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"
43 #include "sd-journal.h"
44 #include "journal-internal.h"
46 #define USER_JOURNALS_MAX 1024
47 #define STDOUT_STREAMS_MAX 4096
49 #define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC)
50 #define DEFAULT_RATE_LIMIT_BURST 200
52 #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
54 #define RECHECK_VAR_AVAILABLE_USEC (30*USEC_PER_SEC)
56 typedef struct StdoutStream StdoutStream;
58 typedef struct Server {
65 JournalFile *runtime_journal;
66 JournalFile *system_journal;
67 Hashmap *user_journals;
74 JournalRateLimit *rate_limit;
76 JournalMetrics metrics;
80 uint64_t cached_available_space;
81 usec_t cached_available_space_timestamp;
83 uint64_t var_available_timestamp;
85 LIST_HEAD(StdoutStream, stdout_streams);
86 unsigned n_stdout_streams;
89 typedef enum StdoutStreamState {
91 STDOUT_STREAM_PRIORITY,
92 STDOUT_STREAM_PRIORITY_PREFIX,
93 STDOUT_STREAM_TEE_CONSOLE,
99 StdoutStreamState state;
107 bool priority_prefix:1;
110 char buffer[LINE_MAX+1];
113 LIST_FIELDS(StdoutStream, stdout_stream);
116 static int server_flush_to_var(Server *s);
118 static uint64_t available_space(Server *s) {
124 uint64_t sum = 0, avail = 0, ss_avail = 0;
127 usec_t ts = now(CLOCK_MONOTONIC);
129 if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts)
130 return s->cached_available_space;
132 r = sd_id128_get_machine(&machine);
136 if (s->system_journal)
137 f = "/var/log/journal/";
139 f = "/run/log/journal/";
141 p = strappend(f, sd_id128_to_string(machine, ids));
151 if (fstatvfs(dirfd(d), &ss) < 0)
156 struct dirent buf, *de;
159 k = readdir_r(d, &buf, &de);
168 if (!dirent_is_file_with_suffix(de, ".journal"))
171 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
174 sum += (uint64_t) st.st_blocks * (uint64_t) st.st_blksize;
177 avail = sum >= s->max_use ? 0 : s->max_use - sum;
179 ss_avail = ss.f_bsize * ss.f_bavail;
181 ss_avail = ss_avail < s->metrics.keep_free ? 0 : ss_avail - s->metrics.keep_free;
183 if (ss_avail < avail)
186 s->cached_available_space = avail;
187 s->cached_available_space_timestamp = ts;
195 static void fix_perms(JournalFile *f, uid_t uid) {
198 acl_permset_t permset;
203 r = fchmod_and_fchown(f->fd, 0640, 0, 0);
205 log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
210 acl = acl_get_fd(f->fd);
212 log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
216 r = acl_find_uid(acl, uid, &entry);
219 if (acl_create_entry(&acl, &entry) < 0 ||
220 acl_set_tag_type(entry, ACL_USER) < 0 ||
221 acl_set_qualifier(entry, &uid) < 0) {
222 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
227 if (acl_get_permset(entry, &permset) < 0 ||
228 acl_add_perm(permset, ACL_READ) < 0 ||
229 acl_calc_mask(&acl) < 0) {
230 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
234 if (acl_set_fd(f->fd, acl) < 0)
235 log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
241 static JournalFile* find_journal(Server *s, uid_t uid) {
250 /* We split up user logs only on /var, not on /run. If the
251 * runtime file is open, we write to it exclusively, in order
252 * to guarantee proper order as soon as we flush /run to
253 * /var and close the runtime file. */
255 if (s->runtime_journal)
256 return s->runtime_journal;
259 return s->system_journal;
261 r = sd_id128_get_machine(&machine);
263 return s->system_journal;
265 f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));
269 if (asprintf(&p, "/var/log/journal/%s/user-%lu.journal", sd_id128_to_string(machine, ids), (unsigned long) uid) < 0)
270 return s->system_journal;
272 while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
273 /* Too many open? Then let's close one */
274 f = hashmap_steal_first(s->user_journals);
276 journal_file_close(f);
279 r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
283 return s->system_journal;
286 f->metrics = s->metrics;
287 f->compress = s->compress;
289 r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
291 journal_file_close(f);
292 return s->system_journal;
298 static void server_vacuum(Server *s) {
307 log_info("Rotating...");
309 if (s->runtime_journal) {
310 r = journal_file_rotate(&s->runtime_journal);
