]>
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" |
958b66ea | 43 | #include "formats-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" |
d025f1e4 | 74 | |
d025f1e4 ZJS |
75 | #define USER_JOURNALS_MAX 1024 |
76 | ||
26687bf8 | 77 | #define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) |
7f1ad696 LP |
78 | #define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC) |
79 | #define DEFAULT_RATE_LIMIT_BURST 1000 | |
e150e820 | 80 | #define DEFAULT_MAX_FILE_USEC USEC_PER_MONTH |
d025f1e4 | 81 | |
8580d1f7 | 82 | #define RECHECK_SPACE_USEC (30*USEC_PER_SEC) |
d025f1e4 | 83 | |
e22aa3d3 LP |
84 | #define NOTIFY_SNDBUF_SIZE (8*1024*1024) |
85 | ||
7a24f3bf VC |
86 | /* The period to insert between posting changes for coalescing */ |
87 | #define POST_CHANGE_TIMER_INTERVAL_USEC (250*USEC_PER_MSEC) | |
88 | ||
8580d1f7 LP |
89 | static int determine_space_for( |
90 | Server *s, | |
91 | JournalMetrics *metrics, | |
92 | const char *path, | |
93 | const char *name, | |
94 | bool verbose, | |
95 | bool patch_min_use, | |
96 | uint64_t *available, | |
97 | uint64_t *limit) { | |
98 | ||
99 | uint64_t sum = 0, ss_avail, avail; | |
7fd1b19b | 100 | _cleanup_closedir_ DIR *d = NULL; |
8580d1f7 LP |
101 | struct dirent *de; |
102 | struct statvfs ss; | |
103 | const char *p; | |
d025f1e4 | 104 | usec_t ts; |
d025f1e4 | 105 | |
8580d1f7 LP |
106 | assert(s); |
107 | assert(metrics); | |
108 | assert(path); | |
109 | assert(name); | |
d025f1e4 | 110 | |
8580d1f7 | 111 | ts = now(CLOCK_MONOTONIC); |
d025f1e4 | 112 | |
8580d1f7 | 113 | if (!verbose && s->cached_space_timestamp + RECHECK_SPACE_USEC > ts) { |
d025f1e4 | 114 | |
8580d1f7 LP |
115 | if (available) |
116 | *available = s->cached_space_available; | |
117 | if (limit) | |
118 | *limit = s->cached_space_limit; | |
d025f1e4 | 119 | |
d025f1e4 | 120 | return 0; |
8580d1f7 | 121 | } |
d025f1e4 | 122 | |
8580d1f7 | 123 | p = strjoina(path, SERVER_MACHINE_ID(s)); |
d025f1e4 | 124 | d = opendir(p); |
d025f1e4 | 125 | if (!d) |
8580d1f7 | 126 | return log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_ERR, errno, "Failed to open %s: %m", p); |
d025f1e4 ZJS |
127 | |
128 | if (fstatvfs(dirfd(d), &ss) < 0) | |
8580d1f7 | 129 | return log_error_errno(errno, "Failed to fstatvfs(%s): %m", p); |
d025f1e4 | 130 | |
8580d1f7 | 131 | FOREACH_DIRENT_ALL(de, d, break) { |
d025f1e4 | 132 | struct stat st; |
d025f1e4 ZJS |
133 | |
134 | if (!endswith(de->d_name, ".journal") && | |
135 | !endswith(de->d_name, ".journal~")) | |
136 | continue; | |
137 | ||
8580d1f7 LP |
138 | if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) { |
139 | log_debug_errno(errno, "Failed to stat %s/%s, ignoring: %m", p, de->d_name); | |
d025f1e4 | 140 | continue; |
8580d1f7 | 141 | } |
d025f1e4 ZJS |
142 | |
143 | if (!S_ISREG(st.st_mode)) | |
144 | continue; | |
145 | ||
146 | sum += (uint64_t) st.st_blocks * 512UL; | |
147 | } | |
148 | ||
8a03c9ef | 149 | /* If requested, then let's bump the min_use limit to the |
8580d1f7 LP |
150 | * current usage on disk. We do this when starting up and |
151 | * first opening the journal files. This way sudden spikes in | |
152 | * disk usage will not cause journald to vacuum files without | |
153 | * bounds. Note that this means that only a restart of | |
154 | * journald will make it reset this value. */ | |
d025f1e4 | 155 | |
8580d1f7 LP |
156 | if (patch_min_use) |
157 | metrics->min_use = MAX(metrics->min_use, sum); | |
348ced90 | 158 | |
8580d1f7 LP |
159 | ss_avail = ss.f_bsize * ss.f_bavail; |
160 | avail = LESS_BY(ss_avail, metrics->keep_free); | |
348ced90 | 161 | |
8580d1f7 LP |
162 | s->cached_space_limit = MIN(MAX(sum + avail, metrics->min_use), metrics->max_use); |
163 | s->cached_space_available = LESS_BY(s->cached_space_limit, sum); | |
164 | s->cached_space_timestamp = ts; | |
d025f1e4 | 165 | |
670b110c ZJS |
166 | if (verbose) { |
167 | char fb1[FORMAT_BYTES_MAX], fb2[FORMAT_BYTES_MAX], fb3[FORMAT_BYTES_MAX], | |
8580d1f7 | 168 | fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX], fb6[FORMAT_BYTES_MAX]; |
282c5c4e ZJS |
169 | format_bytes(fb1, sizeof(fb1), sum); |
170 | format_bytes(fb2, sizeof(fb2), metrics->max_use); | |
171 | format_bytes(fb3, sizeof(fb3), metrics->keep_free); | |
172 | format_bytes(fb4, sizeof(fb4), ss_avail); | |
173 | format_bytes(fb5, sizeof(fb5), s->cached_space_limit); | |
174 | format_bytes(fb6, sizeof(fb6), s->cached_space_available); | |
670b110c ZJS |
175 | |
176 | server_driver_message(s, SD_MESSAGE_JOURNAL_USAGE, | |
282c5c4e ZJS |
177 | LOG_MESSAGE("%s (%s) is %s, max %s, %s free.", |
178 | name, path, fb1, fb5, fb6), | |
179 | "JOURNAL_NAME=%s", name, | |
180 | "JOURNAL_PATH=%s", path, | |
181 | "CURRENT_USE=%"PRIu64, sum, | |
182 | "CURRENT_USE_PRETTY=%s", fb1, | |
183 | "MAX_USE=%"PRIu64, metrics->max_use, | |
184 | "MAX_USE_PRETTY=%s", fb2, | |
185 | "DISK_KEEP_FREE=%"PRIu64, metrics->keep_free, | |
186 | "DISK_KEEP_FREE_PRETTY=%s", fb3, | |
187 | "DISK_AVAILABLE=%"PRIu64, ss_avail, | |
188 | "DISK_AVAILABLE_PRETTY=%s", fb4, | |
189 | "LIMIT=%"PRIu64, s->cached_space_limit, | |
190 | "LIMIT_PRETTY=%s", fb5, | |
191 | "AVAILABLE=%"PRIu64, s->cached_space_available, | |
192 | "AVAILABLE_PRETTY=%s", fb6, | |
8a03c9ef | 193 | NULL); |
8580d1f7 LP |
194 | } |
195 | ||
196 | if (available) | |
197 | *available = s->cached_space_available; | |
198 | if (limit) | |
199 | *limit = s->cached_space_limit; | |
200 | ||
201 | return 1; | |
202 | } | |
203 | ||
204 | static int determine_space(Server *s, bool verbose, bool patch_min_use, uint64_t *available, uint64_t *limit) { | |
205 | JournalMetrics *metrics; | |
206 | const char *path, *name; | |
207 | ||
208 | assert(s); | |
209 | ||
210 | if (s->system_journal) { | |
211 | path = "/var/log/journal/"; | |
212 | metrics = &s->system_metrics; | |
213 | name = "System journal"; | |
214 | } else { | |
215 | path = "/run/log/journal/"; | |
216 | metrics = &s->runtime_metrics; | |
217 | name = "Runtime journal"; | |
670b110c ZJS |
218 | } |
219 | ||
8580d1f7 | 220 | return determine_space_for(s, metrics, path, name, verbose, patch_min_use, available, limit); |
d025f1e4 ZJS |
221 | } |
222 | ||
5c3bde3f | 223 | static void server_add_acls(JournalFile *f, uid_t uid) { |
d025f1e4 | 224 | #ifdef HAVE_ACL |
5c3bde3f | 225 | int r; |
d025f1e4 | 226 | #endif |
d025f1e4 ZJS |
227 | assert(f); |
228 | ||
d025f1e4 | 229 | #ifdef HAVE_ACL |
34c10968 | 230 | if (uid <= SYSTEM_UID_MAX) |
d025f1e4 ZJS |
231 | return; |
232 | ||
5c3bde3f ZJS |
233 | r = add_acls_for_user(f->fd, uid); |
234 | if (r < 0) | |
235 | log_warning_errno(r, "Failed to set ACL on %s, ignoring: %m", f->path); | |
d025f1e4 ZJS |
236 | #endif |
237 | } | |
238 | ||
7a24f3bf VC |
239 | static int open_journal( |
240 | Server *s, | |
241 | bool reliably, | |
242 | const char *fname, | |
243 | int flags, | |
244 | bool seal, | |
245 | JournalMetrics *metrics, | |
7a24f3bf VC |
246 | JournalFile **ret) { |
247 | int r; | |
e167d7fd | 248 | JournalFile *f; |
7a24f3bf VC |
249 | |
250 | assert(s); | |
251 | assert(fname); | |
252 | assert(ret); | |
253 | ||
254 | if (reliably) | |
b58c888f | 255 | r = journal_file_open_reliably(fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f); |
7a24f3bf | 256 | else |
5d1ce257 | 257 | r = journal_file_open(-1, fname, flags, 0640, s->compress, seal, metrics, s->mmap, s->deferred_closes, NULL, &f); |
7a24f3bf VC |
258 | if (r < 0) |
259 | return r; | |
260 | ||
e167d7fd | 261 | r = journal_file_enable_post_change_timer(f, s->event, POST_CHANGE_TIMER_INTERVAL_USEC); |
7a24f3bf | 262 | if (r < 0) { |
69a3a6fd | 263 | (void) journal_file_close(f); |
7a24f3bf VC |
264 | return r; |
265 | } | |
266 | ||
e167d7fd | 267 | *ret = f; |
7a24f3bf VC |
268 | return r; |
269 | } | |
270 | ||
6431c7e2 VC |
271 | static bool flushed_flag_is_set(void) { |
272 | return (access("/run/systemd/journal/flushed", F_OK) >= 0); | |
273 | } | |
274 | ||
105bdb46 | 275 | static int system_journal_open(Server *s, bool flush_requested) { |
929eeb54 | 276 | bool flushed = false; |
105bdb46 VC |
277 | const char *fn; |
278 | int r = 0; | |
279 | ||
280 | if (!s->system_journal && | |
281 | (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) && | |
6431c7e2 | 282 | (flush_requested || (flushed = flushed_flag_is_set()))) { |
105bdb46 VC |
283 | |
284 | /* If in auto mode: first try to create the machine | |
285 | * path, but not the prefix. | |
286 | * | |
287 | * If in persistent mode: create /var/log/journal and | |
288 | * the machine path */ | |
289 | ||
290 | if (s->storage == STORAGE_PERSISTENT) | |
291 | (void) mkdir_p("/var/log/journal/", 0755); | |
292 | ||
293 | fn = strjoina("/var/log/journal/", SERVER_MACHINE_ID(s)); | |
294 | (void) mkdir(fn, 0755); | |
295 | ||
296 | fn = strjoina(fn, "/system.journal"); | |
297 | r = open_journal(s, true, fn, O_RDWR|O_CREAT, s->seal, &s->system_metrics, &s->system_journal); | |
298 | if (r >= 0) { | |
299 | server_add_acls(s->system_journal, 0); | |
300 | (void) determine_space_for(s, &s->system_metrics, "/var/log/journal/", "System journal", true, true, NULL, NULL); | |
301 | } else if (r < 0) { | |
302 | if (r != -ENOENT && r != -EROFS) | |
303 | log_warning_errno(r, "Failed to open system journal: %m"); | |
304 | ||
305 | r = 0; | |
306 | } | |
929eeb54 VC |
307 | |
308 | /* If the runtime journal is open, and we're post-flush, we're | |
309 | * recovering from a failed system journal rotate (ENOSPC) | |
310 | * for which the runtime journal was reopened. | |
311 | * | |
312 | * Perform an implicit flush to var, leaving the runtime | |
313 | * journal closed, now that the system journal is back. | |
314 | */ | |
315 | if (s->runtime_journal && flushed) | |
