]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/journal/journald.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
journald: fix calculation of disk space
[thirdparty/systemd.git] / src / journal / journald.c
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 General Public License as published by
10 the Free Software Foundation; either version 2 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 General Public License for more details.
17
18 You should have received a copy of the GNU General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
20 ***/
21
22 #include <sys/epoll.h>
23 #include <sys/socket.h>
24 #include <errno.h>
25 #include <sys/signalfd.h>
26 #include <unistd.h>
27 #include <fcntl.h>
28 #include <stddef.h>
29 #include <sys/ioctl.h>
30 #include <linux/sockios.h>
31 #include <sys/statvfs.h>
32 #include <sys/user.h>
33
34 #include <systemd/sd-journal.h>
35 #include <systemd/sd-login.h>
36 #include <systemd/sd-messages.h>
37 #include <systemd/sd-daemon.h>
38
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 "journal-rate-limit.h"
45 #include "journal-internal.h"
46 #include "conf-parser.h"
47 #include "journald.h"
48 #include "virt.h"
49 #include "missing.h"
50
51 #ifdef HAVE_ACL
52 #include <sys/acl.h>
53 #include <acl/libacl.h>
54 #include "acl-util.h"
55 #endif
56
57 #ifdef HAVE_SELINUX
58 #include <selinux/selinux.h>
59 #endif
60
61 #define USER_JOURNALS_MAX 1024
62 #define STDOUT_STREAMS_MAX 4096
63
64 #define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC)
65 #define DEFAULT_RATE_LIMIT_BURST 200
66
67 #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
68
69 #define RECHECK_VAR_AVAILABLE_USEC (30*USEC_PER_SEC)
70
71 #define N_IOVEC_META_FIELDS 17
72
73 #define ENTRY_SIZE_MAX (1024*1024*32)
74
75 typedef enum StdoutStreamState {
76 STDOUT_STREAM_IDENTIFIER,
77 STDOUT_STREAM_PRIORITY,
78 STDOUT_STREAM_LEVEL_PREFIX,
79 STDOUT_STREAM_FORWARD_TO_SYSLOG,
80 STDOUT_STREAM_FORWARD_TO_KMSG,
81 STDOUT_STREAM_FORWARD_TO_CONSOLE,
82 STDOUT_STREAM_RUNNING
83 } StdoutStreamState;
84
85 struct StdoutStream {
86 Server *server;
87 StdoutStreamState state;
88
89 int fd;
90
91 struct ucred ucred;
92 #ifdef HAVE_SELINUX
93 security_context_t security_context;
94 #endif
95
96 char *identifier;
97 int priority;
98 bool level_prefix:1;
99 bool forward_to_syslog:1;
100 bool forward_to_kmsg:1;
101 bool forward_to_console:1;
102
103 char buffer[LINE_MAX+1];
104 size_t length;
105
106 LIST_FIELDS(StdoutStream, stdout_stream);
107 };
108
109 static int server_flush_to_var(Server *s);
110
111 static uint64_t available_space(Server *s) {
112 char ids[33], *p;
113 const char *f;
114 sd_id128_t machine;
115 struct statvfs ss;
116 uint64_t sum = 0, avail = 0, ss_avail = 0;
117 int r;
118 DIR *d;
119 usec_t ts;
120 JournalMetrics *m;
121
122 ts = now(CLOCK_MONOTONIC);
123
124 if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts)
125 return s->cached_available_space;
126
127 r = sd_id128_get_machine(&machine);
128 if (r < 0)
129 return 0;
130
131 if (s->system_journal) {
132 f = "/var/log/journal/";
133 m = &s->system_metrics;
134 } else {
135 f = "/run/log/journal/";
136 m = &s->runtime_metrics;
137 }
138
139 assert(m);
140
141 p = strappend(f, sd_id128_to_string(machine, ids));
142 if (!p)
143 return 0;
144
145 d = opendir(p);
146 free(p);
147
148 if (!d)
149 return 0;
150
151 if (fstatvfs(dirfd(d), &ss) < 0)
152 goto finish;
153
154 for (;;) {
155 struct stat st;
156 struct dirent buf, *de;
157 int k;
158
159 k = readdir_r(d, &buf, &de);
160 if (k != 0) {
161 r = -k;
162 goto finish;
163 }
164
165 if (!de)
166 break;
167
168 if (!dirent_is_file_with_suffix(de, ".journal"))
169 continue;
170
171 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
172 continue;
173
174 if (!S_ISREG(st.st_mode))
175 continue;
176
177 sum += (uint64_t) st.st_blocks * 512UL;
178 }
179
180 avail = sum >= m->max_use ? 0 : m->max_use - sum;
181
182 ss_avail = ss.f_bsize * ss.f_bavail;
183
184 ss_avail = ss_avail < m->keep_free ? 0 : ss_avail - m->keep_free;
185
186 if (ss_avail < avail)
187 avail = ss_avail;
188
189 s->cached_available_space = avail;
190 s->cached_available_space_timestamp = ts;
191
192 finish:
193 closedir(d);
194
195 return avail;
196 }
197
198 static void server_read_file_gid(Server *s) {
199 const char *adm = "adm";
200 int r;
201
202 assert(s);
203
204 if (s->file_gid_valid)
205 return;
206
207 r = get_group_creds(&adm, &s->file_gid);
208 if (r < 0)
209 log_warning("Failed to resolve 'adm' group: %s", strerror(-r));
210
211 /* if we couldn't read the gid, then it will be 0, but that's
212 * fine and we shouldn't try to resolve the group again, so
213 * let's just pretend it worked right-away. */
214 s->file_gid_valid = true;
215 }
216
217 static void server_fix_perms(Server *s, JournalFile *f, uid_t uid) {
218 int r;
219 #ifdef HAVE_ACL
220 acl_t acl;
221 acl_entry_t entry;
222 acl_permset_t permset;
223 #endif
224
225 assert(f);
226
227 server_read_file_gid(s);
228
229 r = fchmod_and_fchown(f->fd, 0640, 0, s->file_gid);
230 if (r < 0)
231 log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
232
233 #ifdef HAVE_ACL
234 if (uid <= 0)
235 return;
236
237 acl = acl_get_fd(f->fd);
238 if (!acl) {
239 log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
240 return;
241 }
242
243 r = acl_find_uid(acl, uid, &entry);
244 if (r <= 0) {
245
246 if (acl_create_entry(&acl, &entry) < 0 ||
247 acl_set_tag_type(entry, ACL_USER) < 0 ||
248 acl_set_qualifier(entry, &uid) < 0) {
249 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
250 goto finish;
251 }
252 }
253
254 if (acl_get_permset(entry, &permset) < 0 ||
255 acl_add_perm(permset, ACL_READ) < 0 ||
256 acl_calc_mask(&acl) < 0) {
257 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
258 goto finish;
259 }
260
261 if (acl_set_fd(f->fd, acl) < 0)
262 log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
263
264 finish:
265 acl_free(acl);
266 #endif
267 }
268
269 static JournalFile* find_journal(Server *s, uid_t uid) {
270 char *p;
271 int r;
272 JournalFile *f;
273 char ids[33];
274 sd_id128_t machine;
275
276 assert(s);
277
278 /* We split up user logs only on /var, not on /run. If the
279 * runtime file is open, we write to it exclusively, in order
280 * to guarantee proper order as soon as we flush /run to
281 * /var and close the runtime file. */
282
283 if (s->runtime_journal)
284 return s->runtime_journal;
285
286 if (uid <= 0)
287 return s->system_journal;
288
289 r = sd_id128_get_machine(&machine);
290 if (r < 0)
291 return s->system_journal;
292
293 f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));
294 if (f)
295 return f;
296
297 if (asprintf(&p, "/var/log/journal/%s/user-%lu.journal", sd_id128_to_string(machine, ids), (unsigned long) uid) < 0)
298 return s->system_journal;
299
300 while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
301 /* Too many open? Then let's close one */
302 f = hashmap_steal_first(s->user_journals);
303 assert(f);
304 journal_file_close(f);
305 }
306
307 r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
308 free(p);
309
310 if (r < 0)
311 return s->system_journal;
312
313 server_fix_perms(s, f, uid);
314
315 r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
316 if (r < 0) {
317 journal_file_close(f);
318 return s->system_journal;
319 }
320
321 return f;
322 }
323
324 static void server_rotate(Server *s) {
325 JournalFile *f;
326 void *k;
327 Iterator i;
328 int r;
329
330 log_info("Rotating...");
331
332 if (s->runtime_journal) {
333 r = journal_file_rotate(&s->runtime_journal);
334 if (r < 0)
335 log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
336 }
337
338 if (s->system_journal) {
339 r = journal_file_rotate(&s->system_journal);
340 if (r < 0)
341 log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
342 }
343
344 HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
345 r = journal_file_rotate(&f);
346 if (r < 0)
347 log_error("Failed to rotate %s: %s", f->path, strerror(-r));
348 else
349 hashmap_replace(s->user_journals, k, f);
350 }
351 }
352
353 static void server_vacuum(Server *s) {
354 char *p;
355 char ids[33];
356 sd_id128_t machine;
357 int r;
358
359 log_info("Vacuuming...");
360
361 r = sd_id128_get_machine(&machine);
362 if (r < 0) {
363 log_error("Failed to get machine ID: %s", strerror(-r));
364 return;
365 }
366
367 sd_id128_to_string(machine, ids);
368
369 if (s->system_journal) {
370 if (asprintf(&p, "/var/log/journal/%s", ids) < 0) {
371 log_error("Out of memory.");
372 return;
373 }
374
375 r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free);
376 if (r < 0 && r != -ENOENT)
377 log_error("Failed to vacuum %s: %s", p, strerror(-r));
378 free(p);
379 }
380
381
382 if (s->runtime_journal) {
383 if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
384 log_error("Out of memory.");
385 return;
386 }
387
388 r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free);
389 if (r < 0 && r != -ENOENT)
390 log_error("Failed to vacuum %s: %s", p, strerror(-r));
391 free(p);
392 }
393
394 s->cached_available_space_timestamp = 0;
