]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
cec736d2 | 2 | |
cec736d2 | 3 | #include <errno.h> |
cec736d2 | 4 | #include <fcntl.h> |
11689d2a | 5 | #include <linux/fs.h> |
ac2e41f5 | 6 | #include <pthread.h> |
07630cea LP |
7 | #include <stddef.h> |
8 | #include <sys/mman.h> | |
9 | #include <sys/statvfs.h> | |
10 | #include <sys/uio.h> | |
11 | #include <unistd.h> | |
fb0951b0 | 12 | |
a03d4359 ZJS |
13 | #include "sd-event.h" |
14 | ||
b5efdb8a | 15 | #include "alloc-util.h" |
f27a3864 | 16 | #include "btrfs-util.h" |
c8b3094d | 17 | #include "chattr-util.h" |
07630cea | 18 | #include "compress.h" |
3ffd4af2 | 19 | #include "fd-util.h" |
11b29a96 | 20 | #include "fs-util.h" |
0284adc6 | 21 | #include "journal-authenticate.h" |
cec736d2 LP |
22 | #include "journal-def.h" |
23 | #include "journal-file.h" | |
24 | #include "lookup3.h" | |
6bedfcbb | 25 | #include "parse-util.h" |
5d1ce257 | 26 | #include "path-util.h" |
3df3e884 | 27 | #include "random-util.h" |
b58c888f | 28 | #include "set.h" |
3cc44114 | 29 | #include "stat-util.h" |
07630cea | 30 | #include "string-util.h" |
4761fd0f | 31 | #include "strv.h" |
89a5a90c | 32 | #include "xattr-util.h" |
cec736d2 | 33 | |
4a92baf3 LP |
34 | #define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) |
35 | #define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) | |
cec736d2 | 36 | |
57850536 AG |
37 | #define DEFAULT_COMPRESS_THRESHOLD (512ULL) |
38 | #define MIN_COMPRESS_THRESHOLD (8ULL) | |
807e17f0 | 39 | |
babfc091 | 40 | /* This is the minimum journal file size */ |
16098e93 | 41 | #define JOURNAL_FILE_SIZE_MIN (512ULL*1024ULL) /* 512 KiB */ |
babfc091 LP |
42 | |
43 | /* These are the lower and upper bounds if we deduce the max_use value | |
44 | * from the file system size */ | |
45 | #define DEFAULT_MAX_USE_LOWER (1ULL*1024ULL*1024ULL) /* 1 MiB */ | |
46 | #define DEFAULT_MAX_USE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ | |
47 | ||
8580d1f7 LP |
48 | /* This is the default minimal use limit, how much we'll use even if keep_free suggests otherwise. */ |
49 | #define DEFAULT_MIN_USE (1ULL*1024ULL*1024ULL) /* 1 MiB */ | |
50 | ||
babfc091 | 51 | /* This is the upper bound if we deduce max_size from max_use */ |
71100051 | 52 | #define DEFAULT_MAX_SIZE_UPPER (128ULL*1024ULL*1024ULL) /* 128 MiB */ |
babfc091 LP |
53 | |
54 | /* This is the upper bound if we deduce the keep_free value from the | |
55 | * file system size */ | |
56 | #define DEFAULT_KEEP_FREE_UPPER (4ULL*1024ULL*1024ULL*1024ULL) /* 4 GiB */ | |
57 | ||
58 | /* This is the keep_free value when we can't determine the system | |
59 | * size */ | |
60 | #define DEFAULT_KEEP_FREE (1024ULL*1024ULL) /* 1 MB */ | |
61 | ||
8580d1f7 LP |
62 | /* This is the default maximum number of journal files to keep around. */ |
63 | #define DEFAULT_N_MAX_FILES (100) | |
64 | ||
dca6219e LP |
65 | /* n_data was the first entry we added after the initial file format design */ |
66 | #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) | |
cec736d2 | 67 | |
a4bcff5b LP |
68 | /* How many entries to keep in the entry array chain cache at max */ |
69 | #define CHAIN_CACHE_MAX 20 | |
70 | ||
a676e665 LP |
71 | /* How much to increase the journal file size at once each time we allocate something new. */ |
72 | #define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL) /* 8MB */ | |
73 | ||
2678031a LP |
74 | /* Reread fstat() of the file for detecting deletions at least this often */ |
75 | #define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC) | |
76 | ||
fa6ac760 LP |
77 | /* The mmap context to use for the header we pick as one above the last defined typed */ |
78 | #define CONTEXT_HEADER _OBJECT_TYPE_MAX | |
79 | ||
51804460 ZJS |
80 | #ifdef __clang__ |
81 | # pragma GCC diagnostic ignored "-Waddress-of-packed-member" | |
82 | #endif | |
83 | ||
ac2e41f5 VC |
84 | /* This may be called from a separate thread to prevent blocking the caller for the duration of fsync(). |
85 | * As a result we use atomic operations on f->offline_state for inter-thread communications with | |
86 | * journal_file_set_offline() and journal_file_set_online(). */ | |
87 | static void journal_file_set_offline_internal(JournalFile *f) { | |
26687bf8 | 88 | assert(f); |
ac2e41f5 VC |
89 | assert(f->fd >= 0); |
90 | assert(f->header); | |
91 | ||
92 | for (;;) { | |
93 | switch (f->offline_state) { | |
94 | case OFFLINE_CANCEL: | |
95 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_DONE)) | |
96 | continue; | |
97 | return; | |
98 | ||
99 | case OFFLINE_AGAIN_FROM_SYNCING: | |
100 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_SYNCING)) | |
101 | continue; | |
102 | break; | |
103 | ||
104 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
105 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_SYNCING)) | |
106 | continue; | |
107 | break; | |
108 | ||
109 | case OFFLINE_SYNCING: | |
110 | (void) fsync(f->fd); | |
26687bf8 | 111 | |
ac2e41f5 VC |
112 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_OFFLINING)) |
113 | continue; | |
26687bf8 | 114 | |
8eb85171 | 115 | f->header->state = f->archive ? STATE_ARCHIVED : STATE_OFFLINE; |
ac2e41f5 VC |
116 | (void) fsync(f->fd); |
117 | break; | |
118 | ||
119 | case OFFLINE_OFFLINING: | |
120 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_DONE)) | |
121 | continue; | |
4831981d | 122 | _fallthrough_; |
ac2e41f5 VC |
123 | case OFFLINE_DONE: |
124 | return; | |
125 | ||
126 | case OFFLINE_JOINED: | |
127 | log_debug("OFFLINE_JOINED unexpected offline state for journal_file_set_offline_internal()"); | |
128 | return; | |
129 | } | |
130 | } | |
131 | } | |
132 | ||
133 | static void * journal_file_set_offline_thread(void *arg) { | |
134 | JournalFile *f = arg; | |
135 | ||
fa7ff4cf LP |
136 | (void) pthread_setname_np(pthread_self(), "journal-offline"); |
137 | ||
ac2e41f5 VC |
138 | journal_file_set_offline_internal(f); |
139 | ||
140 | return NULL; | |
141 | } | |
142 | ||
143 | static int journal_file_set_offline_thread_join(JournalFile *f) { | |
144 | int r; | |
145 | ||
146 | assert(f); | |
147 | ||
148 | if (f->offline_state == OFFLINE_JOINED) | |
149 | return 0; | |
150 | ||
151 | r = pthread_join(f->offline_thread, NULL); | |
152 | if (r) | |
153 | return -r; | |
154 | ||
155 | f->offline_state = OFFLINE_JOINED; | |
26687bf8 | 156 | |
be7cdd8e | 157 | if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) |
fa6ac760 LP |
158 | return -EIO; |
159 | ||
ac2e41f5 VC |
160 | return 0; |
161 | } | |
26687bf8 | 162 | |
ac2e41f5 VC |
163 | /* Trigger a restart if the offline thread is mid-flight in a restartable state. */ |
164 | static bool journal_file_set_offline_try_restart(JournalFile *f) { | |
165 | for (;;) { | |
166 | switch (f->offline_state) { | |
167 | case OFFLINE_AGAIN_FROM_SYNCING: | |
168 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
169 | return true; | |
170 | ||
171 | case OFFLINE_CANCEL: | |
172 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_CANCEL, OFFLINE_AGAIN_FROM_SYNCING)) | |
173 | continue; | |
174 | return true; | |
175 | ||
176 | case OFFLINE_SYNCING: | |
177 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_AGAIN_FROM_SYNCING)) | |
178 | continue; | |
179 | return true; | |
180 | ||
181 | case OFFLINE_OFFLINING: | |
182 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_OFFLINING, OFFLINE_AGAIN_FROM_OFFLINING)) | |
183 | continue; | |
184 | return true; | |
26687bf8 OS |
185 | |
186 | default: | |
ac2e41f5 VC |
187 | return false; |
188 | } | |
26687bf8 OS |
189 | } |
190 | } | |
191 | ||
ac2e41f5 VC |
192 | /* Sets a journal offline. |
193 | * | |
194 | * If wait is false then an offline is dispatched in a separate thread for a | |
195 | * subsequent journal_file_set_offline() or journal_file_set_online() of the | |
196 | * same journal to synchronize with. | |
197 | * | |
198 | * If wait is true, then either an existing offline thread will be restarted | |
199 | * and joined, or if none exists the offline is simply performed in this | |
200 | * context without involving another thread. | |
201 | */ | |
202 | int journal_file_set_offline(JournalFile *f, bool wait) { | |
203 | bool restarted; | |
204 | int r; | |
205 | ||
26687bf8 OS |
206 | assert(f); |
207 | ||
208 | if (!f->writable) | |
209 | return -EPERM; | |
210 | ||
846e5418 | 211 | if (f->fd < 0 || !f->header) |
26687bf8 OS |
212 | return -EINVAL; |
213 | ||
b8f99e27 VC |
214 | /* An offlining journal is implicitly online and may modify f->header->state, |
215 | * we must also join any potentially lingering offline thread when not online. */ | |
216 | if (!journal_file_is_offlining(f) && f->header->state != STATE_ONLINE) | |
217 | return journal_file_set_offline_thread_join(f); | |
26687bf8 | 218 | |
ac2e41f5 VC |
219 | /* Restart an in-flight offline thread and wait if needed, or join a lingering done one. */ |
220 | restarted = journal_file_set_offline_try_restart(f); | |
221 | if ((restarted && wait) || !restarted) { | |
222 | r = journal_file_set_offline_thread_join(f); | |
223 | if (r < 0) | |
224 | return r; | |
225 | } | |
26687bf8 | 226 | |
ac2e41f5 VC |
227 | if (restarted) |
228 | return 0; | |
229 | ||
230 | /* Initiate a new offline. */ | |
231 | f->offline_state = OFFLINE_SYNCING; | |
fa6ac760 | 232 | |
ac2e41f5 VC |
233 | if (wait) /* Without using a thread if waiting. */ |
234 | journal_file_set_offline_internal(f); | |
235 | else { | |
5e9f01e8 LP |
236 | sigset_t ss, saved_ss; |
237 | int k; | |
238 | ||
cd2a429e | 239 | assert_se(sigfillset(&ss) >= 0); |
5e9f01e8 LP |
240 | |
241 | r = pthread_sigmask(SIG_BLOCK, &ss, &saved_ss); | |
242 | if (r > 0) | |
243 | return -r; | |
244 | ||
ac2e41f5 | 245 | r = pthread_create(&f->offline_thread, NULL, journal_file_set_offline_thread, f); |
5e9f01e8 LP |
246 | |
247 | k = pthread_sigmask(SIG_SETMASK, &saved_ss, NULL); | |
ec9ffa2c VC |
248 | if (r > 0) { |
249 | f->offline_state = OFFLINE_JOINED; | |
ac2e41f5 | 250 | return -r; |
ec9ffa2c | 251 | } |
5e9f01e8 LP |
252 | if (k > 0) |
253 | return -k; | |
ac2e41f5 VC |
254 | } |
255 | ||
256 | return 0; | |
257 | } | |
258 | ||
259 | static int journal_file_set_online(JournalFile *f) { | |
83bf6b67 | 260 | bool wait = true; |
ac2e41f5 VC |
261 | |
262 | assert(f); | |
263 | ||
264 | if (!f->writable) | |
265 | return -EPERM; | |
266 | ||
846e5418 | 267 | if (f->fd < 0 || !f->header) |
ac2e41f5 VC |
268 | return -EINVAL; |
269 | ||
83bf6b67 | 270 | while (wait) { |
ac2e41f5 VC |
271 | switch (f->offline_state) { |
272 | case OFFLINE_JOINED: | |
273 | /* No offline thread, no need to wait. */ | |
83bf6b67 | 274 | wait = false; |
ac2e41f5 VC |
275 | break; |
276 | ||
277 | case OFFLINE_SYNCING: | |
278 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_SYNCING, OFFLINE_CANCEL)) | |
279 | continue; | |
280 | /* Canceled syncing prior to offlining, no need to wait. */ | |
83bf6b67 | 281 | wait = false; |
ac2e41f5 VC |
282 | break; |
283 | ||
284 | case OFFLINE_AGAIN_FROM_SYNCING: | |
285 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_SYNCING, OFFLINE_CANCEL)) | |
286 | continue; | |
287 | /* Canceled restart from syncing, no need to wait. */ | |
83bf6b67 | 288 | wait = false; |
ac2e41f5 VC |
289 | break; |
290 | ||
291 | case OFFLINE_AGAIN_FROM_OFFLINING: | |
292 | if (!__sync_bool_compare_and_swap(&f->offline_state, OFFLINE_AGAIN_FROM_OFFLINING, OFFLINE_CANCEL)) | |
293 | continue; | |
294 | /* Canceled restart from offlining, must wait for offlining to complete however. */ | |
4831981d | 295 | _fallthrough_; |
ac2e41f5 VC |
296 | default: { |
297 | int r; | |
298 | ||
299 | r = journal_file_set_offline_thread_join(f); | |
300 | if (r < 0) | |
301 | return r; | |
302 | ||
83bf6b67 | 303 | wait = false; |
ac2e41f5 VC |
304 | break; |
305 | } | |
306 | } | |
307 | } | |
26687bf8 | 308 | |
be7cdd8e | 309 | if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) |
fa6ac760 LP |
310 | return -EIO; |
311 | ||
ac2e41f5 VC |
312 | switch (f->header->state) { |
313 | case STATE_ONLINE: | |
314 | return 0; | |
26687bf8 | 315 | |
ac2e41f5 VC |
316 | case STATE_OFFLINE: |
317 | f->header->state = STATE_ONLINE; | |
318 | (void) fsync(f->fd); | |
319 | return 0; | |
320 | ||
321 | default: | |
322 | return -EINVAL; | |
323 | } | |
26687bf8 OS |
324 | } |
325 | ||
b58c888f VC |
326 | bool journal_file_is_offlining(JournalFile *f) { |
327 | assert(f); | |
328 | ||
329 | __sync_synchronize(); | |
330 | ||
3742095b | 331 | if (IN_SET(f->offline_state, OFFLINE_DONE, OFFLINE_JOINED)) |
b58c888f VC |
332 | return false; |
333 | ||
334 | return true; | |
335 | } | |
336 | ||
804ae586 | 337 | JournalFile* journal_file_close(JournalFile *f) { |
de190aef | 338 | assert(f); |
cec736d2 | 339 | |
349cc4a5 | 340 | #if HAVE_GCRYPT |
b0af6f41 | 341 | /* Write the final tag */ |
43cd8794 FB |
342 | if (f->seal && f->writable) { |
343 | int r; | |
344 | ||
345 | r = journal_file_append_tag(f); | |
346 | if (r < 0) | |
347 | log_error_errno(r, "Failed to append tag when closing journal: %m"); | |
348 | } | |
feb12d3e | 349 | #endif |
b0af6f41 | 350 | |
7a24f3bf | 351 | if (f->post_change_timer) { |
b6cdfbe5 ZJS |
352 | if (sd_event_source_get_enabled(f->post_change_timer, NULL) > 0) |
353 | journal_file_post_change(f); | |
7a24f3bf | 354 | |
e167d7fd | 355 | (void) sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_OFF); |
7a24f3bf VC |
356 | sd_event_source_unref(f->post_change_timer); |
357 | } | |
358 | ||
ac2e41f5 | 359 | journal_file_set_offline(f, true); |
cec736d2 | 360 | |
be7cdd8e VC |
361 | if (f->mmap && f->cache_fd) |
362 | mmap_cache_free_fd(f->mmap, f->cache_fd); | |
cec736d2 | 363 | |
11689d2a LP |
364 | if (f->fd >= 0 && f->defrag_on_close) { |
365 | ||
366 | /* Be friendly to btrfs: turn COW back on again now, | |
367 | * and defragment the file. We won't write to the file | |
368 | * ever again, hence remove all fragmentation, and | |
369 | * reenable all the good bits COW usually provides | |
370 | * (such as data checksumming). */ | |
371 | ||
db9a4254 | 372 | (void) chattr_fd(f->fd, 0, FS_NOCOW_FL, NULL); |
11689d2a LP |
373 | (void) btrfs_defrag_fd(f->fd); |
374 | } | |
f27a3864 | 375 | |
5d1ce257 LP |
376 | if (f->close_fd) |
377 | safe_close(f->fd); | |
cec736d2 | 378 | free(f->path); |
807e17f0 | 379 | |
f649045c | 380 | mmap_cache_unref(f->mmap); |
16e9f408 | 381 | |
4743015d | 382 | ordered_hashmap_free_free(f->chain_cache); |
a4bcff5b | 383 | |
349cc4a5 | 384 | #if HAVE_XZ || HAVE_LZ4 |
807e17f0 LP |
385 | free(f->compress_buffer); |
386 | #endif | |
387 | ||
349cc4a5 | 388 | #if HAVE_GCRYPT |
baed47c3 LP |
389 | if (f->fss_file) |
390 | munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size)); | |
dc4ebc07 | 391 | else |
b7c9ae91 LP |
392 | free(f->fsprg_state); |
393 | ||
394 | free(f->fsprg_seed); | |
7560fffc LP |
395 | |
396 | if (f->hmac) | |
397 | gcry_md_close(f->hmac); | |
398 | #endif | |
399 | ||
6b430fdb | 400 | return mfree(f); |
cec736d2 LP |
401 | } |
402 | ||
0ac38b70 | 403 | static int journal_file_init_header(JournalFile *f, JournalFile *template) { |
d89c8fdf | 404 | Header h = {}; |
cec736d2 LP |
405 | ssize_t k; |
406 | int r; | |
407 | ||
408 | assert(f); | |
409 | ||
7560fffc | 410 | memcpy(h.signature, HEADER_SIGNATURE, 8); |
23b0b2b2 | 411 | h.header_size = htole64(ALIGN64(sizeof(h))); |
cec736d2 | 412 | |
d89c8fdf ZJS |
413 | h.incompatible_flags |= htole32( |
414 | f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ | | |
415 | f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4); | |
7560fffc | 416 | |
d89c8fdf ZJS |
417 | h.compatible_flags = htole32( |
418 | f->seal * HEADER_COMPATIBLE_SEALED); | |
7560fffc | 419 | |
cec736d2 LP |
420 | r = sd_id128_randomize(&h.file_id); |
421 | if (r < 0) | |
422 | return r; | |
423 | ||
0ac38b70 LP |
424 | if (template) { |
425 | h.seqnum_id = template->header->seqnum_id; | |
beec0085 | 426 | h.tail_entry_seqnum = template->header->tail_entry_seqnum; |
0ac38b70 LP |
427 | } else |
428 | h.seqnum_id = h.file_id; | |
cec736d2 LP |
429 | |
430 | k = pwrite(f->fd, &h, sizeof(h), 0); | |
431 | if (k < 0) | |
432 | return -errno; | |
433 | ||
434 | if (k != sizeof(h)) | |
435 | return -EIO; | |
436 | ||
437 | return 0; | |
438 | } | |
439 | ||
440 | static int journal_file_refresh_header(JournalFile *f) { | |
de190aef | 441 | sd_id128_t boot_id; |
fa6ac760 | 442 | int r; |
cec736d2 LP |
443 | |
444 | assert(f); | |
c88cc6af | 445 | assert(f->header); |
cec736d2 LP |
446 | |
447 | r = sd_id128_get_machine(&f->header->machine_id); | |
fd4885df ZJS |
448 | if (IN_SET(r, -ENOENT, -ENOMEDIUM)) |
449 | /* We don't have a machine-id, let's continue without */ | |
450 | zero(f->header->machine_id); | |
451 | else if (r < 0) | |
cec736d2 LP |
452 | return r; |
453 | ||
de190aef | 454 | r = sd_id128_get_boot(&boot_id); |
cec736d2 LP |
455 | if (r < 0) |
456 | return r; | |
457 | ||
de190aef LP |
458 | f->header->boot_id = boot_id; |
459 | ||
fa6ac760 | 460 | r = journal_file_set_online(f); |
b788cc23 | 461 | |
7560fffc | 462 | /* Sync the online state to disk */ |
fb426037 | 463 | (void) fsync(f->fd); |
b788cc23 | 464 | |
a0fe2a2d LP |
465 | /* We likely just created a new file, also sync the directory this file is located in. */ |
466 | (void) fsync_directory_of_file(f->fd); | |
467 | ||
fa6ac760 | 468 | return r; |
cec736d2 LP |