312 log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
315 if (s->system_journal) {
316 r = journal_file_rotate(&s->system_journal);
318 log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
321 HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
322 r = journal_file_rotate(&f);
324 log_error("Failed to rotate %s: %s", f->path, strerror(-r));
326 hashmap_replace(s->user_journals, k, f);
329 log_info("Vacuuming...");
331 r = sd_id128_get_machine(&machine);
333 log_error("Failed to get machine ID: %s", strerror(-r));
337 if (asprintf(&p, "/var/log/journal/%s", sd_id128_to_string(machine, ids)) < 0) {
338 log_error("Out of memory.");
342 r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
343 if (r < 0 && r != -ENOENT)
344 log_error("Failed to vacuum %s: %s", p, strerror(-r));
347 if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
348 log_error("Out of memory.");
352 r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
353 if (r < 0 && r != -ENOENT)
354 log_error("Failed to vacuum %s: %s", p, strerror(-r));
357 s->cached_available_space_timestamp = 0;
360 static char *shortened_cgroup_path(pid_t pid) {
362 char *process_path, *init_path, *path;
366 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
370 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
376 if (streq(init_path, "/"))
379 if (startswith(process_path, init_path))
380 path = process_path + strlen(init_path);
389 static void dispatch_message_real(Server *s,
390 struct iovec *iovec, unsigned n, unsigned m,
392 struct timeval *tv) {
394 char *pid = NULL, *uid = NULL, *gid = NULL,
395 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
396 *comm = NULL, *cmdline = NULL, *hostname = NULL,
397 *audit_session = NULL, *audit_loginuid = NULL,
398 *exe = NULL, *cgroup = NULL;
404 uid_t loginuid = 0, realuid = 0;
406 bool vacuumed = false;
417 realuid = ucred->uid;
419 if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
420 IOVEC_SET_STRING(iovec[n++], pid);
422 if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
423 IOVEC_SET_STRING(iovec[n++], uid);
425 if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
426 IOVEC_SET_STRING(iovec[n++], gid);
428 r = get_process_comm(ucred->pid, &t);
430 comm = strappend("_COMM=", t);
432 IOVEC_SET_STRING(iovec[n++], comm);
436 r = get_process_exe(ucred->pid, &t);
438 exe = strappend("_EXE=", t);
440 IOVEC_SET_STRING(iovec[n++], exe);
444 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
446 cmdline = strappend("_CMDLINE=", t);
448 IOVEC_SET_STRING(iovec[n++], cmdline);
452 r = audit_session_from_pid(ucred->pid, &session);
454 if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
455 IOVEC_SET_STRING(iovec[n++], audit_session);
457 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
459 if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
460 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
462 path = shortened_cgroup_path(ucred->pid);
464 cgroup = strappend("_SYSTEMD_CGROUP=", path);
466 IOVEC_SET_STRING(iovec[n++], cgroup);
473 if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
474 (unsigned long long) timeval_load(tv)) >= 0)
475 IOVEC_SET_STRING(iovec[n++], source_time);
478 /* Note that strictly speaking storing the boot id here is
479 * redundant since the entry includes this in-line
480 * anyway. However, we need this indexed, too. */
481 r = sd_id128_get_boot(&id);
483 if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
484 IOVEC_SET_STRING(iovec[n++], boot_id);
486 r = sd_id128_get_machine(&id);
488 if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
489 IOVEC_SET_STRING(iovec[n++], machine_id);
491 t = gethostname_malloc();
493 hostname = strappend("_HOSTNAME=", t);
495 IOVEC_SET_STRING(iovec[n++], hostname);
501 server_flush_to_var(s);
504 f = find_journal(s, realuid == 0 ? 0 : loginuid);
506 log_warning("Dropping message, as we can't find a place to store the data.");
508 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
510 if (r == -E2BIG && !vacuumed) {
511 log_info("Allocation limit reached.");
516 log_info("Retrying write.");
521 log_error("Failed to write entry, ignoring: %s", strerror(-r));
