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