469 | } |
470 | ||
4214009f ZJS |
471 | static bool warn_wrong_flags(const JournalFile *f, bool compatible) { |
472 | const uint32_t any = compatible ? HEADER_COMPATIBLE_ANY : HEADER_INCOMPATIBLE_ANY, | |
473 | supported = compatible ? HEADER_COMPATIBLE_SUPPORTED : HEADER_INCOMPATIBLE_SUPPORTED; | |
474 | const char *type = compatible ? "compatible" : "incompatible"; | |
d89c8fdf ZJS |
475 | uint32_t flags; |
476 | ||
4214009f ZJS |
477 | flags = le32toh(compatible ? f->header->compatible_flags : f->header->incompatible_flags); |
478 | ||
479 | if (flags & ~supported) { | |
480 | if (flags & ~any) | |
4761fd0f | 481 | log_debug("Journal file %s has unknown %s flags 0x%"PRIx32, |
4214009f ZJS |
482 | f->path, type, flags & ~any); |
483 | flags = (flags & any) & ~supported; | |
4761fd0f ZJS |
484 | if (flags) { |
485 | const char* strv[3]; | |
486 | unsigned n = 0; | |
487 | _cleanup_free_ char *t = NULL; | |
488 | ||
489 | if (compatible && (flags & HEADER_COMPATIBLE_SEALED)) | |
490 | strv[n++] = "sealed"; | |
491 | if (!compatible && (flags & HEADER_INCOMPATIBLE_COMPRESSED_XZ)) | |
492 | strv[n++] = "xz-compressed"; | |
493 | if (!compatible && (flags & HEADER_INCOMPATIBLE_COMPRESSED_LZ4)) | |
494 | strv[n++] = "lz4-compressed"; | |
495 | strv[n] = NULL; | |
496 | assert(n < ELEMENTSOF(strv)); | |
497 | ||
498 | t = strv_join((char**) strv, ", "); | |
499 | log_debug("Journal file %s uses %s %s %s disabled at compilation time.", | |
500 | f->path, type, n > 1 ? "flags" : "flag", strnull(t)); | |
501 | } | |
4214009f ZJS |
502 | return true; |
503 | } | |
504 | ||
505 | return false; | |
506 | } | |
507 | ||
508 | static int journal_file_verify_header(JournalFile *f) { | |
6f94e420 TS |
509 | uint64_t arena_size, header_size; |
510 | ||
cec736d2 | 511 | assert(f); |
c88cc6af | 512 | assert(f->header); |
cec736d2 | 513 | |
7560fffc | 514 | if (memcmp(f->header->signature, HEADER_SIGNATURE, 8)) |
cec736d2 LP |
515 | return -EBADMSG; |
516 | ||
4214009f ZJS |
517 | /* In both read and write mode we refuse to open files with incompatible |
518 | * flags we don't know. */ | |
519 | if (warn_wrong_flags(f, false)) | |
cec736d2 LP |
520 | return -EPROTONOSUPPORT; |
521 | ||
4214009f ZJS |
522 | /* When open for writing we refuse to open files with compatible flags, too. */ |
523 | if (f->writable && warn_wrong_flags(f, true)) | |
d89c8fdf | 524 | return -EPROTONOSUPPORT; |
7560fffc | 525 | |
db11ac1a LP |
526 | if (f->header->state >= _STATE_MAX) |
527 | return -EBADMSG; | |
528 | ||
6f94e420 TS |
529 | header_size = le64toh(f->header->header_size); |
530 | ||
dca6219e | 531 | /* The first addition was n_data, so check that we are at least this large */ |
6f94e420 | 532 | if (header_size < HEADER_SIZE_MIN) |
23b0b2b2 LP |
533 | return -EBADMSG; |
534 | ||
8088cbd3 | 535 | if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
beec0085 LP |
536 | return -EBADMSG; |
537 | ||
6f94e420 TS |
538 | arena_size = le64toh(f->header->arena_size); |
539 | ||
540 | if (UINT64_MAX - header_size < arena_size || header_size + arena_size > (uint64_t) f->last_stat.st_size) | |
db11ac1a LP |
541 | return -ENODATA; |
542 | ||
6f94e420 | 543 | if (le64toh(f->header->tail_object_offset) > header_size + arena_size) |
db11ac1a LP |
544 | return -ENODATA; |
545 | ||
7762e02b LP |
546 | if (!VALID64(le64toh(f->header->data_hash_table_offset)) || |
547 | !VALID64(le64toh(f->header->field_hash_table_offset)) || | |
548 | !VALID64(le64toh(f->header->tail_object_offset)) || | |
549 | !VALID64(le64toh(f->header->entry_array_offset))) | |
550 | return -ENODATA; | |
551 | ||
cec736d2 | 552 | if (f->writable) { |
cec736d2 | 553 | sd_id128_t machine_id; |
ae739cc1 | 554 | uint8_t state; |
cec736d2 LP |
555 | int r; |
556 | ||
557 | r = sd_id128_get_machine(&machine_id); | |
558 | if (r < 0) | |
559 | return r; | |
560 | ||
561 | if (!sd_id128_equal(machine_id, f->header->machine_id)) | |
562 | return -EHOSTDOWN; | |
563 | ||
de190aef | 564 | state = f->header->state; |
cec736d2 | 565 | |
b288cdeb ZJS |
566 | if (state == STATE_ARCHIVED) |
567 | return -ESHUTDOWN; /* Already archived */ | |
baaa35ad ZJS |
568 | else if (state == STATE_ONLINE) |
569 | return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), | |
570 | "Journal file %s is already online. Assuming unclean closing.", | |
571 | f->path); | |
572 | else if (state != STATE_OFFLINE) | |
573 | return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), | |
574 | "Journal file %s has unknown state %i.", | |
575 | f->path, state); | |
ae739cc1 | 576 | |
5b3cc0c8 YN |
577 | if (f->header->field_hash_table_size == 0 || f->header->data_hash_table_size == 0) |
578 | return -EBADMSG; | |
579 | ||
ae739cc1 LP |
580 | /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't |
581 | * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks | |
582 | * bisection. */ | |
baaa35ad ZJS |
583 | if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) |
584 | return log_debug_errno(SYNTHETIC_ERRNO(ETXTBSY), | |
585 | "Journal file %s is from the future, refusing to append new data to it that'd be older.", | |
586 | f->path); | |
cec736d2 LP |
587 | } |
588 | ||
d89c8fdf ZJS |
589 | f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header); |
590 | f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header); | |
c586dbf1 | 591 | |
f1889c91 | 592 | f->seal = JOURNAL_HEADER_SEALED(f->header); |
7560fffc | 593 | |
cec736d2 LP |
594 | return 0; |
595 | } | |
596 | ||
2678031a | 597 | static int journal_file_fstat(JournalFile *f) { |
3cc44114 LP |
598 | int r; |
599 | ||
2678031a LP |
600 | assert(f); |
601 | assert(f->fd >= 0); | |
602 | ||
603 | if (fstat(f->fd, &f->last_stat) < 0) | |
604 | return -errno; | |
605 | ||
606 | f->last_stat_usec = now(CLOCK_MONOTONIC); | |
607 | ||
8d6a4d33 | 608 | /* Refuse dealing with with files that aren't regular */ |
3cc44114 LP |
609 | r = stat_verify_regular(&f->last_stat); |
610 | if (r < 0) | |
611 | return r; | |
8d6a4d33 | 612 | |
2678031a LP |
613 | /* Refuse appending to files that are already deleted */ |
614 | if (f->last_stat.st_nlink <= 0) | |
615 | return -EIDRM; | |
616 | ||
617 | return 0; | |
618 | } | |
619 | ||
cec736d2 | 620 | static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { |
eda4b58b | 621 | uint64_t old_size, new_size; |
fec2aa2f | 622 | int r; |
cec736d2 LP |
623 | |
624 | assert(f); | |
c88cc6af | 625 | assert(f->header); |
cec736d2 | 626 | |
cec736d2 | 627 | /* We assume that this file is not sparse, and we know that |
38ac38b2 | 628 | * for sure, since we always call posix_fallocate() |
cec736d2 LP |
629 | * ourselves */ |
630 | ||
be7cdd8e | 631 | if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) |
fa6ac760 LP |
632 | return -EIO; |
633 | ||
cec736d2 | 634 | old_size = |
23b0b2b2 | 635 | le64toh(f->header->header_size) + |
cec736d2 LP |
636 | le64toh(f->header->arena_size); |
637 | ||
bc85bfee | 638 | new_size = PAGE_ALIGN(offset + size); |
23b0b2b2 LP |
639 | if (new_size < le64toh(f->header->header_size)) |
640 | new_size = le64toh(f->header->header_size); | |
bc85bfee | 641 | |
2678031a LP |
642 | if (new_size <= old_size) { |
643 | ||
644 | /* We already pre-allocated enough space, but before | |
645 | * we write to it, let's check with fstat() if the | |
646 | * file got deleted, in order make sure we don't throw | |
647 | * away the data immediately. Don't check fstat() for | |
648 | * all writes though, but only once ever 10s. */ | |
649 | ||
650 | if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC)) | |
651 | return 0; | |
652 | ||
653 | return journal_file_fstat(f); | |
654 | } | |
655 | ||
656 | /* Allocate more space. */ | |
cec736d2 | 657 | |
a676e665 | 658 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) |
bc85bfee | 659 | return -E2BIG; |
cec736d2 | 660 | |
a676e665 | 661 | if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) { |
cec736d2 LP |
662 | struct statvfs svfs; |
663 | ||
664 | if (fstatvfs(f->fd, &svfs) >= 0) { | |
665 | uint64_t available; | |
666 | ||
070052ab | 667 | available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free); |
cec736d2 LP |
668 | |
669 | if (new_size - old_size > available) | |
670 | return -E2BIG; | |
671 | } | |
672 | } | |
673 | ||
eda4b58b | 674 | /* Increase by larger blocks at once */ |
be6b0c21 | 675 | new_size = DIV_ROUND_UP(new_size, FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE; |
eda4b58b LP |
676 | if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) |
677 | new_size = f->metrics.max_size; | |
678 | ||
bc85bfee LP |
679 | /* Note that the glibc fallocate() fallback is very |
680 | inefficient, hence we try to minimize the allocation area | |
681 | as we can. */ | |
fec2aa2f GV |
682 | r = posix_fallocate(f->fd, old_size, new_size - old_size); |
683 | if (r != 0) | |
684 | return -r; | |
cec736d2 | 685 | |
23b0b2b2 | 686 | f->header->arena_size = htole64(new_size - le64toh(f->header->header_size)); |
cec736d2 | 687 | |
2678031a | 688 | return journal_file_fstat(f); |
cec736d2 LP |
689 | } |
690 | ||
78519831 | 691 | static unsigned type_to_context(ObjectType type) { |
d3d3208f | 692 | /* One context for each type, plus one catch-all for the rest */ |
69adae51 | 693 | assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS); |
fa6ac760 | 694 | assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS); |
d05089d8 | 695 | return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0; |
d3d3208f MS |
696 | } |
697 | ||
b439282e | 698 | static int journal_file_move_to(JournalFile *f, ObjectType type, bool keep_always, uint64_t offset, uint64_t size, void **ret, size_t *ret_size) { |
2678031a LP |
699 | int r; |
700 | ||
cec736d2 | 701 | assert(f); |
cec736d2 LP |
702 | assert(ret); |
703 | ||
7762e02b LP |
704 | if (size <= 0) |
705 | return -EINVAL; | |
706 | ||
2a59ea54 | 707 | /* Avoid SIGBUS on invalid accesses */ |
4bbdcdb3 LP |
708 | if (offset + size > (uint64_t) f->last_stat.st_size) { |
709 | /* Hmm, out of range? Let's refresh the fstat() data | |
710 | * first, before we trust that check. */ | |
711 | ||
2678031a LP |
712 | r = journal_file_fstat(f); |
713 | if (r < 0) | |
714 | return r; | |
715 | ||
716 | if (offset + size > (uint64_t) f->last_stat.st_size) | |
4bbdcdb3 LP |
717 | return -EADDRNOTAVAIL; |
718 | } | |
719 | ||
b439282e | 720 | return mmap_cache_get(f->mmap, f->cache_fd, f->prot, type_to_context(type), keep_always, offset, size, &f->last_stat, ret, ret_size); |
cec736d2 LP |
721 | } |
722 | ||
16e9f408 LP |
723 | static uint64_t minimum_header_size(Object *o) { |
724 | ||
b8e891e6 | 725 | static const uint64_t table[] = { |
16e9f408 LP |
726 | [OBJECT_DATA] = sizeof(DataObject), |
727 | [OBJECT_FIELD] = sizeof(FieldObject), | |
728 | [OBJECT_ENTRY] = sizeof(EntryObject), | |
729 | [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), | |
730 | [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), | |
731 | [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), | |
732 | [OBJECT_TAG] = sizeof(TagObject), | |
733 | }; | |
734 | ||
735 | if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) | |
736 | return sizeof(ObjectHeader); | |
737 | ||
738 | return table[o->object.type]; | |
739 | } | |
740 | ||
24754f36 TR |
741 | /* Lightweight object checks. We want this to be fast, so that we won't |
742 | * slowdown every journal_file_move_to_object() call too much. */ | |
743 | static int journal_file_check_object(JournalFile *f, uint64_t offset, Object *o) { | |
744 | assert(f); | |
745 | assert(o); | |
746 | ||
747 | switch (o->object.type) { | |
748 | ||
749 | case OBJECT_DATA: { | |
baaa35ad ZJS |
750 | if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) |
751 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
752 | "Bad n_entries: %" PRIu64 ": %" PRIu64, | |
753 | le64toh(o->data.n_entries), | |
754 | offset); | |
755 | ||
756 | if (le64toh(o->object.size) - offsetof(DataObject, payload) <= 0) | |
757 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
758 | "Bad object size (<= %zu): %" PRIu64 ": %" PRIu64, | |
759 | offsetof(DataObject, payload), | |
760 | le64toh(o->object.size), | |
761 | offset); | |
24754f36 | 762 | |
10e8445b TR |
763 | if (!VALID64(le64toh(o->data.next_hash_offset)) || |
764 | !VALID64(le64toh(o->data.next_field_offset)) || | |
765 | !VALID64(le64toh(o->data.entry_offset)) || | |
baaa35ad ZJS |
766 | !VALID64(le64toh(o->data.entry_array_offset))) |
767 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
768 | "Invalid offset, next_hash_offset=" OFSfmt ", next_field_offset=" OFSfmt ", entry_offset=" OFSfmt ", entry_array_offset=" OFSfmt ": %" PRIu64, | |
769 | le64toh(o->data.next_hash_offset), | |
770 | le64toh(o->data.next_field_offset), | |
771 | le64toh(o->data.entry_offset), | |
772 | le64toh(o->data.entry_array_offset), | |
773 | offset); | |
24754f36 TR |
774 | |
775 | break; | |
776 | } | |
777 | ||
778 | case OBJECT_FIELD: | |
baaa35ad ZJS |
779 | if (le64toh(o->object.size) - offsetof(FieldObject, payload) <= 0) |
780 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
781 | "Bad field size (<= %zu): %" PRIu64 ": %" PRIu64, | |
782 | offsetof(FieldObject, payload), | |
783 | le64toh(o->object.size), | |
784 | offset); | |
24754f36 | 785 | |
10e8445b | 786 | if (!VALID64(le64toh(o->field.next_hash_offset)) || |
baaa35ad ZJS |
787 | !VALID64(le64toh(o->field.head_data_offset))) |
788 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
789 | "Invalid offset, next_hash_offset=" OFSfmt ", head_data_offset=" OFSfmt ": %" PRIu64, | |
790 | le64toh(o->field.next_hash_offset), | |
791 | le64toh(o->field.head_data_offset), | |
792 | offset); | |
24754f36 TR |
793 | break; |
794 | ||
795 | case OBJECT_ENTRY: | |
baaa35ad ZJS |
796 | if ((le64toh(o->object.size) - offsetof(EntryObject, items)) % sizeof(EntryItem) != 0) |
797 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
798 | "Bad entry size (<= %zu): %" PRIu64 ": %" PRIu64, | |
799 | offsetof(EntryObject, items), | |
800 | le64toh(o->object.size), | |
801 | offset); | |
802 | ||
803 | if ((le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem) <= 0) | |
804 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
805 | "Invalid number items in entry: %" PRIu64 ": %" PRIu64, | |
806 | (le64toh(o->object.size) - offsetof(EntryObject, items)) / sizeof(EntryItem), | |
807 | offset); | |
808 | ||
809 | if (le64toh(o->entry.seqnum) <= 0) | |
810 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
811 | "Invalid entry seqnum: %" PRIx64 ": %" PRIu64, | |
812 | le64toh(o->entry.seqnum), | |
813 | offset); | |
814 | ||
815 | if (!VALID_REALTIME(le64toh(o->entry.realtime))) | |
816 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
817 | "Invalid entry realtime timestamp: %" PRIu64 ": %" PRIu64, | |
818 | le64toh(o->entry.realtime), | |
819 | offset); | |
820 | ||
821 | if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) | |
822 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
823 | "Invalid entry monotonic timestamp: %" PRIu64 ": %" PRIu64, | |
824 | le64toh(o->entry.monotonic), | |
825 | offset); | |
24754f36 TR |
826 | |
827 | break; | |
828 | ||
829 | case OBJECT_DATA_HASH_TABLE: | |
830 | case OBJECT_FIELD_HASH_TABLE: | |
831 | if ((le64toh(o->object.size) - offsetof(HashTableObject, items)) % sizeof(HashItem) != 0 || | |
baaa35ad ZJS |
832 | (le64toh(o->object.size) - offsetof(HashTableObject, items)) / sizeof(HashItem) <= 0) |
833 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
834 | "Invalid %s hash table size: %" PRIu64 ": %" PRIu64, | |
835 | o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", | |
836 | le64toh(o->object.size), | |
837 | offset); | |
24754f36 TR |
838 | |
839 | break; | |
840 | ||
841 | case OBJECT_ENTRY_ARRAY: | |
842 | if ((le64toh(o->object.size) - offsetof(EntryArrayObject, items)) % sizeof(le64_t) != 0 || | |
baaa35ad ZJS |
843 | (le64toh(o->object.size) - offsetof(EntryArrayObject, items)) / sizeof(le64_t) <= 0) |
844 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
845 | "Invalid object entry array size: %" PRIu64 ": %" PRIu64, | |
846 | le64toh(o->object.size), | |
847 | offset); | |
848 | ||
849 | if (!VALID64(le64toh(o->entry_array.next_entry_array_offset))) | |
850 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
851 | "Invalid object entry array next_entry_array_offset: " OFSfmt ": %" PRIu64, | |
852 | le64toh(o->entry_array.next_entry_array_offset), | |
853 | offset); | |
24754f36 TR |
854 | |
855 | break; | |
856 | ||
857 | case OBJECT_TAG: | |
baaa35ad ZJS |
858 | if (le64toh(o->object.size) != sizeof(TagObject)) |
859 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
860 | "Invalid object tag size: %" PRIu64 ": %" PRIu64, | |
861 | le64toh(o->object.size), | |
862 | offset); | |
24754f36 | 863 | |
baaa35ad ZJS |
864 | if (!VALID_EPOCH(le64toh(o->tag.epoch))) |
865 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
866 | "Invalid object tag epoch: %" PRIu64 ": %" PRIu64, | |
867 | le64toh(o->tag.epoch), offset); | |
24754f36 TR |
868 | |
869 | break; | |
870 | } | |
871 | ||
872 | return 0; | |
873 | } | |
874 | ||
78519831 | 875 | int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) { |
cec736d2 LP |
876 | int r; |
877 | void *t; | |
b439282e | 878 | size_t tsize; |
cec736d2 LP |
879 | Object *o; |
880 | uint64_t s; | |
