1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2011 Lennart Poettering
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.
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.
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/>.
22 #include <sys/epoll.h>
23 #include <sys/socket.h>
25 #include <sys/signalfd.h>
29 #include <acl/libacl.h>
31 #include <sys/ioctl.h>
32 #include <linux/sockios.h>
33 #include <sys/statvfs.h>
36 #include "journal-file.h"
37 #include "sd-daemon.h"
38 #include "socket-util.h"
40 #include "cgroup-util.h"
42 #include "journal-rate-limit.h"
44 #define USER_JOURNALS_MAX 1024
45 #define STDOUT_STREAMS_MAX 4096
47 typedef struct StdoutStream StdoutStream;
49 typedef struct Server {
56 JournalFile *runtime_journal;
57 JournalFile *system_journal;
58 Hashmap *user_journals;
65 JournalRateLimit *rate_limit;
67 JournalMetrics metrics;
71 LIST_HEAD(StdoutStream, stdout_streams);
72 unsigned n_stdout_streams;
75 typedef enum StdoutStreamState {
77 STDOUT_STREAM_PRIORITY,
78 STDOUT_STREAM_PRIORITY_PREFIX,
79 STDOUT_STREAM_TEE_CONSOLE,
85 StdoutStreamState state;
93 bool priority_prefix:1;
96 char buffer[LINE_MAX+1];
99 LIST_FIELDS(StdoutStream, stdout_stream);
102 static uint64_t available_space(Server *s) {
108 uint64_t sum = 0, avail = 0, ss_avail = 0;
112 r = sd_id128_get_machine(&machine);
116 if (s->system_journal)
117 f = "/var/log/journal/";
119 f = "/run/log/journal/";
121 p = strappend(f, sd_id128_to_string(machine, ids));
131 if (fstatvfs(dirfd(d), &ss) < 0)
136 struct dirent buf, *de;
139 k = readdir_r(d, &buf, &de);
148 if (!dirent_is_file_with_suffix(de, ".journal"))
151 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
154 sum += (uint64_t) st.st_blocks * (uint64_t) st.st_blksize;
157 avail = sum >= s->max_use ? 0 : s->max_use - sum;
159 ss_avail = ss.f_bsize * ss.f_bavail;
161 ss_avail = ss_avail < s->metrics.keep_free ? 0 : ss_avail - s->metrics.keep_free;
163 if (ss_avail < avail)
172 static void fix_perms(JournalFile *f, uid_t uid) {
175 acl_permset_t permset;
180 r = fchmod_and_fchown(f->fd, 0640, 0, 0);
182 log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
187 acl = acl_get_fd(f->fd);
189 log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
193 r = acl_find_uid(acl, uid, &entry);
196 if (acl_create_entry(&acl, &entry) < 0 ||
197 acl_set_tag_type(entry, ACL_USER) < 0 ||
198 acl_set_qualifier(entry, &uid) < 0) {
199 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
204 if (acl_get_permset(entry, &permset) < 0 ||
205 acl_add_perm(permset, ACL_READ) < 0 ||
206 acl_calc_mask(&acl) < 0) {
207 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
211 if (acl_set_fd(f->fd, acl) < 0)
212 log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
218 static JournalFile* find_journal(Server *s, uid_t uid) {
227 /* We split up user logs only on /var, not on /run */
228 if (!s->system_journal)
229 return s->runtime_journal;
232 return s->system_journal;
234 r = sd_id128_get_machine(&machine);
236 return s->system_journal;
238 f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));
242 if (asprintf(&p, "/var/log/journal/%s/user-%lu.journal", sd_id128_to_string(machine, ids), (unsigned long) uid) < 0)
243 return s->system_journal;
245 while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
246 /* Too many open? Then let's close one */
247 f = hashmap_steal_first(s->user_journals);
249 journal_file_close(f);
252 r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
256 return s->system_journal;
259 f->metrics = s->metrics;
260 f->compress = s->compress;
262 r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
264 journal_file_close(f);
265 return s->system_journal;
271 static void server_vacuum(Server *s) {
280 log_info("Rotating...");
282 if (s->runtime_journal) {
283 r = journal_file_rotate(&s->runtime_journal);
