chiark / gitweb /
journald: flush /run to /var as soon as it becomes available
[elogind.git] / src / journal / journald.c
index 9297ca6fb7da390be8877c4b81d6a55308d79fa0..1efe0420dbdb869b35a12841c3f6d95023c333a2 100644 (file)
 #include <sys/signalfd.h>
 #include <unistd.h>
 #include <fcntl.h>
+#include <sys/acl.h>
+#include <acl/libacl.h>
+#include <stddef.h>
+#include <sys/ioctl.h>
+#include <linux/sockios.h>
+#include <sys/statvfs.h>
 
 #include "hashmap.h"
-#include "journal-private.h"
+#include "journal-file.h"
 #include "sd-daemon.h"
 #include "socket-util.h"
+#include "acl-util.h"
+#include "cgroup-util.h"
+#include "list.h"
+#include "journal-rate-limit.h"
+#include "sd-journal.h"
+#include "journal-internal.h"
+
+#define USER_JOURNALS_MAX 1024
+#define STDOUT_STREAMS_MAX 4096
+
+#define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC)
+#define DEFAULT_RATE_LIMIT_BURST 200
+
+#define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
+
+typedef struct StdoutStream StdoutStream;
 
 typedef struct Server {
-        int syslog_fd;
         int epoll_fd;
         int signal_fd;
+        int syslog_fd;
+        int native_fd;
+        int stdout_fd;
 
+        JournalFile *runtime_journal;
         JournalFile *system_journal;
         Hashmap *user_journals;
+
+        uint64_t seqnum;
+
+        char *buffer;
+        size_t buffer_size;
+
+        JournalRateLimit *rate_limit;
+
+        JournalMetrics metrics;
+        uint64_t max_use;
+        bool compress;
+
+        uint64_t cached_available_space;
+        usec_t cached_available_space_timestamp;
+
+        LIST_HEAD(StdoutStream, stdout_streams);
+        unsigned n_stdout_streams;
 } Server;
 