881 | ||
882 | assert(f); | |
883 | assert(ret); | |
884 | ||
db11ac1a | 885 | /* Objects may only be located at multiple of 64 bit */ |
baaa35ad ZJS |
886 | if (!VALID64(offset)) |
887 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
888 | "Attempt to move to object at non-64bit boundary: %" PRIu64, | |
889 | offset); | |
db11ac1a | 890 | |
50809d7a | 891 | /* Object may not be located in the file header */ |
baaa35ad ZJS |
892 | if (offset < le64toh(f->header->header_size)) |
893 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
894 | "Attempt to move to object located in file header: %" PRIu64, | |
895 | offset); | |
50809d7a | 896 | |
b439282e | 897 | r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t, &tsize); |
cec736d2 LP |
898 | if (r < 0) |
899 | return r; | |
900 | ||
901 | o = (Object*) t; | |
902 | s = le64toh(o->object.size); | |
903 | ||
baaa35ad ZJS |
904 | if (s == 0) |
905 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
906 | "Attempt to move to uninitialized object: %" PRIu64, | |
907 | offset); | |
908 | if (s < sizeof(ObjectHeader)) | |
909 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
910 | "Attempt to move to overly short object: %" PRIu64, | |
911 | offset); | |
912 | ||
913 | if (o->object.type <= OBJECT_UNUSED) | |
914 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
915 | "Attempt to move to object with invalid type: %" PRIu64, | |
916 | offset); | |
917 | ||
918 | if (s < minimum_header_size(o)) | |
919 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
920 | "Attempt to move to truncated object: %" PRIu64, | |
921 | offset); | |
922 | ||
923 | if (type > OBJECT_UNUSED && o->object.type != type) | |
924 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
925 | "Attempt to move to object of unexpected type: %" PRIu64, | |
926 | offset); | |
cec736d2 | 927 | |
b439282e VC |
928 | if (s > tsize) { |
929 | r = journal_file_move_to(f, type, false, offset, s, &t, NULL); | |
cec736d2 LP |
930 | if (r < 0) |
931 | return r; | |
932 | ||
933 | o = (Object*) t; | |
934 | } | |
935 | ||
24754f36 TR |
936 | r = journal_file_check_object(f, offset, o); |
937 | if (r < 0) | |
938 | return r; | |
939 | ||
cec736d2 LP |
940 | *ret = o; |
941 | return 0; | |
942 | } | |
943 | ||
d98cc1f2 | 944 | static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) { |
cec736d2 LP |
945 | uint64_t r; |
946 | ||
947 | assert(f); | |
c88cc6af | 948 | assert(f->header); |
cec736d2 | 949 | |
beec0085 | 950 | r = le64toh(f->header->tail_entry_seqnum) + 1; |
c2373f84 LP |
951 | |
952 | if (seqnum) { | |
de190aef | 953 | /* If an external seqnum counter was passed, we update |
c2373f84 LP |
954 | * both the local and the external one, and set it to |
955 | * the maximum of both */ | |
956 | ||
957 | if (*seqnum + 1 > r) | |
958 | r = *seqnum + 1; | |
959 | ||
960 | *seqnum = r; | |
961 | } | |
962 | ||
beec0085 | 963 | f->header->tail_entry_seqnum = htole64(r); |
cec736d2 | 964 | |
beec0085 LP |
965 | if (f->header->head_entry_seqnum == 0) |
966 | f->header->head_entry_seqnum = htole64(r); | |
de190aef | 967 | |
cec736d2 LP |
968 | return r; |
969 | } | |
970 | ||
78519831 | 971 | int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset) { |
cec736d2 LP |
972 | int r; |
973 | uint64_t p; | |
974 | Object *tail, *o; | |
975 | void *t; | |
976 | ||
977 | assert(f); | |
c88cc6af | 978 | assert(f->header); |
d05089d8 | 979 | assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX); |
cec736d2 LP |
980 | assert(size >= sizeof(ObjectHeader)); |
981 | assert(offset); | |
982 | assert(ret); | |
983 | ||
26687bf8 OS |
984 | r = journal_file_set_online(f); |
985 | if (r < 0) | |
986 | return r; | |
987 | ||
cec736d2 | 988 | p = le64toh(f->header->tail_object_offset); |
cec736d2 | 989 | if (p == 0) |
23b0b2b2 | 990 | p = le64toh(f->header->header_size); |
cec736d2 | 991 | else { |
d05089d8 | 992 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail); |
cec736d2 LP |
993 | if (r < 0) |
994 | return r; | |
995 | ||
996 | p += ALIGN64(le64toh(tail->object.size)); | |
997 | } | |
998 | ||
999 | r = journal_file_allocate(f, p, size); | |
1000 | if (r < 0) | |
1001 | return r; | |
1002 | ||
b439282e | 1003 | r = journal_file_move_to(f, type, false, p, size, &t, NULL); |
cec736d2 LP |
1004 | if (r < 0) |
1005 | return r; | |
1006 | ||
1007 | o = (Object*) t; | |
1008 | ||
1009 | zero(o->object); | |
de190aef | 1010 | o->object.type = type; |
cec736d2 LP |
1011 | o->object.size = htole64(size); |
1012 | ||
1013 | f->header->tail_object_offset = htole64(p); | |
cec736d2 LP |
1014 | f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); |
1015 | ||
1016 | *ret = o; | |
1017 | *offset = p; | |
1018 | ||
1019 | return 0; | |
1020 | } | |
1021 | ||
de190aef | 1022 | static int journal_file_setup_data_hash_table(JournalFile *f) { |
cec736d2 LP |
1023 | uint64_t s, p; |
1024 | Object *o; | |
1025 | int r; | |
1026 | ||
1027 | assert(f); | |
c88cc6af | 1028 | assert(f->header); |
cec736d2 | 1029 | |
070052ab LP |
1030 | /* We estimate that we need 1 hash table entry per 768 bytes |
1031 | of journal file and we want to make sure we never get | |
1032 | beyond 75% fill level. Calculate the hash table size for | |
1033 | the maximum file size based on these metrics. */ | |
4a92baf3 | 1034 | |
dfabe643 | 1035 | s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); |
4a92baf3 LP |
1036 | if (s < DEFAULT_DATA_HASH_TABLE_SIZE) |
1037 | s = DEFAULT_DATA_HASH_TABLE_SIZE; | |
1038 | ||
507f22bd | 1039 | log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem)); |
4a92baf3 | 1040 | |
de190aef LP |
1041 | r = journal_file_append_object(f, |
1042 | OBJECT_DATA_HASH_TABLE, | |
1043 | offsetof(Object, hash_table.items) + s, | |
1044 | &o, &p); | |
cec736d2 LP |
1045 | if (r < 0) |
1046 | return r; | |
1047 | ||
29804cc1 | 1048 | memzero(o->hash_table.items, s); |
cec736d2 | 1049 | |
de190aef LP |
1050 | f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
1051 | f->header->data_hash_table_size = htole64(s); | |
cec736d2 LP |
1052 | |
1053 | return 0; | |
1054 | } | |
1055 | ||
de190aef | 1056 | static int journal_file_setup_field_hash_table(JournalFile *f) { |
cec736d2 LP |
1057 | uint64_t s, p; |
1058 | Object *o; | |
1059 | int r; | |
1060 | ||
1061 | assert(f); | |
c88cc6af | 1062 | assert(f->header); |
cec736d2 | 1063 | |
3c1668da LP |
1064 | /* We use a fixed size hash table for the fields as this |
1065 | * number should grow very slowly only */ | |
1066 | ||
de190aef LP |
1067 | s = DEFAULT_FIELD_HASH_TABLE_SIZE; |
1068 | r = journal_file_append_object(f, | |
1069 | OBJECT_FIELD_HASH_TABLE, | |
1070 | offsetof(Object, hash_table.items) + s, | |
1071 | &o, &p); | |
cec736d2 LP |
1072 | if (r < 0) |
1073 | return r; | |
1074 | ||
29804cc1 | 1075 | memzero(o->hash_table.items, s); |
cec736d2 | 1076 | |
de190aef LP |
1077 | f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); |
1078 | f->header->field_hash_table_size = htole64(s); | |
cec736d2 LP |
1079 | |
1080 | return 0; | |
1081 | } | |
1082 | ||
dade37d4 | 1083 | int journal_file_map_data_hash_table(JournalFile *f) { |
cec736d2 LP |
1084 | uint64_t s, p; |
1085 | void *t; | |
1086 | int r; | |
1087 | ||
1088 | assert(f); | |
c88cc6af | 1089 | assert(f->header); |
cec736d2 | 1090 | |
dade37d4 LP |
1091 | if (f->data_hash_table) |
1092 | return 0; | |
1093 | ||
de190aef LP |
1094 | p = le64toh(f->header->data_hash_table_offset); |
1095 | s = le64toh(f->header->data_hash_table_size); | |
cec736d2 | 1096 | |
de190aef | 1097 | r = journal_file_move_to(f, |
16e9f408 | 1098 | OBJECT_DATA_HASH_TABLE, |
fcde2389 | 1099 | true, |
de190aef | 1100 | p, s, |
b42549ad | 1101 | &t, NULL); |
cec736d2 LP |
1102 | if (r < 0) |
1103 | return r; | |
1104 | ||
de190aef | 1105 | f->data_hash_table = t; |
cec736d2 LP |
1106 | return 0; |
1107 | } | |
1108 | ||
dade37d4 | 1109 | int journal_file_map_field_hash_table(JournalFile *f) { |
cec736d2 LP |
1110 | uint64_t s, p; |
1111 | void *t; | |
1112 | int r; | |
1113 | ||
1114 | assert(f); | |
c88cc6af | 1115 | assert(f->header); |
cec736d2 | 1116 | |
dade37d4 LP |
1117 | if (f->field_hash_table) |
1118 | return 0; | |
1119 | ||
de190aef LP |
1120 | p = le64toh(f->header->field_hash_table_offset); |
1121 | s = le64toh(f->header->field_hash_table_size); | |
cec736d2 | 1122 | |
de190aef | 1123 | r = journal_file_move_to(f, |
16e9f408 | 1124 | OBJECT_FIELD_HASH_TABLE, |
fcde2389 | 1125 | true, |
de190aef | 1126 | p, s, |
b42549ad | 1127 | &t, NULL); |
cec736d2 LP |
1128 | if (r < 0) |
1129 | return r; | |
1130 | ||
de190aef | 1131 | f->field_hash_table = t; |
cec736d2 LP |
1132 | return 0; |
1133 | } | |
1134 | ||
3c1668da LP |
1135 | static int journal_file_link_field( |
1136 | JournalFile *f, | |
1137 | Object *o, | |
1138 | uint64_t offset, | |
1139 | uint64_t hash) { | |
1140 | ||
805d1486 | 1141 | uint64_t p, h, m; |
3c1668da LP |
1142 | int r; |
1143 | ||
1144 | assert(f); | |
c88cc6af | 1145 | assert(f->header); |
90d222c1 | 1146 | assert(f->field_hash_table); |
3c1668da LP |
1147 | assert(o); |
1148 | assert(offset > 0); | |
1149 | ||
1150 | if (o->object.type != OBJECT_FIELD) | |
1151 | return -EINVAL; | |
1152 | ||
805d1486 LP |
1153 | m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); |
1154 | if (m <= 0) | |
1155 | return -EBADMSG; | |
3c1668da | 1156 | |
805d1486 | 1157 | /* This might alter the window we are looking at */ |
3c1668da LP |
1158 | o->field.next_hash_offset = o->field.head_data_offset = 0; |
1159 | ||
805d1486 | 1160 | h = hash % m; |
3c1668da LP |
1161 | p = le64toh(f->field_hash_table[h].tail_hash_offset); |
1162 | if (p == 0) | |
1163 | f->field_hash_table[h].head_hash_offset = htole64(offset); | |
1164 | else { | |
1165 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1166 | if (r < 0) | |
1167 | return r; | |
1168 | ||
1169 | o->field.next_hash_offset = htole64(offset); | |
1170 | } | |
1171 | ||
1172 | f->field_hash_table[h].tail_hash_offset = htole64(offset); | |
1173 | ||
1174 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
1175 | f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1); | |
1176 | ||
1177 | return 0; | |
1178 | } | |
1179 | ||
1180 | static int journal_file_link_data( | |
1181 | JournalFile *f, | |
1182 | Object *o, | |
1183 | uint64_t offset, | |
1184 | uint64_t hash) { | |
1185 | ||
805d1486 | 1186 | uint64_t p, h, m; |
cec736d2 LP |
1187 | int r; |
1188 | ||
1189 | assert(f); | |
c88cc6af | 1190 | assert(f->header); |
90d222c1 | 1191 | assert(f->data_hash_table); |
cec736d2 LP |
1192 | assert(o); |
1193 | assert(offset > 0); | |
b588975f LP |
1194 | |
1195 | if (o->object.type != OBJECT_DATA) | |
1196 | return -EINVAL; | |
cec736d2 | 1197 | |
805d1486 LP |
1198 | m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); |
1199 | if (m <= 0) | |
1200 | return -EBADMSG; | |
48496df6 | 1201 | |
805d1486 | 1202 | /* This might alter the window we are looking at */ |
de190aef LP |
1203 | o->data.next_hash_offset = o->data.next_field_offset = 0; |
1204 | o->data.entry_offset = o->data.entry_array_offset = 0; | |
1205 | o->data.n_entries = 0; | |
cec736d2 | 1206 | |
805d1486 | 1207 | h = hash % m; |
8db4213e | 1208 | p = le64toh(f->data_hash_table[h].tail_hash_offset); |
3c1668da | 1209 | if (p == 0) |
cec736d2 | 1210 | /* Only entry in the hash table is easy */ |
de190aef | 1211 | f->data_hash_table[h].head_hash_offset = htole64(offset); |
3c1668da | 1212 | else { |
48496df6 LP |
1213 | /* Move back to the previous data object, to patch in |
1214 | * pointer */ | |
cec736d2 | 1215 | |
de190aef | 1216 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1217 | if (r < 0) |
1218 | return r; | |
1219 | ||
de190aef | 1220 | o->data.next_hash_offset = htole64(offset); |
cec736d2 LP |
1221 | } |
1222 | ||
de190aef | 1223 | f->data_hash_table[h].tail_hash_offset = htole64(offset); |
cec736d2 | 1224 | |
dca6219e LP |
1225 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
1226 | f->header->n_data = htole64(le64toh(f->header->n_data) + 1); | |
1227 | ||
cec736d2 LP |
1228 | return 0; |
1229 | } | |
1230 | ||
3c1668da LP |
1231 | int journal_file_find_field_object_with_hash( |
1232 | JournalFile *f, | |
1233 | const void *field, uint64_t size, uint64_t hash, | |
1234 | Object **ret, uint64_t *offset) { | |
1235 | ||
805d1486 | 1236 | uint64_t p, osize, h, m; |
3c1668da LP |
1237 | int r; |
1238 | ||
1239 | assert(f); | |
c88cc6af | 1240 | assert(f->header); |
3c1668da LP |
1241 | assert(field && size > 0); |
1242 | ||
dade37d4 LP |
1243 | /* If the field hash table is empty, we can't find anything */ |
1244 | if (le64toh(f->header->field_hash_table_size) <= 0) | |
1245 | return 0; | |
1246 | ||
1247 | /* Map the field hash table, if it isn't mapped yet. */ | |
1248 | r = journal_file_map_field_hash_table(f); | |
1249 | if (r < 0) | |
1250 | return r; | |
1251 | ||
3c1668da LP |
1252 | osize = offsetof(Object, field.payload) + size; |
1253 | ||
805d1486 | 1254 | m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); |
805d1486 | 1255 | if (m <= 0) |
3c1668da LP |
1256 | return -EBADMSG; |
1257 | ||
805d1486 | 1258 | h = hash % m; |
3c1668da LP |
1259 | p = le64toh(f->field_hash_table[h].head_hash_offset); |
1260 | ||
1261 | while (p > 0) { | |
1262 | Object *o; | |
1263 | ||
1264 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1265 | if (r < 0) | |
1266 | return r; | |
1267 | ||
1268 | if (le64toh(o->field.hash) == hash && | |
1269 | le64toh(o->object.size) == osize && | |
1270 | memcmp(o->field.payload, field, size) == 0) { | |
1271 | ||
1272 | if (ret) | |
1273 | *ret = o; | |
1274 | if (offset) | |
1275 | *offset = p; | |
1276 | ||
1277 | return 1; | |
1278 | } | |
1279 | ||
1280 | p = le64toh(o->field.next_hash_offset); | |
1281 | } | |
1282 | ||
1283 | return 0; | |
1284 | } | |
1285 | ||
1286 | int journal_file_find_field_object( | |
1287 | JournalFile *f, | |
1288 | const void *field, uint64_t size, | |
1289 | Object **ret, uint64_t *offset) { | |
1290 | ||
1291 | uint64_t hash; | |
1292 | ||
1293 | assert(f); | |
1294 | assert(field && size > 0); | |
1295 | ||
1296 | hash = hash64(field, size); | |
1297 | ||
1298 | return journal_file_find_field_object_with_hash(f, | |
1299 | field, size, hash, | |
1300 | ret, offset); | |
1301 | } | |
1302 | ||
de190aef LP |
1303 | int journal_file_find_data_object_with_hash( |
1304 | JournalFile *f, | |
1305 | const void *data, uint64_t size, uint64_t hash, | |
1306 | Object **ret, uint64_t *offset) { | |
48496df6 | 1307 | |
805d1486 | 1308 | uint64_t p, osize, h, m; |
cec736d2 LP |
1309 | int r; |
1310 | ||
1311 | assert(f); | |
c88cc6af | 1312 | assert(f->header); |
cec736d2 LP |
1313 | assert(data || size == 0); |
1314 | ||
dade37d4 LP |
1315 | /* If there's no data hash table, then there's no entry. */ |
1316 | if (le64toh(f->header->data_hash_table_size) <= 0) | |
1317 | return 0; | |
1318 | ||
1319 | /* Map the data hash table, if it isn't mapped yet. */ | |
1320 | r = journal_file_map_data_hash_table(f); | |
1321 | if (r < 0) | |
1322 | return r; | |
1323 | ||
cec736d2 LP |
1324 | osize = offsetof(Object, data.payload) + size; |
1325 | ||
805d1486 LP |
1326 | m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); |
1327 | if (m <= 0) | |
bc85bfee LP |
1328 | return -EBADMSG; |
1329 | ||
805d1486 | 1330 | h = hash % m; |
de190aef | 1331 | p = le64toh(f->data_hash_table[h].head_hash_offset); |
cec736d2 | 1332 | |
de190aef LP |
1333 | while (p > 0) { |
1334 | Object *o; | |
cec736d2 | 1335 | |
de190aef | 1336 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); |
cec736d2 LP |
1337 | if (r < 0) |
1338 | return r; | |
1339 | ||
807e17f0 | 1340 | if (le64toh(o->data.hash) != hash) |
85a131e8 | 1341 | goto next; |
807e17f0 | 1342 | |
d89c8fdf | 1343 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
349cc4a5 | 1344 | #if HAVE_XZ || HAVE_LZ4 |
fa1c4b51 | 1345 | uint64_t l; |
a7f7d1bd | 1346 | size_t rsize = 0; |
cec736d2 | 1347 | |
807e17f0 LP |
1348 | l = le64toh(o->object.size); |
1349 | if (l <= offsetof(Object, data.payload)) | |
cec736d2 LP |
1350 | return -EBADMSG; |
1351 | ||
807e17f0 LP |
1352 | l -= offsetof(Object, data.payload); |
1353 | ||
d89c8fdf ZJS |
1354 | r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, |
1355 | o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0); | |
1356 | if (r < 0) | |
1357 | return r; | |
807e17f0 | 1358 | |
b785c858 | 1359 | if (rsize == size && |
807e17f0 LP |
1360 | memcmp(f->compress_buffer, data, size) == 0) { |
1361 | ||
1362 | if (ret) | |
1363 | *ret = o; | |
1364 | ||
1365 | if (offset) | |
1366 | *offset = p; | |
1367 | ||
1368 | return 1; | |
1369 | } | |
3b1a55e1 ZJS |
1370 | #else |
1371 | return -EPROTONOSUPPORT; | |
1372 | #endif | |
807e17f0 LP |
1373 | } else if (le64toh(o->object.size) == osize && |
1374 | memcmp(o->data.payload, data, size) == 0) { | |
1375 | ||
cec736d2 LP |
1376 | if (ret) |
1377 | *ret = o; | |
1378 | ||
1379 | if (offset) | |
1380 | *offset = p; | |
1381 | ||
de190aef | 1382 | return 1; |
cec736d2 LP |
1383 | } |
1384 | ||
85a131e8 | 1385 | next: |
cec736d2 LP |
1386 | p = le64toh(o->data.next_hash_offset); |
1387 | } | |
1388 | ||
de190aef LP |
1389 | return 0; |
1390 | } | |
1391 | ||
1392 | int journal_file_find_data_object( | |
1393 | JournalFile *f, | |
1394 | const void *data, uint64_t size, | |