316 | (void) server_flush_to_var(s); | |
105bdb46 VC |
317 | } |
318 | ||
319 | if (!s->runtime_journal && | |
320 | (s->storage != STORAGE_NONE)) { | |
321 | ||
322 | fn = strjoina("/run/log/journal/", SERVER_MACHINE_ID(s), "/system.journal"); | |
323 | ||
324 | if (s->system_journal) { | |
325 | ||
326 | /* Try to open the runtime journal, but only | |
327 | * if it already exists, so that we can flush | |
328 | * it into the system journal */ | |
329 | ||
330 | r = open_journal(s, false, fn, O_RDWR, false, &s->runtime_metrics, &s->runtime_journal); | |
331 | if (r < 0) { | |
332 | if (r != -ENOENT) | |
333 | log_warning_errno(r, "Failed to open runtime journal: %m"); | |
334 | ||
335 | r = 0; | |
336 | } | |
337 | ||
338 | } else { | |
339 | ||
340 | /* OK, we really need the runtime journal, so create | |
341 | * it if necessary. */ | |
342 | ||
343 | (void) mkdir("/run/log", 0755); | |
344 | (void) mkdir("/run/log/journal", 0755); | |
345 | (void) mkdir_parents(fn, 0750); | |
346 | ||
347 | r = open_journal(s, true, fn, O_RDWR|O_CREAT, false, &s->runtime_metrics, &s->runtime_journal); | |
348 | if (r < 0) | |
349 | return log_error_errno(r, "Failed to open runtime journal: %m"); | |
350 | } | |
351 | ||
352 | if (s->runtime_journal) { | |
353 | server_add_acls(s->runtime_journal, 0); | |
354 | (void) determine_space_for(s, &s->runtime_metrics, "/run/log/journal/", "Runtime journal", true, true, NULL, NULL); | |
355 | } | |
356 | } | |
357 | ||
358 | return r; | |
359 | } | |
360 | ||
d025f1e4 | 361 | static JournalFile* find_journal(Server *s, uid_t uid) { |
ed375beb | 362 | _cleanup_free_ char *p = NULL; |
d025f1e4 ZJS |
363 | int r; |
364 | JournalFile *f; | |
365 | sd_id128_t machine; | |
366 | ||
367 | assert(s); | |
368 | ||
105bdb46 VC |
369 | /* A rotate that fails to create the new journal (ENOSPC) leaves the |
370 | * rotated journal as NULL. Unless we revisit opening, even after | |
371 | * space is made available we'll continue to return NULL indefinitely. | |
372 | * | |
373 | * system_journal_open() is a noop if the journals are already open, so | |
374 | * we can just call it here to recover from failed rotates (or anything | |
375 | * else that's left the journals as NULL). | |
376 | * | |
377 | * Fixes https://github.com/systemd/systemd/issues/3968 */ | |
378 | (void) system_journal_open(s, false); | |
379 | ||
d025f1e4 ZJS |
380 | /* We split up user logs only on /var, not on /run. If the |
381 | * runtime file is open, we write to it exclusively, in order | |
382 | * to guarantee proper order as soon as we flush /run to | |
383 | * /var and close the runtime file. */ | |
384 | ||
385 | if (s->runtime_journal) | |
386 | return s->runtime_journal; | |
387 | ||
61755fda | 388 | if (uid <= SYSTEM_UID_MAX || uid_is_dynamic(uid)) |
d025f1e4 ZJS |
389 | return s->system_journal; |
390 | ||
391 | r = sd_id128_get_machine(&machine); | |
392 | if (r < 0) | |
393 | return s->system_journal; | |
394 | ||
4a0b58c4 | 395 | f = ordered_hashmap_get(s->user_journals, UID_TO_PTR(uid)); |
d025f1e4 ZJS |
396 | if (f) |
397 | return f; | |
398 | ||
de0671ee ZJS |
399 | if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-"UID_FMT".journal", |
400 | SD_ID128_FORMAT_VAL(machine), uid) < 0) | |
d025f1e4 ZJS |
401 | return s->system_journal; |
402 | ||
43cf8388 | 403 | while (ordered_hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { |
d025f1e4 | 404 | /* Too many open? Then let's close one */ |
43cf8388 | 405 | f = ordered_hashmap_steal_first(s->user_journals); |
d025f1e4 | 406 | assert(f); |
69a3a6fd | 407 | (void) journal_file_close(f); |
d025f1e4 ZJS |
408 | } |
409 | ||
089ed40b | 410 | r = open_journal(s, true, p, O_RDWR|O_CREAT, s->seal, &s->system_metrics, &f); |
d025f1e4 ZJS |
411 | if (r < 0) |
412 | return s->system_journal; | |
413 | ||
5c3bde3f | 414 | server_add_acls(f, uid); |
d025f1e4 | 415 | |
4a0b58c4 | 416 | r = ordered_hashmap_put(s->user_journals, UID_TO_PTR(uid), f); |
d025f1e4 | 417 | if (r < 0) { |
69a3a6fd | 418 | (void) journal_file_close(f); |
d025f1e4 ZJS |
419 | return s->system_journal; |
420 | } | |
421 | ||
422 | return f; | |
423 | } | |
424 | ||
ea69bd41 LP |
425 | static int do_rotate( |
426 | Server *s, | |
427 | JournalFile **f, | |
428 | const char* name, | |
429 | bool seal, | |
430 | uint32_t uid) { | |
431 | ||
fc55baee ZJS |
432 | int r; |
433 | assert(s); | |
434 | ||
435 | if (!*f) | |
436 | return -EINVAL; | |
437 | ||
b58c888f | 438 | r = journal_file_rotate(f, s->compress, seal, s->deferred_closes); |
fc55baee ZJS |
439 | if (r < 0) |
440 | if (*f) | |
ea69bd41 | 441 | log_error_errno(r, "Failed to rotate %s: %m", (*f)->path); |
fc55baee | 442 | else |
ea69bd41 | 443 | log_error_errno(r, "Failed to create new %s journal: %m", name); |
fc55baee | 444 | else |
5c3bde3f | 445 | server_add_acls(*f, uid); |
2678031a | 446 | |
fc55baee ZJS |
447 | return r; |
448 | } | |
449 | ||
d025f1e4 ZJS |
450 | void server_rotate(Server *s) { |
451 | JournalFile *f; | |
452 | void *k; | |
453 | Iterator i; | |
454 | int r; | |
455 | ||
456 | log_debug("Rotating..."); | |
457 | ||
8580d1f7 LP |
458 | (void) do_rotate(s, &s->runtime_journal, "runtime", false, 0); |
459 | (void) do_rotate(s, &s->system_journal, "system", s->seal, 0); | |
d025f1e4 | 460 | |
43cf8388 | 461 | ORDERED_HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { |
4a0b58c4 | 462 | r = do_rotate(s, &f, "user", s->seal, PTR_TO_UID(k)); |
fc55baee | 463 | if (r >= 0) |
43cf8388 | 464 | ordered_hashmap_replace(s->user_journals, k, f); |
fc55baee ZJS |
465 | else if (!f) |
466 | /* Old file has been closed and deallocated */ | |
43cf8388 | 467 | ordered_hashmap_remove(s->user_journals, k); |
d025f1e4 | 468 | } |
b58c888f VC |
469 | |
470 | /* Perform any deferred closes which aren't still offlining. */ | |
471 | SET_FOREACH(f, s->deferred_closes, i) | |
472 | if (!journal_file_is_offlining(f)) { | |
473 | (void) set_remove(s->deferred_closes, f); | |
474 | (void) journal_file_close(f); | |
475 | } | |
d025f1e4 ZJS |
476 | } |
477 | ||
26687bf8 OS |
478 | void server_sync(Server *s) { |
479 | JournalFile *f; | |
26687bf8 OS |
480 | Iterator i; |
481 | int r; | |
482 | ||
26687bf8 | 483 | if (s->system_journal) { |
ac2e41f5 | 484 | r = journal_file_set_offline(s->system_journal, false); |
26687bf8 | 485 | if (r < 0) |
65089b82 | 486 | log_warning_errno(r, "Failed to sync system journal, ignoring: %m"); |
26687bf8 OS |
487 | } |
488 | ||
65c1d46b | 489 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) { |
ac2e41f5 | 490 | r = journal_file_set_offline(f, false); |
26687bf8 | 491 | if (r < 0) |
65089b82 | 492 | log_warning_errno(r, "Failed to sync user journal, ignoring: %m"); |
26687bf8 OS |
493 | } |
494 | ||
f9a810be LP |
495 | if (s->sync_event_source) { |
496 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF); | |
497 | if (r < 0) | |
da927ba9 | 498 | log_error_errno(r, "Failed to disable sync timer source: %m"); |
f9a810be | 499 | } |
26687bf8 OS |
500 | |
501 | s->sync_scheduled = false; | |
502 | } | |
503 | ||
ea69bd41 LP |
504 | static void do_vacuum( |
505 | Server *s, | |
ea69bd41 | 506 | JournalFile *f, |
8580d1f7 LP |
507 | JournalMetrics *metrics, |
508 | const char *path, | |
509 | const char *name, | |
510 | bool verbose, | |
511 | bool patch_min_use) { | |
ea69bd41 LP |
512 | |
513 | const char *p; | |
8580d1f7 | 514 | uint64_t limit; |
63c8666b ZJS |
515 | int r; |
516 | ||
8580d1f7 LP |
517 | assert(s); |
518 | assert(metrics); | |
519 | assert(path); | |
520 | assert(name); | |
521 | ||
63c8666b ZJS |
522 | if (!f) |
523 | return; | |
524 | ||
8580d1f7 LP |
525 | p = strjoina(path, SERVER_MACHINE_ID(s)); |
526 | ||
527 | limit = metrics->max_use; | |
528 | (void) determine_space_for(s, metrics, path, name, verbose, patch_min_use, NULL, &limit); | |
529 | ||
530 | r = journal_directory_vacuum(p, limit, metrics->n_max_files, s->max_retention_usec, &s->oldest_file_usec, verbose); | |
63c8666b | 531 | if (r < 0 && r != -ENOENT) |
8580d1f7 | 532 | log_warning_errno(r, "Failed to vacuum %s, ignoring: %m", p); |
63c8666b ZJS |
533 | } |
534 | ||
8580d1f7 LP |
535 | int server_vacuum(Server *s, bool verbose, bool patch_min_use) { |
536 | assert(s); | |
d025f1e4 ZJS |
537 | |
538 | log_debug("Vacuuming..."); | |
539 | ||
540 | s->oldest_file_usec = 0; | |
541 | ||
8580d1f7 LP |
542 | do_vacuum(s, s->system_journal, &s->system_metrics, "/var/log/journal/", "System journal", verbose, patch_min_use); |
543 | do_vacuum(s, s->runtime_journal, &s->runtime_metrics, "/run/log/journal/", "Runtime journal", verbose, patch_min_use); | |
d025f1e4 | 544 | |
8580d1f7 LP |
545 | s->cached_space_limit = 0; |
546 | s->cached_space_available = 0; | |
547 | s->cached_space_timestamp = 0; | |
d025f1e4 | 548 | |
8580d1f7 | 549 | return 0; |
d025f1e4 ZJS |
550 | } |
551 | ||
0c24bb23 LP |
552 | static void server_cache_machine_id(Server *s) { |
553 | sd_id128_t id; | |
554 | int r; | |
555 | ||
556 | assert(s); | |
557 | ||
558 | r = sd_id128_get_machine(&id); | |
559 | if (r < 0) | |
560 | return; | |
561 | ||
562 | sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID=")); | |
563 | } | |
564 | ||
565 | static void server_cache_boot_id(Server *s) { | |
566 | sd_id128_t id; | |
567 | int r; | |
568 | ||
569 | assert(s); | |
570 | ||
571 | r = sd_id128_get_boot(&id); | |
572 | if (r < 0) | |
573 | return; | |
574 | ||
575 | sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID=")); | |
576 | } | |
577 | ||
578 | static void server_cache_hostname(Server *s) { | |
579 | _cleanup_free_ char *t = NULL; | |
580 | char *x; | |
581 | ||
582 | assert(s); | |
583 | ||
584 | t = gethostname_malloc(); | |
585 | if (!t) | |
586 | return; | |
587 | ||
588 | x = strappend("_HOSTNAME=", t); | |
