]>
Commit | Line | Data |
---|---|---|
d025f1e4 ZJS |
1 | /*** |
2 | This file is part of systemd. | |
3 | ||
4 | Copyright 2011 Lennart Poettering | |
5 | ||
6 | systemd is free software; you can redistribute it and/or modify it | |
7 | under the terms of the GNU Lesser General Public License as published by | |
8 | the Free Software Foundation; either version 2.1 of the License, or | |
9 | (at your option) any later version. | |
10 | ||
11 | systemd is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | Lesser General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU Lesser General Public License | |
17 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
18 | ***/ | |
19 | ||
24882e06 LP |
20 | #ifdef HAVE_SELINUX |
21 | #include <selinux/selinux.h> | |
22 | #endif | |
8580d1f7 LP |
23 | #include <sys/ioctl.h> |
24 | #include <sys/mman.h> | |
25 | #include <sys/signalfd.h> | |
26 | #include <sys/statvfs.h> | |
07630cea | 27 | #include <linux/sockios.h> |
24882e06 | 28 | |
b4bbcaa9 | 29 | #include "libudev.h" |
8580d1f7 | 30 | #include "sd-daemon.h" |
74df0fca LP |
31 | #include "sd-journal.h" |
32 | #include "sd-messages.h" | |
8580d1f7 LP |
33 | |
34 | #include "acl-util.h" | |
b5efdb8a | 35 | #include "alloc-util.h" |
430f0182 | 36 | #include "audit-util.h" |
d025f1e4 | 37 | #include "cgroup-util.h" |
d025f1e4 | 38 | #include "conf-parser.h" |
a0956174 | 39 | #include "dirent-util.h" |
0dec689b | 40 | #include "extract-word.h" |
3ffd4af2 | 41 | #include "fd-util.h" |
33d52ab9 | 42 | #include "fileio.h" |
f97b34a6 | 43 | #include "format-util.h" |
f4f15635 | 44 | #include "fs-util.h" |
8580d1f7 | 45 | #include "hashmap.h" |
958b66ea | 46 | #include "hostname-util.h" |
4b58153d | 47 | #include "id128-util.h" |
afc5dbf3 | 48 | #include "io-util.h" |
8580d1f7 LP |
49 | #include "journal-authenticate.h" |
50 | #include "journal-file.h" | |
d025f1e4 ZJS |
51 | #include "journal-internal.h" |
52 | #include "journal-vacuum.h" | |
8580d1f7 | 53 | #include "journald-audit.h" |
d025f1e4 | 54 | #include "journald-kmsg.h" |
d025f1e4 | 55 | #include "journald-native.h" |
8580d1f7 | 56 | #include "journald-rate-limit.h" |
3ffd4af2 | 57 | #include "journald-server.h" |
8580d1f7 LP |
58 | #include "journald-stream.h" |
59 | #include "journald-syslog.h" | |
4b58153d | 60 | #include "log.h" |
07630cea LP |
61 | #include "missing.h" |
62 | #include "mkdir.h" | |
6bedfcbb | 63 | #include "parse-util.h" |
4e731273 | 64 | #include "proc-cmdline.h" |
07630cea LP |
65 | #include "process-util.h" |
66 | #include "rm-rf.h" | |
67 | #include "selinux-util.h" | |
68 | #include "signal-util.h" | |
69 | #include "socket-util.h" | |
32917e33 | 70 | #include "stdio-util.h" |
8b43440b | 71 | #include "string-table.h" |
07630cea | 72 | #include "string-util.h" |
4a0b58c4 | 73 | #include "user-util.h" |
863a5610 | 74 | #include "syslog-util.h" |
d025f1e4 | 75 | |
d025f1e4 ZJS |
76 | #define USER_JOURNALS_MAX 1024 |
77 | ||
26687bf8 | 78 | #define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) |
7f1ad696 LP |
79 | #define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC) |
80 | #define DEFAULT_RATE_LIMIT_BURST 1000 | |
e150e820 | 81 | #define DEFAULT_MAX_FILE_USEC USEC_PER_MONTH |
d025f1e4 | 82 | |
8580d1f7 | 83 | #define RECHECK_SPACE_USEC (30*USEC_PER_SEC) |
d025f1e4 | 84 | |
e22aa3d3 LP |
85 | #define NOTIFY_SNDBUF_SIZE (8*1024*1024) |
86 | ||
7a24f3bf VC |
87 | /* The period to insert between posting changes for coalescing */ |
88 | #define POST_CHANGE_TIMER_INTERVAL_USEC (250*USEC_PER_MSEC) | |
89 | ||
e0ed6db9 FB |
90 | static int determine_path_usage(Server *s, const char *path, uint64_t *ret_used, uint64_t *ret_free) { |
91 | _cleanup_closedir_ DIR *d = NULL; | |
92 | struct dirent *de; | |
93 | struct statvfs ss; | |
e0ed6db9 FB |
94 | |
95 | assert(ret_used); | |
96 | assert(ret_free); | |
97 | ||
266a4700 | 98 | d = opendir(path); |
e0ed6db9 FB |
99 | if (!d) |
100 | return log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_ERR, | |
266a4700 | 101 | errno, "Failed to open %s: %m", path); |
e0ed6db9 FB |
102 | |
103 | if (fstatvfs(dirfd(d), &ss) < 0) | |
266a4700 | 104 | return log_error_errno(errno, "Failed to fstatvfs(%s): %m", path); |
e0ed6db9 FB |
105 | |
106 | *ret_free = ss.f_bsize * ss.f_bavail; | |
107 | *ret_used = 0; | |
108 | FOREACH_DIRENT_ALL(de, d, break) { | |
109 | struct stat st; | |
110 | ||
111 | if (!endswith(de->d_name, ".journal") && | |
112 | !endswith(de->d_name, ".journal~")) | |
113 | continue; | |
114 | ||
115 | if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) { | |
266a4700 | 116 | log_debug_errno(errno, "Failed to stat %s/%s, ignoring: %m", path, de->d_name); |
e0ed6db9 FB |
117 | continue; |
118 | } | |
119 | ||
120 | if (!S_ISREG(st.st_mode)) | |
121 | continue; | |
122 | ||
123 | *ret_used += (uint64_t) st.st_blocks * 512UL; | |
124 | } | |
125 | ||
126 | return 0; | |
127 | } | |
128 | ||
a0edc477 FB |
129 | static void cache_space_invalidate(JournalStorageSpace *space) { |
130 | memset(space, 0, sizeof(*space)); | |
131 | } | |
132 | ||
57f443a6 | 133 | static int cache_space_refresh(Server *s, JournalStorage *storage) { |
23aba343 | 134 | JournalStorageSpace *space; |
266a4700 | 135 | JournalMetrics *metrics; |
23aba343 | 136 | uint64_t vfs_used, vfs_avail, avail; |
d025f1e4 | 137 | usec_t ts; |
e0ed6db9 | 138 | int r; |
d025f1e4 | 139 | |
8580d1f7 | 140 | assert(s); |
266a4700 | 141 | |
266a4700 | 142 | metrics = &storage->metrics; |
23aba343 | 143 | space = &storage->space; |
d025f1e4 | 144 | |
8580d1f7 | 145 | ts = now(CLOCK_MONOTONIC); |
d025f1e4 | 146 | |
3099caf2 | 147 | if (space->timestamp != 0 && space->timestamp + RECHECK_SPACE_USEC > ts) |
d025f1e4 ZJS |
148 | return 0; |
149 | ||
23aba343 | 150 | r = determine_path_usage(s, storage->path, &vfs_used, &vfs_avail); |
e0ed6db9 FB |
151 | if (r < 0) |
152 | return r; | |
d025f1e4 | 153 | |
23aba343 FB |
154 | space->vfs_used = vfs_used; |
155 | space->vfs_available = vfs_avail; | |
156 | ||
157 | avail = LESS_BY(vfs_avail, metrics->keep_free); | |
158 | ||
23aba343 FB |
159 | space->limit = MIN(MAX(vfs_used + avail, metrics->min_use), metrics->max_use); |
160 | space->available = LESS_BY(space->limit, vfs_used); | |
161 | space->timestamp = ts; | |
8580d1f7 LP |
162 | return 1; |
163 | } | |
164 | ||
3a19f215 FB |
165 | static void patch_min_use(JournalStorage *storage) { |
166 | assert(storage); | |
167 | ||
168 | /* Let's bump the min_use limit to the current usage on disk. We do | |
169 | * this when starting up and first opening the journal files. This way | |
170 | * sudden spikes in disk usage will not cause journald to vacuum files | |
171 | * without bounds. Note that this means that only a restart of journald | |
172 | * will make it reset this value. */ | |
173 | ||
174 | storage->metrics.min_use = MAX(storage->metrics.min_use, storage->space.vfs_used); | |
175 | } | |
176 | ||
177 | ||
178 | static int determine_space(Server *s, uint64_t *available, uint64_t *limit) { | |
266a4700 | 179 | JournalStorage *js; |
57f443a6 | 180 | int r; |
8580d1f7 LP |
181 | |
182 | assert(s); | |
183 | ||
266a4700 | 184 | js = s->system_journal ? &s->system_storage : &s->runtime_storage; |
57f443a6 FB |
185 | |
186 | r = cache_space_refresh(s, js); | |
187 | if (r >= 0) { | |
188 | if (available) | |
189 | *available = js->space.available; | |
190 | if (limit) | |
191 | *limit = js->space.limit; | |
192 | } | |
193 | return r; | |
d025f1e4 ZJS |
194 | } |
195 | ||
cba5629e FB |
196 | void server_space_usage_message(Server *s, JournalStorage *storage) { |
197 | char fb1[FORMAT_BYTES_MAX], fb2[FORMAT_BYTES_MAX], fb3[FORMAT_BYTES_MAX], | |
198 | fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX], fb6[FORMAT_BYTES_MAX]; | |
199 | JournalMetrics *metrics; | |
cba5629e FB |
200 | |
201 | assert(s); | |
202 | ||
203 | if (!storage) | |
204 | storage = s->system_journal ? &s->system_storage : &s->runtime_storage; | |
205 | ||
57f443a6 | 206 | if (cache_space_refresh(s, storage) < 0) |
cba5629e FB |
207 | return; |
208 | ||
209 | metrics = &storage->metrics; | |
23aba343 | 210 | format_bytes(fb1, sizeof(fb1), storage->space.vfs_used); |
cba5629e FB |
211 | format_bytes(fb2, sizeof(fb2), metrics->max_use); |
212 | format_bytes(fb3, sizeof(fb3), metrics->keep_free); | |
23aba343 | 213 | format_bytes(fb4, sizeof(fb4), storage->space.vfs_available); |
cba5629e FB |
214 | format_bytes(fb5, sizeof(fb5), storage->space.limit); |
215 | format_bytes(fb6, sizeof(fb6), storage->space.available); | |
216 | ||
2b044526 | 217 | server_driver_message(s, "MESSAGE_ID=" SD_MESSAGE_JOURNAL_USAGE_STR, |
cba5629e FB |
218 | LOG_MESSAGE("%s (%s) is %s, max %s, %s free.", |
219 | storage->name, storage->path, fb1, fb5, fb6), | |
220 | "JOURNAL_NAME=%s", storage->name, | |
221 | "JOURNAL_PATH=%s", storage->path, | |
23aba343 | 222 | "CURRENT_USE=%"PRIu64, storage->space.vfs_used, |
cba5629e FB |
223 | "CURRENT_USE_PRETTY=%s", fb1, |
224 | "MAX_USE=%"PRIu64, metrics->max_use, | |
225 | "MAX_USE_PRETTY=%s", fb2, | |
226 | "DISK_KEEP_FREE=%"PRIu64, metrics->keep_free, | |
227 | "DISK_KEEP_FREE_PRETTY=%s", fb3, | |
23aba343 | 228 | "DISK_AVAILABLE=%"PRIu64, storage->space.vfs_available, |
cba5629e FB |
229 | "DISK_AVAILABLE_PRETTY=%s", fb4, |
230 | "LIMIT=%"PRIu64, storage->space.limit, | |
231 | "LIMIT_PRETTY=%s", fb5, | |
232 | "AVAILABLE=%"PRIu64, storage->space.available, | |
233 | "AVAILABLE_PRETTY=%s", fb6, | |
234 | NULL); | |
235 | } | |
236 | ||
5c3bde3f | 237 | static void server_add_acls(JournalFile *f, uid_t uid) { |
d025f1e4 | 238 | #ifdef HAVE_ACL |
5c3bde3f | 239 | int r; |
d025f1e4 | 240 | #endif |
d025f1e4 ZJS |
241 | assert(f); |
242 | ||
d025f1e4 | 243 | #ifdef HAVE_ACL |
34c10968 | 244 | if (uid <= SYSTEM_UID_MAX) |
d025f1e4 ZJS |
245 | return; |
246 | ||
5c3bde3f ZJS |
247 | r = add_acls_for_user(f->fd, uid); |
248 | if (r < 0) | |
249 | log_warning_errno(r, "Failed to set ACL on %s, ignoring: %m", f->path); | |
d025f1e4 ZJS |
250 | #endif |
251 | } | |
252 | ||
7a24f3bf VC |
