]>
Commit | Line | Data |
---|---|---|
d025f1e4 ZJS |
1 | /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ |
2 | ||
3 | /*** | |
4 | This file is part of systemd. | |
5 | ||
6 | Copyright 2011 Lennart Poettering | |
7 | ||
8 | systemd is free software; you can redistribute it and/or modify it | |
9 | under the terms of the GNU Lesser General Public License as published by | |
10 | the Free Software Foundation; either version 2.1 of the License, or | |
11 | (at your option) any later version. | |
12 | ||
13 | systemd is distributed in the hope that it will be useful, but | |
14 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | Lesser General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU Lesser General Public License | |
19 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
20 | ***/ | |
21 | ||
22 | #include <sys/signalfd.h> | |
23 | #include <sys/ioctl.h> | |
24 | #include <linux/sockios.h> | |
25 | #include <sys/statvfs.h> | |
26 | #include <sys/mman.h> | |
26687bf8 | 27 | #include <sys/timerfd.h> |
d025f1e4 ZJS |
28 | |
29 | #include <libudev.h> | |
30 | #include <systemd/sd-journal.h> | |
31 | #include <systemd/sd-messages.h> | |
32 | #include <systemd/sd-daemon.h> | |
33 | ||
a5c32cff | 34 | #include "fileio.h" |
d025f1e4 ZJS |
35 | #include "mkdir.h" |
36 | #include "hashmap.h" | |
37 | #include "journal-file.h" | |
38 | #include "socket-util.h" | |
39 | #include "cgroup-util.h" | |
40 | #include "list.h" | |
41 | #include "virt.h" | |
42 | #include "missing.h" | |
43 | #include "conf-parser.h" | |
44 | #include "journal-internal.h" | |
45 | #include "journal-vacuum.h" | |
46 | #include "journal-authenticate.h" | |
47 | #include "journald-server.h" | |
48 | #include "journald-rate-limit.h" | |
49 | #include "journald-kmsg.h" | |
50 | #include "journald-syslog.h" | |
51 | #include "journald-stream.h" | |
52 | #include "journald-console.h" | |
53 | #include "journald-native.h" | |
54 | ||
55 | #ifdef HAVE_ACL | |
56 | #include <sys/acl.h> | |
57 | #include <acl/libacl.h> | |
58 | #include "acl-util.h" | |
59 | #endif | |
60 | ||
61 | #ifdef HAVE_SELINUX | |
62 | #include <selinux/selinux.h> | |
63 | #endif | |
64 | ||
65 | #define USER_JOURNALS_MAX 1024 | |
66 | ||
26687bf8 | 67 | #define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) |
d025f1e4 ZJS |
68 | #define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC) |
69 | #define DEFAULT_RATE_LIMIT_BURST 200 | |
70 | ||
71 | #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC) | |
72 | ||
73 | static const char* const storage_table[] = { | |
74 | [STORAGE_AUTO] = "auto", | |
75 | [STORAGE_VOLATILE] = "volatile", | |
76 | [STORAGE_PERSISTENT] = "persistent", | |
77 | [STORAGE_NONE] = "none" | |
78 | }; | |
79 | ||
80 | DEFINE_STRING_TABLE_LOOKUP(storage, Storage); | |
81 | DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); | |
82 | ||
83 | static const char* const split_mode_table[] = { | |
84 | [SPLIT_NONE] = "none", | |
85 | [SPLIT_UID] = "uid", | |
86 | [SPLIT_LOGIN] = "login" | |
87 | }; | |
88 | ||
89 | DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode); | |
90 | DEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting"); | |
91 | ||
92 | static uint64_t available_space(Server *s) { | |
db91ea32 | 93 | char ids[33]; |
7fd1b19b | 94 | _cleanup_free_ char *p = NULL; |
d025f1e4 ZJS |
95 | const char *f; |
96 | sd_id128_t machine; | |
97 | struct statvfs ss; | |
98 | uint64_t sum = 0, avail = 0, ss_avail = 0; | |
99 | int r; | |
7fd1b19b | 100 | _cleanup_closedir_ DIR *d = NULL; |
d025f1e4 ZJS |
101 | usec_t ts; |
102 | JournalMetrics *m; | |
103 | ||
104 | ts = now(CLOCK_MONOTONIC); | |
105 | ||
106 | if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts) | |
107 | return s->cached_available_space; | |
108 | ||
109 | r = sd_id128_get_machine(&machine); | |
110 | if (r < 0) | |
111 | return 0; | |
112 | ||
113 | if (s->system_journal) { | |
114 | f = "/var/log/journal/"; | |
115 | m = &s->system_metrics; | |
116 | } else { | |
117 | f = "/run/log/journal/"; | |
118 | m = &s->runtime_metrics; | |
119 | } | |
120 | ||
121 | assert(m); | |
122 | ||
123 | p = strappend(f, sd_id128_to_string(machine, ids)); | |
124 | if (!p) | |
125 | return 0; | |
126 | ||
127 | d = opendir(p); | |
d025f1e4 ZJS |
128 | if (!d) |
129 | return 0; | |
130 | ||
131 | if (fstatvfs(dirfd(d), &ss) < 0) | |
db91ea32 | 132 | return 0; |
d025f1e4 ZJS |
133 | |
134 | for (;;) { | |
135 | struct stat st; | |
136 | struct dirent *de; | |
137 | union dirent_storage buf; | |
138 | ||
139 | r = readdir_r(d, &buf.de, &de); | |
140 | if (r != 0) | |
141 | break; | |
142 | ||
143 | if (!de) | |
144 | break; | |
145 | ||
146 | if (!endswith(de->d_name, ".journal") && | |
147 | !endswith(de->d_name, ".journal~")) | |
148 | continue; | |
149 | ||
150 | if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) | |
151 | continue; | |
152 | ||
153 | if (!S_ISREG(st.st_mode)) | |
154 | continue; | |
155 | ||
156 | sum += (uint64_t) st.st_blocks * 512UL; | |
157 | } | |
158 | ||
159 | avail = sum >= m->max_use ? 0 : m->max_use - sum; | |
160 | ||
161 | ss_avail = ss.f_bsize * ss.f_bavail; | |
162 | ||
163 | ss_avail = ss_avail < m->keep_free ? 0 : ss_avail - m->keep_free; | |
164 | ||
165 | if (ss_avail < avail) | |
166 | avail = ss_avail; | |
167 | ||
168 | s->cached_available_space = avail; | |
169 | s->cached_available_space_timestamp = ts; | |
170 | ||
d025f1e4 ZJS |
171 | return avail; |
172 | } | |
173 | ||
174 | static void server_read_file_gid(Server *s) { | |
a24c64f0 | 175 | const char *g = "systemd-journal"; |
d025f1e4 ZJS |
176 | int r; |
177 | ||
178 | assert(s); | |
179 | ||
180 | if (s->file_gid_valid) | |
181 | return; | |
182 | ||
a24c64f0 | 183 | r = get_group_creds(&g, &s->file_gid); |
d025f1e4 | 184 | if (r < 0) |
a24c64f0 | 185 | log_warning("Failed to resolve '%s' group: %s", g, strerror(-r)); |
d025f1e4 ZJS |
186 | |
187 | /* if we couldn't read the gid, then it will be 0, but that's | |
188 | * fine and we shouldn't try to resolve the group again, so | |
189 | * let's just pretend it worked right-away. */ | |
190 | s->file_gid_valid = true; | |
191 | } | |
192 | ||
193 | void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { | |
194 | int r; | |
195 | #ifdef HAVE_ACL | |
196 | acl_t acl; | |