285 log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
288 if (s->system_journal) {
289 r = journal_file_rotate(&s->system_journal);
291 log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
294 HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
295 r = journal_file_rotate(&f);
297 log_error("Failed to rotate %s: %s", f->path, strerror(-r));
299 hashmap_replace(s->user_journals, k, f);
302 log_info("Vacuuming...");
304 r = sd_id128_get_machine(&machine);
306 log_error("Failed to get machine ID: %s", strerror(-r));
310 if (asprintf(&p, "/var/log/journal/%s", sd_id128_to_string(machine, ids)) < 0) {
311 log_error("Out of memory.");
315 r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
316 if (r < 0 && r != -ENOENT)
317 log_error("Failed to vacuum %s: %s", p, strerror(-r));
320 if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
321 log_error("Out of memory.");
325 r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
326 if (r < 0 && r != -ENOENT)
327 log_error("Failed to vacuum %s: %s", p, strerror(-r));
331 static char *shortened_cgroup_path(pid_t pid) {
333 char *process_path, *init_path, *path;
337 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
341 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
347 if (streq(init_path, "/"))
350 if (startswith(process_path, init_path))
351 path = process_path + strlen(init_path);
360 static void dispatch_message_real(Server *s,
361 struct iovec *iovec, unsigned n, unsigned m,
363 struct timeval *tv) {
365 char *pid = NULL, *uid = NULL, *gid = NULL,
366 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
367 *comm = NULL, *cmdline = NULL, *hostname = NULL,
368 *audit_session = NULL, *audit_loginuid = NULL,
369 *exe = NULL, *cgroup = NULL;
375 uid_t loginuid = 0, realuid = 0;
377 bool vacuumed = false;
388 realuid = ucred->uid;
390 if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
391 IOVEC_SET_STRING(iovec[n++], pid);
393 if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
394 IOVEC_SET_STRING(iovec[n++], uid);
396 if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
397 IOVEC_SET_STRING(iovec[n++], gid);
399 r = get_process_comm(ucred->pid, &t);
401 comm = strappend("_COMM=", t);
403 IOVEC_SET_STRING(iovec[n++], comm);
407 r = get_process_exe(ucred->pid, &t);
409 exe = strappend("_EXE=", t);
411 IOVEC_SET_STRING(iovec[n++], exe);
415 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
417 cmdline = strappend("_CMDLINE=", t);
419 IOVEC_SET_STRING(iovec[n++], cmdline);
423 r = audit_session_from_pid(ucred->pid, &session);
425 if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
426 IOVEC_SET_STRING(iovec[n++], audit_session);
428 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
430 if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
431 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
433 path = shortened_cgroup_path(ucred->pid);
435 cgroup = strappend("_SYSTEMD_CGROUP=", path);
437 IOVEC_SET_STRING(iovec[n++], cgroup);
444 if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
445 (unsigned long long) timeval_load(tv)) >= 0)
446 IOVEC_SET_STRING(iovec[n++], source_time);
449 /* Note that strictly speaking storing the boot id here is
450 * redundant since the entry includes this in-line
451 * anyway. However, we need this indexed, too. */
452 r = sd_id128_get_boot(&id);
454 if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
455 IOVEC_SET_STRING(iovec[n++], boot_id);
457 r = sd_id128_get_machine(&id);
459 if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
460 IOVEC_SET_STRING(iovec[n++], machine_id);
462 t = gethostname_malloc();
464 hostname = strappend("_HOSTNAME=", t);
466 IOVEC_SET_STRING(iovec[n++], hostname);
473 f = find_journal(s, realuid == 0 ? 0 : loginuid);
475 log_warning("Dropping message, as we can't find a place to store the data.");