253 | static int open_journal( |
254 | Server *s, | |
255 | bool reliably, | |
256 | const char *fname, | |
257 | int flags, | |
258 | bool seal, | |
259 | JournalMetrics *metrics, | |
7a24f3bf VC |
260 | JournalFile **ret) { |
261 | int r; | |
e167d7fd | 262 | JournalFile *f; |
7a24f3bf VC |
263 | |
264 | assert(s); | |
265 | assert(fname); | |
266 | assert(ret); | |
267 | ||
268 | if (reliably) | |
b58c888f | 269 | r = journal_file_open_reliably(fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f); |
7a24f3bf | 270 | else |
5d1ce257 | 271 | r = journal_file_open(-1, fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f); |
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); |
105bdb46 VC |
314 | } else if (r < 0) { |
315 | if (r != -ENOENT && r != -EROFS) | |
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 | ||
61755fda | 402 | if (uid <= SYSTEM_UID_MAX || uid_is_dynamic(uid)) |
d025f1e4 ZJS |
403 | return s->system_journal; |
404 | ||
405 | r = sd_id128_get_machine(&machine); | |
406 | if (r < 0) | |
407 | return s->system_journal; | |
408 | ||
4a0b58c4 | 409 | f = ordered_hashmap_get(s->user_journals, UID_TO_PTR(uid)); |
d025f1e4 ZJS |
410 | if (f) |
411 | return f; | |
412 | ||
de0671ee ZJS |
413 | if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-"UID_FMT".journal", |
414 | SD_ID128_FORMAT_VAL(machine), uid) < 0) | |
d025f1e4 ZJS |
415 | return s->system_journal; |
416 | ||
43cf8388 | 417 | while (ordered_hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { |
d025f1e4 | 418 | /* Too many open? Then let's close one */ |
43cf8388 | 419 | f = ordered_hashmap_steal_first(s->user_journals); |
d025f1e4 | 420 | assert(f); |
69a3a6fd | 421 | (void) journal_file_close(f); |
d025f1e4 ZJS |
422 | } |
423 | ||
266a4700 | 424 | r = open_journal(s, true, p, O_RDWR|O_CREAT, s->seal, &s->system_storage.metrics, &f); |
d025f1e4 ZJS |
425 | if (r < 0) |
426 | return s->system_journal; | |
427 | ||
5c3bde3f | 428 | server_add_acls(f, uid); |
d025f1e4 | 429 | |
4a0b58c4 | 430 | r = ordered_hashmap_put(s->user_journals, UID_TO_PTR(uid), f); |
d025f1e4 | 431 | if (r < 0) { |
69a3a6fd | 432 | (void) journal_file_close(f); |
d025f1e4 ZJS |
433 | return s->system_journal; |
434 | } | |
435 | ||
436 | return f; | |
437 | } | |
438 | ||
ea69bd41 LP |
439 | static int do_rotate( |
440 | Server *s, | |
441 | JournalFile **f, | |
442 | const char* name, | |
443 | bool seal, | |
444 | uint32_t uid) { | |
445 | ||
fc55baee ZJS |
446 | int r; |
447 | assert(s); | |
448 | ||
449 | if (!*f) | |
450 | return -EINVAL; | |
451 | ||
b58c888f | 452 | r = journal_file_rotate(f, s->compress, seal, s->deferred_closes); |
fc55baee ZJS |
453 | if (r < 0) |
454 | if (*f) | |
ea69bd41 | 455 | log_error_errno(r, "Failed to rotate %s: %m", (*f)->path); |
fc55baee | 456 | else |
ea69bd41 | 457 | log_error_errno(r, "Failed to create new %s journal: %m", name); |
fc55baee | 458 | else |
5c3bde3f | 459 | server_add_acls(*f, uid); |
2678031a | 460 | |
fc55baee ZJS |
461 | return r; |
462 | } | |
463 | ||
d025f1e4 ZJS |
464 | void server_rotate(Server *s) { |
465 | JournalFile *f; | |
466 | void *k; | |
467 | Iterator i; | |
468 | int r; | |
469 | ||
470 | log_debug("Rotating..."); | |
471 | ||
8580d1f7 LP |
472 | (void) do_rotate(s, &s->runtime_journal, "runtime", false, 0); |
473 | (void) do_rotate(s, &s->system_journal, "system", s->seal, 0); | |
d025f1e4 | 474 | |
43cf8388 | 475 | ORDERED_HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { |
4a0b58c4 | 476 | r = do_rotate(s, &f, "user", s->seal, PTR_TO_UID(k)); |
fc55baee | 477 | if (r >= 0) |
43cf8388 | 478 | ordered_hashmap_replace(s->user_journals, k, f); |
fc55baee ZJS |
479 | else if (!f) |
480 | /* Old file has been closed and deallocated */ | |
43cf8388 | 481 | ordered_hashmap_remove(s->user_journals, k); |
d025f1e4 | 482 | } |
b58c888f VC |
483 | |
484 | /* Perform any deferred closes which aren't still offlining. */ | |
485 | SET_FOREACH(f, s->deferred_closes, i) | |
486 | if (!journal_file_is_offlining(f)) { | |
487 | (void) set_remove(s->deferred_closes, f); | |
488 | (void) journal_file_close(f); | |
489 | } | |
d025f1e4 ZJS |
490 | } |
491 | ||
26687bf8 OS |
492 | void server_sync(Server *s) { |
493 | JournalFile *f; | |
26687bf8 OS |
494 | Iterator i; |
495 | int r; | |
496 | ||
26687bf8 | 497 | if (s->system_journal) { |
ac2e41f5 | 498 | r = journal_file_set_offline(s->system_journal, false); |
26687bf8 | 499 | if (r < 0) |
65089b82 | 500 | log_warning_errno(r, "Failed to sync system journal, ignoring: %m"); |
26687bf8 OS |
501 | } |
502 | ||
65c1d46b | 503 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) { |
ac2e41f5 | 504 | r = journal_file_set_offline(f, false); |
26687bf8 | 505 | if (r < 0) |
65089b82 | 506 | log_warning_errno(r, "Failed to sync user journal, ignoring: %m"); |
26687bf8 OS |
507 | } |
508 | ||
f9a810be LP |
509 | if (s->sync_event_source) { |
510 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF); | |
511 | if (r < 0) | |
da927ba9 | 512 | log_error_errno(r, "Failed to disable sync timer source: %m"); |
f9a810be | 513 | } |
26687bf8 OS |
514 | |
515 | s->sync_scheduled = false; | |
516 | } | |
517 | ||
3a19f215 | 518 | static void do_vacuum(Server *s, JournalStorage *storage, bool verbose) { |
ea69bd41 | 519 | |
63c8666b ZJS |
520 | int r; |
521 | ||
8580d1f7 | 522 | assert(s); |
266a4700 | 523 | assert(storage); |
8580d1f7 | 524 | |
57f443a6 | 525 | (void) cache_space_refresh(s, storage); |
18e758bf FB |
526 | |
527 | if (verbose) | |
528 | server_space_usage_message(s, storage); | |
8580d1f7 | 529 | |
57f443a6 FB |
530 | r = journal_directory_vacuum(storage->path, storage->space.limit, |
531 | storage->metrics.n_max_files, s->max_retention_usec, | |
532 | &s->oldest_file_usec, verbose); | |
63c8666b | 533 | if (r < 0 && r != -ENOENT) |
266a4700 FB |
534 | log_warning_errno(r, "Failed to vacuum %s, ignoring: %m", storage->path); |
535 | ||
a0edc477 | 536 | cache_space_invalidate(&storage->space); |
63c8666b ZJS |
537 | } |
538 | ||
3a19f215 | 539 | int server_vacuum(Server *s, bool verbose) { |
8580d1f7 | 540 | assert(s); |
d025f1e4 ZJS |
541 | |
542 | log_debug("Vacuuming..."); | |
543 | ||
544 | s->oldest_file_usec = 0; | |
545 | ||
266a4700 | 546 | if (s->system_journal) |
3a19f215 | 547 | do_vacuum(s, &s->system_storage, verbose); |
266a4700 | 548 | if (s->runtime_journal) |
3a19f215 | 549 | do_vacuum(s, &s->runtime_storage, verbose); |
d025f1e4 | 550 | |
8580d1f7 | 551 | return 0; |
d025f1e4 ZJS |
552 | } |
553 | ||
0c24bb23 LP |
554 | static void server_cache_machine_id(Server *s) { |
555 | sd_id128_t id; | |
556 | int r; | |
557 | ||
558 | assert(s); | |
559 | ||
560 | r = sd_id128_get_machine(&id); | |
561 | if (r < 0) | |
562 | return; | |
563 | ||
564 | sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID=")); | |
565 | } | |
566 | ||
567 | static void server_cache_boot_id(Server *s) { | |
568 | sd_id128_t id; | |
569 | int r; | |
570 | ||
571 | assert(s); | |
572 | ||
573 | r = sd_id128_get_boot(&id); | |
574 | if (r < 0) | |
575 | return; | |
576 | ||
577 | sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID=")); | |
578 | } | |
579 | ||
580 | static void server_cache_hostname(Server *s) { | |
581 | _cleanup_free_ char *t = NULL; | |
582 | char *x; | |
583 | ||
584 | assert(s); | |
585 | ||
586 | t = gethostname_malloc(); | |
587 | if (!t) | |
588 | return; | |
589 | ||
590 | x = strappend("_HOSTNAME=", t); | |
591 | if (!x) | |
592 | return; | |
593 | ||
594 | free(s->hostname_field); | |
595 | s->hostname_field = x; | |
596 | } | |
597 | ||
8531ae70 | 598 | static bool shall_try_append_again(JournalFile *f, int r) { |
6e1045e5 | 599 | switch(r) { |
ae739cc1 | 600 | |
6e1045e5 ZJS |
601 | case -E2BIG: /* Hit configured limit */ |
602 | case -EFBIG: /* Hit fs limit */ | |
603 | case -EDQUOT: /* Quota limit hit */ | |
604 | case -ENOSPC: /* Disk full */ | |
d025f1e4 | 605 | log_debug("%s: Allocation limit reached, rotating.", f->path); |
6e1045e5 | 606 | return true; |
ae739cc1 | 607 | |
6e1045e5 ZJS |
608 | case -EIO: /* I/O error of some kind (mmap) */ |
609 | log_warning("%s: IO error, rotating.", f->path); | |
610 | return true; | |
ae739cc1 | 611 | |
6e1045e5 | 612 | case -EHOSTDOWN: /* Other machine */ |
d025f1e4 | 613 | log_info("%s: Journal file from other machine, rotating.", f->path); |
6e1045e5 | 614 | return true; |
ae739cc1 | 615 | |
6e1045e5 | 616 | case -EBUSY: /* Unclean shutdown */ |
d025f1e4 | 617 | log_info("%s: Unclean shutdown, rotating.", f->path); |
6e1045e5 | 618 | return true; |
ae739cc1 | 619 | |
6e1045e5 | 620 | case -EPROTONOSUPPORT: /* Unsupported feature */ |
d025f1e4 | 621 | log_info("%s: Unsupported feature, rotating.", f->path); |
6e1045e5 | 622 | return true; |
ae739cc1 | 623 | |
6e1045e5 ZJS |
624 | case -EBADMSG: /* Corrupted */ |
625 | case -ENODATA: /* Truncated */ | |
626 | case -ESHUTDOWN: /* Already archived */ | |
d025f1e4 | 627 | log_warning("%s: Journal file corrupted, rotating.", f->path); |
6e1045e5 | 628 | return true; |
ae739cc1 | 629 | |
6e1045e5 | 630 | case -EIDRM: /* Journal file has been deleted */ |
2678031a | 631 | log_warning("%s: Journal file has been deleted, rotating.", f->path); |
6e1045e5 | 632 | return true; |
ae739cc1 LP |
633 | |
634 | case -ETXTBSY: /* Journal file is from the future */ | |
c1a9199e | 635 | log_warning("%s: Journal file is from the future, rotating.", f->path); |
ae739cc1 LP |
636 | return true; |
637 | ||
6e1045e5 | 638 | default: |
d025f1e4 | 639 | return false; |
6e1045e5 | 640 | } |
d025f1e4 ZJS |
641 | } |
642 | ||
d07f7b9e | 643 | static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n, int priority) { |
7c070017 | 644 | bool vacuumed = false, rotate = false; |
0f972d66 | 645 | struct dual_timestamp ts; |
d025f1e4 | 646 | JournalFile *f; |
d025f1e4 ZJS |
647 | int r; |
648 | ||
649 | assert(s); | |
650 | assert(iovec); | |
651 | assert(n > 0); | |
652 | ||
0f972d66 LP |
653 | /* Get the closest, linearized time we have for this log event from the event loop. (Note that we do not use |