535 free(audit_loginuid);
539 static void dispatch_message(Server *s,
540 struct iovec *iovec, unsigned n, unsigned m,
548 assert(iovec || n == 0);
556 path = shortened_cgroup_path(ucred->pid);
560 /* example: /user/lennart/3/foobar
561 * /system/dbus.service/foobar
563 * So let's cut of everything past the third /, since that is
564 * wher user directories start */
566 c = strchr(path, '/');
568 c = strchr(c+1, '/');
570 c = strchr(c+1, '/');
576 rl = journal_rate_limit_test(s->rate_limit, path, priority, available_space(s));
585 char suppress_message[LINE_MAX];
586 struct iovec suppress_iovec[15];
588 /* Write a suppression message if we suppressed something */
590 snprintf(suppress_message, sizeof(suppress_message), "MESSAGE=Suppressed %u messages from %s", rl - 1, path);
591 char_array_0(suppress_message);
593 IOVEC_SET_STRING(suppress_iovec[j++], "PRIORITY=5");
594 IOVEC_SET_STRING(suppress_iovec[j++], suppress_message);
596 dispatch_message_real(s, suppress_iovec, j, ELEMENTSOF(suppress_iovec), NULL, NULL);
602 dispatch_message_real(s, iovec, n, m, ucred, tv);
605 static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
606 char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL;
607 struct iovec iovec[16];
609 int priority = LOG_USER | LOG_INFO;
614 parse_syslog_priority((char**) &buf, &priority);
615 skip_syslog_date((char**) &buf);
617 if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
618 IOVEC_SET_STRING(iovec[n++], syslog_priority);
620 if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
621 IOVEC_SET_STRING(iovec[n++], syslog_facility);
623 message = strappend("MESSAGE=", buf);
625 IOVEC_SET_STRING(iovec[n++], message);
627 dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority & LOG_PRIMASK);
630 free(syslog_facility);
631 free(syslog_priority);
634 static bool valid_user_field(const char *p, size_t l) {
637 /* We kinda enforce POSIX syntax recommendations for
638 environment variables here, but make a couple of additional
641 http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
643 /* No empty field names */
647 /* Don't allow names longer than 64 chars */
651 /* Variables starting with an underscore are protected */
655 /* Don't allow digits as first character */
656 if (p[0] >= '0' && p[0] <= '9')
659 /* Only allow A-Z0-9 and '_' */
660 for (a = p; a < p + l; a++)
661 if (!((*a >= 'A' && *a <= 'Z') ||
662 (*a >= '0' && *a <= '9') ||
669 static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) {
670 struct iovec *iovec = NULL;
671 unsigned n = 0, m = 0, j;
674 int priority = LOG_INFO;
677 assert(buffer || n == 0);
680 remaining = buffer_size;
682 while (remaining > 0) {
685 e = memchr(p, '\n', remaining);
688 /* Trailing noise, let's ignore it, and flush what we collected */
689 log_debug("Received message with trailing noise, ignoring.");
694 /* Entry separator */
695 dispatch_message(s, iovec, n, m, ucred, tv, priority);
704 if (*p == '.' || *p == '#') {
705 /* Ignore control commands for now, and
707 remaining -= (e - p) + 1;
712 /* A property follows */
718 u = MAX((n+13U) * 2U, 4U);
719 c = realloc(iovec, u * sizeof(struct iovec));
721 log_error("Out of memory");
729 q = memchr(p, '=', e - p);
731 if (valid_user_field(p, q - p)) {
732 /* If the field name starts with an
733 * underscore, skip the variable,
734 * since that indidates a trusted
736 iovec[n].iov_base = (char*) p;
737 iovec[n].iov_len = e - p;
740 /* We need to determine the priority
741 * of this entry for the rate limiting
744 memcmp(p, "PRIORITY=", 10) == 0 &&
747 priority = p[10] - '0';
750 remaining -= (e - p) + 1;
757 if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
758 log_debug("Failed to parse message, ignoring.");
762 memcpy(&l, e + 1, sizeof(uint64_t));
765 if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
766 e[1+sizeof(uint64_t)+l] != '\n') {
767 log_debug("Failed to parse message, ignoring.");
771 k = malloc((e - p) + 1 + l);
773 log_error("Out of memory");
779 memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
781 if (valid_user_field(p, e - p)) {
782 iovec[n].iov_base = k;
783 iovec[n].iov_len = (e - p) + 1 + l;