589 | if (!x) | |
590 | return; | |
591 | ||
592 | free(s->hostname_field); | |
593 | s->hostname_field = x; | |
594 | } | |
595 | ||
8531ae70 | 596 | static bool shall_try_append_again(JournalFile *f, int r) { |
6e1045e5 ZJS |
597 | switch(r) { |
598 | case -E2BIG: /* Hit configured limit */ | |
599 | case -EFBIG: /* Hit fs limit */ | |
600 | case -EDQUOT: /* Quota limit hit */ | |
601 | case -ENOSPC: /* Disk full */ | |
d025f1e4 | 602 | log_debug("%s: Allocation limit reached, rotating.", f->path); |
6e1045e5 ZJS |
603 | return true; |
604 | case -EIO: /* I/O error of some kind (mmap) */ | |
605 | log_warning("%s: IO error, rotating.", f->path); | |
606 | return true; | |
607 | case -EHOSTDOWN: /* Other machine */ | |
d025f1e4 | 608 | log_info("%s: Journal file from other machine, rotating.", f->path); |
6e1045e5 ZJS |
609 | return true; |
610 | case -EBUSY: /* Unclean shutdown */ | |
d025f1e4 | 611 | log_info("%s: Unclean shutdown, rotating.", f->path); |
6e1045e5 ZJS |
612 | return true; |
613 | case -EPROTONOSUPPORT: /* Unsupported feature */ | |
d025f1e4 | 614 | log_info("%s: Unsupported feature, rotating.", f->path); |
6e1045e5 ZJS |
615 | return true; |
616 | case -EBADMSG: /* Corrupted */ | |
617 | case -ENODATA: /* Truncated */ | |
618 | case -ESHUTDOWN: /* Already archived */ | |
d025f1e4 | 619 | log_warning("%s: Journal file corrupted, rotating.", f->path); |
6e1045e5 ZJS |
620 | return true; |
621 | case -EIDRM: /* Journal file has been deleted */ | |
2678031a | 622 | log_warning("%s: Journal file has been deleted, rotating.", f->path); |
6e1045e5 ZJS |
623 | return true; |
624 | default: | |
d025f1e4 | 625 | return false; |
6e1045e5 | 626 | } |
d025f1e4 ZJS |
627 | } |
628 | ||
d07f7b9e | 629 | static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n, int priority) { |
0f972d66 | 630 | struct dual_timestamp ts; |
d025f1e4 | 631 | bool vacuumed = false; |
0f972d66 | 632 | JournalFile *f; |
d025f1e4 ZJS |
633 | int r; |
634 | ||
635 | assert(s); | |
636 | assert(iovec); | |
637 | assert(n > 0); | |
638 | ||
0f972d66 LP |
639 | /* Get the closest, linearized time we have for this log event from the event loop. (Note that we do not use |
640 | * the source time, and not even the time the event was originally seen, but instead simply the time we started | |
641 | * processing it, as we want strictly linear ordering in what we write out.) */ | |
642 | assert_se(sd_event_now(s->event, CLOCK_REALTIME, &ts.realtime) >= 0); | |
643 | assert_se(sd_event_now(s->event, CLOCK_MONOTONIC, &ts.monotonic) >= 0); | |
644 | ||
d025f1e4 ZJS |
645 | f = find_journal(s, uid); |
646 | if (!f) | |
647 | return; | |
648 | ||
649 | if (journal_file_rotate_suggested(f, s->max_file_usec)) { | |
650 | log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path); | |
651 | server_rotate(s); | |
8580d1f7 | 652 | server_vacuum(s, false, false); |
d025f1e4 ZJS |
653 | vacuumed = true; |
654 | ||
655 | f = find_journal(s, uid); | |
656 | if (!f) | |
657 | return; | |
658 | } | |
659 | ||
0f972d66 | 660 | r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL); |
26687bf8 | 661 | if (r >= 0) { |
d07f7b9e | 662 | server_schedule_sync(s, priority); |
d025f1e4 | 663 | return; |
26687bf8 | 664 | } |
d025f1e4 ZJS |
665 | |
666 | if (vacuumed || !shall_try_append_again(f, r)) { | |
8266e1c0 | 667 | log_error_errno(r, "Failed to write entry (%d items, %zu bytes), ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); |
d025f1e4 ZJS |
668 | return; |
669 | } | |
670 | ||
671 | server_rotate(s); | |
8580d1f7 | 672 | server_vacuum(s, false, false); |
d025f1e4 ZJS |
673 | |
674 | f = find_journal(s, uid); | |
675 | if (!f) | |
676 | return; | |
677 | ||
678 | log_debug("Retrying write."); | |
0f972d66 | 679 | r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL); |
8266e1c0 LP |
680 | if (r < 0) |
681 | log_error_errno(r, "Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n)); | |
682 | else | |
d07f7b9e | 683 | server_schedule_sync(s, priority); |
d025f1e4 ZJS |
684 | } |
685 | ||
4b58153d LP |
686 | static int get_invocation_id(const char *cgroup_root, const char *slice, const char *unit, char **ret) { |
687 | _cleanup_free_ char *escaped = NULL, *slice_path = NULL, *p = NULL; | |
688 | char *copy, ids[SD_ID128_STRING_MAX]; | |
689 | int r; | |
690 | ||
691 | /* Read the invocation ID of a unit off a unit. It's stored in the "trusted.invocation_id" extended attribute | |
692 | * on the cgroup path. */ | |
693 | ||
694 | r = cg_slice_to_path(slice, &slice_path); | |
695 | if (r < 0) | |
696 | return r; | |
697 | ||
698 | escaped = cg_escape(unit); | |
699 | if (!escaped) | |
700 | return -ENOMEM; | |
701 | ||
702 | p = strjoin(cgroup_root, "/", slice_path, "/", escaped, NULL); | |
703 | if (!p) | |
704 | return -ENOMEM; | |
705 | ||
706 | r = cg_get_xattr(SYSTEMD_CGROUP_CONTROLLER, p, "trusted.invocation_id", ids, 32); | |
707 | if (r < 0) | |
708 | return r; | |
709 | if (r != 32) | |
710 | return -EINVAL; | |
711 | ids[32] = 0; | |
712 | ||
713 | if (!id128_is_valid(ids)) | |
714 | return -EINVAL; | |
715 | ||
716 | copy = strdup(ids); | |
717 | if (!copy) | |
718 | return -ENOMEM; | |
719 | ||
720 | *ret = copy; | |
721 | return 0; | |
722 | } | |
723 | ||
d025f1e4 ZJS |
724 | static void dispatch_message_real( |
725 | Server *s, | |
726 | struct iovec *iovec, unsigned n, unsigned m, | |
3b3154df LP |
727 | const struct ucred *ucred, |
728 | const struct timeval *tv, | |
d025f1e4 | 729 | const char *label, size_t label_len, |
968f3196 | 730 | const char *unit_id, |
d07f7b9e | 731 | int priority, |
968f3196 | 732 | pid_t object_pid) { |
d025f1e4 | 733 | |
968f3196 | 734 | char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)], |
ae018d9b LP |
735 | uid[sizeof("_UID=") + DECIMAL_STR_MAX(uid_t)], |
736 | gid[sizeof("_GID=") + DECIMAL_STR_MAX(gid_t)], | |
737 | owner_uid[sizeof("_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)], | |
d3789917 | 738 | source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)], |
968f3196 ZJS |
739 | o_uid[sizeof("OBJECT_UID=") + DECIMAL_STR_MAX(uid_t)], |
740 | o_gid[sizeof("OBJECT_GID=") + DECIMAL_STR_MAX(gid_t)], | |
741 | o_owner_uid[sizeof("OBJECT_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)]; | |
742 | uid_t object_uid; | |
743 | gid_t object_gid; | |
968f3196 | 744 | char *x; |
d025f1e4 | 745 | int r; |
ae018d9b | 746 | char *t, *c; |
82499507 LP |
747 | uid_t realuid = 0, owner = 0, journal_uid; |
748 | bool owner_valid = false; | |
ae018d9b | 749 | #ifdef HAVE_AUDIT |
968f3196 ZJS |
750 | char audit_session[sizeof("_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], |
751 | audit_loginuid[sizeof("_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)], | |
752 | o_audit_session[sizeof("OBJECT_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], | |
753 | o_audit_loginuid[sizeof("OBJECT_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)]; | |
ae018d9b LP |
754 | |
755 | uint32_t audit; | |
756 | uid_t loginuid; | |
757 | #endif | |
d025f1e4 ZJS |
758 | |
759 | assert(s); | |
760 | assert(iovec); | |
761 | assert(n > 0); | |
d473176a | 762 | assert(n + N_IOVEC_META_FIELDS + (object_pid > 0 ? N_IOVEC_OBJECT_FIELDS : 0) <= m); |
d025f1e4 ZJS |
763 | |
764 | if (ucred) { | |
d025f1e4 ZJS |
765 | realuid = ucred->uid; |
766 | ||
de0671ee | 767 | sprintf(pid, "_PID="PID_FMT, ucred->pid); |
c2457105 | 768 | IOVEC_SET_STRING(iovec[n++], pid); |
d025f1e4 | 769 | |
de0671ee | 770 | sprintf(uid, "_UID="UID_FMT, ucred->uid); |
c2457105 | 771 | IOVEC_SET_STRING(iovec[n++], uid); |
d025f1e4 | 772 | |
de0671ee | 773 | sprintf(gid, "_GID="GID_FMT, ucred->gid); |
c2457105 | 774 | IOVEC_SET_STRING(iovec[n++], gid); |
d025f1e4 ZJS |
775 | |
776 | r = get_process_comm(ucred->pid, &t); | |
777 | if (r >= 0) { | |
63c372cb | 778 | x = strjoina("_COMM=", t); |
d025f1e4 | 779 | free(t); |
968f3196 | 780 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
781 | } |
782 | ||
783 | r = get_process_exe(ucred->pid, &t); | |
784 | if (r >= 0) { | |
63c372cb | 785 | x = strjoina("_EXE=", t); |
d025f1e4 | 786 | free(t); |
968f3196 | 787 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
788 | } |
789 | ||
9bdbc2e2 | 790 | r = get_process_cmdline(ucred->pid, 0, false, &t); |
d025f1e4 | 791 | if (r >= 0) { |
63c372cb | 792 | x = strjoina("_CMDLINE=", t); |
d025f1e4 | 793 | free(t); |
3a832116 SL |
794 | IOVEC_SET_STRING(iovec[n++], x); |
795 | } | |
796 | ||
797 | r = get_process_capeff(ucred->pid, &t); | |
798 | if (r >= 0) { | |
63c372cb | 799 | x = strjoina("_CAP_EFFECTIVE=", t); |
3a832116 | 800 | free(t); |
968f3196 | 801 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 ZJS |
802 | } |
803 | ||
0a20e3c1 | 804 | #ifdef HAVE_AUDIT |
d025f1e4 | 805 | r = audit_session_from_pid(ucred->pid, &audit); |
ae018d9b | 806 | if (r >= 0) { |
de0671ee | 807 | sprintf(audit_session, "_AUDIT_SESSION=%"PRIu32, audit); |
ae018d9b LP |
808 | IOVEC_SET_STRING(iovec[n++], audit_session); |
809 | } | |
d025f1e4 ZJS |
810 | |
811 | r = audit_loginuid_from_pid(ucred->pid, &loginuid); | |
7027ff61 | 812 | if (r >= 0) { |
de0671ee | 813 | sprintf(audit_loginuid, "_AUDIT_LOGINUID="UID_FMT, loginuid); |
ae018d9b | 814 | IOVEC_SET_STRING(iovec[n++], audit_loginuid); |
d025f1e4 | 815 | } |
ae018d9b | 816 | #endif |
d025f1e4 | 817 | |
e9174f29 | 818 | r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c); |
7027ff61 | 819 | if (r >= 0) { |
4b58153d | 820 | _cleanup_free_ char *raw_unit = NULL, *raw_slice = NULL; |
968f3196 ZJS |
821 | char *session = NULL; |
822 | ||
63c372cb | 823 | x = strjoina("_SYSTEMD_CGROUP=", c); |
968f3196 | 824 | IOVEC_SET_STRING(iovec[n++], x); |
d025f1e4 | 825 | |
ae018d9b LP |
826 | r = cg_path_get_session(c, &t); |
