 | sd_event_source_get_event(template->post_change_timer), | |
3607 | template->post_change_timer_period); | |
7a24f3bf | 3608 | |
7a24f3bf VC |
3609 | if (r < 0) |
3610 | goto fail; | |
3611 | } | |
3612 | ||
f8e2f4d6 | 3613 | /* The file is opened now successfully, thus we take possession of any passed in fd. */ |
5d1ce257 LP |
3614 | f->close_fd = true; |
3615 | ||
0559d3a5 | 3616 | *ret = f; |
cec736d2 LP |
3617 | return 0; |
3618 | ||
3619 | fail: | |
c3bd54bf | 3620 | if (f->cache_fd && mmap_cache_fd_got_sigbus(f->cache_fd)) |
fa6ac760 LP |
3621 | r = -EIO; |
3622 | ||
69a3a6fd | 3623 | (void) journal_file_close(f); |
cec736d2 LP |
3624 | |
3625 | return r; | |
3626 | } | |
0ac38b70 | 3627 | |
7a4d21ad | 3628 | int journal_file_archive(JournalFile *f) { |
57535f47 | 3629 | _cleanup_free_ char *p = NULL; |
0ac38b70 LP |
3630 | |
3631 | assert(f); | |
0ac38b70 | 3632 | |
7a4d21ad | 3633 | if (!f->writable) |
0ac38b70 LP |
3634 | return -EINVAL; |
3635 | ||
5d1ce257 | 3636 | /* Is this a journal file that was passed to us as fd? If so, we synthesized a path name for it, and we refuse |
13e785f7 | 3637 | * rotation, since we don't know the actual path, and couldn't rename the file hence. */ |
7a4d21ad | 3638 | if (path_startswith(f->path, "/proc/self/fd")) |
5d1ce257 LP |
3639 | return -EINVAL; |
3640 | ||
7a4d21ad | 3641 | if (!endswith(f->path, ".journal")) |
0ac38b70 LP |
3642 | return -EINVAL; |
3643 | ||
7a4d21ad LP |
3644 | if (asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal", |
3645 | (int) strlen(f->path) - 8, f->path, | |
3646 | SD_ID128_FORMAT_VAL(f->header->seqnum_id), | |
3647 | le64toh(f->header->head_entry_seqnum), | |
3648 | le64toh(f->header->head_entry_realtime)) < 0) | |
0ac38b70 LP |
3649 | return -ENOMEM; |
3650 | ||
7a4d21ad LP |
3651 | /* Try to rename the file to the archived version. If the file already was deleted, we'll get ENOENT, let's |
3652 | * ignore that case. */ | |
3653 | if (rename(f->path, p) < 0 && errno != ENOENT) | |
0ac38b70 LP |
3654 | return -errno; |
3655 | ||
1fcefd88 | 3656 | /* Sync the rename to disk */ |
7a4d21ad LP |
3657 | (void) fsync_directory_of_file(f->fd); |
3658 | ||
3659 | /* Set as archive so offlining commits w/state=STATE_ARCHIVED. Previously we would set old_file->header->state | |
3660 | * to STATE_ARCHIVED directly here, but journal_file_set_offline() short-circuits when state != STATE_ONLINE, | |
3661 | * which would result in the rotated journal never getting fsync() called before closing. Now we simply queue | |
3662 | * the archive state by setting an archive bit, leaving the state as STATE_ONLINE so proper offlining | |
3663 | * occurs. */ | |
3664 | f->archive = true; | |
3665 | ||
3666 | /* Currently, btrfs is not very good with out write patterns and fragments heavily. Let's defrag our journal | |
3667 | * files when we archive them */ | |
3668 | f->defrag_on_close = true; | |
3669 | ||
3670 | return 0; | |
3671 | } | |
3672 | ||
3673 | JournalFile* journal_initiate_close( | |
3674 | JournalFile *f, | |
3675 | Set *deferred_closes) { | |
