]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
d025f1e4 | 2 | |
349cc4a5 | 3 | #if HAVE_SELINUX |
24882e06 LP |
4 | #include <selinux/selinux.h> |
5 | #endif | |
8580d1f7 LP |
6 | #include <sys/ioctl.h> |
7 | #include <sys/mman.h> | |
8 | #include <sys/signalfd.h> | |
9 | #include <sys/statvfs.h> | |
07630cea | 10 | #include <linux/sockios.h> |
24882e06 | 11 | |
8580d1f7 | 12 | #include "sd-daemon.h" |
74df0fca LP |
13 | #include "sd-journal.h" |
14 | #include "sd-messages.h" | |
8580d1f7 LP |
15 | |
16 | #include "acl-util.h" | |
b5efdb8a | 17 | #include "alloc-util.h" |
430f0182 | 18 | #include "audit-util.h" |
d025f1e4 | 19 | #include "cgroup-util.h" |
d025f1e4 | 20 | #include "conf-parser.h" |
a0956174 | 21 | #include "dirent-util.h" |
0dec689b | 22 | #include "extract-word.h" |
3ffd4af2 | 23 | #include "fd-util.h" |
33d52ab9 | 24 | #include "fileio.h" |
f97b34a6 | 25 | #include "format-util.h" |
f4f15635 | 26 | #include "fs-util.h" |
8580d1f7 | 27 | #include "hashmap.h" |
958b66ea | 28 | #include "hostname-util.h" |
4b58153d | 29 | #include "id128-util.h" |
afc5dbf3 | 30 | #include "io-util.h" |
8580d1f7 LP |
31 | #include "journal-authenticate.h" |
32 | #include "journal-file.h" | |
d025f1e4 ZJS |
33 | #include "journal-internal.h" |
34 | #include "journal-vacuum.h" | |
8580d1f7 | 35 | #include "journald-audit.h" |
22e3a02b | 36 | #include "journald-context.h" |
d025f1e4 | 37 | #include "journald-kmsg.h" |
d025f1e4 | 38 | #include "journald-native.h" |
8580d1f7 | 39 | #include "journald-rate-limit.h" |
3ffd4af2 | 40 | #include "journald-server.h" |
8580d1f7 LP |
41 | #include "journald-stream.h" |
42 | #include "journald-syslog.h" | |
4b58153d | 43 | #include "log.h" |
07630cea LP |
44 | #include "missing.h" |
45 | #include "mkdir.h" | |
6bedfcbb | 46 | #include "parse-util.h" |
4e731273 | 47 | #include "proc-cmdline.h" |
07630cea LP |
48 | #include "process-util.h" |
49 | #include "rm-rf.h" | |
50 | #include "selinux-util.h" | |
51 | #include "signal-util.h" | |
52 | #include "socket-util.h" | |
32917e33 | 53 | #include "stdio-util.h" |
8b43440b | 54 | #include "string-table.h" |
07630cea | 55 | #include "string-util.h" |
863a5610 | 56 | #include "syslog-util.h" |
22e3a02b | 57 | #include "user-util.h" |
d025f1e4 | 58 | |
d025f1e4 ZJS |
59 | #define USER_JOURNALS_MAX 1024 |
60 | ||
26687bf8 | 61 | #define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) |
7f1ad696 | 62 | #define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC) |
3de8ff5a | 63 | #define DEFAULT_RATE_LIMIT_BURST 10000 |
e150e820 | 64 | #define DEFAULT_MAX_FILE_USEC USEC_PER_MONTH |
d025f1e4 | 65 | |
8580d1f7 | 66 | #define RECHECK_SPACE_USEC (30*USEC_PER_SEC) |
d025f1e4 | 67 | |
e22aa3d3 LP |
68 | #define NOTIFY_SNDBUF_SIZE (8*1024*1024) |
69 | ||
7a24f3bf VC |
70 | /* The period to insert between posting changes for coalescing */ |
71 | #define POST_CHANGE_TIMER_INTERVAL_USEC (250*USEC_PER_MSEC) | |
72 | ||
ec20fe5f LP |
73 | /* Pick a good default that is likely to fit into AF_UNIX and AF_INET SOCK_DGRAM datagrams, and even leaves some room |
74 | * for a bit of additional metadata. */ | |
75 | #define DEFAULT_LINE_MAX (48*1024) | |
76 | ||
a33687b7 LP |
77 | #define DEFERRED_CLOSES_MAX (4096) |
78 | ||
e0ed6db9 FB |
79 | static int determine_path_usage(Server *s, const char *path, uint64_t *ret_used, uint64_t *ret_free) { |
80 | _cleanup_closedir_ DIR *d = NULL; | |
81 | struct dirent *de; | |
82 | struct statvfs ss; | |
e0ed6db9 FB |
83 | |
84 | assert(ret_used); | |
85 | assert(ret_free); | |
86 | ||
266a4700 | 87 | d = opendir(path); |
e0ed6db9 FB |
88 | if (!d) |
89 | return log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_ERR, | |
266a4700 | 90 | errno, "Failed to open %s: %m", path); |
e0ed6db9 FB |
91 | |
92 | if (fstatvfs(dirfd(d), &ss) < 0) | |
266a4700 | 93 | return log_error_errno(errno, "Failed to fstatvfs(%s): %m", path); |
e0ed6db9 FB |
94 | |
95 | *ret_free = ss.f_bsize * ss.f_bavail; | |
96 | *ret_used = 0; | |
97 | FOREACH_DIRENT_ALL(de, d, break) { | |
98 | struct stat st; | |
99 | ||
100 | if (!endswith(de->d_name, ".journal") && | |
101 | !endswith(de->d_name, ".journal~")) | |
102 | continue; | |
103 | ||
104 | if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) { | |
266a4700 | 105 | log_debug_errno(errno, "Failed to stat %s/%s, ignoring: %m", path, de->d_name); |
e0ed6db9 FB |
106 | continue; |
107 | } | |
108 | ||
109 | if (!S_ISREG(st.st_mode)) | |
110 | continue; | |
111 | ||
112 | *ret_used += (uint64_t) st.st_blocks * 512UL; | |
113 | } | |
114 | ||
115 | return 0; | |
116 | } | |
117 | ||
a0edc477 | 118 | static void cache_space_invalidate(JournalStorageSpace *space) { |
67319249 | 119 | zero(*space); |
a0edc477 FB |
120 | } |
121 | ||
57f443a6 | 122 | static int cache_space_refresh(Server *s, JournalStorage *storage) { |
23aba343 | 123 | JournalStorageSpace *space; |
266a4700 | 124 | JournalMetrics *metrics; |
23aba343 | 125 | uint64_t vfs_used, vfs_avail, avail; |
d025f1e4 | 126 | usec_t ts; |
e0ed6db9 | 127 | int r; |
d025f1e4 | 128 | |
8580d1f7 | 129 | assert(s); |
266a4700 | 130 | |
266a4700 | 131 | metrics = &storage->metrics; |
23aba343 | 132 | space = &storage->space; |
d025f1e4 | 133 | |
8580d1f7 | 134 | ts = now(CLOCK_MONOTONIC); |
d025f1e4 | 135 | |
3099caf2 | 136 | if (space->timestamp != 0 && space->timestamp + RECHECK_SPACE_USEC > ts) |
d025f1e4 ZJS |
137 | return 0; |
138 | ||
23aba343 | 139 | r = determine_path_usage(s, storage->path, &vfs_used, &vfs_avail); |
e0ed6db9 FB |
140 | if (r < 0) |
141 | return r; | |
d025f1e4 | 142 | |
23aba343 FB |
143 | space->vfs_used = vfs_used; |
144 | space->vfs_available = vfs_avail; | |
145 | ||
146 | avail = LESS_BY(vfs_avail, metrics->keep_free); | |
147 | ||
23aba343 FB |
148 | space->limit = MIN(MAX(vfs_used + avail, metrics->min_use), metrics->max_use); |
149 | space->available = LESS_BY(space->limit, vfs_used); | |
150 | space->timestamp = ts; | |
8580d1f7 LP |
151 | return 1; |
152 | } | |
153 | ||
3a19f215 FB |
154 | static void patch_min_use(JournalStorage *storage) { |
155 | assert(storage); | |
156 | ||
157 | /* Let's bump the min_use limit to the current usage on disk. We do | |
158 | * this when starting up and first opening the journal files. This way | |
159 | * sudden spikes in disk usage will not cause journald to vacuum files | |
160 | * without bounds. Note that this means that only a restart of journald | |
161 | * will make it reset this value. */ | |
162 | ||
163 | storage->metrics.min_use = MAX(storage->metrics.min_use, storage->space.vfs_used); | |
164 | } | |
165 | ||
3a19f215 | 166 | static int determine_space(Server *s, uint64_t *available, uint64_t *limit) { |
266a4700 | 167 | JournalStorage *js; |
57f443a6 | 168 | int r; |
8580d1f7 LP |
169 | |
170 | assert(s); | |
171 | ||
266a4700 | 172 | js = s->system_journal ? &s->system_storage : &s->runtime_storage; |
57f443a6 FB |
173 | |
174 | r = cache_space_refresh(s, js); | |
175 | if (r >= 0) { | |
176 | if (available) | |
177 | *available = js->space.available; | |
178 | if (limit) | |
179 | *limit = js->space.limit; | |
180 | } | |
181 | return r; | |
d025f1e4 ZJS |
182 | } |
183 | ||
cba5629e FB |
184 | void server_space_usage_message(Server *s, JournalStorage *storage) { |
185 | char fb1[FORMAT_BYTES_MAX], fb2[FORMAT_BYTES_MAX], fb3[FORMAT_BYTES_MAX], | |
186 | fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX], fb6[FORMAT_BYTES_MAX]; | |
187 | JournalMetrics *metrics; | |
cba5629e FB |
188 | |
189 | assert(s); | |
190 | ||
191 | if (!storage) | |
192 | storage = s->system_journal ? &s->system_storage : &s->runtime_storage; | |
193 | ||
57f443a6 | 194 | if (cache_space_refresh(s, storage) < 0) |
cba5629e FB |
195 | return; |
196 | ||
197 | metrics = &storage->metrics; | |
23aba343 | 198 | format_bytes(fb1, sizeof(fb1), storage->space.vfs_used); |
cba5629e FB |
199 | format_bytes(fb2, sizeof(fb2), metrics->max_use); |
200 | format_bytes(fb3, sizeof(fb3), metrics->keep_free); | |
23aba343 | 201 | format_bytes(fb4, sizeof(fb4), storage->space.vfs_available); |
cba5629e FB |
202 | format_bytes(fb5, sizeof(fb5), storage->space.limit); |
203 | format_bytes(fb6, sizeof(fb6), storage->space.available); | |
204 | ||
13181942 LP |
205 | server_driver_message(s, 0, |
206 | "MESSAGE_ID=" SD_MESSAGE_JOURNAL_USAGE_STR, | |
cba5629e FB |
207 | LOG_MESSAGE("%s (%s) is %s, max %s, %s free.", |
208 | storage->name, storage->path, fb1, fb5, fb6), | |
209 | "JOURNAL_NAME=%s", storage->name, | |
210 | "JOURNAL_PATH=%s", storage->path, | |
23aba343 | 211 | "CURRENT_USE=%"PRIu64, storage->space.vfs_used, |
cba5629e FB |
212 | "CURRENT_USE_PRETTY=%s", fb1, |
213 | "MAX_USE=%"PRIu64, metrics->max_use, | |
214 | "MAX_USE_PRETTY=%s", fb2, | |
215 | "DISK_KEEP_FREE=%"PRIu64, metrics->keep_free, | |
216 | "DISK_KEEP_FREE_PRETTY=%s", fb3, | |
23aba343 | 217 | "DISK_AVAILABLE=%"PRIu64, storage->space.vfs_available, |
cba5629e FB |
218 | "DISK_AVAILABLE_PRETTY=%s", fb4, |
219 | "LIMIT=%"PRIu64, storage->space.limit, | |
220 | "LIMIT_PRETTY=%s", fb5, | |
221 | "AVAILABLE=%"PRIu64, storage->space.available, | |
222 | "AVAILABLE_PRETTY=%s", fb6, | |
223 | NULL); | |
224 | } | |
225 | ||
2fce06b0 LP |
226 | static bool uid_for_system_journal(uid_t uid) { |
227 | ||
228 | /* Returns true if the specified UID shall get its data stored in the system journal*/ | |
229 | ||
230 | return uid_is_system(uid) || uid_is_dynamic(uid) || uid == UID_NOBODY; | |
231 | } | |
232 | ||
5c3bde3f | 233 | static void server_add_acls(JournalFile *f, uid_t uid) { |
349cc4a5 | 234 | #if HAVE_ACL |
5c3bde3f | 235 | int r; |
d025f1e4 | 236 | #endif |
d025f1e4 ZJS |
237 | assert(f); |
238 | ||
349cc4a5 | 239 | #if HAVE_ACL |
2fce06b0 | 240 | if (uid_for_system_journal(uid)) |
d025f1e4 ZJS |
241 | return; |
242 | ||
5c3bde3f ZJS |
243 | r = add_acls_for_user(f->fd, uid); |
244 | if (r < 0) | |
245 | log_warning_errno(r, "Failed to set ACL on %s, ignoring: %m", f->path); | |
d025f1e4 ZJS |
246 | #endif |
247 | } | |
248 | ||
7a24f3bf VC |
249 | static int open_journal( |
250 | Server *s, | |
251 | bool reliably, | |
252 | const char *fname, | |
253 | int flags, | |
254 | bool seal, | |
255 | JournalMetrics *metrics, | |
7a24f3bf | 256 | JournalFile **ret) { |
e8591544 | 257 | |
e167d7fd | 258 | JournalFile *f; |
e8591544 | 259 | int r; |
7a24f3bf VC |
260 | |
261 | assert(s); | |
262 | assert(fname); | |
263 | assert(ret); | |
264 | ||
265 | if (reliably) | |
1b7cf0e5 AG |
266 | r = journal_file_open_reliably(fname, flags, 0640, s->compress.enabled, s->compress.threshold_bytes, |
267 | seal, metrics, s->mmap, s->deferred_closes, NULL, &f); | |
7a24f3bf | 268 | else |
1b7cf0e5 AG |
269 | r = journal_file_open(-1, fname, flags, 0640, s->compress.enabled, s->compress.threshold_bytes, seal, |
270 | metrics, s->mmap, s->deferred_closes, NULL, &f); | |
271 | ||
7a24f3bf VC |
272 | if (r < 0) |
273 | return r; | |
274 | ||
e167d7fd | 275 | r = journal_file_enable_post_change_timer(f, s->event, POST_CHANGE_TIMER_INTERVAL_USEC); |
7a24f3bf | 276 | if (r < 0) { |
69a3a6fd | 277 | (void) journal_file_close(f); |
7a24f3bf VC |
278 | return r; |