654 | * the source time, and not even the time the event was originally seen, but instead simply the time we started | |
655 | * processing it, as we want strictly linear ordering in what we write out.) */ | |
656 | assert_se(sd_event_now(s->event, CLOCK_REALTIME, &ts.realtime) >= 0); | |
657 | assert_se(sd_event_now(s->event, CLOCK_MONOTONIC, &ts.monotonic) >= 0); | |
658 | ||
7c070017 LP |
659 | if (ts.realtime < s->last_realtime_clock) { |
660 | /* When the time jumps backwards, let's immediately rotate. Of course, this should not happen during | |
661 | * regular operation. However, when it does happen, then we should make sure that we start fresh files | |
662 | * to ensure that the entries in the journal files are strictly ordered by time, in order to ensure | |
663 | * bisection works correctly. */ | |
d025f1e4 | 664 | |
7c070017 LP |
665 | log_debug("Time jumped backwards, rotating."); |
666 | rotate = true; | |
667 | } else { | |
668 | ||
669 | f = find_journal(s, uid); | |
670 | if (!f) | |
671 | return; | |
672 | ||
673 | if (journal_file_rotate_suggested(f, s->max_file_usec)) { | |
674 | log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path); | |
675 | rotate = true; | |
676 | } | |
677 | } | |
d025f1e4 | 678 | |
7c070017 | 679 | if (rotate) { |
d025f1e4 | 680 | server_rotate(s); |
3a19f215 | 681 | server_vacuum(s, false); |
d025f1e4 ZJS |
682 | vacuumed = true; |
683 | ||
684 | f = find_journal(s, uid); | |
685 | if (!f) | |
686 | return; | |
687 | } | |
688 | ||
7c070017 LP |
689 | s->last_realtime_clock = ts.realtime; |
690 | ||
0f972d66 | 691 | r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL); |
26687bf8 | 692 | if (r >= 0) { |
d07f7b9e | 693 | server_schedule_sync(s, priority); |
d025f1e4 | 694 | return; |
26687bf8 | 695 | } |
d025f1e4 ZJS |
696 | |
697 | if (vacuumed || !shall_try_append_again(f, r)) { | |
8266e1c0 | 698 | log_error_errno(r, "Failed to write entry (%d items, %zu bytes), ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); |
d025f1e4 ZJS |
699 | return; |
700 | } | |
701 | ||
702 | server_rotate(s); | |
3a19f215 | 703 | server_vacuum(s, false); |
d025f1e4 ZJS |
704 | |
705 | f = find_journal(s, uid); | |
706 | if (!f) | |
707 | return; | |
708 | ||
709 | log_debug("Retrying write."); | |
0f972d66 | 710 | r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL); |
8266e1c0 LP |
711 | if (r < 0) |
712 | log_error_errno(r, "Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); | |
713 | else | |
d07f7b9e | 714 | server_schedule_sync(s, priority); |
d025f1e4 ZJS |
715 | } |
716 | ||
4b58153d LP |
717 | static int get_invocation_id(const char *cgroup_root, const char *slice, const char *unit, char **ret) { |
718 | _cleanup_free_ char *escaped = NULL, *slice_path = NULL, *p = NULL; | |
719 | char *copy, ids[SD_ID128_STRING_MAX]; | |
720 | int r; | |
721 | ||
722 | /* Read the invocation ID of a unit off a unit. It's stored in the "trusted.invocation_id" extended attribute | |
723 | * on the cgroup path. */ | |
724 | ||
725 | r = cg_slice_to_path(slice, &slice_path); | |
726 | if (r < 0) | |
727 | return r; | |
728 | ||
729 | escaped = cg_escape(unit); | |
730 | if (!escaped) | |
731 | return -ENOMEM; | |
732 | ||
605405c6 | 733 | p = strjoin(cgroup_root, "/", slice_path, "/", escaped); |
4b58153d LP |
734 | if (!p) |
735 | return -ENOMEM; | |
736 | ||
737 | r = cg_get_xattr(SYSTEMD_CGROUP_CONTROLLER, p, "trusted.invocation_id", ids, 32); | |
738 | if (r < 0) | |
739 | return r; | |
740 | if (r != 32) | |
741 | return -EINVAL; | |
742 | ids[32] = 0; | |
743 | ||
744 | if (!id128_is_valid(ids)) | |
745 | return -EINVAL; | |
746 | ||
747 | copy = strdup(ids); | |
748 | if (!copy) | |
749 | return -ENOMEM; | |
750 | ||
751 | *ret = copy; | |
752 | return 0; | |
753 | } | |
754 | ||
d025f1e4 ZJS |
755 | static void dispatch_message_real( |
756 | Server *s, | |
757 | struct iovec *iovec, unsigned n, unsigned m, | |
3b3154df LP |
758 | const struct ucred *ucred, |
759 | const struct timeval *tv, | |
d025f1e4 | 760 | const char *label, size_t label_len, |
968f3196 | 761 | const char *unit_id, |
d07f7b9e | 762 | int priority, |
b4e7bdcb NK |
763 | pid_t object_pid, |
764 | char *cgroup) { | |
d025f1e4 | 765 | |
968f3196 | 766 | char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)], |
ae018d9b LP |
767 | uid[sizeof("_UID=") + DECIMAL_STR_MAX(uid_t)], |
768 | gid[sizeof("_GID=") + DECIMAL_STR_MAX(gid_t)], | |
769 | owner_uid[sizeof("_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)], | |
d3789917 | 770 | source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)], |
968f3196 ZJS |
771 | o_uid[sizeof("OBJECT_UID=") + DECIMAL_STR_MAX(uid_t)], |
772 | o_gid[sizeof("OBJECT_GID=") + DECIMAL_STR_MAX(gid_t)], | |
773 | o_owner_uid[sizeof("OBJECT_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)]; | |
774 | uid_t object_uid; | |
775 | gid_t object_gid; | |
968f3196 | 776 | char *x; |
d025f1e4 | 777 | int r; |
ae018d9b | 778 | char *t, *c; |
82499507 LP |
779 | uid_t realuid = 0, owner = 0, journal_uid; |
780 | bool owner_valid = false; | |
ae018d9b | 781 | #ifdef HAVE_AUDIT |
968f3196 ZJS |
782 | char audit_session[sizeof("_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], |
783 | audit_loginuid[sizeof("_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)], | |
784 | o_audit_session[sizeof("OBJECT_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], | |
785 | o_audit_loginuid[sizeof("OBJECT_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)]; | |
ae018d9b LP |
786 | |
787 | uint32_t audit; | |
788 | uid_t loginuid; | |
789 | #endif | |
d025f1e4 ZJS |
790 | |
791 | assert(s); | |
792 | assert(iovec); | |
793 | assert(n > 0); | |
d473176a | 794 | assert(n + N_IOVEC_META_FIELDS + (object_pid > 0 ? N_IOVEC_OBJECT_FIELDS : 0) <= m); |
d025f1e4 ZJS |
795 | |
796 | if (ucred) { | |
d025f1e4 ZJS |
797 | realuid = ucred->uid; |
798 | ||
de0671ee | 799 | sprintf(pid, "_PID="PID_FMT, ucred->pid); |
c2457105 | 800 | IOVEC_SET_STRING(iovec[n++], pid); |
d025f1e4 | 801 | |
de0671ee | 802 | sprintf(uid, "_UID="UID_FMT, ucred->uid); |
c2457105 | 803 | IOVEC_SET_STRING(iovec[n++], uid); |
d025f1e4 | 804 | |
de0671ee | 805 | sprintf(gid, "_GID="GID_FMT, ucred->gid); |
c2457105 | 806 | IOVEC_SET_STRING(iovec[n++], gid); |
d025f1e4 ZJS |
807 | |
808 | r = get_process_comm(ucred->pid, &t); | |
809 | if (r >= 0) { | |
63c372cb | 810 | x = strjoina("_COMM=", t); |
d025f1e4 | 811 | free(t); |
968f3196 | 812 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
813 | } |
814 | ||
815 | r = get_process_exe(ucred->pid, &t); | |
816 | if (r >= 0) { | |
63c372cb | 817 | x = strjoina("_EXE=", t); |
d025f1e4 | 818 | free(t); |
968f3196 | 819 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
820 | } |
821 | ||
9bdbc2e2 | 822 | r = get_process_cmdline(ucred->pid, 0, false, &t); |
d025f1e4 | 823 | if (r >= 0) { |
63c372cb | 824 | x = strjoina("_CMDLINE=", t); |
d025f1e4 | 825 | free(t); |
3a832116 SL |
826 | IOVEC_SET_STRING(iovec[n++], x); |
827 | } | |
828 | ||
829 | r = get_process_capeff(ucred->pid, &t); | |
830 | if (r >= 0) { | |
63c372cb | 831 | x = strjoina("_CAP_EFFECTIVE=", t); |
3a832116 | 832 | free(t); |
968f3196 | 833 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
834 | } |
835 | ||
0a20e3c1 | 836 | #ifdef HAVE_AUDIT |
d025f1e4 | 837 | r = audit_session_from_pid(ucred->pid, &audit); |
ae018d9b | 838 | if (r >= 0) { |
de0671ee | 839 | sprintf(audit_session, "_AUDIT_SESSION=%"PRIu32, audit); |
ae018d9b LP |
840 | IOVEC_SET_STRING(iovec[n++], audit_session); |
841 | } | |
d025f1e4 ZJS |
842 | |
843 | r = audit_loginuid_from_pid(ucred->pid, &loginuid); | |
7027ff61 | 844 | if (r >= 0) { |
de0671ee | 845 | sprintf(audit_loginuid, "_AUDIT_LOGINUID="UID_FMT, loginuid); |
ae018d9b | 846 | IOVEC_SET_STRING(iovec[n++], audit_loginuid); |
d025f1e4 | 847 | } |
ae018d9b | 848 | #endif |
d025f1e4 | 849 | |
b4e7bdcb NK |
850 | r = 0; |
851 | if (cgroup) | |
852 | c = cgroup; | |
853 | else | |
854 | r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c); | |
855 | ||
7027ff61 | 856 | if (r >= 0) { |
4b58153d | 857 | _cleanup_free_ char *raw_unit = NULL, *raw_slice = NULL; |
968f3196 ZJS |
858 | char *session = NULL; |
859 | ||
63c372cb | 860 | x = strjoina("_SYSTEMD_CGROUP=", c); |
968f3196 | 861 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 | 862 | |
ae018d9b LP |
863 | r = cg_path_get_session(c, &t); |
864 | if (r >= 0) { | |
63c372cb | 865 | session = strjoina("_SYSTEMD_SESSION=", t); |
ae018d9b | 866 | free(t); |
d025f1e4 | 867 | IOVEC_SET_STRING(iovec[n++], session); |
ae018d9b LP |
868 | } |
869 | ||
870 | if (cg_path_get_owner_uid(c, &owner) >= 0) { | |
871 | owner_valid = true; | |
d025f1e4 | 872 | |
de0671ee | 873 | sprintf(owner_uid, "_SYSTEMD_OWNER_UID="UID_FMT, owner); |
d025f1e4 | 874 | IOVEC_SET_STRING(iovec[n++], owner_uid); |
ae018d9b | 875 | } |
d025f1e4 | 876 | |
4b58153d LP |
877 | if (cg_path_get_unit(c, &raw_unit) >= 0) { |
878 | x = strjoina("_SYSTEMD_UNIT=", raw_unit); | |
19cace37 LP |
879 | IOVEC_SET_STRING(iovec[n++], x); |
880 | } else if (unit_id && !session) { | |
63c372cb | 881 | x = strjoina("_SYSTEMD_UNIT=", unit_id); |
19cace37 LP |
882 | IOVEC_SET_STRING(iovec[n++], x); |
883 | } | |
884 | ||
885 | if (cg_path_get_user_unit(c, &t) >= 0) { | |
63c372cb | 886 | x = strjoina("_SYSTEMD_USER_UNIT=", t); |
ae018d9b | 887 | free(t); |
968f3196 | 888 | IOVEC_SET_STRING(iovec[n++], x); |
19cace37 | 889 | } else if (unit_id && session) { |
63c372cb | 890 | x = strjoina("_SYSTEMD_USER_UNIT=", unit_id); |
19cace37 LP |
891 | IOVEC_SET_STRING(iovec[n++], x); |
892 | } | |
ae018d9b | 893 | |
4b58153d LP |
894 | if (cg_path_get_slice(c, &raw_slice) >= 0) { |
895 | x = strjoina("_SYSTEMD_SLICE=", raw_slice); | |
0a244b8e LP |
896 | IOVEC_SET_STRING(iovec[n++], x); |
897 | } | |
898 | ||
d473176a LP |
899 | if (cg_path_get_user_slice(c, &t) >= 0) { |
900 | x = strjoina("_SYSTEMD_USER_SLICE=", t); | |
901 | free(t); | |
902 | IOVEC_SET_STRING(iovec[n++], x); | |
903 | } | |
904 | ||
4b58153d LP |