1395 | Object **ret, uint64_t *offset) { | |
1396 | ||
1397 | uint64_t hash; | |
1398 | ||
1399 | assert(f); | |
1400 | assert(data || size == 0); | |
1401 | ||
1402 | hash = hash64(data, size); | |
1403 | ||
1404 | return journal_file_find_data_object_with_hash(f, | |
1405 | data, size, hash, | |
1406 | ret, offset); | |
1407 | } | |
1408 | ||
3c1668da LP |
1409 | static int journal_file_append_field( |
1410 | JournalFile *f, | |
1411 | const void *field, uint64_t size, | |
1412 | Object **ret, uint64_t *offset) { | |
1413 | ||
1414 | uint64_t hash, p; | |
1415 | uint64_t osize; | |
1416 | Object *o; | |
1417 | int r; | |
1418 | ||
1419 | assert(f); | |
1420 | assert(field && size > 0); | |
1421 | ||
1422 | hash = hash64(field, size); | |
1423 | ||
1424 | r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p); | |
1425 | if (r < 0) | |
1426 | return r; | |
1427 | else if (r > 0) { | |
1428 | ||
1429 | if (ret) | |
1430 | *ret = o; | |
1431 | ||
1432 | if (offset) | |
1433 | *offset = p; | |
1434 | ||
1435 | return 0; | |
1436 | } | |
1437 | ||
1438 | osize = offsetof(Object, field.payload) + size; | |
1439 | r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p); | |
8c92d4bb LP |
1440 | if (r < 0) |
1441 | return r; | |
3c1668da LP |
1442 | |
1443 | o->field.hash = htole64(hash); | |
1444 | memcpy(o->field.payload, field, size); | |
1445 | ||
1446 | r = journal_file_link_field(f, o, p, hash); | |
1447 | if (r < 0) | |
1448 | return r; | |
1449 | ||
1450 | /* The linking might have altered the window, so let's | |
1451 | * refresh our pointer */ | |
1452 | r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); | |
1453 | if (r < 0) | |
1454 | return r; | |
1455 | ||
349cc4a5 | 1456 | #if HAVE_GCRYPT |
3c1668da LP |
1457 | r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p); |
1458 | if (r < 0) | |
1459 | return r; | |
1460 | #endif | |
1461 | ||
1462 | if (ret) | |
1463 | *ret = o; | |
1464 | ||
1465 | if (offset) | |
1466 | *offset = p; | |
1467 | ||
1468 | return 0; | |
1469 | } | |
1470 | ||
48496df6 LP |
1471 | static int journal_file_append_data( |
1472 | JournalFile *f, | |
1473 | const void *data, uint64_t size, | |
1474 | Object **ret, uint64_t *offset) { | |
1475 | ||
de190aef LP |
1476 | uint64_t hash, p; |
1477 | uint64_t osize; | |
1478 | Object *o; | |
d89c8fdf | 1479 | int r, compression = 0; |
3c1668da | 1480 | const void *eq; |
de190aef LP |
1481 | |
1482 | assert(f); | |
1483 | assert(data || size == 0); | |
1484 | ||
1485 | hash = hash64(data, size); | |
1486 | ||
1487 | r = journal_file_find_data_object_with_hash(f, data, size, hash, &o, &p); | |
1488 | if (r < 0) | |
1489 | return r; | |
0240c603 | 1490 | if (r > 0) { |
de190aef LP |
1491 | |
1492 | if (ret) | |
1493 | *ret = o; | |
1494 | ||
1495 | if (offset) | |
1496 | *offset = p; | |
1497 | ||
1498 | return 0; | |
1499 | } | |
1500 | ||
1501 | osize = offsetof(Object, data.payload) + size; | |
1502 | r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); | |
cec736d2 LP |
1503 | if (r < 0) |
1504 | return r; | |
1505 | ||
cec736d2 | 1506 | o->data.hash = htole64(hash); |
807e17f0 | 1507 | |
349cc4a5 | 1508 | #if HAVE_XZ || HAVE_LZ4 |
57850536 | 1509 | if (JOURNAL_FILE_COMPRESS(f) && size >= f->compress_threshold_bytes) { |
a7f7d1bd | 1510 | size_t rsize = 0; |
807e17f0 | 1511 | |
5d6f46b6 | 1512 | compression = compress_blob(data, size, o->data.payload, size - 1, &rsize); |
807e17f0 | 1513 | |
d1afbcd2 | 1514 | if (compression >= 0) { |
807e17f0 | 1515 | o->object.size = htole64(offsetof(Object, data.payload) + rsize); |
d89c8fdf | 1516 | o->object.flags |= compression; |
807e17f0 | 1517 | |
fa1c4b51 | 1518 | log_debug("Compressed data object %"PRIu64" -> %zu using %s", |
d89c8fdf | 1519 | size, rsize, object_compressed_to_string(compression)); |
d1afbcd2 LP |
1520 | } else |
1521 | /* Compression didn't work, we don't really care why, let's continue without compression */ | |
1522 | compression = 0; | |
807e17f0 LP |
1523 | } |
1524 | #endif | |
1525 | ||
75f32f04 ZJS |
1526 | if (compression == 0) |
1527 | memcpy_safe(o->data.payload, data, size); | |
cec736d2 | 1528 | |
de190aef | 1529 | r = journal_file_link_data(f, o, p, hash); |
cec736d2 LP |
1530 | if (r < 0) |
1531 | return r; | |
1532 | ||
349cc4a5 | 1533 | #if HAVE_GCRYPT |
33685a5a FB |
1534 | r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p); |
1535 | if (r < 0) | |
1536 | return r; | |
1537 | #endif | |
1538 | ||
48496df6 LP |
1539 | /* The linking might have altered the window, so let's |
1540 | * refresh our pointer */ | |
1541 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
1542 | if (r < 0) | |
1543 | return r; | |
1544 | ||
08c6f819 SL |
1545 | if (!data) |
1546 | eq = NULL; | |
1547 | else | |
1548 | eq = memchr(data, '=', size); | |
3c1668da | 1549 | if (eq && eq > data) { |
748db592 | 1550 | Object *fo = NULL; |
3c1668da | 1551 | uint64_t fp; |
3c1668da LP |
1552 | |
1553 | /* Create field object ... */ | |
1554 | r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp); | |
1555 | if (r < 0) | |
1556 | return r; | |
1557 | ||
1558 | /* ... and link it in. */ | |
1559 | o->data.next_field_offset = fo->field.head_data_offset; | |
1560 | fo->field.head_data_offset = le64toh(p); | |
1561 | } | |
1562 | ||
cec736d2 LP |
1563 | if (ret) |
1564 | *ret = o; | |
1565 | ||
1566 | if (offset) | |
de190aef | 1567 | *offset = p; |
cec736d2 LP |
1568 | |
1569 | return 0; | |
1570 | } | |
1571 | ||
1572 | uint64_t journal_file_entry_n_items(Object *o) { | |
1573 | assert(o); | |
b588975f LP |
1574 | |
1575 | if (o->object.type != OBJECT_ENTRY) | |
1576 | return 0; | |
cec736d2 LP |
1577 | |
1578 | return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem); | |
1579 | } | |
1580 | ||
0284adc6 | 1581 | uint64_t journal_file_entry_array_n_items(Object *o) { |
de190aef | 1582 | assert(o); |
b588975f LP |
1583 | |
1584 | if (o->object.type != OBJECT_ENTRY_ARRAY) | |
1585 | return 0; | |
de190aef LP |
1586 | |
1587 | return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t); | |
1588 | } | |
1589 | ||
fb9a24b6 LP |
1590 | uint64_t journal_file_hash_table_n_items(Object *o) { |
1591 | assert(o); | |
b588975f | 1592 | |
ec2ce0c5 | 1593 | if (!IN_SET(o->object.type, OBJECT_DATA_HASH_TABLE, OBJECT_FIELD_HASH_TABLE)) |
b588975f | 1594 | return 0; |
fb9a24b6 LP |
1595 | |
1596 | return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem); | |
1597 | } | |
1598 | ||
de190aef | 1599 | static int link_entry_into_array(JournalFile *f, |
4fd052ae FC |
1600 | le64_t *first, |
1601 | le64_t *idx, | |
de190aef | 1602 | uint64_t p) { |
cec736d2 | 1603 | int r; |
de190aef LP |
1604 | uint64_t n = 0, ap = 0, q, i, a, hidx; |
1605 | Object *o; | |
1606 | ||
cec736d2 | 1607 | assert(f); |
c88cc6af | 1608 | assert(f->header); |
de190aef LP |
1609 | assert(first); |
1610 | assert(idx); | |
1611 | assert(p > 0); | |
cec736d2 | 1612 | |
de190aef LP |
1613 | a = le64toh(*first); |
1614 | i = hidx = le64toh(*idx); | |
1615 | while (a > 0) { | |
1616 | ||
1617 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); | |
1618 | if (r < 0) | |
1619 | return r; | |
cec736d2 | 1620 | |
de190aef LP |
1621 | n = journal_file_entry_array_n_items(o); |
1622 | if (i < n) { | |
1623 | o->entry_array.items[i] = htole64(p); | |
1624 | *idx = htole64(hidx + 1); | |
1625 | return 0; | |
1626 | } | |
cec736d2 | 1627 | |
de190aef LP |
1628 | i -= n; |
1629 | ap = a; | |
1630 | a = le64toh(o->entry_array.next_entry_array_offset); | |
1631 | } | |
1632 | ||
1633 | if (hidx > n) | |
1634 | n = (hidx+1) * 2; | |
1635 | else | |
1636 | n = n * 2; | |
1637 | ||
1638 | if (n < 4) | |
1639 | n = 4; | |
1640 | ||
1641 | r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, | |
1642 | offsetof(Object, entry_array.items) + n * sizeof(uint64_t), | |
1643 | &o, &q); | |
cec736d2 LP |
1644 | if (r < 0) |
1645 | return r; | |
1646 | ||
349cc4a5 | 1647 | #if HAVE_GCRYPT |
5996c7c2 | 1648 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q); |
b0af6f41 LP |
1649 | if (r < 0) |
1650 | return r; | |
feb12d3e | 1651 | #endif |
b0af6f41 | 1652 | |
de190aef | 1653 | o->entry_array.items[i] = htole64(p); |
cec736d2 | 1654 | |
de190aef | 1655 | if (ap == 0) |
7be3aa17 | 1656 | *first = htole64(q); |
cec736d2 | 1657 | else { |
de190aef | 1658 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); |
cec736d2 LP |
1659 | if (r < 0) |
1660 | return r; | |
1661 | ||
de190aef LP |
1662 | o->entry_array.next_entry_array_offset = htole64(q); |
1663 | } | |
cec736d2 | 1664 | |
2dee23eb LP |
1665 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
1666 | f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); | |
1667 | ||
de190aef LP |
1668 | *idx = htole64(hidx + 1); |
1669 | ||
1670 | return 0; | |
1671 | } | |
cec736d2 | 1672 | |
de190aef | 1673 | static int link_entry_into_array_plus_one(JournalFile *f, |
4fd052ae FC |
1674 | le64_t *extra, |
1675 | le64_t *first, | |
1676 | le64_t *idx, | |
de190aef LP |
1677 | uint64_t p) { |
1678 | ||
1679 | int r; | |
1680 | ||
1681 | assert(f); | |
1682 | assert(extra); | |
1683 | assert(first); | |
1684 | assert(idx); | |
1685 | assert(p > 0); | |
1686 | ||
1687 | if (*idx == 0) | |
1688 | *extra = htole64(p); | |
1689 | else { | |
4fd052ae | 1690 | le64_t i; |
de190aef | 1691 | |
7be3aa17 | 1692 | i = htole64(le64toh(*idx) - 1); |
de190aef LP |
1693 | r = link_entry_into_array(f, first, &i, p); |
1694 | if (r < 0) | |
1695 | return r; | |
cec736d2 LP |
1696 | } |
1697 | ||
de190aef LP |
1698 | *idx = htole64(le64toh(*idx) + 1); |
1699 | return 0; | |
1700 | } | |
1701 | ||
1702 | static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) { | |
1703 | uint64_t p; | |
1704 | int r; | |
1705 | assert(f); | |
1706 | assert(o); | |
1707 | assert(offset > 0); | |
1708 | ||
1709 | p = le64toh(o->entry.items[i].object_offset); | |
1710 | if (p == 0) | |
1711 | return -EINVAL; | |
1712 | ||
1713 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
cec736d2 LP |
1714 | if (r < 0) |
1715 | return r; | |
1716 | ||
de190aef LP |
1717 | return link_entry_into_array_plus_one(f, |
1718 | &o->data.entry_offset, | |
1719 | &o->data.entry_array_offset, | |
1720 | &o->data.n_entries, | |
1721 | offset); | |
cec736d2 LP |
1722 | } |
1723 | ||
1724 | static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) { | |
de190aef | 1725 | uint64_t n, i; |
cec736d2 LP |
1726 | int r; |
1727 | ||
1728 | assert(f); | |
c88cc6af | 1729 | assert(f->header); |
cec736d2 LP |
1730 | assert(o); |
1731 | assert(offset > 0); | |
b588975f LP |
1732 | |
1733 | if (o->object.type != OBJECT_ENTRY) | |
1734 | return -EINVAL; | |
cec736d2 | 1735 | |
b788cc23 LP |
1736 | __sync_synchronize(); |
1737 | ||
cec736d2 | 1738 | /* Link up the entry itself */ |
de190aef LP |
1739 | r = link_entry_into_array(f, |
1740 | &f->header->entry_array_offset, | |
1741 | &f->header->n_entries, | |
1742 | offset); | |
1743 | if (r < 0) | |
1744 | return r; | |
cec736d2 | 1745 | |
507f22bd | 1746 | /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */ |
cec736d2 | 1747 | |
de190aef | 1748 | if (f->header->head_entry_realtime == 0) |
0ac38b70 | 1749 | f->header->head_entry_realtime = o->entry.realtime; |
cec736d2 | 1750 | |
0ac38b70 | 1751 | f->header->tail_entry_realtime = o->entry.realtime; |
de190aef LP |
1752 | f->header->tail_entry_monotonic = o->entry.monotonic; |
1753 | ||
cec736d2 LP |
1754 | /* Link up the items */ |
1755 | n = journal_file_entry_n_items(o); | |
1756 | for (i = 0; i < n; i++) { | |
1757 | r = journal_file_link_entry_item(f, o, offset, i); | |
1758 | if (r < 0) | |
1759 | return r; | |
1760 | } | |
1761 | ||
cec736d2 LP |
1762 | return 0; |
1763 | } | |
1764 | ||
1765 | static int journal_file_append_entry_internal( | |
1766 | JournalFile *f, | |
1767 | const dual_timestamp *ts, | |
d180c349 | 1768 | const sd_id128_t *boot_id, |
cec736d2 LP |
1769 | uint64_t xor_hash, |
1770 | const EntryItem items[], unsigned n_items, | |
de190aef | 1771 | uint64_t *seqnum, |
cec736d2 LP |
1772 | Object **ret, uint64_t *offset) { |
1773 | uint64_t np; | |
1774 | uint64_t osize; | |
1775 | Object *o; | |
1776 | int r; | |
1777 | ||
1778 | assert(f); | |
c88cc6af | 1779 | assert(f->header); |
cec736d2 | 1780 | assert(items || n_items == 0); |
de190aef | 1781 | assert(ts); |
cec736d2 LP |
1782 | |
1783 | osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem)); | |
1784 | ||
de190aef | 1785 | r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); |
cec736d2 LP |
1786 | if (r < 0) |
1787 | return r; | |
1788 | ||
d98cc1f2 | 1789 | o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); |
75f32f04 | 1790 | memcpy_safe(o->entry.items, items, n_items * sizeof(EntryItem)); |
de190aef LP |
1791 | o->entry.realtime = htole64(ts->realtime); |
1792 | o->entry.monotonic = htole64(ts->monotonic); | |
cec736d2 | 1793 | o->entry.xor_hash = htole64(xor_hash); |
d180c349 | 1794 | o->entry.boot_id = boot_id ? *boot_id : f->header->boot_id; |
cec736d2 | 1795 | |
349cc4a5 | 1796 | #if HAVE_GCRYPT |
5996c7c2 | 1797 | r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np); |
b0af6f41 LP |
1798 | if (r < 0) |
1799 | return r; | |
feb12d3e | 1800 | #endif |
b0af6f41 | 1801 | |
cec736d2 LP |
1802 | r = journal_file_link_entry(f, o, np); |
1803 | if (r < 0) | |
1804 | return r; | |
1805 | ||
1806 | if (ret) | |
1807 | *ret = o; | |
1808 | ||
1809 | if (offset) | |
1810 | *offset = np; | |
1811 | ||
1812 | return 0; | |
1813 | } | |
1814 | ||
cf244689 | 1815 | void journal_file_post_change(JournalFile *f) { |
50f20cfd LP |
1816 | assert(f); |
1817 | ||
c5236850 DT |
1818 | if (f->fd < 0) |
1819 | return; | |
1820 | ||
50f20cfd LP |
1821 | /* inotify() does not receive IN_MODIFY events from file |
1822 | * accesses done via mmap(). After each access we hence | |
1823 | * trigger IN_MODIFY by truncating the journal file to its | |
1824 | * current size which triggers IN_MODIFY. */ | |
1825 | ||
bc85bfee LP |
1826 | __sync_synchronize(); |
1827 | ||
50f20cfd | 1828 | if (ftruncate(f->fd, f->last_stat.st_size) < 0) |
e167d7fd | 1829 | log_debug_errno(errno, "Failed to truncate file to its own size: %m"); |
50f20cfd LP |
1830 | } |
1831 | ||
7a24f3bf VC |
1832 | static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) { |
1833 | assert(userdata); | |
1834 | ||
1835 | journal_file_post_change(userdata); | |
1836 | ||
1837 | return 1; | |
1838 | } | |
1839 | ||
1840 | static void schedule_post_change(JournalFile *f) { | |
7a24f3bf | 1841 | uint64_t now; |
b6cdfbe5 | 1842 | int r; |
7a24f3bf VC |
1843 | |
1844 | assert(f); | |
1845 | assert(f->post_change_timer); | |
1846 | ||
b6cdfbe5 | 1847 | r = sd_event_source_get_enabled(f->post_change_timer, NULL); |
7a24f3bf | 1848 | if (r < 0) { |
e167d7fd LP |
1849 | log_debug_errno(r, "Failed to get ftruncate timer state: %m"); |
1850 | goto fail; | |
7a24f3bf | 1851 | } |
b6cdfbe5 | 1852 | if (r > 0) |
7a24f3bf VC |
1853 | return; |
1854 | ||
ca5d90d4 | 1855 | r = sd_event_now(sd_event_source_get_event(f->post_change_timer), CLOCK_MONOTONIC, &now); |
7a24f3bf | 1856 | if (r < 0) { |
e167d7fd LP |
1857 | log_debug_errno(r, "Failed to get clock's now for scheduling ftruncate: %m"); |
1858 | goto fail; | |
7a24f3bf VC |
1859 | } |
1860 | ||
ca5d90d4 | 1861 | r = sd_event_source_set_time(f->post_change_timer, now + f->post_change_timer_period); |
7a24f3bf | 1862 | if (r < 0) { |
e167d7fd LP |
1863 | log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m"); |
1864 | goto fail; | |
7a24f3bf VC |
1865 | } |
1866 | ||
ca5d90d4 | 1867 | r = sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_ONESHOT); |
7a24f3bf | 1868 | if (r < 0) { |
e167d7fd LP |
1869 | log_debug_errno(r, "Failed to enable scheduled ftruncate: %m"); |
1870 | goto fail; | |
7a24f3bf | 1871 | } |
e167d7fd LP |
1872 | |
1873 | return; | |
1874 | ||
1875 | fail: | |
1876 | /* On failure, let's simply post the change immediately. */ | |
1877 | journal_file_post_change(f); | |
7a24f3bf VC |
1878 | } |
1879 | ||
1880 | /* Enable coalesced change posting in a timer on the provided sd_event instance */ | |
1881 | int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) { | |
1882 | _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL; | |
1883 | int r; | |
1884 | ||
1885 | assert(f); | |
1886 | assert_return(!f->post_change_timer, -EINVAL); | |
1887 | assert(e); | |
1888 | assert(t); | |
1889 | ||
1890 | r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f); | |
1891 | if (r < 0) | |
1892 | return r; | |
1893 | ||
1894 | r = sd_event_source_set_enabled(timer, SD_EVENT_OFF); | |
1895 | if (r < 0) | |
1896 | return r; | |
1897 | ||
1cc6c93a | 1898 | f->post_change_timer = TAKE_PTR(timer); |
7a24f3bf VC |
1899 | f->post_change_timer_period = t; |
1900 | ||
1901 | return r; | |
1902 | } | |
1903 | ||
93bab288 YW |
1904 | static int entry_item_cmp(const EntryItem *a, const EntryItem *b) { |
1905 | return CMP(le64toh(a->object_offset), le64toh(b->object_offset)); | |
1f2da9ec LP |
1906 | } |
1907 | ||
d180c349 ZJS |
1908 | int journal_file_append_entry( |
1909 | JournalFile *f, | |
1910 | const dual_timestamp *ts, | |
1911 | const sd_id128_t *boot_id, | |
1912 | const struct iovec iovec[], unsigned n_iovec, | |
1913 | uint64_t *seqnum, | |
1914 | Object **ret, uint64_t *offset) { | |
1915 | ||
cec736d2 LP |
1916 | unsigned i; |
1917 | EntryItem *items; | |