197 | acl_entry_t entry; | |
198 | acl_permset_t permset; | |
199 | #endif | |
200 | ||
201 | assert(f); | |
202 | ||
203 | server_read_file_gid(s); | |
204 | ||
205 | r = fchmod_and_fchown(f->fd, 0640, 0, s->file_gid); | |
206 | if (r < 0) | |
207 | log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r)); | |
208 | ||
209 | #ifdef HAVE_ACL | |
210 | if (uid <= 0) | |
211 | return; | |
212 | ||
213 | acl = acl_get_fd(f->fd); | |
214 | if (!acl) { | |
215 | log_warning("Failed to read ACL on %s, ignoring: %m", f->path); | |
216 | return; | |
217 | } | |
218 | ||
219 | r = acl_find_uid(acl, uid, &entry); | |
220 | if (r <= 0) { | |
221 | ||
222 | if (acl_create_entry(&acl, &entry) < 0 || | |
223 | acl_set_tag_type(entry, ACL_USER) < 0 || | |
224 | acl_set_qualifier(entry, &uid) < 0) { | |
225 | log_warning("Failed to patch ACL on %s, ignoring: %m", f->path); | |
226 | goto finish; | |
227 | } | |
228 | } | |
229 | ||
230 | if (acl_get_permset(entry, &permset) < 0 || | |
231 | acl_add_perm(permset, ACL_READ) < 0 || | |
232 | acl_calc_mask(&acl) < 0) { | |
233 | log_warning("Failed to patch ACL on %s, ignoring: %m", f->path); | |
234 | goto finish; | |
235 | } | |
236 | ||
237 | if (acl_set_fd(f->fd, acl) < 0) | |
238 | log_warning("Failed to set ACL on %s, ignoring: %m", f->path); | |
239 | ||
240 | finish: | |
241 | acl_free(acl); | |
242 | #endif | |
243 | } | |
244 | ||
245 | static JournalFile* find_journal(Server *s, uid_t uid) { | |
246 | char *p; | |
247 | int r; | |
248 | JournalFile *f; | |
249 | sd_id128_t machine; | |
250 | ||
251 | assert(s); | |
252 | ||
253 | /* We split up user logs only on /var, not on /run. If the | |
254 | * runtime file is open, we write to it exclusively, in order | |
255 | * to guarantee proper order as soon as we flush /run to | |
256 | * /var and close the runtime file. */ | |
257 | ||
258 | if (s->runtime_journal) | |
259 | return s->runtime_journal; | |
260 | ||
261 | if (uid <= 0) | |
262 | return s->system_journal; | |
263 | ||
264 | r = sd_id128_get_machine(&machine); | |
265 | if (r < 0) | |
266 | return s->system_journal; | |
267 | ||
268 | f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid)); | |
269 | if (f) | |
270 | return f; | |
271 | ||
272 | if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal", | |
273 | SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0) | |
274 | return s->system_journal; | |
275 | ||
276 | while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { | |
277 | /* Too many open? Then let's close one */ | |
278 | f = hashmap_steal_first(s->user_journals); | |
279 | assert(f); | |
280 | journal_file_close(f); | |
281 | } | |
282 | ||
283 | r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, s->system_journal, &f); | |
284 | free(p); | |
285 | ||
286 | if (r < 0) | |
287 | return s->system_journal; | |
288 | ||
289 | server_fix_perms(s, f, uid); | |
290 | ||
291 | r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f); | |
292 | if (r < 0) { | |
293 | journal_file_close(f); | |
294 | return s->system_journal; | |
295 | } | |
296 | ||
297 | return f; | |
298 | } | |
299 | ||
300 | void server_rotate(Server *s) { | |
301 | JournalFile *f; | |
302 | void *k; | |
303 | Iterator i; | |
304 | int r; | |
305 | ||
306 | log_debug("Rotating..."); | |
307 | ||
308 | if (s->runtime_journal) { | |
309 | r = journal_file_rotate(&s->runtime_journal, s->compress, false); | |
310 | if (r < 0) | |
311 | if (s->runtime_journal) | |
312 | log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r)); | |
313 | else | |
314 | log_error("Failed to create new runtime journal: %s", strerror(-r)); | |
315 | else | |
316 | server_fix_perms(s, s->runtime_journal, 0); | |
317 | } | |
318 | ||
319 | if (s->system_journal) { | |
320 | r = journal_file_rotate(&s->system_journal, s->compress, s->seal); | |
321 | if (r < 0) | |
322 | if (s->system_journal) | |
323 | log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r)); | |
324 | else | |
325 | log_error("Failed to create new system journal: %s", strerror(-r)); | |
326 | ||
327 | else | |
328 | server_fix_perms(s, s->system_journal, 0); | |
329 | } | |
330 | ||
331 | HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { | |
332 | r = journal_file_rotate(&f, s->compress, s->seal); | |
333 | if (r < 0) | |
7d73c134 | 334 | if (f) |
d025f1e4 ZJS |
335 | log_error("Failed to rotate %s: %s", f->path, strerror(-r)); |
336 | else | |
337 | log_error("Failed to create user journal: %s", strerror(-r)); | |
338 | else { | |
339 | hashmap_replace(s->user_journals, k, f); | |
340 | server_fix_perms(s, f, PTR_TO_UINT32(k)); | |
341 | } | |
342 | } | |
343 | } | |
344 | ||
26687bf8 OS |
345 | void server_sync(Server *s) { |
346 | JournalFile *f; | |
347 | void *k; | |
348 | Iterator i; | |
349 | int r; | |
350 | ||
351 | static const struct itimerspec sync_timer_disable = {}; | |
352 | ||
353 | if (s->system_journal) { | |
354 | r = journal_file_set_offline(s->system_journal); | |
355 | if (r < 0) | |
356 | log_error("Failed to sync system journal: %s", strerror(-r)); | |
357 | } | |
358 | ||
359 | HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { | |
360 | r = journal_file_set_offline(f); | |
361 | if (r < 0) | |
362 | log_error("Failed to sync user journal: %s", strerror(-r)); | |
363 | } | |
364 | ||
365 | r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_disable, NULL); | |
366 | if (r < 0) | |
367 | log_error("Failed to disable max timer: %m"); | |
368 | ||
369 | s->sync_scheduled = false; | |
370 | } | |
371 | ||
d025f1e4 ZJS |
372 | void server_vacuum(Server *s) { |
373 | char *p; | |
374 | char ids[33]; | |
375 | sd_id128_t machine; | |
376 | int r; | |
377 | ||
378 | log_debug("Vacuuming..."); | |
379 | ||
380 | s->oldest_file_usec = 0; | |
381 | ||
382 | r = sd_id128_get_machine(&machine); | |
383 | if (r < 0) { | |
384 | log_error("Failed to get machine ID: %s", strerror(-r)); | |
385 | return; | |
386 | } | |
387 | ||
388 | sd_id128_to_string(machine, ids); | |
389 | ||
390 | if (s->system_journal) { | |
391 | p = strappend("/var/log/journal/", ids); | |
392 | if (!p) { | |
393 | log_oom(); | |
394 | return; | |
395 | } | |
396 | ||
397 | r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); | |
398 | if (r < 0 && r != -ENOENT) | |
399 | log_error("Failed to vacuum %s: %s", p, strerror(-r)); | |