3676 | ||
3677 | int r; | |
3678 | ||
3679 | assert(f); | |
3680 | ||
3681 | if (deferred_closes) { | |
0ac38b70 | 3682 | |
7a4d21ad LP |
3683 | r = set_put(deferred_closes, f); |
3684 | if (r < 0) | |
3685 | log_debug_errno(r, "Failed to add file to deferred close set, closing immediately."); | |
3686 | else { | |
3687 | (void) journal_file_set_offline(f, false); | |
3688 | return NULL; | |
3689 | } | |
3690 | } | |
3691 | ||
3692 | return journal_file_close(f); | |
3693 | } | |
3694 | ||
3695 | int journal_file_rotate( | |
3696 | JournalFile **f, | |
3697 | bool compress, | |
3698 | uint64_t compress_threshold_bytes, | |
3699 | bool seal, | |
3700 | Set *deferred_closes) { | |
3701 | ||
3702 | JournalFile *new_file = NULL; | |
3703 | int r; | |
3704 | ||
3705 | assert(f); | |
3706 | assert(*f); | |
3707 | ||
3708 | r = journal_file_archive(*f); | |
3709 | if (r < 0) | |
3710 | return r; | |
3711 | ||
3712 | r = journal_file_open( | |
3713 | -1, | |
3714 | (*f)->path, | |
3715 | (*f)->flags, | |
3716 | (*f)->mode, | |
3717 | compress, | |
3718 | compress_threshold_bytes, | |
3719 | seal, | |
3720 | NULL, /* metrics */ | |
3721 | (*f)->mmap, | |
3722 | deferred_closes, | |
3723 | *f, /* template */ | |
3724 | &new_file); | |
3725 | ||
3726 | journal_initiate_close(*f, deferred_closes); | |
0ac38b70 | 3727 | *f = new_file; |
7a4d21ad | 3728 | |
0ac38b70 LP |
3729 | return r; |
3730 | } | |
3731 | ||
68127658 LP |
3732 | int journal_file_dispose(int dir_fd, const char *fname) { |
3733 | _cleanup_free_ char *p = NULL; | |
3734 | _cleanup_close_ int fd = -1; | |
3735 | ||
3736 | assert(fname); | |
3737 | ||
24ee0f9d | 3738 | /* Renames a journal file to *.journal~, i.e. to mark it as corrupted or otherwise uncleanly shutdown. Note that |
68127658 LP |
3739 | * this is done without looking into the file or changing any of its contents. The idea is that this is called |
3740 | * whenever something is suspicious and we want to move the file away and make clear that it is not accessed | |
3741 | * for writing anymore. */ | |
3742 | ||
3743 | if (!endswith(fname, ".journal")) | |
3744 | return -EINVAL; | |
3745 | ||
3746 | if (asprintf(&p, "%.*s@%016" PRIx64 "-%016" PRIx64 ".journal~", | |
3747 | (int) strlen(fname) - 8, fname, | |
3748 | now(CLOCK_REALTIME), | |
3749 | random_u64()) < 0) | |
3750 | return -ENOMEM; | |
3751 | ||
3752 | if (renameat(dir_fd, fname, dir_fd, p) < 0) | |
3753 | return -errno; | |
3754 | ||
3755 | /* btrfs doesn't cope well with our write pattern and fragments heavily. Let's defrag all files we rotate */ | |
3756 | fd = openat(dir_fd, p, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
3757 | if (fd < 0) | |
3758 | log_debug_errno(errno, "Failed to open file for defragmentation/FS_NOCOW_FL, ignoring: %m"); | |
3759 | else { | |
3760 | (void) chattr_fd(fd, 0, FS_NOCOW_FL, NULL); | |
3761 | (void) btrfs_defrag_fd(fd); | |
3762 | } | |
3763 | ||
3764 | return 0; | |
3765 | } | |
3766 | ||
9447a7f1 LP |
3767 | int journal_file_open_reliably( |
3768 | const char *fname, | |