279 | } | |
280 | ||
e167d7fd | 281 | *ret = f; |
7a24f3bf VC |
282 | return r; |
283 | } | |
284 | ||
6431c7e2 | 285 | static bool flushed_flag_is_set(void) { |
f78273c8 | 286 | return access("/run/systemd/journal/flushed", F_OK) >= 0; |
6431c7e2 VC |
287 | } |
288 | ||
105bdb46 VC |
289 | static int system_journal_open(Server *s, bool flush_requested) { |
290 | const char *fn; | |
291 | int r = 0; | |
292 | ||
293 | if (!s->system_journal && | |
f78273c8 LP |
294 | IN_SET(s->storage, STORAGE_PERSISTENT, STORAGE_AUTO) && |
295 | (flush_requested || flushed_flag_is_set())) { | |
105bdb46 VC |
296 | |
297 | /* If in auto mode: first try to create the machine | |
298 | * path, but not the prefix. | |
299 | * | |
300 | * If in persistent mode: create /var/log/journal and | |
301 | * the machine path */ | |
302 | ||
303 | if (s->storage == STORAGE_PERSISTENT) | |
304 | (void) mkdir_p("/var/log/journal/", 0755); | |
305 | ||
266a4700 | 306 | (void) mkdir(s->system_storage.path, 0755); |
105bdb46 | 307 | |
266a4700 FB |
308 | fn = strjoina(s->system_storage.path, "/system.journal"); |
309 | r = open_journal(s, true, fn, O_RDWR|O_CREAT, s->seal, &s->system_storage.metrics, &s->system_journal); | |
105bdb46 VC |
310 | if (r >= 0) { |
311 | server_add_acls(s->system_journal, 0); | |
57f443a6 | 312 | (void) cache_space_refresh(s, &s->system_storage); |
3a19f215 | 313 | patch_min_use(&s->system_storage); |
29bfb683 | 314 | } else { |
4c701096 | 315 | if (!IN_SET(r, -ENOENT, -EROFS)) |
105bdb46 VC |
316 | log_warning_errno(r, "Failed to open system journal: %m"); |
317 | ||
318 | r = 0; | |
319 | } | |
929eeb54 VC |
320 | |
321 | /* If the runtime journal is open, and we're post-flush, we're | |
322 | * recovering from a failed system journal rotate (ENOSPC) | |
323 | * for which the runtime journal was reopened. | |
324 | * | |
325 | * Perform an implicit flush to var, leaving the runtime | |
326 | * journal closed, now that the system journal is back. | |
327 | */ | |
f78273c8 LP |
328 | if (!flush_requested) |
329 | (void) server_flush_to_var(s, true); | |
105bdb46 VC |
330 | } |
331 | ||
332 | if (!s->runtime_journal && | |
333 | (s->storage != STORAGE_NONE)) { | |
334 | ||
266a4700 | 335 | fn = strjoina(s->runtime_storage.path, "/system.journal"); |
105bdb46 VC |
336 | |
337 | if (s->system_journal) { | |
338 | ||
339 | /* Try to open the runtime journal, but only | |
340 | * if it already exists, so that we can flush | |
341 | * it into the system journal */ | |
342 | ||
266a4700 | 343 | r = open_journal(s, false, fn, O_RDWR, false, &s->runtime_storage.metrics, &s->runtime_journal); |
105bdb46 VC |
344 | if (r < 0) { |
345 | if (r != -ENOENT) | |
346 | log_warning_errno(r, "Failed to open runtime journal: %m"); | |
347 | ||
348 | r = 0; | |
349 | } | |
350 | ||
351 | } else { | |
352 | ||
353 | /* OK, we really need the runtime journal, so create | |
354 | * it if necessary. */ | |
355 | ||
356 | (void) mkdir("/run/log", 0755); | |
357 | (void) mkdir("/run/log/journal", 0755); | |
358 | (void) mkdir_parents(fn, 0750); | |
359 | ||
266a4700 | 360 | r = open_journal(s, true, fn, O_RDWR|O_CREAT, false, &s->runtime_storage.metrics, &s->runtime_journal); |
105bdb46 VC |
361 | if (r < 0) |
362 | return log_error_errno(r, "Failed to open runtime journal: %m"); | |
363 | } | |
364 | ||
365 | if (s->runtime_journal) { | |
366 | server_add_acls(s->runtime_journal, 0); | |
57f443a6 | 367 | (void) cache_space_refresh(s, &s->runtime_storage); |
3a19f215 | 368 | patch_min_use(&s->runtime_storage); |
105bdb46 VC |
369 | } |
370 | } | |
371 | ||
372 | return r; | |
373 | } | |
374 | ||
d025f1e4 | 375 | static JournalFile* find_journal(Server *s, uid_t uid) { |
ed375beb | 376 | _cleanup_free_ char *p = NULL; |
d025f1e4 ZJS |
377 | int r; |
378 | JournalFile *f; | |
379 | sd_id128_t machine; | |
380 | ||
381 | assert(s); | |
382 | ||
105bdb46 VC |
383 | /* A rotate that fails to create the new journal (ENOSPC) leaves the |
384 | * rotated journal as NULL. Unless we revisit opening, even after | |
385 | * space is made available we'll continue to return NULL indefinitely. | |
386 | * | |
387 | * system_journal_open() is a noop if the journals are already open, so | |
388 | * we can just call it here to recover from failed rotates (or anything | |
389 | * else that's left the journals as NULL). | |
390 | * | |
391 | * Fixes https://github.com/systemd/systemd/issues/3968 */ | |
392 | (void) system_journal_open(s, false); | |
393 | ||
d025f1e4 ZJS |
394 | /* We split up user logs only on /var, not on /run. If the |
395 | * runtime file is open, we write to it exclusively, in order | |
396 | * to guarantee proper order as soon as we flush /run to | |
397 | * /var and close the runtime file. */ | |
398 | ||
399 | if (s->runtime_journal) | |
400 | return s->runtime_journal; | |
401 | ||
2fce06b0 | 402 | if (uid_for_system_journal(uid)) |
d025f1e4 ZJS |
403 | return s->system_journal; |
404 | ||
4a0b58c4 | 405 | f = ordered_hashmap_get(s->user_journals, UID_TO_PTR(uid)); |
d025f1e4 ZJS |
406 | if (f) |
407 | return f; | |
408 | ||
e8591544 LP |
409 | r = sd_id128_get_machine(&machine); |
410 | if (r < 0) { | |
411 | log_debug_errno(r, "Failed to determine machine ID, using system log: %m"); | |
412 | return s->system_journal; | |
413 | } | |
414 | ||
de0671ee | 415 | if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-"UID_FMT".journal", |
d0307775 LP |
416 | SD_ID128_FORMAT_VAL(machine), uid) < 0) { |
417 | log_oom(); | |
d025f1e4 | 418 | return s->system_journal; |
d0307775 | 419 | } |
d025f1e4 | 420 | |
43cf8388 | 421 | while (ordered_hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { |
d025f1e4 | 422 | /* Too many open? Then let's close one */ |
43cf8388 | 423 | f = ordered_hashmap_steal_first(s->user_journals); |
d025f1e4 | 424 | assert(f); |
69a3a6fd | 425 | (void) journal_file_close(f); |
d025f1e4 ZJS |
426 | } |
427 | ||
266a4700 | 428 | r = open_journal(s, true, p, O_RDWR|O_CREAT, s->seal, &s->system_storage.metrics, &f); |
d025f1e4 ZJS |
429 | if (r < 0) |
430 | return s->system_journal; | |
431 | ||
5c3bde3f | 432 | server_add_acls(f, uid); |
d025f1e4 | 433 | |
4a0b58c4 | 434 | r = ordered_hashmap_put(s->user_journals, UID_TO_PTR(uid), f); |
d025f1e4 | 435 | if (r < 0) { |
69a3a6fd | 436 | (void) journal_file_close(f); |
d025f1e4 ZJS |
437 | return s->system_journal; |
438 | } | |
439 | ||
440 | return f; | |
441 | } | |
442 | ||
ea69bd41 LP |
443 | static int do_rotate( |
444 | Server *s, | |
445 | JournalFile **f, | |
446 | const char* name, | |
447 | bool seal, | |
448 | uint32_t uid) { | |
449 | ||
fc55baee ZJS |
450 | int r; |
451 | assert(s); | |
452 | ||
453 | if (!*f) | |
454 | return -EINVAL; | |
455 | ||
1b7cf0e5 | 456 | r = journal_file_rotate(f, s->compress.enabled, s->compress.threshold_bytes, seal, s->deferred_closes); |
bb6b922f | 457 | if (r < 0) { |
fc55baee | 458 | if (*f) |
bb6b922f | 459 | return log_error_errno(r, "Failed to rotate %s: %m", (*f)->path); |
fc55baee | 460 | else |
bb6b922f YW |
461 | return log_error_errno(r, "Failed to create new %s journal: %m", name); |
462 | } | |
463 | ||
464 | server_add_acls(*f, uid); | |
2678031a | 465 | |
fc55baee ZJS |
466 | return r; |
467 | } | |
468 | ||
f760d8a8 LP |
469 | static void server_process_deferred_closes(Server *s) { |
470 | JournalFile *f; | |
471 | Iterator i; | |
472 | ||
473 | /* Perform any deferred closes which aren't still offlining. */ | |
a33687b7 LP |
474 | SET_FOREACH(f, s->deferred_closes, i) { |
475 | if (journal_file_is_offlining(f)) | |
476 | continue; | |
477 | ||
478 | (void) set_remove(s->deferred_closes, f); | |
479 | (void) journal_file_close(f); | |
480 | } | |
481 | } | |
482 | ||
483 | static void server_vacuum_deferred_closes(Server *s) { | |
484 | assert(s); | |
485 | ||
486 | /* Make some room in the deferred closes list, so that it doesn't grow without bounds */ | |
487 | if (set_size(s->deferred_closes) < DEFERRED_CLOSES_MAX) | |
488 | return; | |
489 | ||
490 | /* Let's first remove all journal files that might already have completed closing */ | |
491 | server_process_deferred_closes(s); | |
492 | ||
493 | /* And now, let's close some more until we reach the limit again. */ | |
494 | while (set_size(s->deferred_closes) >= DEFERRED_CLOSES_MAX) { | |
495 | JournalFile *f; | |
496 | ||
497 | assert_se(f = set_steal_first(s->deferred_closes)); | |
498 | journal_file_close(f); | |
499 | } | |
500 | } | |
501 | ||
502 | static int open_user_journal_directory(Server *s, DIR **ret_dir, char **ret_path) { | |
503 | _cleanup_closedir_ DIR *dir = NULL; | |
504 | _cleanup_free_ char *path = NULL; | |
505 | sd_id128_t machine; | |
506 | int r; | |
507 | ||
508 | assert(s); | |
509 | ||
510 | r = sd_id128_get_machine(&machine); | |
511 | if (r < 0) | |
512 | return log_error_errno(r, "Failed to determine machine ID, ignoring: %m"); | |
513 | ||
514 | if (asprintf(&path, "/var/log/journal/" SD_ID128_FORMAT_STR "/", SD_ID128_FORMAT_VAL(machine)) < 0) | |
515 | return log_oom(); | |
516 | ||
517 | dir = opendir(path); | |
518 | if (!dir) | |
519 | return log_error_errno(errno, "Failed to open user journal directory '%s': %m", path); | |
520 | ||
521 | if (ret_dir) | |
522 | *ret_dir = TAKE_PTR(dir); | |
523 | if (ret_path) | |
524 | *ret_path = TAKE_PTR(path); | |
525 | ||
526 | return 0; | |
f760d8a8 LP |
527 | } |
528 | ||
d025f1e4 | 529 | void server_rotate(Server *s) { |
a33687b7 LP |
530 | _cleanup_free_ char *path = NULL; |
531 | _cleanup_closedir_ DIR *d = NULL; | |
d025f1e4 | 532 | JournalFile *f; |
d025f1e4 | 533 | Iterator i; |
a33687b7 | 534 | void *k; |
d025f1e4 ZJS |
535 | int r; |
536 | ||
537 | log_debug("Rotating..."); | |
538 | ||
a33687b7 | 539 | /* First, rotate the system journal (either in its runtime flavour or in its runtime flavour) */ |
8580d1f7 LP |
540 | (void) do_rotate(s, &s->runtime_journal, "runtime", false, 0); |
541 | (void) do_rotate(s, &s->system_journal, "system", s->seal, 0); | |
d025f1e4 | 542 | |
a33687b7 | 543 | /* Then, rotate all user journals we have open (keeping them open) */ |
43cf8388 | 544 | ORDERED_HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { |
4a0b58c4 | 545 | r = do_rotate(s, &f, "user", s->seal, PTR_TO_UID(k)); |
fc55baee | 546 | if (r >= 0) |
43cf8388 | 547 | ordered_hashmap_replace(s->user_journals, k, f); |
fc55baee ZJS |
548 | else if (!f) |
549 | /* Old file has been closed and deallocated */ | |
43cf8388 | 550 | ordered_hashmap_remove(s->user_journals, k); |
d025f1e4 | 551 | } |
b58c888f | 552 | |
30acbadc LP |
553 | /* Finally, also rotate all user journals we currently do not have open. (But do so only if we actually have |
554 | * access to /var, i.e. are not in the log-to-runtime-journal mode). */ | |
555 | if (!s->runtime_journal && | |
556 | open_user_journal_directory(s, &d, &path) >= 0) { | |
557 | ||
a33687b7 LP |
558 | struct dirent *de; |
559 | ||
560 | FOREACH_DIRENT(de, d, log_warning_errno(errno, "Failed to enumerate %s, ignoring: %m", path)) { | |
561 | _cleanup_free_ char *u = NULL, *full = NULL; | |
562 | _cleanup_close_ int fd = -1; | |
563 | const char *a, *b; | |