400 | free(p); | |
401 | } | |
402 | ||
403 | if (s->runtime_journal) { | |
404 | p = strappend("/run/log/journal/", ids); | |
405 | if (!p) { | |
406 | log_oom(); | |
407 | return; | |
408 | } | |
409 | ||
410 | r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); | |
411 | if (r < 0 && r != -ENOENT) | |
412 | log_error("Failed to vacuum %s: %s", p, strerror(-r)); | |
413 | free(p); | |
414 | } | |
415 | ||
416 | s->cached_available_space_timestamp = 0; | |
417 | } | |
418 | ||
d025f1e4 ZJS |
419 | bool shall_try_append_again(JournalFile *f, int r) { |
420 | ||
421 | /* -E2BIG Hit configured limit | |
422 | -EFBIG Hit fs limit | |
423 | -EDQUOT Quota limit hit | |
424 | -ENOSPC Disk full | |
425 | -EHOSTDOWN Other machine | |
426 | -EBUSY Unclean shutdown | |
427 | -EPROTONOSUPPORT Unsupported feature | |
428 | -EBADMSG Corrupted | |
429 | -ENODATA Truncated | |
430 | -ESHUTDOWN Already archived */ | |
431 | ||
432 | if (r == -E2BIG || r == -EFBIG || r == -EDQUOT || r == -ENOSPC) | |
433 | log_debug("%s: Allocation limit reached, rotating.", f->path); | |
434 | else if (r == -EHOSTDOWN) | |
435 | log_info("%s: Journal file from other machine, rotating.", f->path); | |
436 | else if (r == -EBUSY) | |
437 | log_info("%s: Unclean shutdown, rotating.", f->path); | |
438 | else if (r == -EPROTONOSUPPORT) | |
439 | log_info("%s: Unsupported feature, rotating.", f->path); | |
440 | else if (r == -EBADMSG || r == -ENODATA || r == ESHUTDOWN) | |
441 | log_warning("%s: Journal file corrupted, rotating.", f->path); | |
442 | else | |
443 | return false; | |
444 | ||
445 | return true; | |
446 | } | |
447 | ||
448 | static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n) { | |
449 | JournalFile *f; | |
450 | bool vacuumed = false; | |
451 | int r; | |
452 | ||
453 | assert(s); | |
454 | assert(iovec); | |
455 | assert(n > 0); | |
456 | ||
457 | f = find_journal(s, uid); | |
458 | if (!f) | |
459 | return; | |
460 | ||
461 | if (journal_file_rotate_suggested(f, s->max_file_usec)) { | |
462 | log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path); | |
463 | server_rotate(s); | |
464 | server_vacuum(s); | |
465 | vacuumed = true; | |
466 | ||
467 | f = find_journal(s, uid); | |
468 | if (!f) | |
469 | return; | |
470 | } | |
471 | ||
472 | r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); | |
26687bf8 OS |
473 | if (r >= 0) { |
474 | server_schedule_sync(s); | |
d025f1e4 | 475 | return; |
26687bf8 | 476 | } |
d025f1e4 ZJS |
477 | |
478 | if (vacuumed || !shall_try_append_again(f, r)) { | |
479 | log_error("Failed to write entry, ignoring: %s", strerror(-r)); | |
480 | return; | |
481 | } | |
482 | ||
483 | server_rotate(s); | |
484 | server_vacuum(s); | |
485 | ||
486 | f = find_journal(s, uid); | |
487 | if (!f) | |
488 | return; | |
489 | ||
490 | log_debug("Retrying write."); | |
491 | r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); | |
492 | if (r < 0) | |
493 | log_error("Failed to write entry, ignoring: %s", strerror(-r)); | |
494 | } | |
495 | ||
496 | static void dispatch_message_real( | |
497 | Server *s, | |
498 | struct iovec *iovec, unsigned n, unsigned m, | |
499 | struct ucred *ucred, | |
500 | struct timeval *tv, | |
501 | const char *label, size_t label_len, | |
502 | const char *unit_id) { | |
503 | ||
ae018d9b LP |
504 | char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)], |
505 | uid[sizeof("_UID=") + DECIMAL_STR_MAX(uid_t)], | |
506 | gid[sizeof("_GID=") + DECIMAL_STR_MAX(gid_t)], | |
507 | owner_uid[sizeof("_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)], | |
d3789917 | 508 | source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)], |
adb435bb LP |
509 | boot_id[sizeof("_BOOT_ID=") + 32] = "_BOOT_ID=", |
510 | machine_id[sizeof("_MACHINE_ID=") + 32] = "_MACHINE_ID="; | |
ae018d9b | 511 | char *comm, *exe, *cmdline, *cgroup, *session, *unit, *selinux_context, *hostname; |
d025f1e4 ZJS |
512 | sd_id128_t id; |
513 | int r; | |
ae018d9b | 514 | char *t, *c; |
82499507 LP |
515 | uid_t realuid = 0, owner = 0, journal_uid; |
516 | bool owner_valid = false; | |
ae018d9b LP |
517 | #ifdef HAVE_AUDIT |
518 | char audit_session[sizeof("_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], | |
519 | audit_loginuid[sizeof("_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)]; | |
520 | ||
521 | uint32_t audit; | |
522 | uid_t loginuid; | |
523 | #endif | |
d025f1e4 ZJS |
524 | |
525 | assert(s); | |
526 | assert(iovec); | |
527 | assert(n > 0); | |
528 | assert(n + N_IOVEC_META_FIELDS <= m); | |
529 | ||
530 | if (ucred) { | |
d025f1e4 ZJS |
531 | realuid = ucred->uid; |
532 | ||
ae018d9b | 533 | sprintf(pid, "_PID=%lu", (unsigned long) ucred->pid); |
c2457105 | 534 | IOVEC_SET_STRING(iovec[n++], pid); |
d025f1e4 | 535 | |
ae018d9b | 536 | sprintf(uid, "_UID=%lu", (unsigned long) ucred->uid); |
c2457105 | 537 | IOVEC_SET_STRING(iovec[n++], uid); |
d025f1e4 | 538 | |
ae018d9b | 539 | sprintf(gid, "_GID=%lu", (unsigned long) ucred->gid); |
c2457105 | 540 | IOVEC_SET_STRING(iovec[n++], gid); |
d025f1e4 ZJS |
541 | |
542 | r = get_process_comm(ucred->pid, &t); | |
543 | if (r >= 0) { | |
ae018d9b | 544 | comm = strappenda("_COMM=", t); |
d025f1e4 | 545 | free(t); |
ae018d9b | 546 | IOVEC_SET_STRING(iovec[n++], comm); |
d025f1e4 ZJS |
547 | } |
548 | ||
549 | r = get_process_exe(ucred->pid, &t); | |
550 | if (r >= 0) { | |
ae018d9b | 551 | exe = strappenda("_EXE=", t); |
d025f1e4 | 552 | free(t); |
ae018d9b | 553 | IOVEC_SET_STRING(iovec[n++], exe); |
d025f1e4 ZJS |
554 | } |
555 | ||
9bdbc2e2 | 556 | r = get_process_cmdline(ucred->pid, 0, false, &t); |
d025f1e4 | 557 | if (r >= 0) { |
ae018d9b | 558 | cmdline = strappenda("_CMDLINE=", t); |
d025f1e4 | 559 | free(t); |
ae018d9b | 560 | IOVEC_SET_STRING(iovec[n++], cmdline); |
d025f1e4 ZJS |
561 | } |
562 | ||
0a20e3c1 | 563 | #ifdef HAVE_AUDIT |
d025f1e4 | 564 | r = audit_session_from_pid(ucred->pid, &audit); |
ae018d9b LP |
565 | if (r >= 0) { |
566 | sprintf(audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit); | |
567 | IOVEC_SET_STRING(iovec[n++], audit_session); | |
568 | } | |
d025f1e4 ZJS |
569 | |
570 | r = audit_loginuid_from_pid(ucred->pid, &loginuid); | |
7027ff61 | 571 | if (r >= 0) { |
ae018d9b LP |
572 | sprintf(audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid); |
