chiark / gitweb /
journald: splitt of syslog protocol support into its own file
[elogind.git] / src / journal / journald.c
index e66bb07f84156c9b572a635cc962150e381e5c6f..4b0ff53125a0148bedbc5b79b2e50a3cd0f89413 100644 (file)
@@ -29,7 +29,9 @@
 #include <sys/ioctl.h>
 #include <linux/sockios.h>
 #include <sys/statvfs.h>
+#include <sys/mman.h>
 
+#include <libudev.h>
 #include <systemd/sd-journal.h>
 #include <systemd/sd-messages.h>
 #include <systemd/sd-daemon.h>
 #include "socket-util.h"
 #include "cgroup-util.h"
 #include "list.h"
+#include "virt.h"
+#include "missing.h"
+#include "conf-parser.h"
 #include "journal-rate-limit.h"
 #include "journal-internal.h"
-#include "conf-parser.h"
+#include "journal-vacuum.h"
+#include "journal-authenticate.h"
 #include "journald.h"
-#include "virt.h"
-#include "missing.h"
+#include "journald-kmsg.h"
+#include "journald-syslog.h"
 
 #ifdef HAVE_ACL
 #include <sys/acl.h>
@@ -69,8 +75,6 @@
 
 #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
 
-#define N_IOVEC_META_FIELDS 17
-
 #define ENTRY_SIZE_MAX (1024*1024*32)
 
 typedef enum StdoutStreamState {
@@ -279,7 +283,6 @@ static JournalFile* find_journal(Server *s, uid_t uid) {
         char *p;
         int r;
         JournalFile *f;
-        char ids[33];
         sd_id128_t machine;
 
         assert(s);
@@ -303,7 +306,8 @@ static JournalFile* find_journal(Server *s, uid_t uid) {
         if (f)
                 return f;
 
-        if (asprintf(&p, "/var/log/journal/%s/user-%lu.journal", sd_id128_to_string(machine, ids), (unsigned long) uid) < 0)
+        if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal",
+                     SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0)
                 return s->system_journal;
 
         while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
@@ -313,7 +317,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) {
                 journal_file_close(f);
         }
 
-        r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
+        r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, s->system_journal, &f);
         free(p);
 
         if (r < 0)
@@ -339,7 +343,7 @@ static void server_rotate(Server *s) {
         log_info("Rotating...");
 
         if (s->runtime_journal) {
-                r = journal_file_rotate(&s->runtime_journal);
+                r = journal_file_rotate(&s->runtime_journal, s->compress, false);
                 if (r < 0)
                         if (s->runtime_journal)
                                 log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
@@ -350,7 +354,7 @@ static void server_rotate(Server *s) {
         }
 
         if (s->system_journal) {
-                r = journal_file_rotate(&s->system_journal);
+                r = journal_file_rotate(&s->system_journal, s->compress, s->seal);
                 if (r < 0)
                         if (s->system_journal)
                                 log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
@@ -362,7 +366,7 @@ static void server_rotate(Server *s) {
         }
 
         HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
-                r = journal_file_rotate(&f);
+                r = journal_file_rotate(&f, s->compress, s->seal);
                 if (r < 0)
                         if (f->path)
                                 log_error("Failed to rotate %s: %s", f->path, strerror(-r));
@@ -393,7 +397,7 @@ static void server_vacuum(Server *s) {
 
         if (s->system_journal) {
                 if (asprintf(&p, "/var/log/journal/%s", ids) < 0) {
-                        log_error("Out of memory.");
+                        log_oom();
                         return;
                 }
 
@@ -405,7 +409,7 @@ static void server_vacuum(Server *s) {
 
         if (s->runtime_journal) {
                 if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
-                        log_error("Out of memory.");
+                        log_oom();
                         return;
                 }
 
@@ -478,6 +482,10 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
                 server_rotate(s);
                 server_vacuum(s);
                 vacuumed = true;
+
+                f = find_journal(s, uid);
+                if (!f)
+                        return;
         }
 
         for (;;) {
@@ -493,7 +501,9 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
                      r != -EBADMSG && /* corrupted */
                      r != -ENODATA && /* truncated */
                      r != -EHOSTDOWN && /* other machine */
-                     r != -EPROTONOSUPPORT /* unsupported feature */)) {
+                     r != -EPROTONOSUPPORT && /* unsupported feature */
+                     r != -EBUSY && /* unclean shutdown */
+                     r != -ESHUTDOWN /* already archived */)) {
                         log_error("Failed to write entry, ignoring: %s", strerror(-r));
                         return;
                 }
@@ -502,6 +512,8 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
                         log_info("Allocation limit reached, rotating.");
                 else if (r == -EHOSTDOWN)
                         log_info("Journal file from other machine, rotating.");
+                else if (r == -EBUSY)
+                        log_info("Unlcean shutdown, rotating.");
                 else
                         log_warning("Journal file corrupted, rotating.");
 
@@ -509,6 +521,10 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
                 server_vacuum(s);
                 vacuumed = true;
 