395 }
396
397 static char *shortened_cgroup_path(pid_t pid) {
398 int r;
399 char *process_path, *init_path, *path;
400
401 assert(pid > 0);
402
403 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
404 if (r < 0)
405 return NULL;
406
407 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
408 if (r < 0) {
409 free(process_path);
410 return NULL;
411 }
412
413 if (endswith(init_path, "/system"))
414 init_path[strlen(init_path) - 7] = 0;
415 else if (streq(init_path, "/"))
416 init_path[0] = 0;
417
418 if (startswith(process_path, init_path)) {
419 char *p;
420
421 p = strdup(process_path + strlen(init_path));
422 if (!p) {
423 free(process_path);
424 free(init_path);
425 return NULL;
426 }
427 path = p;
428 } else {
429 path = process_path;
430 process_path = NULL;
431 }
432
433 free(process_path);
434 free(init_path);
435
436 return path;
437 }
438
439 static void dispatch_message_real(
440 Server *s,
441 struct iovec *iovec, unsigned n, unsigned m,
442 struct ucred *ucred,
443 struct timeval *tv,
444 const char *label, size_t label_len) {
445
446 char *pid = NULL, *uid = NULL, *gid = NULL,
447 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
448 *comm = NULL, *cmdline = NULL, *hostname = NULL,
449 *audit_session = NULL, *audit_loginuid = NULL,
450 *exe = NULL, *cgroup = NULL, *session = NULL,
451 *owner_uid = NULL, *unit = NULL, *selinux_context = NULL;
452
453 char idbuf[33];
454 sd_id128_t id;
455 int r;
456 char *t;
457 uid_t loginuid = 0, realuid = 0;
458 JournalFile *f;
459 bool vacuumed = false;
460
461 assert(s);
462 assert(iovec);
463 assert(n > 0);
464 assert(n + N_IOVEC_META_FIELDS <= m);
465
466 if (ucred) {
467 uint32_t audit;
468 uid_t owner;
469
470 realuid = ucred->uid;
471
472 if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
473 IOVEC_SET_STRING(iovec[n++], pid);
474
475 if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
476 IOVEC_SET_STRING(iovec[n++], uid);
477
478 if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
479 IOVEC_SET_STRING(iovec[n++], gid);
480
481 r = get_process_comm(ucred->pid, &t);
482 if (r >= 0) {
483 comm = strappend("_COMM=", t);
484 free(t);
485
486 if (comm)
487 IOVEC_SET_STRING(iovec[n++], comm);
488 }
489
490 r = get_process_exe(ucred->pid, &t);
491 if (r >= 0) {
492 exe = strappend("_EXE=", t);
493 free(t);
494
495 if (exe)
496 IOVEC_SET_STRING(iovec[n++], exe);
497 }
498
499 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
500 if (r >= 0) {
501 cmdline = strappend("_CMDLINE=", t);
502 free(t);
503
504 if (cmdline)
505 IOVEC_SET_STRING(iovec[n++], cmdline);
506 }
507
508 r = audit_session_from_pid(ucred->pid, &audit);
509 if (r >= 0)
510 if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit) >= 0)
511 IOVEC_SET_STRING(iovec[n++], audit_session);
512
513 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
514 if (r >= 0)
515 if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
516 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
517
518 t = shortened_cgroup_path(ucred->pid);
519 if (t) {
520 cgroup = strappend("_SYSTEMD_CGROUP=", t);
521 free(t);
522
523 if (cgroup)
524 IOVEC_SET_STRING(iovec[n++], cgroup);
525 }
526
527 if (sd_pid_get_session(ucred->pid, &t) >= 0) {
528 session = strappend("_SYSTEMD_SESSION=", t);
529 free(t);
530
531 if (session)
532 IOVEC_SET_STRING(iovec[n++], session);
533 }
534
535 if (sd_pid_get_unit(ucred->pid, &t) >= 0) {
536 unit = strappend("_SYSTEMD_UNIT=", t);
537 free(t);
538
539 if (unit)
540 IOVEC_SET_STRING(iovec[n++], unit);
541 }
542
543 if (sd_pid_get_owner_uid(ucred->uid, &owner) >= 0)
544 if (asprintf(&owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner) >= 0)
545 IOVEC_SET_STRING(iovec[n++], owner_uid);
546
547 #ifdef HAVE_SELINUX
548 if (label) {
549 selinux_context = malloc(sizeof("_SELINUX_CONTEXT=") + label_len);
550 if (selinux_context) {
551 memcpy(selinux_context, "_SELINUX_CONTEXT=", sizeof("_SELINUX_CONTEXT=")-1);
552 memcpy(selinux_context+sizeof("_SELINUX_CONTEXT=")-1, label, label_len);
553 selinux_context[sizeof("_SELINUX_CONTEXT=")-1+label_len] = 0;
554 IOVEC_SET_STRING(iovec[n++], selinux_context);
555 }
556 } else {
557 security_context_t con;
558
559 if (getpidcon(ucred->pid, &con) >= 0) {
560 selinux_context = strappend("_SELINUX_CONTEXT=", con);
561 if (selinux_context)
562 IOVEC_SET_STRING(iovec[n++], selinux_context);
563
564 freecon(con);
565 }
566 }
567 #endif
568 }
569
570 if (tv) {
571 if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
572 (unsigned long long) timeval_load(tv)) >= 0)
573 IOVEC_SET_STRING(iovec[n++], source_time);
574 }
575
576 /* Note that strictly speaking storing the boot id here is
577 * redundant since the entry includes this in-line
578 * anyway. However, we need this indexed, too. */
579 r = sd_id128_get_boot(&id);
580 if (r >= 0)
581 if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
582 IOVEC_SET_STRING(iovec[n++], boot_id);
583
584 r = sd_id128_get_machine(&id);
585 if (r >= 0)
586 if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
587 IOVEC_SET_STRING(iovec[n++], machine_id);
588
589 t = gethostname_malloc();
590 if (t) {
591 hostname = strappend("_HOSTNAME=", t);
592 free(t);
593 if (hostname)
594 IOVEC_SET_STRING(iovec[n++], hostname);
595 }
596
597 assert(n <= m);
598
599 server_flush_to_var(s);
600
601 retry:
602 f = find_journal(s, realuid == 0 ? 0 : loginuid);
603 if (!f)
604 log_warning("Dropping message, as we can't find a place to store the data.");
605 else {
606 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
607
608 if ((r == -EBADMSG || r == -E2BIG) && !vacuumed) {
609
610 if (r == -E2BIG)
611 log_info("Allocation limit reached, rotating.");
612 else
613 log_warning("Journal file corrupted, rotating.");
614
615 server_rotate(s);
616 server_vacuum(s);
617 vacuumed = true;
618
619 log_info("Retrying write.");
620 goto retry;
621 }
622
623 if (r < 0)
624 log_error("Failed to write entry, ignoring: %s", strerror(-r));
625 }
626
627 free(pid);
628 free(uid);
629 free(gid);
630 free(comm);
631 free(exe);
632 free(cmdline);
633 free(source_time);
634 free(boot_id);
635 free(machine_id);
636 free(hostname);
637 free(audit_session);
638 free(audit_loginuid);
639 free(cgroup);
640 free(session);
641 free(owner_uid);
642 free(unit);
643 free(selinux_context);
644 }
645
646 static void driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {
647 char mid[11 + 32 + 1];
648 char buffer[16 + LINE_MAX + 1];
649 struct iovec iovec[N_IOVEC_META_FIELDS + 4];
650 int n = 0;
651 va_list ap;
652 struct ucred ucred;
653
654 assert(s);
655 assert(format);
656
657 IOVEC_SET_STRING(iovec[n++], "PRIORITY=5");
658 IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver");
659
660 memcpy(buffer, "MESSAGE=", 8);
661 va_start(ap, format);
662 vsnprintf(buffer + 8, sizeof(buffer) - 8, format, ap);
663 va_end(ap);
664 char_array_0(buffer);
665 IOVEC_SET_STRING(iovec[n++], buffer);
666
667 snprintf(mid, sizeof(mid), "MESSAGE_ID=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(message_id));
668 char_array_0(mid);
669 IOVEC_SET_STRING(iovec[n++], mid);
670
671 zero(ucred);
672 ucred.pid = getpid();
673 ucred.uid = getuid();
674 ucred.gid = getgid();
675
676 dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0);
677 }
678
679 static void dispatch_message(Server *s,
680 struct iovec *iovec, unsigned n, unsigned m,
681 struct ucred *ucred,
682 struct timeval *tv,
683 const char *label, size_t label_len,
684 int priority) {
685 int rl;
686 char *path = NULL, *c;
687
688 assert(s);
689 assert(iovec || n == 0);
690
691 if (n == 0)
692 return;
693
694 if (!ucred)
695 goto finish;
696
697 path = shortened_cgroup_path(ucred->pid);
698 if (!path)
699 goto finish;
700
701 /* example: /user/lennart/3/foobar
702 * /system/dbus.service/foobar
703 *
704 * So let's cut of everything past the third /, since that is
705 * wher user directories start */
706
707 c = strchr(path, '/');
708 if (c) {
709 c = strchr(c+1, '/');
710 if (c) {
711 c = strchr(c+1, '/');
712 if (c)
713 *c = 0;
714 }
715 }
716
717 rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available_space(s));
718
719 if (rl == 0) {
720 free(path);
721 return;
722 }
723
724 /* Write a suppression message if we suppressed something */
725 if (rl > 1)
726 driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path);
727
728 free(path);
729
730 finish:
731 dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len);
732 }
733
734 static void forward_syslog_iovec(Server *s, const struct iovec *iovec, unsigned n_iovec, struct ucred *ucred, struct timeval *tv) {
735 struct msghdr msghdr;
736 struct cmsghdr *cmsg;
737 union {
738 struct cmsghdr cmsghdr;
739 uint8_t buf[CMSG_SPACE(sizeof(struct ucred))];
740 } control;
741 union sockaddr_union sa;
742
743 assert(s);
744 assert(iovec);
745 assert(n_iovec > 0);
746
747 zero(msghdr);
748 msghdr.msg_iov = (struct iovec*) iovec;
749 msghdr.msg_iovlen = n_iovec;
750
751 zero(sa);
752 sa.un.sun_family = AF_UNIX;
753 strncpy(sa.un.sun_path, "/run/systemd/journal/syslog", sizeof(sa.un.sun_path));
754 msghdr.msg_name = &sa;