573 | IOVEC_SET_STRING(iovec[n++], audit_loginuid); | |
d025f1e4 | 574 | } |
ae018d9b | 575 | #endif |
d025f1e4 | 576 | |
ae018d9b | 577 | r = cg_pid_get_path_shifted(ucred->pid, NULL, &c); |
7027ff61 | 578 | if (r >= 0) { |
ae018d9b LP |
579 | cgroup = strappenda("_SYSTEMD_CGROUP=", c); |
580 | IOVEC_SET_STRING(iovec[n++], cgroup); | |
d025f1e4 | 581 | |
ae018d9b LP |
582 | r = cg_path_get_session(c, &t); |
583 | if (r >= 0) { | |
584 | session = strappenda("_SYSTEMD_SESSION=", t); | |
585 | free(t); | |
d025f1e4 | 586 | IOVEC_SET_STRING(iovec[n++], session); |
ae018d9b LP |
587 | } |
588 | ||
589 | if (cg_path_get_owner_uid(c, &owner) >= 0) { | |
590 | owner_valid = true; | |
d025f1e4 | 591 | |
ae018d9b | 592 | sprintf(owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner); |
d025f1e4 | 593 | IOVEC_SET_STRING(iovec[n++], owner_uid); |
ae018d9b | 594 | } |
d025f1e4 | 595 | |
ae018d9b LP |
596 | if (cg_path_get_unit(c, &t) >= 0) { |
597 | unit = strappenda("_SYSTEMD_UNIT=", t); | |
598 | free(t); | |
599 | } else if (cg_path_get_user_unit(c, &t) >= 0) { | |
600 | unit = strappenda("_SYSTEMD_USER_UNIT=", t); | |
601 | free(t); | |
602 | } else if (unit_id) { | |
603 | if (session) | |
604 | unit = strappenda("_SYSTEMD_USER_UNIT=", unit_id); | |
605 | else | |
606 | unit = strappenda("_SYSTEMD_UNIT=", unit_id); | |
607 | } else | |
608 | unit = NULL; | |
609 | ||
610 | if (unit) | |
611 | IOVEC_SET_STRING(iovec[n++], unit); | |
612 | ||
613 | free(c); | |
ef1673d1 | 614 | } |
d025f1e4 | 615 | |
d025f1e4 ZJS |
616 | #ifdef HAVE_SELINUX |
617 | if (label) { | |
ae018d9b LP |
618 | selinux_context = alloca(sizeof("_SELINUX_CONTEXT=") + label_len); |
619 | ||
620 | *((char*) mempcpy(stpcpy(selinux_context, "_SELINUX_CONTEXT="), label, label_len)) = 0; | |
621 | IOVEC_SET_STRING(iovec[n++], selinux_context); | |
d025f1e4 ZJS |
622 | } else { |
623 | security_context_t con; | |
624 | ||
625 | if (getpidcon(ucred->pid, &con) >= 0) { | |
ae018d9b | 626 | selinux_context = strappenda("_SELINUX_CONTEXT=", con); |
d025f1e4 | 627 | freecon(con); |
ae018d9b | 628 | IOVEC_SET_STRING(iovec[n++], selinux_context); |
d025f1e4 ZJS |
629 | } |
630 | } | |
631 | #endif | |
632 | } | |
633 | ||
634 | if (tv) { | |
ae018d9b | 635 | sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu", (unsigned long long) timeval_load(tv)); |
a5693989 | 636 | IOVEC_SET_STRING(iovec[n++], source_time); |
d025f1e4 ZJS |
637 | } |
638 | ||
639 | /* Note that strictly speaking storing the boot id here is | |
640 | * redundant since the entry includes this in-line | |
641 | * anyway. However, we need this indexed, too. */ | |
642 | r = sd_id128_get_boot(&id); | |
adb435bb LP |
643 | if (r >= 0) { |
644 | sd_id128_to_string(id, boot_id + sizeof("_BOOT_ID=") - 1); | |
645 | IOVEC_SET_STRING(iovec[n++], boot_id); | |
646 | } | |
d025f1e4 ZJS |
647 | |
648 | r = sd_id128_get_machine(&id); | |
adb435bb | 649 | if (r >= 0) { |
d3789917 | 650 | sd_id128_to_string(id, machine_id + sizeof("_MACHINE_ID=") - 1); |
adb435bb LP |
651 | IOVEC_SET_STRING(iovec[n++], machine_id); |
652 | } | |
d025f1e4 ZJS |
653 | |
654 | t = gethostname_malloc(); | |
655 | if (t) { | |
ae018d9b | 656 | hostname = strappenda("_HOSTNAME=", t); |
d025f1e4 | 657 | free(t); |
ae018d9b | 658 | IOVEC_SET_STRING(iovec[n++], hostname); |
d025f1e4 ZJS |
659 | } |
660 | ||
661 | assert(n <= m); | |
662 | ||
da499392 | 663 | if (s->split_mode == SPLIT_UID && realuid > 0) |
40adcda8 | 664 | /* Split up strictly by any UID */ |
759c945a | 665 | journal_uid = realuid; |
82499507 | 666 | else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0) |
40adcda8 LP |
667 | /* Split up by login UIDs, this avoids creation of |
668 | * individual journals for system UIDs. We do this | |
669 | * only if the realuid is not root, in order not to | |
82499507 LP |
670 | * accidentally leak privileged information to the |
671 | * user that is logged by a privileged process that is | |
672 | * part of an unprivileged session.*/ | |
8a0889df | 673 | journal_uid = owner; |
da499392 KS |
674 | else |
675 | journal_uid = 0; | |
759c945a CW |
676 | |
677 | write_to_journal(s, journal_uid, iovec, n); | |
d025f1e4 ZJS |
678 | } |
679 | ||
680 | void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) { | |
681 | char mid[11 + 32 + 1]; | |
682 | char buffer[16 + LINE_MAX + 1]; | |
683 | struct iovec iovec[N_IOVEC_META_FIELDS + 4]; | |
684 | int n = 0; | |
685 | va_list ap; | |
b92bea5d | 686 | struct ucred ucred = {}; |
d025f1e4 ZJS |
687 | |
688 | assert(s); | |
689 | assert(format); | |
690 | ||
691 | IOVEC_SET_STRING(iovec[n++], "PRIORITY=6"); | |
692 | IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver"); | |
693 | ||
694 | memcpy(buffer, "MESSAGE=", 8); | |
695 | va_start(ap, format); | |
696 | vsnprintf(buffer + 8, sizeof(buffer) - 8, format, ap); | |
697 | va_end(ap); | |
698 | char_array_0(buffer); | |
699 | IOVEC_SET_STRING(iovec[n++], buffer); | |
700 | ||
701 | if (!sd_id128_equal(message_id, SD_ID128_NULL)) { | |
702 | snprintf(mid, sizeof(mid), MESSAGE_ID(message_id)); | |
703 | char_array_0(mid); | |
704 | IOVEC_SET_STRING(iovec[n++], mid); | |
705 | } | |
706 | ||
d025f1e4 ZJS |
707 | ucred.pid = getpid(); |
708 | ucred.uid = getuid(); | |
709 | ucred.gid = getgid(); | |
710 | ||
711 | dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL); | |
712 | } | |
713 | ||
714 | void server_dispatch_message( | |
715 | Server *s, | |
716 | struct iovec *iovec, unsigned n, unsigned m, | |
717 | struct ucred *ucred, | |
718 | struct timeval *tv, | |
719 | const char *label, size_t label_len, | |
720 | const char *unit_id, | |
721 | int priority) { | |
722 | ||
7027ff61 | 723 | int rl, r; |
7fd1b19b | 724 | _cleanup_free_ char *path = NULL; |
db91ea32 | 725 | char *c; |
d025f1e4 ZJS |
726 | |
727 | assert(s); | |
728 | assert(iovec || n == 0); | |
729 | ||
730 | if (n == 0) | |
731 | return; | |
732 | ||
733 | if (LOG_PRI(priority) > s->max_level_store) | |
734 | return; | |
735 | ||
736 | if (!ucred) | |
737 | goto finish; | |
738 | ||
7027ff61 LP |
739 | r = cg_pid_get_path_shifted(ucred->pid, NULL, &path); |
740 | if (r < 0) | |
d025f1e4 ZJS |
741 | goto finish; |
742 | ||
743 | /* example: /user/lennart/3/foobar | |
744 | * /system/dbus.service/foobar | |