905 | if (raw_slice && raw_unit) { |
906 | if (get_invocation_id(s->cgroup_root, raw_slice, raw_unit, &t) >= 0) { | |
907 | x = strjoina("_SYSTEMD_INVOCATION_ID=", t); | |
908 | free(t); | |
909 | IOVEC_SET_STRING(iovec[n++], x); | |
910 | } | |
911 | } | |
912 | ||
b4e7bdcb NK |
913 | if (!cgroup) |
914 | free(c); | |
2d43b190 | 915 | } else if (unit_id) { |
63c372cb | 916 | x = strjoina("_SYSTEMD_UNIT=", unit_id); |
2d43b190 | 917 | IOVEC_SET_STRING(iovec[n++], x); |
ef1673d1 | 918 | } |
d025f1e4 | 919 | |
d025f1e4 | 920 | #ifdef HAVE_SELINUX |
6d395665 | 921 | if (mac_selinux_use()) { |
d682b3a7 | 922 | if (label) { |
f8294e41 | 923 | x = alloca(strlen("_SELINUX_CONTEXT=") + label_len + 1); |
ae018d9b | 924 | |
d682b3a7 LP |
925 | *((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0; |
926 | IOVEC_SET_STRING(iovec[n++], x); | |
927 | } else { | |
2ed96880 | 928 | char *con; |
d025f1e4 | 929 | |
d682b3a7 | 930 | if (getpidcon(ucred->pid, &con) >= 0) { |
63c372cb | 931 | x = strjoina("_SELINUX_CONTEXT=", con); |
e7ff4e7f | 932 | |
d682b3a7 LP |
933 | freecon(con); |
934 | IOVEC_SET_STRING(iovec[n++], x); | |
935 | } | |
d025f1e4 ZJS |
936 | } |
937 | } | |
938 | #endif | |
939 | } | |
968f3196 ZJS |
940 | assert(n <= m); |
941 | ||
942 | if (object_pid) { | |
943 | r = get_process_uid(object_pid, &object_uid); | |
944 | if (r >= 0) { | |
de0671ee | 945 | sprintf(o_uid, "OBJECT_UID="UID_FMT, object_uid); |
968f3196 ZJS |
946 | IOVEC_SET_STRING(iovec[n++], o_uid); |
947 | } | |
948 | ||
949 | r = get_process_gid(object_pid, &object_gid); | |
950 | if (r >= 0) { | |
de0671ee | 951 | sprintf(o_gid, "OBJECT_GID="GID_FMT, object_gid); |
968f3196 ZJS |
952 | IOVEC_SET_STRING(iovec[n++], o_gid); |
953 | } | |
954 | ||
955 | r = get_process_comm(object_pid, &t); | |
956 | if (r >= 0) { | |
63c372cb | 957 | x = strjoina("OBJECT_COMM=", t); |
968f3196 ZJS |
958 | free(t); |
959 | IOVEC_SET_STRING(iovec[n++], x); | |
960 | } | |
961 | ||
962 | r = get_process_exe(object_pid, &t); | |
963 | if (r >= 0) { | |
63c372cb | 964 | x = strjoina("OBJECT_EXE=", t); |
968f3196 ZJS |
965 | free(t); |
966 | IOVEC_SET_STRING(iovec[n++], x); | |
967 | } | |
968 | ||
969 | r = get_process_cmdline(object_pid, 0, false, &t); | |
970 | if (r >= 0) { | |
63c372cb | 971 | x = strjoina("OBJECT_CMDLINE=", t); |
968f3196 ZJS |
972 | free(t); |
973 | IOVEC_SET_STRING(iovec[n++], x); | |
974 | } | |
975 | ||
976 | #ifdef HAVE_AUDIT | |
977 | r = audit_session_from_pid(object_pid, &audit); | |
978 | if (r >= 0) { | |
de0671ee | 979 | sprintf(o_audit_session, "OBJECT_AUDIT_SESSION=%"PRIu32, audit); |
968f3196 ZJS |
980 | IOVEC_SET_STRING(iovec[n++], o_audit_session); |
981 | } | |
982 | ||
983 | r = audit_loginuid_from_pid(object_pid, &loginuid); | |
984 | if (r >= 0) { | |
de0671ee | 985 | sprintf(o_audit_loginuid, "OBJECT_AUDIT_LOGINUID="UID_FMT, loginuid); |
968f3196 ZJS |
986 | IOVEC_SET_STRING(iovec[n++], o_audit_loginuid); |
987 | } | |
988 | #endif | |
989 | ||
e9174f29 | 990 | r = cg_pid_get_path_shifted(object_pid, s->cgroup_root, &c); |
968f3196 | 991 | if (r >= 0) { |
63c372cb | 992 | x = strjoina("OBJECT_SYSTEMD_CGROUP=", c); |
968f3196 ZJS |
993 | IOVEC_SET_STRING(iovec[n++], x); |
994 | ||
995 | r = cg_path_get_session(c, &t); | |
996 | if (r >= 0) { | |
63c372cb | 997 | x = strjoina("OBJECT_SYSTEMD_SESSION=", t); |
968f3196 ZJS |
998 | free(t); |
999 | IOVEC_SET_STRING(iovec[n++], x); | |
1000 | } | |
1001 | ||
1002 | if (cg_path_get_owner_uid(c, &owner) >= 0) { | |
de0671ee | 1003 | sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID="UID_FMT, owner); |
968f3196 ZJS |
1004 | IOVEC_SET_STRING(iovec[n++], o_owner_uid); |
1005 | } | |
1006 | ||
1007 | if (cg_path_get_unit(c, &t) >= 0) { | |
63c372cb | 1008 | x = strjoina("OBJECT_SYSTEMD_UNIT=", t); |
968f3196 | 1009 | free(t); |
19cace37 LP |
1010 | IOVEC_SET_STRING(iovec[n++], x); |
1011 | } | |
1012 | ||
1013 | if (cg_path_get_user_unit(c, &t) >= 0) { | |
63c372cb | 1014 | x = strjoina("OBJECT_SYSTEMD_USER_UNIT=", t); |
968f3196 | 1015 | free(t); |
968f3196 | 1016 | IOVEC_SET_STRING(iovec[n++], x); |
19cace37 | 1017 | } |
968f3196 | 1018 | |
d473176a LP |
1019 | if (cg_path_get_slice(c, &t) >= 0) { |
1020 | x = strjoina("OBJECT_SYSTEMD_SLICE=", t); | |
1021 | free(t); | |
1022 | IOVEC_SET_STRING(iovec[n++], x); | |
1023 | } | |
1024 | ||
1025 | if (cg_path_get_user_slice(c, &t) >= 0) { | |
1026 | x = strjoina("OBJECT_SYSTEMD_USER_SLICE=", t); | |
1027 | free(t); | |
1028 | IOVEC_SET_STRING(iovec[n++], x); | |
1029 | } | |
1030 | ||
968f3196 ZJS |
1031 | free(c); |
1032 | } | |
1033 | } | |
1034 | assert(n <= m); | |
d025f1e4 ZJS |
1035 | |
1036 | if (tv) { | |
398a50cd | 1037 | sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=" USEC_FMT, timeval_load(tv)); |
a5693989 | 1038 | IOVEC_SET_STRING(iovec[n++], source_time); |
d025f1e4 ZJS |
1039 | } |
1040 | ||
1041 | /* Note that strictly speaking storing the boot id here is | |
1042 | * redundant since the entry includes this in-line | |
1043 | * anyway. However, we need this indexed, too. */ | |
0c24bb23 LP |
1044 | if (!isempty(s->boot_id_field)) |
1045 | IOVEC_SET_STRING(iovec[n++], s->boot_id_field); | |
d025f1e4 | 1046 | |
0c24bb23 LP |
1047 | if (!isempty(s->machine_id_field)) |
1048 | IOVEC_SET_STRING(iovec[n++], s->machine_id_field); | |
d025f1e4 | 1049 | |
0c24bb23 LP |
1050 | if (!isempty(s->hostname_field)) |
1051 | IOVEC_SET_STRING(iovec[n++], s->hostname_field); | |
d025f1e4 ZJS |
1052 | |
1053 | assert(n <= m); | |
1054 | ||
da499392 | 1055 | if (s->split_mode == SPLIT_UID && realuid > 0) |
40adcda8 | 1056 | /* Split up strictly by any UID */ |
759c945a | 1057 | journal_uid = realuid; |
82499507 | 1058 | else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0) |
edc3797f LP |
1059 | /* Split up by login UIDs. We do this only if the |
1060 | * realuid is not root, in order not to accidentally | |
1061 | * leak privileged information to the user that is | |
1062 | * logged by a privileged process that is part of an | |
7517e174 | 1063 | * unprivileged session. */ |
8a0889df | 1064 | journal_uid = owner; |
da499392 KS |
1065 | else |
1066 | journal_uid = 0; | |
759c945a | 1067 | |
d07f7b9e | 1068 | write_to_journal(s, journal_uid, iovec, n, priority); |
d025f1e4 ZJS |
1069 | } |
1070 | ||
2b044526 | 1071 | void server_driver_message(Server *s, const char *message_id, const char *format, ...) { |
8a03c9ef ZJS |
1072 | struct iovec iovec[N_IOVEC_META_FIELDS + 5 + N_IOVEC_PAYLOAD_FIELDS]; |
1073 | unsigned n = 0, m; | |
32917e33 | 1074 | int r; |
d025f1e4 | 1075 | va_list ap; |
b92bea5d | 1076 | struct ucred ucred = {}; |
d025f1e4 ZJS |
1077 | |
1078 | assert(s); | |
1079 | assert(format); | |
1080 | ||
4850d39a | 1081 | assert_cc(3 == LOG_FAC(LOG_DAEMON)); |
b6fa2555 EV |
1082 | IOVEC_SET_STRING(iovec[n++], "SYSLOG_FACILITY=3"); |
1083 | IOVEC_SET_STRING(iovec[n++], "SYSLOG_IDENTIFIER=systemd-journald"); | |
1084 | ||
d025f1e4 | 1085 | IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver"); |
4850d39a | 1086 | assert_cc(6 == LOG_INFO); |
32917e33 | 1087 | IOVEC_SET_STRING(iovec[n++], "PRIORITY=6"); |
d025f1e4 | 1088 | |
2b044526 ZJS |
1089 | if (message_id) |
1090 | IOVEC_SET_STRING(iovec[n++], message_id); | |
8a03c9ef ZJS |
1091 | m = n; |
1092 | ||
1093 | va_start(ap, format); | |
32917e33 ZJS |
1094 | r = log_format_iovec(iovec, ELEMENTSOF(iovec), &n, false, 0, format, ap); |
1095 | /* Error handling below */ | |
8a03c9ef ZJS |
1096 | va_end(ap); |
1097 | ||
d025f1e4 ZJS |
1098 | ucred.pid = getpid(); |
1099 | ucred.uid = getuid(); | |
1100 | ucred.gid = getgid(); | |
1101 | ||
32917e33 | 1102 | if (r >= 0) |
b4e7bdcb | 1103 | dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0, NULL); |
8a03c9ef ZJS |
1104 | |
1105 | while (m < n) | |
1106 | free(iovec[m++].iov_base); | |
32917e33 ZJS |
1107 | |
1108 | if (r < 0) { | |
1109 | /* We failed to format the message. Emit a warning instead. */ | |
1110 | char buf[LINE_MAX]; | |
1111 | ||
1112 | xsprintf(buf, "MESSAGE=Entry printing failed: %s", strerror(-r)); | |
1113 | ||
1114 | n = 3; | |
1115 | IOVEC_SET_STRING(iovec[n++], "PRIORITY=4"); | |
1116 | IOVEC_SET_STRING(iovec[n++], buf); | |
b4e7bdcb | 1117 | dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0, NULL); |
32917e33 | 1118 | } |
d025f1e4 ZJS |
1119 | } |
1120 | ||
1121 | void server_dispatch_message( | |
1122 | Server *s, | |
1123 | struct iovec *iovec, unsigned n, unsigned m, | |
3b3154df LP |
1124 | const struct ucred *ucred, |
1125 | const struct timeval *tv, | |
d025f1e4 ZJS |
1126 | const char *label, size_t label_len, |
1127 | const char *unit_id, | |
968f3196 ZJS |
1128 | int priority, |
1129 | pid_t object_pid) { | |
d025f1e4 | 1130 | |
7027ff61 | 1131 | int rl, r; |
7fd1b19b | 1132 | _cleanup_free_ char *path = NULL; |
8580d1f7 | 1133 | uint64_t available = 0; |
b4e7bdcb | 1134 | char *c = NULL; |
d025f1e4 ZJS |
1135 | |
1136 | assert(s); | |
1137 | assert(iovec || n == 0); | |
1138 | ||
1139 | if (n == 0) | |
1140 | return; | |
1141 | ||
1142 | if (LOG_PRI(priority) > s->max_level_store) | |
1143 | return; | |
1144 | ||
2f5df74a HHPF |
1145 | /* Stop early in case the information will not be stored |
1146 | * in a journal. */ | |
1147 | if (s->storage == STORAGE_NONE) | |
1148 | return; | |
1149 | ||
d025f1e4 ZJS |
1150 | if (!ucred) |
1151 | goto finish; | |
1152 | ||
e9174f29 | 1153 | r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path); |
7027ff61 | 1154 | if (r < 0) |
d025f1e4 ZJS |
1155 | goto finish; |
1156 | ||
1157 | /* example: /user/lennart/3/foobar | |
1158 | * /system/dbus.service/foobar | |
1159 | * | |
1160 | * So let's cut of everything past the third /, since that is | |
1161 | * where user directories start */ | |
1162 | ||
1163 | c = strchr(path, '/'); | |
1164 | if (c) { | |
1165 | c = strchr(c+1, '/'); | |
1166 | if (c) { | |
1167 | c = strchr(c+1, '/'); | |
1168 | if (c) | |
1169 | *c = 0; | |
1170 | } | |
1171 | } | |
1172 | ||
3a19f215 | 1173 | (void) determine_space(s, &available, NULL); |