564 | uid_t uid; | |
565 | ||
566 | a = startswith(de->d_name, "user-"); | |
567 | if (!a) | |
568 | continue; | |
569 | b = endswith(de->d_name, ".journal"); | |
570 | if (!b) | |
571 | continue; | |
572 | ||
573 | u = strndup(a, b-a); | |
574 | if (!u) { | |
575 | log_oom(); | |
576 | break; | |
577 | } | |
578 | ||
579 | r = parse_uid(u, &uid); | |
580 | if (r < 0) { | |
581 | log_debug_errno(r, "Failed to parse UID from file name '%s', ignoring: %m", de->d_name); | |
582 | continue; | |
583 | } | |
584 | ||
585 | /* Already rotated in the above loop? i.e. is it an open user journal? */ | |
586 | if (ordered_hashmap_contains(s->user_journals, UID_TO_PTR(uid))) | |
587 | continue; | |
588 | ||
589 | full = strjoin(path, de->d_name); | |
590 | if (!full) { | |
591 | log_oom(); | |
592 | break; | |
593 | } | |
594 | ||
595 | fd = openat(dirfd(d), de->d_name, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW|O_NONBLOCK); | |
596 | if (fd < 0) { | |
597 | log_full_errno(IN_SET(errno, ELOOP, ENOENT) ? LOG_DEBUG : LOG_WARNING, errno, | |
598 | "Failed to open journal file '%s' for rotation: %m", full); | |
599 | continue; | |
600 | } | |
601 | ||
602 | /* Make some room in the set of deferred close()s */ | |
603 | server_vacuum_deferred_closes(s); | |
604 | ||
605 | /* Open the file briefly, so that we can archive it */ | |
606 | r = journal_file_open(fd, | |
607 | full, | |
608 | O_RDWR, | |
609 | 0640, | |
610 | s->compress.enabled, | |
611 | s->compress.threshold_bytes, | |
612 | s->seal, | |
613 | &s->system_storage.metrics, | |
614 | s->mmap, | |
615 | s->deferred_closes, | |
616 | NULL, | |
617 | &f); | |
618 | if (r < 0) { | |
619 | log_warning_errno(r, "Failed to read journal file %s for rotation, trying to move it out of the way: %m", full); | |
620 | ||
621 | r = journal_file_dispose(dirfd(d), de->d_name); | |
622 | if (r < 0) | |
623 | log_warning_errno(r, "Failed to move %s out of the way, ignoring: %m", full); | |
624 | else | |
625 | log_debug("Successfully moved %s out of the way.", full); | |
626 | ||
627 | continue; | |
628 | } | |
629 | ||
630 | TAKE_FD(fd); /* Donated to journal_file_open() */ | |
631 | ||
632 | r = journal_file_archive(f); | |
633 | if (r < 0) | |
634 | log_debug_errno(r, "Failed to archive journal file '%s', ignoring: %m", full); | |
635 | ||
636 | f = journal_initiate_close(f, s->deferred_closes); | |
637 | } | |
638 | } | |
639 | ||
f760d8a8 | 640 | server_process_deferred_closes(s); |
d025f1e4 ZJS |
641 | } |
642 | ||
26687bf8 OS |
643 | void server_sync(Server *s) { |
644 | JournalFile *f; | |
26687bf8 OS |
645 | Iterator i; |
646 | int r; | |
647 | ||
26687bf8 | 648 | if (s->system_journal) { |
ac2e41f5 | 649 | r = journal_file_set_offline(s->system_journal, false); |
26687bf8 | 650 | if (r < 0) |
65089b82 | 651 | log_warning_errno(r, "Failed to sync system journal, ignoring: %m"); |
26687bf8 OS |
652 | } |
653 | ||
65c1d46b | 654 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) { |
ac2e41f5 | 655 | r = journal_file_set_offline(f, false); |
26687bf8 | 656 | if (r < 0) |
65089b82 | 657 | log_warning_errno(r, "Failed to sync user journal, ignoring: %m"); |
26687bf8 OS |
658 | } |
659 | ||
f9a810be LP |
660 | if (s->sync_event_source) { |
661 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF); | |
662 | if (r < 0) | |
da927ba9 | 663 | log_error_errno(r, "Failed to disable sync timer source: %m"); |
f9a810be | 664 | } |
26687bf8 OS |
665 | |
666 | s->sync_scheduled = false; | |
667 | } | |
668 | ||
3a19f215 | 669 | static void do_vacuum(Server *s, JournalStorage *storage, bool verbose) { |
ea69bd41 | 670 | |
63c8666b ZJS |
671 | int r; |
672 | ||
8580d1f7 | 673 | assert(s); |
266a4700 | 674 | assert(storage); |
8580d1f7 | 675 | |
57f443a6 | 676 | (void) cache_space_refresh(s, storage); |
18e758bf FB |
677 | |
678 | if (verbose) | |
679 | server_space_usage_message(s, storage); | |
8580d1f7 | 680 | |
57f443a6 FB |
681 | r = journal_directory_vacuum(storage->path, storage->space.limit, |
682 | storage->metrics.n_max_files, s->max_retention_usec, | |
683 | &s->oldest_file_usec, verbose); | |
63c8666b | 684 | if (r < 0 && r != -ENOENT) |
266a4700 FB |
685 | log_warning_errno(r, "Failed to vacuum %s, ignoring: %m", storage->path); |
686 | ||
a0edc477 | 687 | cache_space_invalidate(&storage->space); |
63c8666b ZJS |
688 | } |
689 | ||
3a19f215 | 690 | int server_vacuum(Server *s, bool verbose) { |
8580d1f7 | 691 | assert(s); |
d025f1e4 ZJS |
692 | |
693 | log_debug("Vacuuming..."); | |
694 | ||
695 | s->oldest_file_usec = 0; | |
696 | ||
266a4700 | 697 | if (s->system_journal) |
3a19f215 | 698 | do_vacuum(s, &s->system_storage, verbose); |
266a4700 | 699 | if (s->runtime_journal) |
3a19f215 | 700 | do_vacuum(s, &s->runtime_storage, verbose); |
d025f1e4 | 701 | |
8580d1f7 | 702 | return 0; |
d025f1e4 ZJS |
703 | } |
704 | ||
0c24bb23 LP |
705 | static void server_cache_machine_id(Server *s) { |
706 | sd_id128_t id; | |
707 | int r; | |
708 | ||
709 | assert(s); | |
710 | ||
711 | r = sd_id128_get_machine(&id); | |
712 | if (r < 0) | |
713 | return; | |
714 | ||
715 | sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID=")); | |
716 | } | |
717 | ||
718 | static void server_cache_boot_id(Server *s) { | |
719 | sd_id128_t id; | |
720 | int r; | |
721 | ||
722 | assert(s); | |
723 | ||
724 | r = sd_id128_get_boot(&id); | |
725 | if (r < 0) | |
726 | return; | |
727 | ||
728 | sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID=")); | |
729 | } | |
730 | ||
731 | static void server_cache_hostname(Server *s) { | |
732 | _cleanup_free_ char *t = NULL; | |
733 | char *x; | |
734 | ||
735 | assert(s); | |
736 | ||
737 | t = gethostname_malloc(); | |
738 | if (!t) | |
739 | return; | |
740 | ||
741 | x = strappend("_HOSTNAME=", t); | |
742 | if (!x) | |
743 | return; | |
744 | ||
745 | free(s->hostname_field); | |
746 | s->hostname_field = x; | |
747 | } | |
748 | ||
8531ae70 | 749 | static bool shall_try_append_again(JournalFile *f, int r) { |
6e1045e5 | 750 | switch(r) { |
ae739cc1 | 751 | |
6e1045e5 ZJS |
752 | case -E2BIG: /* Hit configured limit */ |
753 | case -EFBIG: /* Hit fs limit */ | |
754 | case -EDQUOT: /* Quota limit hit */ | |
755 | case -ENOSPC: /* Disk full */ | |
d025f1e4 | 756 | log_debug("%s: Allocation limit reached, rotating.", f->path); |
6e1045e5 | 757 | return true; |
ae739cc1 | 758 | |
6e1045e5 ZJS |
759 | case -EIO: /* I/O error of some kind (mmap) */ |
760 | log_warning("%s: IO error, rotating.", f->path); | |
761 | return true; | |
ae739cc1 | 762 | |
6e1045e5 | 763 | case -EHOSTDOWN: /* Other machine */ |
d025f1e4 | 764 | log_info("%s: Journal file from other machine, rotating.", f->path); |
6e1045e5 | 765 | return true; |
ae739cc1 | 766 | |
6e1045e5 | 767 | case -EBUSY: /* Unclean shutdown */ |
d025f1e4 | 768 | log_info("%s: Unclean shutdown, rotating.", f->path); |
6e1045e5 | 769 | return true; |
ae739cc1 | 770 | |
6e1045e5 | 771 | case -EPROTONOSUPPORT: /* Unsupported feature */ |
d025f1e4 | 772 | log_info("%s: Unsupported feature, rotating.", f->path); |
6e1045e5 | 773 | return true; |
ae739cc1 | 774 | |
6e1045e5 ZJS |
775 | case -EBADMSG: /* Corrupted */ |
776 | case -ENODATA: /* Truncated */ | |
777 | case -ESHUTDOWN: /* Already archived */ | |
d025f1e4 | 778 | log_warning("%s: Journal file corrupted, rotating.", f->path); |
6e1045e5 | 779 | return true; |
ae739cc1 | 780 | |
6e1045e5 | 781 | case -EIDRM: /* Journal file has been deleted */ |
2678031a | 782 | log_warning("%s: Journal file has been deleted, rotating.", f->path); |
6e1045e5 | 783 | return true; |
ae739cc1 LP |
784 | |
785 | case -ETXTBSY: /* Journal file is from the future */ | |
c1a9199e | 786 | log_warning("%s: Journal file is from the future, rotating.", f->path); |
ae739cc1 LP |
787 | return true; |
788 | ||
6e1045e5 | 789 | default: |
d025f1e4 | 790 | return false; |
6e1045e5 | 791 | } |
d025f1e4 ZJS |
792 | } |
793 | ||
da6053d0 | 794 | static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, size_t n, int priority) { |
7c070017 | 795 | bool vacuumed = false, rotate = false; |
0f972d66 | 796 | struct dual_timestamp ts; |
d025f1e4 | 797 | JournalFile *f; |
d025f1e4 ZJS |
798 | int r; |
799 | ||
800 | assert(s); | |
801 | assert(iovec); | |
802 | assert(n > 0); | |
803 | ||
0f972d66 LP |
804 | /* Get the closest, linearized time we have for this log event from the event loop. (Note that we do not use |
805 | * the source time, and not even the time the event was originally seen, but instead simply the time we started | |
806 | * processing it, as we want strictly linear ordering in what we write out.) */ | |
807 | assert_se(sd_event_now(s->event, CLOCK_REALTIME, &ts.realtime) >= 0); | |
808 | assert_se(sd_event_now(s->event, CLOCK_MONOTONIC, &ts.monotonic) >= 0); | |
809 | ||
7c070017 LP |
810 | if (ts.realtime < s->last_realtime_clock) { |
811 | /* When the time jumps backwards, let's immediately rotate. Of course, this should not happen during | |
812 | * regular operation. However, when it does happen, then we should make sure that we start fresh files | |
813 | * to ensure that the entries in the journal files are strictly ordered by time, in order to ensure | |
814 | * bisection works correctly. */ | |
d025f1e4 | 815 | |
7c070017 LP |
816 | log_debug("Time jumped backwards, rotating."); |
817 | rotate = true; | |
818 | } else { | |
819 | ||
820 | f = find_journal(s, uid); | |
821 | if (!f) | |
822 | return; | |
823 | ||
824 | if (journal_file_rotate_suggested(f, s->max_file_usec)) { | |
825 | log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path); | |
826 | rotate = true; | |
827 | } | |
828 | } | |
d025f1e4 | 829 | |
7c070017 | 830 | if (rotate) { |
d025f1e4 | 831 | server_rotate(s); |
3a19f215 | 832 | server_vacuum(s, false); |
d025f1e4 ZJS |
833 | vacuumed = true; |
834 | ||
835 | f = find_journal(s, uid); | |
836 | if (!f) | |
837 | return; | |
838 | } | |
839 | ||
7c070017 LP |
840 | s->last_realtime_clock = ts.realtime; |
841 | ||
d180c349 | 842 | r = journal_file_append_entry(f, &ts, NULL, iovec, n, &s->seqnum, NULL, NULL); |
26687bf8 | 843 | if (r >= 0) { |
d07f7b9e | 844 | server_schedule_sync(s, priority); |
d025f1e4 | 845 | return; |
26687bf8 | 846 | } |
d025f1e4 ZJS |
847 | |
848 | if (vacuumed || !shall_try_append_again(f, r)) { | |
da6053d0 | 849 | log_error_errno(r, "Failed to write entry (%zu items, %zu bytes), ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); |
d025f1e4 ZJS |
850 | return; |
851 | } | |
852 | ||
853 | server_rotate(s); | |
3a19f215 | 854 | server_vacuum(s, false); |
d025f1e4 ZJS |
855 | |
856 | f = find_journal(s, uid); | |
857 | if (!f) | |
858 | return; | |
859 | ||
860 | log_debug("Retrying write."); | |
d180c349 | 861 | r = journal_file_append_entry(f, &ts, NULL, iovec, n, &s->seqnum, NULL, NULL); |
8266e1c0 | 862 | if (r < 0) |
da6053d0 | 863 | log_error_errno(r, "Failed to write entry (%zu items, %zu bytes) despite vacuuming, ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); |
8266e1c0 | 864 | else |
d07f7b9e | 865 | server_schedule_sync(s, priority); |
d025f1e4 ZJS |
866 | } |
867 | ||
22e3a02b LP |