827 | if (r >= 0) { | |
63c372cb | 828 | session = strjoina("_SYSTEMD_SESSION=", t); |
ae018d9b | 829 | free(t); |
d025f1e4 | 830 | IOVEC_SET_STRING(iovec[n++], session); |
ae018d9b LP |
831 | } |
832 | ||
833 | if (cg_path_get_owner_uid(c, &owner) >= 0) { | |
834 | owner_valid = true; | |
d025f1e4 | 835 | |
de0671ee | 836 | sprintf(owner_uid, "_SYSTEMD_OWNER_UID="UID_FMT, owner); |
d025f1e4 | 837 | IOVEC_SET_STRING(iovec[n++], owner_uid); |
ae018d9b | 838 | } |
d025f1e4 | 839 | |
4b58153d LP |
840 | if (cg_path_get_unit(c, &raw_unit) >= 0) { |
841 | x = strjoina("_SYSTEMD_UNIT=", raw_unit); | |
19cace37 LP |
842 | IOVEC_SET_STRING(iovec[n++], x); |
843 | } else if (unit_id && !session) { | |
63c372cb | 844 | x = strjoina("_SYSTEMD_UNIT=", unit_id); |
19cace37 LP |
845 | IOVEC_SET_STRING(iovec[n++], x); |
846 | } | |
847 | ||
848 | if (cg_path_get_user_unit(c, &t) >= 0) { | |
63c372cb | 849 | x = strjoina("_SYSTEMD_USER_UNIT=", t); |
ae018d9b | 850 | free(t); |
968f3196 | 851 | IOVEC_SET_STRING(iovec[n++], x); |
19cace37 | 852 | } else if (unit_id && session) { |
63c372cb | 853 | x = strjoina("_SYSTEMD_USER_UNIT=", unit_id); |
19cace37 LP |
854 | IOVEC_SET_STRING(iovec[n++], x); |
855 | } | |
ae018d9b | 856 | |
4b58153d LP |
857 | if (cg_path_get_slice(c, &raw_slice) >= 0) { |
858 | x = strjoina("_SYSTEMD_SLICE=", raw_slice); | |
0a244b8e LP |
859 | IOVEC_SET_STRING(iovec[n++], x); |
860 | } | |
861 | ||
d473176a LP |
862 | if (cg_path_get_user_slice(c, &t) >= 0) { |
863 | x = strjoina("_SYSTEMD_USER_SLICE=", t); | |
864 | free(t); | |
865 | IOVEC_SET_STRING(iovec[n++], x); | |
866 | } | |
867 | ||
4b58153d LP |
868 | if (raw_slice && raw_unit) { |
869 | if (get_invocation_id(s->cgroup_root, raw_slice, raw_unit, &t) >= 0) { | |
870 | x = strjoina("_SYSTEMD_INVOCATION_ID=", t); | |
871 | free(t); | |
872 | IOVEC_SET_STRING(iovec[n++], x); | |
873 | } | |
874 | } | |
875 | ||
ae018d9b | 876 | free(c); |
2d43b190 | 877 | } else if (unit_id) { |
63c372cb | 878 | x = strjoina("_SYSTEMD_UNIT=", unit_id); |
2d43b190 | 879 | IOVEC_SET_STRING(iovec[n++], x); |
ef1673d1 | 880 | } |
d025f1e4 | 881 | |
d025f1e4 | 882 | #ifdef HAVE_SELINUX |
6355e756 | 883 | if (mac_selinux_have()) { |
d682b3a7 | 884 | if (label) { |
f8294e41 | 885 | x = alloca(strlen("_SELINUX_CONTEXT=") + label_len + 1); |
ae018d9b | 886 | |
d682b3a7 LP |
887 | *((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0; |
888 | IOVEC_SET_STRING(iovec[n++], x); | |
889 | } else { | |
2ed96880 | 890 | char *con; |
d025f1e4 | 891 | |
d682b3a7 | 892 | if (getpidcon(ucred->pid, &con) >= 0) { |
63c372cb | 893 | x = strjoina("_SELINUX_CONTEXT=", con); |
e7ff4e7f | 894 | |
d682b3a7 LP |
895 | freecon(con); |
896 | IOVEC_SET_STRING(iovec[n++], x); | |
897 | } | |
d025f1e4 ZJS |
898 | } |
899 | } | |
900 | #endif | |
901 | } | |
968f3196 ZJS |
902 | assert(n <= m); |
903 | ||
904 | if (object_pid) { | |
905 | r = get_process_uid(object_pid, &object_uid); | |
906 | if (r >= 0) { | |
de0671ee | 907 | sprintf(o_uid, "OBJECT_UID="UID_FMT, object_uid); |
968f3196 ZJS |
908 | IOVEC_SET_STRING(iovec[n++], o_uid); |
909 | } | |
910 | ||
911 | r = get_process_gid(object_pid, &object_gid); | |
912 | if (r >= 0) { | |
de0671ee | 913 | sprintf(o_gid, "OBJECT_GID="GID_FMT, object_gid); |
968f3196 ZJS |
914 | IOVEC_SET_STRING(iovec[n++], o_gid); |
915 | } | |
916 | ||
917 | r = get_process_comm(object_pid, &t); | |
918 | if (r >= 0) { | |
63c372cb | 919 | x = strjoina("OBJECT_COMM=", t); |
968f3196 ZJS |
920 | free(t); |
921 | IOVEC_SET_STRING(iovec[n++], x); | |
922 | } | |
923 | ||
924 | r = get_process_exe(object_pid, &t); | |
925 | if (r >= 0) { | |
63c372cb | 926 | x = strjoina("OBJECT_EXE=", t); |
968f3196 ZJS |
927 | free(t); |
928 | IOVEC_SET_STRING(iovec[n++], x); | |
929 | } | |
930 | ||
931 | r = get_process_cmdline(object_pid, 0, false, &t); | |
932 | if (r >= 0) { | |
63c372cb | 933 | x = strjoina("OBJECT_CMDLINE=", t); |
968f3196 ZJS |
934 | free(t); |
935 | IOVEC_SET_STRING(iovec[n++], x); | |
936 | } | |
937 | ||
938 | #ifdef HAVE_AUDIT | |
939 | r = audit_session_from_pid(object_pid, &audit); | |
940 | if (r >= 0) { | |
de0671ee | 941 | sprintf(o_audit_session, "OBJECT_AUDIT_SESSION=%"PRIu32, audit); |
968f3196 ZJS |
942 | IOVEC_SET_STRING(iovec[n++], o_audit_session); |
943 | } | |
944 | ||
945 | r = audit_loginuid_from_pid(object_pid, &loginuid); | |
946 | if (r >= 0) { | |
de0671ee | 947 | sprintf(o_audit_loginuid, "OBJECT_AUDIT_LOGINUID="UID_FMT, loginuid); |
968f3196 ZJS |
948 | IOVEC_SET_STRING(iovec[n++], o_audit_loginuid); |
949 | } | |
950 | #endif | |
951 | ||
e9174f29 | 952 | r = cg_pid_get_path_shifted(object_pid, s->cgroup_root, &c); |
968f3196 | 953 | if (r >= 0) { |
63c372cb | 954 | x = strjoina("OBJECT_SYSTEMD_CGROUP=", c); |
968f3196 ZJS |
955 | IOVEC_SET_STRING(iovec[n++], x); |
956 | ||
957 | r = cg_path_get_session(c, &t); | |
958 | if (r >= 0) { | |
63c372cb | 959 | x = strjoina("OBJECT_SYSTEMD_SESSION=", t); |
968f3196 ZJS |
960 | free(t); |
961 | IOVEC_SET_STRING(iovec[n++], x); | |
962 | } | |
963 | ||
964 | if (cg_path_get_owner_uid(c, &owner) >= 0) { | |
de0671ee | 965 | sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID="UID_FMT, owner); |
968f3196 ZJS |
966 | IOVEC_SET_STRING(iovec[n++], o_owner_uid); |
967 | } | |
968 | ||
969 | if (cg_path_get_unit(c, &t) >= 0) { | |
63c372cb | 970 | x = strjoina("OBJECT_SYSTEMD_UNIT=", t); |
968f3196 | 971 | free(t); |
19cace37 LP |
972 | IOVEC_SET_STRING(iovec[n++], x); |
973 | } | |
974 | ||
975 | if (cg_path_get_user_unit(c, &t) >= 0) { | |
63c372cb | 976 | x = strjoina("OBJECT_SYSTEMD_USER_UNIT=", t); |
968f3196 | 977 | free(t); |
968f3196 | 978 | IOVEC_SET_STRING(iovec[n++], x); |
19cace37 | 979 | } |
968f3196 | 980 | |
d473176a LP |
981 | if (cg_path_get_slice(c, &t) >= 0) { |
982 | x = strjoina("OBJECT_SYSTEMD_SLICE=", t); | |
983 | free(t); | |
984 | IOVEC_SET_STRING(iovec[n++], x); | |
985 | } | |
986 | ||
987 | if (cg_path_get_user_slice(c, &t) >= 0) { | |
988 | x = strjoina("OBJECT_SYSTEMD_USER_SLICE=", t); | |
989 | free(t); | |
990 | IOVEC_SET_STRING(iovec[n++], x); | |
991 | } | |
992 | ||
968f3196 ZJS |
993 | free(c); |
994 | } | |
995 | } | |
996 | assert(n <= m); | |
d025f1e4 ZJS |
997 | |
998 | if (tv) { | |
398a50cd | 999 | sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=" USEC_FMT, timeval_load(tv)); |
a5693989 | 1000 | IOVEC_SET_STRING(iovec[n++], source_time); |
d025f1e4 ZJS |
1001 | } |
1002 | ||
1003 | /* Note that strictly speaking storing the boot id here is | |
1004 | * redundant since the entry includes this in-line | |
1005 | * anyway. However, we need this indexed, too. */ | |
0c24bb23 LP |
1006 | if (!isempty(s->boot_id_field)) |
1007 | IOVEC_SET_STRING(iovec[n++], s->boot_id_field); | |
d025f1e4 | 1008 | |
0c24bb23 LP |
1009 | if (!isempty(s->machine_id_field)) |
1010 | IOVEC_SET_STRING(iovec[n++], s->machine_id_field); | |
d025f1e4 | 1011 | |
0c24bb23 LP |
1012 | if (!isempty(s->hostname_field)) |
1013 | IOVEC_SET_STRING(iovec[n++], s->hostname_field); | |
d025f1e4 ZJS |
1014 | |
1015 | assert(n <= m); | |
1016 | ||
da499392 | 1017 | if (s->split_mode == SPLIT_UID && realuid > 0) |
40adcda8 | 1018 | /* Split up strictly by any UID */ |
759c945a | 1019 | journal_uid = realuid; |
82499507 | 1020 | else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0) |
edc3797f LP |
1021 | /* Split up by login UIDs. We do this only if the |
1022 | * realuid is not root, in order not to accidentally | |
1023 | * leak privileged information to the user that is | |
1024 | * logged by a privileged process that is part of an | |
7517e174 | 1025 | * unprivileged session. */ |
8a0889df | 1026 | journal_uid = owner; |
da499392 KS |
1027 | else |
1028 | journal_uid = 0; | |
759c945a | 1029 | |
d07f7b9e | 1030 | write_to_journal(s, journal_uid, iovec, n, priority); |
d025f1e4 ZJS |
1031 | } |
1032 | ||
1033 | void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) { | |
1034 | char mid[11 + 32 + 1]; | |
8a03c9ef ZJS |
1035 | struct iovec iovec[N_IOVEC_META_FIELDS + 5 + N_IOVEC_PAYLOAD_FIELDS]; |
1036 | unsigned n = 0, m; | |
32917e33 | 1037 | int r; |
d025f1e4 | 1038 | va_list ap; |
b92bea5d | 1039 | struct ucred ucred = {}; |
d025f1e4 ZJS |
1040 | |
1041 | assert(s); | |
1042 | assert(format); | |
1043 | ||
4850d39a | 1044 | assert_cc(3 == LOG_FAC(LOG_DAEMON)); |
b6fa2555 EV |
1045 | IOVEC_SET_STRING(iovec[n++], "SYSLOG_FACILITY=3"); |
1046 | IOVEC_SET_STRING(iovec[n++], "SYSLOG_IDENTIFIER=systemd-journald"); | |
1047 | ||
d025f1e4 | 1048 | IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver"); |
4850d39a | 1049 | assert_cc(6 == LOG_INFO); |
32917e33 | 1050 | IOVEC_SET_STRING(iovec[n++], "PRIORITY=6"); |
d025f1e4 | 1051 | |
3bbaff3e | 1052 | if (!sd_id128_is_null(message_id)) { |
e2cc6eca | 1053 | snprintf(mid, sizeof(mid), LOG_MESSAGE_ID(message_id)); |
d025f1e4 ZJS |
1054 | IOVEC_SET_STRING(iovec[n++], mid); |
1055 | } | |
1056 | ||
8a03c9ef ZJS |
1057 | m = n; |
1058 | ||
1059 | va_start(ap, format); | |
32917e33 ZJS |
1060 | r = log_format_iovec(iovec, ELEMENTSOF(iovec), &n, false, 0, format, ap); |
1061 | /* Error handling below */ | |
8a03c9ef ZJS |
1062 | va_end(ap); |
1063 | ||
d025f1e4 ZJS |
1064 | ucred.pid = getpid(); |
1065 | ucred.uid = getuid(); | |
1066 | ucred.gid = getgid(); | |
1067 | ||
32917e33 ZJS |
1068 | if (r >= 0) |
1069 | dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0); | |
8a03c9ef ZJS |
1070 | |
1071 | while (m < n) | |
1072 | free(iovec[m++].iov_base); | |
32917e33 ZJS |
1073 | |
1074 | if (r < 0) { | |
1075 | /* We failed to format the message. Emit a warning instead. */ | |
1076 | char buf[LINE_MAX]; | |
1077 | ||
1078 | xsprintf(buf, "MESSAGE=Entry printing failed: %s", strerror(-r)); | |