3769 | int flags, | |
3770 | mode_t mode, | |
7560fffc | 3771 | bool compress, |
57850536 | 3772 | uint64_t compress_threshold_bytes, |
baed47c3 | 3773 | bool seal, |
4a92baf3 | 3774 | JournalMetrics *metrics, |
27370278 | 3775 | MMapCache *mmap_cache, |
b58c888f | 3776 | Set *deferred_closes, |
9447a7f1 LP |
3777 | JournalFile *template, |
3778 | JournalFile **ret) { | |
3779 | ||
68127658 | 3780 | int r; |
9447a7f1 | 3781 | |
57850536 AG |
3782 | r = journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache, |
3783 | deferred_closes, template, ret); | |
288359db | 3784 | if (!IN_SET(r, |
b288cdeb ZJS |
3785 | -EBADMSG, /* Corrupted */ |
3786 | -ENODATA, /* Truncated */ | |
3787 | -EHOSTDOWN, /* Other machine */ | |
3788 | -EPROTONOSUPPORT, /* Incompatible feature */ | |
3789 | -EBUSY, /* Unclean shutdown */ | |
3790 | -ESHUTDOWN, /* Already archived */ | |
288359db | 3791 | -EIO, /* IO error, including SIGBUS on mmap */ |
ae739cc1 LP |
3792 | -EIDRM, /* File has been deleted */ |
3793 | -ETXTBSY)) /* File is from the future */ | |
9447a7f1 LP |
3794 | return r; |
3795 | ||
3796 | if ((flags & O_ACCMODE) == O_RDONLY) | |
3797 | return r; | |
3798 | ||
3799 | if (!(flags & O_CREAT)) | |
3800 | return r; | |
3801 | ||
7560fffc LP |
3802 | if (!endswith(fname, ".journal")) |
3803 | return r; | |
3804 | ||
5c70eab4 | 3805 | /* The file is corrupted. Rotate it away and try it again (but only once) */ |
65089b82 | 3806 | log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname); |
9447a7f1 | 3807 | |
68127658 LP |
3808 | r = journal_file_dispose(AT_FDCWD, fname); |
3809 | if (r < 0) | |
3810 | return r; | |
3811 | ||
57850536 AG |
3812 | return journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache, |
3813 | deferred_closes, template, ret); | |
9447a7f1 LP |
3814 | } |
3815 | ||
5a271b08 | 3816 | int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p) { |
f6a0cfa5 | 3817 | uint64_t q, n, xor_hash = 0; |
d180c349 | 3818 | const sd_id128_t *boot_id; |
f6a0cfa5 SS |
3819 | dual_timestamp ts; |
3820 | EntryItem *items; | |
3821 | int r; | |
cf244689 LP |
3822 | |
3823 | assert(from); | |
3824 | assert(to); | |
3825 | assert(o); | |
3826 | assert(p); | |
3827 | ||
3828 | if (!to->writable) | |
3829 | return -EPERM; | |
3830 | ||
d164ac77 DDM |
3831 | ts = (dual_timestamp) { |
3832 | .monotonic = le64toh(o->entry.monotonic), | |
3833 | .realtime = le64toh(o->entry.realtime), | |
3834 | }; | |
d180c349 | 3835 | boot_id = &o->entry.boot_id; |
cf244689 | 3836 | |
cf244689 | 3837 | n = journal_file_entry_n_items(o); |
5222651e | 3838 | items = newa(EntryItem, n); |
cf244689 | 3839 | |
f6a0cfa5 | 3840 | for (uint64_t i = 0; i < n; i++) { |
4fd052ae FC |
3841 | uint64_t l, h; |
3842 | le64_t le_hash; | |
cf244689 LP |
3843 | size_t t; |
3844 | void *data; | |
3845 | Object *u; | |
3846 | ||
3847 | q = le64toh(o->entry.items[i].object_offset); | |
3848 | le_hash = o->entry.items[i].hash; | |
3849 | ||
3850 | r = journal_file_move_to_object(from, OBJECT_DATA, q, &o); | |
3851 | if (r < 0) | |
3852 | return r; | |