8580d1f7 | 1174 | rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available); |
db91ea32 | 1175 | if (rl == 0) |
d025f1e4 | 1176 | return; |
d025f1e4 ZJS |
1177 | |
1178 | /* Write a suppression message if we suppressed something */ | |
1179 | if (rl > 1) | |
2b044526 | 1180 | server_driver_message(s, "MESSAGE_ID=" SD_MESSAGE_JOURNAL_DROPPED_STR, |
8a03c9ef ZJS |
1181 | LOG_MESSAGE("Suppressed %u messages from %s", rl - 1, path), |
1182 | NULL); | |
d025f1e4 ZJS |
1183 | |
1184 | finish: | |
b4e7bdcb NK |
1185 | /* restore cgroup path for logging */ |
1186 | if (c) | |
1187 | *c = '/'; | |
1188 | dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid, path); | |
d025f1e4 ZJS |
1189 | } |
1190 | ||
f78273c8 | 1191 | int server_flush_to_var(Server *s, bool require_flag_file) { |
d025f1e4 ZJS |
1192 | sd_id128_t machine; |
1193 | sd_journal *j = NULL; | |
fbb63411 LP |
1194 | char ts[FORMAT_TIMESPAN_MAX]; |
1195 | usec_t start; | |
1196 | unsigned n = 0; | |
1197 | int r; | |
d025f1e4 ZJS |
1198 | |
1199 | assert(s); | |
1200 | ||
f78273c8 | 1201 | if (!IN_SET(s->storage, STORAGE_AUTO, STORAGE_PERSISTENT)) |
d025f1e4 ZJS |
1202 | return 0; |
1203 | ||
1204 | if (!s->runtime_journal) | |
1205 | return 0; | |
1206 | ||
f78273c8 LP |
1207 | if (require_flag_file && !flushed_flag_is_set()) |
1208 | return 0; | |
1209 | ||
8580d1f7 | 1210 | (void) system_journal_open(s, true); |
d025f1e4 ZJS |
1211 | |
1212 | if (!s->system_journal) | |
1213 | return 0; | |
1214 | ||
1215 | log_debug("Flushing to /var..."); | |
1216 | ||
fbb63411 LP |
1217 | start = now(CLOCK_MONOTONIC); |
1218 | ||
d025f1e4 | 1219 | r = sd_id128_get_machine(&machine); |
00a16861 | 1220 | if (r < 0) |
d025f1e4 | 1221 | return r; |
d025f1e4 ZJS |
1222 | |
1223 | r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); | |
23bbb0de MS |
1224 | if (r < 0) |
1225 | return log_error_errno(r, "Failed to read runtime journal: %m"); | |
d025f1e4 | 1226 | |
93b73b06 LP |
1227 | sd_journal_set_data_threshold(j, 0); |
1228 | ||
d025f1e4 ZJS |
1229 | SD_JOURNAL_FOREACH(j) { |
1230 | Object *o = NULL; | |
1231 | JournalFile *f; | |
1232 | ||
1233 | f = j->current_file; | |
1234 | assert(f && f->current_offset > 0); | |
1235 | ||
fbb63411 LP |
1236 | n++; |
1237 | ||
d025f1e4 ZJS |
1238 | r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); |
1239 | if (r < 0) { | |
da927ba9 | 1240 | log_error_errno(r, "Can't read entry: %m"); |
d025f1e4 ZJS |
1241 | goto finish; |
1242 | } | |
1243 | ||
1244 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
1245 | if (r >= 0) | |
1246 | continue; | |
1247 | ||
1248 | if (!shall_try_append_again(s->system_journal, r)) { | |
da927ba9 | 1249 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1250 | goto finish; |
1251 | } | |
1252 | ||
1253 | server_rotate(s); | |
3a19f215 | 1254 | server_vacuum(s, false); |
d025f1e4 | 1255 | |
253f59df LP |
1256 | if (!s->system_journal) { |
1257 | log_notice("Didn't flush runtime journal since rotation of system journal wasn't successful."); | |
1258 | r = -EIO; | |
1259 | goto finish; | |
1260 | } | |
1261 | ||
d025f1e4 ZJS |
1262 | log_debug("Retrying write."); |
1263 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
1264 | if (r < 0) { | |
da927ba9 | 1265 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1266 | goto finish; |
1267 | } | |
1268 | } | |
1269 | ||
804ae586 LP |
1270 | r = 0; |
1271 | ||
d025f1e4 ZJS |
1272 | finish: |
1273 | journal_file_post_change(s->system_journal); | |
1274 | ||
804ae586 | 1275 | s->runtime_journal = journal_file_close(s->runtime_journal); |
d025f1e4 ZJS |
1276 | |
1277 | if (r >= 0) | |
c6878637 | 1278 | (void) rm_rf("/run/log/journal", REMOVE_ROOT); |
d025f1e4 | 1279 | |
763c7aa2 | 1280 | sd_journal_close(j); |
d025f1e4 | 1281 | |
2b044526 | 1282 | server_driver_message(s, NULL, |
8a03c9ef ZJS |
1283 | LOG_MESSAGE("Time spent on flushing to /var is %s for %u entries.", |
1284 | format_timespan(ts, sizeof(ts), now(CLOCK_MONOTONIC) - start, 0), | |
1285 | n), | |
1286 | NULL); | |
fbb63411 | 1287 | |
d025f1e4 ZJS |
1288 | return r; |
1289 | } | |
1290 | ||
8531ae70 | 1291 | int server_process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
f9a810be | 1292 | Server *s = userdata; |
a315ac4e LP |
1293 | struct ucred *ucred = NULL; |
1294 | struct timeval *tv = NULL; | |
1295 | struct cmsghdr *cmsg; | |
1296 | char *label = NULL; | |
1297 | size_t label_len = 0, m; | |
1298 | struct iovec iovec; | |
1299 | ssize_t n; | |
1300 | int *fds = NULL, v = 0; | |
1301 | unsigned n_fds = 0; | |
1302 | ||
1303 | union { | |
1304 | struct cmsghdr cmsghdr; | |
1305 | ||
1306 | /* We use NAME_MAX space for the SELinux label | |
1307 | * here. The kernel currently enforces no | |
1308 | * limit, but according to suggestions from | |
1309 | * the SELinux people this will change and it | |
1310 | * will probably be identical to NAME_MAX. For | |
1311 | * now we use that, but this should be updated | |
1312 | * one day when the final limit is known. */ | |
1313 | uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) + | |
1314 | CMSG_SPACE(sizeof(struct timeval)) + | |
1315 | CMSG_SPACE(sizeof(int)) + /* fd */ | |
1316 | CMSG_SPACE(NAME_MAX)]; /* selinux label */ | |
1317 | } control = {}; | |
1318 | ||
1319 | union sockaddr_union sa = {}; | |
1320 | ||
1321 | struct msghdr msghdr = { | |
1322 | .msg_iov = &iovec, | |
1323 | .msg_iovlen = 1, | |
1324 | .msg_control = &control, | |
1325 | .msg_controllen = sizeof(control), | |
1326 | .msg_name = &sa, | |
1327 | .msg_namelen = sizeof(sa), | |
1328 | }; | |
f9a810be | 1329 | |
d025f1e4 | 1330 | assert(s); |
875c2e22 | 1331 | assert(fd == s->native_fd || fd == s->syslog_fd || fd == s->audit_fd); |
f9a810be LP |
1332 | |
1333 | if (revents != EPOLLIN) { | |
1334 | log_error("Got invalid event from epoll for datagram fd: %"PRIx32, revents); | |
1335 | return -EIO; | |
1336 | } | |
1337 | ||
a315ac4e LP |
1338 | /* Try to get the right size, if we can. (Not all |
1339 | * sockets support SIOCINQ, hence we just try, but | |
1340 | * don't rely on it. */ | |
1341 | (void) ioctl(fd, SIOCINQ, &v); | |
d025f1e4 | 1342 | |
a315ac4e LP |
1343 | /* Fix it up, if it is too small. We use the same fixed value as auditd here. Awful! */ |
1344 | m = PAGE_ALIGN(MAX3((size_t) v + 1, | |
1345 | (size_t) LINE_MAX, | |
1346 | ALIGN(sizeof(struct nlmsghdr)) + ALIGN((size_t) MAX_AUDIT_MESSAGE_LENGTH)) + 1); | |
d025f1e4 | 1347 | |
a315ac4e LP |
1348 | if (!GREEDY_REALLOC(s->buffer, s->buffer_size, m)) |
1349 | return log_oom(); | |
875c2e22 | 1350 | |
a315ac4e LP |
1351 | iovec.iov_base = s->buffer; |
1352 | iovec.iov_len = s->buffer_size - 1; /* Leave room for trailing NUL we add later */ | |
d025f1e4 | 1353 | |
a315ac4e LP |
1354 | n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); |
1355 | if (n < 0) { | |
1356 | if (errno == EINTR || errno == EAGAIN) | |
1357 | return 0; | |
875c2e22 | 1358 | |
a315ac4e LP |
1359 | return log_error_errno(errno, "recvmsg() failed: %m"); |
1360 | } | |
875c2e22 | 1361 | |
a315ac4e LP |
1362 | CMSG_FOREACH(cmsg, &msghdr) { |
1363 | ||
1364 | if (cmsg->cmsg_level == SOL_SOCKET && | |
1365 | cmsg->cmsg_type == SCM_CREDENTIALS && | |
1366 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) | |
1367 | ucred = (struct ucred*) CMSG_DATA(cmsg); | |
1368 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1369 | cmsg->cmsg_type == SCM_SECURITY) { | |
1370 | label = (char*) CMSG_DATA(cmsg); | |
1371 | label_len = cmsg->cmsg_len - CMSG_LEN(0); | |
1372 | } else if (cmsg->cmsg_level == SOL_SOCKET && | |
1373 | cmsg->cmsg_type == SO_TIMESTAMP && | |
1374 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval))) | |
1375 | tv = (struct timeval*) CMSG_DATA(cmsg); | |
1376 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1377 | cmsg->cmsg_type == SCM_RIGHTS) { | |
1378 | fds = (int*) CMSG_DATA(cmsg); | |
1379 | n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); | |
d025f1e4 | 1380 | } |
a315ac4e | 1381 | } |
d025f1e4 | 1382 | |
a315ac4e LP |
1383 | /* And a trailing NUL, just in case */ |
1384 | s->buffer[n] = 0; | |
1385 | ||
1386 | if (fd == s->syslog_fd) { | |
1387 | if (n > 0 && n_fds == 0) | |
1388 | server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len); | |
1389 | else if (n_fds > 0) | |
1390 | log_warning("Got file descriptors via syslog socket. Ignoring."); | |
1391 | ||
1392 | } else if (fd == s->native_fd) { | |
1393 | if (n > 0 && n_fds == 0) | |
1394 | server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len); | |
1395 | else if (n == 0 && n_fds == 1) | |
1396 | server_process_native_file(s, fds[0], ucred, tv, label, label_len); | |
1397 | else if (n_fds > 0) | |
1398 | log_warning("Got too many file descriptors via native socket. Ignoring."); | |
1399 | ||
1400 | } else { | |
1401 | assert(fd == s->audit_fd); | |
1402 | ||
1403 | if (n > 0 && n_fds == 0) | |
1404 | server_process_audit_message(s, s->buffer, n, ucred, &sa, msghdr.msg_namelen); | |
1405 | else if (n_fds > 0) | |
1406 | log_warning("Got file descriptors via audit socket. Ignoring."); | |
f9a810be | 1407 | } |
a315ac4e LP |
1408 | |
1409 | close_many(fds, n_fds); | |
1410 | return 0; | |
f9a810be | 1411 | } |
d025f1e4 | 1412 | |
f9a810be LP |
1413 | static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1414 | Server *s = userdata; | |
33d52ab9 | 1415 | int r; |
d025f1e4 | 1416 | |
f9a810be | 1417 | assert(s); |
d025f1e4 | 1418 | |
94b65516 | 1419 | log_info("Received request to flush runtime journal from PID " PID_FMT, si->ssi_pid); |
d025f1e4 | 1420 | |
f78273c8 | 1421 | (void) server_flush_to_var(s, false); |
f9a810be | 1422 | server_sync(s); |
3a19f215 | 1423 | server_vacuum(s, false); |
d025f1e4 | 1424 | |
33d52ab9 LP |
1425 | r = touch("/run/systemd/journal/flushed"); |
1426 | if (r < 0) | |
1427 | log_warning_errno(r, "Failed to touch /run/systemd/journal/flushed, ignoring: %m"); | |
74055aa7 | 1428 | |
18e758bf | 1429 | server_space_usage_message(s, NULL); |
f9a810be LP |
1430 | return 0; |
1431 | } | |
d025f1e4 | 1432 | |
f9a810be LP |