1079 | ||
1080 | n = 3; | |
1081 | IOVEC_SET_STRING(iovec[n++], "PRIORITY=4"); | |
1082 | IOVEC_SET_STRING(iovec[n++], buf); | |
1083 | dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0); | |
1084 | } | |
d025f1e4 ZJS |
1085 | } |
1086 | ||
1087 | void server_dispatch_message( | |
1088 | Server *s, | |
1089 | struct iovec *iovec, unsigned n, unsigned m, | |
3b3154df LP |
1090 | const struct ucred *ucred, |
1091 | const struct timeval *tv, | |
d025f1e4 ZJS |
1092 | const char *label, size_t label_len, |
1093 | const char *unit_id, | |
968f3196 ZJS |
1094 | int priority, |
1095 | pid_t object_pid) { | |
d025f1e4 | 1096 | |
7027ff61 | 1097 | int rl, r; |
7fd1b19b | 1098 | _cleanup_free_ char *path = NULL; |
8580d1f7 | 1099 | uint64_t available = 0; |
db91ea32 | 1100 | char *c; |
d025f1e4 ZJS |
1101 | |
1102 | assert(s); | |
1103 | assert(iovec || n == 0); | |
1104 | ||
1105 | if (n == 0) | |
1106 | return; | |
1107 | ||
1108 | if (LOG_PRI(priority) > s->max_level_store) | |
1109 | return; | |
1110 | ||
2f5df74a HHPF |
1111 | /* Stop early in case the information will not be stored |
1112 | * in a journal. */ | |
1113 | if (s->storage == STORAGE_NONE) | |
1114 | return; | |
1115 | ||
d025f1e4 ZJS |
1116 | if (!ucred) |
1117 | goto finish; | |
1118 | ||
e9174f29 | 1119 | r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path); |
7027ff61 | 1120 | if (r < 0) |
d025f1e4 ZJS |
1121 | goto finish; |
1122 | ||
1123 | /* example: /user/lennart/3/foobar | |
1124 | * /system/dbus.service/foobar | |
1125 | * | |
1126 | * So let's cut of everything past the third /, since that is | |
1127 | * where user directories start */ | |
1128 | ||
1129 | c = strchr(path, '/'); | |
1130 | if (c) { | |
1131 | c = strchr(c+1, '/'); | |
1132 | if (c) { | |
1133 | c = strchr(c+1, '/'); | |
1134 | if (c) | |
1135 | *c = 0; | |
1136 | } | |
1137 | } | |
1138 | ||
8580d1f7 LP |
1139 | (void) determine_space(s, false, false, &available, NULL); |
1140 | rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available); | |
db91ea32 | 1141 | if (rl == 0) |
d025f1e4 | 1142 | return; |
d025f1e4 ZJS |
1143 | |
1144 | /* Write a suppression message if we suppressed something */ | |
1145 | if (rl > 1) | |
db91ea32 | 1146 | server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, |
8a03c9ef ZJS |
1147 | LOG_MESSAGE("Suppressed %u messages from %s", rl - 1, path), |
1148 | NULL); | |
d025f1e4 ZJS |
1149 | |
1150 | finish: | |
d07f7b9e | 1151 | dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid); |
d025f1e4 ZJS |
1152 | } |
1153 | ||
d025f1e4 | 1154 | int server_flush_to_var(Server *s) { |
d025f1e4 ZJS |
1155 | sd_id128_t machine; |
1156 | sd_journal *j = NULL; | |
fbb63411 LP |
1157 | char ts[FORMAT_TIMESPAN_MAX]; |
1158 | usec_t start; | |
1159 | unsigned n = 0; | |
1160 | int r; | |
d025f1e4 ZJS |
1161 | |
1162 | assert(s); | |
1163 | ||
1164 | if (s->storage != STORAGE_AUTO && | |
1165 | s->storage != STORAGE_PERSISTENT) | |
1166 | return 0; | |
1167 | ||
1168 | if (!s->runtime_journal) | |
1169 | return 0; | |
1170 | ||
8580d1f7 | 1171 | (void) system_journal_open(s, true); |
d025f1e4 ZJS |
1172 | |
1173 | if (!s->system_journal) | |
1174 | return 0; | |
1175 | ||
1176 | log_debug("Flushing to /var..."); | |
1177 | ||
fbb63411 LP |
1178 | start = now(CLOCK_MONOTONIC); |
1179 | ||
d025f1e4 | 1180 | r = sd_id128_get_machine(&machine); |
00a16861 | 1181 | if (r < 0) |
d025f1e4 | 1182 | return r; |
d025f1e4 ZJS |
1183 | |
1184 | r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); | |
23bbb0de MS |
1185 | if (r < 0) |
1186 | return log_error_errno(r, "Failed to read runtime journal: %m"); | |
d025f1e4 | 1187 | |
93b73b06 LP |
1188 | sd_journal_set_data_threshold(j, 0); |
1189 | ||
d025f1e4 ZJS |
1190 | SD_JOURNAL_FOREACH(j) { |
1191 | Object *o = NULL; | |
1192 | JournalFile *f; | |
1193 | ||
1194 | f = j->current_file; | |
1195 | assert(f && f->current_offset > 0); | |
1196 | ||
fbb63411 LP |
1197 | n++; |
1198 | ||
d025f1e4 ZJS |
1199 | r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); |
1200 | if (r < 0) { | |
da927ba9 | 1201 | log_error_errno(r, "Can't read entry: %m"); |
d025f1e4 ZJS |
1202 | goto finish; |
1203 | } | |
1204 | ||
1205 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
1206 | if (r >= 0) | |
1207 | continue; | |
1208 | ||
1209 | if (!shall_try_append_again(s->system_journal, r)) { | |
da927ba9 | 1210 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1211 | goto finish; |
1212 | } | |
1213 | ||
1214 | server_rotate(s); | |
8580d1f7 | 1215 | server_vacuum(s, false, false); |
d025f1e4 | 1216 | |
253f59df LP |
1217 | if (!s->system_journal) { |
1218 | log_notice("Didn't flush runtime journal since rotation of system journal wasn't successful."); | |
1219 | r = -EIO; | |
1220 | goto finish; | |
1221 | } | |
1222 | ||
d025f1e4 ZJS |
1223 | log_debug("Retrying write."); |
1224 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
1225 | if (r < 0) { | |
da927ba9 | 1226 | log_error_errno(r, "Can't write entry: %m"); |
d025f1e4 ZJS |
1227 | goto finish; |
1228 | } | |
1229 | } | |
1230 | ||
804ae586 LP |
1231 | r = 0; |
1232 | ||
d025f1e4 ZJS |
1233 | finish: |
1234 | journal_file_post_change(s->system_journal); | |
1235 | ||
804ae586 | 1236 | s->runtime_journal = journal_file_close(s->runtime_journal); |
d025f1e4 ZJS |
1237 | |
1238 | if (r >= 0) | |
c6878637 | 1239 | (void) rm_rf("/run/log/journal", REMOVE_ROOT); |
d025f1e4 | 1240 | |
763c7aa2 | 1241 | sd_journal_close(j); |
d025f1e4 | 1242 | |
8a03c9ef ZJS |
1243 | server_driver_message(s, SD_ID128_NULL, |
1244 | LOG_MESSAGE("Time spent on flushing to /var is %s for %u entries.", | |
1245 | format_timespan(ts, sizeof(ts), now(CLOCK_MONOTONIC) - start, 0), | |
1246 | n), | |
1247 | NULL); | |
fbb63411 | 1248 | |
d025f1e4 ZJS |
1249 | return r; |
1250 | } | |
1251 | ||
8531ae70 | 1252 | int server_process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
f9a810be | 1253 | Server *s = userdata; |
a315ac4e LP |
1254 | struct ucred *ucred = NULL; |
1255 | struct timeval *tv = NULL; | |
1256 | struct cmsghdr *cmsg; | |
1257 | char *label = NULL; | |
1258 | size_t label_len = 0, m; | |
1259 | struct iovec iovec; | |
1260 | ssize_t n; | |
1261 | int *fds = NULL, v = 0; | |
1262 | unsigned n_fds = 0; | |
1263 | ||
1264 | union { | |
1265 | struct cmsghdr cmsghdr; | |
1266 | ||
1267 | /* We use NAME_MAX space for the SELinux label | |
1268 | * here. The kernel currently enforces no | |
1269 | * limit, but according to suggestions from | |
1270 | * the SELinux people this will change and it | |
1271 | * will probably be identical to NAME_MAX. For | |
1272 | * now we use that, but this should be updated | |
1273 | * one day when the final limit is known. */ | |
1274 | uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) + | |
1275 | CMSG_SPACE(sizeof(struct timeval)) + | |
1276 | CMSG_SPACE(sizeof(int)) + /* fd */ | |
1277 | CMSG_SPACE(NAME_MAX)]; /* selinux label */ | |
1278 | } control = {}; | |
1279 | ||
1280 | union sockaddr_union sa = {}; | |
1281 | ||
1282 | struct msghdr msghdr = { | |
1283 | .msg_iov = &iovec, | |
1284 | .msg_iovlen = 1, | |
1285 | .msg_control = &control, | |
1286 | .msg_controllen = sizeof(control), | |
1287 | .msg_name = &sa, | |
1288 | .msg_namelen = sizeof(sa), | |
1289 | }; | |
f9a810be | 1290 | |
d025f1e4 | 1291 | assert(s); |
875c2e22 | 1292 | assert(fd == s->native_fd || fd == s->syslog_fd || fd == s->audit_fd); |
f9a810be LP |
1293 | |
1294 | if (revents != EPOLLIN) { | |
1295 | log_error("Got invalid event from epoll for datagram fd: %"PRIx32, revents); | |
1296 | return -EIO; | |
1297 | } | |
1298 | ||
a315ac4e LP |
1299 | /* Try to get the right size, if we can. (Not all |
1300 | * sockets support SIOCINQ, hence we just try, but | |
1301 | * don't rely on it. */ | |
1302 | (void) ioctl(fd, SIOCINQ, &v); | |
d025f1e4 | 1303 | |
a315ac4e LP |
1304 | /* Fix it up, if it is too small. We use the same fixed value as auditd here. Awful! */ |
1305 | m = PAGE_ALIGN(MAX3((size_t) v + 1, | |
1306 | (size_t) LINE_MAX, | |
1307 | ALIGN(sizeof(struct nlmsghdr)) + ALIGN((size_t) MAX_AUDIT_MESSAGE_LENGTH)) + 1); | |
d025f1e4 | 1308 | |
a315ac4e LP |
1309 | if (!GREEDY_REALLOC(s->buffer, s->buffer_size, m)) |
1310 | return log_oom(); | |
875c2e22 | 1311 | |
a315ac4e LP |
1312 | iovec.iov_base = s->buffer; |
1313 | iovec.iov_len = s->buffer_size - 1; /* Leave room for trailing NUL we add later */ | |
d025f1e4 | 1314 | |
a315ac4e LP |
1315 | n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); |
1316 | if (n < 0) { | |
1317 | if (errno == EINTR || errno == EAGAIN) | |
1318 | return 0; | |
875c2e22 | 1319 | |
a315ac4e LP |
1320 | return log_error_errno(errno, "recvmsg() failed: %m"); |
1321 | } | |
875c2e22 | 1322 | |
a315ac4e LP |
1323 | CMSG_FOREACH(cmsg, &msghdr) { |
1324 | ||
1325 | if (cmsg->cmsg_level == SOL_SOCKET && | |
1326 | cmsg->cmsg_type == SCM_CREDENTIALS && | |
1327 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) | |
1328 | ucred = (struct ucred*) CMSG_DATA(cmsg); | |
1329 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1330 | cmsg->cmsg_type == SCM_SECURITY) { | |
1331 | label = (char*) CMSG_DATA(cmsg); | |
1332 | label_len = cmsg->cmsg_len - CMSG_LEN(0); | |
1333 | } else if (cmsg->cmsg_level == SOL_SOCKET && | |
1334 | cmsg->cmsg_type == SO_TIMESTAMP && | |
1335 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval))) | |
1336 | tv = (struct timeval*) CMSG_DATA(cmsg); | |
1337 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1338 | cmsg->cmsg_type == SCM_RIGHTS) { | |
1339 | fds = (int*) CMSG_DATA(cmsg); | |
1340 | n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); | |
d025f1e4 | 1341 | } |
a315ac4e | 1342 | } |
d025f1e4 | 1343 | |