3853 | ||
3854 | if (le_hash != o->data.hash) | |
3855 | return -EBADMSG; | |
3856 | ||
893e0f8f LP |
3857 | l = le64toh(READ_NOW(o->object.size)); |
3858 | if (l < offsetof(Object, data.payload)) | |
3859 | return -EBADMSG; | |
3860 | ||
3861 | l -= offsetof(Object, data.payload); | |
cf244689 LP |
3862 | t = (size_t) l; |
3863 | ||
3864 | /* We hit the limit on 32bit machines */ | |
3865 | if ((uint64_t) t != l) | |
3866 | return -E2BIG; | |
3867 | ||
d89c8fdf | 3868 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
d80b051c | 3869 | #if HAVE_COMPRESSION |
a7f7d1bd | 3870 | size_t rsize = 0; |
cf244689 | 3871 | |
319a4f4b LP |
3872 | r = decompress_blob( |
3873 | o->object.flags & OBJECT_COMPRESSION_MASK, | |
3874 | o->data.payload, l, | |
3875 | &from->compress_buffer, &rsize, | |
3876 | 0); | |
d89c8fdf ZJS |
3877 | if (r < 0) |
3878 | return r; | |
cf244689 LP |
3879 | |
3880 | data = from->compress_buffer; | |
3881 | l = rsize; | |
3b1a55e1 ZJS |
3882 | #else |
3883 | return -EPROTONOSUPPORT; | |
3884 | #endif | |
cf244689 LP |
3885 | } else |
3886 | data = o->data.payload; | |
3887 | ||
bc6b326d DDM |
3888 | if (l == 0) |
3889 | return -EBADMSG; | |
3890 | ||
cf244689 LP |
3891 | r = journal_file_append_data(to, data, l, &u, &h); |
3892 | if (r < 0) | |
3893 | return r; | |
3894 | ||
4ce534f4 LP |
3895 | if (JOURNAL_HEADER_KEYED_HASH(to->header)) |
3896 | xor_hash ^= jenkins_hash64(data, l); | |
3897 | else | |
3898 | xor_hash ^= le64toh(u->data.hash); | |
3899 | ||
d164ac77 DDM |
3900 | items[i] = (EntryItem) { |
3901 | .object_offset = htole64(h), | |
3902 | .hash = u->data.hash, | |
3903 | }; | |
cf244689 LP |
3904 | |
3905 | r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); | |
3906 | if (r < 0) | |
3907 | return r; | |
3908 | } | |
3909 | ||
bc6b326d | 3910 | r = journal_file_append_entry_internal(to, &ts, boot_id, xor_hash, items, n, NULL, NULL, NULL); |
fa6ac760 | 3911 | |
c3bd54bf | 3912 | if (mmap_cache_fd_got_sigbus(to->cache_fd)) |
fa6ac760 LP |
3913 | return -EIO; |
3914 | ||
3915 | return r; | |
cf244689 | 3916 | } |
babfc091 | 3917 | |
8580d1f7 LP |
3918 | void journal_reset_metrics(JournalMetrics *m) { |
3919 | assert(m); | |
3920 | ||
3921 | /* Set everything to "pick automatic values". */ | |
3922 | ||
3923 | *m = (JournalMetrics) { | |
f5fbe71d YW |
3924 | .min_use = UINT64_MAX, |
3925 | .max_use = UINT64_MAX, | |
3926 | .min_size = UINT64_MAX, | |
3927 | .max_size = UINT64_MAX, | |
3928 | .keep_free = UINT64_MAX, | |
3929 | .n_max_files = UINT64_MAX, | |
8580d1f7 LP |
3930 | }; |
3931 | } | |
3932 | ||
babfc091 | 3933 | void journal_default_metrics(JournalMetrics *m, int fd) { |
babfc091 | 3934 | struct statvfs ss; |
6aae0b1a | 3935 | uint64_t fs_size = 0; |
babfc091 LP |
3936 | |
3937 | assert(m); | |
3938 | assert(fd >= 0); | |
3939 | ||
3940 | if (fstatvfs(fd, &ss) >= 0) | |
3941 | fs_size = ss.f_frsize * ss.f_blocks; | |
6aae0b1a | 3942 | else |
8fc58f1a | 3943 | log_debug_errno(errno, "Failed to determine disk size: %m"); |
babfc091 | 3944 | |