788 remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
789 p = e + 1 + sizeof(uint64_t) + l + 1;
793 dispatch_message(s, iovec, n, m, ucred, tv, priority);
795 for (j = 0; j < n; j++)
796 if (iovec[j].iov_base < buffer ||
797 (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
798 free(iovec[j].iov_base);
801 static int stdout_stream_log(StdoutStream *s, const char *p, size_t l) {
802 struct iovec iovec[15];
803 char *message = NULL, *syslog_priority = NULL;
811 priority = s->priority;
813 if (s->priority_prefix &&
816 p[1] >= '0' && p[1] <= '7' &&
819 priority = p[1] - '0';
827 if (asprintf(&syslog_priority, "PRIORITY=%i", priority) >= 0)
828 IOVEC_SET_STRING(iovec[n++], syslog_priority);
830 tag_len = s->tag ? strlen(s->tag) + 2: 0;
831 message = malloc(8 + tag_len + l);
833 memcpy(message, "MESSAGE=", 8);
836 memcpy(message+8, s->tag, tag_len-2);
837 memcpy(message+8+tag_len-2, ": ", 2);
840 memcpy(message+8+tag_len, p, l);
841 iovec[n].iov_base = message;
842 iovec[n].iov_len = 8+tag_len+l;
846 dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority);
848 if (s->tee_console) {
851 console = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
855 IOVEC_SET_STRING(iovec[n++], s->tag);
856 IOVEC_SET_STRING(iovec[n++], ": ");
859 iovec[n].iov_base = (void*) p;
860 iovec[n].iov_len = l;
863 IOVEC_SET_STRING(iovec[n++], (char*) "\n");
865 writev(console, iovec, n);
870 free(syslog_priority);
875 static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) {
879 while (l > 0 && strchr(WHITESPACE, *p)) {
884 while (l > 0 && strchr(WHITESPACE, *(p+l-1)))
889 case STDOUT_STREAM_TAG:
892 s->tag = strndup(p, l);
894 log_error("Out of memory");
899 s->state = STDOUT_STREAM_PRIORITY;
902 case STDOUT_STREAM_PRIORITY:
903 if (l != 1 || *p < '0' || *p > '7') {
904 log_warning("Failed to parse log priority line.");
908 s->priority = *p - '0';
909 s->state = STDOUT_STREAM_PRIORITY_PREFIX;
912 case STDOUT_STREAM_PRIORITY_PREFIX:
913 if (l != 1 || *p < '0' || *p > '1') {
914 log_warning("Failed to parse priority prefix line.");
918 s->priority_prefix = *p - '0';
919 s->state = STDOUT_STREAM_TEE_CONSOLE;
922 case STDOUT_STREAM_TEE_CONSOLE:
923 if (l != 1 || *p < '0' || *p > '1') {
924 log_warning("Failed to parse tee to console line.");
928 s->tee_console = *p - '0';
929 s->state = STDOUT_STREAM_RUNNING;
932 case STDOUT_STREAM_RUNNING:
933 return stdout_stream_log(s, p, l);
936 assert_not_reached("Unknown stream state");
939 static int stdout_stream_scan(StdoutStream *s, bool force_flush) {
947 remaining = s->length;
952 end = memchr(p, '\n', remaining);
954 if (remaining >= LINE_MAX) {
962 r = stdout_stream_line(s, p, end - p);
970 if (force_flush && remaining > 0) {
971 r = stdout_stream_line(s, p, remaining);
980 memmove(s->buffer, p, remaining);
981 s->length = remaining;
987 static int stdout_stream_process(StdoutStream *s) {
993 l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length);
999 log_warning("Failed to read from stream: %m");
1004 r = stdout_stream_scan(s, true);
1012 r = stdout_stream_scan(s, false);
1020 static void stdout_stream_free(StdoutStream *s) {
1024 assert(s->server->n_stdout_streams > 0);
1025 s->server->n_stdout_streams --;
1026 LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s);
1031 epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL);
1033 close_nointr_nofail(s->fd);
1040 static int stdout_stream_new(Server *s) {
1041 StdoutStream *stream;
1044 struct epoll_event ev;
1048 fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC);
1050 if (errno == EAGAIN)
1053 log_error("Failed to accept stdout connection: %m");
1057 if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) {
1058 log_warning("Too many stdout streams, refusing connection.");
1059 close_nointr_nofail(fd);
1063 stream = new0(StdoutStream, 1);
1065 log_error("Out of memory.");
1066 close_nointr_nofail(fd);
1072 len = sizeof(stream->ucred);
1073 if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) {
1074 log_error("Failed to determine peer credentials: %m");
1079 if (shutdown(fd, SHUT_WR) < 0) {
1080 log_error("Failed to shutdown writing side of socket: %m");
1086 ev.data.ptr = stream;