477 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
479 if (r == -E2BIG && !vacuumed) {
480 log_info("Allocation limit reached.");
485 log_info("Retrying write.");
490 log_error("Failed to write entry, ignoring: %s", strerror(-r));
504 free(audit_loginuid);
508 static void dispatch_message(Server *s,
509 struct iovec *iovec, unsigned n, unsigned m,
517 assert(iovec || n == 0);
525 path = shortened_cgroup_path(ucred->pid);
529 /* example: /user/lennart/3/foobar
530 * /system/dbus.service/foobar
532 * So let's cut of everything past the third /, since that is
533 * wher user directories start */
535 c = strchr(path, '/');
537 c = strchr(c+1, '/');
539 c = strchr(c+1, '/');
545 rl = journal_rate_limit_test(s->rate_limit, path, priority, available_space(s));
554 char suppress_message[LINE_MAX];
555 struct iovec suppress_iovec[15];
557 /* Write a suppression message if we suppressed something */
559 snprintf(suppress_message, sizeof(suppress_message), "MESSAGE=Suppressed %u messages from %s", rl - 1, path);
560 char_array_0(suppress_message);
562 IOVEC_SET_STRING(suppress_iovec[j++], "PRIORITY=5");
563 IOVEC_SET_STRING(suppress_iovec[j++], suppress_message);
565 dispatch_message_real(s, suppress_iovec, j, ELEMENTSOF(suppress_iovec), NULL, NULL);
571 dispatch_message_real(s, iovec, n, m, ucred, tv);
574 static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
575 char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL;
576 struct iovec iovec[16];
578 int priority = LOG_USER | LOG_INFO;
583 parse_syslog_priority((char**) &buf, &priority);
584 skip_syslog_date((char**) &buf);
586 if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
587 IOVEC_SET_STRING(iovec[n++], syslog_priority);
589 if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
590 IOVEC_SET_STRING(iovec[n++], syslog_facility);
592 message = strappend("MESSAGE=", buf);
594 IOVEC_SET_STRING(iovec[n++], message);
596 dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority & LOG_PRIMASK);
599 free(syslog_facility);
600 free(syslog_priority);
603 static bool valid_user_field(const char *p, size_t l) {
606 /* We kinda enforce POSIX syntax recommendations for
607 environment variables here, but make a couple of additional
610 http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
612 /* No empty field names */
616 /* Don't allow names longer than 64 chars */
620 /* Variables starting with an underscore are protected */
624 /* Don't allow digits as first character */
625 if (p[0] >= '0' && p[0] <= '9')
628 /* Only allow A-Z0-9 and '_' */
629 for (a = p; a < p + l; a++)
630 if (!((*a >= 'A' && *a <= 'Z') ||
631 (*a >= '0' && *a <= '9') ||
638 static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) {
639 struct iovec *iovec = NULL;
640 unsigned n = 0, m = 0, j;
643 int priority = LOG_INFO;
646 assert(buffer || n == 0);
649 remaining = buffer_size;
651 while (remaining > 0) {
654 e = memchr(p, '\n', remaining);
657 /* Trailing noise, let's ignore it, and flush what we collected */
658 log_debug("Received message with trailing noise, ignoring.");
663 /* Entry separator */
664 dispatch_message(s, iovec, n, m, ucred, tv, priority);
673 if (*p == '.' || *p == '#') {
674 /* Ignore control commands for now, and
676 remaining -= (e - p) + 1;
681 /* A property follows */
687 u = MAX((n+13U) * 2U, 4U);
688 c = realloc(iovec, u * sizeof(struct iovec));
690 log_error("Out of memory");
698 q = memchr(p, '=', e - p);
700 if (valid_user_field(p, q - p)) {
701 /* If the field name starts with an
702 * underscore, skip the variable,
703 * since that indidates a trusted
705 iovec[n].iov_base = (char*) p;
706 iovec[n].iov_len = e - p;
709 /* We need to determine the priority
710 * of this entry for the rate limiting
713 memcmp(p, "PRIORITY=", 10) == 0 &&
716 priority = p[10] - '0';