f5fbe71d | 3945 | if (m->max_use == UINT64_MAX) { |
babfc091 | 3946 | |
6aae0b1a ZJS |
3947 | if (fs_size > 0) |
3948 | m->max_use = CLAMP(PAGE_ALIGN(fs_size / 10), /* 10% of file system size */ | |
3949 | MAX_USE_LOWER, MAX_USE_UPPER); | |
3950 | else | |
3951 | m->max_use = MAX_USE_LOWER; | |
babfc091 LP |
3952 | } else { |
3953 | m->max_use = PAGE_ALIGN(m->max_use); | |
3954 | ||
8580d1f7 | 3955 | if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2) |
babfc091 LP |
3956 | m->max_use = JOURNAL_FILE_SIZE_MIN*2; |
3957 | } | |
3958 | ||
f5fbe71d | 3959 | if (m->min_use == UINT64_MAX) { |
6aae0b1a ZJS |
3960 | if (fs_size > 0) |
3961 | m->min_use = CLAMP(PAGE_ALIGN(fs_size / 50), /* 2% of file system size */ | |
3962 | MIN_USE_LOW, MIN_USE_HIGH); | |
3963 | else | |
3964 | m->min_use = MIN_USE_LOW; | |
3965 | } | |
8580d1f7 LP |
3966 | |
3967 | if (m->min_use > m->max_use) | |
3968 | m->min_use = m->max_use; | |
3969 | ||
f5fbe71d | 3970 | if (m->max_size == UINT64_MAX) |
6aae0b1a ZJS |
3971 | m->max_size = MIN(PAGE_ALIGN(m->max_use / 8), /* 8 chunks */ |
3972 | MAX_SIZE_UPPER); | |
3973 | else | |
babfc091 LP |
3974 | m->max_size = PAGE_ALIGN(m->max_size); |
3975 | ||
8580d1f7 LP |
3976 | if (m->max_size != 0) { |
3977 | if (m->max_size < JOURNAL_FILE_SIZE_MIN) | |
3978 | m->max_size = JOURNAL_FILE_SIZE_MIN; | |
babfc091 | 3979 | |
8580d1f7 LP |
3980 | if (m->max_use != 0 && m->max_size*2 > m->max_use) |
3981 | m->max_use = m->max_size*2; | |
3982 | } | |
babfc091 | 3983 | |
f5fbe71d | 3984 | if (m->min_size == UINT64_MAX) |
babfc091 | 3985 | m->min_size = JOURNAL_FILE_SIZE_MIN; |
6aae0b1a ZJS |
3986 | else |
3987 | m->min_size = CLAMP(PAGE_ALIGN(m->min_size), | |
3988 | JOURNAL_FILE_SIZE_MIN, | |
3989 | m->max_size ?: UINT64_MAX); | |
babfc091 | 3990 | |
f5fbe71d | 3991 | if (m->keep_free == UINT64_MAX) { |
6aae0b1a ZJS |
3992 | if (fs_size > 0) |
3993 | m->keep_free = MIN(PAGE_ALIGN(fs_size / 20), /* 5% of file system size */ | |
3994 | KEEP_FREE_UPPER); | |
3995 | else | |
babfc091 LP |
3996 | m->keep_free = DEFAULT_KEEP_FREE; |
3997 | } | |
3998 | ||
f5fbe71d | 3999 | if (m->n_max_files == UINT64_MAX) |
8580d1f7 LP |
4000 | m->n_max_files = DEFAULT_N_MAX_FILES; |
4001 | ||
4002 | log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64, | |
2b59bf51 ZJS |
4003 | FORMAT_BYTES(m->min_use), |
4004 | FORMAT_BYTES(m->max_use), | |
4005 | FORMAT_BYTES(m->max_size), | |
4006 | FORMAT_BYTES(m->min_size), | |
4007 | FORMAT_BYTES(m->keep_free), | |
8580d1f7 | 4008 | m->n_max_files); |
babfc091 | 4009 | } |
08984293 LP |
4010 | |
4011 | int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { | |
08984293 | 4012 | assert(f); |
c88cc6af | 4013 | assert(f->header); |
08984293 LP |
4014 | assert(from || to); |
4015 | ||
4016 | if (from) { | |
162566a4 LP |
4017 | if (f->header->head_entry_realtime == 0) |
4018 | return -ENOENT; | |
08984293 | 4019 | |
162566a4 | 4020 | *from = le64toh(f->header->head_entry_realtime); |
08984293 LP |
4021 | } |
4022 | ||
4023 | if (to) { | |
162566a4 LP |
4024 | if (f->header->tail_entry_realtime == 0) |