755 msghdr.msg_namelen = offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path);
756
757 if (ucred) {
758 zero(control);
759 msghdr.msg_control = &control;
760 msghdr.msg_controllen = sizeof(control);
761
762 cmsg = CMSG_FIRSTHDR(&msghdr);
763 cmsg->cmsg_level = SOL_SOCKET;
764 cmsg->cmsg_type = SCM_CREDENTIALS;
765 cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
766 memcpy(CMSG_DATA(cmsg), ucred, sizeof(struct ucred));
767 msghdr.msg_controllen = cmsg->cmsg_len;
768 }
769
770 /* Forward the syslog message we received via /dev/log to
771 * /run/systemd/syslog. Unfortunately we currently can't set
772 * the SO_TIMESTAMP auxiliary data, and hence we don't. */
773
774 if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
775 return;
776
777 /* The socket is full? I guess the syslog implementation is
778 * too slow, and we shouldn't wait for that... */
779 if (errno == EAGAIN)
780 return;
781
782 if (ucred && errno == ESRCH) {
783 struct ucred u;
784
785 /* Hmm, presumably the sender process vanished
786 * by now, so let's fix it as good as we
787 * can, and retry */
788
789 u = *ucred;
790 u.pid = getpid();
791 memcpy(CMSG_DATA(cmsg), &u, sizeof(struct ucred));
792
793 if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
794 return;
795
796 if (errno == EAGAIN)
797 return;
798 }
799
800 log_debug("Failed to forward syslog message: %m");
801 }
802
803 static void forward_syslog_raw(Server *s, const char *buffer, struct ucred *ucred, struct timeval *tv) {
804 struct iovec iovec;
805
806 assert(s);
807 assert(buffer);
808
809 IOVEC_SET_STRING(iovec, buffer);
810 forward_syslog_iovec(s, &iovec, 1, ucred, tv);
811 }
812
813 static void forward_syslog(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred, struct timeval *tv) {
814 struct iovec iovec[5];
815 char header_priority[6], header_time[64], header_pid[16];
816 int n = 0;
817 time_t t;
818 struct tm *tm;
819 char *ident_buf = NULL;
820
821 assert(s);
822 assert(priority >= 0);
823 assert(priority <= 999);
824 assert(message);
825
826 /* First: priority field */
827 snprintf(header_priority, sizeof(header_priority), "<%i>", priority);
828 char_array_0(header_priority);
829 IOVEC_SET_STRING(iovec[n++], header_priority);
830
831 /* Second: timestamp */
832 t = tv ? tv->tv_sec : ((time_t) (now(CLOCK_REALTIME) / USEC_PER_SEC));
833 tm = localtime(&t);
834 if (!tm)
835 return;
836 if (strftime(header_time, sizeof(header_time), "%h %e %T ", tm) <= 0)
837 return;
838 IOVEC_SET_STRING(iovec[n++], header_time);
839
840 /* Third: identifier and PID */
841 if (ucred) {
842 if (!identifier) {
843 get_process_comm(ucred->pid, &ident_buf);
844 identifier = ident_buf;
845 }
846
847 snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid);
848 char_array_0(header_pid);
849
850 if (identifier)
851 IOVEC_SET_STRING(iovec[n++], identifier);
852
853 IOVEC_SET_STRING(iovec[n++], header_pid);
854 } else if (identifier) {
855 IOVEC_SET_STRING(iovec[n++], identifier);
856 IOVEC_SET_STRING(iovec[n++], ": ");
857 }
858
859 /* Fourth: message */
860 IOVEC_SET_STRING(iovec[n++], message);
861
862 forward_syslog_iovec(s, iovec, n, ucred, tv);
863
864 free(ident_buf);
865 }
866
867 static int fixup_priority(int priority) {
868
869 if ((priority & LOG_FACMASK) == 0)
870 return (priority & LOG_PRIMASK) | LOG_USER;
871
872 return priority;
873 }
874
875 static void forward_kmsg(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) {
876 struct iovec iovec[5];
877 char header_priority[6], header_pid[16];
878 int n = 0;
879 char *ident_buf = NULL;
880 int fd;
881
882 assert(s);
883 assert(priority >= 0);
884 assert(priority <= 999);
885 assert(message);
886
887 /* Never allow messages with kernel facility to be written to
888 * kmsg, regardless where the data comes from. */
889 priority = fixup_priority(priority);
890
891 /* First: priority field */
892 snprintf(header_priority, sizeof(header_priority), "<%i>", priority);
893 char_array_0(header_priority);
894 IOVEC_SET_STRING(iovec[n++], header_priority);
895
896 /* Second: identifier and PID */
897 if (ucred) {
898 if (!identifier) {
899 get_process_comm(ucred->pid, &ident_buf);
900 identifier = ident_buf;
901 }
902
903 snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid);
904 char_array_0(header_pid);
905
906 if (identifier)
907 IOVEC_SET_STRING(iovec[n++], identifier);
908
909 IOVEC_SET_STRING(iovec[n++], header_pid);
910 } else if (identifier) {
911 IOVEC_SET_STRING(iovec[n++], identifier);
912 IOVEC_SET_STRING(iovec[n++], ": ");
913 }
914
915 /* Fourth: message */
916 IOVEC_SET_STRING(iovec[n++], message);
917 IOVEC_SET_STRING(iovec[n++], "\n");
918
919 fd = open("/dev/kmsg", O_WRONLY|O_NOCTTY|O_CLOEXEC);
920 if (fd < 0) {
921 log_debug("Failed to open /dev/kmsg for logging: %s", strerror(errno));
922 goto finish;
923 }
924
925 if (writev(fd, iovec, n) < 0)
926 log_debug("Failed to write to /dev/kmsg for logging: %s", strerror(errno));
927
928 close_nointr_nofail(fd);
929
930 finish:
931 free(ident_buf);
932 }
933
934 static void forward_console(Server *s, const char *identifier, const char *message, struct ucred *ucred) {
935 struct iovec iovec[4];
936 char header_pid[16];
937 int n = 0, fd;
938 char *ident_buf = NULL;
939
940 assert(s);
941 assert(message);
942
943 /* First: identifier and PID */
944 if (ucred) {
945 if (!identifier) {
946 get_process_comm(ucred->pid, &ident_buf);
947 identifier = ident_buf;
948 }
949
950 snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid);
951 char_array_0(header_pid);
952
953 if (identifier)
954 IOVEC_SET_STRING(iovec[n++], identifier);
955
956 IOVEC_SET_STRING(iovec[n++], header_pid);
957 } else if (identifier) {
958 IOVEC_SET_STRING(iovec[n++], identifier);
959 IOVEC_SET_STRING(iovec[n++], ": ");
960 }
961
962 /* Third: message */
963 IOVEC_SET_STRING(iovec[n++], message);
964 IOVEC_SET_STRING(iovec[n++], "\n");
965
966 fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
967 if (fd < 0) {
968 log_debug("Failed to open /dev/console for logging: %s", strerror(errno));
969 goto finish;
970 }
971
972 if (writev(fd, iovec, n) < 0)
973 log_debug("Failed to write to /dev/console for logging: %s", strerror(errno));
974
975 close_nointr_nofail(fd);
976
977 finish:
978 free(ident_buf);
979 }
980
981 static void read_identifier(const char **buf, char **identifier, char **pid) {
982 const char *p;
983 char *t;
984 size_t l, e;
985
986 assert(buf);
987 assert(identifier);
988 assert(pid);
989
990 p = *buf;
991
992 p += strspn(p, WHITESPACE);
993 l = strcspn(p, WHITESPACE);
994
995 if (l <= 0 ||
996 p[l-1] != ':')
997 return;
998
999 e = l;
1000 l--;
1001
1002 if (p[l-1] == ']') {
1003 size_t k = l-1;
1004
1005 for (;;) {
1006
1007 if (p[k] == '[') {
1008 t = strndup(p+k+1, l-k-2);
1009 if (t)
1010 *pid = t;
1011
1012 l = k;
1013 break;
1014 }
1015
1016 if (k == 0)
1017 break;
1018
1019 k--;
1020 }
1021 }
1022
1023 t = strndup(p, l);
1024 if (t)
1025 *identifier = t;
1026
1027 *buf = p + e;
1028 *buf += strspn(*buf, WHITESPACE);
1029 }
1030
1031 static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv, const char *label, size_t label_len) {
1032 char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *syslog_pid = NULL;
1033 struct iovec iovec[N_IOVEC_META_FIELDS + 6];
1034 unsigned n = 0;
1035 int priority = LOG_USER | LOG_INFO;
1036 char *identifier = NULL, *pid = NULL;
1037
1038 assert(s);
1039 assert(buf);
1040
1041 if (s->forward_to_syslog)
1042 forward_syslog_raw(s, buf, ucred, tv);
1043
1044 parse_syslog_priority((char**) &buf, &priority);
1045 skip_syslog_date((char**) &buf);
1046 read_identifier(&buf, &identifier, &pid);
1047
1048 if (s->forward_to_kmsg)
1049 forward_kmsg(s, priority, identifier, buf, ucred);
1050
1051 if (s->forward_to_console)
1052 forward_console(s, identifier, buf, ucred);
1053
1054 IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=syslog");
1055
1056 if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
1057 IOVEC_SET_STRING(iovec[n++], syslog_priority);
1058
1059 if (priority & LOG_FACMASK)
1060 if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
1061 IOVEC_SET_STRING(iovec[n++], syslog_facility);
1062
1063 if (identifier) {
1064 syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier);
1065 if (syslog_identifier)
1066 IOVEC_SET_STRING(iovec[n++], syslog_identifier);
1067 }
1068
1069 if (pid) {
1070 syslog_pid = strappend("SYSLOG_PID=", pid);
1071 if (syslog_pid)
1072 IOVEC_SET_STRING(iovec[n++], syslog_pid);
1073 }
1074
1075 message = strappend("MESSAGE=", buf);
1076 if (message)
1077 IOVEC_SET_STRING(iovec[n++], message);
1078
1079 dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, label, label_len, priority);
1080
1081 free(message);
1082 free(identifier);
1083 free(pid);
1084 free(syslog_priority);
1085 free(syslog_facility);
1086 free(syslog_identifier);
1087 }
1088
1089 static bool valid_user_field(const char *p, size_t l) {
1090 const char *a;
1091
1092 /* We kinda enforce POSIX syntax recommendations for
1093 environment variables here, but make a couple of additional
1094 requirements.