1087 ev.events = EPOLLIN;
1088 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) {
1089 log_error("Failed to add stream to event loop: %m");
1095 LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream);
1096 s->n_stdout_streams ++;
1101 stdout_stream_free(stream);
1105 static int system_journal_open(Server *s) {
1111 r = sd_id128_get_machine(&machine);
1115 sd_id128_to_string(machine, ids);
1117 if (!s->system_journal) {
1119 /* First try to create the machine path, but not the prefix */
1120 fn = strappend("/var/log/journal/", ids);
1123 (void) mkdir(fn, 0755);
1126 /* The create the system journal file */
1127 fn = join("/var/log/journal/", ids, "/system.journal", NULL);
1131 r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
1135 s->system_journal->metrics = s->metrics;
1136 s->system_journal->compress = s->compress;
1138 fix_perms(s->system_journal, 0);
1144 log_error("Failed to open system journal: %s", strerror(-r));
1150 if (!s->runtime_journal) {
1152 fn = join("/run/log/journal/", ids, "/system.journal", NULL);
1156 if (s->system_journal) {
1158 /* Try to open the runtime journal, but only
1159 * if it already exists, so that we can flush
1160 * it into the system journal */
1162 r = journal_file_open(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
1170 log_error("Failed to open runtime journal: %s", strerror(-r));
1177 /* OK, we really need the runtime journal, so create
1178 * it if necessary. */
1180 (void) mkdir_parents(fn, 0755);
1181 r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
1185 log_error("Failed to open runtime journal: %s", strerror(-r));
1190 if (s->runtime_journal) {
1191 s->runtime_journal->metrics = s->metrics;
1192 s->runtime_journal->compress = s->compress;
1194 fix_perms(s->runtime_journal, 0);
1201 static int server_flush_to_var(Server *s) {
1202 char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
1211 if (!s->runtime_journal)
1214 ts = now(CLOCK_MONOTONIC);
1215 if (s->var_available_timestamp + RECHECK_VAR_AVAILABLE_USEC > ts)
1218 s->var_available_timestamp = ts;
1220 system_journal_open(s);
1222 if (!s->system_journal)
1225 r = sd_id128_get_machine(&machine);
1227 log_error("Failed to get machine id: %s", strerror(-r));
1231 r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);
1233 log_error("Failed to read runtime journal: %s", strerror(-r));
1237 SD_JOURNAL_FOREACH(j) {
1240 f = j->current_file;
1241 assert(f && f->current_offset > 0);
1243 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1245 log_error("Can't read entry: %s", strerror(-r));
1249 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
1251 log_info("Allocation limit reached.");
1253 journal_file_post_change(s->system_journal);
1256 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
1260 log_error("Can't write entry: %s", strerror(-r));
1266 journal_file_post_change(s->system_journal);
1268 journal_file_close(s->runtime_journal);
1269 s->runtime_journal = NULL;
1272 sd_id128_to_string(machine, path + 17);
1273 rm_rf(path, false, true, false);
1279 static int process_event(Server *s, struct epoll_event *ev) {
1282 if (ev->data.fd == s->signal_fd) {
1283 struct signalfd_siginfo sfsi;
1286 if (ev->events != EPOLLIN) {
1287 log_info("Got invalid event from epoll.");
1291 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
1292 if (n != sizeof(sfsi)) {
1297 if (errno == EINTR || errno == EAGAIN)
1303 if (sfsi.ssi_signo == SIGUSR1) {
1304 server_flush_to_var(s);
1308 log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
1311 } else if (ev->data.fd == s->native_fd ||
1312 ev->data.fd == s->syslog_fd) {
1314 if (ev->events != EPOLLIN) {
1315 log_info("Got invalid event from epoll.");
1320 struct msghdr msghdr;
1322 struct ucred *ucred = NULL;
1323 struct timeval *tv = NULL;
1324 struct cmsghdr *cmsg;
1326 struct cmsghdr cmsghdr;
1327 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
1328 CMSG_SPACE(sizeof(struct timeval))];
1333 if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
1334 log_error("SIOCINQ failed: %m");
1341 if (s->buffer_size < (size_t) v) {
1345 l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