+                f = find_journal(s, uid);
+                if (!f)
+                        return;
+
                 log_info("Retrying write.");
         }
 }
@@ -698,7 +714,7 @@ static void dispatch_message_real(
         free(selinux_context);
 }
 
-static void driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {
+void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {
         char mid[11 + 32 + 1];
         char buffer[16 + LINE_MAX + 1];
         struct iovec iovec[N_IOVEC_META_FIELDS + 4];
@@ -709,7 +725,7 @@ static void driver_message(Server *s, sd_id128_t message_id, const char *format,
         assert(s);
         assert(format);
 
-        IOVEC_SET_STRING(iovec[n++], "PRIORITY=5");
+        IOVEC_SET_STRING(iovec[n++], "PRIORITY=6");
         IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver");
 
         memcpy(buffer, "MESSAGE=", 8);
@@ -731,13 +747,15 @@ static void driver_message(Server *s, sd_id128_t message_id, const char *format,
         dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL);
 }
 
-static void dispatch_message(Server *s,
-                             struct iovec *iovec, unsigned n, unsigned m,
-                             struct ucred *ucred,
-                             struct timeval *tv,
-                             const char *label, size_t label_len,
-                             const char *unit_id,
-                             int priority) {
+void server_dispatch_message(
+                Server *s,
+                struct iovec *iovec, unsigned n, unsigned m,
+                struct ucred *ucred,
+                struct timeval *tv,
+                const char *label, size_t label_len,
+                const char *unit_id,
+                int priority) {
+
         int rl;
         char *path = NULL, *c;
 
@@ -782,7 +800,7 @@ static void dispatch_message(Server *s,
 
         /* Write a suppression message if we suppressed something */
         if (rl > 1)
-                driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path);
+                server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path);
 
         free(path);
 
@@ -790,216 +808,7 @@ finish:
         dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id);
 }
 