745 | * | |
746 | * So let's cut of everything past the third /, since that is | |
747 | * where user directories start */ | |
748 | ||
749 | c = strchr(path, '/'); | |
750 | if (c) { | |
751 | c = strchr(c+1, '/'); | |
752 | if (c) { | |
753 | c = strchr(c+1, '/'); | |
754 | if (c) | |
755 | *c = 0; | |
756 | } | |
757 | } | |
758 | ||
db91ea32 ZJS |
759 | rl = journal_rate_limit_test(s->rate_limit, path, |
760 | priority & LOG_PRIMASK, available_space(s)); | |
d025f1e4 | 761 | |
db91ea32 | 762 | if (rl == 0) |
d025f1e4 | 763 | return; |
d025f1e4 ZJS |
764 | |
765 | /* Write a suppression message if we suppressed something */ | |
766 | if (rl > 1) | |
db91ea32 ZJS |
767 | server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, |
768 | "Suppressed %u messages from %s", rl - 1, path); | |
d025f1e4 ZJS |
769 | |
770 | finish: | |
771 | dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id); | |
772 | } | |
773 | ||
774 | ||
775 | static int system_journal_open(Server *s) { | |
776 | int r; | |
777 | char *fn; | |
778 | sd_id128_t machine; | |
779 | char ids[33]; | |
780 | ||
781 | r = sd_id128_get_machine(&machine); | |
782 | if (r < 0) | |
783 | return r; | |
784 | ||
785 | sd_id128_to_string(machine, ids); | |
786 | ||
787 | if (!s->system_journal && | |
788 | (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) && | |
789 | access("/run/systemd/journal/flushed", F_OK) >= 0) { | |
790 | ||
791 | /* If in auto mode: first try to create the machine | |
792 | * path, but not the prefix. | |
793 | * | |
794 | * If in persistent mode: create /var/log/journal and | |
795 | * the machine path */ | |
796 | ||
797 | if (s->storage == STORAGE_PERSISTENT) | |
798 | (void) mkdir("/var/log/journal/", 0755); | |
799 | ||
800 | fn = strappend("/var/log/journal/", ids); | |
801 | if (!fn) | |
802 | return -ENOMEM; | |
803 | ||
804 | (void) mkdir(fn, 0755); | |
805 | free(fn); | |
806 | ||
807 | fn = strjoin("/var/log/journal/", ids, "/system.journal", NULL); | |
808 | if (!fn) | |
809 | return -ENOMEM; | |
810 | ||
811 | r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &s->system_journal); | |
812 | free(fn); | |
813 | ||
814 | if (r >= 0) { | |
815 | char fb[FORMAT_BYTES_MAX]; | |
816 | ||
817 | server_fix_perms(s, s->system_journal, 0); | |
818 | server_driver_message(s, SD_ID128_NULL, "Allowing system journal files to grow to %s.", | |
819 | format_bytes(fb, sizeof(fb), s->system_metrics.max_use)); | |
820 | ||
821 | } else if (r < 0) { | |
822 | ||
823 | if (r != -ENOENT && r != -EROFS) | |
824 | log_warning("Failed to open system journal: %s", strerror(-r)); | |
825 | ||
826 | r = 0; | |
827 | } | |
828 | } | |
829 | ||
830 | if (!s->runtime_journal && | |
831 | (s->storage != STORAGE_NONE)) { | |
832 | ||
833 | fn = strjoin("/run/log/journal/", ids, "/system.journal", NULL); | |
834 | if (!fn) | |
835 | return -ENOMEM; | |
836 | ||
837 | if (s->system_journal) { | |
838 | ||
839 | /* Try to open the runtime journal, but only | |
840 | * if it already exists, so that we can flush | |
841 | * it into the system journal */ | |
842 | ||
843 | r = journal_file_open(fn, O_RDWR, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal); | |
844 | free(fn); | |
845 | ||
846 | if (r < 0) { | |
847 | if (r != -ENOENT) | |
848 | log_warning("Failed to open runtime journal: %s", strerror(-r)); | |
849 | ||
850 | r = 0; | |
851 | } | |
852 | ||
853 | } else { | |
854 | ||
855 | /* OK, we really need the runtime journal, so create | |
856 | * it if necessary. */ | |
857 | ||
858 | (void) mkdir_parents(fn, 0755); | |
859 | r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal); | |
860 | free(fn); | |
861 | ||
862 | if (r < 0) { | |
863 | log_error("Failed to open runtime journal: %s", strerror(-r)); | |
864 | return r; | |
865 | } | |
866 | } | |
867 | ||
868 | if (s->runtime_journal) { | |
869 | char fb[FORMAT_BYTES_MAX]; | |
870 | ||
871 | server_fix_perms(s, s->runtime_journal, 0); | |
872 | server_driver_message(s, SD_ID128_NULL, "Allowing runtime journal files to grow to %s.", | |
873 | format_bytes(fb, sizeof(fb), s->runtime_metrics.max_use)); | |
874 | } | |
875 | } | |
876 | ||
877 | return r; | |
878 | } | |
879 | ||
880 | int server_flush_to_var(Server *s) { | |
881 | int r; | |
882 | sd_id128_t machine; | |
883 | sd_journal *j = NULL; | |
884 | ||
885 | assert(s); | |
886 | ||
887 | if (s->storage != STORAGE_AUTO && | |
888 | s->storage != STORAGE_PERSISTENT) | |
889 | return 0; | |
890 | ||
891 | if (!s->runtime_journal) | |
892 | return 0; | |
893 | ||
894 | system_journal_open(s); | |
895 | ||
896 | if (!s->system_journal) | |
897 | return 0; | |
898 | ||
899 | log_debug("Flushing to /var..."); | |
900 | ||
901 | r = sd_id128_get_machine(&machine); | |
902 | if (r < 0) { | |
903 | log_error("Failed to get machine id: %s", strerror(-r)); | |
904 | return r; | |
905 | } | |
906 | ||
907 | r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); | |
908 | if (r < 0) { | |
909 | log_error("Failed to read runtime journal: %s", strerror(-r)); | |
910 | return r; | |
911 | } | |
912 | ||
93b73b06 LP |
913 | sd_journal_set_data_threshold(j, 0); |
914 | ||
d025f1e4 ZJS |
915 | SD_JOURNAL_FOREACH(j) { |
916 | Object *o = NULL; | |
917 | JournalFile *f; | |
918 | ||
919 | f = j->current_file; | |
920 | assert(f && f->current_offset > 0); | |
921 | ||
922 | r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); | |
923 | if (r < 0) { | |
924 | log_error("Can't read entry: %s", strerror(-r)); | |
925 | goto finish; | |
926 | } | |
927 | ||
928 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
929 | if (r >= 0) | |
930 | continue; | |
931 | ||
932 | if (!shall_try_append_again(s->system_journal, r)) { | |
933 | log_error("Can't write entry: %s", strerror(-r)); | |
934 | goto finish; | |
935 | } | |
936 | ||
937 | server_rotate(s); | |
938 | server_vacuum(s); | |
939 | ||
940 | log_debug("Retrying write."); | |
941 | r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); | |
942 | if (r < 0) { | |
943 | log_error("Can't write entry: %s", strerror(-r)); | |
944 | goto finish; | |
945 | } | |
946 | } | |
947 | ||
948 | finish: | |
949 | journal_file_post_change(s->system_journal); | |
950 | ||
951 | journal_file_close(s->runtime_journal); | |
952 | s->runtime_journal = NULL; | |
953 | ||