a315ac4e LP |
1344 | /* And a trailing NUL, just in case */ |
1345 | s->buffer[n] = 0; | |
1346 | ||
1347 | if (fd == s->syslog_fd) { | |
1348 | if (n > 0 && n_fds == 0) | |
1349 | server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len); | |
1350 | else if (n_fds > 0) | |
1351 | log_warning("Got file descriptors via syslog socket. Ignoring."); | |
1352 | ||
1353 | } else if (fd == s->native_fd) { | |
1354 | if (n > 0 && n_fds == 0) | |
1355 | server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len); | |
1356 | else if (n == 0 && n_fds == 1) | |
1357 | server_process_native_file(s, fds[0], ucred, tv, label, label_len); | |
1358 | else if (n_fds > 0) | |
1359 | log_warning("Got too many file descriptors via native socket. Ignoring."); | |
1360 | ||
1361 | } else { | |
1362 | assert(fd == s->audit_fd); | |
1363 | ||
1364 | if (n > 0 && n_fds == 0) | |
1365 | server_process_audit_message(s, s->buffer, n, ucred, &sa, msghdr.msg_namelen); | |
1366 | else if (n_fds > 0) | |
1367 | log_warning("Got file descriptors via audit socket. Ignoring."); | |
f9a810be | 1368 | } |
a315ac4e LP |
1369 | |
1370 | close_many(fds, n_fds); | |
1371 | return 0; | |
f9a810be | 1372 | } |
d025f1e4 | 1373 | |
f9a810be LP |
1374 | static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1375 | Server *s = userdata; | |
33d52ab9 | 1376 | int r; |
d025f1e4 | 1377 | |
f9a810be | 1378 | assert(s); |
d025f1e4 | 1379 | |
94b65516 | 1380 | log_info("Received request to flush runtime journal from PID " PID_FMT, si->ssi_pid); |
d025f1e4 | 1381 | |
929eeb54 | 1382 | (void) server_flush_to_var(s); |
f9a810be | 1383 | server_sync(s); |
8580d1f7 | 1384 | server_vacuum(s, false, false); |
d025f1e4 | 1385 | |
33d52ab9 LP |
1386 | r = touch("/run/systemd/journal/flushed"); |
1387 | if (r < 0) | |
1388 | log_warning_errno(r, "Failed to touch /run/systemd/journal/flushed, ignoring: %m"); | |
74055aa7 | 1389 | |
f9a810be LP |
1390 | return 0; |
1391 | } | |
d025f1e4 | 1392 | |
f9a810be LP |
1393 | static int dispatch_sigusr2(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1394 | Server *s = userdata; | |
33d52ab9 | 1395 | int r; |
d025f1e4 | 1396 | |
f9a810be | 1397 | assert(s); |
d025f1e4 | 1398 | |
94b65516 | 1399 | log_info("Received request to rotate journal from PID " PID_FMT, si->ssi_pid); |
f9a810be | 1400 | server_rotate(s); |
8580d1f7 | 1401 | server_vacuum(s, true, true); |
d025f1e4 | 1402 | |
dbd6e31c | 1403 | /* Let clients know when the most recent rotation happened. */ |
33d52ab9 LP |
1404 | r = write_timestamp_file_atomic("/run/systemd/journal/rotated", now(CLOCK_MONOTONIC)); |
1405 | if (r < 0) | |
1406 | log_warning_errno(r, "Failed to write /run/systemd/journal/rotated, ignoring: %m"); | |
dbd6e31c | 1407 | |
f9a810be LP |
1408 | return 0; |
1409 | } | |
d025f1e4 | 1410 | |
f9a810be LP |
1411 | static int dispatch_sigterm(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1412 | Server *s = userdata; | |
d025f1e4 | 1413 | |
f9a810be | 1414 | assert(s); |
d025f1e4 | 1415 | |
4daf54a8 | 1416 | log_received_signal(LOG_INFO, si); |
d025f1e4 | 1417 | |
6203e07a | 1418 | sd_event_exit(s->event, 0); |
d025f1e4 ZJS |
1419 | return 0; |
1420 | } | |
1421 | ||
94b65516 LP |
1422 | static int dispatch_sigrtmin1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { |
1423 | Server *s = userdata; | |
33d52ab9 | 1424 | int r; |
94b65516 LP |
1425 | |
1426 | assert(s); | |
1427 | ||
1428 | log_debug("Received request to sync from PID " PID_FMT, si->ssi_pid); | |
1429 | ||
1430 | server_sync(s); | |
1431 | ||
1432 | /* Let clients know when the most recent sync happened. */ | |
33d52ab9 LP |
1433 | r = write_timestamp_file_atomic("/run/systemd/journal/synced", now(CLOCK_MONOTONIC)); |
1434 | if (r < 0) | |
1435 | log_warning_errno(r, "Failed to write /run/systemd/journal/synced, ignoring: %m"); | |
94b65516 LP |
1436 | |
1437 | return 0; | |
1438 | } | |
1439 | ||
f9a810be | 1440 | static int setup_signals(Server *s) { |
f9a810be | 1441 | int r; |
d025f1e4 ZJS |
1442 | |
1443 | assert(s); | |
1444 | ||
94b65516 | 1445 | assert(sigprocmask_many(SIG_SETMASK, NULL, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, SIGRTMIN+1, -1) >= 0); |
d025f1e4 | 1446 | |
151b9b96 | 1447 | r = sd_event_add_signal(s->event, &s->sigusr1_event_source, SIGUSR1, dispatch_sigusr1, s); |
f9a810be LP |
1448 | if (r < 0) |
1449 | return r; | |
1450 | ||
151b9b96 | 1451 | r = sd_event_add_signal(s->event, &s->sigusr2_event_source, SIGUSR2, dispatch_sigusr2, s); |
f9a810be LP |
1452 | if (r < 0) |
1453 | return r; | |
d025f1e4 | 1454 | |
151b9b96 | 1455 | r = sd_event_add_signal(s->event, &s->sigterm_event_source, SIGTERM, dispatch_sigterm, s); |
f9a810be LP |
1456 | if (r < 0) |
1457 | return r; | |
d025f1e4 | 1458 | |
b374689c LP |
1459 | /* Let's process SIGTERM late, so that we flush all queued |
1460 | * messages to disk before we exit */ | |
1461 | r = sd_event_source_set_priority(s->sigterm_event_source, SD_EVENT_PRIORITY_NORMAL+20); | |
1462 | if (r < 0) | |
1463 | return r; | |
1464 | ||
1465 | /* When journald is invoked on the terminal (when debugging), | |
1466 | * it's useful if C-c is handled equivalent to SIGTERM. */ | |
151b9b96 | 1467 | r = sd_event_add_signal(s->event, &s->sigint_event_source, SIGINT, dispatch_sigterm, s); |
f9a810be LP |
1468 | if (r < 0) |
1469 | return r; | |
d025f1e4 | 1470 | |
b374689c LP |
1471 | r = sd_event_source_set_priority(s->sigint_event_source, SD_EVENT_PRIORITY_NORMAL+20); |
1472 | if (r < 0) | |
1473 | return r; | |
1474 | ||
94b65516 LP |
1475 | /* SIGRTMIN+1 causes an immediate sync. We process this very |
1476 | * late, so that everything else queued at this point is | |
1477 | * really written to disk. Clients can watch | |
1478 | * /run/systemd/journal/synced with inotify until its mtime | |
1479 | * changes to see when a sync happened. */ | |
1480 | r = sd_event_add_signal(s->event, &s->sigrtmin1_event_source, SIGRTMIN+1, dispatch_sigrtmin1, s); | |
1481 | if (r < 0) | |
1482 | return r; | |
1483 | ||
1484 | r = sd_event_source_set_priority(s->sigrtmin1_event_source, SD_EVENT_PRIORITY_NORMAL+15); | |
1485 | if (r < 0) | |
1486 | return r; | |
1487 | ||
d025f1e4 ZJS |
1488 | return 0; |
1489 | } | |
1490 | ||
1491 | static int server_parse_proc_cmdline(Server *s) { | |
7fd1b19b | 1492 | _cleanup_free_ char *line = NULL; |
d581d9d9 | 1493 | const char *p; |
74df0fca | 1494 | int r; |
d025f1e4 | 1495 | |
74df0fca | 1496 | r = proc_cmdline(&line); |
b5884878 | 1497 | if (r < 0) { |
da927ba9 | 1498 | log_warning_errno(r, "Failed to read /proc/cmdline, ignoring: %m"); |
d025f1e4 | 1499 | return 0; |
b5884878 | 1500 | } |
d025f1e4 | 1501 | |
d581d9d9 | 1502 | p = line; |
9ed794a3 | 1503 | for (;;) { |
ff82c36c | 1504 | _cleanup_free_ char *word = NULL; |
d025f1e4 | 1505 | |
d581d9d9 SS |
1506 | r = extract_first_word(&p, &word, NULL, 0); |
1507 | if (r < 0) | |
1508 | return log_error_errno(r, "Failed to parse journald syntax \"%s\": %m", line); | |
1509 | ||
1510 | if (r == 0) | |
1511 | break; | |
d025f1e4 ZJS |
1512 | |
1513 | if (startswith(word, "systemd.journald.forward_to_syslog=")) { | |
1514 | r = parse_boolean(word + 35); | |
1515 | if (r < 0) | |
1516 | log_warning("Failed to parse forward to syslog switch %s. Ignoring.", word + 35); | |
1517 | else | |
1518 | s->forward_to_syslog = r; | |
1519 | } else if (startswith(word, "systemd.journald.forward_to_kmsg=")) { | |
1520 | r = parse_boolean(word + 33); | |
1521 | if (r < 0) | |
1522 | log_warning("Failed to parse forward to kmsg switch %s. Ignoring.", word + 33); | |
1523 | else | |
1524 | s->forward_to_kmsg = r; | |
1525 | } else if (startswith(word, "systemd.journald.forward_to_console=")) { | |
1526 | r = parse_boolean(word + 36); | |
1527 | if (r < 0) | |
1528 | log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36); | |
1529 | else | |
1530 | s->forward_to_console = r; | |
40b71e89 ST |
1531 | } else if (startswith(word, "systemd.journald.forward_to_wall=")) { |
1532 | r = parse_boolean(word + 33); | |
1533 | if (r < 0) | |
1534 | log_warning("Failed to parse forward to wall switch %s. Ignoring.", word + 33); | |
1535 | else | |
1536 | s->forward_to_wall = r; | |
d025f1e4 ZJS |
1537 | } else if (startswith(word, "systemd.journald")) |
1538 | log_warning("Invalid systemd.journald parameter. Ignoring."); | |
d025f1e4 ZJS |
1539 | } |
1540 | ||
804ae586 | 1541 | /* do not warn about state here, since probably systemd already did */ |
db91ea32 | 1542 | return 0; |
d025f1e4 ZJS |
1543 | } |
1544 | ||
1545 | static int server_parse_config_file(Server *s) { | |
d025f1e4 ZJS |
1546 | assert(s); |
1547 | ||
43688c49 | 1548 | return config_parse_many_nulstr(PKGSYSCONFDIR "/journald.conf", |
75eb6154 | 1549 | CONF_PATHS_NULSTR("systemd/journald.conf.d"), |
a9edaeff JT |
1550 | "Journal\0", |
1551 | config_item_perf_lookup, journald_gperf_lookup, | |
1552 | false, s); | |
d025f1e4 ZJS |
1553 | } |
1554 | ||
f9a810be LP |
1555 | static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) { |
1556 | Server *s = userdata; | |
26687bf8 OS |
1557 | |
1558 | assert(s); | |
1559 | ||
f9a810be | 1560 | server_sync(s); |
26687bf8 OS |
1561 | return 0; |
1562 | } | |
1563 | ||
d07f7b9e | 1564 | int server_schedule_sync(Server *s, int priority) { |
26687bf8 OS |
1565 | int r; |
1566 | ||
26687bf8 OS |
1567 | assert(s); |
1568 | ||
d07f7b9e LP |
1569 | if (priority <= LOG_CRIT) { |
1570 | /* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */ | |
1571 | server_sync(s); | |
1572 | return 0; | |
1573 | } | |
1574 | ||
26687bf8 OS |
1575 | if (s->sync_scheduled) |
1576 | return 0; | |
1577 | ||
f9a810be LP |
1578 | if (s->sync_interval_usec > 0) { |
1579 | usec_t when; | |
ca267016 | 1580 | |