-static void forward_syslog_iovec(Server *s, const struct iovec *iovec, unsigned n_iovec, struct ucred *ucred, struct timeval *tv) {
-        struct msghdr msghdr;
-        struct cmsghdr *cmsg;
-        union {
-                struct cmsghdr cmsghdr;
-                uint8_t buf[CMSG_SPACE(sizeof(struct ucred))];
-        } control;
-        union sockaddr_union sa;
-
-        assert(s);
-        assert(iovec);
-        assert(n_iovec > 0);
-
-        zero(msghdr);
-        msghdr.msg_iov = (struct iovec*) iovec;
-        msghdr.msg_iovlen = n_iovec;
-
-        zero(sa);
-        sa.un.sun_family = AF_UNIX;
-        strncpy(sa.un.sun_path, "/run/systemd/journal/syslog", sizeof(sa.un.sun_path));
-        msghdr.msg_name = &sa;
-        msghdr.msg_namelen = offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path);
-
-        if (ucred) {
-                zero(control);
-                msghdr.msg_control = &control;
-                msghdr.msg_controllen = sizeof(control);
-
-                cmsg = CMSG_FIRSTHDR(&msghdr);
-                cmsg->cmsg_level = SOL_SOCKET;
-                cmsg->cmsg_type = SCM_CREDENTIALS;
-                cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
-                memcpy(CMSG_DATA(cmsg), ucred, sizeof(struct ucred));
-                msghdr.msg_controllen = cmsg->cmsg_len;
-        }
-
-        /* Forward the syslog message we received via /dev/log to
-         * /run/systemd/syslog. Unfortunately we currently can't set
-         * the SO_TIMESTAMP auxiliary data, and hence we don't. */
-
-        if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
-                return;
-
-        /* The socket is full? I guess the syslog implementation is
-         * too slow, and we shouldn't wait for that... */
-        if (errno == EAGAIN)
-                return;
-
-        if (ucred && errno == ESRCH) {
-                struct ucred u;
-
-                /* Hmm, presumably the sender process vanished
-                 * by now, so let's fix it as good as we
-                 * can, and retry */
-
-                u = *ucred;
-                u.pid = getpid();
-                memcpy(CMSG_DATA(cmsg), &u, sizeof(struct ucred));
-
-                if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
-                        return;
-
-                if (errno == EAGAIN)
-                        return;
-        }
-
-        log_debug("Failed to forward syslog message: %m");
-}
-
-static void forward_syslog_raw(Server *s, int priority, const char *buffer, struct ucred *ucred, struct timeval *tv) {
-        struct iovec iovec;
-
-        assert(s);
-        assert(buffer);
-
-        if (LOG_PRI(priority) > s->max_level_syslog)
-                return;
-
-        IOVEC_SET_STRING(iovec, buffer);
-        forward_syslog_iovec(s, &iovec, 1, ucred, tv);
-}
-
-static void forward_syslog(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred, struct timeval *tv) {
-        struct iovec iovec[5];
-        char header_priority[6], header_time[64], header_pid[16];
-        int n = 0;
-        time_t t;
-        struct tm *tm;
-        char *ident_buf = NULL;
-
-        assert(s);
-        assert(priority >= 0);
-        assert(priority <= 999);
-        assert(message);
-
-        if (LOG_PRI(priority) > s->max_level_syslog)
-                return;
-
-        /* First: priority field */
-        snprintf(header_priority, sizeof(header_priority), "<%i>", priority);
-        char_array_0(header_priority);
-        IOVEC_SET_STRING(iovec[n++], header_priority);
-
-        /* Second: timestamp */
-        t = tv ? tv->tv_sec : ((time_t) (now(CLOCK_REALTIME) / USEC_PER_SEC));
-        tm = localtime(&t);
-        if (!tm)
-                return;
-        if (strftime(header_time, sizeof(header_time), "%h %e %T ", tm) <= 0)
-                return;
-        IOVEC_SET_STRING(iovec[n++], header_time);
-
-        /* Third: identifier and PID */
-        if (ucred) {
-                if (!identifier) {
-                        get_process_comm(ucred->pid, &ident_buf);
-                        identifier = ident_buf;
-                }
-
-                snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid);
-                char_array_0(header_pid);
-
-                if (identifier)
-                        IOVEC_SET_STRING(iovec[n++], identifier);
-
-                IOVEC_SET_STRING(iovec[n++], header_pid);
-        } else if (identifier) {
-                IOVEC_SET_STRING(iovec[n++], identifier);
-                IOVEC_SET_STRING(iovec[n++], ": ");
-        }
-
-        /* Fourth: message */
-        IOVEC_SET_STRING(iovec[n++], message);
-
-        forward_syslog_iovec(s, iovec, n, ucred, tv);
-
-        free(ident_buf);
-}
-
-static int fixup_priority(int priority) {
-
-        if ((priority & LOG_FACMASK) == 0)
-                return (priority & LOG_PRIMASK) | LOG_USER;
-
-        return priority;
-}
-
-static void forward_kmsg(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) {
-        struct iovec iovec[5];
-        char header_priority[6], header_pid[16];
-        int n = 0;
-        char *ident_buf = NULL;
-        int fd;
-
-        assert(s);
-        assert(priority >= 0);
-        assert(priority <= 999);
-        assert(message);
-
-        if (LOG_PRI(priority) > s->max_level_kmsg)
-                return;
-
-        /* Never allow messages with kernel facility to be written to
-         * kmsg, regardless where the data comes from. */
-        priority = fixup_priority(priority);
-
-        /* First: priority field */
-        snprintf(header_priority, sizeof(header_priority), "<%i>", priority);
-        char_array_0(header_priority);
-        IOVEC_SET_STRING(iovec[n++], header_priority);
-
-        /* Second: identifier and PID */
-        if (ucred) {
-                if (!identifier) {
-                        get_process_comm(ucred->pid, &ident_buf);
-                        identifier = ident_buf;
-                }
-
-                snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid);
-                char_array_0(header_pid);
-
-                if (identifier)
-                        IOVEC_SET_STRING(iovec[n++], identifier);
-
-                IOVEC_SET_STRING(iovec[n++], header_pid);
-        } else if (identifier) {
-                IOVEC_SET_STRING(iovec[n++], identifier);
-                IOVEC_SET_STRING(iovec[n++], ": ");
-        }
-
-        /* Fourth: message */
-        IOVEC_SET_STRING(iovec[n++], message);
-        IOVEC_SET_STRING(iovec[n++], "\n");
-
-        fd = open("/dev/kmsg", O_WRONLY|O_NOCTTY|O_CLOEXEC);
-        if (fd < 0) {
-                log_debug("Failed to open /dev/kmsg for logging: %s", strerror(errno));
-                goto finish;
-        }
-
-        if (writev(fd, iovec, n) < 0)
-                log_debug("Failed to write to /dev/kmsg for logging: %s", strerror(errno));
-
-        close_nointr_nofail(fd);
-
-finish:
-        free(ident_buf);
-}
-
-static void forward_console(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) {
+void server_forward_console(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) {
         struct iovec iovec[4];
         char header_pid[16];
         int n = 0, fd;
@@ -1052,116 +861,7 @@ finish:
         free(ident_buf);
 }
 