954 | if (r >= 0) | |
955 | rm_rf("/run/log/journal", false, true, false); | |
956 | ||
763c7aa2 | 957 | sd_journal_close(j); |
d025f1e4 ZJS |
958 | |
959 | return r; | |
960 | } | |
961 | ||
962 | int process_event(Server *s, struct epoll_event *ev) { | |
963 | assert(s); | |
964 | assert(ev); | |
965 | ||
966 | if (ev->data.fd == s->signal_fd) { | |
967 | struct signalfd_siginfo sfsi; | |
968 | ssize_t n; | |
969 | ||
970 | if (ev->events != EPOLLIN) { | |
971 | log_error("Got invalid event from epoll."); | |
972 | return -EIO; | |
973 | } | |
974 | ||
975 | n = read(s->signal_fd, &sfsi, sizeof(sfsi)); | |
976 | if (n != sizeof(sfsi)) { | |
977 | ||
978 | if (n >= 0) | |
979 | return -EIO; | |
980 | ||
981 | if (errno == EINTR || errno == EAGAIN) | |
982 | return 1; | |
983 | ||
984 | return -errno; | |
985 | } | |
986 | ||
d025f1e4 ZJS |
987 | if (sfsi.ssi_signo == SIGUSR1) { |
988 | touch("/run/systemd/journal/flushed"); | |
989 | server_flush_to_var(s); | |
26687bf8 | 990 | server_sync(s); |
d025f1e4 ZJS |
991 | return 1; |
992 | } | |
993 | ||
994 | if (sfsi.ssi_signo == SIGUSR2) { | |
995 | server_rotate(s); | |
996 | server_vacuum(s); | |
997 | return 1; | |
998 | } | |
999 | ||
26687bf8 OS |
1000 | log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo)); |
1001 | ||
d025f1e4 ZJS |
1002 | return 0; |
1003 | ||
26687bf8 OS |
1004 | } else if (ev->data.fd == s->sync_timer_fd) { |
1005 | int r; | |
1006 | uint64_t t; | |
1007 | ||
1008 | log_debug("Got sync request from epoll."); | |
1009 | ||
1010 | r = read(ev->data.fd, (void *)&t, sizeof(t)); | |
1011 | if (r < 0) | |
1012 | return 0; | |
1013 | ||
1014 | server_sync(s); | |
1015 | return 1; | |
1016 | ||
d025f1e4 ZJS |
1017 | } else if (ev->data.fd == s->dev_kmsg_fd) { |
1018 | int r; | |
1019 | ||
1020 | if (ev->events != EPOLLIN) { | |
1021 | log_error("Got invalid event from epoll."); | |
1022 | return -EIO; | |
1023 | } | |
1024 | ||
1025 | r = server_read_dev_kmsg(s); | |
1026 | if (r < 0) | |
1027 | return r; | |
1028 | ||
1029 | return 1; | |
1030 | ||
1031 | } else if (ev->data.fd == s->native_fd || | |
1032 | ev->data.fd == s->syslog_fd) { | |
1033 | ||
1034 | if (ev->events != EPOLLIN) { | |
1035 | log_error("Got invalid event from epoll."); | |
1036 | return -EIO; | |
1037 | } | |
1038 | ||
1039 | for (;;) { | |
1040 | struct msghdr msghdr; | |
1041 | struct iovec iovec; | |
1042 | struct ucred *ucred = NULL; | |
1043 | struct timeval *tv = NULL; | |
1044 | struct cmsghdr *cmsg; | |
1045 | char *label = NULL; | |
1046 | size_t label_len = 0; | |
1047 | union { | |
1048 | struct cmsghdr cmsghdr; | |
1049 | ||
1050 | /* We use NAME_MAX space for the | |
1051 | * SELinux label here. The kernel | |
1052 | * currently enforces no limit, but | |
1053 | * according to suggestions from the | |
1054 | * SELinux people this will change and | |
1055 | * it will probably be identical to | |
1056 | * NAME_MAX. For now we use that, but | |
1057 | * this should be updated one day when | |
1058 | * the final limit is known.*/ | |
1059 | uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) + | |
1060 | CMSG_SPACE(sizeof(struct timeval)) + | |
1061 | CMSG_SPACE(sizeof(int)) + /* fd */ | |
1062 | CMSG_SPACE(NAME_MAX)]; /* selinux label */ | |
1063 | } control; | |
1064 | ssize_t n; | |
1065 | int v; | |
1066 | int *fds = NULL; | |
1067 | unsigned n_fds = 0; | |
1068 | ||
1069 | if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) { | |
1070 | log_error("SIOCINQ failed: %m"); | |
1071 | return -errno; | |
1072 | } | |
1073 | ||
1074 | if (s->buffer_size < (size_t) v) { | |
1075 | void *b; | |
1076 | size_t l; | |
1077 | ||
1078 | l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2); | |
1079 | b = realloc(s->buffer, l+1); | |
1080 | ||
1081 | if (!b) { | |
1082 | log_error("Couldn't increase buffer."); | |
1083 | return -ENOMEM; | |
1084 | } | |
1085 | ||
1086 | s->buffer_size = l; | |
1087 | s->buffer = b; | |
1088 | } | |
1089 | ||
1090 | zero(iovec); | |
1091 | iovec.iov_base = s->buffer; | |
1092 | iovec.iov_len = s->buffer_size; | |
1093 | ||
1094 | zero(control); | |
1095 | zero(msghdr); | |
1096 | msghdr.msg_iov = &iovec; | |
1097 | msghdr.msg_iovlen = 1; | |
1098 | msghdr.msg_control = &control; | |
1099 | msghdr.msg_controllen = sizeof(control); | |
1100 | ||
1101 | n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); | |
1102 | if (n < 0) { | |
1103 | ||
1104 | if (errno == EINTR || errno == EAGAIN) | |
1105 | return 1; | |
1106 | ||
1107 | log_error("recvmsg() failed: %m"); | |
1108 | return -errno; | |
1109 | } | |
1110 | ||
1111 | for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) { | |
1112 | ||
1113 | if (cmsg->cmsg_level == SOL_SOCKET && | |
1114 | cmsg->cmsg_type == SCM_CREDENTIALS && | |
1115 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) | |
1116 | ucred = (struct ucred*) CMSG_DATA(cmsg); | |
1117 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1118 | cmsg->cmsg_type == SCM_SECURITY) { | |
1119 | label = (char*) CMSG_DATA(cmsg); | |
1120 | label_len = cmsg->cmsg_len - CMSG_LEN(0); | |
1121 | } else if (cmsg->cmsg_level == SOL_SOCKET && | |
1122 | cmsg->cmsg_type == SO_TIMESTAMP && | |
1123 | cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval))) | |
1124 | tv = (struct timeval*) CMSG_DATA(cmsg); | |
1125 | else if (cmsg->cmsg_level == SOL_SOCKET && | |
1126 | cmsg->cmsg_type == SCM_RIGHTS) { | |
1127 | fds = (int*) CMSG_DATA(cmsg); | |
1128 | n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); | |
1129 | } | |
1130 | } | |
1131 | ||
1132 | if (ev->data.fd == s->syslog_fd) { | |
1133 | char *e; | |
1134 | ||
1135 | if (n > 0 && n_fds == 0) { | |
1136 | e = memchr(s->buffer, '\n', n); | |
1137 | if (e) | |
1138 | *e = 0; | |
1139 | else | |
1140 | s->buffer[n] = 0; | |
1141 | ||
1142 | server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len); | |
1143 | } else if (n_fds > 0) | |
1144 | log_warning("Got file descriptors via syslog socket. Ignoring."); | |
1145 | ||
1146 | } else { | |
1147 | if (n > 0 && n_fds == 0) | |
1148 | server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len); | |
1149 | else if (n == 0 && n_fds == 1) | |
1150 | server_process_native_file(s, fds[0], ucred, tv, label, label_len); | |
1151 | else if (n_fds > 0) | |