1918 | int r; | |
1919 | uint64_t xor_hash = 0; | |
de190aef | 1920 | struct dual_timestamp _ts; |
cec736d2 LP |
1921 | |
1922 | assert(f); | |
c88cc6af | 1923 | assert(f->header); |
cec736d2 LP |
1924 | assert(iovec || n_iovec == 0); |
1925 | ||
c6273953 | 1926 | if (ts) { |
baaa35ad ZJS |
1927 | if (!VALID_REALTIME(ts->realtime)) |
1928 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
1929 | "Invalid realtime timestamp %" PRIu64 ", refusing entry.", | |
1930 | ts->realtime); | |
1931 | if (!VALID_MONOTONIC(ts->monotonic)) | |
1932 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
1933 | "Invalid monotomic timestamp %" PRIu64 ", refusing entry.", | |
1934 | ts->monotonic); | |
c6273953 | 1935 | } else { |
de190aef LP |
1936 | dual_timestamp_get(&_ts); |
1937 | ts = &_ts; | |
1938 | } | |
1939 | ||
349cc4a5 | 1940 | #if HAVE_GCRYPT |
7560fffc LP |
1941 | r = journal_file_maybe_append_tag(f, ts->realtime); |
1942 | if (r < 0) | |
1943 | return r; | |
feb12d3e | 1944 | #endif |
7560fffc | 1945 | |
64825d3c | 1946 | /* alloca() can't take 0, hence let's allocate at least one */ |
cf409d15 | 1947 | items = newa(EntryItem, MAX(1u, n_iovec)); |
cec736d2 LP |
1948 | |
1949 | for (i = 0; i < n_iovec; i++) { | |
1950 | uint64_t p; | |
1951 | Object *o; | |
1952 | ||
1953 | r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); | |
1954 | if (r < 0) | |
cf244689 | 1955 | return r; |
cec736d2 LP |
1956 | |
1957 | xor_hash ^= le64toh(o->data.hash); | |
1958 | items[i].object_offset = htole64(p); | |
de7b95cd | 1959 | items[i].hash = o->data.hash; |
cec736d2 LP |
1960 | } |
1961 | ||
1f2da9ec LP |
1962 | /* Order by the position on disk, in order to improve seek |
1963 | * times for rotating media. */ | |
93bab288 | 1964 | typesafe_qsort(items, n_iovec, entry_item_cmp); |
1f2da9ec | 1965 | |
d180c349 | 1966 | r = journal_file_append_entry_internal(f, ts, boot_id, xor_hash, items, n_iovec, seqnum, ret, offset); |
cec736d2 | 1967 | |
fa6ac760 LP |
1968 | /* If the memory mapping triggered a SIGBUS then we return an |
1969 | * IO error and ignore the error code passed down to us, since | |
1970 | * it is very likely just an effect of a nullified replacement | |
1971 | * mapping page */ | |
1972 | ||
be7cdd8e | 1973 | if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) |
fa6ac760 LP |
1974 | r = -EIO; |
1975 | ||
7a24f3bf VC |
1976 | if (f->post_change_timer) |
1977 | schedule_post_change(f); | |
1978 | else | |
1979 | journal_file_post_change(f); | |
50f20cfd | 1980 | |
cec736d2 LP |
1981 | return r; |
1982 | } | |
1983 | ||
a4bcff5b | 1984 | typedef struct ChainCacheItem { |
fb099c8d | 1985 | uint64_t first; /* the array at the beginning of the chain */ |
a4bcff5b LP |
1986 | uint64_t array; /* the cached array */ |
1987 | uint64_t begin; /* the first item in the cached array */ | |
1988 | uint64_t total; /* the total number of items in all arrays before this one in the chain */ | |
f268980d | 1989 | uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */ |
a4bcff5b LP |
1990 | } ChainCacheItem; |
1991 | ||
1992 | static void chain_cache_put( | |
4743015d | 1993 | OrderedHashmap *h, |
a4bcff5b LP |
1994 | ChainCacheItem *ci, |
1995 | uint64_t first, | |
1996 | uint64_t array, | |
1997 | uint64_t begin, | |
f268980d LP |
1998 | uint64_t total, |
1999 | uint64_t last_index) { | |
a4bcff5b LP |
2000 | |
2001 | if (!ci) { | |
34741aa3 LP |
2002 | /* If the chain item to cache for this chain is the |
2003 | * first one it's not worth caching anything */ | |
2004 | if (array == first) | |
2005 | return; | |
2006 | ||
29433089 | 2007 | if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) { |
4743015d | 2008 | ci = ordered_hashmap_steal_first(h); |
29433089 LP |
2009 | assert(ci); |
2010 | } else { | |
a4bcff5b LP |
2011 | ci = new(ChainCacheItem, 1); |
2012 | if (!ci) | |
2013 | return; | |
2014 | } | |
2015 | ||
2016 | ci->first = first; | |
2017 | ||
4743015d | 2018 | if (ordered_hashmap_put(h, &ci->first, ci) < 0) { |
a4bcff5b LP |
2019 | free(ci); |
2020 | return; | |
2021 | } | |
2022 | } else | |
2023 | assert(ci->first == first); | |
2024 | ||
2025 | ci->array = array; | |
2026 | ci->begin = begin; | |
2027 | ci->total = total; | |
f268980d | 2028 | ci->last_index = last_index; |
a4bcff5b LP |
2029 | } |
2030 | ||
f268980d LP |
2031 | static int generic_array_get( |
2032 | JournalFile *f, | |
2033 | uint64_t first, | |
2034 | uint64_t i, | |
2035 | Object **ret, uint64_t *offset) { | |
de190aef | 2036 | |
cec736d2 | 2037 | Object *o; |
a4bcff5b | 2038 | uint64_t p = 0, a, t = 0; |
cec736d2 | 2039 | int r; |
a4bcff5b | 2040 | ChainCacheItem *ci; |
cec736d2 LP |
2041 | |
2042 | assert(f); | |
2043 | ||
de190aef | 2044 | a = first; |
a4bcff5b LP |
2045 | |
2046 | /* Try the chain cache first */ | |
4743015d | 2047 | ci = ordered_hashmap_get(f->chain_cache, &first); |
a4bcff5b LP |
2048 | if (ci && i > ci->total) { |
2049 | a = ci->array; | |
2050 | i -= ci->total; | |
2051 | t = ci->total; | |
2052 | } | |
2053 | ||
de190aef | 2054 | while (a > 0) { |
a4bcff5b | 2055 | uint64_t k; |
cec736d2 | 2056 | |
de190aef LP |
2057 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); |
2058 | if (r < 0) | |
2059 | return r; | |
cec736d2 | 2060 | |
a4bcff5b LP |
2061 | k = journal_file_entry_array_n_items(o); |
2062 | if (i < k) { | |
de190aef | 2063 | p = le64toh(o->entry_array.items[i]); |
a4bcff5b | 2064 | goto found; |
cec736d2 LP |
2065 | } |
2066 | ||
a4bcff5b LP |
2067 | i -= k; |
2068 | t += k; | |
de190aef LP |
2069 | a = le64toh(o->entry_array.next_entry_array_offset); |
2070 | } | |
2071 | ||
a4bcff5b LP |
2072 | return 0; |
2073 | ||
2074 | found: | |
2075 | /* Let's cache this item for the next invocation */ | |
af13a6b0 | 2076 | chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i); |
de190aef LP |
2077 | |
2078 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2079 | if (r < 0) | |
2080 | return r; | |
2081 | ||
2082 | if (ret) | |
2083 | *ret = o; | |
2084 | ||
2085 | if (offset) | |
2086 | *offset = p; | |
2087 | ||
2088 | return 1; | |
2089 | } | |
2090 | ||
f268980d LP |
2091 | static int generic_array_get_plus_one( |
2092 | JournalFile *f, | |
2093 | uint64_t extra, | |
2094 | uint64_t first, | |
2095 | uint64_t i, | |
2096 | Object **ret, uint64_t *offset) { | |
de190aef LP |
2097 | |
2098 | Object *o; | |
2099 | ||
2100 | assert(f); | |
2101 | ||
2102 | if (i == 0) { | |
2103 | int r; | |
2104 | ||
2105 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
cec736d2 LP |
2106 | if (r < 0) |
2107 | return r; | |
2108 | ||
de190aef LP |
2109 | if (ret) |
2110 | *ret = o; | |
cec736d2 | 2111 | |
de190aef LP |
2112 | if (offset) |
2113 | *offset = extra; | |
cec736d2 | 2114 | |
de190aef | 2115 | return 1; |
cec736d2 LP |
2116 | } |
2117 | ||
de190aef LP |
2118 | return generic_array_get(f, first, i-1, ret, offset); |
2119 | } | |
cec736d2 | 2120 | |
de190aef LP |
2121 | enum { |
2122 | TEST_FOUND, | |
2123 | TEST_LEFT, | |
2124 | TEST_RIGHT | |
2125 | }; | |
cec736d2 | 2126 | |
f268980d LP |
2127 | static int generic_array_bisect( |
2128 | JournalFile *f, | |
2129 | uint64_t first, | |
2130 | uint64_t n, | |
2131 | uint64_t needle, | |
2132 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
2133 | direction_t direction, | |
2134 | Object **ret, | |
2135 | uint64_t *offset, | |
2136 | uint64_t *idx) { | |
2137 | ||
2138 | uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1; | |
de190aef LP |
2139 | bool subtract_one = false; |
2140 | Object *o, *array = NULL; | |
2141 | int r; | |
a4bcff5b | 2142 | ChainCacheItem *ci; |
cec736d2 | 2143 | |
de190aef LP |
2144 | assert(f); |
2145 | assert(test_object); | |
cec736d2 | 2146 | |
a4bcff5b | 2147 | /* Start with the first array in the chain */ |
de190aef | 2148 | a = first; |
a4bcff5b | 2149 | |
4743015d | 2150 | ci = ordered_hashmap_get(f->chain_cache, &first); |
96d4d024 | 2151 | if (ci && n > ci->total && ci->begin != 0) { |
a4bcff5b LP |
2152 | /* Ah, we have iterated this bisection array chain |
2153 | * previously! Let's see if we can skip ahead in the | |
2154 | * chain, as far as the last time. But we can't jump | |
2155 | * backwards in the chain, so let's check that | |
2156 | * first. */ | |
2157 | ||
2158 | r = test_object(f, ci->begin, needle); | |
2159 | if (r < 0) | |
2160 | return r; | |
2161 | ||
2162 | if (r == TEST_LEFT) { | |
f268980d | 2163 | /* OK, what we are looking for is right of the |
a4bcff5b LP |
2164 | * begin of this EntryArray, so let's jump |
2165 | * straight to previously cached array in the | |
2166 | * chain */ | |
2167 | ||
2168 | a = ci->array; | |
2169 | n -= ci->total; | |
2170 | t = ci->total; | |
f268980d | 2171 | last_index = ci->last_index; |
a4bcff5b LP |
2172 | } |
2173 | } | |
2174 | ||
de190aef LP |
2175 | while (a > 0) { |
2176 | uint64_t left, right, k, lp; | |
2177 | ||
2178 | r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); | |
cec736d2 LP |
2179 | if (r < 0) |
2180 | return r; | |
2181 | ||
de190aef LP |
2182 | k = journal_file_entry_array_n_items(array); |
2183 | right = MIN(k, n); | |
2184 | if (right <= 0) | |
2185 | return 0; | |
cec736d2 | 2186 | |
de190aef LP |
2187 | i = right - 1; |
2188 | lp = p = le64toh(array->entry_array.items[i]); | |
2189 | if (p <= 0) | |
bee6a291 LP |
2190 | r = -EBADMSG; |
2191 | else | |
2192 | r = test_object(f, p, needle); | |
2193 | if (r == -EBADMSG) { | |
2194 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)"); | |
2195 | n = i; | |
2196 | continue; | |
2197 | } | |
de190aef LP |
2198 | if (r < 0) |
2199 | return r; | |
cec736d2 | 2200 | |
de190aef LP |
2201 | if (r == TEST_FOUND) |
2202 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2203 | ||
2204 | if (r == TEST_RIGHT) { | |
2205 | left = 0; | |
2206 | right -= 1; | |
f268980d LP |
2207 | |
2208 | if (last_index != (uint64_t) -1) { | |
2209 | assert(last_index <= right); | |
2210 | ||
2211 | /* If we cached the last index we | |
2212 | * looked at, let's try to not to jump | |
2213 | * too wildly around and see if we can | |
2214 | * limit the range to look at early to | |
2215 | * the immediate neighbors of the last | |
2216 | * index we looked at. */ | |
2217 | ||
2218 | if (last_index > 0) { | |
2219 | uint64_t x = last_index - 1; | |
2220 | ||
2221 | p = le64toh(array->entry_array.items[x]); | |
2222 | if (p <= 0) | |
2223 | return -EBADMSG; | |
2224 | ||
2225 | r = test_object(f, p, needle); | |
2226 | if (r < 0) | |
2227 | return r; | |
2228 | ||
2229 | if (r == TEST_FOUND) | |
2230 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2231 | ||
2232 | if (r == TEST_RIGHT) | |
2233 | right = x; | |
2234 | else | |
2235 | left = x + 1; | |
2236 | } | |
2237 | ||
2238 | if (last_index < right) { | |
2239 | uint64_t y = last_index + 1; | |
2240 | ||
2241 | p = le64toh(array->entry_array.items[y]); | |
2242 | if (p <= 0) | |
2243 | return -EBADMSG; | |
2244 | ||
2245 | r = test_object(f, p, needle); | |
2246 | if (r < 0) | |
2247 | return r; | |
2248 | ||
2249 | if (r == TEST_FOUND) | |
2250 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2251 | ||
2252 | if (r == TEST_RIGHT) | |
2253 | right = y; | |
2254 | else | |
2255 | left = y + 1; | |
2256 | } | |
f268980d LP |
2257 | } |
2258 | ||
de190aef LP |
2259 | for (;;) { |
2260 | if (left == right) { | |
2261 | if (direction == DIRECTION_UP) | |
2262 | subtract_one = true; | |
2263 | ||
2264 | i = left; | |
2265 | goto found; | |
2266 | } | |
2267 | ||
2268 | assert(left < right); | |
de190aef | 2269 | i = (left + right) / 2; |
f268980d | 2270 | |
de190aef LP |
2271 | p = le64toh(array->entry_array.items[i]); |
2272 | if (p <= 0) | |
bee6a291 LP |
2273 | r = -EBADMSG; |
2274 | else | |
2275 | r = test_object(f, p, needle); | |
2276 | if (r == -EBADMSG) { | |
2277 | log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)"); | |
2278 | right = n = i; | |
2279 | continue; | |
2280 | } | |
de190aef LP |
2281 | if (r < 0) |
2282 | return r; | |
cec736d2 | 2283 | |
de190aef LP |
2284 | if (r == TEST_FOUND) |
2285 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2286 | ||
2287 | if (r == TEST_RIGHT) | |
2288 | right = i; | |
2289 | else | |
2290 | left = i + 1; | |
2291 | } | |
2292 | } | |
2293 | ||
2173cbf8 | 2294 | if (k >= n) { |
cbdca852 LP |
2295 | if (direction == DIRECTION_UP) { |
2296 | i = n; | |
2297 | subtract_one = true; | |
2298 | goto found; | |
2299 | } | |
2300 | ||
cec736d2 | 2301 | return 0; |
cbdca852 | 2302 | } |
cec736d2 | 2303 | |
de190aef LP |
2304 | last_p = lp; |
2305 | ||
2306 | n -= k; | |
2307 | t += k; | |
f268980d | 2308 | last_index = (uint64_t) -1; |
de190aef | 2309 | a = le64toh(array->entry_array.next_entry_array_offset); |
cec736d2 LP |
2310 | } |
2311 | ||
2312 | return 0; | |
de190aef LP |
2313 | |
2314 | found: | |
2315 | if (subtract_one && t == 0 && i == 0) | |
2316 | return 0; | |
2317 | ||
a4bcff5b | 2318 | /* Let's cache this item for the next invocation */ |
af13a6b0 | 2319 | chain_cache_put(f->chain_cache, ci, first, a, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i); |
a4bcff5b | 2320 | |
de190aef LP |
2321 | if (subtract_one && i == 0) |
2322 | p = last_p; | |
2323 | else if (subtract_one) | |
2324 | p = le64toh(array->entry_array.items[i-1]); | |
2325 | else | |
2326 | p = le64toh(array->entry_array.items[i]); | |
2327 | ||
2328 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2329 | if (r < 0) | |
2330 | return r; | |
2331 | ||
2332 | if (ret) | |
2333 | *ret = o; | |
2334 | ||
2335 | if (offset) | |
2336 | *offset = p; | |
2337 | ||
2338 | if (idx) | |
cbdca852 | 2339 | *idx = t + i + (subtract_one ? -1 : 0); |
de190aef LP |
2340 | |
2341 | return 1; | |
cec736d2 LP |
2342 | } |
2343 | ||
f268980d LP |
2344 | static int generic_array_bisect_plus_one( |
2345 | JournalFile *f, | |
2346 | uint64_t extra, | |
2347 | uint64_t first, | |
2348 | uint64_t n, | |
2349 | uint64_t needle, | |
2350 | int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), | |
2351 | direction_t direction, | |
2352 | Object **ret, | |
2353 | uint64_t *offset, | |
2354 | uint64_t *idx) { | |
de190aef | 2355 | |
cec736d2 | 2356 | int r; |
cbdca852 LP |
2357 | bool step_back = false; |
2358 | Object *o; | |
cec736d2 LP |
2359 | |
2360 | assert(f); | |
de190aef | 2361 | assert(test_object); |
cec736d2 | 2362 | |
de190aef LP |
2363 | if (n <= 0) |
2364 | return 0; | |
cec736d2 | 2365 | |
de190aef LP |
2366 | /* This bisects the array in object 'first', but first checks |
2367 | * an extra */ | |
de190aef LP |
2368 | r = test_object(f, extra, needle); |
2369 | if (r < 0) | |
2370 | return r; | |
a536e261 LP |
2371 | |
2372 | if (r == TEST_FOUND) | |
2373 | r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; | |
2374 | ||
cbdca852 LP |
2375 | /* if we are looking with DIRECTION_UP then we need to first |
2376 | see if in the actual array there is a matching entry, and | |
2377 | return the last one of that. But if there isn't any we need | |
2378 | to return this one. Hence remember this, and return it | |
2379 | below. */ | |
2380 | if (r == TEST_LEFT) | |
2381 | step_back = direction == DIRECTION_UP; | |
de190aef | 2382 | |
cbdca852 LP |
2383 | if (r == TEST_RIGHT) { |
2384 | if (direction == DIRECTION_DOWN) | |
2385 | goto found; | |
2386 | else | |
2387 | return 0; | |
a536e261 | 2388 | } |
cec736d2 | 2389 | |
de190aef LP |
2390 | r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx); |
2391 | ||
cbdca852 LP |
2392 | if (r == 0 && step_back) |
2393 | goto found; | |
2394 | ||
ecf68b1d | 2395 | if (r > 0 && idx) |
313cefa1 | 2396 | (*idx)++; |
de190aef LP |
2397 | |
2398 | return r; | |
cbdca852 LP |
2399 | |
2400 | found: | |
2401 | r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o); | |
2402 | if (r < 0) | |
2403 | return r; | |
2404 | ||
2405 | if (ret) | |
2406 | *ret = o; | |
2407 | ||
2408 | if (offset) | |
2409 | *offset = extra; | |
2410 | ||
2411 | if (idx) | |
2412 | *idx = 0; | |
2413 | ||
2414 | return 1; | |
2415 | } | |
2416 | ||
44a6b1b6 | 2417 | _pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { |
cbdca852 LP |
2418 | assert(f); |
2419 | assert(p > 0); | |
2420 | ||
2421 | if (p == needle) | |
2422 | return TEST_FOUND; | |
2423 | else if (p < needle) | |
2424 | return TEST_LEFT; | |
2425 | else | |
2426 | return TEST_RIGHT; | |
2427 | } | |
2428 | ||
de190aef LP |
2429 | static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { |
2430 | Object *o; | |
2431 | int r; | |
2432 | ||
2433 | assert(f); | |
2434 | assert(p > 0); | |
2435 | ||
2436 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
cec736d2 LP |
2437 | if (r < 0) |
2438 | return r; | |
2439 | ||
de190aef LP |
2440 | if (le64toh(o->entry.seqnum) == needle) |
2441 | return TEST_FOUND; | |
2442 | else if (le64toh(o->entry.seqnum) < needle) | |
2443 | return TEST_LEFT; | |
2444 | else | |
2445 | return TEST_RIGHT; | |
2446 | } | |
cec736d2 | 2447 | |
de190aef LP |
2448 | int journal_file_move_to_entry_by_seqnum( |
2449 | JournalFile *f, | |
2450 | uint64_t seqnum, | |
2451 | direction_t direction, | |
2452 | Object **ret, | |
2453 | uint64_t *offset) { | |
c88cc6af VC |
2454 | assert(f); |
2455 | assert(f->header); | |
de190aef LP |
2456 | |
2457 | return generic_array_bisect(f, | |
2458 | le64toh(f->header->entry_array_offset), | |
2459 | le64toh(f->header->n_entries), | |