-static void read_identifier(const char **buf, char **identifier, char **pid) {
-        const char *p;
-        char *t;
-        size_t l, e;
-
-        assert(buf);
-        assert(identifier);
-        assert(pid);
-
-        p = *buf;
-
-        p += strspn(p, WHITESPACE);
-        l = strcspn(p, WHITESPACE);
-
-        if (l <= 0 ||
-            p[l-1] != ':')
-                return;
-
-        e = l;
-        l--;
-
-        if (p[l-1] == ']') {
-                size_t k = l-1;
-
-                for (;;) {
-
-                        if (p[k] == '[') {
-                                t = strndup(p+k+1, l-k-2);
-                                if (t)
-                                        *pid = t;
-
-                                l = k;
-                                break;
-                        }
-
-                        if (k == 0)
-                                break;
-
-                        k--;
-                }
-        }
-
-        t = strndup(p, l);
-        if (t)
-                *identifier = t;
-
-        *buf = p + e;
-        *buf += strspn(*buf, WHITESPACE);
-}
-
-static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv, const char *label, size_t label_len) {
-        char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *syslog_pid = NULL;
-        struct iovec iovec[N_IOVEC_META_FIELDS + 6];
-        unsigned n = 0;
-        int priority = LOG_USER | LOG_INFO;
-        char *identifier = NULL, *pid = NULL;
-        const char *orig;
-
-        assert(s);
-        assert(buf);
-
-        orig = buf;
-        parse_syslog_priority((char**) &buf, &priority);
-
-        if (s->forward_to_syslog)
-                forward_syslog_raw(s, priority, orig, ucred, tv);
-
-        skip_syslog_date((char**) &buf);
-        read_identifier(&buf, &identifier, &pid);
-
-        if (s->forward_to_kmsg)
-                forward_kmsg(s, priority, identifier, buf, ucred);
-
-        if (s->forward_to_console)
-                forward_console(s, priority, identifier, buf, ucred);
-
-        IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=syslog");
 
-        if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
-                IOVEC_SET_STRING(iovec[n++], syslog_priority);
-
-        if (priority & LOG_FACMASK)
-                if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], syslog_facility);
-
-        if (identifier) {
-                syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier);
-                if (syslog_identifier)
-                        IOVEC_SET_STRING(iovec[n++], syslog_identifier);
-        }
-
-        if (pid) {
-                syslog_pid = strappend("SYSLOG_PID=", pid);
-                if (syslog_pid)
-                        IOVEC_SET_STRING(iovec[n++], syslog_pid);
-        }
-
-        message = strappend("MESSAGE=", buf);
-        if (message)
-                IOVEC_SET_STRING(iovec[n++], message);
-
-        dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, label, label_len, NULL, priority);
-
-        free(message);
-        free(identifier);
-        free(pid);
-        free(syslog_priority);
-        free(syslog_facility);
-        free(syslog_identifier);
-}
 
 static bool valid_user_field(const char *p, size_t l) {
         const char *a;
@@ -1231,7 +931,7 @@ static void process_native_message(
 
                 if (e == p) {
                         /* Entry separator */
-                        dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority);
+                        server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority);
                         n = 0;
                         priority = LOG_INFO;
 
@@ -1257,7 +957,7 @@ static void process_native_message(
                         u = MAX((n+N_IOVEC_META_FIELDS+1) * 2U, 4U);
                         c = realloc(iovec, u * sizeof(struct iovec));
                         if (!c) {
-                                log_error("Out of memory");
+                                log_oom();
                                 break;
                         }
 
@@ -1344,7 +1044,7 @@ static void process_native_message(
 
                         k = malloc((e - p) + 1 + l);
                         if (!k) {
-                                log_error("Out of memory");
+                                log_oom();
                                 break;
                         }
 
@@ -1372,16 +1072,16 @@ static void process_native_message(
 
         if (message) {
                 if (s->forward_to_syslog)
-                        forward_syslog(s, priority, identifier, message, ucred, tv);
+                        server_forward_syslog(s, priority, identifier, message, ucred, tv);
 
                 if (s->forward_to_kmsg)
-                        forward_kmsg(s, priority, identifier, message, ucred);
+                        server_forward_kmsg(s, priority, identifier, message, ucred);
 
                 if (s->forward_to_console)
-                        forward_console(s, priority, identifier, message, ucred);
+                        server_forward_console(s, priority, identifier, message, ucred);
         }
 
-        dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority);
+        server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority);
 
 finish:
         for (j = 0; j < n; j++)  {
@@ -1437,7 +1137,7 @@ static void process_native_file(
 
         p = malloc(st.st_size);
         if (!p) {
-                log_error("Out of memory");
+                log_oom();
                 return;
         }
 
@@ -1467,16 +1167,16 @@ static int stdout_stream_log(StdoutStream *s, const char *p) {
         priority = s->priority;
 
         if (s->level_prefix)
-                parse_syslog_priority((char**) &p, &priority);
+                syslog_parse_priority((char**) &p, &priority);
 
         if (s->forward_to_syslog || s->server->forward_to_syslog)
-                forward_syslog(s->server, fixup_priority(priority), s->identifier, p, &s->ucred, NULL);
+                server_forward_syslog(s->server, syslog_fixup_facility(priority), s->identifier, p, &s->ucred, NULL);
 
         if (s->forward_to_kmsg || s->server->forward_to_kmsg)
-                forward_kmsg(s->server, priority, s->identifier, p, &s->ucred);
+                server_forward_kmsg(s->server, priority, s->identifier, p, &s->ucred);
 
         if (s->forward_to_console || s->server->forward_to_console)
-                forward_console(s->server, priority, s->identifier, p, &s->ucred);
+                server_forward_console(s->server, priority, s->identifier, p, &s->ucred);
 
         IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=stdout");
 
@@ -1504,7 +1204,7 @@ static int stdout_stream_log(StdoutStream *s, const char *p) {
         }
 #endif
 
-        dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, label, label_len, s->unit_id, priority);
+        server_dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, label, label_len, s->unit_id, priority);
 
         free(message);
         free(syslog_priority);
@@ -1529,10 +1229,8 @@ static int stdout_stream_line(StdoutStream *s, char *p) {
                         s->identifier = NULL;
                 else  {
                         s->identifier = strdup(p);
-                        if (!s->identifier) {
-                                log_error("Out of memory");
-                                return -ENOMEM;
-                        }
+                        if (!s->identifier)
+                                return log_oom();
                 }
 
                 s->state = STDOUT_STREAM_UNIT_ID;
@@ -1544,10 +1242,8 @@ static int stdout_stream_line(StdoutStream *s, char *p) {
                                 s->unit_id = NULL;
                         else  {
                                 s->unit_id = strdup(p);
-                                if (!s->unit_id) {
-                                        log_error("Out of memory");
-                                        return -ENOMEM;
-                                }
+                                if (!s->unit_id)
+                                        return log_oom();
                         }
                 }
 
@@ -1748,9 +1444,8 @@ static int stdout_stream_new(Server *s) {
 
         stream = new0(StdoutStream, 1);
         if (!stream) {
-                log_error("Out of memory.");
                 close_nointr_nofail(fd);
-                return -ENOMEM;
+                return log_oom();
         }
 
         stream->fd = fd;
@@ -1793,178 +1488,6 @@ fail:
         return r;
 }
 
-static int parse_kernel_timestamp(char **_p, usec_t *t) {
-        usec_t r;
-        int k, i;
-        char *p;
-
-        assert(_p);
-        assert(*_p);
-        assert(t);
-
-        p = *_p;
-
-        if (strlen(p) < 14 || p[0] != '[' || p[13] != ']' || p[6] != '.')
-                return 0;
-
-        r = 0;
-
-        for (i = 1; i <= 5; i++) {
-                r *= 10;
-
-                if (p[i] == ' ')
-                        continue;
-
-                k = undecchar(p[i]);
-                if (k < 0)
-                        return 0;
-
-                r += k;
-        }
-
-        for (i = 7; i <= 12; i++) {
-                r *= 10;
-
-                k = undecchar(p[i]);
-                if (k < 0)
-                        return 0;
-
-                r += k;
-        }
-
-        *t = r;
-        *_p += 14;
-        *_p += strspn(*_p, WHITESPACE);
-
-        return 1;
-}
-
-static bool is_us(const char *pid) {
-        pid_t t;
-
-        assert(pid);
-
-        if (parse_pid(pid, &t) < 0)
-                return false;
-
-        return t == getpid();
-}
-
-static void proc_kmsg_line(Server *s, const char *p) {
-        struct iovec iovec[N_IOVEC_META_FIELDS + 7];
-        char *message = NULL, *syslog_priority = NULL, *syslog_pid = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *source_time = NULL;
-        int priority = LOG_KERN | LOG_INFO;
-        unsigned n = 0;
-        usec_t usec;
-        char *identifier = NULL, *pid = NULL;
-
-        assert(s);
-        assert(p);
-
-        if (isempty(p))
-                return;
-
-        parse_syslog_priority((char **) &p, &priority);
-
-        if (s->forward_to_kmsg && (priority & LOG_FACMASK) != LOG_KERN)
-                return;
-
-        if (parse_kernel_timestamp((char **) &p, &usec) > 0) {
-                if (asprintf(&source_time, "_SOURCE_MONOTONIC_TIMESTAMP=%llu",
-                             (unsigned long long) usec) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], source_time);
-        }
-
-        IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=kernel");
-
-        if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
-                IOVEC_SET_STRING(iovec[n++], syslog_priority);
-
-        if ((priority & LOG_FACMASK) == LOG_KERN) {
-
-                if (s->forward_to_syslog)
-                        forward_syslog(s, priority, "kernel", p, NULL, NULL);
-
-                IOVEC_SET_STRING(iovec[n++], "SYSLOG_IDENTIFIER=kernel");
-        } else {
-                read_identifier(&p, &identifier, &pid);
-
-                /* Avoid any messages we generated ourselves via
-                 * log_info() and friends. */
-                if (pid && is_us(pid))
-                        goto finish;
-
-                if (s->forward_to_syslog)
-                        forward_syslog(s, priority, identifier, p, NULL, NULL);
-
-                if (identifier) {
-                        syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier);
-                        if (syslog_identifier)
-                                IOVEC_SET_STRING(iovec[n++], syslog_identifier);
-                }
-
-                if (pid) {
-                        syslog_pid = strappend("SYSLOG_PID=", pid);
-                        if (syslog_pid)
-                                IOVEC_SET_STRING(iovec[n++], syslog_pid);
-                }
-
-                if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], syslog_facility);
-        }
-
-        message = strappend("MESSAGE=", p);
-        if (message)
-                IOVEC_SET_STRING(iovec[n++], message);
-
-        dispatch_message(s, iovec, n, ELEMENTSOF(iovec), NULL, NULL, NULL, 0, NULL, priority);
-
-finish:
-        free(message);
-        free(syslog_priority);
-        free(syslog_identifier);
-        free(syslog_pid);
-        free(syslog_facility);
-        free(source_time);
-        free(identifier);
-        free(pid);
-}
-
-static void proc_kmsg_scan(Server *s) {
-        char *p;
-        size_t remaining;
-
-        assert(s);
-
-        p = s->proc_kmsg_buffer;
-        remaining = s->proc_kmsg_length;
-        for (;;) {
-                char *end;
-                size_t skip;
-
-                end = memchr(p, '\n', remaining);
-                if (end)
-                        skip = end - p + 1;
-                else if (remaining >= sizeof(s->proc_kmsg_buffer) - 1) {
-                        end = p + sizeof(s->proc_kmsg_buffer) - 1;
-                        skip = remaining;
-                } else
-                        break;
-
-                *end = 0;
-
-                proc_kmsg_line(s, p);
-
-                remaining -= skip;
-                p += skip;
-        }
-
-        if (p > s->proc_kmsg_buffer) {
-                memmove(s->proc_kmsg_buffer, p, remaining);
-                s->proc_kmsg_length = remaining;
-        }
-}
-
 static int system_journal_open(Server *s) {
         int r;
         char *fn;
@@ -2001,17 +1524,12 @@ static int system_journal_open(Server *s) {
                 if (!fn)
                         return -ENOMEM;
 