868 | #define IOVEC_ADD_NUMERIC_FIELD(iovec, n, value, type, isset, format, field) \ |
869 | if (isset(value)) { \ | |
870 | char *k; \ | |
fbd0b64f | 871 | k = newa(char, STRLEN(field "=") + DECIMAL_STR_MAX(type) + 1); \ |
22e3a02b | 872 | sprintf(k, field "=" format, value); \ |
e6a7ec4b | 873 | iovec[n++] = IOVEC_MAKE_STRING(k); \ |
22e3a02b | 874 | } |
4b58153d | 875 | |
22e3a02b LP |
876 | #define IOVEC_ADD_STRING_FIELD(iovec, n, value, field) \ |
877 | if (!isempty(value)) { \ | |
878 | char *k; \ | |
879 | k = strjoina(field "=", value); \ | |
e6a7ec4b | 880 | iovec[n++] = IOVEC_MAKE_STRING(k); \ |
22e3a02b | 881 | } |
4b58153d | 882 | |
22e3a02b LP |
883 | #define IOVEC_ADD_ID128_FIELD(iovec, n, value, field) \ |
884 | if (!sd_id128_is_null(value)) { \ | |
885 | char *k; \ | |
fbd0b64f | 886 | k = newa(char, STRLEN(field "=") + SD_ID128_STRING_MAX); \ |
22e3a02b | 887 | sd_id128_to_string(value, stpcpy(k, field "=")); \ |
e6a7ec4b | 888 | iovec[n++] = IOVEC_MAKE_STRING(k); \ |
22e3a02b | 889 | } |
4b58153d | 890 | |
22e3a02b LP |
891 | #define IOVEC_ADD_SIZED_FIELD(iovec, n, value, value_size, field) \ |
892 | if (value_size > 0) { \ | |
893 | char *k; \ | |
fbd0b64f | 894 | k = newa(char, STRLEN(field "=") + value_size + 1); \ |
22e3a02b | 895 | *((char*) mempcpy(stpcpy(k, field "="), value, value_size)) = 0; \ |
e6a7ec4b | 896 | iovec[n++] = IOVEC_MAKE_STRING(k); \ |
22e3a02b | 897 | } \ |
4b58153d | 898 | |
d025f1e4 ZJS |
899 | static void dispatch_message_real( |
900 | Server *s, | |
d3070fbd | 901 | struct iovec *iovec, size_t n, size_t m, |
22e3a02b | 902 | const ClientContext *c, |
3b3154df | 903 | const struct timeval *tv, |
d07f7b9e | 904 | int priority, |
22e3a02b LP |
905 | pid_t object_pid) { |
906 | ||
907 | char source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)]; | |
084eeb86 | 908 | _cleanup_free_ char *cmdline1 = NULL, *cmdline2 = NULL; |
22e3a02b LP |
909 | uid_t journal_uid; |
910 | ClientContext *o; | |
d025f1e4 ZJS |
911 | |
912 | assert(s); | |
913 | assert(iovec); | |
914 | assert(n > 0); | |
d3070fbd LP |
915 | assert(n + |
916 | N_IOVEC_META_FIELDS + | |
917 | (pid_is_valid(object_pid) ? N_IOVEC_OBJECT_FIELDS : 0) + | |
918 | client_context_extra_fields_n_iovec(c) <= m); | |
19cace37 | 919 | |
22e3a02b LP |
920 | if (c) { |
921 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->pid, pid_t, pid_is_valid, PID_FMT, "_PID"); | |
922 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->uid, uid_t, uid_is_valid, UID_FMT, "_UID"); | |
923 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->gid, gid_t, gid_is_valid, GID_FMT, "_GID"); | |
4b58153d | 924 | |
084eeb86 ZJS |
925 | IOVEC_ADD_STRING_FIELD(iovec, n, c->comm, "_COMM"); /* At most TASK_COMM_LENGTH (16 bytes) */ |
926 | IOVEC_ADD_STRING_FIELD(iovec, n, c->exe, "_EXE"); /* A path, so at most PATH_MAX (4096 bytes) */ | |
d025f1e4 | 927 | |
084eeb86 ZJS |
928 | if (c->cmdline) |
929 | /* At most _SC_ARG_MAX (2MB usually), which is too much to put on stack. | |
930 | * Let's use a heap allocation for this one. */ | |
931 | cmdline1 = set_iovec_string_field(iovec, &n, "_CMDLINE=", c->cmdline); | |
ae018d9b | 932 | |
084eeb86 ZJS |
933 | IOVEC_ADD_STRING_FIELD(iovec, n, c->capeff, "_CAP_EFFECTIVE"); /* Read from /proc/.../status */ |
934 | IOVEC_ADD_SIZED_FIELD(iovec, n, c->label, c->label_size, "_SELINUX_CONTEXT"); | |
22e3a02b LP |
935 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->auditid, uint32_t, audit_session_is_valid, "%" PRIu32, "_AUDIT_SESSION"); |
936 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->loginuid, uid_t, uid_is_valid, UID_FMT, "_AUDIT_LOGINUID"); | |
d025f1e4 | 937 | |
084eeb86 | 938 | IOVEC_ADD_STRING_FIELD(iovec, n, c->cgroup, "_SYSTEMD_CGROUP"); /* A path */ |
22e3a02b LP |
939 | IOVEC_ADD_STRING_FIELD(iovec, n, c->session, "_SYSTEMD_SESSION"); |
940 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, c->owner_uid, uid_t, uid_is_valid, UID_FMT, "_SYSTEMD_OWNER_UID"); | |
084eeb86 | 941 | IOVEC_ADD_STRING_FIELD(iovec, n, c->unit, "_SYSTEMD_UNIT"); /* Unit names are bounded by UNIT_NAME_MAX */ |
22e3a02b LP |
942 | IOVEC_ADD_STRING_FIELD(iovec, n, c->user_unit, "_SYSTEMD_USER_UNIT"); |
943 | IOVEC_ADD_STRING_FIELD(iovec, n, c->slice, "_SYSTEMD_SLICE"); | |
944 | IOVEC_ADD_STRING_FIELD(iovec, n, c->user_slice, "_SYSTEMD_USER_SLICE"); | |
e7ff4e7f | 945 | |
22e3a02b | 946 | IOVEC_ADD_ID128_FIELD(iovec, n, c->invocation_id, "_SYSTEMD_INVOCATION_ID"); |
d3070fbd LP |
947 | |
948 | if (c->extra_fields_n_iovec > 0) { | |
949 | memcpy(iovec + n, c->extra_fields_iovec, c->extra_fields_n_iovec * sizeof(struct iovec)); | |
950 | n += c->extra_fields_n_iovec; | |
951 | } | |
d025f1e4 | 952 | } |
968f3196 | 953 | |
22e3a02b | 954 | assert(n <= m); |
968f3196 | 955 | |
22e3a02b | 956 | if (pid_is_valid(object_pid) && client_context_get(s, object_pid, NULL, NULL, 0, NULL, &o) >= 0) { |
968f3196 | 957 | |
22e3a02b LP |
958 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->pid, pid_t, pid_is_valid, PID_FMT, "OBJECT_PID"); |
959 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->uid, uid_t, uid_is_valid, UID_FMT, "OBJECT_UID"); | |
960 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->gid, gid_t, gid_is_valid, GID_FMT, "OBJECT_GID"); | |
968f3196 | 961 | |
084eeb86 | 962 | /* See above for size limits, only ->cmdline may be large, so use a heap allocation for it. */ |
22e3a02b LP |
963 | IOVEC_ADD_STRING_FIELD(iovec, n, o->comm, "OBJECT_COMM"); |
964 | IOVEC_ADD_STRING_FIELD(iovec, n, o->exe, "OBJECT_EXE"); | |
084eeb86 ZJS |
965 | if (o->cmdline) |
966 | cmdline2 = set_iovec_string_field(iovec, &n, "OBJECT_CMDLINE=", o->cmdline); | |
968f3196 | 967 | |
084eeb86 | 968 | IOVEC_ADD_STRING_FIELD(iovec, n, o->capeff, "OBJECT_CAP_EFFECTIVE"); |
22e3a02b | 969 | IOVEC_ADD_SIZED_FIELD(iovec, n, o->label, o->label_size, "OBJECT_SELINUX_CONTEXT"); |
22e3a02b LP |
970 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->auditid, uint32_t, audit_session_is_valid, "%" PRIu32, "OBJECT_AUDIT_SESSION"); |
971 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->loginuid, uid_t, uid_is_valid, UID_FMT, "OBJECT_AUDIT_LOGINUID"); | |
d473176a | 972 | |
22e3a02b LP |
973 | IOVEC_ADD_STRING_FIELD(iovec, n, o->cgroup, "OBJECT_SYSTEMD_CGROUP"); |
974 | IOVEC_ADD_STRING_FIELD(iovec, n, o->session, "OBJECT_SYSTEMD_SESSION"); | |
975 | IOVEC_ADD_NUMERIC_FIELD(iovec, n, o->owner_uid, uid_t, uid_is_valid, UID_FMT, "OBJECT_SYSTEMD_OWNER_UID"); | |
976 | IOVEC_ADD_STRING_FIELD(iovec, n, o->unit, "OBJECT_SYSTEMD_UNIT"); | |
977 | IOVEC_ADD_STRING_FIELD(iovec, n, o->user_unit, "OBJECT_SYSTEMD_USER_UNIT"); | |
978 | IOVEC_ADD_STRING_FIELD(iovec, n, o->slice, "OBJECT_SYSTEMD_SLICE"); | |
979 | IOVEC_ADD_STRING_FIELD(iovec, n, o->user_slice, "OBJECT_SYSTEMD_USER_SLICE"); | |
d473176a | 980 | |
22e3a02b | 981 | IOVEC_ADD_ID128_FIELD(iovec, n, o->invocation_id, "OBJECT_SYSTEMD_INVOCATION_ID="); |
968f3196 | 982 | } |
22e3a02b | 983 | |
968f3196 | 984 | assert(n <= m); |
d025f1e4 ZJS |
985 | |
986 | if (tv) { | |
398a50cd | 987 | sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=" USEC_FMT, timeval_load(tv)); |
e6a7ec4b | 988 | iovec[n++] = IOVEC_MAKE_STRING(source_time); |
d025f1e4 ZJS |
989 | } |
990 | ||
991 | /* Note that strictly speaking storing the boot id here is | |
992 | * redundant since the entry includes this in-line | |
993 | * anyway. However, we need this indexed, too. */ | |
0c24bb23 | 994 | if (!isempty(s->boot_id_field)) |
e6a7ec4b | 995 | iovec[n++] = IOVEC_MAKE_STRING(s->boot_id_field); |
d025f1e4 | 996 | |
0c24bb23 | 997 | if (!isempty(s->machine_id_field)) |
e6a7ec4b | 998 | iovec[n++] = IOVEC_MAKE_STRING(s->machine_id_field); |
d025f1e4 | 999 | |
0c24bb23 | 1000 | if (!isempty(s->hostname_field)) |
e6a7ec4b | 1001 | iovec[n++] = IOVEC_MAKE_STRING(s->hostname_field); |
d025f1e4 ZJS |
1002 | |
1003 | assert(n <= m); | |
1004 | ||
22e3a02b LP |
1005 | if (s->split_mode == SPLIT_UID && c && uid_is_valid(c->uid)) |
1006 | /* Split up strictly by (non-root) UID */ | |
1007 | journal_uid = c->uid; | |
1008 | else if (s->split_mode == SPLIT_LOGIN && c && c->uid > 0 && uid_is_valid(c->owner_uid)) | |
edc3797f LP |
1009 | /* Split up by login UIDs. We do this only if the |
1010 | * realuid is not root, in order not to accidentally | |
1011 | * leak privileged information to the user that is | |
1012 | * logged by a privileged process that is part of an | |
7517e174 | 1013 | * unprivileged session. */ |
22e3a02b | 1014 | journal_uid = c->owner_uid; |
da499392 KS |
1015 | else |
1016 | journal_uid = 0; | |
759c945a | 1017 | |
d07f7b9e | 1018 | write_to_journal(s, journal_uid, iovec, n, priority); |
d025f1e4 ZJS |
1019 | } |
1020 | ||
13181942 | 1021 | void server_driver_message(Server *s, pid_t object_pid, const char *message_id, const char *format, ...) { |
22e3a02b | 1022 | |
d3070fbd LP |
1023 | struct iovec *iovec; |
1024 | size_t n = 0, k, m; | |
d025f1e4 | 1025 | va_list ap; |
22e3a02b | 1026 | int r; |
d025f1e4 ZJS |
1027 | |
1028 | assert(s); | |
1029 | assert(format); | |
1030 | ||
f643ae71 | 1031 | m = N_IOVEC_META_FIELDS + 5 + N_IOVEC_PAYLOAD_FIELDS + client_context_extra_fields_n_iovec(s->my_context) + N_IOVEC_OBJECT_FIELDS; |
d3070fbd LP |
1032 | iovec = newa(struct iovec, m); |
1033 | ||
4850d39a | 1034 | assert_cc(3 == LOG_FAC(LOG_DAEMON)); |
e6a7ec4b LP |
1035 | iovec[n++] = IOVEC_MAKE_STRING("SYSLOG_FACILITY=3"); |
1036 | iovec[n++] = IOVEC_MAKE_STRING("SYSLOG_IDENTIFIER=systemd-journald"); | |
b6fa2555 | 1037 | |
e6a7ec4b | 1038 | iovec[n++] = IOVEC_MAKE_STRING("_TRANSPORT=driver"); |
4850d39a | 1039 | assert_cc(6 == LOG_INFO); |
e6a7ec4b | 1040 | iovec[n++] = IOVEC_MAKE_STRING("PRIORITY=6"); |
d025f1e4 | 1041 | |
2b044526 | 1042 | if (message_id) |
e6a7ec4b | 1043 | iovec[n++] = IOVEC_MAKE_STRING(message_id); |
d3070fbd | 1044 | k = n; |
8a03c9ef ZJS |
1045 | |
1046 | va_start(ap, format); | |
d3070fbd | 1047 | r = log_format_iovec(iovec, m, &n, false, 0, format, ap); |
32917e33 | 1048 | /* Error handling below */ |
8a03c9ef ZJS |
1049 | va_end(ap); |
1050 | ||
32917e33 | 1051 | if (r >= 0) |
d3070fbd | 1052 | dispatch_message_real(s, iovec, n, m, s->my_context, NULL, LOG_INFO, object_pid); |
8a03c9ef | 1053 | |
d3070fbd LP |
1054 | while (k < n) |
1055 | free(iovec[k++].iov_base); | |
32917e33 ZJS |
1056 | |
1057 | if (r < 0) { | |
1058 | /* We failed to format the message. Emit a warning instead. */ | |
1059 | char buf[LINE_MAX]; | |
1060 | ||
1061 | xsprintf(buf, "MESSAGE=Entry printing failed: %s", strerror(-r)); | |
1062 | ||
1063 | n = 3; | |
e6a7ec4b LP |
1064 | iovec[n++] = IOVEC_MAKE_STRING("PRIORITY=4"); |
1065 | iovec[n++] = IOVEC_MAKE_STRING(buf); | |
d3070fbd | 1066 | dispatch_message_real(s, iovec, n, m, s->my_context, NULL, LOG_INFO, object_pid); |
32917e33 | 1067 | } |
d025f1e4 ZJS |
1068 | } |
1069 | ||
1070 | void server_dispatch_message( | |
1071 | Server *s, | |
d3070fbd | 1072 | struct iovec *iovec, size_t n, size_t m, |
22e3a02b | 1073 | ClientContext *c, |
3b3154df | 1074 | const struct timeval *tv, |
968f3196 ZJS |
1075 | int priority, |
1076 | pid_t object_pid) { | |
d025f1e4 | 1077 | |