2460 | seqnum, | |
2461 | test_object_seqnum, | |
2462 | direction, | |
2463 | ret, offset, NULL); | |
2464 | } | |
cec736d2 | 2465 | |
de190aef LP |
2466 | static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { |
2467 | Object *o; | |
2468 | int r; | |
2469 | ||
2470 | assert(f); | |
2471 | assert(p > 0); | |
2472 | ||
2473 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2474 | if (r < 0) | |
2475 | return r; | |
2476 | ||
2477 | if (le64toh(o->entry.realtime) == needle) | |
2478 | return TEST_FOUND; | |
2479 | else if (le64toh(o->entry.realtime) < needle) | |
2480 | return TEST_LEFT; | |
2481 | else | |
2482 | return TEST_RIGHT; | |
cec736d2 LP |
2483 | } |
2484 | ||
de190aef LP |
2485 | int journal_file_move_to_entry_by_realtime( |
2486 | JournalFile *f, | |
2487 | uint64_t realtime, | |
2488 | direction_t direction, | |
2489 | Object **ret, | |
2490 | uint64_t *offset) { | |
c88cc6af VC |
2491 | assert(f); |
2492 | assert(f->header); | |
de190aef LP |
2493 | |
2494 | return generic_array_bisect(f, | |
2495 | le64toh(f->header->entry_array_offset), | |
2496 | le64toh(f->header->n_entries), | |
2497 | realtime, | |
2498 | test_object_realtime, | |
2499 | direction, | |
2500 | ret, offset, NULL); | |
2501 | } | |
2502 | ||
2503 | static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { | |
2504 | Object *o; | |
2505 | int r; | |
2506 | ||
2507 | assert(f); | |
2508 | assert(p > 0); | |
2509 | ||
2510 | r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); | |
2511 | if (r < 0) | |
2512 | return r; | |
2513 | ||
2514 | if (le64toh(o->entry.monotonic) == needle) | |
2515 | return TEST_FOUND; | |
2516 | else if (le64toh(o->entry.monotonic) < needle) | |
2517 | return TEST_LEFT; | |
2518 | else | |
2519 | return TEST_RIGHT; | |
2520 | } | |
2521 | ||
2a560338 | 2522 | static int find_data_object_by_boot_id( |
47838ab3 ZJS |
2523 | JournalFile *f, |
2524 | sd_id128_t boot_id, | |
2525 | Object **o, | |
2526 | uint64_t *b) { | |
2a560338 | 2527 | |
fbd0b64f | 2528 | char t[STRLEN("_BOOT_ID=") + 32 + 1] = "_BOOT_ID="; |
47838ab3 ZJS |
2529 | |
2530 | sd_id128_to_string(boot_id, t + 9); | |
2531 | return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b); | |
2532 | } | |
2533 | ||
de190aef LP |
2534 | int journal_file_move_to_entry_by_monotonic( |
2535 | JournalFile *f, | |
2536 | sd_id128_t boot_id, | |
2537 | uint64_t monotonic, | |
2538 | direction_t direction, | |
2539 | Object **ret, | |
2540 | uint64_t *offset) { | |
2541 | ||
de190aef LP |
2542 | Object *o; |
2543 | int r; | |
2544 | ||
cbdca852 | 2545 | assert(f); |
de190aef | 2546 | |
47838ab3 | 2547 | r = find_data_object_by_boot_id(f, boot_id, &o, NULL); |
de190aef LP |
2548 | if (r < 0) |
2549 | return r; | |
cbdca852 | 2550 | if (r == 0) |
de190aef LP |
2551 | return -ENOENT; |
2552 | ||
2553 | return generic_array_bisect_plus_one(f, | |
2554 | le64toh(o->data.entry_offset), | |
2555 | le64toh(o->data.entry_array_offset), | |
2556 | le64toh(o->data.n_entries), | |
2557 | monotonic, | |
2558 | test_object_monotonic, | |
2559 | direction, | |
2560 | ret, offset, NULL); | |
2561 | } | |
2562 | ||
1fc605b0 | 2563 | void journal_file_reset_location(JournalFile *f) { |
6573ef05 | 2564 | f->location_type = LOCATION_HEAD; |
1fc605b0 | 2565 | f->current_offset = 0; |
6573ef05 MS |
2566 | f->current_seqnum = 0; |
2567 | f->current_realtime = 0; | |
2568 | f->current_monotonic = 0; | |
2569 | zero(f->current_boot_id); | |
2570 | f->current_xor_hash = 0; | |
2571 | } | |
2572 | ||
950c07d4 | 2573 | void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) { |
6573ef05 MS |
2574 | f->location_type = LOCATION_SEEK; |
2575 | f->current_offset = offset; | |
2576 | f->current_seqnum = le64toh(o->entry.seqnum); | |
2577 | f->current_realtime = le64toh(o->entry.realtime); | |
2578 | f->current_monotonic = le64toh(o->entry.monotonic); | |
2579 | f->current_boot_id = o->entry.boot_id; | |
2580 | f->current_xor_hash = le64toh(o->entry.xor_hash); | |
1fc605b0 MS |
2581 | } |
2582 | ||
d8ae66d7 | 2583 | int journal_file_compare_locations(JournalFile *af, JournalFile *bf) { |
90c88092 YW |
2584 | int r; |
2585 | ||
d8ae66d7 | 2586 | assert(af); |
c88cc6af | 2587 | assert(af->header); |
d8ae66d7 | 2588 | assert(bf); |
c88cc6af | 2589 | assert(bf->header); |
d8ae66d7 MS |
2590 | assert(af->location_type == LOCATION_SEEK); |
2591 | assert(bf->location_type == LOCATION_SEEK); | |
2592 | ||
2593 | /* If contents and timestamps match, these entries are | |
2594 | * identical, even if the seqnum does not match */ | |
2595 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) && | |
2596 | af->current_monotonic == bf->current_monotonic && | |
2597 | af->current_realtime == bf->current_realtime && | |
2598 | af->current_xor_hash == bf->current_xor_hash) | |
2599 | return 0; | |
2600 | ||
2601 | if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) { | |
2602 | ||
2603 | /* If this is from the same seqnum source, compare | |
2604 | * seqnums */ | |
90c88092 YW |
2605 | r = CMP(af->current_seqnum, bf->current_seqnum); |
2606 | if (r != 0) | |
2607 | return r; | |
d8ae66d7 MS |
2608 | |
2609 | /* Wow! This is weird, different data but the same | |
2610 | * seqnums? Something is borked, but let's make the | |
2611 | * best of it and compare by time. */ | |
2612 | } | |
2613 | ||
2614 | if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) { | |
2615 | ||
2616 | /* If the boot id matches, compare monotonic time */ | |
90c88092 YW |
2617 | r = CMP(af->current_monotonic, bf->current_monotonic); |
2618 | if (r != 0) | |
2619 | return r; | |
d8ae66d7 MS |
2620 | } |
2621 | ||
2622 | /* Otherwise, compare UTC time */ | |
90c88092 YW |
2623 | r = CMP(af->current_realtime, bf->current_realtime); |
2624 | if (r != 0) | |
2625 | return r; | |
d8ae66d7 MS |
2626 | |
2627 | /* Finally, compare by contents */ | |
6dd91b36 | 2628 | return CMP(af->current_xor_hash, bf->current_xor_hash); |
d8ae66d7 MS |
2629 | } |
2630 | ||
aa598ba5 LP |
2631 | static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) { |
2632 | ||
2633 | /* Increase or decrease the specified index, in the right direction. */ | |
2634 | ||
2635 | if (direction == DIRECTION_DOWN) { | |
2636 | if (*i >= n - 1) | |
2637 | return 0; | |
2638 | ||
2639 | (*i) ++; | |
2640 | } else { | |
2641 | if (*i <= 0) | |
2642 | return 0; | |
2643 | ||
2644 | (*i) --; | |
2645 | } | |
2646 | ||
2647 | return 1; | |
2648 | } | |
2649 | ||
b6da4ed0 LP |
2650 | static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) { |
2651 | ||
2652 | /* Consider it an error if any of the two offsets is uninitialized */ | |
2653 | if (old_offset == 0 || new_offset == 0) | |
2654 | return false; | |
2655 | ||
2656 | /* If we go down, the new offset must be larger than the old one. */ | |
2657 | return direction == DIRECTION_DOWN ? | |
2658 | new_offset > old_offset : | |
2659 | new_offset < old_offset; | |
2660 | } | |
2661 | ||
de190aef LP |
2662 | int journal_file_next_entry( |
2663 | JournalFile *f, | |
f534928a | 2664 | uint64_t p, |
de190aef LP |
2665 | direction_t direction, |
2666 | Object **ret, uint64_t *offset) { | |
2667 | ||
fb099c8d | 2668 | uint64_t i, n, ofs; |
cec736d2 LP |
2669 | int r; |
2670 | ||
2671 | assert(f); | |
c88cc6af | 2672 | assert(f->header); |
de190aef LP |
2673 | |
2674 | n = le64toh(f->header->n_entries); | |
2675 | if (n <= 0) | |
2676 | return 0; | |
cec736d2 | 2677 | |
f534928a | 2678 | if (p == 0) |
de190aef | 2679 | i = direction == DIRECTION_DOWN ? 0 : n - 1; |
cec736d2 | 2680 | else { |
de190aef LP |
2681 | r = generic_array_bisect(f, |
2682 | le64toh(f->header->entry_array_offset), | |
2683 | le64toh(f->header->n_entries), | |
2684 | p, | |
2685 | test_object_offset, | |
2686 | DIRECTION_DOWN, | |
2687 | NULL, NULL, | |
2688 | &i); | |
2689 | if (r <= 0) | |
2690 | return r; | |
2691 | ||
aa598ba5 LP |
2692 | r = bump_array_index(&i, direction, n); |
2693 | if (r <= 0) | |
2694 | return r; | |
cec736d2 LP |
2695 | } |
2696 | ||
de190aef | 2697 | /* And jump to it */ |
989793d3 LP |
2698 | for (;;) { |
2699 | r = generic_array_get(f, | |
2700 | le64toh(f->header->entry_array_offset), | |
2701 | i, | |
2702 | ret, &ofs); | |
2703 | if (r > 0) | |
2704 | break; | |
2705 | if (r != -EBADMSG) | |
2706 | return r; | |
2707 | ||
2708 | /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if | |
2709 | * the next one might work for us instead. */ | |
2710 | log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i); | |
2711 | ||
2712 | r = bump_array_index(&i, direction, n); | |
2713 | if (r <= 0) | |
2714 | return r; | |
caeab8f6 | 2715 | } |
fb099c8d | 2716 | |
b6da4ed0 | 2717 | /* Ensure our array is properly ordered. */ |
baaa35ad ZJS |
2718 | if (p > 0 && !check_properly_ordered(ofs, p, direction)) |
2719 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2720 | "%s: entry array not properly ordered at entry %" PRIu64, | |
2721 | f->path, i); | |
fb099c8d ZJS |
2722 | |
2723 | if (offset) | |
2724 | *offset = ofs; | |
2725 | ||
2726 | return 1; | |
de190aef | 2727 | } |
cec736d2 | 2728 | |
de190aef LP |
2729 | int journal_file_next_entry_for_data( |
2730 | JournalFile *f, | |
2731 | Object *o, uint64_t p, | |
2732 | uint64_t data_offset, | |
2733 | direction_t direction, | |
2734 | Object **ret, uint64_t *offset) { | |
2735 | ||
ded5034e | 2736 | uint64_t i, n, ofs; |
de190aef | 2737 | Object *d; |
989793d3 | 2738 | int r; |
cec736d2 LP |
2739 | |
2740 | assert(f); | |
de190aef | 2741 | assert(p > 0 || !o); |
cec736d2 | 2742 | |
de190aef | 2743 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
466ccd92 | 2744 | if (r < 0) |
de190aef | 2745 | return r; |
cec736d2 | 2746 | |
de190aef LP |
2747 | n = le64toh(d->data.n_entries); |
2748 | if (n <= 0) | |
2749 | return n; | |
cec736d2 | 2750 | |
de190aef LP |
2751 | if (!o) |
2752 | i = direction == DIRECTION_DOWN ? 0 : n - 1; | |
2753 | else { | |
2754 | if (o->object.type != OBJECT_ENTRY) | |
2755 | return -EINVAL; | |
cec736d2 | 2756 | |
de190aef LP |
2757 | r = generic_array_bisect_plus_one(f, |
2758 | le64toh(d->data.entry_offset), | |
2759 | le64toh(d->data.entry_array_offset), | |
2760 | le64toh(d->data.n_entries), | |
2761 | p, | |
2762 | test_object_offset, | |
2763 | DIRECTION_DOWN, | |
2764 | NULL, NULL, | |
2765 | &i); | |
2766 | ||
2767 | if (r <= 0) | |
cec736d2 LP |
2768 | return r; |
2769 | ||
aa598ba5 LP |
2770 | r = bump_array_index(&i, direction, n); |
2771 | if (r <= 0) | |
2772 | return r; | |
de190aef | 2773 | } |
cec736d2 | 2774 | |
989793d3 LP |
2775 | for (;;) { |
2776 | r = generic_array_get_plus_one(f, | |
2777 | le64toh(d->data.entry_offset), | |
2778 | le64toh(d->data.entry_array_offset), | |
2779 | i, | |
2780 | ret, &ofs); | |
2781 | if (r > 0) | |
2782 | break; | |
2783 | if (r != -EBADMSG) | |
2784 | return r; | |
2785 | ||
2786 | log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i); | |
2787 | ||
2788 | r = bump_array_index(&i, direction, n); | |
2789 | if (r <= 0) | |
2790 | return r; | |
2791 | } | |
ded5034e LP |
2792 | |
2793 | /* Ensure our array is properly ordered. */ | |
baaa35ad ZJS |
2794 | if (p > 0 && check_properly_ordered(ofs, p, direction)) |
2795 | return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), | |
2796 | "%s data entry array not properly ordered at entry %" PRIu64, | |
2797 | f->path, i); | |
ded5034e LP |
2798 | |
2799 | if (offset) | |
2800 | *offset = ofs; | |
2801 | ||
2802 | return 1; | |
de190aef | 2803 | } |
cec736d2 | 2804 | |
cbdca852 LP |
2805 | int journal_file_move_to_entry_by_offset_for_data( |
2806 | JournalFile *f, | |
2807 | uint64_t data_offset, | |
2808 | uint64_t p, | |
2809 | direction_t direction, | |
2810 | Object **ret, uint64_t *offset) { | |
2811 | ||
2812 | int r; | |
2813 | Object *d; | |
2814 | ||
2815 | assert(f); | |
2816 | ||
2817 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
2818 | if (r < 0) | |
2819 | return r; | |
2820 | ||
2821 | return generic_array_bisect_plus_one(f, | |
2822 | le64toh(d->data.entry_offset), | |
2823 | le64toh(d->data.entry_array_offset), | |
2824 | le64toh(d->data.n_entries), | |
2825 | p, | |
2826 | test_object_offset, | |
2827 | direction, | |
2828 | ret, offset, NULL); | |
2829 | } | |
2830 | ||
2831 | int journal_file_move_to_entry_by_monotonic_for_data( | |
2832 | JournalFile *f, | |
2833 | uint64_t data_offset, | |
2834 | sd_id128_t boot_id, | |
2835 | uint64_t monotonic, | |
2836 | direction_t direction, | |
2837 | Object **ret, uint64_t *offset) { | |
2838 | ||
cbdca852 LP |
2839 | Object *o, *d; |
2840 | int r; | |
2841 | uint64_t b, z; | |
2842 | ||
2843 | assert(f); | |
2844 | ||
2845 | /* First, seek by time */ | |
47838ab3 | 2846 | r = find_data_object_by_boot_id(f, boot_id, &o, &b); |
cbdca852 LP |
2847 | if (r < 0) |
2848 | return r; | |
2849 | if (r == 0) | |
2850 | return -ENOENT; | |
2851 | ||
2852 | r = generic_array_bisect_plus_one(f, | |
2853 | le64toh(o->data.entry_offset), | |
2854 | le64toh(o->data.entry_array_offset), | |
2855 | le64toh(o->data.n_entries), | |
2856 | monotonic, | |
2857 | test_object_monotonic, | |
2858 | direction, | |
2859 | NULL, &z, NULL); | |
2860 | if (r <= 0) | |
2861 | return r; | |
2862 | ||
2863 | /* And now, continue seeking until we find an entry that | |
2864 | * exists in both bisection arrays */ | |
2865 | ||
2866 | for (;;) { | |
2867 | Object *qo; | |
2868 | uint64_t p, q; | |
2869 | ||
2870 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); | |
2871 | if (r < 0) | |
2872 | return r; | |
2873 | ||
2874 | r = generic_array_bisect_plus_one(f, | |
2875 | le64toh(d->data.entry_offset), | |
2876 | le64toh(d->data.entry_array_offset), | |
2877 | le64toh(d->data.n_entries), | |
2878 | z, | |
2879 | test_object_offset, | |
2880 | direction, | |
2881 | NULL, &p, NULL); | |
2882 | if (r <= 0) | |
2883 | return r; | |
2884 | ||
2885 | r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); | |
2886 | if (r < 0) | |
2887 | return r; | |
2888 | ||
2889 | r = generic_array_bisect_plus_one(f, | |
2890 | le64toh(o->data.entry_offset), | |
2891 | le64toh(o->data.entry_array_offset), | |
2892 | le64toh(o->data.n_entries), | |
2893 | p, | |
2894 | test_object_offset, | |
2895 | direction, | |
2896 | &qo, &q, NULL); | |
2897 | ||
2898 | if (r <= 0) | |
2899 | return r; | |
2900 | ||
2901 | if (p == q) { | |
2902 | if (ret) | |
2903 | *ret = qo; | |
2904 | if (offset) | |
2905 | *offset = q; | |
2906 | ||
2907 | return 1; | |
2908 | } | |
2909 | ||
2910 | z = q; | |
2911 | } | |
cbdca852 LP |
2912 | } |
2913 | ||
de190aef LP |
2914 | int journal_file_move_to_entry_by_seqnum_for_data( |
2915 | JournalFile *f, | |
2916 | uint64_t data_offset, | |
2917 | uint64_t seqnum, | |
2918 | direction_t direction, | |
2919 | Object **ret, uint64_t *offset) { | |
cec736d2 | 2920 | |
de190aef LP |
2921 | Object *d; |
2922 | int r; | |
cec736d2 | 2923 | |
91a31dde LP |
2924 | assert(f); |
2925 | ||
de190aef | 2926 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 2927 | if (r < 0) |
de190aef | 2928 | return r; |
cec736d2 | 2929 | |
de190aef LP |
2930 | return generic_array_bisect_plus_one(f, |
2931 | le64toh(d->data.entry_offset), | |
2932 | le64toh(d->data.entry_array_offset), | |
2933 | le64toh(d->data.n_entries), | |
2934 | seqnum, | |
2935 | test_object_seqnum, | |
2936 | direction, | |
2937 | ret, offset, NULL); | |
2938 | } | |
cec736d2 | 2939 | |
de190aef LP |
2940 | int journal_file_move_to_entry_by_realtime_for_data( |
2941 | JournalFile *f, | |
2942 | uint64_t data_offset, | |
2943 | uint64_t realtime, | |
2944 | direction_t direction, | |
2945 | Object **ret, uint64_t *offset) { | |
2946 | ||
2947 | Object *d; | |
2948 | int r; | |
2949 | ||
91a31dde LP |
2950 | assert(f); |
2951 | ||
de190aef | 2952 | r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d); |
91a31dde | 2953 | if (r < 0) |
de190aef LP |
2954 | return r; |
2955 | ||
2956 | return generic_array_bisect_plus_one(f, | |
2957 | le64toh(d->data.entry_offset), | |
2958 | le64toh(d->data.entry_array_offset), | |
2959 | le64toh(d->data.n_entries), | |
2960 | realtime, | |
2961 | test_object_realtime, | |
2962 | direction, | |
2963 | ret, offset, NULL); | |
cec736d2 LP |
2964 | } |
2965 | ||
0284adc6 | 2966 | void journal_file_dump(JournalFile *f) { |
7560fffc | 2967 | Object *o; |
7560fffc | 2968 | int r; |
0284adc6 | 2969 | uint64_t p; |
7560fffc LP |
2970 | |
2971 | assert(f); | |
c88cc6af | 2972 | assert(f->header); |
7560fffc | 2973 | |
0284adc6 | 2974 | journal_file_print_header(f); |
7560fffc | 2975 | |
0284adc6 LP |
2976 | p = le64toh(f->header->header_size); |
2977 | while (p != 0) { | |
d05089d8 | 2978 | r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); |
0284adc6 LP |
2979 | if (r < 0) |
2980 | goto fail; | |
7560fffc | 2981 | |
0284adc6 | 2982 | switch (o->object.type) { |
d98cc1f2 | 2983 | |
0284adc6 LP |
2984 | case OBJECT_UNUSED: |
2985 | printf("Type: OBJECT_UNUSED\n"); | |
2986 | break; | |
d98cc1f2 | 2987 | |
0284adc6 LP |
2988 | case OBJECT_DATA: |
2989 | printf("Type: OBJECT_DATA\n"); | |
2990 | break; | |
7560fffc | 2991 | |
3c1668da LP |
2992 | case OBJECT_FIELD: |
2993 | printf("Type: OBJECT_FIELD\n"); | |
2994 | break; | |
2995 | ||
0284adc6 | 2996 | case OBJECT_ENTRY: |
507f22bd ZJS |
2997 | printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n", |
2998 | le64toh(o->entry.seqnum), | |