4025 | return -ENOENT; | |
08984293 | 4026 | |
162566a4 | 4027 | *to = le64toh(f->header->tail_entry_realtime); |
08984293 LP |
4028 | } |
4029 | ||
4030 | return 1; | |
4031 | } | |
4032 | ||
4033 | int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) { | |
08984293 LP |
4034 | Object *o; |
4035 | uint64_t p; | |
4036 | int r; | |
4037 | ||
4038 | assert(f); | |
4039 | assert(from || to); | |
4040 | ||
47838ab3 | 4041 | r = find_data_object_by_boot_id(f, boot_id, &o, &p); |
08984293 LP |
4042 | if (r <= 0) |
4043 | return r; | |
4044 | ||
4045 | if (le64toh(o->data.n_entries) <= 0) | |
4046 | return 0; | |
4047 | ||
4048 | if (from) { | |
4049 | r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); | |
4050 | if (r < 0) | |
4051 | return r; | |
4052 | ||
4053 | *from = le64toh(o->entry.monotonic); | |
4054 | } | |
4055 | ||
4056 | if (to) { | |
4057 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
4058 | if (r < 0) | |
4059 | return r; | |
4060 | ||
4061 | r = generic_array_get_plus_one(f, | |
4062 | le64toh(o->data.entry_offset), | |
4063 | le64toh(o->data.entry_array_offset), | |
4064 | le64toh(o->data.n_entries)-1, | |
4065 | &o, NULL); | |
4066 | if (r <= 0) | |
4067 | return r; | |
4068 | ||
4069 | *to = le64toh(o->entry.monotonic); | |
4070 | } | |
4071 | ||
4072 | return 1; | |
4073 | } | |
dca6219e | 4074 | |
c8e6e1f1 | 4075 | bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec, int log_level) { |
dca6219e | 4076 | assert(f); |
c88cc6af | 4077 | assert(f->header); |
dca6219e LP |
4078 | |
4079 | /* If we gained new header fields we gained new features, | |
4080 | * hence suggest a rotation */ | |
361f9cbc | 4081 | if (le64toh(f->header->header_size) < sizeof(Header)) { |
c8e6e1f1 | 4082 | log_full(log_level, "%s uses an outdated header, suggesting rotation.", f->path); |
dca6219e | 4083 | return true; |
361f9cbc | 4084 | } |
dca6219e | 4085 | |
0dbe57ee LP |
4086 | /* Let's check if the hash tables grew over a certain fill level (75%, borrowing this value from |
4087 | * Java's hash table implementation), and if so suggest a rotation. To calculate the fill level we | |
4088 | * need the n_data field, which only exists in newer versions. */ | |
dca6219e LP |
4089 | |
4090 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) | |
361f9cbc | 4091 | if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
c8e6e1f1 DDM |
4092 | log_full(log_level, |
4093 | "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.", | |
4094 | f->path, | |
4095 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))), | |
4096 | le64toh(f->header->n_data), | |
4097 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
4098 | (unsigned long long) f->last_stat.st_size, | |
4099 | f->last_stat.st_size / le64toh(f->header->n_data)); | |
dca6219e | 4100 | return true; |
361f9cbc | 4101 | } |
dca6219e LP |
4102 | |
4103 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
361f9cbc | 4104 | if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
c8e6e1f1 DDM |
4105 | log_full(log_level, |