-                r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
+                r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &s->system_journal);
                 free(fn);
 
-                if (r >= 0) {
-                        journal_default_metrics(&s->system_metrics, s->system_journal->fd);
-
-                        s->system_journal->metrics = s->system_metrics;
-                        s->system_journal->compress = s->compress;
-
+                if (r >= 0)
                         server_fix_perms(s, s->system_journal, 0);
-                else if (r < 0) {
+                else if (r < 0) {
 
                         if (r != -ENOENT && r != -EROFS)
                                 log_warning("Failed to open system journal: %s", strerror(-r));
@@ -2033,7 +1551,7 @@ static int system_journal_open(Server *s) {
                          * if it already exists, so that we can flush
                          * it into the system journal */
 
-                        r = journal_file_open(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
+                        r = journal_file_open(fn, O_RDWR, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);
                         free(fn);
 
                         if (r < 0) {
@@ -2049,7 +1567,7 @@ static int system_journal_open(Server *s) {
                          * it if necessary. */
 
                         (void) mkdir_parents(fn, 0755);
-                        r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
+                        r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);
                         free(fn);
 
                         if (r < 0) {
@@ -2058,14 +1576,8 @@ static int system_journal_open(Server *s) {
                         }
                 }
 
-                if (s->runtime_journal) {
-                        journal_default_metrics(&s->runtime_metrics, s->runtime_journal->fd);
-
-                        s->runtime_journal->metrics = s->runtime_metrics;
-                        s->runtime_journal->compress = s->compress;
-
+                if (s->runtime_journal)
                         server_fix_perms(s, s->runtime_journal, 0);
-                }
         }
 
         return r;
@@ -2140,62 +1652,12 @@ finish:
         journal_file_close(s->runtime_journal);
         s->runtime_journal = NULL;
 
-        if (r >= 0) {
-                char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
-                sd_id128_to_string(machine, path + 17);
-                rm_rf(path, false, true, false);
-        }
+        if (r >= 0)
+                rm_rf("/run/log/journal", false, true, false);
 
         return r;
 }
 
-static int server_read_proc_kmsg(Server *s) {
-        ssize_t l;
-        assert(s);
-        assert(s->proc_kmsg_fd >= 0);
-
-        l = read(s->proc_kmsg_fd, s->proc_kmsg_buffer + s->proc_kmsg_length, sizeof(s->proc_kmsg_buffer) - 1 - s->proc_kmsg_length);
-        if (l == 0) /* the kernel is stupid and in some race
-                     * conditions returns 0 in the middle of the
-                     * stream. */
-                return 0;
-        if (l < 0) {
-
-                if (errno == EAGAIN || errno == EINTR)
-                        return 0;
-
-                log_error("Failed to read from kernel: %m");
-                return -errno;
-        }
-
-        s->proc_kmsg_length += l;
-
-        proc_kmsg_scan(s);
-        return 1;
-}
-
-static int server_flush_proc_kmsg(Server *s) {
-        int r;
-
-        assert(s);
-
-        if (s->proc_kmsg_fd < 0)
-                return 0;
-
-        log_info("Flushing /proc/kmsg...");
-
-        for (;;) {
-                r = server_read_proc_kmsg(s);
-                if (r < 0)
-                        return r;
-
-                if (r == 0)
-                        break;
-        }
-
-        return 0;
-}
-
 static int process_event(Server *s, struct epoll_event *ev) {
         assert(s);
         assert(ev);
@@ -2221,6 +1683,8 @@ static int process_event(Server *s, struct epoll_event *ev) {
                         return -errno;
                 }
 