1433 | static int dispatch_sigusr2(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1434 | Server *s = userdata; | |
33d52ab9 | 1435 | int r; |
d025f1e4 | 1436 | |
f9a810be | 1437 | assert(s); |
d025f1e4 | 1438 | |
94b65516 | 1439 | log_info("Received request to rotate journal from PID " PID_FMT, si->ssi_pid); |
f9a810be | 1440 | server_rotate(s); |
3a19f215 FB |
1441 | server_vacuum(s, true); |
1442 | ||
1443 | if (s->system_journal) | |
1444 | patch_min_use(&s->system_storage); | |
1445 | if (s->runtime_journal) | |
1446 | patch_min_use(&s->runtime_storage); | |
d025f1e4 | 1447 | |
dbd6e31c | 1448 | /* Let clients know when the most recent rotation happened. */ |
33d52ab9 LP |
1449 | r = write_timestamp_file_atomic("/run/systemd/journal/rotated", now(CLOCK_MONOTONIC)); |
1450 | if (r < 0) | |
1451 | log_warning_errno(r, "Failed to write /run/systemd/journal/rotated, ignoring: %m"); | |
dbd6e31c | 1452 | |
f9a810be LP |
1453 | return 0; |
1454 | } | |
d025f1e4 | 1455 | |
f9a810be LP |
1456 | static int dispatch_sigterm(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1457 | Server *s = userdata; | |
d025f1e4 | 1458 | |
f9a810be | 1459 | assert(s); |
d025f1e4 | 1460 | |
4daf54a8 | 1461 | log_received_signal(LOG_INFO, si); |
d025f1e4 | 1462 | |
6203e07a | 1463 | sd_event_exit(s->event, 0); |
d025f1e4 ZJS |
1464 | return 0; |
1465 | } | |
1466 | ||
94b65516 LP |
1467 | static int dispatch_sigrtmin1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1468 | Server *s = userdata; | |
33d52ab9 | 1469 | int r; |
94b65516 LP |
1470 | |
1471 | assert(s); | |
1472 | ||
1473 | log_debug("Received request to sync from PID " PID_FMT, si->ssi_pid); | |
1474 | ||
1475 | server_sync(s); | |
1476 | ||
1477 | /* Let clients know when the most recent sync happened. */ | |
33d52ab9 LP |
1478 | r = write_timestamp_file_atomic("/run/systemd/journal/synced", now(CLOCK_MONOTONIC)); |
1479 | if (r < 0) | |
1480 | log_warning_errno(r, "Failed to write /run/systemd/journal/synced, ignoring: %m"); | |
94b65516 LP |
1481 | |
1482 | return 0; | |
1483 | } | |
1484 | ||
f9a810be | 1485 | static int setup_signals(Server *s) { |
f9a810be | 1486 | int r; |
d025f1e4 ZJS |
1487 | |
1488 | assert(s); | |
1489 | ||
9bab3b65 | 1490 | assert_se(sigprocmask_many(SIG_SETMASK, NULL, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, SIGRTMIN+1, -1) >= 0); |
d025f1e4 | 1491 | |
151b9b96 | 1492 | r = sd_event_add_signal(s->event, &s->sigusr1_event_source, SIGUSR1, dispatch_sigusr1, s); |
f9a810be LP |
1493 | if (r < 0) |
1494 | return r; | |
1495 | ||
151b9b96 | 1496 | r = sd_event_add_signal(s->event, &s->sigusr2_event_source, SIGUSR2, dispatch_sigusr2, s); |
f9a810be LP |
1497 | if (r < 0) |
1498 | return r; | |
d025f1e4 | 1499 | |
151b9b96 | 1500 | r = sd_event_add_signal(s->event, &s->sigterm_event_source, SIGTERM, dispatch_sigterm, s); |
f9a810be LP |
1501 | if (r < 0) |
1502 | return r; | |
d025f1e4 | 1503 | |
b374689c LP |
1504 | /* Let's process SIGTERM late, so that we flush all queued |
1505 | * messages to disk before we exit */ | |
1506 | r = sd_event_source_set_priority(s->sigterm_event_source, SD_EVENT_PRIORITY_NORMAL+20); | |
1507 | if (r < 0) | |
1508 | return r; | |
1509 | ||
1510 | /* When journald is invoked on the terminal (when debugging), | |
1511 | * it's useful if C-c is handled equivalent to SIGTERM. */ | |
151b9b96 | 1512 | r = sd_event_add_signal(s->event, &s->sigint_event_source, SIGINT, dispatch_sigterm, s); |
f9a810be LP |
1513 | if (r < 0) |
1514 | return r; | |
d025f1e4 | 1515 | |
b374689c LP |
1516 | r = sd_event_source_set_priority(s->sigint_event_source, SD_EVENT_PRIORITY_NORMAL+20); |
1517 | if (r < 0) | |
1518 | return r; | |
1519 | ||
94b65516 LP |
1520 | /* SIGRTMIN+1 causes an immediate sync. We process this very |
1521 | * late, so that everything else queued at this point is | |
1522 | * really written to disk. Clients can watch | |
1523 | * /run/systemd/journal/synced with inotify until its mtime | |
1524 | * changes to see when a sync happened. */ | |
1525 | r = sd_event_add_signal(s->event, &s->sigrtmin1_event_source, SIGRTMIN+1, dispatch_sigrtmin1, s); | |
1526 | if (r < 0) | |
1527 | return r; | |
1528 | ||
1529 | r = sd_event_source_set_priority(s->sigrtmin1_event_source, SD_EVENT_PRIORITY_NORMAL+15); | |
1530 | if (r < 0) | |
1531 | return r; | |
1532 | ||
d025f1e4 ZJS |
1533 | return 0; |
1534 | } | |
1535 | ||
5707ecf3 ZJS |
1536 | static int parse_proc_cmdline_item(const char *key, const char *value, void *data) { |
1537 | Server *s = data; | |
74df0fca | 1538 | int r; |
d025f1e4 | 1539 | |
5707ecf3 | 1540 | assert(s); |
d025f1e4 | 1541 | |
1d84ad94 LP |
1542 | if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_syslog")) { |
1543 | ||
5707ecf3 | 1544 | r = value ? parse_boolean(value) : true; |
d581d9d9 | 1545 | if (r < 0) |
5707ecf3 ZJS |
1546 | log_warning("Failed to parse forward to syslog switch \"%s\". Ignoring.", value); |
1547 | else | |
1548 | s->forward_to_syslog = r; | |
1d84ad94 LP |
1549 | |
1550 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_kmsg")) { | |
1551 | ||
5707ecf3 ZJS |
1552 | r = value ? parse_boolean(value) : true; |
1553 | if (r < 0) | |
1554 | log_warning("Failed to parse forward to kmsg switch \"%s\". Ignoring.", value); | |
1555 | else | |
1556 | s->forward_to_kmsg = r; | |
1d84ad94 LP |
1557 | |
1558 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_console")) { | |
1559 | ||
5707ecf3 ZJS |
1560 | r = value ? parse_boolean(value) : true; |
1561 | if (r < 0) | |
1562 | log_warning("Failed to parse forward to console switch \"%s\". Ignoring.", value); | |
1563 | else | |
1564 | s->forward_to_console = r; | |
1d84ad94 LP |
1565 | |
1566 | } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_wall")) { | |
1567 | ||
5707ecf3 ZJS |
1568 | r = value ? parse_boolean(value) : true; |
1569 | if (r < 0) | |
1570 | log_warning("Failed to parse forward to wall switch \"%s\". Ignoring.", value); | |
1571 | else | |
1572 | s->forward_to_wall = r; | |
1d84ad94 LP |
1573 | |
1574 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_console")) { | |
1575 | ||
1576 | if (proc_cmdline_value_missing(key, value)) | |
1577 | return 0; | |
1578 | ||
5707ecf3 ZJS |
1579 | r = log_level_from_string(value); |
1580 | if (r < 0) | |
1581 | log_warning("Failed to parse max level console value \"%s\". Ignoring.", value); | |
1582 | else | |
1583 | s->max_level_console = r; | |
1d84ad94 LP |
1584 | |
1585 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_store")) { | |
1586 | ||
1587 | if (proc_cmdline_value_missing(key, value)) | |
1588 | return 0; | |
1589 | ||
5707ecf3 ZJS |
1590 | r = log_level_from_string(value); |
1591 | if (r < 0) | |
1592 | log_warning("Failed to parse max level store value \"%s\". Ignoring.", value); | |
1593 | else | |
1594 | s->max_level_store = r; | |
1d84ad94 LP |
1595 | |
1596 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_syslog")) { | |
1597 | ||
1598 | if (proc_cmdline_value_missing(key, value)) | |
1599 | return 0; | |
1600 | ||
5707ecf3 ZJS |
1601 | r = log_level_from_string(value); |
1602 | if (r < 0) | |
1603 | log_warning("Failed to parse max level syslog value \"%s\". Ignoring.", value); | |
1604 | else | |
1605 | s->max_level_syslog = r; | |
1d84ad94 LP |
1606 | |
1607 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_kmsg")) { | |
1608 | ||
1609 | if (proc_cmdline_value_missing(key, value)) | |
1610 | return 0; | |
1611 | ||
5707ecf3 ZJS |
1612 | r = log_level_from_string(value); |
1613 | if (r < 0) | |
1614 | log_warning("Failed to parse max level kmsg value \"%s\". Ignoring.", value); | |
1615 | else | |
1616 | s->max_level_kmsg = r; | |
1d84ad94 LP |
1617 | |
1618 | } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_wall")) { | |
1619 | ||
1620 | if (proc_cmdline_value_missing(key, value)) | |
1621 | return 0; | |
1622 | ||
5707ecf3 ZJS |
1623 | r = log_level_from_string(value); |
1624 | if (r < 0) | |
1625 | log_warning("Failed to parse max level wall value \"%s\". Ignoring.", value); | |
1626 | else | |
1627 | s->max_level_wall = r; | |
1d84ad94 | 1628 | |
5707ecf3 ZJS |
1629 | } else if (startswith(key, "systemd.journald")) |
1630 | log_warning("Unknown journald kernel command line option \"%s\". Ignoring.", key); | |
d025f1e4 | 1631 | |
804ae586 | 1632 | /* do not warn about state here, since probably systemd already did */ |
db91ea32 | 1633 | return 0; |
d025f1e4 ZJS |
1634 | } |
1635 | ||
1636 | static int server_parse_config_file(Server *s) { | |
d025f1e4 ZJS |
1637 | assert(s); |
1638 | ||
43688c49 | 1639 | return config_parse_many_nulstr(PKGSYSCONFDIR "/journald.conf", |
da412854 YW |
1640 | CONF_PATHS_NULSTR("systemd/journald.conf.d"), |
1641 | "Journal\0", | |
1642 | config_item_perf_lookup, journald_gperf_lookup, | |
1643 | false, s); | |
d025f1e4 ZJS |
1644 | } |
1645 | ||
f9a810be LP |
1646 | static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) { |
1647 | Server *s = userdata; | |
26687bf8 OS |
1648 | |
1649 | assert(s); | |
1650 | ||
f9a810be | 1651 | server_sync(s); |
26687bf8 OS |
1652 | return 0; |
1653 | } | |
1654 | ||
d07f7b9e | 1655 | int server_schedule_sync(Server *s, int priority) { |
26687bf8 OS |
1656 | int r; |
1657 | ||
26687bf8 OS |
1658 | assert(s); |
1659 | ||
d07f7b9e LP |
1660 | if (priority <= LOG_CRIT) { |
1661 | /* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */ | |
1662 | server_sync(s); | |
1663 | return 0; | |
1664 | } | |
1665 | ||
26687bf8 OS |
1666 | if (s->sync_scheduled) |
1667 | return 0; | |
1668 | ||
f9a810be LP |
1669 | if (s->sync_interval_usec > 0) { |
1670 | usec_t when; | |
ca267016 | 1671 | |
6a0f1f6d | 1672 | r = sd_event_now(s->event, CLOCK_MONOTONIC, &when); |
f9a810be LP |
1673 | if (r < 0) |
1674 | return r; | |
26687bf8 | 1675 | |
f9a810be LP |
1676 | when += s->sync_interval_usec; |
1677 | ||
1678 | if (!s->sync_event_source) { | |
6a0f1f6d LP |
1679 | r = sd_event_add_time( |
1680 | s->event, | |
1681 | &s->sync_event_source, | |
1682 | CLOCK_MONOTONIC, | |
1683 | when, 0, | |
1684 | server_dispatch_sync, s); | |
f9a810be LP |