8580d1f7 | 1078 | uint64_t available = 0; |
22e3a02b | 1079 | int rl; |
d025f1e4 ZJS |
1080 | |
1081 | assert(s); | |
1082 | assert(iovec || n == 0); | |
1083 | ||
1084 | if (n == 0) | |
1085 | return; | |
1086 | ||
1087 | if (LOG_PRI(priority) > s->max_level_store) | |
1088 | return; | |
1089 | ||
2f5df74a HHPF |
1090 | /* Stop early in case the information will not be stored |
1091 | * in a journal. */ | |
1092 | if (s->storage == STORAGE_NONE) | |
1093 | return; | |
1094 | ||
22e3a02b LP |
1095 | if (c && c->unit) { |
1096 | (void) determine_space(s, &available, NULL); | |
d025f1e4 | 1097 | |
90fc172e | 1098 | rl = journal_rate_limit_test(s->rate_limit, c->unit, c->log_rate_limit_interval, c->log_rate_limit_burst, priority & LOG_PRIMASK, available); |
22e3a02b LP |
1099 | if (rl == 0) |
1100 | return; | |
d025f1e4 | 1101 | |
22e3a02b LP |
1102 | /* Write a suppression message if we suppressed something */ |
1103 | if (rl > 1) | |
13181942 LP |
1104 | server_driver_message(s, c->pid, |
1105 | "MESSAGE_ID=" SD_MESSAGE_JOURNAL_DROPPED_STR, | |
1106 | LOG_MESSAGE("Suppressed %i messages from %s", rl - 1, c->unit), | |
5908ff1c | 1107 | "N_DROPPED=%i", rl - 1, |
22e3a02b | 1108 | NULL); |
d025f1e4 ZJS |
1109 | } |
1110 | ||
22e3a02b | 1111 | dispatch_message_real(s, iovec, n, m, c, tv, priority, object_pid); |
d025f1e4 ZJS |
1112 | } |
1113 | ||
f78273c8 | 1114 | int server_flush_to_var(Server *s, bool require_flag_file) { |
d025f1e4 ZJS |
1115 | sd_id128_t machine; |
1116 | sd_journal *j = NULL; | |
fbb63411 LP |
1117 | char ts[FORMAT_TIMESPAN_MAX]; |
1118 | usec_t start; | |
1119 | unsigned n = 0; | |
1120 | int r; | |
d025f1e4 ZJS |
1121 | |
1122 | assert(s); | |
1123 | ||
f78273c8 | 1124 | if (!IN_SET(s->storage, STORAGE_AUTO, STORAGE_PERSISTENT)) |
d025f1e4 ZJS |
1125 | return 0; |
1126 | ||
1127 | if (!s->runtime_journal) | |
1128 | return 0; | |
1129 | ||
f78273c8 LP |
1130 | if (require_flag_file && !flushed_flag_is_set()) |
1131 | return 0; | |
1132 | ||
8580d1f7 | 1133 | (void) system_journal_open(s, true); |
d025f1e4 ZJS |
1134 | |
1135 | if (!s->system_journal) | |
1136 | return 0; | |
1137 | ||
1138 | log_debug("Flushing to /var..."); | |
1139 | ||
fbb63411 LP |
1140 | start = now(CLOCK_MONOTONIC); |
1141 | ||
d025f1e4 | 1142 | r = sd_id128_get_machine(&machine); |
00a16861 | 1143 | if (r < 0) |
d025f1e4 | 1144 | return r; |
d025f1e4 ZJS |
1145 | |
1146 | r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); | |
23bbb0de MS |
1147 | if (r < 0) |
1148 | return log_error_errno(r, "Failed to read runtime journal: %m"); | |
d025f1e4 | 1149 | |
93b73b06 LP |
1150 | sd_journal_set_data_threshold(j, 0); |
1151 | ||
d025f1e4 ZJS |
1152 | SD_JOURNAL_FOREACH(j) { |
1153 | Object *o = NULL; | |
1154 | JournalFile *f; | |
1155 | ||
1156 | f = j->current_file; | |
1157 | assert(f && f->current_offset > 0); | |
1158 | ||
fbb63411 LP |
1159 | n++; |
1160 | ||
d025f1e4 ZJS |
1161 | r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); |
1162 | if (r < 0) { | |
da927ba9 | 1163 | log_error_errno(r, "Can't read entry: %m"); |
d025f1e4 ZJS |
1164 | goto finish; |
1165 | } | |
1166 | ||
5a271b08 | 1167 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset); |
d025f1e4 ZJS |
1168 | if (r >= 0) |
1169 | continue; | |
1170 | ||
1171 | if (!shall_try_append_again(s->system_journal, r)) { | |
da927ba9 | 1172 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1173 | goto finish; |
1174 | } | |
1175 | ||
1176 | server_rotate(s); | |
3a19f215 | 1177 | server_vacuum(s, false); |
d025f1e4 | 1178 | |
253f59df LP |
1179 | if (!s->system_journal) { |
1180 | log_notice("Didn't flush runtime journal since rotation of system journal wasn't successful."); | |
1181 | r = -EIO; | |
1182 | goto finish; | |
1183 | } | |
1184 | ||
d025f1e4 | 1185 | log_debug("Retrying write."); |
5a271b08 | 1186 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset); |
d025f1e4 | 1187 | if (r < 0) { |
da927ba9 | 1188 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1189 | goto finish; |
1190 | } | |
1191 | } | |
1192 | ||
804ae586 LP |
1193 | r = 0; |
1194 | ||
d025f1e4 | 1195 | finish: |
fd790d6f RM |
1196 | if (s->system_journal) |
1197 | journal_file_post_change(s->system_journal); | |
d025f1e4 | 1198 | |
804ae586 | 1199 | s->runtime_journal = journal_file_close(s->runtime_journal); |
d025f1e4 ZJS |
1200 | |
1201 | if (r >= 0) | |
c6878637 | 1202 | (void) rm_rf("/run/log/journal", REMOVE_ROOT); |
d025f1e4 | 1203 | |
763c7aa2 | 1204 | sd_journal_close(j); |
d025f1e4 | 1205 | |
13181942 | 1206 | server_driver_message(s, 0, NULL, |
8a03c9ef ZJS |
1207 | LOG_MESSAGE("Time spent on flushing to /var is %s for %u entries.", |
1208 | format_timespan(ts, sizeof(ts), now(CLOCK_MONOTONIC) - start, 0), | |
1209 | n), | |
1210 | NULL); | |
fbb63411 | 1211 | |
d025f1e4 ZJS |
1212 | return r; |
1213 | } | |
1214 | ||
8531ae70 | 1215 | int server_process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
f9a810be | 1216 | Server *s = userdata; |
a315ac4e LP |
1217 | struct ucred *ucred = NULL; |
1218 | struct timeval *tv = NULL; | |
1219 | struct cmsghdr *cmsg; | |
1220 | char *label = NULL; | |
1221 | size_t label_len = 0, m; | |
1222 | struct iovec iovec; | |
1223 | ssize_t n; | |
1224 | int *fds = NULL, v = 0; | |
da6053d0 | 1225 | size_t n_fds = 0; |
a315ac4e LP |
1226 | |
1227 | union { | |
1228 | struct cmsghdr cmsghdr; | |
1229 | ||
1230 | /* We use NAME_MAX space for the SELinux label | |
1231 | * here. The kernel currently enforces no | |
1232 | * limit, but according to suggestions from | |
1233 | * the SELinux people this will change and it | |
1234 | * will probably be identical to NAME_MAX. For | |
1235 | * now we use that, but this should be updated | |
1236 | * one day when the final limit is known. */ | |
1237 | uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) + | |
1238 | CMSG_SPACE(sizeof(struct timeval)) + | |
1239 | CMSG_SPACE(sizeof(int)) + /* fd */ | |
1240 | CMSG_SPACE(NAME_MAX)]; /* selinux label */ | |
1241 | } control = {}; | |
1242 | ||
1243 | union sockaddr_union sa = {}; | |
1244 | ||
1245 | struct msghdr msghdr = { | |
1246 | .msg_iov = &iovec, | |
1247 | .msg_iovlen = 1, | |
1248 | .msg_control = &control, | |
1249 | .msg_controllen = sizeof(control), | |
1250 | .msg_name = &sa, | |
1251 | .msg_namelen = sizeof(sa), | |
1252 | }; | |
f9a810be | 1253 | |
d025f1e4 | 1254 | assert(s); |
875c2e22 | 1255 | assert(fd == s->native_fd || fd == s->syslog_fd || fd == s->audit_fd); |
f9a810be | 1256 | |
baaa35ad ZJS |
1257 | if (revents != EPOLLIN) |
1258 | return log_error_errno(SYNTHETIC_ERRNO(EIO), | |
1259 | "Got invalid event from epoll for datagram fd: %" PRIx32, | |
1260 | revents); | |
f9a810be | 1261 | |
22e3a02b LP |
1262 | /* Try to get the right size, if we can. (Not all sockets support SIOCINQ, hence we just try, but don't rely on |
1263 | * it.) */ | |
a315ac4e | 1264 | (void) ioctl(fd, SIOCINQ, &v); |
d025f1e4 | 1265 | |
a315ac4e LP |
1266 | /* Fix it up, if it is too small. We use the same fixed value as auditd here. Awful! */ |
1267 | m = PAGE_ALIGN(MAX3((size_t) v + 1, | |
1268 | (size_t) LINE_MAX, | |
1269 | ALIGN(sizeof(struct nlmsghdr)) + ALIGN((size_t) MAX_AUDIT_MESSAGE_LENGTH)) + 1); | |
d025f1e4 | 1270 | |
a315ac4e LP |
1271 | if (!GREEDY_REALLOC(s->buffer, s->buffer_size, m)) |
1272 | return log_oom(); | |
875c2e22 | 1273 | |
5cfa2c3d | 1274 | iovec = IOVEC_MAKE(s->buffer, s->buffer_size - 1); /* Leave room for trailing NUL we add later */ |
d025f1e4 | 1275 | |
a315ac4e LP |
1276 | n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); |
1277 | if (n < 0) { | |
3742095b | 1278 | if (IN_SET(errno, EINTR, EAGAIN)) |
a315ac4e | 1279 | return 0; |
875c2e22 | 1280 | |
a315ac4e LP |
1281 | return log_error_errno(errno, "recvmsg() failed: %m"); |
1282 | } | |
875c2e22 | 1283 | |
bc2762a3 | 1284 | CMSG_FOREACH(cmsg, &msghdr) |
a315ac4e LP |
1285 | if (cmsg->cmsg_level == SOL_SOCKET && |
1286 | cmsg->cmsg_type == SCM_CREDENTIALS && | |
1287 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) | |
1288 | ucred = (struct ucred*) CMSG_DATA(cmsg); | |
1289 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1290 | cmsg->cmsg_type == SCM_SECURITY) { | |
1291 | label = (char*) CMSG_DATA(cmsg); | |
1292 | label_len = cmsg->cmsg_len - CMSG_LEN(0); | |
1293 | } else if (cmsg->cmsg_level == SOL_SOCKET && | |
1294 | cmsg->cmsg_type == SO_TIMESTAMP && | |
1295 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval))) | |
1296 | tv = (struct timeval*) CMSG_DATA(cmsg); | |
1297 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1298 | cmsg->cmsg_type == SCM_RIGHTS) { | |
1299 | fds = (int*) CMSG_DATA(cmsg); | |
1300 | n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); | |
d025f1e4 ZJS |
1301 | } |
1302 | ||
a315ac4e LP |
1303 | /* And a trailing NUL, just in case */ |
1304 | s->buffer[n] = 0; | |
1305 | ||
1306 | if (fd == s->syslog_fd) { | |
1307 | if (n > 0 && n_fds == 0) | |
bb3ff70a | 1308 | server_process_syslog_message(s, s->buffer, n, ucred, tv, label, label_len); |
a315ac4e LP |
1309 | else if (n_fds > 0) |
1310 | log_warning("Got file descriptors via syslog socket. Ignoring."); | |
1311 | ||
1312 | } else if (fd == s->native_fd) { | |
1313 | if (n > 0 && n_fds == 0) | |
1314 | server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len); | |
1315 | else if (n == 0 && n_fds == 1) | |
1316 | server_process_native_file(s, fds[0], ucred, tv, label, label_len); | |
1317 | else if (n_fds > 0) | |
1318 | log_warning("Got too many file descriptors via native socket. Ignoring."); | |
1319 | ||
1320 | } else { | |
1321 | assert(fd == s->audit_fd); | |
1322 | ||
1323 | if (n > 0 && n_fds == 0) | |
1324 | server_process_audit_message(s, s->buffer, n, ucred, &sa, msghdr.msg_namelen); | |
1325 | else if (n_fds > 0) | |
1326 | log_warning("Got file descriptors via audit socket. Ignoring."); | |
f9a810be | 1327 | } |
a315ac4e LP |
1328 | |
1329 | close_many(fds, n_fds); | |
1330 | return 0; | |
f9a810be | 1331 | } |
d025f1e4 | 1332 | |
f9a810be LP |
1333 | static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1334 | Server *s = userdata; | |
33d52ab9 | 1335 | int r; |
d025f1e4 | 1336 | |
f9a810be | 1337 | assert(s); |
d025f1e4 | 1338 | |
94b65516 | 1339 | log_info("Received request to flush runtime journal from PID " PID_FMT, si->ssi_pid); |
d025f1e4 | 1340 | |
f78273c8 | 1341 | (void) server_flush_to_var(s, false); |
f9a810be | 1342 | server_sync(s); |
3a19f215 | 1343 | server_vacuum(s, false); |
d025f1e4 | 1344 | |
33d52ab9 LP |
1345 | r = touch("/run/systemd/journal/flushed"); |
1346 | if (r < 0) | |
1347 | log_warning_errno(r, "Failed to touch /run/systemd/journal/flushed, ignoring: %m"); | |
74055aa7 | 1348 | |
18e758bf | 1349 | server_space_usage_message(s, NULL); |
f9a810be LP |
1350 | return 0; |
1351 | } | |
d025f1e4 | 1352 | |
f9a810be LP |
1353 | static int dispatch_sigusr2(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1354 | Server *s = userdata; | |
33d52ab9 | 1355 | int r; |
d025f1e4 | 1356 | |
f9a810be | 1357 | assert(s); |
d025f1e4 | 1358 | |
94b65516 | 1359 | log_info("Received request to rotate journal from PID " PID_FMT, si->ssi_pid); |