1346 b = realloc(s->buffer, l+1);
1349 log_error("Couldn't increase buffer.");
1358 iovec.iov_base = s->buffer;
1359 iovec.iov_len = s->buffer_size;
1363 msghdr.msg_iov = &iovec;
1364 msghdr.msg_iovlen = 1;
1365 msghdr.msg_control = &control;
1366 msghdr.msg_controllen = sizeof(control);
1368 n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT);
1371 if (errno == EINTR || errno == EAGAIN)
1374 log_error("recvmsg() failed: %m");
1378 for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {
1380 if (cmsg->cmsg_level == SOL_SOCKET &&
1381 cmsg->cmsg_type == SCM_CREDENTIALS &&
1382 cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))
1383 ucred = (struct ucred*) CMSG_DATA(cmsg);
1384 else if (cmsg->cmsg_level == SOL_SOCKET &&
1385 cmsg->cmsg_type == SO_TIMESTAMP &&
1386 cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))
1387 tv = (struct timeval*) CMSG_DATA(cmsg);
1390 if (ev->data.fd == s->syslog_fd) {
1393 e = memchr(s->buffer, '\n', n);
1399 process_syslog_message(s, strstrip(s->buffer), ucred, tv);
1401 process_native_message(s, s->buffer, n, ucred, tv);
1406 } else if (ev->data.fd == s->stdout_fd) {
1408 if (ev->events != EPOLLIN) {
1409 log_info("Got invalid event from epoll.");
1413 stdout_stream_new(s);
1417 StdoutStream *stream;
1419 if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
1420 log_info("Got invalid event from epoll.");
1424 /* If it is none of the well-known fds, it must be an
1425 * stdout stream fd. Note that this is a bit ugly here
1426 * (since we rely that none of the well-known fds
1427 * could be interpreted as pointer), but nonetheless
1428 * safe, since the well-known fds would never get an
1429 * fd > 4096, i.e. beyond the first memory page */
1431 stream = ev->data.ptr;
1433 if (stdout_stream_process(stream) <= 0)
1434 stdout_stream_free(stream);
1439 log_error("Unknown event.");
1443 static int open_syslog_socket(Server *s) {
1444 union sockaddr_union sa;
1446 struct epoll_event ev;
1450 if (s->syslog_fd < 0) {
1452 s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1453 if (s->syslog_fd < 0) {
1454 log_error("socket() failed: %m");
1459 sa.un.sun_family = AF_UNIX;
1460 strncpy(sa.un.sun_path, "/run/systemd/syslog", sizeof(sa.un.sun_path));
1462 unlink(sa.un.sun_path);
1464 r = bind(s->syslog_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1466 log_error("bind() failed: %m");
1470 chmod(sa.un.sun_path, 0666);
1474 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1476 log_error("SO_PASSCRED failed: %m");
1481 r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1483 log_error("SO_TIMESTAMP failed: %m");
1488 ev.events = EPOLLIN;
1489 ev.data.fd = s->syslog_fd;
1490 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->syslog_fd, &ev) < 0) {
1491 log_error("Failed to add syslog server fd to epoll object: %m");
1498 static int open_native_socket(Server*s) {
1499 union sockaddr_union sa;
1501 struct epoll_event ev;
1505 if (s->native_fd < 0) {
1507 s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1508 if (s->native_fd < 0) {
1509 log_error("socket() failed: %m");
1514 sa.un.sun_family = AF_UNIX;
1515 strncpy(sa.un.sun_path, "/run/systemd/journal", sizeof(sa.un.sun_path));
1517 unlink(sa.un.sun_path);
1519 r = bind(s->native_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1521 log_error("bind() failed: %m");
1525 chmod(sa.un.sun_path, 0666);
1529 r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1531 log_error("SO_PASSCRED failed: %m");
1536 r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1538 log_error("SO_TIMESTAMP failed: %m");
1543 ev.events = EPOLLIN;
1544 ev.data.fd = s->native_fd;
1545 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->native_fd, &ev) < 0) {
1546 log_error("Failed to add native server fd to epoll object: %m");
1553 static int open_stdout_socket(Server *s) {
1554 union sockaddr_union sa;
1556 struct epoll_event ev;
1560 if (s->stdout_fd < 0) {
1562 s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0);
1563 if (s->stdout_fd < 0) {
1564 log_error("socket() failed: %m");
1569 sa.un.sun_family = AF_UNIX;
1570 strncpy(sa.un.sun_path, "/run/systemd/stdout", sizeof(sa.un.sun_path));