1152 | log_warning("Got too many file descriptors via native socket. Ignoring."); | |
1153 | } | |
1154 | ||
1155 | close_many(fds, n_fds); | |
1156 | } | |
1157 | ||
1158 | return 1; | |
1159 | ||
1160 | } else if (ev->data.fd == s->stdout_fd) { | |
1161 | ||
1162 | if (ev->events != EPOLLIN) { | |
1163 | log_error("Got invalid event from epoll."); | |
1164 | return -EIO; | |
1165 | } | |
1166 | ||
1167 | stdout_stream_new(s); | |
1168 | return 1; | |
1169 | ||
1170 | } else { | |
1171 | StdoutStream *stream; | |
1172 | ||
1173 | if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) { | |
1174 | log_error("Got invalid event from epoll."); | |
1175 | return -EIO; | |
1176 | } | |
1177 | ||
1178 | /* If it is none of the well-known fds, it must be an | |
1179 | * stdout stream fd. Note that this is a bit ugly here | |
1180 | * (since we rely that none of the well-known fds | |
1181 | * could be interpreted as pointer), but nonetheless | |
1182 | * safe, since the well-known fds would never get an | |
1183 | * fd > 4096, i.e. beyond the first memory page */ | |
1184 | ||
1185 | stream = ev->data.ptr; | |
1186 | ||
1187 | if (stdout_stream_process(stream) <= 0) | |
1188 | stdout_stream_free(stream); | |
1189 | ||
1190 | return 1; | |
1191 | } | |
1192 | ||
1193 | log_error("Unknown event."); | |
1194 | return 0; | |
1195 | } | |
1196 | ||
1197 | static int open_signalfd(Server *s) { | |
1198 | sigset_t mask; | |
1199 | struct epoll_event ev; | |
1200 | ||
1201 | assert(s); | |
1202 | ||
1203 | assert_se(sigemptyset(&mask) == 0); | |
1204 | sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, -1); | |
1205 | assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); | |
1206 | ||
1207 | s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC); | |
1208 | if (s->signal_fd < 0) { | |
1209 | log_error("signalfd(): %m"); | |
1210 | return -errno; | |
1211 | } | |
1212 | ||
1213 | zero(ev); | |
1214 | ev.events = EPOLLIN; | |
1215 | ev.data.fd = s->signal_fd; | |
1216 | ||
1217 | if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) { | |
1218 | log_error("epoll_ctl(): %m"); | |
1219 | return -errno; | |
1220 | } | |
1221 | ||
1222 | return 0; | |
1223 | } | |
1224 | ||
1225 | static int server_parse_proc_cmdline(Server *s) { | |
7fd1b19b | 1226 | _cleanup_free_ char *line = NULL; |
db91ea32 | 1227 | char *w, *state; |
d025f1e4 ZJS |
1228 | int r; |
1229 | size_t l; | |
1230 | ||
1231 | if (detect_container(NULL) > 0) | |
1232 | return 0; | |
1233 | ||
1234 | r = read_one_line_file("/proc/cmdline", &line); | |
1235 | if (r < 0) { | |
1236 | log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); | |
1237 | return 0; | |
1238 | } | |
1239 | ||
1240 | FOREACH_WORD_QUOTED(w, l, line, state) { | |
7fd1b19b | 1241 | _cleanup_free_ char *word; |
d025f1e4 ZJS |
1242 | |
1243 | word = strndup(w, l); | |
db91ea32 ZJS |
1244 | if (!word) |
1245 | return -ENOMEM; | |
d025f1e4 ZJS |
1246 | |
1247 | if (startswith(word, "systemd.journald.forward_to_syslog=")) { | |
1248 | r = parse_boolean(word + 35); | |
1249 | if (r < 0) | |
1250 | log_warning("Failed to parse forward to syslog switch %s. Ignoring.", word + 35); | |
1251 | else | |
1252 | s->forward_to_syslog = r; | |
1253 | } else if (startswith(word, "systemd.journald.forward_to_kmsg=")) { | |
1254 | r = parse_boolean(word + 33); | |
1255 | if (r < 0) | |
1256 | log_warning("Failed to parse forward to kmsg switch %s. Ignoring.", word + 33); | |
1257 | else | |
1258 | s->forward_to_kmsg = r; | |
1259 | } else if (startswith(word, "systemd.journald.forward_to_console=")) { | |
1260 | r = parse_boolean(word + 36); | |
1261 | if (r < 0) | |
1262 | log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36); | |
1263 | else | |
1264 | s->forward_to_console = r; | |
1265 | } else if (startswith(word, "systemd.journald")) | |
1266 | log_warning("Invalid systemd.journald parameter. Ignoring."); | |
d025f1e4 ZJS |
1267 | } |
1268 | ||
db91ea32 | 1269 | return 0; |
d025f1e4 ZJS |
1270 | } |
1271 | ||
1272 | static int server_parse_config_file(Server *s) { | |
db91ea32 | 1273 | static const char *fn = "/etc/systemd/journald.conf"; |
7fd1b19b | 1274 | _cleanup_fclose_ FILE *f = NULL; |
d025f1e4 ZJS |
1275 | int r; |
1276 | ||
1277 | assert(s); | |
1278 | ||
d025f1e4 ZJS |
1279 | f = fopen(fn, "re"); |
1280 | if (!f) { | |
1281 | if (errno == ENOENT) | |
1282 | return 0; | |
1283 | ||
1284 | log_warning("Failed to open configuration file %s: %m", fn); | |
1285 | return -errno; | |
1286 | } | |
1287 | ||
e8e581bf | 1288 | r = config_parse(NULL, fn, f, "Journal\0", config_item_perf_lookup, |
db91ea32 | 1289 | (void*) journald_gperf_lookup, false, s); |
d025f1e4 ZJS |
1290 | if (r < 0) |
1291 | log_warning("Failed to parse configuration file: %s", strerror(-r)); | |
1292 | ||
d025f1e4 ZJS |
1293 | return r; |
1294 | } | |
1295 | ||
26687bf8 OS |
1296 | static int server_open_sync_timer(Server *s) { |
1297 | int r; | |
1298 | struct epoll_event ev; | |
1299 | ||
1300 | assert(s); | |
1301 | ||
1302 | s->sync_timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_CLOEXEC); | |
1303 | if (s->sync_timer_fd < 0) | |
1304 | return -errno; | |
1305 | ||
1306 | zero(ev); | |
1307 | ev.events = EPOLLIN; | |
1308 | ev.data.fd = s->sync_timer_fd; | |
1309 | ||
1310 | r = epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->sync_timer_fd, &ev); | |
1311 | if (r < 0) { | |
1312 | log_error("Failed to add idle timer fd to epoll object: %m"); | |
1313 | return -errno; | |
1314 | } | |
1315 | ||
1316 | return 0; | |
1317 | } | |
1318 | ||
1319 | int server_schedule_sync(Server *s) { | |
1320 | int r; | |
1321 | ||
26687bf8 OS |
1322 | assert(s); |
1323 | ||
1324 | if (s->sync_scheduled) | |
1325 | return 0; | |
1326 | ||
1327 | if (s->sync_interval_usec) { | |
b92bea5d ZJS |
1328 | struct itimerspec sync_timer_enable = { |
1329 | .it_value.tv_sec = s->sync_interval_usec / USEC_PER_SEC, | |
1330 | .it_value.tv_nsec = s->sync_interval_usec % MSEC_PER_SEC, | |
1331 | }; | |
26687bf8 OS |
1332 | |
1333 | r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_enable, NULL); | |
1334 | if (r < 0) | |
1335 | return -errno; | |
1336 | } | |
1337 | ||
1338 | s->sync_scheduled = true; | |
1339 | ||
1340 | return 0; | |
1341 | } | |
1342 | ||
d025f1e4 ZJS |
1343 | int server_init(Server *s) { |
1344 | int n, r, fd; | |