f9a810be | 1360 | server_rotate(s); |
3a19f215 FB |
1361 | server_vacuum(s, true); |
1362 | ||
1363 | if (s->system_journal) | |
1364 | patch_min_use(&s->system_storage); | |
1365 | if (s->runtime_journal) | |
1366 | patch_min_use(&s->runtime_storage); | |
d025f1e4 | 1367 | |
dbd6e31c | 1368 | /* Let clients know when the most recent rotation happened. */ |
33d52ab9 LP |
1369 | r = write_timestamp_file_atomic("/run/systemd/journal/rotated", now(CLOCK_MONOTONIC)); |
1370 | if (r < 0) | |
1371 | log_warning_errno(r, "Failed to write /run/systemd/journal/rotated, ignoring: %m"); | |
dbd6e31c | 1372 | |
f9a810be LP |
1373 | return 0; |
1374 | } | |
d025f1e4 | 1375 | |
f9a810be LP |
1376 | static int dispatch_sigterm(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1377 | Server *s = userdata; | |
d025f1e4 | 1378 | |
f9a810be | 1379 | assert(s); |
d025f1e4 | 1380 | |
4daf54a8 | 1381 | log_received_signal(LOG_INFO, si); |
d025f1e4 | 1382 | |
6203e07a | 1383 | sd_event_exit(s->event, 0); |
d025f1e4 ZJS |
1384 | return 0; |
1385 | } | |
1386 | ||
94b65516 LP |
1387 | static int dispatch_sigrtmin1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1388 | Server *s = userdata; | |
33d52ab9 | 1389 | int r; |
94b65516 LP |
1390 | |
1391 | assert(s); | |
1392 | ||
1393 | log_debug("Received request to sync from PID " PID_FMT, si->ssi_pid); | |
1394 | ||
1395 | server_sync(s); | |
1396 | ||
1397 | /* Let clients know when the most recent sync happened. */ | |
33d52ab9 LP |
1398 | r = write_timestamp_file_atomic("/run/systemd/journal/synced", now(CLOCK_MONOTONIC)); |
1399 | if (r < 0) | |
1400 | log_warning_errno(r, "Failed to write /run/systemd/journal/synced, ignoring: %m"); | |
94b65516 LP |
1401 | |
1402 | return 0; | |
1403 | } | |
1404 | ||
f9a810be | 1405 | static int setup_signals(Server *s) { |
f9a810be | 1406 | int r; |
d025f1e4 ZJS |
1407 | |
1408 | assert(s); | |
1409 | ||
9bab3b65 | 1410 | assert_se(sigprocmask_many(SIG_SETMASK, NULL, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, SIGRTMIN+1, -1) >= 0); |
d025f1e4 | 1411 | |
151b9b96 | 1412 | r = sd_event_add_signal(s->event, &s->sigusr1_event_source, SIGUSR1, dispatch_sigusr1, s); |
f9a810be LP |
1413 | if (r < 0) |
1414 | return r; | |
1415 | ||
151b9b96 | 1416 | r = sd_event_add_signal(s->event, &s->sigusr2_event_source, SIGUSR2, dispatch_sigusr2, s); |
f9a810be LP |
1417 | if (r < 0) |
1418 | return r; | |
d025f1e4 | 1419 | |
151b9b96 | 1420 | r = sd_event_add_signal(s->event, &s->sigterm_event_source, SIGTERM, dispatch_sigterm, s); |
f9a810be LP |
1421 | if (r < 0) |
1422 | return r; | |
d025f1e4 | 1423 | |
337fabf7 | 1424 | /* Let's process SIGTERM late, so that we flush all queued messages to disk before we exit */ |
b374689c LP |
1425 | r = sd_event_source_set_priority(s->sigterm_event_source, SD_EVENT_PRIORITY_NORMAL+20); |
1426 | if (r < 0) | |
1427 | return r; | |
1428 | ||
337fabf7 LP |
1429 | /* When journald is invoked on the terminal (when debugging), it's useful if C-c is handled |
1430 | * equivalent to SIGTERM. */ | |
151b9b96 | 1431 | r = sd_event_add_signal(s->event, &s->sigint_event_source, SIGINT, dispatch_sigterm, s); |
f9a810be LP |
1432 | if (r < 0) |
1433 | return r; | |
d025f1e4 | 1434 | |
b374689c LP |
1435 | r = sd_event_source_set_priority(s->sigint_event_source, SD_EVENT_PRIORITY_NORMAL+20); |
1436 | if (r < 0) | |
1437 | return r; | |
1438 | ||
337fabf7 LP |
1439 | /* SIGRTMIN+1 causes an immediate sync. We process this very late, so that everything else queued at |
1440 | * this point is really written to disk. Clients can watch /run/systemd/journal/synced with inotify | |
1441 | * until its mtime changes to see when a sync happened. */ | |
94b65516 LP |
1442 | r = sd_event_add_signal(s->event, &s->sigrtmin1_event_source, SIGRTMIN+1, dispatch_sigrtmin1, s); |
1443 | if (r < 0) | |
1444 | return r; | |
1445 | ||
1446 | r = sd_event_source_set_priority(s->sigrtmin1_event_source, SD_EVENT_PRIORITY_NORMAL+15); | |
1447 | if (r < 0) | |
1448 | return r; | |
1449 | ||
d025f1e4 ZJS |
1450 | return 0; |
1451 | } | |
1452 | ||
5707ecf3 ZJS |
1453 | static int parse_proc_cmdline_item(const char *key, const char *value, void *data) { |
1454 | Server *s = data; | |
74df0fca | 1455 | int r; |
d025f1e4 | 1456 | |
5707ecf3 | 1457 | assert(s); |
d025f1e4 | 1458 | |
1d84ad94 LP |
1459 | if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_syslog")) { |
1460 | ||
5707ecf3 | 1461 | r = value ? parse_boolean(value) : true; |
d581d9d9 | 1462 | if (r < 0) |
5707ecf3 ZJS |
1463 | log_warning("Failed to parse forward to syslog switch \"%s\". Ignoring.", value); |
1464 | else | |
1465 | s->forward_to_syslog = r; | |
1d84ad94 LP |
1466 | |
1467 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_kmsg")) { | |
1468 | ||
5707ecf3 ZJS |
1469 | r = value ? parse_boolean(value) : true; |
1470 | if (r < 0) | |
1471 | log_warning("Failed to parse forward to kmsg switch \"%s\". Ignoring.", value); | |
1472 | else | |
1473 | s->forward_to_kmsg = r; | |
1d84ad94 LP |
1474 | |
1475 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_console")) { | |
1476 | ||
5707ecf3 ZJS |
1477 | r = value ? parse_boolean(value) : true; |
1478 | if (r < 0) | |
1479 | log_warning("Failed to parse forward to console switch \"%s\". Ignoring.", value); | |
1480 | else | |
1481 | s->forward_to_console = r; | |
1d84ad94 LP |
1482 | |
1483 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_wall")) { | |
1484 | ||
5707ecf3 ZJS |
1485 | r = value ? parse_boolean(value) : true; |
1486 | if (r < 0) | |
1487 | log_warning("Failed to parse forward to wall switch \"%s\". Ignoring.", value); | |
1488 | else | |
1489 | s->forward_to_wall = r; | |
1d84ad94 LP |
1490 | |
1491 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_console")) { | |
1492 | ||
1493 | if (proc_cmdline_value_missing(key, value)) | |
1494 | return 0; | |
1495 | ||
5707ecf3 ZJS |
1496 | r = log_level_from_string(value); |
1497 | if (r < 0) | |
1498 | log_warning("Failed to parse max level console value \"%s\". Ignoring.", value); | |
1499 | else | |
1500 | s->max_level_console = r; | |
1d84ad94 LP |
1501 | |
1502 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_store")) { | |
1503 | ||
1504 | if (proc_cmdline_value_missing(key, value)) | |
1505 | return 0; | |
1506 | ||
5707ecf3 ZJS |
1507 | r = log_level_from_string(value); |
1508 | if (r < 0) | |
1509 | log_warning("Failed to parse max level store value \"%s\". Ignoring.", value); | |
1510 | else | |
1511 | s->max_level_store = r; | |
1d84ad94 LP |
1512 | |
1513 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_syslog")) { | |
1514 | ||
1515 | if (proc_cmdline_value_missing(key, value)) | |
1516 | return 0; | |
1517 | ||
5707ecf3 ZJS |
1518 | r = log_level_from_string(value); |
1519 | if (r < 0) | |
1520 | log_warning("Failed to parse max level syslog value \"%s\". Ignoring.", value); | |
1521 | else | |
1522 | s->max_level_syslog = r; | |
1d84ad94 LP |
1523 | |
1524 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_kmsg")) { | |
1525 | ||
1526 | if (proc_cmdline_value_missing(key, value)) | |
1527 | return 0; | |
1528 | ||
5707ecf3 ZJS |
1529 | r = log_level_from_string(value); |
1530 | if (r < 0) | |
1531 | log_warning("Failed to parse max level kmsg value \"%s\". Ignoring.", value); | |
1532 | else | |
1533 | s->max_level_kmsg = r; | |
1d84ad94 LP |
1534 | |
1535 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_wall")) { | |
1536 | ||
1537 | if (proc_cmdline_value_missing(key, value)) | |
1538 | return 0; | |
1539 | ||
5707ecf3 ZJS |
1540 | r = log_level_from_string(value); |
1541 | if (r < 0) | |
1542 | log_warning("Failed to parse max level wall value \"%s\". Ignoring.", value); | |
1543 | else | |
1544 | s->max_level_wall = r; | |
1d84ad94 | 1545 | |
5707ecf3 ZJS |
1546 | } else if (startswith(key, "systemd.journald")) |
1547 | log_warning("Unknown journald kernel command line option \"%s\". Ignoring.", key); | |
d025f1e4 | 1548 | |
804ae586 | 1549 | /* do not warn about state here, since probably systemd already did */ |
db91ea32 | 1550 | return 0; |
d025f1e4 ZJS |
1551 | } |
1552 | ||
1553 | static int server_parse_config_file(Server *s) { | |
d025f1e4 ZJS |
1554 | assert(s); |
1555 | ||
43688c49 | 1556 | return config_parse_many_nulstr(PKGSYSCONFDIR "/journald.conf", |
da412854 YW |
1557 | CONF_PATHS_NULSTR("systemd/journald.conf.d"), |
1558 | "Journal\0", | |
1559 | config_item_perf_lookup, journald_gperf_lookup, | |
bcde742e | 1560 | CONFIG_PARSE_WARN, s); |
d025f1e4 ZJS |
1561 | } |
1562 | ||
f9a810be LP |
1563 | static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) { |
1564 | Server *s = userdata; | |
26687bf8 OS |
1565 | |
1566 | assert(s); | |
1567 | ||
f9a810be | 1568 | server_sync(s); |
26687bf8 OS |
1569 | return 0; |
1570 | } | |
1571 | ||
d07f7b9e | 1572 | int server_schedule_sync(Server *s, int priority) { |
26687bf8 OS |
1573 | int r; |
1574 | ||
26687bf8 OS |
1575 | assert(s); |
1576 | ||
d07f7b9e LP |
1577 | if (priority <= LOG_CRIT) { |
1578 | /* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */ | |
1579 | server_sync(s); | |
1580 | return 0; | |
1581 | } | |
1582 | ||
26687bf8 OS |
1583 | if (s->sync_scheduled) |
1584 | return 0; | |
1585 | ||
f9a810be LP |
1586 | if (s->sync_interval_usec > 0) { |
1587 | usec_t when; | |
ca267016 | 1588 | |
6a0f1f6d | 1589 | r = sd_event_now(s->event, CLOCK_MONOTONIC, &when); |
f9a810be LP |
1590 | if (r < 0) |
1591 | return r; | |
26687bf8 | 1592 | |
f9a810be LP |
1593 | when += s->sync_interval_usec; |
1594 | ||
1595 | if (!s->sync_event_source) { | |
6a0f1f6d LP |
1596 | r = sd_event_add_time( |
1597 | s->event, | |
1598 | &s->sync_event_source, | |
1599 | CLOCK_MONOTONIC, | |
1600 | when, 0, | |
1601 | server_dispatch_sync, s); | |
f9a810be LP |
1602 | if (r < 0) |
1603 | return r; | |
1604 | ||
1605 | r = sd_event_source_set_priority(s->sync_event_source, SD_EVENT_PRIORITY_IMPORTANT); | |
1606 | } else { | |
1607 | r = sd_event_source_set_time(s->sync_event_source, when); | |
1608 | if (r < 0) | |
1609 | return r; | |
1610 | ||
1611 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_ONESHOT); | |
1612 | } | |
26687bf8 | 1613 | if (r < 0) |
f9a810be | 1614 | return r; |
26687bf8 | 1615 | |
f9a810be LP |
1616 | s->sync_scheduled = true; |
1617 | } | |
26687bf8 OS |
1618 | |
1619 | return 0; | |
1620 | } | |
1621 | ||
0c24bb23 LP |
1622 | static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1623 | Server *s = userdata; | |
1624 | ||
1625 | assert(s); | |
1626 | ||
1627 | server_cache_hostname(s); | |
1628 | return 0; | |
1629 | } | |
1630 | ||
1631 | static int server_open_hostname(Server *s) { | |
1632 | int r; | |
1633 | ||
1634 | assert(s); | |
1635 | ||
db4a47e9 LP |
1636 | s->hostname_fd = open("/proc/sys/kernel/hostname", |
1637 | O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); | |
4a62c710 MS |
1638 | if (s->hostname_fd < 0) |
1639 | return log_error_errno(errno, "Failed to open /proc/sys/kernel/hostname: %m"); | |
0c24bb23 | 1640 | |
151b9b96 | 1641 | r = sd_event_add_io(s->event, &s->hostname_event_source, s->hostname_fd, 0, dispatch_hostname_change, s); |