1685 | if (r < 0) |
1686 | return r; | |
1687 | ||
1688 | r = sd_event_source_set_priority(s->sync_event_source, SD_EVENT_PRIORITY_IMPORTANT); | |
1689 | } else { | |
1690 | r = sd_event_source_set_time(s->sync_event_source, when); | |
1691 | if (r < 0) | |
1692 | return r; | |
1693 | ||
1694 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_ONESHOT); | |
1695 | } | |
26687bf8 | 1696 | if (r < 0) |
f9a810be | 1697 | return r; |
26687bf8 | 1698 | |
f9a810be LP |
1699 | s->sync_scheduled = true; |
1700 | } | |
26687bf8 OS |
1701 | |
1702 | return 0; | |
1703 | } | |
1704 | ||
0c24bb23 LP |
1705 | static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1706 | Server *s = userdata; | |
1707 | ||
1708 | assert(s); | |
1709 | ||
1710 | server_cache_hostname(s); | |
1711 | return 0; | |
1712 | } | |
1713 | ||
1714 | static int server_open_hostname(Server *s) { | |
1715 | int r; | |
1716 | ||
1717 | assert(s); | |
1718 | ||
1719 | s->hostname_fd = open("/proc/sys/kernel/hostname", O_RDONLY|O_CLOEXEC|O_NDELAY|O_NOCTTY); | |
4a62c710 MS |
1720 | if (s->hostname_fd < 0) |
1721 | return log_error_errno(errno, "Failed to open /proc/sys/kernel/hostname: %m"); | |
0c24bb23 | 1722 | |
151b9b96 | 1723 | r = sd_event_add_io(s->event, &s->hostname_event_source, s->hostname_fd, 0, dispatch_hostname_change, s); |
0c24bb23 | 1724 | if (r < 0) { |
28def94c DR |
1725 | /* kernels prior to 3.2 don't support polling this file. Ignore |
1726 | * the failure. */ | |
1727 | if (r == -EPERM) { | |
e53fc357 | 1728 | log_warning_errno(r, "Failed to register hostname fd in event loop, ignoring: %m"); |
03e334a1 | 1729 | s->hostname_fd = safe_close(s->hostname_fd); |
28def94c DR |
1730 | return 0; |
1731 | } | |
1732 | ||
23bbb0de | 1733 | return log_error_errno(r, "Failed to register hostname fd in event loop: %m"); |
0c24bb23 LP |
1734 | } |
1735 | ||
1736 | r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10); | |
23bbb0de MS |
1737 | if (r < 0) |
1738 | return log_error_errno(r, "Failed to adjust priority of host name event source: %m"); | |
0c24bb23 LP |
1739 | |
1740 | return 0; | |
1741 | } | |
1742 | ||
e22aa3d3 LP |
1743 | static int dispatch_notify_event(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1744 | Server *s = userdata; | |
1745 | int r; | |
1746 | ||
1747 | assert(s); | |
1748 | assert(s->notify_event_source == es); | |
1749 | assert(s->notify_fd == fd); | |
1750 | ||
e22aa3d3 | 1751 | /* The $NOTIFY_SOCKET is writable again, now send exactly one |
dd835265 | 1752 | * message on it. Either it's the watchdog event, the initial |
119e9655 LP |
1753 | * READY=1 event or an stdout stream event. If there's nothing |
1754 | * to write anymore, turn our event source off. The next time | |
1755 | * there's something to send it will be turned on again. */ | |
e22aa3d3 LP |
1756 | |
1757 | if (!s->sent_notify_ready) { | |
1758 | static const char p[] = | |
1759 | "READY=1\n" | |
1760 | "STATUS=Processing requests..."; | |
1761 | ssize_t l; | |
1762 | ||
1763 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1764 | if (l < 0) { | |
1765 | if (errno == EAGAIN) | |
1766 | return 0; | |
1767 | ||
1768 | return log_error_errno(errno, "Failed to send READY=1 notification message: %m"); | |
1769 | } | |
1770 | ||
1771 | s->sent_notify_ready = true; | |
1772 | log_debug("Sent READY=1 notification."); | |
1773 | ||
119e9655 LP |
1774 | } else if (s->send_watchdog) { |
1775 | ||
1776 | static const char p[] = | |
1777 | "WATCHDOG=1"; | |
1778 | ||
1779 | ssize_t l; | |
1780 | ||
1781 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1782 | if (l < 0) { | |
1783 | if (errno == EAGAIN) | |
1784 | return 0; | |
1785 | ||
1786 | return log_error_errno(errno, "Failed to send WATCHDOG=1 notification message: %m"); | |
1787 | } | |
1788 | ||
1789 | s->send_watchdog = false; | |
1790 | log_debug("Sent WATCHDOG=1 notification."); | |
1791 | ||
e22aa3d3 LP |
1792 | } else if (s->stdout_streams_notify_queue) |
1793 | /* Dispatch one stream notification event */ | |
1794 | stdout_stream_send_notify(s->stdout_streams_notify_queue); | |
1795 | ||
61233823 | 1796 | /* Leave us enabled if there's still more to do. */ |
119e9655 | 1797 | if (s->send_watchdog || s->stdout_streams_notify_queue) |
e22aa3d3 LP |
1798 | return 0; |
1799 | ||
1800 | /* There was nothing to do anymore, let's turn ourselves off. */ | |
1801 | r = sd_event_source_set_enabled(es, SD_EVENT_OFF); | |
1802 | if (r < 0) | |
1803 | return log_error_errno(r, "Failed to turn off notify event source: %m"); | |
1804 | ||
1805 | return 0; | |
1806 | } | |
1807 | ||
119e9655 LP |
1808 | static int dispatch_watchdog(sd_event_source *es, uint64_t usec, void *userdata) { |
1809 | Server *s = userdata; | |
1810 | int r; | |
1811 | ||
1812 | assert(s); | |
1813 | ||
1814 | s->send_watchdog = true; | |
1815 | ||
1816 | r = sd_event_source_set_enabled(s->notify_event_source, SD_EVENT_ON); | |
1817 | if (r < 0) | |
1818 | log_warning_errno(r, "Failed to turn on notify event source: %m"); | |
1819 | ||
1820 | r = sd_event_source_set_time(s->watchdog_event_source, usec + s->watchdog_usec / 2); | |
1821 | if (r < 0) | |
1822 | return log_error_errno(r, "Failed to restart watchdog event source: %m"); | |
1823 | ||
1824 | r = sd_event_source_set_enabled(s->watchdog_event_source, SD_EVENT_ON); | |
1825 | if (r < 0) | |
1826 | return log_error_errno(r, "Failed to enable watchdog event source: %m"); | |
1827 | ||
1828 | return 0; | |
1829 | } | |
1830 | ||
e22aa3d3 LP |
1831 | static int server_connect_notify(Server *s) { |
1832 | union sockaddr_union sa = { | |
1833 | .un.sun_family = AF_UNIX, | |
1834 | }; | |
1835 | const char *e; | |
1836 | int r; | |
1837 | ||
1838 | assert(s); | |
1839 | assert(s->notify_fd < 0); | |
1840 | assert(!s->notify_event_source); | |
1841 | ||
1842 | /* | |
1843 | So here's the problem: we'd like to send notification | |
1844 | messages to PID 1, but we cannot do that via sd_notify(), | |
1845 | since that's synchronous, and we might end up blocking on | |
1846 | it. Specifically: given that PID 1 might block on | |
1847 | dbus-daemon during IPC, and dbus-daemon is logging to us, | |
1848 | and might hence block on us, we might end up in a deadlock | |
ccddd104 | 1849 | if we block on sending PID 1 notification messages — by |
e22aa3d3 LP |
1850 | generating a full blocking circle. To avoid this, let's |
1851 | create a non-blocking socket, and connect it to the | |
1852 | notification socket, and then wait for POLLOUT before we | |
1853 | send anything. This should efficiently avoid any deadlocks, | |
1854 | as we'll never block on PID 1, hence PID 1 can safely block | |
1855 | on dbus-daemon which can safely block on us again. | |
1856 | ||
1857 | Don't think that this issue is real? It is, see: | |
1858 | https://github.com/systemd/systemd/issues/1505 | |
1859 | */ | |
1860 | ||
1861 | e = getenv("NOTIFY_SOCKET"); | |
1862 | if (!e) | |
1863 | return 0; | |
1864 | ||
1865 | if ((e[0] != '@' && e[0] != '/') || e[1] == 0) { | |
1866 | log_error("NOTIFY_SOCKET set to an invalid value: %s", e); | |
1867 | return -EINVAL; | |
1868 | } | |
1869 | ||
1870 | if (strlen(e) > sizeof(sa.un.sun_path)) { | |
1871 | log_error("NOTIFY_SOCKET path too long: %s", e); | |
1872 | return -EINVAL; | |
1873 | } | |
1874 | ||
1875 | s->notify_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); | |
1876 | if (s->notify_fd < 0) | |
1877 | return log_error_errno(errno, "Failed to create notify socket: %m"); | |
1878 | ||
1879 | (void) fd_inc_sndbuf(s->notify_fd, NOTIFY_SNDBUF_SIZE); | |
1880 | ||
1881 | strncpy(sa.un.sun_path, e, sizeof(sa.un.sun_path)); | |
1882 | if (sa.un.sun_path[0] == '@') | |
1883 | sa.un.sun_path[0] = 0; | |
1884 | ||
fc2fffe7 | 1885 | r = connect(s->notify_fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); |
e22aa3d3 LP |
1886 | if (r < 0) |
1887 | return log_error_errno(errno, "Failed to connect to notify socket: %m"); | |
1888 | ||
1889 | r = sd_event_add_io(s->event, &s->notify_event_source, s->notify_fd, EPOLLOUT, dispatch_notify_event, s); | |
1890 | if (r < 0) | |
1891 | return log_error_errno(r, "Failed to watch notification socket: %m"); | |
1892 | ||
119e9655 LP |
1893 | if (sd_watchdog_enabled(false, &s->watchdog_usec) > 0) { |
1894 | s->send_watchdog = true; | |
1895 | ||
4de2402b | 1896 | 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 |
1897 | if (r < 0) |
1898 | return log_error_errno(r, "Failed to add watchdog time event: %m"); | |
1899 | } | |
1900 | ||
e22aa3d3 LP |
1901 | /* This should fire pretty soon, which we'll use to send the |
1902 | * READY=1 event. */ | |
1903 | ||
1904 | return 0; | |
1905 | } | |
1906 | ||
d025f1e4 | 1907 | int server_init(Server *s) { |
13790add | 1908 | _cleanup_fdset_free_ FDSet *fds = NULL; |
d025f1e4 | 1909 | int n, r, fd; |
7d18d348 | 1910 | bool no_sockets; |
d025f1e4 ZJS |
1911 | |
1912 | assert(s); | |
1913 | ||
1914 | zero(*s); | |
e22aa3d3 | 1915 | s->syslog_fd = s->native_fd = s->stdout_fd = s->dev_kmsg_fd = s->audit_fd = s->hostname_fd = s->notify_fd = -1; |
d025f1e4 ZJS |
1916 | s->compress = true; |
1917 | s->seal = true; | |
b2392ff3 | 1918 | s->read_kmsg = true; |
d025f1e4 | 1919 | |
119e9655 LP |
1920 | s->watchdog_usec = USEC_INFINITY; |
1921 | ||
26687bf8 OS |
1922 | s->sync_interval_usec = DEFAULT_SYNC_INTERVAL_USEC; |
1923 | s->sync_scheduled = false; | |
1924 | ||
d025f1e4 ZJS |
1925 | s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL; |
1926 | s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST; | |
1927 | ||
40b71e89 | 1928 | s->forward_to_wall = true; |
d025f1e4 | 1929 | |
e150e820 MB |
1930 | s->max_file_usec = DEFAULT_MAX_FILE_USEC; |
1931 | ||
d025f1e4 ZJS |
1932 | s->max_level_store = LOG_DEBUG; |
1933 | s->max_level_syslog = LOG_DEBUG; | |
1934 | s->max_level_kmsg = LOG_NOTICE; | |
1935 | s->max_level_console = LOG_INFO; | |
40b71e89 | 1936 | s->max_level_wall = LOG_EMERG; |
d025f1e4 | 1937 | |
266a4700 FB |
1938 | journal_reset_metrics(&s->system_storage.metrics); |
1939 | journal_reset_metrics(&s->runtime_storage.metrics); | |