2999 | le64toh(o->entry.monotonic), | |
3000 | le64toh(o->entry.realtime)); | |
0284adc6 | 3001 | break; |
7560fffc | 3002 | |
0284adc6 LP |
3003 | case OBJECT_FIELD_HASH_TABLE: |
3004 | printf("Type: OBJECT_FIELD_HASH_TABLE\n"); | |
3005 | break; | |
7560fffc | 3006 | |
0284adc6 LP |
3007 | case OBJECT_DATA_HASH_TABLE: |
3008 | printf("Type: OBJECT_DATA_HASH_TABLE\n"); | |
3009 | break; | |
7560fffc | 3010 | |
0284adc6 LP |
3011 | case OBJECT_ENTRY_ARRAY: |
3012 | printf("Type: OBJECT_ENTRY_ARRAY\n"); | |
3013 | break; | |
7560fffc | 3014 | |
0284adc6 | 3015 | case OBJECT_TAG: |
507f22bd ZJS |
3016 | printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n", |
3017 | le64toh(o->tag.seqnum), | |
3018 | le64toh(o->tag.epoch)); | |
0284adc6 | 3019 | break; |
3c1668da LP |
3020 | |
3021 | default: | |
8facc349 | 3022 | printf("Type: unknown (%i)\n", o->object.type); |
3c1668da | 3023 | break; |
0284adc6 | 3024 | } |
7560fffc | 3025 | |
d89c8fdf ZJS |
3026 | if (o->object.flags & OBJECT_COMPRESSION_MASK) |
3027 | printf("Flags: %s\n", | |
3028 | object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK)); | |
7560fffc | 3029 | |
0284adc6 LP |
3030 | if (p == le64toh(f->header->tail_object_offset)) |
3031 | p = 0; | |
3032 | else | |
3033 | p = p + ALIGN64(le64toh(o->object.size)); | |
3034 | } | |
7560fffc | 3035 | |
0284adc6 LP |
3036 | return; |
3037 | fail: | |
3038 | log_error("File corrupt"); | |
7560fffc LP |
3039 | } |
3040 | ||
718fe4b1 ZJS |
3041 | static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) { |
3042 | const char *x; | |
3043 | ||
3044 | x = format_timestamp(buf, l, t); | |
3045 | if (x) | |
3046 | return x; | |
3047 | return " --- "; | |
3048 | } | |
3049 | ||
0284adc6 | 3050 | void journal_file_print_header(JournalFile *f) { |
2765b7bb | 3051 | char a[33], b[33], c[33], d[33]; |
ed375beb | 3052 | char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX]; |
a1a03e30 LP |
3053 | struct stat st; |
3054 | char bytes[FORMAT_BYTES_MAX]; | |
7560fffc LP |
3055 | |
3056 | assert(f); | |
c88cc6af | 3057 | assert(f->header); |
7560fffc | 3058 | |
0284adc6 LP |
3059 | printf("File Path: %s\n" |
3060 | "File ID: %s\n" | |
3061 | "Machine ID: %s\n" | |
3062 | "Boot ID: %s\n" | |
3063 | "Sequential Number ID: %s\n" | |
3064 | "State: %s\n" | |
3065 | "Compatible Flags:%s%s\n" | |
d89c8fdf | 3066 | "Incompatible Flags:%s%s%s\n" |
507f22bd ZJS |
3067 | "Header size: %"PRIu64"\n" |
3068 | "Arena size: %"PRIu64"\n" | |
3069 | "Data Hash Table Size: %"PRIu64"\n" | |
3070 | "Field Hash Table Size: %"PRIu64"\n" | |
0284adc6 | 3071 | "Rotate Suggested: %s\n" |
0808b92f LP |
3072 | "Head Sequential Number: %"PRIu64" (%"PRIx64")\n" |
3073 | "Tail Sequential Number: %"PRIu64" (%"PRIx64")\n" | |
3074 | "Head Realtime Timestamp: %s (%"PRIx64")\n" | |
3075 | "Tail Realtime Timestamp: %s (%"PRIx64")\n" | |
3076 | "Tail Monotonic Timestamp: %s (%"PRIx64")\n" | |
507f22bd ZJS |
3077 | "Objects: %"PRIu64"\n" |
3078 | "Entry Objects: %"PRIu64"\n", | |
0284adc6 LP |
3079 | f->path, |
3080 | sd_id128_to_string(f->header->file_id, a), | |
3081 | sd_id128_to_string(f->header->machine_id, b), | |
3082 | sd_id128_to_string(f->header->boot_id, c), | |
2765b7bb | 3083 | sd_id128_to_string(f->header->seqnum_id, d), |
3223f44f LP |
3084 | f->header->state == STATE_OFFLINE ? "OFFLINE" : |
3085 | f->header->state == STATE_ONLINE ? "ONLINE" : | |
3086 | f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN", | |
8088cbd3 | 3087 | JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "", |
d89c8fdf ZJS |
3088 | (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "", |
3089 | JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "", | |
3090 | JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "", | |
3091 | (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "", | |
507f22bd ZJS |
3092 | le64toh(f->header->header_size), |
3093 | le64toh(f->header->arena_size), | |
3094 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
3095 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem), | |
fb0951b0 | 3096 | yes_no(journal_file_rotate_suggested(f, 0)), |
0808b92f LP |
3097 | le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum), |
3098 | le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum), | |
3099 | format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime), | |
3100 | format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime), | |
3101 | format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic), | |
507f22bd ZJS |
3102 | le64toh(f->header->n_objects), |
3103 | le64toh(f->header->n_entries)); | |
7560fffc | 3104 | |
0284adc6 | 3105 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) |
507f22bd | 3106 | printf("Data Objects: %"PRIu64"\n" |
0284adc6 | 3107 | "Data Hash Table Fill: %.1f%%\n", |
507f22bd | 3108 | le64toh(f->header->n_data), |
0284adc6 | 3109 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)))); |
7560fffc | 3110 | |
0284adc6 | 3111 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) |
507f22bd | 3112 | printf("Field Objects: %"PRIu64"\n" |
0284adc6 | 3113 | "Field Hash Table Fill: %.1f%%\n", |
507f22bd | 3114 | le64toh(f->header->n_fields), |
0284adc6 | 3115 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); |
3223f44f LP |
3116 | |
3117 | if (JOURNAL_HEADER_CONTAINS(f->header, n_tags)) | |
507f22bd ZJS |
3118 | printf("Tag Objects: %"PRIu64"\n", |
3119 | le64toh(f->header->n_tags)); | |
3223f44f | 3120 | if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) |
507f22bd ZJS |
3121 | printf("Entry Array Objects: %"PRIu64"\n", |
3122 | le64toh(f->header->n_entry_arrays)); | |
a1a03e30 LP |
3123 | |
3124 | if (fstat(f->fd, &st) >= 0) | |
59f448cf | 3125 | printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (uint64_t) st.st_blocks * 512ULL)); |
7560fffc LP |
3126 | } |
3127 | ||
fc68c929 LP |
3128 | static int journal_file_warn_btrfs(JournalFile *f) { |
3129 | unsigned attrs; | |
3130 | int r; | |
3131 | ||
3132 | assert(f); | |
3133 | ||
3134 | /* Before we write anything, check if the COW logic is turned | |
3135 | * off on btrfs. Given our write pattern that is quite | |
3136 | * unfriendly to COW file systems this should greatly improve | |
3137 | * performance on COW file systems, such as btrfs, at the | |
3138 | * expense of data integrity features (which shouldn't be too | |
3139 | * bad, given that we do our own checksumming). */ | |
3140 | ||
3141 | r = btrfs_is_filesystem(f->fd); | |
3142 | if (r < 0) | |
3143 | return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m"); | |
3144 | if (!r) | |
3145 | return 0; | |
3146 | ||
3147 | r = read_attr_fd(f->fd, &attrs); | |
3148 | if (r < 0) | |
3149 | return log_warning_errno(r, "Failed to read file attributes: %m"); | |
3150 | ||
3151 | if (attrs & FS_NOCOW_FL) { | |
3152 | log_debug("Detected btrfs file system with copy-on-write disabled, all is good."); | |
3153 | return 0; | |
3154 | } | |
3155 | ||
3156 | log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. " | |
3157 | "This is likely to slow down journal access substantially, please consider turning " | |
3158 | "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path); | |
3159 | ||
3160 | return 1; | |
3161 | } | |
3162 | ||
0284adc6 | 3163 | int journal_file_open( |
5d1ce257 | 3164 | int fd, |
0284adc6 LP |
3165 | const char *fname, |
3166 | int flags, | |
3167 | mode_t mode, | |
3168 | bool compress, | |
57850536 | 3169 | uint64_t compress_threshold_bytes, |
baed47c3 | 3170 | bool seal, |
0284adc6 LP |
3171 | JournalMetrics *metrics, |
3172 | MMapCache *mmap_cache, | |
b58c888f | 3173 | Set *deferred_closes, |
0284adc6 LP |
3174 | JournalFile *template, |
3175 | JournalFile **ret) { | |
7560fffc | 3176 | |
fa6ac760 | 3177 | bool newly_created = false; |
0284adc6 | 3178 | JournalFile *f; |
fa6ac760 | 3179 | void *h; |
0284adc6 | 3180 | int r; |
57850536 | 3181 | char bytes[FORMAT_BYTES_MAX]; |
7560fffc | 3182 | |
0559d3a5 | 3183 | assert(ret); |
5d1ce257 | 3184 | assert(fd >= 0 || fname); |
7560fffc | 3185 | |
ec2ce0c5 | 3186 | if (!IN_SET((flags & O_ACCMODE), O_RDONLY, O_RDWR)) |
0284adc6 | 3187 | return -EINVAL; |
7560fffc | 3188 | |
6eda13d3 LP |
3189 | if (fname && (flags & O_CREAT) && !endswith(fname, ".journal")) |
3190 | return -EINVAL; | |
7560fffc | 3191 | |
971b52c4 | 3192 | f = new(JournalFile, 1); |
0284adc6 LP |
3193 | if (!f) |
3194 | return -ENOMEM; | |
7560fffc | 3195 | |
971b52c4 LP |
3196 | *f = (JournalFile) { |
3197 | .fd = fd, | |
3198 | .mode = mode, | |
3199 | ||
3200 | .flags = flags, | |
3201 | .prot = prot_from_flags(flags), | |
3202 | .writable = (flags & O_ACCMODE) != O_RDONLY, | |
7560fffc | 3203 | |
349cc4a5 | 3204 | #if HAVE_LZ4 |
971b52c4 | 3205 | .compress_lz4 = compress, |
349cc4a5 | 3206 | #elif HAVE_XZ |
971b52c4 | 3207 | .compress_xz = compress, |
48b61739 | 3208 | #endif |
971b52c4 LP |
3209 | .compress_threshold_bytes = compress_threshold_bytes == (uint64_t) -1 ? |
3210 | DEFAULT_COMPRESS_THRESHOLD : | |
3211 | MAX(MIN_COMPRESS_THRESHOLD, compress_threshold_bytes), | |
349cc4a5 | 3212 | #if HAVE_GCRYPT |
971b52c4 | 3213 | .seal = seal, |
49a32d43 | 3214 | #endif |
971b52c4 | 3215 | }; |
7560fffc | 3216 | |
57850536 AG |
3217 | log_debug("Journal effective settings seal=%s compress=%s compress_threshold_bytes=%s", |
3218 | yes_no(f->seal), yes_no(JOURNAL_FILE_COMPRESS(f)), | |
3219 | format_bytes(bytes, sizeof(bytes), f->compress_threshold_bytes)); | |
3220 | ||
0284adc6 LP |
3221 | if (mmap_cache) |
3222 | f->mmap = mmap_cache_ref(mmap_cache); | |
3223 | else { | |
84168d80 | 3224 | f->mmap = mmap_cache_new(); |
0284adc6 LP |
3225 | if (!f->mmap) { |
3226 | r = -ENOMEM; | |
3227 | goto fail; | |
3228 | } | |
3229 | } | |
7560fffc | 3230 | |
7645c77b | 3231 | if (fname) { |
5d1ce257 | 3232 | f->path = strdup(fname); |
7645c77b ZJS |
3233 | if (!f->path) { |
3234 | r = -ENOMEM; | |
3235 | goto fail; | |
3236 | } | |
3237 | } else { | |
817b1c5b LP |
3238 | assert(fd >= 0); |
3239 | ||
7645c77b ZJS |
3240 | /* If we don't know the path, fill in something explanatory and vaguely useful */ |
3241 | if (asprintf(&f->path, "/proc/self/%i", fd) < 0) { | |
3242 | r = -ENOMEM; | |
3243 | goto fail; | |
3244 | } | |
0284adc6 | 3245 | } |
7560fffc | 3246 | |
4743015d | 3247 | f->chain_cache = ordered_hashmap_new(&uint64_hash_ops); |
a4bcff5b LP |
3248 | if (!f->chain_cache) { |
3249 | r = -ENOMEM; | |
3250 | goto fail; | |
3251 | } | |
3252 | ||
0284adc6 | 3253 | if (f->fd < 0) { |
817b1c5b LP |
3254 | /* We pass O_NONBLOCK here, so that in case somebody pointed us to some character device node or FIFO |
3255 | * or so, we likely fail quickly than block for long. For regular files O_NONBLOCK has no effect, hence | |
3256 | * it doesn't hurt in that case. */ | |
3257 | ||
3258 | f->fd = open(f->path, f->flags|O_CLOEXEC|O_NONBLOCK, f->mode); | |
5d1ce257 LP |
3259 | if (f->fd < 0) { |
3260 | r = -errno; | |
3261 | goto fail; | |
3262 | } | |
3263 | ||
3264 | /* fds we opened here by us should also be closed by us. */ | |
3265 | f->close_fd = true; | |
817b1c5b LP |
3266 | |
3267 | r = fd_nonblock(f->fd, false); | |
3268 | if (r < 0) | |
3269 | goto fail; | |
7560fffc | 3270 | } |
7560fffc | 3271 | |
be7cdd8e VC |
3272 | f->cache_fd = mmap_cache_add_fd(f->mmap, f->fd); |
3273 | if (!f->cache_fd) { | |
3274 | r = -ENOMEM; | |
3275 | goto fail; | |
3276 | } | |
3277 | ||
2678031a LP |
3278 | r = journal_file_fstat(f); |
3279 | if (r < 0) | |
0284adc6 | 3280 | goto fail; |
7560fffc | 3281 | |
0284adc6 | 3282 | if (f->last_stat.st_size == 0 && f->writable) { |
11689d2a | 3283 | |
fc68c929 | 3284 | (void) journal_file_warn_btrfs(f); |
11689d2a | 3285 | |
4c2e1b39 LP |
3286 | /* Let's attach the creation time to the journal file, so that the vacuuming code knows the age of this |
3287 | * file even if the file might end up corrupted one day... Ideally we'd just use the creation time many | |
3288 | * file systems maintain for each file, but the API to query this is very new, hence let's emulate this | |
3289 | * via extended attributes. If extended attributes are not supported we'll just skip this, and rely | |
3290 | * solely on mtime/atime/ctime of the file. */ | |
3291 | (void) fd_setcrtime(f->fd, 0); | |
7560fffc | 3292 | |
349cc4a5 | 3293 | #if HAVE_GCRYPT |
0284adc6 | 3294 | /* Try to load the FSPRG state, and if we can't, then |
baed47c3 | 3295 | * just don't do sealing */ |
49a32d43 LP |
3296 | if (f->seal) { |
3297 | r = journal_file_fss_load(f); | |
3298 | if (r < 0) | |
3299 | f->seal = false; | |
3300 | } | |
feb12d3e | 3301 | #endif |
7560fffc | 3302 | |
0284adc6 LP |
3303 | r = journal_file_init_header(f, template); |
3304 | if (r < 0) | |
3305 | goto fail; | |
7560fffc | 3306 | |
2678031a LP |
3307 | r = journal_file_fstat(f); |
3308 | if (r < 0) | |
0284adc6 | 3309 | goto fail; |
fb0951b0 LP |
3310 | |
3311 | newly_created = true; | |
0284adc6 | 3312 | } |
7560fffc | 3313 | |
0284adc6 | 3314 | if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { |
cfb571f3 | 3315 | r = -ENODATA; |
0284adc6 LP |
3316 | goto fail; |
3317 | } | |
7560fffc | 3318 | |
b42549ad | 3319 | r = mmap_cache_get(f->mmap, f->cache_fd, f->prot, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h, NULL); |
977eaa1e | 3320 | if (r < 0) |
0284adc6 | 3321 | goto fail; |
7560fffc | 3322 | |
fa6ac760 LP |
3323 | f->header = h; |
3324 | ||
0284adc6 | 3325 | if (!newly_created) { |
f9168190 | 3326 | set_clear_with_destructor(deferred_closes, journal_file_close); |
b58c888f | 3327 | |
0284adc6 LP |
3328 | r = journal_file_verify_header(f); |
3329 | if (r < 0) | |
3330 | goto fail; | |
3331 | } | |
7560fffc | 3332 | |
349cc4a5 | 3333 | #if HAVE_GCRYPT |
0284adc6 | 3334 | if (!newly_created && f->writable) { |
baed47c3 | 3335 | r = journal_file_fss_load(f); |
0284adc6 LP |
3336 | if (r < 0) |
3337 | goto fail; | |
3338 | } | |
feb12d3e | 3339 | #endif |
cec736d2 LP |
3340 | |
3341 | if (f->writable) { | |
4a92baf3 LP |
3342 | if (metrics) { |
3343 | journal_default_metrics(metrics, f->fd); | |
3344 | f->metrics = *metrics; | |
3345 | } else if (template) | |
3346 | f->metrics = template->metrics; | |
3347 | ||
cec736d2 LP |
3348 | r = journal_file_refresh_header(f); |
3349 | if (r < 0) | |
3350 | goto fail; | |
3351 | } | |
3352 | ||
349cc4a5 | 3353 | #if HAVE_GCRYPT |
baed47c3 | 3354 | r = journal_file_hmac_setup(f); |
14d10188 LP |
3355 | if (r < 0) |
3356 | goto fail; | |
feb12d3e | 3357 | #endif |
14d10188 | 3358 | |
cec736d2 | 3359 | if (newly_created) { |
de190aef | 3360 | r = journal_file_setup_field_hash_table(f); |
cec736d2 LP |
3361 | if (r < 0) |
3362 | goto fail; | |
3363 | ||
de190aef | 3364 | r = journal_file_setup_data_hash_table(f); |
cec736d2 LP |
3365 | if (r < 0) |
3366 | goto fail; | |
7560fffc | 3367 | |
349cc4a5 | 3368 | #if HAVE_GCRYPT |
7560fffc LP |
3369 | r = journal_file_append_first_tag(f); |
3370 | if (r < 0) | |
3371 | goto fail; | |
feb12d3e | 3372 | #endif |
cec736d2 LP |
3373 | } |
3374 | ||
be7cdd8e | 3375 | if (mmap_cache_got_sigbus(f->mmap, f->cache_fd)) { |
fa6ac760 LP |
3376 | r = -EIO; |
3377 | goto fail; | |
3378 | } | |
3379 | ||
7a24f3bf | 3380 | if (template && template->post_change_timer) { |
e167d7fd LP |
3381 | r = journal_file_enable_post_change_timer( |
3382 | f, | |
3383 | sd_event_source_get_event(template->post_change_timer), | |
3384 | template->post_change_timer_period); | |
7a24f3bf | 3385 | |
7a24f3bf VC |
3386 | if (r < 0) |
3387 | goto fail; | |
3388 | } | |
3389 | ||
f8e2f4d6 | 3390 | /* The file is opened now successfully, thus we take possession of any passed in fd. */ |
5d1ce257 LP |
3391 | f->close_fd = true; |
3392 | ||
0559d3a5 | 3393 | *ret = f; |
cec736d2 LP |
3394 | return 0; |
3395 | ||
3396 | fail: | |
be7cdd8e | 3397 | if (f->cache_fd && mmap_cache_got_sigbus(f->mmap, f->cache_fd)) |
fa6ac760 LP |
3398 | r = -EIO; |
3399 | ||
69a3a6fd | 3400 | (void) journal_file_close(f); |
cec736d2 LP |
3401 | |
3402 | return r; | |
3403 | } | |
0ac38b70 | 3404 | |
7a4d21ad | 3405 | int journal_file_archive(JournalFile *f) { |
57535f47 | 3406 | _cleanup_free_ char *p = NULL; |
0ac38b70 LP |
3407 | |
3408 | assert(f); | |
0ac38b70 | 3409 | |
7a4d21ad | 3410 | if (!f->writable) |
0ac38b70 LP |
3411 | return -EINVAL; |
3412 | ||
5d1ce257 | 3413 | /* 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 | 3414 | * rotation, since we don't know the actual path, and couldn't rename the file hence. */ |
7a4d21ad | 3415 | if (path_startswith(f->path, "/proc/self/fd")) |
5d1ce257 LP |
3416 | return -EINVAL; |
3417 | ||
7a4d21ad | 3418 | if (!endswith(f->path, ".journal")) |
0ac38b70 LP |
3419 | return -EINVAL; |
3420 | ||
7a4d21ad LP |
3421 | if (asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal", |
3422 | (int) strlen(f->path) - 8, f->path, | |
3423 | SD_ID128_FORMAT_VAL(f->header->seqnum_id), | |
3424 | le64toh(f->header->head_entry_seqnum), | |
3425 | le64toh(f->header->head_entry_realtime)) < 0) | |
0ac38b70 LP |
3426 | return -ENOMEM; |
3427 | ||
7a4d21ad LP |
3428 | /* Try to rename the file to the archived version. If the file already was deleted, we'll get ENOENT, let's |
3429 | * ignore that case. */ | |
3430 | if (rename(f->path, p) < 0 && errno != ENOENT) | |
0ac38b70 LP |
3431 | return -errno; |
3432 | ||
1fcefd88 | 3433 | /* Sync the rename to disk */ |
7a4d21ad LP |