1095
1096 http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
1097
1098 /* No empty field names */
1099 if (l <= 0)
1100 return false;
1101
1102 /* Don't allow names longer than 64 chars */
1103 if (l > 64)
1104 return false;
1105
1106 /* Variables starting with an underscore are protected */
1107 if (p[0] == '_')
1108 return false;
1109
1110 /* Don't allow digits as first character */
1111 if (p[0] >= '0' && p[0] <= '9')
1112 return false;
1113
1114 /* Only allow A-Z0-9 and '_' */
1115 for (a = p; a < p + l; a++)
1116 if (!((*a >= 'A' && *a <= 'Z') ||
1117 (*a >= '0' && *a <= '9') ||
1118 *a == '_'))
1119 return false;
1120
1121 return true;
1122 }
1123
1124 static void process_native_message(
1125 Server *s,
1126 const void *buffer, size_t buffer_size,
1127 struct ucred *ucred,
1128 struct timeval *tv,
1129 const char *label, size_t label_len) {
1130
1131 struct iovec *iovec = NULL;
1132 unsigned n = 0, m = 0, j, tn = (unsigned) -1;
1133 const char *p;
1134 size_t remaining;
1135 int priority = LOG_INFO;
1136 char *identifier = NULL, *message = NULL;
1137
1138 assert(s);
1139 assert(buffer || n == 0);
1140
1141 p = buffer;
1142 remaining = buffer_size;
1143
1144 while (remaining > 0) {
1145 const char *e, *q;
1146
1147 e = memchr(p, '\n', remaining);
1148
1149 if (!e) {
1150 /* Trailing noise, let's ignore it, and flush what we collected */
1151 log_debug("Received message with trailing noise, ignoring.");
1152 break;
1153 }
1154
1155 if (e == p) {
1156 /* Entry separator */
1157 dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, priority);
1158 n = 0;
1159 priority = LOG_INFO;
1160
1161 p++;
1162 remaining--;
1163 continue;
1164 }
1165
1166 if (*p == '.' || *p == '#') {
1167 /* Ignore control commands for now, and
1168 * comments too. */
1169 remaining -= (e - p) + 1;
1170 p = e + 1;
1171 continue;
1172 }
1173
1174 /* A property follows */
1175
1176 if (n+N_IOVEC_META_FIELDS >= m) {
1177 struct iovec *c;
1178 unsigned u;
1179
1180 u = MAX((n+N_IOVEC_META_FIELDS+1) * 2U, 4U);
1181 c = realloc(iovec, u * sizeof(struct iovec));
1182 if (!c) {
1183 log_error("Out of memory");
1184 break;
1185 }
1186
1187 iovec = c;
1188 m = u;
1189 }
1190
1191 q = memchr(p, '=', e - p);
1192 if (q) {
1193 if (valid_user_field(p, q - p)) {
1194 size_t l;
1195
1196 l = e - p;
1197
1198 /* If the field name starts with an
1199 * underscore, skip the variable,
1200 * since that indidates a trusted
1201 * field */
1202 iovec[n].iov_base = (char*) p;
1203 iovec[n].iov_len = l;
1204 n++;
1205
1206 /* We need to determine the priority
1207 * of this entry for the rate limiting
1208 * logic */
1209 if (l == 10 &&
1210 memcmp(p, "PRIORITY=", 9) == 0 &&
1211 p[9] >= '0' && p[9] <= '9')
1212 priority = (priority & LOG_FACMASK) | (p[9] - '0');
1213
1214 else if (l == 17 &&
1215 memcmp(p, "SYSLOG_FACILITY=", 16) == 0 &&
1216 p[16] >= '0' && p[16] <= '9')
1217 priority = (priority & LOG_PRIMASK) | ((p[16] - '0') << 3);
1218
1219 else if (l == 18 &&
1220 memcmp(p, "SYSLOG_FACILITY=", 16) == 0 &&
1221 p[16] >= '0' && p[16] <= '9' &&
1222 p[17] >= '0' && p[17] <= '9')
1223 priority = (priority & LOG_PRIMASK) | (((p[16] - '0')*10 + (p[17] - '0')) << 3);
1224
1225 else if (l >= 12 &&
1226 memcmp(p, "SYSLOG_IDENTIFIER=", 11) == 0) {
1227 char *t;
1228
1229 t = strndup(p + 11, l - 11);
1230 if (t) {
1231 free(identifier);
1232 identifier = t;
1233 }
1234 } else if (l >= 8 &&
1235 memcmp(p, "MESSAGE=", 8) == 0) {
1236 char *t;
1237
1238 t = strndup(p + 8, l - 8);
1239 if (t) {
1240 free(message);
1241 message = t;
1242 }
1243 }
1244 }
1245
1246 remaining -= (e - p) + 1;
1247 p = e + 1;
1248 continue;
1249 } else {
1250 uint64_t l;
1251 char *k;
1252
1253 if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
1254 log_debug("Failed to parse message, ignoring.");
1255 break;
1256 }
1257
1258 memcpy(&l, e + 1, sizeof(uint64_t));
1259 l = le64toh(l);
1260
1261 if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
1262 e[1+sizeof(uint64_t)+l] != '\n') {
1263 log_debug("Failed to parse message, ignoring.");
1264 break;
1265 }
1266
1267 k = malloc((e - p) + 1 + l);
1268 if (!k) {
1269 log_error("Out of memory");
1270 break;
1271 }
1272
1273 memcpy(k, p, e - p);
1274 k[e - p] = '=';
1275 memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
1276
1277 if (valid_user_field(p, e - p)) {
1278 iovec[n].iov_base = k;
1279 iovec[n].iov_len = (e - p) + 1 + l;
1280 n++;
1281 } else
1282 free(k);
1283
1284 remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
1285 p = e + 1 + sizeof(uint64_t) + l + 1;
1286 }
1287 }
1288
1289 if (n <= 0)
1290 goto finish;
1291
1292 tn = n++;
1293 IOVEC_SET_STRING(iovec[tn], "_TRANSPORT=journal");
1294
1295 if (message) {
1296 if (s->forward_to_syslog)
1297 forward_syslog(s, priority, identifier, message, ucred, tv);
1298
1299 if (s->forward_to_kmsg)
1300 forward_kmsg(s, priority, identifier, message, ucred);
1301
1302 if (s->forward_to_console)
1303 forward_console(s, identifier, message, ucred);
1304 }
1305
1306 dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, priority);
1307
1308 finish:
1309 for (j = 0; j < n; j++) {
1310 if (j == tn)
1311 continue;
1312
1313 if (iovec[j].iov_base < buffer ||
1314 (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
1315 free(iovec[j].iov_base);
1316 }
1317
1318 free(iovec);
1319 free(identifier);
1320 free(message);
1321 }
1322
1323 static void process_native_file(
1324 Server *s,
1325 int fd,
1326 struct ucred *ucred,
1327 struct timeval *tv,
1328 const char *label, size_t label_len) {
1329
1330 struct stat st;
1331 void *p;
1332 ssize_t n;
1333
1334 assert(s);
1335 assert(fd >= 0);
1336
1337 /* Data is in the passed file, since it didn't fit in a
1338 * datagram. We can't map the file here, since clients might
1339 * then truncate it and trigger a SIGBUS for us. So let's
1340 * stupidly read it */
1341
1342 if (fstat(fd, &st) < 0) {
1343 log_error("Failed to stat passed file, ignoring: %m");
1344 return;
1345 }
1346
1347 if (!S_ISREG(st.st_mode)) {
1348 log_error("File passed is not regular. Ignoring.");
1349 return;
1350 }
1351
1352 if (st.st_size <= 0)
1353 return;
1354
1355 if (st.st_size > ENTRY_SIZE_MAX) {
1356 log_error("File passed too large. Ignoring.");
1357 return;
1358 }
1359
1360 p = malloc(st.st_size);
1361 if (!p) {
1362 log_error("Out of memory");
1363 return;
1364 }
1365
1366 n = pread(fd, p, st.st_size, 0);
1367 if (n < 0)
1368 log_error("Failed to read file, ignoring: %s", strerror(-n));
1369 else if (n > 0)
1370 process_native_message(s, p, n, ucred, tv, label, label_len);
1371
1372 free(p);
1373 }
1374
1375 static int stdout_stream_log(StdoutStream *s, const char *p) {
1376 struct iovec iovec[N_IOVEC_META_FIELDS + 5];
1377 char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL, *syslog_identifier = NULL;
1378 unsigned n = 0;
1379 int priority;
1380 char *label = NULL;
1381 size_t label_len = 0;
1382
1383 assert(s);
1384 assert(p);
1385
1386 if (isempty(p))
1387 return 0;
1388
1389 priority = s->priority;
1390
1391 if (s->level_prefix)
1392 parse_syslog_priority((char**) &p, &priority);
1393
1394 if (s->forward_to_syslog || s->server->forward_to_syslog)
1395 forward_syslog(s->server, fixup_priority(priority), s->identifier, p, &s->ucred, NULL);
1396
1397 if (s->forward_to_kmsg || s->server->forward_to_kmsg)