1345 | ||
1346 | assert(s); | |
1347 | ||
1348 | zero(*s); | |
26687bf8 OS |
1349 | s->sync_timer_fd = s->syslog_fd = s->native_fd = s->stdout_fd = |
1350 | s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1; | |
d025f1e4 ZJS |
1351 | s->compress = true; |
1352 | s->seal = true; | |
1353 | ||
26687bf8 OS |
1354 | s->sync_interval_usec = DEFAULT_SYNC_INTERVAL_USEC; |
1355 | s->sync_scheduled = false; | |
1356 | ||
d025f1e4 ZJS |
1357 | s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL; |
1358 | s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST; | |
1359 | ||
1360 | s->forward_to_syslog = true; | |
1361 | ||
1362 | s->max_level_store = LOG_DEBUG; | |
1363 | s->max_level_syslog = LOG_DEBUG; | |
1364 | s->max_level_kmsg = LOG_NOTICE; | |
1365 | s->max_level_console = LOG_INFO; | |
1366 | ||
1367 | memset(&s->system_metrics, 0xFF, sizeof(s->system_metrics)); | |
1368 | memset(&s->runtime_metrics, 0xFF, sizeof(s->runtime_metrics)); | |
1369 | ||
1370 | server_parse_config_file(s); | |
1371 | server_parse_proc_cmdline(s); | |
d288f79f ZJS |
1372 | if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { |
1373 | log_debug("Setting both rate limit interval and burst from %llu,%u to 0,0", | |
1374 | (long long unsigned) s->rate_limit_interval, | |
1375 | s->rate_limit_burst); | |
1376 | s->rate_limit_interval = s->rate_limit_burst = 0; | |
1377 | } | |
d025f1e4 ZJS |
1378 | |
1379 | mkdir_p("/run/systemd/journal", 0755); | |
1380 | ||
1381 | s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func); | |
1382 | if (!s->user_journals) | |
1383 | return log_oom(); | |
1384 | ||
1385 | s->mmap = mmap_cache_new(); | |
1386 | if (!s->mmap) | |
1387 | return log_oom(); | |
1388 | ||
1389 | s->epoll_fd = epoll_create1(EPOLL_CLOEXEC); | |
1390 | if (s->epoll_fd < 0) { | |
1391 | log_error("Failed to create epoll object: %m"); | |
1392 | return -errno; | |
1393 | } | |
1394 | ||
1395 | n = sd_listen_fds(true); | |
1396 | if (n < 0) { | |
1397 | log_error("Failed to read listening file descriptors from environment: %s", strerror(-n)); | |
1398 | return n; | |
1399 | } | |
1400 | ||
1401 | for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { | |
1402 | ||
1403 | if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) { | |
1404 | ||
1405 | if (s->native_fd >= 0) { | |
1406 | log_error("Too many native sockets passed."); | |
1407 | return -EINVAL; | |
1408 | } | |
1409 | ||
1410 | s->native_fd = fd; | |
1411 | ||
1412 | } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) { | |
1413 | ||
1414 | if (s->stdout_fd >= 0) { | |
1415 | log_error("Too many stdout sockets passed."); | |
1416 | return -EINVAL; | |
1417 | } | |
1418 | ||
1419 | s->stdout_fd = fd; | |
1420 | ||
1421 | } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) { | |
1422 | ||
1423 | if (s->syslog_fd >= 0) { | |
1424 | log_error("Too many /dev/log sockets passed."); | |
1425 | return -EINVAL; | |
1426 | } | |
1427 | ||
1428 | s->syslog_fd = fd; | |
1429 | ||
1430 | } else { | |
1431 | log_error("Unknown socket passed."); | |
1432 | return -EINVAL; | |
1433 | } | |
1434 | } | |
1435 | ||
1436 | r = server_open_syslog_socket(s); | |
1437 | if (r < 0) | |
1438 | return r; | |
1439 | ||
1440 | r = server_open_native_socket(s); | |
1441 | if (r < 0) | |
1442 | return r; | |
1443 | ||
1444 | r = server_open_stdout_socket(s); | |
1445 | if (r < 0) | |
1446 | return r; | |
1447 | ||
1448 | r = server_open_dev_kmsg(s); | |
1449 | if (r < 0) | |
1450 | return r; | |
1451 | ||
1452 | r = server_open_kernel_seqnum(s); | |
1453 | if (r < 0) | |
1454 | return r; | |
1455 | ||
26687bf8 OS |
1456 | r = server_open_sync_timer(s); |
1457 | if (r < 0) | |
1458 | return r; | |
1459 | ||
d025f1e4 ZJS |
1460 | r = open_signalfd(s); |
1461 | if (r < 0) | |
1462 | return r; | |
1463 | ||
1464 | s->udev = udev_new(); | |
1465 | if (!s->udev) | |
1466 | return -ENOMEM; | |
1467 | ||
d288f79f ZJS |
1468 | s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, |
1469 | s->rate_limit_burst); | |
d025f1e4 ZJS |
1470 | if (!s->rate_limit) |
1471 | return -ENOMEM; | |
1472 | ||
1473 | r = system_journal_open(s); | |
1474 | if (r < 0) | |
1475 | return r; | |
1476 | ||
1477 | return 0; | |
1478 | } | |
1479 | ||
1480 | void server_maybe_append_tags(Server *s) { | |
1481 | #ifdef HAVE_GCRYPT | |
1482 | JournalFile *f; | |
1483 | Iterator i; | |
1484 | usec_t n; | |
1485 | ||
1486 | n = now(CLOCK_REALTIME); | |
1487 | ||
1488 | if (s->system_journal) | |
1489 | journal_file_maybe_append_tag(s->system_journal, n); | |
1490 | ||
1491 | HASHMAP_FOREACH(f, s->user_journals, i) | |
1492 | journal_file_maybe_append_tag(f, n); | |
1493 | #endif | |
1494 | } | |
1495 | ||
1496 | void server_done(Server *s) { | |
1497 | JournalFile *f; | |
1498 | assert(s); | |
1499 | ||
1500 | while (s->stdout_streams) | |
1501 | stdout_stream_free(s->stdout_streams); | |
1502 | ||
1503 | if (s->system_journal) | |
1504 | journal_file_close(s->system_journal); | |
1505 | ||
1506 | if (s->runtime_journal) | |
1507 | journal_file_close(s->runtime_journal); | |
1508 | ||
1509 | while ((f = hashmap_steal_first(s->user_journals))) | |
1510 | journal_file_close(f); | |
1511 | ||
1512 | hashmap_free(s->user_journals); | |
1513 | ||
1514 | if (s->epoll_fd >= 0) | |
1515 | close_nointr_nofail(s->epoll_fd); | |
1516 | ||
1517 | if (s->signal_fd >= 0) | |
1518 | close_nointr_nofail(s->signal_fd); | |
1519 | ||
1520 | if (s->syslog_fd >= 0) | |
1521 | close_nointr_nofail(s->syslog_fd); | |
1522 | ||
1523 | if (s->native_fd >= 0) | |
1524 | close_nointr_nofail(s->native_fd); | |
1525 | ||
1526 | if (s->stdout_fd >= 0) | |
1527 | close_nointr_nofail(s->stdout_fd); | |
1528 | ||
1529 | if (s->dev_kmsg_fd >= 0) | |
1530 | close_nointr_nofail(s->dev_kmsg_fd); | |
1531 | ||
26687bf8 OS |
1532 | if (s->sync_timer_fd >= 0) |
1533 | close_nointr_nofail(s->sync_timer_fd); | |
1534 | ||
d025f1e4 ZJS |
1535 | if (s->rate_limit) |
1536 | journal_rate_limit_free(s->rate_limit); | |
1537 | ||
1538 | if (s->kernel_seqnum) | |
1539 | munmap(s->kernel_seqnum, sizeof(uint64_t)); | |
1540 | ||
1541 | free(s->buffer); | |
1542 | free(s->tty_path); | |
1543 | ||
1544 | if (s->mmap) | |
1545 | mmap_cache_unref(s->mmap); | |
1546 | ||
1547 | if (s->udev) | |
1548 | udev_unref(s->udev); | |
1549 | } |