4106 | "Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.", | |
4107 | f->path, | |
4108 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), | |
4109 | le64toh(f->header->n_fields), | |
4110 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem)); | |
dca6219e | 4111 | return true; |
361f9cbc | 4112 | } |
dca6219e | 4113 | |
0dbe57ee LP |
4114 | /* If there are too many hash collisions somebody is most likely playing games with us. Hence, if our |
4115 | * longest chain is longer than some threshold, let's suggest rotation. */ | |
4116 | if (JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth) && | |
4117 | le64toh(f->header->data_hash_chain_depth) > HASH_CHAIN_DEPTH_MAX) { | |
c8e6e1f1 DDM |
4118 | log_full(log_level, |
4119 | "Data hash table of %s has deepest hash chain of length %" PRIu64 ", suggesting rotation.", | |
4120 | f->path, le64toh(f->header->data_hash_chain_depth)); | |
0dbe57ee LP |
4121 | return true; |
4122 | } | |
4123 | ||
4124 | if (JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth) && | |
4125 | le64toh(f->header->field_hash_chain_depth) > HASH_CHAIN_DEPTH_MAX) { | |
c8e6e1f1 DDM |
4126 | log_full(log_level, |
4127 | "Field hash table of %s has deepest hash chain of length at %" PRIu64 ", suggesting rotation.", | |
4128 | f->path, le64toh(f->header->field_hash_chain_depth)); | |
0dbe57ee LP |
4129 | return true; |
4130 | } | |
4131 | ||
0598fd4a LP |
4132 | /* Are the data objects properly indexed by field objects? */ |
4133 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && | |
4134 | JOURNAL_HEADER_CONTAINS(f->header, n_fields) && | |
4135 | le64toh(f->header->n_data) > 0 && | |
012181ea DDM |
4136 | le64toh(f->header->n_fields) == 0) { |
4137 | log_full(log_level, | |
4138 | "Data objects of %s are not indexed by field objects, suggesting rotation.", | |
4139 | f->path); | |
0598fd4a | 4140 | return true; |
012181ea | 4141 | } |
0598fd4a | 4142 | |
fb0951b0 LP |
4143 | if (max_file_usec > 0) { |
4144 | usec_t t, h; | |
4145 | ||
4146 | h = le64toh(f->header->head_entry_realtime); | |
4147 | t = now(CLOCK_REALTIME); | |
4148 | ||
012181ea DDM |
4149 | if (h > 0 && t > h + max_file_usec) { |
4150 | log_full(log_level, | |
4151 | "Oldest entry in %s is older than the configured file retention duration (%s), suggesting rotation.", | |
4152 | f->path, FORMAT_TIMESPAN(max_file_usec, USEC_PER_SEC)); | |
fb0951b0 | 4153 | return true; |
012181ea | 4154 | } |
fb0951b0 LP |
4155 | } |
4156 | ||
dca6219e LP |
4157 | return false; |
4158 | } | |
363b2b9a DDM |
4159 | |
4160 | static const char * const journal_object_type_table[] = { | |
4161 | [OBJECT_UNUSED] = "unused", | |
4162 | [OBJECT_DATA] = "data", | |
4163 | [OBJECT_FIELD] = "field", | |
4164 | [OBJECT_ENTRY] = "entry", | |
4165 | [OBJECT_DATA_HASH_TABLE] = "data hash table", | |
4166 | [OBJECT_FIELD_HASH_TABLE] = "field hash table", | |
4167 | [OBJECT_ENTRY_ARRAY] = "entry array", | |
4168 | [OBJECT_TAG] = "tag", | |
4169 | }; | |
4170 | ||
4171 | DEFINE_STRING_TABLE_LOOKUP_TO_STRING(journal_object_type, ObjectType); |