719 remaining -= (e - p) + 1;
726 if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
727 log_debug("Failed to parse message, ignoring.");
731 memcpy(&l, e + 1, sizeof(uint64_t));
734 if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
735 e[1+sizeof(uint64_t)+l] != '\n') {
736 log_debug("Failed to parse message, ignoring.");
740 k = malloc((e - p) + 1 + l);
742 log_error("Out of memory");
748 memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
750 if (valid_user_field(p, e - p)) {
751 iovec[n].iov_base = k;
752 iovec[n].iov_len = (e - p) + 1 + l;
757 remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
758 p = e + 1 + sizeof(uint64_t) + l + 1;
762 dispatch_message(s, iovec, n, m, ucred, tv, priority);
764 for (j = 0; j < n; j++)
765 if (iovec[j].iov_base < buffer ||
766 (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
767 free(iovec[j].iov_base);
770 static int stdout_stream_log(StdoutStream *s, const char *p, size_t l) {
771 struct iovec iovec[15];
772 char *message = NULL, *syslog_priority = NULL;
780 priority = s->priority;
782 if (s->priority_prefix &&
785 p[1] >= '0' && p[1] <= '7' &&
788 priority = p[1] - '0';
796 if (asprintf(&syslog_priority, "PRIORITY=%i", priority) >= 0)
797 IOVEC_SET_STRING(iovec[n++], syslog_priority);
799 tag_len = s->tag ? strlen(s->tag) + 2: 0;
800 message = malloc(8 + tag_len + l);
802 memcpy(message, "MESSAGE=", 8);
805 memcpy(message+8, s->tag, tag_len-2);
806 memcpy(message+8+tag_len-2, ": ", 2);
809 memcpy(message+8+tag_len, p, l);
810 iovec[n].iov_base = message;
811 iovec[n].iov_len = 8+tag_len+l;
815 dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority);
817 if (s->tee_console) {
820 console = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
824 IOVEC_SET_STRING(iovec[n++], s->tag);
825 IOVEC_SET_STRING(iovec[n++], ": ");
828 iovec[n].iov_base = (void*) p;
829 iovec[n].iov_len = l;
832 IOVEC_SET_STRING(iovec[n++], (char*) "\n");
834 writev(console, iovec, n);
839 free(syslog_priority);
844 static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) {
848 while (l > 0 && strchr(WHITESPACE, *p)) {
853 while (l > 0 && strchr(WHITESPACE, *(p+l-1)))
858 case STDOUT_STREAM_TAG:
861 s->tag = strndup(p, l);
863 log_error("Out of memory");
868 s->state = STDOUT_STREAM_PRIORITY;
871 case STDOUT_STREAM_PRIORITY:
872 if (l != 1 || *p < '0' || *p > '7') {
873 log_warning("Failed to parse log priority line.");
877 s->priority = *p - '0';
878 s->state = STDOUT_STREAM_PRIORITY_PREFIX;
881 case STDOUT_STREAM_PRIORITY_PREFIX:
882 if (l != 1 || *p < '0' || *p > '1') {
883 log_warning("Failed to parse priority prefix line.");
887 s->priority_prefix = *p - '0';
888 s->state = STDOUT_STREAM_TEE_CONSOLE;
891 case STDOUT_STREAM_TEE_CONSOLE:
892 if (l != 1 || *p < '0' || *p > '1') {
893 log_warning("Failed to parse tee to console line.");
897 s->tee_console = *p - '0';
898 s->state = STDOUT_STREAM_RUNNING;
901 case STDOUT_STREAM_RUNNING:
902 return stdout_stream_log(s, p, l);
905 assert_not_reached("Unknown stream state");
908 static int stdout_stream_scan(StdoutStream *s, bool force_flush) {
916 remaining = s->length;
921 end = memchr(p, '\n', remaining);
923 if (remaining >= LINE_MAX) {
931 r = stdout_stream_line(s, p, end - p);
939 if (force_flush && remaining > 0) {
940 r = stdout_stream_line(s, p, remaining);
949 memmove(s->buffer, p, remaining);
950 s->length = remaining;
956 static int stdout_stream_process(StdoutStream *s) {
962 l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length);
968 log_warning("Failed to read from stream: %m");
973 r = stdout_stream_scan(s, true);
981 r = stdout_stream_scan(s, false);
989 static void stdout_stream_free(StdoutStream *s) {
993 assert(s->server->n_stdout_streams > 0);
994 s->server->n_stdout_streams --;
995 LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s);