1398 forward_kmsg(s->server, priority, s->identifier, p, &s->ucred);
1399
1400 if (s->forward_to_console || s->server->forward_to_console)
1401 forward_console(s->server, s->identifier, p, &s->ucred);
1402
1403 IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=stdout");
1404
1405 if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
1406 IOVEC_SET_STRING(iovec[n++], syslog_priority);
1407
1408 if (priority & LOG_FACMASK)
1409 if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
1410 IOVEC_SET_STRING(iovec[n++], syslog_facility);
1411
1412 if (s->identifier) {
1413 syslog_identifier = strappend("SYSLOG_IDENTIFIER=", s->identifier);
1414 if (syslog_identifier)
1415 IOVEC_SET_STRING(iovec[n++], syslog_identifier);
1416 }
1417
1418 message = strappend("MESSAGE=", p);
1419 if (message)
1420 IOVEC_SET_STRING(iovec[n++], message);
1421
1422 #ifdef HAVE_SELINUX
1423 if (s->security_context) {
1424 label = (char*) s->security_context;
1425 label_len = strlen((char*) s->security_context);
1426 }
1427 #endif
1428
1429 dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, label, label_len, priority);
1430
1431 free(message);
1432 free(syslog_priority);
1433 free(syslog_facility);
1434 free(syslog_identifier);
1435
1436 return 0;
1437 }
1438
1439 static int stdout_stream_line(StdoutStream *s, char *p) {
1440 int r;
1441
1442 assert(s);
1443 assert(p);
1444
1445 p = strstrip(p);
1446
1447 switch (s->state) {
1448
1449 case STDOUT_STREAM_IDENTIFIER:
1450 if (isempty(p))
1451 s->identifier = NULL;
1452 else {
1453 s->identifier = strdup(p);
1454 if (!s->identifier) {
1455 log_error("Out of memory");
1456 return -ENOMEM;
1457 }
1458 }
1459
1460 s->state = STDOUT_STREAM_PRIORITY;
1461 return 0;
1462
1463 case STDOUT_STREAM_PRIORITY:
1464 r = safe_atoi(p, &s->priority);
1465 if (r < 0 || s->priority <= 0 || s->priority >= 999) {
1466 log_warning("Failed to parse log priority line.");
1467 return -EINVAL;
1468 }
1469
1470 s->state = STDOUT_STREAM_LEVEL_PREFIX;
1471 return 0;
1472
1473 case STDOUT_STREAM_LEVEL_PREFIX:
1474 r = parse_boolean(p);
1475 if (r < 0) {
1476 log_warning("Failed to parse level prefix line.");
1477 return -EINVAL;
1478 }
1479
1480 s->level_prefix = !!r;
1481 s->state = STDOUT_STREAM_FORWARD_TO_SYSLOG;
1482 return 0;
1483
1484 case STDOUT_STREAM_FORWARD_TO_SYSLOG:
1485 r = parse_boolean(p);
1486 if (r < 0) {
1487 log_warning("Failed to parse forward to syslog line.");
1488 return -EINVAL;
1489 }
1490
1491 s->forward_to_syslog = !!r;
1492 s->state = STDOUT_STREAM_FORWARD_TO_KMSG;
1493 return 0;
1494
1495 case STDOUT_STREAM_FORWARD_TO_KMSG:
1496 r = parse_boolean(p);
1497 if (r < 0) {
1498 log_warning("Failed to parse copy to kmsg line.");
1499 return -EINVAL;
1500 }
1501
1502 s->forward_to_kmsg = !!r;
1503 s->state = STDOUT_STREAM_FORWARD_TO_CONSOLE;
1504 return 0;
1505
1506 case STDOUT_STREAM_FORWARD_TO_CONSOLE:
1507 r = parse_boolean(p);
1508 if (r < 0) {
1509 log_warning("Failed to parse copy to console line.");
1510 return -EINVAL;
1511 }
1512
1513 s->forward_to_console = !!r;
1514 s->state = STDOUT_STREAM_RUNNING;
1515 return 0;
1516
1517 case STDOUT_STREAM_RUNNING:
1518 return stdout_stream_log(s, p);
1519 }
1520
1521 assert_not_reached("Unknown stream state");
1522 }
1523
1524 static int stdout_stream_scan(StdoutStream *s, bool force_flush) {
1525 char *p;
1526 size_t remaining;
1527 int r;
1528
1529 assert(s);
1530
1531 p = s->buffer;
1532 remaining = s->length;
1533 for (;;) {
1534 char *end;
1535 size_t skip;
1536
1537 end = memchr(p, '\n', remaining);
1538 if (end)
1539 skip = end - p + 1;
1540 else if (remaining >= sizeof(s->buffer) - 1) {
1541 end = p + sizeof(s->buffer) - 1;
1542 skip = remaining;
1543 } else
1544 break;
1545
1546 *end = 0;
1547
1548 r = stdout_stream_line(s, p);
1549 if (r < 0)
1550 return r;
1551
1552 remaining -= skip;
1553 p += skip;
1554 }
1555
1556 if (force_flush && remaining > 0) {
1557 p[remaining] = 0;
1558 r = stdout_stream_line(s, p);
1559 if (r < 0)
1560 return r;
1561
1562 p += remaining;
1563 remaining = 0;
1564 }
1565
1566 if (p > s->buffer) {
1567 memmove(s->buffer, p, remaining);
1568 s->length = remaining;
1569 }
1570
1571 return 0;
1572 }
1573
1574 static int stdout_stream_process(StdoutStream *s) {
1575 ssize_t l;
1576 int r;
1577
1578 assert(s);
1579
1580 l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length);
1581 if (l < 0) {
1582
1583 if (errno == EAGAIN)
1584 return 0;
1585
1586 log_warning("Failed to read from stream: %m");
1587 return -errno;
1588 }
1589
1590 if (l == 0) {
1591 r = stdout_stream_scan(s, true);
1592 if (r < 0)
1593 return r;
1594
1595 return 0;
1596 }
1597
1598 s->length += l;
1599 r = stdout_stream_scan(s, false);
1600 if (r < 0)
1601 return r;
1602
1603 return 1;
1604
1605 }
1606
1607 static void stdout_stream_free(StdoutStream *s) {
1608 assert(s);
1609
1610 if (s->server) {
1611 assert(s->server->n_stdout_streams > 0);
1612 s->server->n_stdout_streams --;
1613 LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s);
1614 }
1615
1616 if (s->fd >= 0) {
1617 if (s->server)
1618 epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL);
1619
1620 close_nointr_nofail(s->fd);
1621 }
1622
1623 #ifdef HAVE_SELINUX
1624 if (s->security_context)
1625 freecon(s->security_context);
1626 #endif
1627
1628 free(s->identifier);
1629 free(s);
1630 }
1631
1632 static int stdout_stream_new(Server *s) {
1633 StdoutStream *stream;
1634 int fd, r;
1635 socklen_t len;
1636 struct epoll_event ev;
1637
1638 assert(s);
1639
1640 fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC);
1641 if (fd < 0) {
1642 if (errno == EAGAIN)
1643 return 0;
1644
1645 log_error("Failed to accept stdout connection: %m");
1646 return -errno;
1647 }
1648
1649 if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) {
1650 log_warning("Too many stdout streams, refusing connection.");
1651 close_nointr_nofail(fd);
1652 return 0;
1653 }
1654
1655 stream = new0(StdoutStream, 1);
1656 if (!stream) {
1657 log_error("Out of memory.");
1658 close_nointr_nofail(fd);
1659 return -ENOMEM;
1660 }
1661
1662 stream->fd = fd;
1663
1664 len = sizeof(stream->ucred);
1665 if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) {
1666 log_error("Failed to determine peer credentials: %m");
1667 r = -errno;
1668 goto fail;
1669 }
1670
1671 #ifdef HAVE_SELINUX
1672 if (getpeercon(fd, &stream->security_context) < 0)
1673 log_error("Failed to determine peer security context.");
1674 #endif
1675
1676 if (shutdown(fd, SHUT_WR) < 0) {
1677 log_error("Failed to shutdown writing side of socket: %m");
1678 r = -errno;
1679 goto fail;
1680 }
1681
1682 zero(ev);
1683 ev.data.ptr = stream;
1684 ev.events = EPOLLIN;
1685 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) {
1686 log_error("Failed to add stream to event loop: %m");
1687 r = -errno;
1688 goto fail;
1689 }
1690
1691 stream->server = s;
1692 LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream);
1693 s->n_stdout_streams ++;
1694
1695 return 0;
1696
1697 fail:
1698 stdout_stream_free(stream);
1699 return r;
1700 }
1701
1702 static int parse_kernel_timestamp(char **_p, usec_t *t) {
1703 usec_t r;
1704 int k, i;
1705 char *p;
1706
1707 assert(_p);
1708 assert(*_p);
1709 assert(t);
1710
1711 p = *_p;
1712
1713 if (strlen(p) < 14 || p[0] != '[' || p[13] != ']' || p[6] != '.')