0c24bb23 | 1642 | if (r < 0) { |
28def94c DR |
1643 | /* kernels prior to 3.2 don't support polling this file. Ignore |
1644 | * the failure. */ | |
1645 | if (r == -EPERM) { | |
e53fc357 | 1646 | log_warning_errno(r, "Failed to register hostname fd in event loop, ignoring: %m"); |
03e334a1 | 1647 | s->hostname_fd = safe_close(s->hostname_fd); |
28def94c DR |
1648 | return 0; |
1649 | } | |
1650 | ||
23bbb0de | 1651 | return log_error_errno(r, "Failed to register hostname fd in event loop: %m"); |
0c24bb23 LP |
1652 | } |
1653 | ||
1654 | r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10); | |
23bbb0de MS |
1655 | if (r < 0) |
1656 | return log_error_errno(r, "Failed to adjust priority of host name event source: %m"); | |
0c24bb23 LP |
1657 | |
1658 | return 0; | |
1659 | } | |
1660 | ||
e22aa3d3 LP |
1661 | static int dispatch_notify_event(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1662 | Server *s = userdata; | |
1663 | int r; | |
1664 | ||
1665 | assert(s); | |
1666 | assert(s->notify_event_source == es); | |
1667 | assert(s->notify_fd == fd); | |
1668 | ||
e22aa3d3 | 1669 | /* The $NOTIFY_SOCKET is writable again, now send exactly one |
dd835265 | 1670 | * message on it. Either it's the watchdog event, the initial |
119e9655 LP |
1671 | * READY=1 event or an stdout stream event. If there's nothing |
1672 | * to write anymore, turn our event source off. The next time | |
1673 | * there's something to send it will be turned on again. */ | |
e22aa3d3 LP |
1674 | |
1675 | if (!s->sent_notify_ready) { | |
1676 | static const char p[] = | |
1677 | "READY=1\n" | |
1678 | "STATUS=Processing requests..."; | |
1679 | ssize_t l; | |
1680 | ||
1681 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1682 | if (l < 0) { | |
1683 | if (errno == EAGAIN) | |
1684 | return 0; | |
1685 | ||
1686 | return log_error_errno(errno, "Failed to send READY=1 notification message: %m"); | |
1687 | } | |
1688 | ||
1689 | s->sent_notify_ready = true; | |
1690 | log_debug("Sent READY=1 notification."); | |
1691 | ||
119e9655 LP |
1692 | } else if (s->send_watchdog) { |
1693 | ||
1694 | static const char p[] = | |
1695 | "WATCHDOG=1"; | |
1696 | ||
1697 | ssize_t l; | |
1698 | ||
1699 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1700 | if (l < 0) { | |
1701 | if (errno == EAGAIN) | |
1702 | return 0; | |
1703 | ||
1704 | return log_error_errno(errno, "Failed to send WATCHDOG=1 notification message: %m"); | |
1705 | } | |
1706 | ||
1707 | s->send_watchdog = false; | |
1708 | log_debug("Sent WATCHDOG=1 notification."); | |
1709 | ||
e22aa3d3 LP |
1710 | } else if (s->stdout_streams_notify_queue) |
1711 | /* Dispatch one stream notification event */ | |
1712 | stdout_stream_send_notify(s->stdout_streams_notify_queue); | |
1713 | ||
61233823 | 1714 | /* Leave us enabled if there's still more to do. */ |
119e9655 | 1715 | if (s->send_watchdog || s->stdout_streams_notify_queue) |
e22aa3d3 LP |
1716 | return 0; |
1717 | ||
1718 | /* There was nothing to do anymore, let's turn ourselves off. */ | |
1719 | r = sd_event_source_set_enabled(es, SD_EVENT_OFF); | |
1720 | if (r < 0) | |
1721 | return log_error_errno(r, "Failed to turn off notify event source: %m"); | |
1722 | ||
1723 | return 0; | |
1724 | } | |
1725 | ||
119e9655 LP |
1726 | static int dispatch_watchdog(sd_event_source *es, uint64_t usec, void *userdata) { |
1727 | Server *s = userdata; | |
1728 | int r; | |
1729 | ||
1730 | assert(s); | |
1731 | ||
1732 | s->send_watchdog = true; | |
1733 | ||
1734 | r = sd_event_source_set_enabled(s->notify_event_source, SD_EVENT_ON); | |
1735 | if (r < 0) | |
1736 | log_warning_errno(r, "Failed to turn on notify event source: %m"); | |
1737 | ||
1738 | r = sd_event_source_set_time(s->watchdog_event_source, usec + s->watchdog_usec / 2); | |
1739 | if (r < 0) | |
1740 | return log_error_errno(r, "Failed to restart watchdog event source: %m"); | |
1741 | ||
1742 | r = sd_event_source_set_enabled(s->watchdog_event_source, SD_EVENT_ON); | |
1743 | if (r < 0) | |
1744 | return log_error_errno(r, "Failed to enable watchdog event source: %m"); | |
1745 | ||
1746 | return 0; | |
1747 | } | |
1748 | ||
e22aa3d3 | 1749 | static int server_connect_notify(Server *s) { |
15a3e96f | 1750 | union sockaddr_union sa = {}; |
e22aa3d3 | 1751 | const char *e; |
15a3e96f | 1752 | int r, salen; |
e22aa3d3 LP |
1753 | |
1754 | assert(s); | |
1755 | assert(s->notify_fd < 0); | |
1756 | assert(!s->notify_event_source); | |
1757 | ||
1758 | /* | |
337fabf7 LP |
1759 | * So here's the problem: we'd like to send notification messages to PID 1, but we cannot do that via |
1760 | * sd_notify(), since that's synchronous, and we might end up blocking on it. Specifically: given | |
1761 | * that PID 1 might block on dbus-daemon during IPC, and dbus-daemon is logging to us, and might | |
1762 | * hence block on us, we might end up in a deadlock if we block on sending PID 1 notification | |
1763 | * messages — by generating a full blocking circle. To avoid this, let's create a non-blocking | |
1764 | * socket, and connect it to the notification socket, and then wait for POLLOUT before we send | |
1765 | * anything. This should efficiently avoid any deadlocks, as we'll never block on PID 1, hence PID 1 | |
1766 | * can safely block on dbus-daemon which can safely block on us again. | |
1767 | * | |
1768 | * Don't think that this issue is real? It is, see: https://github.com/systemd/systemd/issues/1505 | |
1769 | */ | |
e22aa3d3 LP |
1770 | |
1771 | e = getenv("NOTIFY_SOCKET"); | |
1772 | if (!e) | |
1773 | return 0; | |
1774 | ||
15a3e96f LP |
1775 | salen = sockaddr_un_set_path(&sa.un, e); |
1776 | if (salen < 0) | |
1777 | return log_error_errno(salen, "NOTIFY_SOCKET set to invalid value '%s': %m", e); | |
e22aa3d3 LP |
1778 | |
1779 | s->notify_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); | |
1780 | if (s->notify_fd < 0) | |
1781 | return log_error_errno(errno, "Failed to create notify socket: %m"); | |
1782 | ||
1783 | (void) fd_inc_sndbuf(s->notify_fd, NOTIFY_SNDBUF_SIZE); | |
1784 | ||
15a3e96f | 1785 | r = connect(s->notify_fd, &sa.sa, salen); |
e22aa3d3 LP |
1786 | if (r < 0) |
1787 | return log_error_errno(errno, "Failed to connect to notify socket: %m"); | |
1788 | ||
1789 | r = sd_event_add_io(s->event, &s->notify_event_source, s->notify_fd, EPOLLOUT, dispatch_notify_event, s); | |
1790 | if (r < 0) | |
1791 | return log_error_errno(r, "Failed to watch notification socket: %m"); | |
1792 | ||
119e9655 LP |
1793 | if (sd_watchdog_enabled(false, &s->watchdog_usec) > 0) { |
1794 | s->send_watchdog = true; | |
1795 | ||
4de2402b | 1796 | r = sd_event_add_time(s->event, &s->watchdog_event_source, CLOCK_MONOTONIC, now(CLOCK_MONOTONIC) + s->watchdog_usec/2, s->watchdog_usec/4, dispatch_watchdog, s); |
119e9655 LP |
1797 | if (r < 0) |
1798 | return log_error_errno(r, "Failed to add watchdog time event: %m"); | |
1799 | } | |
1800 | ||
337fabf7 | 1801 | /* This should fire pretty soon, which we'll use to send the READY=1 event. */ |
e22aa3d3 LP |
1802 | |
1803 | return 0; | |
1804 | } | |
1805 | ||
d025f1e4 | 1806 | int server_init(Server *s) { |
13790add | 1807 | _cleanup_fdset_free_ FDSet *fds = NULL; |
d025f1e4 | 1808 | int n, r, fd; |
7d18d348 | 1809 | bool no_sockets; |
d025f1e4 ZJS |
1810 | |
1811 | assert(s); | |
1812 | ||
e4d9c985 LP |
1813 | *s = (Server) { |
1814 | .syslog_fd = -1, | |
1815 | .native_fd = -1, | |
1816 | .stdout_fd = -1, | |
1817 | .dev_kmsg_fd = -1, | |
1818 | .audit_fd = -1, | |
1819 | .hostname_fd = -1, | |
1820 | .notify_fd = -1, | |
d025f1e4 | 1821 | |
e4d9c985 LP |
1822 | .compress.enabled = true, |
1823 | .compress.threshold_bytes = (uint64_t) -1, | |
1824 | .seal = true, | |
1825 | .read_kmsg = true, | |
119e9655 | 1826 | |
e4d9c985 LP |
1827 | .watchdog_usec = USEC_INFINITY, |
1828 | ||
1829 | .sync_interval_usec = DEFAULT_SYNC_INTERVAL_USEC, | |
1830 | .sync_scheduled = false, | |
26687bf8 | 1831 | |
e4d9c985 LP |
1832 | .rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL, |
1833 | .rate_limit_burst = DEFAULT_RATE_LIMIT_BURST, | |
d025f1e4 | 1834 | |
e4d9c985 | 1835 | .forward_to_wall = true, |
d025f1e4 | 1836 | |
e4d9c985 | 1837 | .max_file_usec = DEFAULT_MAX_FILE_USEC, |
e150e820 | 1838 | |
e4d9c985 LP |
1839 | .max_level_store = LOG_DEBUG, |
1840 | .max_level_syslog = LOG_DEBUG, | |
1841 | .max_level_kmsg = LOG_NOTICE, | |
1842 | .max_level_console = LOG_INFO, | |
1843 | .max_level_wall = LOG_EMERG, | |
d025f1e4 | 1844 | |
e4d9c985 LP |
1845 | .line_max = DEFAULT_LINE_MAX, |
1846 | ||
1847 | .runtime_storage.name = "Runtime Journal", | |
1848 | .system_storage.name = "System Journal", | |
1849 | }; | |
ec20fe5f | 1850 | |
266a4700 FB |
1851 | journal_reset_metrics(&s->system_storage.metrics); |
1852 | journal_reset_metrics(&s->runtime_storage.metrics); | |
d025f1e4 ZJS |
1853 | |
1854 | server_parse_config_file(s); | |
1d84ad94 LP |
1855 | |
1856 | r = proc_cmdline_parse(parse_proc_cmdline_item, s, PROC_CMDLINE_STRIP_RD_PREFIX); | |
1857 | if (r < 0) | |
1858 | log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); | |
8580d1f7 | 1859 | |
d288f79f | 1860 | if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { |
b1389b0d ZJS |
1861 | log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0", |
1862 | s->rate_limit_interval, s->rate_limit_burst); | |
d288f79f ZJS |
1863 | s->rate_limit_interval = s->rate_limit_burst = 0; |
1864 | } | |
d025f1e4 | 1865 | |
8580d1f7 | 1866 | (void) mkdir_p("/run/systemd/journal", 0755); |
d025f1e4 | 1867 | |
43cf8388 | 1868 | s->user_journals = ordered_hashmap_new(NULL); |
d025f1e4 ZJS |
1869 | if (!s->user_journals) |
1870 | return log_oom(); | |
1871 | ||
1872 | s->mmap = mmap_cache_new(); | |
1873 | if (!s->mmap) | |
1874 | return log_oom(); | |
1875 | ||
b58c888f VC |
1876 | s->deferred_closes = set_new(NULL); |
1877 | if (!s->deferred_closes) | |
1878 | return log_oom(); | |
1879 | ||
f9a810be | 1880 | r = sd_event_default(&s->event); |
23bbb0de MS |
1881 | if (r < 0) |
1882 | return log_error_errno(r, "Failed to create event loop: %m"); | |
d025f1e4 ZJS |
1883 | |
1884 | n = sd_listen_fds(true); | |
23bbb0de MS |
1885 | if (n < 0) |
1886 | return log_error_errno(n, "Failed to read listening file descriptors from environment: %m"); | |
d025f1e4 ZJS |
1887 | |
1888 | for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { | |
1889 | ||
1890 | if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) { | |
1891 | ||
baaa35ad ZJS |
1892 | if (s->native_fd >= 0) |
1893 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), | |
1894 | "Too many native sockets passed."); | |
d025f1e4 ZJS |
1895 | |
1896 | s->native_fd = fd; | |
1897 | ||
1898 | } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) { | |
1899 | ||
baaa35ad ZJS |
1900 | if (s->stdout_fd >= 0) |
1901 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), | |
1902 | "Too many stdout sockets passed."); | |
d025f1e4 ZJS |
1903 | |
1904 | s->stdout_fd = fd; | |
1905 | ||
03ee5c38 LP |
1906 | } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0 || |
1907 | sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/dev-log", 0) > 0) { | |