1000 epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL);
1002 close_nointr_nofail(s->fd);
1009 static int stdout_stream_new(Server *s) {
1010 StdoutStream *stream;
1013 struct epoll_event ev;
1017 fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC);
1019 if (errno == EAGAIN)
1022 log_error("Failed to accept stdout connection: %m");
1026 if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) {
1027 log_warning("Too many stdout streams, refusing connection.");
1028 close_nointr_nofail(fd);
1032 stream = new0(StdoutStream, 1);
1034 log_error("Out of memory.");
1035 close_nointr_nofail(fd);
1041 len = sizeof(stream->ucred);
1042 if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) {
1043 log_error("Failed to determine peer credentials: %m");
1048 if (shutdown(fd, SHUT_WR) < 0) {
1049 log_error("Failed to shutdown writing side of socket: %m");
1055 ev.data.ptr = stream;
1056 ev.events = EPOLLIN;
1057 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) {
1058 log_error("Failed to add stream to event loop: %m");
1064 LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream);
1065 s->n_stdout_streams ++;
1070 stdout_stream_free(stream);
1074 static int process_event(Server *s, struct epoll_event *ev) {
1077 if (ev->data.fd == s->signal_fd) {
1078 struct signalfd_siginfo sfsi;
1081 if (ev->events != EPOLLIN) {
1082 log_info("Got invalid event from epoll.");
1086 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
1087 if (n != sizeof(sfsi)) {
1092 if (errno == EINTR || errno == EAGAIN)
1098 log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
1101 } else if (ev->data.fd == s->native_fd ||
1102 ev->data.fd == s->syslog_fd) {
1104 if (ev->events != EPOLLIN) {
1105 log_info("Got invalid event from epoll.");
1110 struct msghdr msghdr;
1112 struct ucred *ucred = NULL;
1113 struct timeval *tv = NULL;
1114 struct cmsghdr *cmsg;
1116 struct cmsghdr cmsghdr;
1117 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
1118 CMSG_SPACE(sizeof(struct timeval))];
1123 if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
1124 log_error("SIOCINQ failed: %m");
1131 if (s->buffer_size < (size_t) v) {
1135 l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
1136 b = realloc(s->buffer, l+1);
1139 log_error("Couldn't increase buffer.");
1148 iovec.iov_base = s->buffer;
1149 iovec.iov_len = s->buffer_size;
1153 msghdr.msg_iov = &iovec;
1154 msghdr.msg_iovlen = 1;
1155 msghdr.msg_control = &control;
1156 msghdr.msg_controllen = sizeof(control);
1158 n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT);
1161 if (errno == EINTR || errno == EAGAIN)
1164 log_error("recvmsg() failed: %m");
1168 for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {
1170 if (cmsg->cmsg_level == SOL_SOCKET &&
1171 cmsg->cmsg_type == SCM_CREDENTIALS &&
1172 cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))
1173 ucred = (struct ucred*) CMSG_DATA(cmsg);
1174 else if (cmsg->cmsg_level == SOL_SOCKET &&
1175 cmsg->cmsg_type == SO_TIMESTAMP &&
1176 cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))
1177 tv = (struct timeval*) CMSG_DATA(cmsg);
1180 if (ev->data.fd == s->syslog_fd) {
1183 e = memchr(s->buffer, '\n', n);
1189 process_syslog_message(s, strstrip(s->buffer), ucred, tv);
1191 process_native_message(s, s->buffer, n, ucred, tv);
1196 } else if (ev->data.fd == s->stdout_fd) {
1198 if (ev->events != EPOLLIN) {
1199 log_info("Got invalid event from epoll.");
1203 stdout_stream_new(s);
1207 StdoutStream *stream;
1209 if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
1210 log_info("Got invalid event from epoll.");
1214 /* If it is none of the well-known fds, it must be an
1215 * stdout stream fd. Note that this is a bit ugly here
1216 * (since we rely that none of the well-known fds
1217 * could be interpreted as pointer), but nonetheless