1714 return 0;
1715
1716 r = 0;
1717
1718 for (i = 1; i <= 5; i++) {
1719 r *= 10;
1720
1721 if (p[i] == ' ')
1722 continue;
1723
1724 k = undecchar(p[i]);
1725 if (k < 0)
1726 return 0;
1727
1728 r += k;
1729 }
1730
1731 for (i = 7; i <= 12; i++) {
1732 r *= 10;
1733
1734 k = undecchar(p[i]);
1735 if (k < 0)
1736 return 0;
1737
1738 r += k;
1739 }
1740
1741 *t = r;
1742 *_p += 14;
1743 *_p += strspn(*_p, WHITESPACE);
1744
1745 return 1;
1746 }
1747
1748 static void proc_kmsg_line(Server *s, const char *p) {
1749 struct iovec iovec[N_IOVEC_META_FIELDS + 7];
1750 char *message = NULL, *syslog_priority = NULL, *syslog_pid = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *source_time = NULL;
1751 int priority = LOG_KERN | LOG_INFO;
1752 unsigned n = 0;
1753 usec_t usec;
1754 char *identifier = NULL, *pid = NULL;
1755
1756 assert(s);
1757 assert(p);
1758
1759 if (isempty(p))
1760 return;
1761
1762 parse_syslog_priority((char **) &p, &priority);
1763
1764 if (s->forward_to_kmsg && (priority & LOG_FACMASK) != LOG_KERN)
1765 return;
1766
1767 if (parse_kernel_timestamp((char **) &p, &usec) > 0) {
1768 if (asprintf(&source_time, "_SOURCE_MONOTONIC_TIMESTAMP=%llu",
1769 (unsigned long long) usec) >= 0)
1770 IOVEC_SET_STRING(iovec[n++], source_time);
1771 }
1772
1773 IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=kernel");
1774
1775 if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
1776 IOVEC_SET_STRING(iovec[n++], syslog_priority);
1777
1778 if ((priority & LOG_FACMASK) == LOG_KERN) {
1779
1780 if (s->forward_to_syslog)
1781 forward_syslog(s, priority, "kernel", p, NULL, NULL);
1782
1783 IOVEC_SET_STRING(iovec[n++], "SYSLOG_IDENTIFIER=kernel");
1784 } else {
1785 read_identifier(&p, &identifier, &pid);
1786
1787 if (s->forward_to_syslog)
1788 forward_syslog(s, priority, identifier, p, NULL, NULL);
1789
1790 if (identifier) {
1791 syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier);
1792 if (syslog_identifier)
1793 IOVEC_SET_STRING(iovec[n++], syslog_identifier);
1794 }
1795
1796 if (pid) {
1797 syslog_pid = strappend("SYSLOG_PID=", pid);
1798 if (syslog_pid)
1799 IOVEC_SET_STRING(iovec[n++], syslog_pid);
1800 }
1801
1802 if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
1803 IOVEC_SET_STRING(iovec[n++], syslog_facility);
1804 }
1805
1806 message = strappend("MESSAGE=", p);
1807 if (message)
1808 IOVEC_SET_STRING(iovec[n++], message);
1809
1810 dispatch_message(s, iovec, n, ELEMENTSOF(iovec), NULL, NULL, NULL, 0, priority);
1811
1812 free(message);
1813 free(syslog_priority);
1814 free(syslog_identifier);
1815 free(syslog_pid);
1816 free(syslog_facility);
1817 free(source_time);
1818 free(identifier);
1819 free(pid);
1820 }
1821
1822 static void proc_kmsg_scan(Server *s) {
1823 char *p;
1824 size_t remaining;
1825
1826 assert(s);
1827
1828 p = s->proc_kmsg_buffer;
1829 remaining = s->proc_kmsg_length;
1830 for (;;) {
1831 char *end;
1832 size_t skip;
1833
1834 end = memchr(p, '\n', remaining);
1835 if (end)
1836 skip = end - p + 1;
1837 else if (remaining >= sizeof(s->proc_kmsg_buffer) - 1) {
1838 end = p + sizeof(s->proc_kmsg_buffer) - 1;
1839 skip = remaining;
1840 } else
1841 break;
1842
1843 *end = 0;
1844
1845 proc_kmsg_line(s, p);
1846
1847 remaining -= skip;
1848 p += skip;
1849 }
1850
1851 if (p > s->proc_kmsg_buffer) {
1852 memmove(s->proc_kmsg_buffer, p, remaining);
1853 s->proc_kmsg_length = remaining;
1854 }
1855 }
1856
1857 static int system_journal_open(Server *s) {
1858 int r;
1859 char *fn;
1860 sd_id128_t machine;
1861 char ids[33];
1862
1863 r = sd_id128_get_machine(&machine);
1864 if (r < 0)
1865 return r;
1866
1867 sd_id128_to_string(machine, ids);
1868
1869 if (!s->system_journal) {
1870
1871 /* First try to create the machine path, but not the prefix */
1872 fn = strappend("/var/log/journal/", ids);
1873 if (!fn)
1874 return -ENOMEM;
1875 (void) mkdir(fn, 0755);
1876 free(fn);
1877
1878 /* The create the system journal file */
1879 fn = join("/var/log/journal/", ids, "/system.journal", NULL);
1880 if (!fn)
1881 return -ENOMEM;
1882
1883 r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
1884 free(fn);
1885
1886 if (r >= 0) {
1887 journal_default_metrics(&s->system_metrics, s->system_journal->fd);
1888
1889 s->system_journal->metrics = s->system_metrics;
1890 s->system_journal->compress = s->compress;
1891
1892 server_fix_perms(s, s->system_journal, 0);
1893 } else if (r < 0) {
1894
1895 if (r != -ENOENT && r != -EROFS)
1896 log_warning("Failed to open system journal: %s", strerror(-r));
1897
1898 r = 0;
1899 }
1900 }
1901
1902 if (!s->runtime_journal) {
1903
1904 fn = join("/run/log/journal/", ids, "/system.journal", NULL);
1905 if (!fn)
1906 return -ENOMEM;
1907
1908 if (s->system_journal) {
1909
1910 /* Try to open the runtime journal, but only
1911 * if it already exists, so that we can flush
1912 * it into the system journal */
1913
1914 r = journal_file_open_reliably(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
1915 free(fn);
1916
1917 if (r < 0) {
1918 if (r != -ENOENT)
1919 log_warning("Failed to open runtime journal: %s", strerror(-r));
1920
1921 r = 0;
1922 }
1923
1924 } else {
1925
1926 /* OK, we really need the runtime journal, so create
1927 * it if necessary. */
1928
1929 (void) mkdir_parents(fn, 0755);
1930 r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
1931 free(fn);
1932
1933 if (r < 0) {
1934 log_error("Failed to open runtime journal: %s", strerror(-r));
1935 return r;
1936 }
1937 }
1938
1939 if (s->runtime_journal) {
1940 journal_default_metrics(&s->runtime_metrics, s->runtime_journal->fd);
1941
1942 s->runtime_journal->metrics = s->runtime_metrics;
1943 s->runtime_journal->compress = s->compress;
1944
1945 server_fix_perms(s, s->runtime_journal, 0);
1946 }
1947 }
1948
1949 return r;
1950 }
1951
1952 static int server_flush_to_var(Server *s) {
1953 char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
1954 Object *o = NULL;
1955 int r;
1956 sd_id128_t machine;
1957 sd_journal *j;
1958 usec_t ts;
1959
1960 assert(s);
1961
1962 if (!s->runtime_journal)
1963 return 0;
1964
1965 ts = now(CLOCK_MONOTONIC);
1966 if (s->var_available_timestamp + RECHECK_VAR_AVAILABLE_USEC > ts)
1967 return 0;
1968
1969 s->var_available_timestamp = ts;
1970
1971 system_journal_open(s);
1972
1973 if (!s->system_journal)
1974 return 0;
1975
1976 log_info("Flushing to /var...");
1977
1978 r = sd_id128_get_machine(&machine);
1979 if (r < 0) {
1980 log_error("Failed to get machine id: %s", strerror(-r));
1981 return r;
1982 }
1983
1984 r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);
1985 if (r < 0) {
1986 log_error("Failed to read runtime journal: %s", strerror(-r));
1987 return r;
1988 }
1989
1990 SD_JOURNAL_FOREACH(j) {
1991 JournalFile *f;
1992
1993 f = j->current_file;
1994 assert(f && f->current_offset > 0);
1995
1996 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1997 if (r < 0) {
1998 log_error("Can't read entry: %s", strerror(-r));
1999 goto finish;
2000 }
2001
2002 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
2003 if (r == -E2BIG) {
2004 log_info("Allocation limit reached.");
2005
2006 journal_file_post_change(s->system_journal);
2007 server_rotate(s);
2008 server_vacuum(s);
2009
2010 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
2011 }
2012
2013 if (r < 0) {
2014 log_error("Can't write entry: %s", strerror(-r));
2015 goto finish;
2016 }
2017 }
2018
2019 finish:
2020 journal_file_post_change(s->system_journal);
2021
2022 journal_file_close(s->runtime_journal);
2023 s->runtime_journal = NULL;
2024
2025 if (r >= 0) {
2026 sd_id128_to_string(machine, path + 17);
2027 rm_rf(path, false, true, false);
2028 }
2029
2030 return r;
2031 }
2032
2033 static int server_read_proc_kmsg(Server *s) {
2034 ssize_t l;
2035 assert(s);
2036 assert(s->proc_kmsg_fd >= 0);
2037
2038 l = read(s->proc_kmsg_fd, s->proc_kmsg_buffer + s->proc_kmsg_length, sizeof(s->proc_kmsg_buffer) - 1 - s->proc_kmsg_length);
2039 if (l < 0) {
2040
2041 if (errno == EAGAIN || errno == EINTR)
2042 return 0;
2043
2044 log_error("Failed to read from kernel: %m");
2045 return -errno;
2046 }
2047
2048 s->proc_kmsg_length += l;
2049
2050 proc_kmsg_scan(s);
2051 return 1;
2052 }
2053
2054 static int server_flush_proc_kmsg(Server *s) {
2055 int r;
2056
2057 assert(s);
2058
2059 if (s->proc_kmsg_fd < 0)
2060 return 0;
2061
2062 log_info("Flushing /proc/kmsg...");
2063
2064 for (;;) {
2065 r = server_read_proc_kmsg(s);
2066 if (r < 0)
2067 return r;
2068
2069 if (r == 0)
2070 break;
2071 }
2072
2073 return 0;
2074 }
2075
2076 static int process_event(Server *s, struct epoll_event *ev) {
2077 assert(s);
2078
2079 if (ev->data.fd == s->signal_fd) {
2080 struct signalfd_siginfo sfsi;
2081 ssize_t n;
2082
2083 if (ev->events != EPOLLIN) {
2084 log_info("Got invalid event from epoll.");