6a0f1f6d | 1581 | r = sd_event_now(s->event, CLOCK_MONOTONIC, &when); |
f9a810be LP |
1582 | if (r < 0) |
1583 | return r; | |
26687bf8 | 1584 | |
f9a810be LP |
1585 | when += s->sync_interval_usec; |
1586 | ||
1587 | if (!s->sync_event_source) { | |
6a0f1f6d LP |
1588 | r = sd_event_add_time( |
1589 | s->event, | |
1590 | &s->sync_event_source, | |
1591 | CLOCK_MONOTONIC, | |
1592 | when, 0, | |
1593 | server_dispatch_sync, s); | |
f9a810be LP |
1594 | if (r < 0) |
1595 | return r; | |
1596 | ||
1597 | r = sd_event_source_set_priority(s->sync_event_source, SD_EVENT_PRIORITY_IMPORTANT); | |
1598 | } else { | |
1599 | r = sd_event_source_set_time(s->sync_event_source, when); | |
1600 | if (r < 0) | |
1601 | return r; | |
1602 | ||
1603 | r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_ONESHOT); | |
1604 | } | |
26687bf8 | 1605 | if (r < 0) |
f9a810be | 1606 | return r; |
26687bf8 | 1607 | |
f9a810be LP |
1608 | s->sync_scheduled = true; |
1609 | } | |
26687bf8 OS |
1610 | |
1611 | return 0; | |
1612 | } | |
1613 | ||
0c24bb23 LP |
1614 | static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1615 | Server *s = userdata; | |
1616 | ||
1617 | assert(s); | |
1618 | ||
1619 | server_cache_hostname(s); | |
1620 | return 0; | |
1621 | } | |
1622 | ||
1623 | static int server_open_hostname(Server *s) { | |
1624 | int r; | |
1625 | ||
1626 | assert(s); | |
1627 | ||
1628 | s->hostname_fd = open("/proc/sys/kernel/hostname", O_RDONLY|O_CLOEXEC|O_NDELAY|O_NOCTTY); | |
4a62c710 MS |
1629 | if (s->hostname_fd < 0) |
1630 | return log_error_errno(errno, "Failed to open /proc/sys/kernel/hostname: %m"); | |
0c24bb23 | 1631 | |
151b9b96 | 1632 | r = sd_event_add_io(s->event, &s->hostname_event_source, s->hostname_fd, 0, dispatch_hostname_change, s); |
0c24bb23 | 1633 | if (r < 0) { |
28def94c DR |
1634 | /* kernels prior to 3.2 don't support polling this file. Ignore |
1635 | * the failure. */ | |
1636 | if (r == -EPERM) { | |
e53fc357 | 1637 | log_warning_errno(r, "Failed to register hostname fd in event loop, ignoring: %m"); |
03e334a1 | 1638 | s->hostname_fd = safe_close(s->hostname_fd); |
28def94c DR |
1639 | return 0; |
1640 | } | |
1641 | ||
23bbb0de | 1642 | return log_error_errno(r, "Failed to register hostname fd in event loop: %m"); |
0c24bb23 LP |
1643 | } |
1644 | ||
1645 | r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10); | |
23bbb0de MS |
1646 | if (r < 0) |
1647 | return log_error_errno(r, "Failed to adjust priority of host name event source: %m"); | |
0c24bb23 LP |
1648 | |
1649 | return 0; | |
1650 | } | |
1651 | ||
e22aa3d3 LP |
1652 | static int dispatch_notify_event(sd_event_source *es, int fd, uint32_t revents, void *userdata) { |
1653 | Server *s = userdata; | |
1654 | int r; | |
1655 | ||
1656 | assert(s); | |
1657 | assert(s->notify_event_source == es); | |
1658 | assert(s->notify_fd == fd); | |
1659 | ||
e22aa3d3 | 1660 | /* The $NOTIFY_SOCKET is writable again, now send exactly one |
dd835265 | 1661 | * message on it. Either it's the watchdog event, the initial |
119e9655 LP |
1662 | * READY=1 event or an stdout stream event. If there's nothing |
1663 | * to write anymore, turn our event source off. The next time | |
1664 | * there's something to send it will be turned on again. */ | |
e22aa3d3 LP |
1665 | |
1666 | if (!s->sent_notify_ready) { | |
1667 | static const char p[] = | |
1668 | "READY=1\n" | |
1669 | "STATUS=Processing requests..."; | |
1670 | ssize_t l; | |
1671 | ||
1672 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1673 | if (l < 0) { | |
1674 | if (errno == EAGAIN) | |
1675 | return 0; | |
1676 | ||
1677 | return log_error_errno(errno, "Failed to send READY=1 notification message: %m"); | |
1678 | } | |
1679 | ||
1680 | s->sent_notify_ready = true; | |
1681 | log_debug("Sent READY=1 notification."); | |
1682 | ||
119e9655 LP |
1683 | } else if (s->send_watchdog) { |
1684 | ||
1685 | static const char p[] = | |
1686 | "WATCHDOG=1"; | |
1687 | ||
1688 | ssize_t l; | |
1689 | ||
1690 | l = send(s->notify_fd, p, strlen(p), MSG_DONTWAIT); | |
1691 | if (l < 0) { | |
1692 | if (errno == EAGAIN) | |
1693 | return 0; | |
1694 | ||
1695 | return log_error_errno(errno, "Failed to send WATCHDOG=1 notification message: %m"); | |
1696 | } | |
1697 | ||
1698 | s->send_watchdog = false; | |
1699 | log_debug("Sent WATCHDOG=1 notification."); | |
1700 | ||
e22aa3d3 LP |
1701 | } else if (s->stdout_streams_notify_queue) |
1702 | /* Dispatch one stream notification event */ | |
1703 | stdout_stream_send_notify(s->stdout_streams_notify_queue); | |
1704 | ||
61233823 | 1705 | /* Leave us enabled if there's still more to do. */ |
119e9655 | 1706 | if (s->send_watchdog || s->stdout_streams_notify_queue) |
e22aa3d3 LP |
1707 | return 0; |
1708 | ||
1709 | /* There was nothing to do anymore, let's turn ourselves off. */ | |
1710 | r = sd_event_source_set_enabled(es, SD_EVENT_OFF); | |
1711 | if (r < 0) | |
1712 | return log_error_errno(r, "Failed to turn off notify event source: %m"); | |
1713 | ||
1714 | return 0; | |
1715 | } | |
1716 | ||
119e9655 LP |
1717 | static int dispatch_watchdog(sd_event_source *es, uint64_t usec, void *userdata) { |
1718 | Server *s = userdata; | |
1719 | int r; | |
1720 | ||
1721 | assert(s); | |
1722 | ||
1723 | s->send_watchdog = true; | |
1724 | ||
1725 | r = sd_event_source_set_enabled(s->notify_event_source, SD_EVENT_ON); | |
1726 | if (r < 0) | |
1727 | log_warning_errno(r, "Failed to turn on notify event source: %m"); | |
1728 | ||
1729 | r = sd_event_source_set_time(s->watchdog_event_source, usec + s->watchdog_usec / 2); | |
1730 | if (r < 0) | |
1731 | return log_error_errno(r, "Failed to restart watchdog event source: %m"); | |
1732 | ||
1733 | r = sd_event_source_set_enabled(s->watchdog_event_source, SD_EVENT_ON); | |
1734 | if (r < 0) | |
1735 | return log_error_errno(r, "Failed to enable watchdog event source: %m"); | |
1736 | ||
1737 | return 0; | |
1738 | } | |
1739 | ||
e22aa3d3 LP |
1740 | static int server_connect_notify(Server *s) { |
1741 | union sockaddr_union sa = { | |
1742 | .un.sun_family = AF_UNIX, | |
1743 | }; | |
1744 | const char *e; | |
1745 | int r; | |
1746 | ||
1747 | assert(s); | |
1748 | assert(s->notify_fd < 0); | |
1749 | assert(!s->notify_event_source); | |
1750 | ||
1751 | /* | |
1752 | So here's the problem: we'd like to send notification | |
1753 | messages to PID 1, but we cannot do that via sd_notify(), | |
1754 | since that's synchronous, and we might end up blocking on | |
1755 | it. Specifically: given that PID 1 might block on | |
1756 | dbus-daemon during IPC, and dbus-daemon is logging to us, | |
1757 | and might hence block on us, we might end up in a deadlock | |
ccddd104 | 1758 | if we block on sending PID 1 notification messages — by |
e22aa3d3 LP |
1759 | generating a full blocking circle. To avoid this, let's |
1760 | create a non-blocking socket, and connect it to the | |
1761 | notification socket, and then wait for POLLOUT before we | |
1762 | send anything. This should efficiently avoid any deadlocks, | |
1763 | as we'll never block on PID 1, hence PID 1 can safely block | |
1764 | on dbus-daemon which can safely block on us again. | |
1765 | ||
1766 | Don't think that this issue is real? It is, see: | |
1767 | https://github.com/systemd/systemd/issues/1505 | |
1768 | */ | |
1769 | ||
1770 | e = getenv("NOTIFY_SOCKET"); | |
1771 | if (!e) | |
1772 | return 0; | |
1773 | ||
1774 | if ((e[0] != '@' && e[0] != '/') || e[1] == 0) { | |
1775 | log_error("NOTIFY_SOCKET set to an invalid value: %s", e); | |
1776 | return -EINVAL; | |
1777 | } | |
1778 | ||
1779 | if (strlen(e) > sizeof(sa.un.sun_path)) { | |
1780 | log_error("NOTIFY_SOCKET path too long: %s", e); | |
1781 | return -EINVAL; | |
1782 | } | |
1783 | ||
1784 | s->notify_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); | |
1785 | if (s->notify_fd < 0) | |
1786 | return log_error_errno(errno, "Failed to create notify socket: %m"); | |
1787 | ||
1788 | (void) fd_inc_sndbuf(s->notify_fd, NOTIFY_SNDBUF_SIZE); | |
1789 | ||
1790 | strncpy(sa.un.sun_path, e, sizeof(sa.un.sun_path)); | |
1791 | if (sa.un.sun_path[0] == '@') | |
1792 | sa.un.sun_path[0] = 0; | |
1793 | ||
fc2fffe7 | 1794 | r = connect(s->notify_fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); |
e22aa3d3 LP |
1795 | if (r < 0) |
1796 | return log_error_errno(errno, "Failed to connect to notify socket: %m"); | |
1797 | ||
1798 | r = sd_event_add_io(s->event, &s->notify_event_source, s->notify_fd, EPOLLOUT, dispatch_notify_event, s); | |
1799 | if (r < 0) | |
1800 | return log_error_errno(r, "Failed to watch notification socket: %m"); | |
1801 | ||
119e9655 LP |
1802 | if (sd_watchdog_enabled(false, &s->watchdog_usec) > 0) { |
1803 | s->send_watchdog = true; | |
1804 | ||
4de2402b | 1805 | 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 |
1806 | if (r < 0) |
1807 | return log_error_errno(r, "Failed to add watchdog time event: %m"); | |
1808 | } | |
1809 | ||
e22aa3d3 LP |
1810 | /* This should fire pretty soon, which we'll use to send the |
1811 | * READY=1 event. */ | |
1812 | ||
1813 | return 0; | |
1814 | } | |
1815 | ||
d025f1e4 | 1816 | int server_init(Server *s) { |
13790add | 1817 | _cleanup_fdset_free_ FDSet *fds = NULL; |
d025f1e4 | 1818 | int n, r, fd; |
7d18d348 | 1819 | bool no_sockets; |
d025f1e4 ZJS |
1820 | |
1821 | assert(s); | |
1822 | ||
1823 | zero(*s); | |
e22aa3d3 | 1824 | 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 |
1825 | s->compress = true; |
1826 | s->seal = true; | |
1827 | ||
119e9655 LP |
1828 | s->watchdog_usec = USEC_INFINITY; |
1829 | ||
26687bf8 OS |
1830 | s->sync_interval_usec = DEFAULT_SYNC_INTERVAL_USEC; |
1831 | s->sync_scheduled = false; | |
1832 | ||
d025f1e4 ZJS |