1218 * safe, since the well-known fds would never get an
1219 * fd > 4096, i.e. beyond the first memory page */
1221 stream = ev->data.ptr;
1223 if (stdout_stream_process(stream) <= 0)
1224 stdout_stream_free(stream);
1229 log_error("Unknown event.");
1233 static int system_journal_open(Server *s) {
1239 r = sd_id128_get_machine(&machine);
1243 /* First try to create the machine path, but not the prefix */
1244 fn = strappend("/var/log/journal/", sd_id128_to_string(machine, ids));
1247 (void) mkdir(fn, 0755);
1250 /* The create the system journal file */
1251 fn = join("/var/log/journal/", ids, "/system.journal", NULL);
1255 r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
1259 s->system_journal->metrics = s->metrics;
1260 s->system_journal->compress = s->compress;
1262 fix_perms(s->system_journal, 0);
1266 if (r < 0 && r != -ENOENT) {
1267 log_error("Failed to open system journal: %s", strerror(-r));
1271 /* /var didn't work, so try /run, but this time we
1272 * create the prefix too */
1273 fn = join("/run/log/journal/", ids, "/system.journal", NULL);
1277 (void) mkdir_parents(fn, 0755);
1278 r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
1282 log_error("Failed to open runtime journal: %s", strerror(-r));
1286 s->runtime_journal->metrics = s->metrics;
1287 s->runtime_journal->compress = s->compress;
1289 fix_perms(s->runtime_journal, 0);
1293 static int open_syslog_socket(Server *s) {
1294 union sockaddr_union sa;
1296 struct epoll_event ev;
1300 if (s->syslog_fd < 0) {
1302 s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1303 if (s->syslog_fd < 0) {
1304 log_error("socket() failed: %m");
1309 sa.un.sun_family = AF_UNIX;
1310 strncpy(sa.un.sun_path, "/run/systemd/syslog", sizeof(sa.un.sun_path));
1312 unlink(sa.un.sun_path);
1314 r = bind(s->syslog_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1316 log_error("bind() failed: %m");
1320 chmod(sa.un.sun_path, 0666);
1324 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1326 log_error("SO_PASSCRED failed: %m");
1331 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1333 log_error("SO_TIMESTAMP failed: %m");
1338 ev.events = EPOLLIN;
1339 ev.data.fd = s->syslog_fd;
1340 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->syslog_fd, &ev) < 0) {
1341 log_error("Failed to add syslog server fd to epoll object: %m");
1348 static int open_native_socket(Server*s) {
1349 union sockaddr_union sa;
1351 struct epoll_event ev;
1355 if (s->native_fd < 0) {
1357 s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1358 if (s->native_fd < 0) {
1359 log_error("socket() failed: %m");
1364 sa.un.sun_family = AF_UNIX;
1365 strncpy(sa.un.sun_path, "/run/systemd/journal", sizeof(sa.un.sun_path));
1367 unlink(sa.un.sun_path);
1369 r = bind(s->native_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1371 log_error("bind() failed: %m");
1375 chmod(sa.un.sun_path, 0666);
1379 r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1381 log_error("SO_PASSCRED failed: %m");
1386 r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1388 log_error("SO_TIMESTAMP failed: %m");
1393 ev.events = EPOLLIN;
1394 ev.data.fd = s->native_fd;
1395 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->native_fd, &ev) < 0) {
1396 log_error("Failed to add native server fd to epoll object: %m");
1403 static int open_stdout_socket(Server *s) {
1404 union sockaddr_union sa;
1406 struct epoll_event ev;
1410 if (s->stdout_fd < 0) {
1412 s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0);
1413 if (s->stdout_fd < 0) {
1414 log_error("socket() failed: %m");
1419 sa.un.sun_family = AF_UNIX;
1420 strncpy(sa.un.sun_path, "/run/systemd/stdout", sizeof(sa.un.sun_path));
1422 unlink(sa.un.sun_path);
1424 r = bind(s->stdout_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1426 log_error("bind() failed: %m");
1430 chmod(sa.un.sun_path, 0666);