2085 return -EIO;
2086 }
2087
2088 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
2089 if (n != sizeof(sfsi)) {
2090
2091 if (n >= 0)
2092 return -EIO;
2093
2094 if (errno == EINTR || errno == EAGAIN)
2095 return 1;
2096
2097 return -errno;
2098 }
2099
2100 if (sfsi.ssi_signo == SIGUSR1) {
2101 server_flush_to_var(s);
2102 return 0;
2103 }
2104
2105 log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
2106 return 0;
2107
2108 } else if (ev->data.fd == s->proc_kmsg_fd) {
2109 int r;
2110
2111 if (ev->events != EPOLLIN) {
2112 log_info("Got invalid event from epoll.");
2113 return -EIO;
2114 }
2115
2116 r = server_read_proc_kmsg(s);
2117 if (r < 0)
2118 return r;
2119
2120 return 1;
2121
2122 } else if (ev->data.fd == s->native_fd ||
2123 ev->data.fd == s->syslog_fd) {
2124
2125 if (ev->events != EPOLLIN) {
2126 log_info("Got invalid event from epoll.");
2127 return -EIO;
2128 }
2129
2130 for (;;) {
2131 struct msghdr msghdr;
2132 struct iovec iovec;
2133 struct ucred *ucred = NULL;
2134 struct timeval *tv = NULL;
2135 struct cmsghdr *cmsg;
2136 char *label = NULL;
2137 size_t label_len = 0;
2138 union {
2139 struct cmsghdr cmsghdr;
2140 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
2141 CMSG_SPACE(sizeof(struct timeval)) +
2142 CMSG_SPACE(sizeof(int)) +
2143 CMSG_SPACE(PAGE_SIZE)]; /* selinux label */
2144 } control;
2145 ssize_t n;
2146 int v;
2147 int *fds = NULL;
2148 unsigned n_fds = 0;
2149
2150 if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
2151 log_error("SIOCINQ failed: %m");
2152 return -errno;
2153 }
2154
2155 if (s->buffer_size < (size_t) v) {
2156 void *b;
2157 size_t l;
2158
2159 l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
2160 b = realloc(s->buffer, l+1);
2161
2162 if (!b) {
2163 log_error("Couldn't increase buffer.");
2164 return -ENOMEM;
2165 }
2166
2167 s->buffer_size = l;
2168 s->buffer = b;
2169 }
2170
2171 zero(iovec);
2172 iovec.iov_base = s->buffer;
2173 iovec.iov_len = s->buffer_size;
2174
2175 zero(control);
2176 zero(msghdr);
2177 msghdr.msg_iov = &iovec;
2178 msghdr.msg_iovlen = 1;
2179 msghdr.msg_control = &control;
2180 msghdr.msg_controllen = sizeof(control);
2181
2182 n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
2183 if (n < 0) {
2184
2185 if (errno == EINTR || errno == EAGAIN)
2186 return 1;
2187
2188 log_error("recvmsg() failed: %m");
2189 return -errno;
2190 }
2191
2192 for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {
2193
2194 if (cmsg->cmsg_level == SOL_SOCKET &&
2195 cmsg->cmsg_type == SCM_CREDENTIALS &&
2196 cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))
2197 ucred = (struct ucred*) CMSG_DATA(cmsg);
2198 else if (cmsg->cmsg_level == SOL_SOCKET &&
2199 cmsg->cmsg_type == SCM_SECURITY) {
2200 label = (char*) CMSG_DATA(cmsg);
2201 label_len = cmsg->cmsg_len - CMSG_LEN(0);
2202 } else if (cmsg->cmsg_level == SOL_SOCKET &&
2203 cmsg->cmsg_type == SO_TIMESTAMP &&
2204 cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))
2205 tv = (struct timeval*) CMSG_DATA(cmsg);
2206 else if (cmsg->cmsg_level == SOL_SOCKET &&
2207 cmsg->cmsg_type == SCM_RIGHTS) {
2208 fds = (int*) CMSG_DATA(cmsg);
2209 n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int);
2210 }
2211 }
2212
2213 if (ev->data.fd == s->syslog_fd) {
2214 char *e;
2215
2216 if (n > 0 && n_fds == 0) {
2217 e = memchr(s->buffer, '\n', n);
2218 if (e)
2219 *e = 0;
2220 else
2221 s->buffer[n] = 0;
2222
2223 process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len);
2224 } else if (n_fds > 0)
2225 log_warning("Got file descriptors via syslog socket. Ignoring.");
2226
2227 } else {
2228 if (n > 0 && n_fds == 0)
2229 process_native_message(s, s->buffer, n, ucred, tv, label, label_len);
2230 else if (n == 0 && n_fds == 1)
2231 process_native_file(s, fds[0], ucred, tv, label, label_len);
2232 else if (n_fds > 0)
2233 log_warning("Got too many file descriptors via native socket. Ignoring.");
2234 }
2235
2236 close_many(fds, n_fds);
2237 }
2238
2239 return 1;
2240
2241 } else if (ev->data.fd == s->stdout_fd) {
2242
2243 if (ev->events != EPOLLIN) {
2244 log_info("Got invalid event from epoll.");
2245 return -EIO;
2246 }
2247
2248 stdout_stream_new(s);
2249 return 1;
2250
2251 } else {
2252 StdoutStream *stream;
2253
2254 if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
2255 log_info("Got invalid event from epoll.");
2256 return -EIO;
2257 }
2258
2259 /* If it is none of the well-known fds, it must be an
2260 * stdout stream fd. Note that this is a bit ugly here
2261 * (since we rely that none of the well-known fds
2262 * could be interpreted as pointer), but nonetheless
2263 * safe, since the well-known fds would never get an
2264 * fd > 4096, i.e. beyond the first memory page */
2265
2266 stream = ev->data.ptr;
2267
2268 if (stdout_stream_process(stream) <= 0)
2269 stdout_stream_free(stream);
2270
2271 return 1;
2272 }
2273
2274 log_error("Unknown event.");
2275 return 0;
2276 }
2277
2278 static int open_syslog_socket(Server *s) {
2279 union sockaddr_union sa;
2280 int one, r;
2281 struct epoll_event ev;
2282
2283 assert(s);
2284
2285 if (s->syslog_fd < 0) {
2286
2287 s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
2288 if (s->syslog_fd < 0) {
2289 log_error("socket() failed: %m");
2290 return -errno;
2291 }
2292
2293 zero(sa);
2294 sa.un.sun_family = AF_UNIX;
2295 strncpy(sa.un.sun_path, "/dev/log", sizeof(sa.un.sun_path));
2296
2297 unlink(sa.un.sun_path);
2298
2299 r = bind(s->syslog_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
2300 if (r < 0) {
2301 log_error("bind() failed: %m");
2302 return -errno;
2303 }
2304
2305 chmod(sa.un.sun_path, 0666);
2306 } else
2307 fd_nonblock(s->syslog_fd, 1);
2308
2309 one = 1;
2310 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
2311 if (r < 0) {
2312 log_error("SO_PASSCRED failed: %m");
2313 return -errno;
2314 }
2315
2316 #ifdef HAVE_SELINUX
2317 one = 1;
2318 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
2319 if (r < 0)
2320 log_warning("SO_PASSSEC failed: %m");
2321 #endif
2322
2323 one = 1;
2324 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
2325 if (r < 0) {
2326 log_error("SO_TIMESTAMP failed: %m");
2327 return -errno;
2328 }
2329
2330 zero(ev);
2331 ev.events = EPOLLIN;
2332 ev.data.fd = s->syslog_fd;
2333 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->syslog_fd, &ev) < 0) {
2334 log_error("Failed to add syslog server fd to epoll object: %m");
2335 return -errno;
2336 }
2337
2338 return 0;
2339 }
2340
2341 static int open_native_socket(Server*s) {
2342 union sockaddr_union sa;
2343 int one, r;
2344 struct epoll_event ev;
2345
2346 assert(s);
2347
2348 if (s->native_fd < 0) {
2349
2350 s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
2351 if (s->native_fd < 0) {
2352 log_error("socket() failed: %m");
2353 return -errno;
2354 }
2355
2356 zero(sa);
2357 sa.un.sun_family = AF_UNIX;
2358 strncpy(sa.un.sun_path, "/run/systemd/journal/socket", sizeof(sa.un.sun_path));
2359
2360 unlink(sa.un.sun_path);
2361
2362 r = bind(s->native_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
2363 if (r < 0) {
2364 log_error("bind() failed: %m");
2365 return -errno;
2366 }
2367
2368 chmod(sa.un.sun_path, 0666);
2369 } else
2370 fd_nonblock(s->native_fd, 1);
2371
2372 one = 1;
2373 r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
2374 if (r < 0) {
2375 log_error("SO_PASSCRED failed: %m");
2376 return -errno;
2377 }
2378
2379 #ifdef HAVE_SELINUX
2380 one = 1;
2381 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
2382 if (r < 0)
2383 log_warning("SO_PASSSEC failed: %m");
2384 #endif
2385
2386 one = 1;
2387 r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
2388 if (r < 0) {
2389 log_error("SO_TIMESTAMP failed: %m");
2390 return -errno;
2391 }
2392
2393 zero(ev);
2394 ev.events = EPOLLIN;
2395 ev.data.fd = s->native_fd;
2396 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->native_fd, &ev) < 0) {
2397 log_error("Failed to add native server fd to epoll object: %m");
2398 return -errno;
2399 }
2400
2401 return 0;
2402 }
2403
2404 static int open_stdout_socket(Server *s) {
2405 union sockaddr_union sa;
2406 int r;
2407 struct epoll_event ev;
2408
2409 assert(s);
2410
2411 if (s->stdout_fd < 0) {
2412
2413 s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
2414 if (s->stdout_fd < 0) {
2415 log_error("socket() failed: %m");
2416 return -errno;
2417 }
2418
2419 zero(sa);
2420 sa.un.sun_family = AF_UNIX;
2421 strncpy(sa.un.sun_path, "/run/systemd/journal/stdout", sizeof(sa.un.sun_path));
2422
2423 unlink(sa.un.sun_path);
2424
2425 r = bind(s->stdout_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
2426 if (r < 0) {
2427 log_error("bind() failed: %m");
2428 return -errno;
2429 }
2430
2431 chmod(sa.un.sun_path, 0666);
2432
2433 if (listen(s->stdout_fd, SOMAXCONN) < 0) {