3434 | (void) fsync_directory_of_file(f->fd); |
3435 | ||
3436 | /* Set as archive so offlining commits w/state=STATE_ARCHIVED. Previously we would set old_file->header->state | |
3437 | * to STATE_ARCHIVED directly here, but journal_file_set_offline() short-circuits when state != STATE_ONLINE, | |
3438 | * which would result in the rotated journal never getting fsync() called before closing. Now we simply queue | |
3439 | * the archive state by setting an archive bit, leaving the state as STATE_ONLINE so proper offlining | |
3440 | * occurs. */ | |
3441 | f->archive = true; | |
3442 | ||
3443 | /* Currently, btrfs is not very good with out write patterns and fragments heavily. Let's defrag our journal | |
3444 | * files when we archive them */ | |
3445 | f->defrag_on_close = true; | |
3446 | ||
3447 | return 0; | |
3448 | } | |
3449 | ||
3450 | JournalFile* journal_initiate_close( | |
3451 | JournalFile *f, | |
3452 | Set *deferred_closes) { | |
3453 | ||
3454 | int r; | |
3455 | ||
3456 | assert(f); | |
3457 | ||
3458 | if (deferred_closes) { | |
0ac38b70 | 3459 | |
7a4d21ad LP |
3460 | r = set_put(deferred_closes, f); |
3461 | if (r < 0) | |
3462 | log_debug_errno(r, "Failed to add file to deferred close set, closing immediately."); | |
3463 | else { | |
3464 | (void) journal_file_set_offline(f, false); | |
3465 | return NULL; | |
3466 | } | |
3467 | } | |
3468 | ||
3469 | return journal_file_close(f); | |
3470 | } | |
3471 | ||
3472 | int journal_file_rotate( | |
3473 | JournalFile **f, | |
3474 | bool compress, | |
3475 | uint64_t compress_threshold_bytes, | |
3476 | bool seal, | |
3477 | Set *deferred_closes) { | |
3478 | ||
3479 | JournalFile *new_file = NULL; | |
3480 | int r; | |
3481 | ||
3482 | assert(f); | |
3483 | assert(*f); | |
3484 | ||
3485 | r = journal_file_archive(*f); | |
3486 | if (r < 0) | |
3487 | return r; | |
3488 | ||
3489 | r = journal_file_open( | |
3490 | -1, | |
3491 | (*f)->path, | |
3492 | (*f)->flags, | |
3493 | (*f)->mode, | |
3494 | compress, | |
3495 | compress_threshold_bytes, | |
3496 | seal, | |
3497 | NULL, /* metrics */ | |
3498 | (*f)->mmap, | |
3499 | deferred_closes, | |
3500 | *f, /* template */ | |
3501 | &new_file); | |
3502 | ||
3503 | journal_initiate_close(*f, deferred_closes); | |
0ac38b70 | 3504 | *f = new_file; |
7a4d21ad | 3505 | |
0ac38b70 LP |
3506 | return r; |
3507 | } | |
3508 | ||
68127658 LP |
3509 | int journal_file_dispose(int dir_fd, const char *fname) { |
3510 | _cleanup_free_ char *p = NULL; | |
3511 | _cleanup_close_ int fd = -1; | |
3512 | ||
3513 | assert(fname); | |
3514 | ||
3515 | /* Renames a journal file to *.journal~, i.e. to mark it as corruped or otherwise uncleanly shutdown. Note that | |
3516 | * this is done without looking into the file or changing any of its contents. The idea is that this is called | |
3517 | * whenever something is suspicious and we want to move the file away and make clear that it is not accessed | |
3518 | * for writing anymore. */ | |
3519 | ||
3520 | if (!endswith(fname, ".journal")) | |
3521 | return -EINVAL; | |
3522 | ||
3523 | if (asprintf(&p, "%.*s@%016" PRIx64 "-%016" PRIx64 ".journal~", | |
3524 | (int) strlen(fname) - 8, fname, | |
3525 | now(CLOCK_REALTIME), | |
3526 | random_u64()) < 0) | |
3527 | return -ENOMEM; | |
3528 | ||
3529 | if (renameat(dir_fd, fname, dir_fd, p) < 0) | |
3530 | return -errno; | |
3531 | ||
3532 | /* btrfs doesn't cope well with our write pattern and fragments heavily. Let's defrag all files we rotate */ | |
3533 | fd = openat(dir_fd, p, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
3534 | if (fd < 0) | |
3535 | log_debug_errno(errno, "Failed to open file for defragmentation/FS_NOCOW_FL, ignoring: %m"); | |
3536 | else { | |
3537 | (void) chattr_fd(fd, 0, FS_NOCOW_FL, NULL); | |
3538 | (void) btrfs_defrag_fd(fd); | |
3539 | } | |
3540 | ||
3541 | return 0; | |
3542 | } | |
3543 | ||
9447a7f1 LP |
3544 | int journal_file_open_reliably( |
3545 | const char *fname, | |
3546 | int flags, | |
3547 | mode_t mode, | |
7560fffc | 3548 | bool compress, |
57850536 | 3549 | uint64_t compress_threshold_bytes, |
baed47c3 | 3550 | bool seal, |
4a92baf3 | 3551 | JournalMetrics *metrics, |
27370278 | 3552 | MMapCache *mmap_cache, |
b58c888f | 3553 | Set *deferred_closes, |
9447a7f1 LP |
3554 | JournalFile *template, |
3555 | JournalFile **ret) { | |
3556 | ||
68127658 | 3557 | int r; |
9447a7f1 | 3558 | |
57850536 AG |
3559 | r = journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache, |
3560 | deferred_closes, template, ret); | |
288359db | 3561 | if (!IN_SET(r, |
b288cdeb ZJS |
3562 | -EBADMSG, /* Corrupted */ |
3563 | -ENODATA, /* Truncated */ | |
3564 | -EHOSTDOWN, /* Other machine */ | |
3565 | -EPROTONOSUPPORT, /* Incompatible feature */ | |
3566 | -EBUSY, /* Unclean shutdown */ | |
3567 | -ESHUTDOWN, /* Already archived */ | |
288359db | 3568 | -EIO, /* IO error, including SIGBUS on mmap */ |
ae739cc1 LP |
3569 | -EIDRM, /* File has been deleted */ |
3570 | -ETXTBSY)) /* File is from the future */ | |
9447a7f1 LP |
3571 | return r; |
3572 | ||
3573 | if ((flags & O_ACCMODE) == O_RDONLY) | |
3574 | return r; | |
3575 | ||
3576 | if (!(flags & O_CREAT)) | |
3577 | return r; | |
3578 | ||
7560fffc LP |
3579 | if (!endswith(fname, ".journal")) |
3580 | return r; | |
3581 | ||
5c70eab4 | 3582 | /* The file is corrupted. Rotate it away and try it again (but only once) */ |
65089b82 | 3583 | log_warning_errno(r, "File %s corrupted or uncleanly shut down, renaming and replacing.", fname); |
9447a7f1 | 3584 | |
68127658 LP |
3585 | r = journal_file_dispose(AT_FDCWD, fname); |
3586 | if (r < 0) | |
3587 | return r; | |
3588 | ||
57850536 AG |
3589 | return journal_file_open(-1, fname, flags, mode, compress, compress_threshold_bytes, seal, metrics, mmap_cache, |
3590 | deferred_closes, template, ret); | |
9447a7f1 LP |
3591 | } |
3592 | ||
5a271b08 | 3593 | int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p) { |
cf244689 LP |
3594 | uint64_t i, n; |
3595 | uint64_t q, xor_hash = 0; | |
3596 | int r; | |
3597 | EntryItem *items; | |
3598 | dual_timestamp ts; | |
d180c349 | 3599 | const sd_id128_t *boot_id; |
cf244689 LP |
3600 | |
3601 | assert(from); | |
3602 | assert(to); | |
3603 | assert(o); | |
3604 | assert(p); | |
3605 | ||
3606 | if (!to->writable) | |
3607 | return -EPERM; | |
3608 | ||
3609 | ts.monotonic = le64toh(o->entry.monotonic); | |
3610 | ts.realtime = le64toh(o->entry.realtime); | |
d180c349 | 3611 | boot_id = &o->entry.boot_id; |
cf244689 | 3612 | |
cf244689 | 3613 | n = journal_file_entry_n_items(o); |
4faa7004 | 3614 | /* alloca() can't take 0, hence let's allocate at least one */ |
cf409d15 | 3615 | items = newa(EntryItem, MAX(1u, n)); |
cf244689 LP |
3616 | |
3617 | for (i = 0; i < n; i++) { | |
4fd052ae FC |
3618 | uint64_t l, h; |
3619 | le64_t le_hash; | |
cf244689 LP |
3620 | size_t t; |
3621 | void *data; | |
3622 | Object *u; | |
3623 | ||
3624 | q = le64toh(o->entry.items[i].object_offset); | |
3625 | le_hash = o->entry.items[i].hash; | |
3626 | ||
3627 | r = journal_file_move_to_object(from, OBJECT_DATA, q, &o); | |
3628 | if (r < 0) | |
3629 | return r; | |
3630 | ||
3631 | if (le_hash != o->data.hash) | |
3632 | return -EBADMSG; | |
3633 | ||
3634 | l = le64toh(o->object.size) - offsetof(Object, data.payload); | |
3635 | t = (size_t) l; | |
3636 | ||
3637 | /* We hit the limit on 32bit machines */ | |
3638 | if ((uint64_t) t != l) | |
3639 | return -E2BIG; | |
3640 | ||
d89c8fdf | 3641 | if (o->object.flags & OBJECT_COMPRESSION_MASK) { |
349cc4a5 | 3642 | #if HAVE_XZ || HAVE_LZ4 |
a7f7d1bd | 3643 | size_t rsize = 0; |
cf244689 | 3644 | |
d89c8fdf ZJS |
3645 | r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK, |
3646 | o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0); | |
3647 | if (r < 0) | |
3648 | return r; | |
cf244689 LP |
3649 | |
3650 | data = from->compress_buffer; | |
3651 | l = rsize; | |
3b1a55e1 ZJS |
3652 | #else |
3653 | return -EPROTONOSUPPORT; | |
3654 | #endif | |
cf244689 LP |
3655 | } else |
3656 | data = o->data.payload; | |
3657 | ||
3658 | r = journal_file_append_data(to, data, l, &u, &h); | |
3659 | if (r < 0) | |
3660 | return r; | |
3661 | ||
3662 | xor_hash ^= le64toh(u->data.hash); | |
3663 | items[i].object_offset = htole64(h); | |
3664 | items[i].hash = u->data.hash; | |
3665 | ||
3666 | r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); | |
3667 | if (r < 0) | |
3668 | return r; | |
3669 | } | |
3670 | ||
d180c349 ZJS |
3671 | r = journal_file_append_entry_internal(to, &ts, boot_id, xor_hash, items, n, |
3672 | NULL, NULL, NULL); | |
fa6ac760 | 3673 | |
be7cdd8e | 3674 | if (mmap_cache_got_sigbus(to->mmap, to->cache_fd)) |
fa6ac760 LP |
3675 | return -EIO; |
3676 | ||
3677 | return r; | |
cf244689 | 3678 | } |
babfc091 | 3679 | |
8580d1f7 LP |
3680 | void journal_reset_metrics(JournalMetrics *m) { |
3681 | assert(m); | |
3682 | ||
3683 | /* Set everything to "pick automatic values". */ | |
3684 | ||
3685 | *m = (JournalMetrics) { | |
3686 | .min_use = (uint64_t) -1, | |
3687 | .max_use = (uint64_t) -1, | |
3688 | .min_size = (uint64_t) -1, | |
3689 | .max_size = (uint64_t) -1, | |
3690 | .keep_free = (uint64_t) -1, | |
3691 | .n_max_files = (uint64_t) -1, | |
3692 | }; | |
3693 | } | |
3694 | ||
babfc091 | 3695 | void journal_default_metrics(JournalMetrics *m, int fd) { |
8580d1f7 | 3696 | char a[FORMAT_BYTES_MAX], b[FORMAT_BYTES_MAX], c[FORMAT_BYTES_MAX], d[FORMAT_BYTES_MAX], e[FORMAT_BYTES_MAX]; |
babfc091 | 3697 | struct statvfs ss; |
8580d1f7 | 3698 | uint64_t fs_size; |
babfc091 LP |
3699 | |
3700 | assert(m); | |
3701 | assert(fd >= 0); | |
3702 | ||
3703 | if (fstatvfs(fd, &ss) >= 0) | |
3704 | fs_size = ss.f_frsize * ss.f_blocks; | |
8580d1f7 | 3705 | else { |
8fc58f1a | 3706 | log_debug_errno(errno, "Failed to determine disk size: %m"); |
8580d1f7 LP |
3707 | fs_size = 0; |
3708 | } | |
babfc091 LP |
3709 | |
3710 | if (m->max_use == (uint64_t) -1) { | |
3711 | ||
3712 | if (fs_size > 0) { | |
3713 | m->max_use = PAGE_ALIGN(fs_size / 10); /* 10% of file system size */ | |
3714 | ||
3715 | if (m->max_use > DEFAULT_MAX_USE_UPPER) | |
3716 | m->max_use = DEFAULT_MAX_USE_UPPER; | |
3717 | ||
3718 | if (m->max_use < DEFAULT_MAX_USE_LOWER) | |
3719 | m->max_use = DEFAULT_MAX_USE_LOWER; | |
3720 | } else | |
3721 | m->max_use = DEFAULT_MAX_USE_LOWER; | |
3722 | } else { | |
3723 | m->max_use = PAGE_ALIGN(m->max_use); | |
3724 | ||
8580d1f7 | 3725 | if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2) |
babfc091 LP |
3726 | m->max_use = JOURNAL_FILE_SIZE_MIN*2; |
3727 | } | |
3728 | ||
8580d1f7 LP |
3729 | if (m->min_use == (uint64_t) -1) |
3730 | m->min_use = DEFAULT_MIN_USE; | |
3731 | ||
3732 | if (m->min_use > m->max_use) | |
3733 | m->min_use = m->max_use; | |
3734 | ||
babfc091 LP |
3735 | if (m->max_size == (uint64_t) -1) { |
3736 | m->max_size = PAGE_ALIGN(m->max_use / 8); /* 8 chunks */ | |
3737 | ||
3738 | if (m->max_size > DEFAULT_MAX_SIZE_UPPER) | |
3739 | m->max_size = DEFAULT_MAX_SIZE_UPPER; | |
3740 | } else | |
3741 | m->max_size = PAGE_ALIGN(m->max_size); | |
3742 | ||
8580d1f7 LP |
3743 | if (m->max_size != 0) { |
3744 | if (m->max_size < JOURNAL_FILE_SIZE_MIN) | |
3745 | m->max_size = JOURNAL_FILE_SIZE_MIN; | |
babfc091 | 3746 | |
8580d1f7 LP |
3747 | if (m->max_use != 0 && m->max_size*2 > m->max_use) |
3748 | m->max_use = m->max_size*2; | |
3749 | } | |
babfc091 LP |
3750 | |
3751 | if (m->min_size == (uint64_t) -1) | |
3752 | m->min_size = JOURNAL_FILE_SIZE_MIN; | |
3753 | else { | |
3754 | m->min_size = PAGE_ALIGN(m->min_size); | |
3755 | ||
3756 | if (m->min_size < JOURNAL_FILE_SIZE_MIN) | |
3757 | m->min_size = JOURNAL_FILE_SIZE_MIN; | |
3758 | ||
8580d1f7 | 3759 | if (m->max_size != 0 && m->min_size > m->max_size) |
babfc091 LP |
3760 | m->max_size = m->min_size; |
3761 | } | |
3762 | ||
3763 | if (m->keep_free == (uint64_t) -1) { | |
3764 | ||
3765 | if (fs_size > 0) { | |
8621b110 | 3766 | m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */ |
babfc091 LP |
3767 | |
3768 | if (m->keep_free > DEFAULT_KEEP_FREE_UPPER) | |
3769 | m->keep_free = DEFAULT_KEEP_FREE_UPPER; | |
3770 | ||
3771 | } else | |
3772 | m->keep_free = DEFAULT_KEEP_FREE; | |
3773 | } | |
3774 | ||
8580d1f7 LP |
3775 | if (m->n_max_files == (uint64_t) -1) |
3776 | m->n_max_files = DEFAULT_N_MAX_FILES; | |
3777 | ||
3778 | log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64, | |
3779 | format_bytes(a, sizeof(a), m->min_use), | |
3780 | format_bytes(b, sizeof(b), m->max_use), | |
3781 | format_bytes(c, sizeof(c), m->max_size), | |
3782 | format_bytes(d, sizeof(d), m->min_size), | |
3783 | format_bytes(e, sizeof(e), m->keep_free), | |
3784 | m->n_max_files); | |
babfc091 | 3785 | } |
08984293 LP |
3786 | |
3787 | int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) { | |
08984293 | 3788 | assert(f); |
c88cc6af | 3789 | assert(f->header); |
08984293 LP |
3790 | assert(from || to); |
3791 | ||
3792 | if (from) { | |
162566a4 LP |
3793 | if (f->header->head_entry_realtime == 0) |
3794 | return -ENOENT; | |
08984293 | 3795 | |
162566a4 | 3796 | *from = le64toh(f->header->head_entry_realtime); |
08984293 LP |
3797 | } |
3798 | ||
3799 | if (to) { | |
162566a4 LP |
3800 | if (f->header->tail_entry_realtime == 0) |
3801 | return -ENOENT; | |
08984293 | 3802 | |
162566a4 | 3803 | *to = le64toh(f->header->tail_entry_realtime); |
08984293 LP |
3804 | } |
3805 | ||
3806 | return 1; | |
3807 | } | |
3808 | ||
3809 | int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) { | |
08984293 LP |
3810 | Object *o; |
3811 | uint64_t p; | |
3812 | int r; | |
3813 | ||
3814 | assert(f); | |
3815 | assert(from || to); | |
3816 | ||
47838ab3 | 3817 | r = find_data_object_by_boot_id(f, boot_id, &o, &p); |
08984293 LP |
3818 | if (r <= 0) |
3819 | return r; | |
3820 | ||
3821 | if (le64toh(o->data.n_entries) <= 0) | |
3822 | return 0; | |
3823 | ||
3824 | if (from) { | |
3825 | r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); | |
3826 | if (r < 0) | |
3827 | return r; | |
3828 | ||
3829 | *from = le64toh(o->entry.monotonic); | |
3830 | } | |
3831 | ||
3832 | if (to) { | |
3833 | r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); | |
3834 | if (r < 0) | |
3835 | return r; | |
3836 | ||
3837 | r = generic_array_get_plus_one(f, | |
3838 | le64toh(o->data.entry_offset), | |
3839 | le64toh(o->data.entry_array_offset), | |
3840 | le64toh(o->data.n_entries)-1, | |
3841 | &o, NULL); | |
3842 | if (r <= 0) | |
3843 | return r; | |
3844 | ||
3845 | *to = le64toh(o->entry.monotonic); | |
3846 | } | |
3847 | ||
3848 | return 1; | |
3849 | } | |
dca6219e | 3850 | |
fb0951b0 | 3851 | bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) { |
dca6219e | 3852 | assert(f); |
c88cc6af | 3853 | assert(f->header); |
dca6219e LP |
3854 | |
3855 | /* If we gained new header fields we gained new features, | |
3856 | * hence suggest a rotation */ | |
361f9cbc LP |
3857 | if (le64toh(f->header->header_size) < sizeof(Header)) { |
3858 | log_debug("%s uses an outdated header, suggesting rotation.", f->path); | |
dca6219e | 3859 | return true; |
361f9cbc | 3860 | } |
dca6219e LP |
3861 | |
3862 | /* Let's check if the hash tables grew over a certain fill | |
3863 | * level (75%, borrowing this value from Java's hash table | |
3864 | * implementation), and if so suggest a rotation. To calculate | |
3865 | * the fill level we need the n_data field, which only exists | |
3866 | * in newer versions. */ | |
3867 | ||
3868 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) | |
361f9cbc | 3869 | if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
507f22bd | 3870 | log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.", |
361f9cbc LP |
3871 | f->path, |
3872 | 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))), | |
507f22bd ZJS |
3873 | le64toh(f->header->n_data), |
3874 | le64toh(f->header->data_hash_table_size) / sizeof(HashItem), | |
3875 | (unsigned long long) f->last_stat.st_size, | |
3876 | f->last_stat.st_size / le64toh(f->header->n_data)); | |
dca6219e | 3877 | return true; |
361f9cbc | 3878 | } |
dca6219e LP |
3879 | |
3880 | if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) | |
361f9cbc | 3881 | if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) { |
507f22bd | 3882 | log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.", |
361f9cbc LP |
3883 | f->path, |
3884 | 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), | |
507f22bd ZJS |
3885 | le64toh(f->header->n_fields), |
3886 | le64toh(f->header->field_hash_table_size) / sizeof(HashItem)); | |
dca6219e | 3887 | return true; |
361f9cbc | 3888 | } |
dca6219e | 3889 | |
0598fd4a LP |
3890 | /* Are the data objects properly indexed by field objects? */ |
3891 | if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && | |
3892 | JOURNAL_HEADER_CONTAINS(f->header, n_fields) && | |
3893 | le64toh(f->header->n_data) > 0 && | |
3894 | le64toh(f->header->n_fields) == 0) | |
3895 | return true; | |
3896 | ||
fb0951b0 LP |
3897 | if (max_file_usec > 0) { |
3898 | usec_t t, h; | |
3899 | ||
3900 | h = le64toh(f->header->head_entry_realtime); | |
3901 | t = now(CLOCK_REALTIME); | |
3902 | ||
3903 | if (h > 0 && t > h + max_file_usec) | |
3904 | return true; | |
3905 | } | |
3906 | ||
dca6219e LP |
3907 | return false; |
3908 | } |