+                log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo));
+
                 if (sfsi.ssi_signo == SIGUSR1) {
                         touch("/run/systemd/journal/flushed");
                         server_flush_to_var(s);
@@ -2233,10 +1697,9 @@ static int process_event(Server *s, struct epoll_event *ev) {
                         return 1;
                 }
 
-                log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
                 return 0;
 
-        } else if (ev->data.fd == s->proc_kmsg_fd) {
+        } else if (ev->data.fd == s->dev_kmsg_fd) {
                 int r;
 
                 if (ev->events != EPOLLIN) {
@@ -2244,7 +1707,7 @@ static int process_event(Server *s, struct epoll_event *ev) {
                         return -EIO;
                 }
 
-                r = server_read_proc_kmsg(s);
+                r = server_read_dev_kmsg(s);
                 if (r < 0)
                         return r;
 
@@ -2361,7 +1824,7 @@ static int process_event(Server *s, struct epoll_event *ev) {
                                         else
                                                 s->buffer[n] = 0;
 
-                                        process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len);
+                                        server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len);
                                 } else if (n_fds > 0)
                                         log_warning("Got file descriptors via syslog socket. Ignoring.");
 
@@ -2416,68 +1879,6 @@ static int process_event(Server *s, struct epoll_event *ev) {
         return 0;
 }
 
-static int open_syslog_socket(Server *s) {
-        union sockaddr_union sa;
-        int one, r;
-        struct epoll_event ev;
-
-        assert(s);
-
-        if (s->syslog_fd < 0) {
-
-                s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
-                if (s->syslog_fd < 0) {
-                        log_error("socket() failed: %m");
-                        return -errno;
-                }
-
-                zero(sa);
-                sa.un.sun_family = AF_UNIX;
-                strncpy(sa.un.sun_path, "/dev/log", sizeof(sa.un.sun_path));
-
-                unlink(sa.un.sun_path);
-
-                r = bind(s->syslog_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
-                if (r < 0) {
-                        log_error("bind() failed: %m");
-                        return -errno;
-                }
-
-                chmod(sa.un.sun_path, 0666);
-        } else
-                fd_nonblock(s->syslog_fd, 1);
-
-        one = 1;
-        r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
-        if (r < 0) {
-                log_error("SO_PASSCRED failed: %m");
-                return -errno;
-        }
-
-#ifdef HAVE_SELINUX
-        one = 1;
-        r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
-        if (r < 0)
-                log_warning("SO_PASSSEC failed: %m");
-#endif
-
-        one = 1;
-        r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
-        if (r < 0) {
-                log_error("SO_TIMESTAMP failed: %m");
-                return -errno;
-        }
-
-        zero(ev);
-        ev.events = EPOLLIN;
-        ev.data.fd = s->syslog_fd;
-        if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->syslog_fd, &ev) < 0) {
-                log_error("Failed to add syslog server fd to epoll object: %m");
-                return -errno;
-        }
-
-        return 0;
-}
 
 static int open_native_socket(Server*s) {
         union sockaddr_union sa;
@@ -2589,31 +1990,6 @@ static int open_stdout_socket(Server *s) {
         return 0;
 }
 
-static int open_proc_kmsg(Server *s) {
-        struct epoll_event ev;
-
-        assert(s);
-
-        if (!s->import_proc_kmsg)
-                return 0;
-
-        s->proc_kmsg_fd = open("/proc/kmsg", O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY);
-        if (s->proc_kmsg_fd < 0) {
-                log_warning("Failed to open /proc/kmsg, ignoring: %m");
-                return 0;
-        }
-
-        zero(ev);
-        ev.events = EPOLLIN;
-        ev.data.fd = s->proc_kmsg_fd;
-        if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->proc_kmsg_fd, &ev) < 0) {
-                log_error("Failed to add /proc/kmsg fd to epoll object: %m");
-                return -errno;
-        }
-
-        return 0;
-}
-
 static int open_signalfd(Server *s) {
         sigset_t mask;
         struct epoll_event ev;
@@ -2728,8 +2104,9 @@ static int server_init(Server *s) {
         assert(s);
 
         zero(*s);
-        s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->proc_kmsg_fd = -1;
+        s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1;
         s->compress = true;
+        s->seal = true;
 
         s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL;
         s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST;
@@ -2747,11 +2124,15 @@ static int server_init(Server *s) {
         server_parse_config_file(s);
         server_parse_proc_cmdline(s);
 
+        mkdir_p("/run/systemd/journal", 0755);
+
         s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
-        if (!s->user_journals) {
-                log_error("Out of memory.");
-                return -ENOMEM;
-        }
+        if (!s->user_journals)
+                return log_oom();
+
+        s->mmap = mmap_cache_new();
+        if (!s->mmap)
+                return log_oom();
 
         s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
         if (s->epoll_fd < 0) {
@@ -2800,7 +2181,7 @@ static int server_init(Server *s) {
                 }
         }
 