1833 | s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL; |
1834 | s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST; | |
1835 | ||
40b71e89 | 1836 | s->forward_to_wall = true; |
d025f1e4 | 1837 | |
e150e820 MB |
1838 | s->max_file_usec = DEFAULT_MAX_FILE_USEC; |
1839 | ||
d025f1e4 ZJS |
1840 | s->max_level_store = LOG_DEBUG; |
1841 | s->max_level_syslog = LOG_DEBUG; | |
1842 | s->max_level_kmsg = LOG_NOTICE; | |
1843 | s->max_level_console = LOG_INFO; | |
40b71e89 | 1844 | s->max_level_wall = LOG_EMERG; |
d025f1e4 | 1845 | |
8580d1f7 LP |
1846 | journal_reset_metrics(&s->system_metrics); |
1847 | journal_reset_metrics(&s->runtime_metrics); | |
d025f1e4 ZJS |
1848 | |
1849 | server_parse_config_file(s); | |
1850 | server_parse_proc_cmdline(s); | |
8580d1f7 | 1851 | |
d288f79f | 1852 | if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { |
b1389b0d ZJS |
1853 | log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0", |
1854 | s->rate_limit_interval, s->rate_limit_burst); | |
d288f79f ZJS |
1855 | s->rate_limit_interval = s->rate_limit_burst = 0; |
1856 | } | |
d025f1e4 | 1857 | |
8580d1f7 | 1858 | (void) mkdir_p("/run/systemd/journal", 0755); |
d025f1e4 | 1859 | |
43cf8388 | 1860 | s->user_journals = ordered_hashmap_new(NULL); |
d025f1e4 ZJS |
1861 | if (!s->user_journals) |
1862 | return log_oom(); | |
1863 | ||
1864 | s->mmap = mmap_cache_new(); | |
1865 | if (!s->mmap) | |
1866 | return log_oom(); | |
1867 | ||
b58c888f VC |
1868 | s->deferred_closes = set_new(NULL); |
1869 | if (!s->deferred_closes) | |
1870 | return log_oom(); | |
1871 | ||
f9a810be | 1872 | r = sd_event_default(&s->event); |
23bbb0de MS |
1873 | if (r < 0) |
1874 | return log_error_errno(r, "Failed to create event loop: %m"); | |
d025f1e4 ZJS |
1875 | |
1876 | n = sd_listen_fds(true); | |
23bbb0de MS |
1877 | if (n < 0) |
1878 | return log_error_errno(n, "Failed to read listening file descriptors from environment: %m"); | |
d025f1e4 ZJS |
1879 | |
1880 | for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { | |
1881 | ||
1882 | if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) { | |
1883 | ||
1884 | if (s->native_fd >= 0) { | |
1885 | log_error("Too many native sockets passed."); | |
1886 | return -EINVAL; | |
1887 | } | |
1888 | ||
1889 | s->native_fd = fd; | |
1890 | ||
1891 | } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) { | |
1892 | ||
1893 | if (s->stdout_fd >= 0) { | |
1894 | log_error("Too many stdout sockets passed."); | |
1895 | return -EINVAL; | |
1896 | } | |
1897 | ||
1898 | s->stdout_fd = fd; | |
1899 | ||
03ee5c38 LP |
1900 | } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0 || |
1901 | sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/dev-log", 0) > 0) { | |
d025f1e4 ZJS |
1902 | |
1903 | if (s->syslog_fd >= 0) { | |
1904 | log_error("Too many /dev/log sockets passed."); | |
1905 | return -EINVAL; | |
1906 | } | |
1907 | ||
1908 | s->syslog_fd = fd; | |
1909 | ||
875c2e22 LP |
1910 | } else if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1) > 0) { |
1911 | ||
1912 | if (s->audit_fd >= 0) { | |
1913 | log_error("Too many audit sockets passed."); | |
1914 | return -EINVAL; | |
1915 | } | |
1916 | ||
1917 | s->audit_fd = fd; | |
1918 | ||
4ec3cd73 | 1919 | } else { |
4ec3cd73 | 1920 | |
13790add LP |
1921 | if (!fds) { |
1922 | fds = fdset_new(); | |
1923 | if (!fds) | |
1924 | return log_oom(); | |
1925 | } | |
4ec3cd73 | 1926 | |
13790add LP |
1927 | r = fdset_put(fds, fd); |
1928 | if (r < 0) | |
1929 | return log_oom(); | |
4ec3cd73 | 1930 | } |
d025f1e4 ZJS |
1931 | } |
1932 | ||
15d91bff ZJS |
1933 | /* Try to restore streams, but don't bother if this fails */ |
1934 | (void) server_restore_streams(s, fds); | |
d025f1e4 | 1935 | |
13790add LP |
1936 | if (fdset_size(fds) > 0) { |
1937 | log_warning("%u unknown file descriptors passed, closing.", fdset_size(fds)); | |
1938 | fds = fdset_free(fds); | |
1939 | } | |
1940 | ||
7d18d348 ZJS |
1941 | no_sockets = s->native_fd < 0 && s->stdout_fd < 0 && s->syslog_fd < 0 && s->audit_fd < 0; |
1942 | ||
1943 | /* always open stdout, syslog, native, and kmsg sockets */ | |
37b7affe ZJS |
1944 | |
1945 | /* systemd-journald.socket: /run/systemd/journal/stdout */ | |
15d91bff ZJS |
1946 | r = server_open_stdout_socket(s); |
1947 | if (r < 0) | |
1948 | return r; | |
1949 | ||
37b7affe | 1950 | /* systemd-journald-dev-log.socket: /run/systemd/journal/dev-log */ |
13790add | 1951 | r = server_open_syslog_socket(s); |
d025f1e4 ZJS |
1952 | if (r < 0) |
1953 | return r; | |
1954 | ||
37b7affe | 1955 | /* systemd-journald.socket: /run/systemd/journal/socket */ |
13790add | 1956 | r = server_open_native_socket(s); |
d025f1e4 ZJS |
1957 | if (r < 0) |
1958 | return r; | |
1959 | ||
37b7affe | 1960 | /* /dev/ksmg */ |
d025f1e4 ZJS |
1961 | r = server_open_dev_kmsg(s); |
1962 | if (r < 0) | |
1963 | return r; | |
1964 | ||
7d18d348 ZJS |
1965 | /* Unless we got *some* sockets and not audit, open audit socket */ |
1966 | if (s->audit_fd >= 0 || no_sockets) { | |
1967 | r = server_open_audit(s); | |
1968 | if (r < 0) | |
1969 | return r; | |
1970 | } | |
875c2e22 | 1971 | |
d025f1e4 ZJS |
1972 | r = server_open_kernel_seqnum(s); |
1973 | if (r < 0) | |
1974 | return r; | |
1975 | ||
0c24bb23 LP |
1976 | r = server_open_hostname(s); |
1977 | if (r < 0) | |
1978 | return r; | |
1979 | ||
f9a810be | 1980 | r = setup_signals(s); |
d025f1e4 ZJS |
1981 | if (r < 0) |
1982 | return r; | |
1983 | ||
1984 | s->udev = udev_new(); | |
1985 | if (!s->udev) | |
1986 | return -ENOMEM; | |
1987 | ||
f9a810be | 1988 | s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst); |
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 | ||
e22aa3d3 LP |
2000 | (void) server_connect_notify(s); |
2001 | ||
804ae586 | 2002 | return system_journal_open(s, false); |
d025f1e4 ZJS |
2003 | } |
2004 | ||
2005 | void server_maybe_append_tags(Server *s) { | |
2006 | #ifdef HAVE_GCRYPT | |
2007 | JournalFile *f; | |
2008 | Iterator i; | |
2009 | usec_t n; | |
2010 | ||
2011 | n = now(CLOCK_REALTIME); | |
2012 | ||
2013 | if (s->system_journal) | |
2014 | journal_file_maybe_append_tag(s->system_journal, n); | |
2015 | ||
43cf8388 | 2016 | ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) |
d025f1e4 ZJS |
2017 | journal_file_maybe_append_tag(f, n); |
2018 | #endif | |
2019 | } | |
2020 | ||
2021 | void server_done(Server *s) { | |
2022 | JournalFile *f; | |
2023 | assert(s); | |
2024 | ||
b58c888f VC |
2025 | if (s->deferred_closes) { |
2026 | journal_file_close_set(s->deferred_closes); | |
2027 | set_free(s->deferred_closes); | |
2028 | } | |
2029 | ||
d025f1e4 ZJS |
2030 | while (s->stdout_streams) |
2031 | stdout_stream_free(s->stdout_streams); | |
2032 | ||
2033 | if (s->system_journal) | |
69a3a6fd | 2034 | (void) journal_file_close(s->system_journal); |
d025f1e4 ZJS |
2035 | |
2036 | if (s->runtime_journal) | |
69a3a6fd | 2037 | (void) journal_file_close(s->runtime_journal); |
d025f1e4 | 2038 | |
43cf8388 | 2039 | while ((f = ordered_hashmap_steal_first(s->user_journals))) |
69a3a6fd | 2040 | (void) journal_file_close(f); |
d025f1e4 | 2041 | |
43cf8388 | 2042 | ordered_hashmap_free(s->user_journals); |
d025f1e4 | 2043 | |
f9a810be LP |
2044 | sd_event_source_unref(s->syslog_event_source); |
2045 | sd_event_source_unref(s->native_event_source); | |
2046 | sd_event_source_unref(s->stdout_event_source); | |
2047 | sd_event_source_unref(s->dev_kmsg_event_source); | |
875c2e22 | 2048 | sd_event_source_unref(s->audit_event_source); |
f9a810be LP |
2049 | sd_event_source_unref(s->sync_event_source); |
2050 | sd_event_source_unref(s->sigusr1_event_source); | |
2051 | sd_event_source_unref(s->sigusr2_event_source); | |
2052 | sd_event_source_unref(s->sigterm_event_source); | |
2053 | sd_event_source_unref(s->sigint_event_source); | |
94b65516 | 2054 | sd_event_source_unref(s->sigrtmin1_event_source); |
0c24bb23 | 2055 | sd_event_source_unref(s->hostname_event_source); |
e22aa3d3 | 2056 | sd_event_source_unref(s->notify_event_source); |
119e9655 | 2057 | sd_event_source_unref(s->watchdog_event_source); |
f9a810be | 2058 | sd_event_unref(s->event); |
d025f1e4 | 2059 | |
03e334a1 LP |
2060 | safe_close(s->syslog_fd); |
2061 | safe_close(s->native_fd); | |
2062 | safe_close(s->stdout_fd); | |
2063 | safe_close(s->dev_kmsg_fd); | |
875c2e22 | 2064 | safe_close(s->audit_fd); |
03e334a1 | 2065 | safe_close(s->hostname_fd); |
e22aa3d3 | 2066 | safe_close(s->notify_fd); |
0c24bb23 | 2067 | |
d025f1e4 ZJS |
2068 | if (s->rate_limit) |
2069 | journal_rate_limit_free(s->rate_limit); | |
2070 | ||
2071 | if (s->kernel_seqnum) | |
2072 | munmap(s->kernel_seqnum, sizeof(uint64_t)); | |
2073 | ||
2074 | free(s->buffer); | |
2075 | free(s->tty_path); | |
e9174f29 | 2076 | free(s->cgroup_root); |
99d0966e | 2077 | free(s->hostname_field); |
d025f1e4 ZJS |
2078 | |
2079 | if (s->mmap) | |
2080 | mmap_cache_unref(s->mmap); | |
2081 | ||
3e044c49 | 2082 | udev_unref(s->udev); |
d025f1e4 | 2083 | } |
8580d1f7 LP |
2084 | |
2085 | static const char* const storage_table[_STORAGE_MAX] = { | |
2086 | [STORAGE_AUTO] = "auto", | |
2087 | [STORAGE_VOLATILE] = "volatile", | |
2088 | [STORAGE_PERSISTENT] = "persistent", | |
2089 | [STORAGE_NONE] = "none" | |
2090 | }; | |
2091 | ||
2092 | DEFINE_STRING_TABLE_LOOKUP(storage, Storage); | |
2093 | DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); | |
2094 | ||
2095 | static const char* const split_mode_table[_SPLIT_MAX] = { | |
2096 | [SPLIT_LOGIN] = "login", | |
2097 | [SPLIT_UID] = "uid", | |
2098 | [SPLIT_NONE] = "none", | |
2099 | }; | |
2100 | ||
2101 | DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode); | |
2102 | DEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting"); |