2434 log_error("liste() failed: %m");
2435 return -errno;
2436 }
2437 } else
2438 fd_nonblock(s->stdout_fd, 1);
2439
2440 zero(ev);
2441 ev.events = EPOLLIN;
2442 ev.data.fd = s->stdout_fd;
2443 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->stdout_fd, &ev) < 0) {
2444 log_error("Failed to add stdout server fd to epoll object: %m");
2445 return -errno;
2446 }
2447
2448 return 0;
2449 }
2450
2451 static int open_proc_kmsg(Server *s) {
2452 struct epoll_event ev;
2453
2454 assert(s);
2455
2456 if (!s->import_proc_kmsg)
2457 return 0;
2458
2459
2460 s->proc_kmsg_fd = open("/proc/kmsg", O_CLOEXEC|O_NONBLOCK|O_NOCTTY);
2461 if (s->proc_kmsg_fd < 0) {
2462 log_warning("Failed to open /proc/kmsg, ignoring: %m");
2463 return 0;
2464 }
2465
2466 zero(ev);
2467 ev.events = EPOLLIN;
2468 ev.data.fd = s->proc_kmsg_fd;
2469 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->proc_kmsg_fd, &ev) < 0) {
2470 log_error("Failed to add /proc/kmsg fd to epoll object: %m");
2471 return -errno;
2472 }
2473
2474 return 0;
2475 }
2476
2477 static int open_signalfd(Server *s) {
2478 sigset_t mask;
2479 struct epoll_event ev;
2480
2481 assert(s);
2482
2483 assert_se(sigemptyset(&mask) == 0);
2484 sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, -1);
2485 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
2486
2487 s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
2488 if (s->signal_fd < 0) {
2489 log_error("signalfd(): %m");
2490 return -errno;
2491 }
2492
2493 zero(ev);
2494 ev.events = EPOLLIN;
2495 ev.data.fd = s->signal_fd;
2496
2497 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
2498 log_error("epoll_ctl(): %m");
2499 return -errno;
2500 }
2501
2502 return 0;
2503 }
2504
2505 static int server_parse_proc_cmdline(Server *s) {
2506 char *line, *w, *state;
2507 int r;
2508 size_t l;
2509
2510 if (detect_container(NULL) > 0)
2511 return 0;
2512
2513 r = read_one_line_file("/proc/cmdline", &line);
2514 if (r < 0) {
2515 log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));
2516 return 0;
2517 }
2518
2519 FOREACH_WORD_QUOTED(w, l, line, state) {
2520 char *word;
2521
2522 word = strndup(w, l);
2523 if (!word) {
2524 r = -ENOMEM;
2525 goto finish;
2526 }
2527
2528 if (startswith(word, "systemd_journald.forward_to_syslog=")) {
2529 r = parse_boolean(word + 35);
2530 if (r < 0)
2531 log_warning("Failed to parse forward to syslog switch %s. Ignoring.", word + 35);
2532 else
2533 s->forward_to_syslog = r;
2534 } else if (startswith(word, "systemd_journald.forward_to_kmsg=")) {
2535 r = parse_boolean(word + 33);
2536 if (r < 0)
2537 log_warning("Failed to parse forward to kmsg switch %s. Ignoring.", word + 33);
2538 else
2539 s->forward_to_kmsg = r;
2540 } else if (startswith(word, "systemd_journald.forward_to_console=")) {
2541 r = parse_boolean(word + 36);
2542 if (r < 0)
2543 log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36);
2544 else
2545 s->forward_to_console = r;
2546 }
2547
2548 free(word);
2549 }
2550
2551 r = 0;
2552
2553 finish:
2554 free(line);
2555 return r;
2556 }
2557
2558 static int server_parse_config_file(Server *s) {
2559 FILE *f;
2560 const char *fn;
2561 int r;
2562
2563 assert(s);
2564
2565 fn = "/etc/systemd/systemd-journald.conf";
2566 f = fopen(fn, "re");
2567 if (!f) {
2568 if (errno == ENOENT)
2569 return 0;
2570
2571 log_warning("Failed to open configuration file %s: %m", fn);
2572 return -errno;
2573 }
2574
2575 r = config_parse(fn, f, "Journal\0", config_item_perf_lookup, (void*) journald_gperf_lookup, false, s);
2576 if (r < 0)
2577 log_warning("Failed to parse configuration file: %s", strerror(-r));
2578
2579 fclose(f);
2580
2581 return r;
2582 }
2583
2584 static int server_init(Server *s) {
2585 int n, r, fd;
2586
2587 assert(s);
2588
2589 zero(*s);
2590 s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->proc_kmsg_fd = -1;
2591 s->compress = true;
2592
2593 s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL;
2594 s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST;
2595
2596 s->forward_to_syslog = true;
2597 s->import_proc_kmsg = true;
2598
2599 memset(&s->system_metrics, 0xFF, sizeof(s->system_metrics));
2600 memset(&s->runtime_metrics, 0xFF, sizeof(s->runtime_metrics));
2601
2602 server_parse_config_file(s);
2603 server_parse_proc_cmdline(s);
2604
2605 s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
2606 if (!s->user_journals) {
2607 log_error("Out of memory.");
2608 return -ENOMEM;
2609 }
2610
2611 s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
2612 if (s->epoll_fd < 0) {
2613 log_error("Failed to create epoll object: %m");
2614 return -errno;
2615 }
2616
2617 n = sd_listen_fds(true);
2618 if (n < 0) {
2619 log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
2620 return n;
2621 }
2622
2623 for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
2624
2625 if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) {
2626
2627 if (s->native_fd >= 0) {
2628 log_error("Too many native sockets passed.");
2629 return -EINVAL;
2630 }
2631
2632 s->native_fd = fd;
2633
2634 } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) {
2635
2636 if (s->stdout_fd >= 0) {
2637 log_error("Too many stdout sockets passed.");
2638 return -EINVAL;
2639 }
2640
2641 s->stdout_fd = fd;
2642
2643 } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
2644
2645 if (s->syslog_fd >= 0) {
2646 log_error("Too many /dev/log sockets passed.");
2647 return -EINVAL;
2648 }
2649
2650 s->syslog_fd = fd;
2651
2652 } else {
2653 log_error("Unknown socket passed.");
2654 return -EINVAL;
2655 }
2656 }
2657
2658 r = open_syslog_socket(s);
2659 if (r < 0)
2660 return r;
2661
2662 r = open_native_socket(s);
2663 if (r < 0)
2664 return r;
2665
2666 r = open_stdout_socket(s);
2667 if (r < 0)
2668 return r;
2669
2670 r = open_proc_kmsg(s);
2671 if (r < 0)
2672 return r;
2673
2674 r = open_signalfd(s);
2675 if (r < 0)
2676 return r;
2677
2678 s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst);
2679 if (!s->rate_limit)
2680 return -ENOMEM;
2681
2682 r = system_journal_open(s);
2683 if (r < 0)
2684 return r;
2685
2686 return 0;
2687 }
2688
2689 static void server_done(Server *s) {
2690 JournalFile *f;
2691 assert(s);
2692
2693 while (s->stdout_streams)
2694 stdout_stream_free(s->stdout_streams);
2695
2696 if (s->system_journal)
2697 journal_file_close(s->system_journal);
2698
2699 if (s->runtime_journal)
2700 journal_file_close(s->runtime_journal);
2701
2702 while ((f = hashmap_steal_first(s->user_journals)))
2703 journal_file_close(f);
2704
2705 hashmap_free(s->user_journals);
2706
2707 if (s->epoll_fd >= 0)
2708 close_nointr_nofail(s->epoll_fd);
2709
2710 if (s->signal_fd >= 0)
2711 close_nointr_nofail(s->signal_fd);
2712
2713 if (s->syslog_fd >= 0)
2714 close_nointr_nofail(s->syslog_fd);
2715
2716 if (s->native_fd >= 0)
2717 close_nointr_nofail(s->native_fd);
2718
2719 if (s->stdout_fd >= 0)
2720 close_nointr_nofail(s->stdout_fd);
2721
2722 if (s->proc_kmsg_fd >= 0)
2723 close_nointr_nofail(s->proc_kmsg_fd);
2724
2725 if (s->rate_limit)
2726 journal_rate_limit_free(s->rate_limit);
2727
2728 free(s->buffer);
2729 }
2730
2731 int main(int argc, char *argv[]) {
2732 Server server;
2733 int r;
2734
2735 /* if (getppid() != 1) { */
2736 /* log_error("This program should be invoked by init only."); */
2737 /* return EXIT_FAILURE; */
2738 /* } */
2739
2740 if (argc > 1) {
2741 log_error("This program does not take arguments.");
2742 return EXIT_FAILURE;
2743 }
2744
2745 log_set_target(LOG_TARGET_CONSOLE);
2746 log_parse_environment();
2747 log_open();
2748
2749 umask(0022);
2750
2751 r = server_init(&server);
2752 if (r < 0)
2753 goto finish;
2754
2755 server_vacuum(&server);
2756 server_flush_to_var(&server);
2757 server_flush_proc_kmsg(&server);
2758
2759 log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
2760 driver_message(&server, SD_MESSAGE_JOURNAL_START, "Journal started");
2761
2762 sd_notify(false,
2763 "READY=1\n"
2764 "STATUS=Processing requests...");
2765
2766 for (;;) {
2767 struct epoll_event event;
2768
2769 r = epoll_wait(server.epoll_fd, &event, 1, -1);
2770 if (r < 0) {
2771
2772 if (errno == EINTR)
2773 continue;
2774
2775 log_error("epoll_wait() failed: %m");
2776 r = -errno;
2777 goto finish;
2778 } else if (r == 0)
2779 break;
2780
2781 r = process_event(&server, &event);
2782 if (r < 0)
2783 goto finish;
2784 else if (r == 0)
2785 break;
2786 }
2787
2788 log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
2789 driver_message(&server, SD_MESSAGE_JOURNAL_STOP, "Journal stopped");
2790
2791 finish:
2792 sd_notify(false,
2793 "STATUS=Shutting down...");
2794
2795 server_done(&server);
2796
2797 return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2798 }
Unnamed repository; edit this file 'description' to name the repository.