1572 unlink(sa.un.sun_path);
1574 r = bind(s->stdout_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1576 log_error("bind() failed: %m");
1580 chmod(sa.un.sun_path, 0666);
1582 if (listen(s->stdout_fd, SOMAXCONN) < 0) {
1583 log_error("liste() failed: %m");
1589 ev.events = EPOLLIN;
1590 ev.data.fd = s->stdout_fd;
1591 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->stdout_fd, &ev) < 0) {
1592 log_error("Failed to add stdout server fd to epoll object: %m");
1599 static int open_signalfd(Server *s) {
1601 struct epoll_event ev;
1605 assert_se(sigemptyset(&mask) == 0);
1606 sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, -1);
1607 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
1609 s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
1610 if (s->signal_fd < 0) {
1611 log_error("signalfd(): %m");
1616 ev.events = EPOLLIN;
1617 ev.data.fd = s->signal_fd;
1619 if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
1620 log_error("epoll_ctl(): %m");
1627 static int server_init(Server *s) {
1633 s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = -1;
1634 s->metrics.max_size = DEFAULT_MAX_SIZE;
1635 s->metrics.min_size = DEFAULT_MIN_SIZE;
1636 s->metrics.keep_free = DEFAULT_KEEP_FREE;
1637 s->max_use = DEFAULT_MAX_USE;
1640 s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
1641 if (!s->user_journals) {
1642 log_error("Out of memory.");
1646 s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
1647 if (s->epoll_fd < 0) {
1648 log_error("Failed to create epoll object: %m");
1652 n = sd_listen_fds(true);
1654 log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
1658 for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
1660 if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/native", 0) > 0) {
1662 if (s->native_fd >= 0) {
1663 log_error("Too many native sockets passed.");
1669 } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/stdout", 0) > 0) {
1671 if (s->stdout_fd >= 0) {
1672 log_error("Too many stdout sockets passed.");
1678 } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
1680 if (s->syslog_fd >= 0) {
1681 log_error("Too many /dev/log sockets passed.");
1688 log_error("Unknown socket passed.");
1693 r = open_syslog_socket(s);
1697 r = open_native_socket(s);
1701 r = open_stdout_socket(s);
1705 r = system_journal_open(s);
1709 r = open_signalfd(s);
1713 s->rate_limit = journal_rate_limit_new(DEFAULT_RATE_LIMIT_INTERVAL, DEFAULT_RATE_LIMIT_BURST);
1720 static void server_done(Server *s) {
1724 while (s->stdout_streams)
1725 stdout_stream_free(s->stdout_streams);
1727 if (s->system_journal)
1728 journal_file_close(s->system_journal);
1730 if (s->runtime_journal)
1731 journal_file_close(s->runtime_journal);
1733 while ((f = hashmap_steal_first(s->user_journals)))
1734 journal_file_close(f);
1736 hashmap_free(s->user_journals);
1738 if (s->epoll_fd >= 0)
1739 close_nointr_nofail(s->epoll_fd);
1741 if (s->signal_fd >= 0)
1742 close_nointr_nofail(s->signal_fd);
1744 if (s->syslog_fd >= 0)
1745 close_nointr_nofail(s->syslog_fd);
1747 if (s->native_fd >= 0)
1748 close_nointr_nofail(s->native_fd);
1750 if (s->stdout_fd >= 0)
1751 close_nointr_nofail(s->stdout_fd);
1754 journal_rate_limit_free(s->rate_limit);
1757 int main(int argc, char *argv[]) {
1761 /* if (getppid() != 1) { */
1762 /* log_error("This program should be invoked by init only."); */
1763 /* return EXIT_FAILURE; */
1767 log_error("This program does not take arguments.");
1768 return EXIT_FAILURE;
1771 log_set_target(LOG_TARGET_CONSOLE);
1772 log_set_max_level(LOG_DEBUG);
1773 log_parse_environment();
1778 r = server_init(&server);
1782 log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
1786 "STATUS=Processing requests...");
1788 server_vacuum(&server);
1789 server_flush_to_var(&server);
1792 struct epoll_event event;
1794 r = epoll_wait(server.epoll_fd, &event, 1, -1);
1800 log_error("epoll_wait() failed: %m");
1806 r = process_event(&server, &event);
1813 log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
1817 "STATUS=Shutting down...");
1819 server_done(&server);
1821 return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;