-        r = open_syslog_socket(s);
+        r = server_open_syslog_socket(s);
         if (r < 0)
                 return r;
 
@@ -2812,7 +2193,11 @@ static int server_init(Server *s) {
         if (r < 0)
                 return r;
 
-        r = open_proc_kmsg(s);
+        r = server_open_dev_kmsg(s);
+        if (r < 0)
+                return r;
+
+        r = server_open_kernel_seqnum(s);
         if (r < 0)
                 return r;
 
@@ -2820,6 +2205,10 @@ static int server_init(Server *s) {
         if (r < 0)
                 return r;
 
+        s->udev = udev_new();
+        if (!s->udev)
+                return -ENOMEM;
+
         s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst);
         if (!s->rate_limit)
                 return -ENOMEM;
@@ -2831,6 +2220,22 @@ static int server_init(Server *s) {
         return 0;
 }
 
+static void maybe_append_tags(Server *s) {
+#ifdef HAVE_GCRYPT
+        JournalFile *f;
+        Iterator i;
+        usec_t n;
+
+        n = now(CLOCK_REALTIME);
+
+        if (s->system_journal)
+                journal_file_maybe_append_tag(s->system_journal, n);
+
+        HASHMAP_FOREACH(f, s->user_journals, i)
+                journal_file_maybe_append_tag(f, n);
+#endif
+}
+
 static void server_done(Server *s) {
         JournalFile *f;
         assert(s);
@@ -2864,14 +2269,23 @@ static void server_done(Server *s) {
         if (s->stdout_fd >= 0)
                 close_nointr_nofail(s->stdout_fd);
 
-        if (s->proc_kmsg_fd >= 0)
-                close_nointr_nofail(s->proc_kmsg_fd);
+        if (s->dev_kmsg_fd >= 0)
+                close_nointr_nofail(s->dev_kmsg_fd);
 
         if (s->rate_limit)
                 journal_rate_limit_free(s->rate_limit);
 
+        if (s->kernel_seqnum)
+                munmap(s->kernel_seqnum, sizeof(uint64_t));
+
         free(s->buffer);
         free(s->tty_path);
+
+        if (s->mmap)
+                mmap_cache_unref(s->mmap);
+
+        if (s->udev)
+                udev_unref(s->udev);
 }
 
 int main(int argc, char *argv[]) {
@@ -2890,6 +2304,7 @@ int main(int argc, char *argv[]) {
 
         log_set_target(LOG_TARGET_SAFE);
         log_set_facility(LOG_SYSLOG);
+        log_set_max_level(LOG_DEBUG);
         log_parse_environment();
         log_open();
 
@@ -2901,10 +2316,10 @@ int main(int argc, char *argv[]) {
 
         server_vacuum(&server);
         server_flush_to_var(&server);
-        server_flush_proc_kmsg(&server);
+        server_flush_dev_kmsg(&server);
 
         log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
-        driver_message(&server, SD_MESSAGE_JOURNAL_START, "Journal started");
+        server_driver_message(&server, SD_MESSAGE_JOURNAL_START, "Journal started");
 
         sd_notify(false,
                   "READY=1\n"
@@ -2912,8 +2327,26 @@ int main(int argc, char *argv[]) {
 
         for (;;) {
                 struct epoll_event event;
+                int t;
+
+#ifdef HAVE_GCRYPT
+                usec_t u;
+
+                if (server.system_journal &&
+                    journal_file_next_evolve_usec(server.system_journal, &u)) {
+                        usec_t n;
 
-                r = epoll_wait(server.epoll_fd, &event, 1, -1);
+                        n = now(CLOCK_REALTIME);
+
+                        if (n >= u)
+                                t = 0;
+                        else
+                                t = (int) ((u - n + USEC_PER_MSEC - 1) / USEC_PER_MSEC);
+                } else
+#endif
+                        t = -1;
+
+                r = epoll_wait(server.epoll_fd, &event, 1, t);
                 if (r < 0) {
 
                         if (errno == EINTR)
@@ -2922,18 +2355,21 @@ int main(int argc, char *argv[]) {
                         log_error("epoll_wait() failed: %m");
                         r = -errno;
                         goto finish;
-                } else if (r == 0)
-                        break;
+                }
 
-                r = process_event(&server, &event);
-                if (r < 0)
-                        goto finish;
-                else if (r == 0)
-                        break;
+                if (r > 0) {
+                        r = process_event(&server, &event);
+                        if (r < 0)
+                                goto finish;
+                        else if (r == 0)
+                                break;
+                }
+
+                maybe_append_tags(&server);
         }
 
         log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
-        driver_message(&server, SD_MESSAGE_JOURNAL_STOP, "Journal stopped");
+        server_driver_message(&server, SD_MESSAGE_JOURNAL_STOP, "Journal stopped");
 
 finish:
         sd_notify(false,