d025f1e4 ZJS |
1940 | |
1941 | server_parse_config_file(s); | |
1d84ad94 LP |
1942 | |
1943 | r = proc_cmdline_parse(parse_proc_cmdline_item, s, PROC_CMDLINE_STRIP_RD_PREFIX); | |
1944 | if (r < 0) | |
1945 | log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); | |
8580d1f7 | 1946 | |
d288f79f | 1947 | if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { |
b1389b0d ZJS |
1948 | log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0", |
1949 | s->rate_limit_interval, s->rate_limit_burst); | |
d288f79f ZJS |
1950 | s->rate_limit_interval = s->rate_limit_burst = 0; |
1951 | } | |
d025f1e4 | 1952 | |
8580d1f7 | 1953 | (void) mkdir_p("/run/systemd/journal", 0755); |
d025f1e4 | 1954 | |
43cf8388 | 1955 | s->user_journals = ordered_hashmap_new(NULL); |
d025f1e4 ZJS |
1956 | if (!s->user_journals) |
1957 | return log_oom(); | |
1958 | ||
1959 | s->mmap = mmap_cache_new(); | |
1960 | if (!s->mmap) | |
1961 | return log_oom(); | |
1962 | ||
b58c888f VC |
1963 | s->deferred_closes = set_new(NULL); |
1964 | if (!s->deferred_closes) | |
1965 | return log_oom(); | |
1966 | ||
f9a810be | 1967 | r = sd_event_default(&s->event); |
23bbb0de MS |
1968 | if (r < 0) |
1969 | return log_error_errno(r, "Failed to create event loop: %m"); | |
d025f1e4 ZJS |
1970 | |
1971 | n = sd_listen_fds(true); | |
23bbb0de MS |
1972 | if (n < 0) |
1973 | return log_error_errno(n, "Failed to read listening file descriptors from environment: %m"); | |
d025f1e4 ZJS |
1974 | |
1975 | for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { | |
1976 | ||
1977 | if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) { | |
1978 | ||
1979 | if (s->native_fd >= 0) { | |
1980 | log_error("Too many native sockets passed."); | |
1981 | return -EINVAL; | |
1982 | } | |
1983 | ||
1984 | s->native_fd = fd; | |
1985 | ||
1986 | } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) { | |
1987 | ||
1988 | if (s->stdout_fd >= 0) { | |
1989 | log_error("Too many stdout sockets passed."); | |
1990 | return -EINVAL; | |
1991 | } | |
1992 | ||
1993 | s->stdout_fd = fd; | |
1994 | ||
03ee5c38 LP |
1995 | } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0 || |
1996 | sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/dev-log", 0) > 0) { | |
d025f1e4 ZJS |
1997 | |
1998 | if (s->syslog_fd >= 0) { | |
1999 | log_error("Too many /dev/log sockets passed."); | |
2000 | return -EINVAL; | |
2001 | } | |
2002 | ||
2003 | s->syslog_fd = fd; | |
2004 | ||
875c2e22 LP |
2005 | } else if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1) > 0) { |
2006 | ||
2007 | if (s->audit_fd >= 0) { | |
2008 | log_error("Too many audit sockets passed."); | |
2009 | return -EINVAL; | |
2010 | } | |
2011 | ||
2012 | s->audit_fd = fd; | |
2013 | ||
4ec3cd73 | 2014 | } else { |
4ec3cd73 | 2015 | |
13790add LP |
2016 | if (!fds) { |
2017 | fds = fdset_new(); | |
2018 | if (!fds) | |
2019 | return log_oom(); | |
2020 | } | |
4ec3cd73 | 2021 | |
13790add LP |
2022 | r = fdset_put(fds, fd); |
2023 | if (r < 0) | |
2024 | return log_oom(); | |
4ec3cd73 | 2025 | } |
d025f1e4 ZJS |
2026 | } |
2027 | ||
15d91bff ZJS |
2028 | /* Try to restore streams, but don't bother if this fails */ |
2029 | (void) server_restore_streams(s, fds); | |
d025f1e4 | 2030 | |
13790add LP |
2031 | if (fdset_size(fds) > 0) { |
2032 | log_warning("%u unknown file descriptors passed, closing.", fdset_size(fds)); | |
2033 | fds = fdset_free(fds); | |
2034 | } | |
2035 | ||
7d18d348 ZJS |
2036 | no_sockets = s->native_fd < 0 && s->stdout_fd < 0 && s->syslog_fd < 0 && s->audit_fd < 0; |
2037 | ||
2038 | /* always open stdout, syslog, native, and kmsg sockets */ | |
37b7affe ZJS |
2039 | |
2040 | /* systemd-journald.socket: /run/systemd/journal/stdout */ | |
15d91bff ZJS |
2041 | r = server_open_stdout_socket(s); |
2042 | if (r < 0) | |
2043 | return r; | |
2044 | ||
37b7affe | 2045 | /* systemd-journald-dev-log.socket: /run/systemd/journal/dev-log */ |
13790add | 2046 | r = server_open_syslog_socket(s); |
d025f1e4 ZJS |
2047 | if (r < 0) |
2048 | return r; | |
2049 | ||
37b7affe | 2050 | /* systemd-journald.socket: /run/systemd/journal/socket */ |
13790add | 2051 | r = server_open_native_socket(s); |
d025f1e4 ZJS |
2052 | if (r < 0) |
2053 | return r; | |
2054 | ||
b2392ff3 | 2055 | /* /dev/kmsg */ |
d025f1e4 ZJS |
2056 | r = server_open_dev_kmsg(s); |
2057 | if (r < 0) | |
2058 | return r; | |
2059 | ||
7d18d348 ZJS |
2060 | /* Unless we got *some* sockets and not audit, open audit socket */ |
2061 | if (s->audit_fd >= 0 || no_sockets) { | |
2062 | r = server_open_audit(s); | |
2063 | if (r < 0) | |
2064 | return r; | |
2065 | } | |
875c2e22 | 2066 | |
d025f1e4 ZJS |
2067 | r = server_open_kernel_seqnum(s); |
2068 | if (r < 0) | |
2069 | return r; | |
2070 | ||
0c24bb23 LP |
2071 | r = server_open_hostname(s); |
2072 | if (r < 0) | |
2073 | return r; | |
2074 | ||
f9a810be | 2075 | r = setup_signals(s); |
d025f1e4 ZJS |
2076 | if (r < 0) |
2077 | return r; | |
2078 | ||
2079 | s->udev = udev_new(); | |
2080 | if (!s->udev) | |
2081 | return -ENOMEM; | |
2082 | ||
f9a810be | 2083 | s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst); |
d025f1e4 ZJS |
2084 | if (!s->rate_limit) |
2085 | return -ENOMEM; | |
2086 | ||
e9174f29 LP |
2087 | r = cg_get_root_path(&s->cgroup_root); |
2088 | if (r < 0) | |
2089 | return r; | |
2090 | ||
0c24bb23 LP |
2091 | server_cache_hostname(s); |
2092 | server_cache_boot_id(s); | |
2093 | server_cache_machine_id(s); | |
2094 | ||
266a4700 FB |
2095 | s->runtime_storage.name = "Runtime journal"; |
2096 | s->system_storage.name = "System journal"; | |
2097 | ||
605405c6 ZJS |
2098 | s->runtime_storage.path = strjoin("/run/log/journal/", SERVER_MACHINE_ID(s)); |
2099 | s->system_storage.path = strjoin("/var/log/journal/", SERVER_MACHINE_ID(s)); | |
266a4700 FB |
2100 | if (!s->runtime_storage.path || !s->system_storage.path) |
2101 | return -ENOMEM; | |
2102 | ||
e22aa3d3 LP |
2103 | (void) server_connect_notify(s); |
2104 | ||
804ae586 | 2105 | return system_journal_open(s, false); |
d025f1e4 ZJS |
2106 | } |
2107 | ||
2108 | void server_maybe_append_tags(Server *s) { | |
2109 | #ifdef HAVE_GCRYPT | |
2110 | JournalFile *f; | |
2111 | Iterator i; | |
2112 | usec_t n; | |
2113 | ||
2114 | n = now(CLOCK_REALTIME); | |
2115 | ||
2116 | if (s->system_journal) | |
2117 | journal_file_maybe_append_tag(s->system_journal, n); | |
2118 | ||
43cf8388 | 2119 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) |
d025f1e4 ZJS |
2120 | journal_file_maybe_append_tag(f, n); |
2121 | #endif | |
2122 | } | |
2123 | ||
2124 | void server_done(Server *s) { | |
2125 | JournalFile *f; | |
2126 | assert(s); | |
2127 | ||
b58c888f VC |
2128 | if (s->deferred_closes) { |
2129 | journal_file_close_set(s->deferred_closes); | |
2130 | set_free(s->deferred_closes); | |
2131 | } | |
2132 | ||
d025f1e4 ZJS |
2133 | while (s->stdout_streams) |
2134 | stdout_stream_free(s->stdout_streams); | |
2135 | ||
2136 | if (s->system_journal) | |
69a3a6fd | 2137 | (void) journal_file_close(s->system_journal); |
d025f1e4 ZJS |
2138 | |
2139 | if (s->runtime_journal) | |
69a3a6fd | 2140 | (void) journal_file_close(s->runtime_journal); |
d025f1e4 | 2141 | |
43cf8388 | 2142 | while ((f = ordered_hashmap_steal_first(s->user_journals))) |
69a3a6fd | 2143 | (void) journal_file_close(f); |
d025f1e4 | 2144 | |
43cf8388 | 2145 | ordered_hashmap_free(s->user_journals); |
d025f1e4 | 2146 | |
f9a810be LP |
2147 | sd_event_source_unref(s->syslog_event_source); |
2148 | sd_event_source_unref(s->native_event_source); | |
2149 | sd_event_source_unref(s->stdout_event_source); | |
2150 | sd_event_source_unref(s->dev_kmsg_event_source); | |
875c2e22 | 2151 | sd_event_source_unref(s->audit_event_source); |
f9a810be LP |
2152 | sd_event_source_unref(s->sync_event_source); |
2153 | sd_event_source_unref(s->sigusr1_event_source); | |
2154 | sd_event_source_unref(s->sigusr2_event_source); | |
2155 | sd_event_source_unref(s->sigterm_event_source); | |
2156 | sd_event_source_unref(s->sigint_event_source); | |
94b65516 | 2157 | sd_event_source_unref(s->sigrtmin1_event_source); |
0c24bb23 | 2158 | sd_event_source_unref(s->hostname_event_source); |
e22aa3d3 | 2159 | sd_event_source_unref(s->notify_event_source); |
119e9655 | 2160 | sd_event_source_unref(s->watchdog_event_source); |
f9a810be | 2161 | sd_event_unref(s->event); |
d025f1e4 | 2162 | |
03e334a1 LP |
2163 | safe_close(s->syslog_fd); |
2164 | safe_close(s->native_fd); | |
2165 | safe_close(s->stdout_fd); | |
2166 | safe_close(s->dev_kmsg_fd); | |
875c2e22 | 2167 | safe_close(s->audit_fd); |
03e334a1 | 2168 | safe_close(s->hostname_fd); |
e22aa3d3 | 2169 | safe_close(s->notify_fd); |
0c24bb23 | 2170 | |
d025f1e4 ZJS |
2171 | if (s->rate_limit) |
2172 | journal_rate_limit_free(s->rate_limit); | |
2173 | ||
2174 | if (s->kernel_seqnum) | |
2175 | munmap(s->kernel_seqnum, sizeof(uint64_t)); | |
2176 | ||
2177 | free(s->buffer); | |
2178 | free(s->tty_path); | |
e9174f29 | 2179 | free(s->cgroup_root); |
99d0966e | 2180 | free(s->hostname_field); |
c6e9e16f ZJS |
2181 | free(s->runtime_storage.path); |
2182 | free(s->system_storage.path); | |
d025f1e4 ZJS |
2183 | |
2184 | if (s->mmap) | |
2185 | mmap_cache_unref(s->mmap); | |
2186 | ||
3e044c49 | 2187 | udev_unref(s->udev); |
d025f1e4 | 2188 | } |
8580d1f7 LP |
2189 | |
2190 | static const char* const storage_table[_STORAGE_MAX] = { | |
2191 | [STORAGE_AUTO] = "auto", | |
2192 | [STORAGE_VOLATILE] = "volatile", | |
2193 | [STORAGE_PERSISTENT] = "persistent", | |
2194 | [STORAGE_NONE] = "none" | |
2195 | }; | |
2196 | ||
2197 | DEFINE_STRING_TABLE_LOOKUP(storage, Storage); | |
2198 | DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); | |
2199 | ||
2200 | static const char* const split_mode_table[_SPLIT_MAX] = { | |
2201 | [SPLIT_LOGIN] = "login", | |
2202 | [SPLIT_UID] = "uid", | |
2203 | [SPLIT_NONE] = "none", | |
2204 | }; | |
2205 | ||
2206 | DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode); | |
2207 | DEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting"); |