1432 if (listen(s->stdout_fd, SOMAXCONN) < 0) {
1433 log_error("liste() failed: %m");
1439 ev.events = EPOLLIN;
1440 ev.data.fd = s->stdout_fd;
1441 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->stdout_fd, &ev) < 0) {
1442 log_error("Failed to add stdout server fd to epoll object: %m");
1449 static int open_signalfd(Server *s) {
1451 struct epoll_event ev;
1455 assert_se(sigemptyset(&mask) == 0);
1456 sigset_add_many(&mask, SIGINT, SIGTERM, -1);
1457 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
1459 s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
1460 if (s->signal_fd < 0) {
1461 log_error("signalfd(): %m");
1466 ev.events = EPOLLIN;
1467 ev.data.fd = s->signal_fd;
1469 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
1470 log_error("epoll_ctl(): %m");
1477 static int server_init(Server *s) {
1483 s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = -1;
1484 s->metrics.max_size = DEFAULT_MAX_SIZE;
1485 s->metrics.min_size = DEFAULT_MIN_SIZE;
1486 s->metrics.keep_free = DEFAULT_KEEP_FREE;
1487 s->max_use = DEFAULT_MAX_USE;
1490 s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
1491 if (!s->user_journals) {
1492 log_error("Out of memory.");
1496 s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
1497 if (s->epoll_fd < 0) {
1498 log_error("Failed to create epoll object: %m");
1502 n = sd_listen_fds(true);
1504 log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
1508 for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
1510 if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/native", 0) > 0) {
1512 if (s->native_fd >= 0) {
1513 log_error("Too many native sockets passed.");
1519 } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/stdout", 0) > 0) {
1521 if (s->stdout_fd >= 0) {
1522 log_error("Too many stdout sockets passed.");
1528 } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
1530 if (s->syslog_fd >= 0) {
1531 log_error("Too many /dev/log sockets passed.");
1538 log_error("Unknown socket passed.");
1543 r = open_syslog_socket(s);
1547 r = open_native_socket(s);
1551 r = open_stdout_socket(s);
1555 r = system_journal_open(s);
1559 r = open_signalfd(s);
1563 s->rate_limit = journal_rate_limit_new(10*USEC_PER_SEC, 2);
1570 static void server_done(Server *s) {
1574 while (s->stdout_streams)
1575 stdout_stream_free(s->stdout_streams);
1577 if (s->system_journal)
1578 journal_file_close(s->system_journal);
1580 if (s->runtime_journal)
1581 journal_file_close(s->runtime_journal);
1583 while ((f = hashmap_steal_first(s->user_journals)))
1584 journal_file_close(f);
1586 hashmap_free(s->user_journals);
1588 if (s->epoll_fd >= 0)
1589 close_nointr_nofail(s->epoll_fd);
1591 if (s->signal_fd >= 0)
1592 close_nointr_nofail(s->signal_fd);
1594 if (s->syslog_fd >= 0)
1595 close_nointr_nofail(s->syslog_fd);
1597 if (s->native_fd >= 0)
1598 close_nointr_nofail(s->native_fd);
1600 if (s->stdout_fd >= 0)
1601 close_nointr_nofail(s->stdout_fd);
1604 journal_rate_limit_free(s->rate_limit);
1607 int main(int argc, char *argv[]) {
1611 /* if (getppid() != 1) { */
1612 /* log_error("This program should be invoked by init only."); */
1613 /* return EXIT_FAILURE; */
1617 log_error("This program does not take arguments.");
1618 return EXIT_FAILURE;
1621 log_set_target(LOG_TARGET_CONSOLE);
1622 log_set_max_level(LOG_DEBUG);
1623 log_parse_environment();
1628 r = server_init(&server);
1632 log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
1636 "STATUS=Processing requests...");
1639 struct epoll_event event;
1641 r = epoll_wait(server.epoll_fd, &event, 1, -1);
1647 log_error("epoll_wait() failed: %m");
1653 r = process_event(&server, &event);
1660 log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
1664 "STATUS=Shutting down...");
1666 server_done(&server);
1668 return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;