d025f1e4 | 1908 | |
baaa35ad ZJS |
1909 | if (s->syslog_fd >= 0) |
1910 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), | |
1911 | "Too many /dev/log sockets passed."); | |
d025f1e4 ZJS |
1912 | |
1913 | s->syslog_fd = fd; | |
1914 | ||
875c2e22 LP |
1915 | } else if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1) > 0) { |
1916 | ||
baaa35ad ZJS |
1917 | if (s->audit_fd >= 0) |
1918 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), | |
1919 | "Too many audit sockets passed."); | |
875c2e22 LP |
1920 | |
1921 | s->audit_fd = fd; | |
1922 | ||
4ec3cd73 | 1923 | } else { |
4ec3cd73 | 1924 | |
13790add LP |
1925 | if (!fds) { |
1926 | fds = fdset_new(); | |
1927 | if (!fds) | |
1928 | return log_oom(); | |
1929 | } | |
4ec3cd73 | 1930 | |
13790add LP |
1931 | r = fdset_put(fds, fd); |
1932 | if (r < 0) | |
1933 | return log_oom(); | |
4ec3cd73 | 1934 | } |
d025f1e4 ZJS |
1935 | } |
1936 | ||
15d91bff ZJS |
1937 | /* Try to restore streams, but don't bother if this fails */ |
1938 | (void) server_restore_streams(s, fds); | |
d025f1e4 | 1939 | |
13790add LP |
1940 | if (fdset_size(fds) > 0) { |
1941 | log_warning("%u unknown file descriptors passed, closing.", fdset_size(fds)); | |
1942 | fds = fdset_free(fds); | |
1943 | } | |
1944 | ||
7d18d348 ZJS |
1945 | no_sockets = s->native_fd < 0 && s->stdout_fd < 0 && s->syslog_fd < 0 && s->audit_fd < 0; |
1946 | ||
1947 | /* always open stdout, syslog, native, and kmsg sockets */ | |
37b7affe ZJS |
1948 | |
1949 | /* systemd-journald.socket: /run/systemd/journal/stdout */ | |
15d91bff ZJS |
1950 | r = server_open_stdout_socket(s); |
1951 | if (r < 0) | |
1952 | return r; | |
1953 | ||
37b7affe | 1954 | /* systemd-journald-dev-log.socket: /run/systemd/journal/dev-log */ |
13790add | 1955 | r = server_open_syslog_socket(s); |
d025f1e4 ZJS |
1956 | if (r < 0) |
1957 | return r; | |
1958 | ||
37b7affe | 1959 | /* systemd-journald.socket: /run/systemd/journal/socket */ |
13790add | 1960 | r = server_open_native_socket(s); |
d025f1e4 ZJS |
1961 | if (r < 0) |
1962 | return r; | |
1963 | ||
b2392ff3 | 1964 | /* /dev/kmsg */ |
d025f1e4 ZJS |
1965 | r = server_open_dev_kmsg(s); |
1966 | if (r < 0) | |
1967 | return r; | |
1968 | ||
7d18d348 ZJS |
1969 | /* Unless we got *some* sockets and not audit, open audit socket */ |
1970 | if (s->audit_fd >= 0 || no_sockets) { | |
1971 | r = server_open_audit(s); | |
1972 | if (r < 0) | |
1973 | return r; | |
1974 | } | |
875c2e22 | 1975 | |
d025f1e4 ZJS |
1976 | r = server_open_kernel_seqnum(s); |
1977 | if (r < 0) | |
1978 | return r; | |
1979 | ||
0c24bb23 LP |
1980 | r = server_open_hostname(s); |
1981 | if (r < 0) | |
1982 | return r; | |
1983 | ||
f9a810be | 1984 | r = setup_signals(s); |
d025f1e4 ZJS |
1985 | if (r < 0) |
1986 | return r; | |
1987 | ||
90fc172e | 1988 | s->rate_limit = journal_rate_limit_new(); |
d025f1e4 ZJS |
1989 | if (!s->rate_limit) |
1990 | return -ENOMEM; | |
1991 | ||
e9174f29 LP |
1992 | r = cg_get_root_path(&s->cgroup_root); |
1993 | if (r < 0) | |
1994 | return r; | |
1995 | ||
0c24bb23 LP |
1996 | server_cache_hostname(s); |
1997 | server_cache_boot_id(s); | |
1998 | server_cache_machine_id(s); | |
1999 | ||
605405c6 ZJS |
2000 | s->runtime_storage.path = strjoin("/run/log/journal/", SERVER_MACHINE_ID(s)); |
2001 | s->system_storage.path = strjoin("/var/log/journal/", SERVER_MACHINE_ID(s)); | |
266a4700 FB |
2002 | if (!s->runtime_storage.path || !s->system_storage.path) |
2003 | return -ENOMEM; | |
2004 | ||
e22aa3d3 LP |
2005 | (void) server_connect_notify(s); |
2006 | ||
22e3a02b LP |
2007 | (void) client_context_acquire_default(s); |
2008 | ||
804ae586 | 2009 | return system_journal_open(s, false); |
d025f1e4 ZJS |
2010 | } |
2011 | ||
2012 | void server_maybe_append_tags(Server *s) { | |
349cc4a5 | 2013 | #if HAVE_GCRYPT |
d025f1e4 ZJS |
2014 | JournalFile *f; |
2015 | Iterator i; | |
2016 | usec_t n; | |
2017 | ||
2018 | n = now(CLOCK_REALTIME); | |
2019 | ||
2020 | if (s->system_journal) | |
2021 | journal_file_maybe_append_tag(s->system_journal, n); | |
2022 | ||
43cf8388 | 2023 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) |
d025f1e4 ZJS |
2024 | journal_file_maybe_append_tag(f, n); |
2025 | #endif | |
2026 | } | |
2027 | ||
2028 | void server_done(Server *s) { | |
d025f1e4 ZJS |
2029 | assert(s); |
2030 | ||
f9168190 | 2031 | set_free_with_destructor(s->deferred_closes, journal_file_close); |
b58c888f | 2032 | |
d025f1e4 ZJS |
2033 | while (s->stdout_streams) |
2034 | stdout_stream_free(s->stdout_streams); | |
2035 | ||
22e3a02b LP |
2036 | client_context_flush_all(s); |
2037 | ||
d025f1e4 | 2038 | if (s->system_journal) |
69a3a6fd | 2039 | (void) journal_file_close(s->system_journal); |
d025f1e4 ZJS |
2040 | |
2041 | if (s->runtime_journal) | |
69a3a6fd | 2042 | (void) journal_file_close(s->runtime_journal); |
d025f1e4 | 2043 | |
f9168190 | 2044 | ordered_hashmap_free_with_destructor(s->user_journals, journal_file_close); |
d025f1e4 | 2045 | |
f9a810be LP |
2046 | sd_event_source_unref(s->syslog_event_source); |
2047 | sd_event_source_unref(s->native_event_source); | |
2048 | sd_event_source_unref(s->stdout_event_source); | |
2049 | sd_event_source_unref(s->dev_kmsg_event_source); | |
875c2e22 | 2050 | sd_event_source_unref(s->audit_event_source); |
f9a810be LP |
2051 | sd_event_source_unref(s->sync_event_source); |
2052 | sd_event_source_unref(s->sigusr1_event_source); | |
2053 | sd_event_source_unref(s->sigusr2_event_source); | |
2054 | sd_event_source_unref(s->sigterm_event_source); | |
2055 | sd_event_source_unref(s->sigint_event_source); | |
94b65516 | 2056 | sd_event_source_unref(s->sigrtmin1_event_source); |
0c24bb23 | 2057 | sd_event_source_unref(s->hostname_event_source); |
e22aa3d3 | 2058 | sd_event_source_unref(s->notify_event_source); |
119e9655 | 2059 | sd_event_source_unref(s->watchdog_event_source); |
f9a810be | 2060 | sd_event_unref(s->event); |
d025f1e4 | 2061 | |
03e334a1 LP |
2062 | safe_close(s->syslog_fd); |
2063 | safe_close(s->native_fd); | |
2064 | safe_close(s->stdout_fd); | |
2065 | safe_close(s->dev_kmsg_fd); | |
875c2e22 | 2066 | safe_close(s->audit_fd); |
03e334a1 | 2067 | safe_close(s->hostname_fd); |
e22aa3d3 | 2068 | safe_close(s->notify_fd); |
0c24bb23 | 2069 | |
d025f1e4 ZJS |
2070 | if (s->rate_limit) |
2071 | journal_rate_limit_free(s->rate_limit); | |
2072 | ||
2073 | if (s->kernel_seqnum) | |
2074 | munmap(s->kernel_seqnum, sizeof(uint64_t)); | |
2075 | ||
2076 | free(s->buffer); | |
2077 | free(s->tty_path); | |
e9174f29 | 2078 | free(s->cgroup_root); |
99d0966e | 2079 | free(s->hostname_field); |
c6e9e16f ZJS |
2080 | free(s->runtime_storage.path); |
2081 | free(s->system_storage.path); | |
d025f1e4 | 2082 | |
e3d78cb1 | 2083 | mmap_cache_unref(s->mmap); |
d025f1e4 | 2084 | } |
8580d1f7 LP |
2085 | |
2086 | static const char* const storage_table[_STORAGE_MAX] = { | |
2087 | [STORAGE_AUTO] = "auto", | |
2088 | [STORAGE_VOLATILE] = "volatile", | |
2089 | [STORAGE_PERSISTENT] = "persistent", | |
2090 | [STORAGE_NONE] = "none" | |
2091 | }; | |
2092 | ||
2093 | DEFINE_STRING_TABLE_LOOKUP(storage, Storage); | |
2094 | DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); | |
2095 | ||
2096 | static const char* const split_mode_table[_SPLIT_MAX] = { | |
2097 | [SPLIT_LOGIN] = "login", | |
2098 | [SPLIT_UID] = "uid", | |
2099 | [SPLIT_NONE] = "none", | |
2100 | }; | |
2101 | ||
2102 | DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode); | |
2103 | DEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting"); | |
ec20fe5f LP |
2104 | |
2105 | int config_parse_line_max( | |
2106 | const char* unit, | |
2107 | const char *filename, | |
2108 | unsigned line, | |
2109 | const char *section, | |
2110 | unsigned section_line, | |
2111 | const char *lvalue, | |
2112 | int ltype, | |
2113 | const char *rvalue, | |
2114 | void *data, | |
2115 | void *userdata) { | |
2116 | ||
2117 | size_t *sz = data; | |
2118 | int r; | |
2119 | ||
2120 | assert(filename); | |
2121 | assert(lvalue); | |
2122 | assert(rvalue); | |
2123 | assert(data); | |
2124 | ||
2125 | if (isempty(rvalue)) | |
2126 | /* Empty assignment means default */ | |
2127 | *sz = DEFAULT_LINE_MAX; | |
2128 | else { | |
2129 | uint64_t v; | |
2130 | ||
2131 | r = parse_size(rvalue, 1024, &v); | |
2132 | if (r < 0) { | |
2133 | log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse LineMax= value, ignoring: %s", rvalue); | |
2134 | return 0; | |
2135 | } | |
2136 | ||
2137 | if (v < 79) { | |
2138 | /* Why specify 79 here as minimum line length? Simply, because the most common traditional | |
2139 | * terminal size is 80ch, and it might make sense to break one character before the natural | |
2140 | * line break would occur on that. */ | |
2141 | log_syntax(unit, LOG_WARNING, filename, line, 0, "LineMax= too small, clamping to 79: %s", rvalue); | |
2142 | *sz = 79; | |
2143 | } else if (v > (uint64_t) (SSIZE_MAX-1)) { | |
2144 | /* So, why specify SSIZE_MAX-1 here? Because that's one below the largest size value read() | |
2145 | * can return, and we need one extra byte for the trailing NUL byte. Of course IRL such large | |
2146 | * memory allocations will fail anyway, hence this limit is mostly theoretical anyway, as we'll | |
2147 | * fail much earlier anyway. */ | |
2148 | log_syntax(unit, LOG_WARNING, filename, line, 0, "LineMax= too large, clamping to %" PRIu64 ": %s", (uint64_t) (SSIZE_MAX-1), rvalue); | |
2149 | *sz = SSIZE_MAX-1; | |
2150 | } else | |
2151 | *sz = (size_t) v; | |
2152 | } | |
2153 | ||
2154 | return 0; | |
2155 | } | |
1b7cf0e5 | 2156 | |
e3d36a8d LP |
2157 | int config_parse_compress( |
2158 | const char* unit, | |
2159 | const char *filename, | |
2160 | unsigned line, | |
2161 | const char *section, | |
2162 | unsigned section_line, | |
2163 | const char *lvalue, | |
2164 | int ltype, | |
2165 | const char *rvalue, | |
2166 | void *data, | |
2167 | void *userdata) { | |
2168 | ||
1b7cf0e5 AG |
2169 | JournalCompressOptions* compress = data; |
2170 | int r; | |
2171 | ||
e3d36a8d LP |
2172 | if (isempty(rvalue)) { |
2173 | compress->enabled = true; | |
2174 | compress->threshold_bytes = (uint64_t) -1; | |
2175 | } else if (streq(rvalue, "1")) { | |
1b7cf0e5 AG |
2176 | log_syntax(unit, LOG_WARNING, filename, line, 0, |
2177 | "Compress= ambiguously specified as 1, enabling compression with default threshold"); | |
2178 | compress->enabled = true; | |
2179 | } else if (streq(rvalue, "0")) { | |
2180 | log_syntax(unit, LOG_WARNING, filename, line, 0, | |
2181 | "Compress= ambiguously specified as 0, disabling compression"); | |
2182 | compress->enabled = false; | |
e3d36a8d LP |
2183 | } else { |
2184 | r = parse_boolean(rvalue); | |
2185 | if (r < 0) { | |
2186 | r = parse_size(rvalue, 1024, &compress->threshold_bytes); | |
2187 | if (r < 0) | |
2188 | log_syntax(unit, LOG_ERR, filename, line, r, | |
2189 | "Failed to parse Compress= value, ignoring: %s", rvalue); | |
2190 | else | |
2191 | compress->enabled = true; | |
2192 | } else | |
2193 | compress->enabled = r; | |
2194 | } | |
1b7cf0e5 AG |
2195 | |
2196 | return 0; | |
2197 | } |