-static void process_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
-        char *message = NULL, *pid = NULL, *uid = NULL, *gid = NULL,
+typedef enum StdoutStreamState {
+        STDOUT_STREAM_TAG,
+        STDOUT_STREAM_PRIORITY,
+        STDOUT_STREAM_PRIORITY_PREFIX,
+        STDOUT_STREAM_TEE_CONSOLE,
+        STDOUT_STREAM_RUNNING
+} StdoutStreamState;
+
+struct StdoutStream {
+        Server *server;
+        StdoutStreamState state;
+
+        int fd;
+
+        struct ucred ucred;
+
+        char *tag;
+        int priority;
+        bool priority_prefix:1;
+        bool tee_console:1;
+
+        char buffer[LINE_MAX+1];
+        size_t length;
+
+        LIST_FIELDS(StdoutStream, stdout_stream);
+};
+
+static int server_flush_to_var(Server *s);
+
+static uint64_t available_space(Server *s) {
+        char ids[33];
+        sd_id128_t machine;
+        char *p;
+        const char *f;
+        struct statvfs ss;
+        uint64_t sum = 0, avail = 0, ss_avail = 0;
+        int r;
+        DIR *d;
+        usec_t ts = now(CLOCK_MONOTONIC);
+
+        if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts)
+                return s->cached_available_space;
+
+        r = sd_id128_get_machine(&machine);
+        if (r < 0)
+                return 0;
+
+        if (s->system_journal)
+                f = "/var/log/journal/";
+        else
+                f = "/run/log/journal/";
+
+        p = strappend(f, sd_id128_to_string(machine, ids));
+        if (!p)
+                return 0;
+
+        d = opendir(p);
+        free(p);
+
+        if (!d)
+                return 0;
+
+        if (fstatvfs(dirfd(d), &ss) < 0)
+                goto finish;
+
+        for (;;) {
+                struct stat st;
+                struct dirent buf, *de;
+                int k;
+
+                k = readdir_r(d, &buf, &de);
+                if (k != 0) {
+                        r = -k;
+                        goto finish;
+                }
+
+                if (!de)
+                        break;
+
+                if (!dirent_is_file_with_suffix(de, ".journal"))
+                        continue;
+
+                if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
+                        continue;
+
+                sum += (uint64_t) st.st_blocks * (uint64_t) st.st_blksize;
+        }
+
+        avail = sum >= s->max_use ? 0 : s->max_use - sum;
+
+        ss_avail = ss.f_bsize * ss.f_bavail;
+
+        ss_avail = ss_avail < s->metrics.keep_free ? 0 : ss_avail - s->metrics.keep_free;
+
+        if (ss_avail < avail)
+                avail = ss_avail;
+
+        s->cached_available_space = avail;
+        s->cached_available_space_timestamp = ts;
+
+finish:
+        closedir(d);
+
+        return avail;
+}
+
+static void fix_perms(JournalFile *f, uid_t uid) {
+        acl_t acl;
+        acl_entry_t entry;
+        acl_permset_t permset;
+        int r;
+
+        assert(f);
+
+        r = fchmod_and_fchown(f->fd, 0640, 0, 0);
+        if (r < 0)
+                log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
+
+        if (uid <= 0)
+                return;
+
+        acl = acl_get_fd(f->fd);
+        if (!acl) {
+                log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
+                return;
+        }
+
+        r = acl_find_uid(acl, uid, &entry);
+        if (r <= 0) {
+
+                if (acl_create_entry(&acl, &entry) < 0 ||
+                    acl_set_tag_type(entry, ACL_USER) < 0 ||
+                    acl_set_qualifier(entry, &uid) < 0) {
+                        log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
+                        goto finish;
+                }
+        }
+
+        if (acl_get_permset(entry, &permset) < 0 ||
+            acl_add_perm(permset, ACL_READ) < 0 ||
+            acl_calc_mask(&acl) < 0) {
+                log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
+                goto finish;
+        }
+
+        if (acl_set_fd(f->fd, acl) < 0)
+                log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
+
+finish:
+        acl_free(acl);
+}
+
+static JournalFile* find_journal(Server *s, uid_t uid) {
+        char *p;
+        int r;
+        JournalFile *f;
+        char ids[33];
+        sd_id128_t machine;
+
+        assert(s);
+
+        /* We split up user logs only on /var, not on /run. If the
+         * runtime file is open, we write to it exclusively, in order
+         * to guarantee proper order as soon as we flush /run to
+         * /var and close the runtime file. */
+
+        if (s->runtime_journal)
+                return s->runtime_journal;
+
+        if (uid <= 0)
+                return s->system_journal;
+
+        r = sd_id128_get_machine(&machine);
+        if (r < 0)
+                return s->system_journal;
+
+        f = hashmap_get(s->user_journals, UINT32_TO_PTR(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)
+                return s->system_journal;
+
+        while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
+                /* Too many open? Then let's close one */
+                f = hashmap_steal_first(s->user_journals);
+                assert(f);
+                journal_file_close(f);
+        }
+
+        r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
+        free(p);
+
+        if (r < 0)
+                return s->system_journal;
+
+        fix_perms(f, uid);
+        f->metrics = s->metrics;
+        f->compress = s->compress;
+
+        r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
+        if (r < 0) {
+                journal_file_close(f);
+                return s->system_journal;
+        }
+
+        return f;
+}
+
+static void server_vacuum(Server *s) {
+        Iterator i;
+        void *k;
+        char *p;
+        char ids[33];
+        sd_id128_t machine;
+        int r;
+        JournalFile *f;
+
+        log_info("Rotating...");
+
+        if (s->runtime_journal) {
+                r = journal_file_rotate(&s->runtime_journal);
+                if (r < 0)
+                        log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
+        }
+
+        if (s->system_journal) {
+                r = journal_file_rotate(&s->system_journal);
+                if (r < 0)
+                        log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
+        }
+
+        HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
+                r = journal_file_rotate(&f);
+                if (r < 0)
+                        log_error("Failed to rotate %s: %s", f->path, strerror(-r));
+                else
+                        hashmap_replace(s->user_journals, k, f);
+        }
+
+        log_info("Vacuuming...");
+
+        r = sd_id128_get_machine(&machine);
+        if (r < 0) {
+                log_error("Failed to get machine ID: %s", strerror(-r));
+                return;
+        }
+
+        if (asprintf(&p, "/var/log/journal/%s", sd_id128_to_string(machine, ids)) < 0) {
+                log_error("Out of memory.");
+                return;
+        }
+
+        r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
+        if (r < 0 && r != -ENOENT)
+                log_error("Failed to vacuum %s: %s", p, strerror(-r));
+        free(p);
+
+        if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
+                log_error("Out of memory.");
+                return;
+        }
+
+        r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
+        if (r < 0 && r != -ENOENT)
+                log_error("Failed to vacuum %s: %s", p, strerror(-r));
+        free(p);
+
+        s->cached_available_space_timestamp = 0;
+}
+
+static char *shortened_cgroup_path(pid_t pid) {
+        int r;
+        char *process_path, *init_path, *path;
+
+        assert(pid > 0);
+
+        r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
+        if (r < 0)
+                return NULL;
+
+        r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
+        if (r < 0) {
+                free(process_path);
+                return NULL;
+        }
+
+        if (streq(init_path, "/"))
+                init_path[0] = 0;
+
+        if (startswith(process_path, init_path))
+                path = process_path + strlen(init_path);
+        else
+                path = process_path;
+
+        free(init_path);
+
+        return path;
+}
+
+static void dispatch_message_real(Server *s,
+                             struct iovec *iovec, unsigned n, unsigned m,
+                             struct ucred *ucred,
+                             struct timeval *tv) {
+
+        char *pid = NULL, *uid = NULL, *gid = NULL,
                 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
                 *comm = NULL, *cmdline = NULL, *hostname = NULL,
                 *audit_session = NULL, *audit_loginuid = NULL,
-                *syslog_priority = NULL, *syslog_facility = NULL,
-                *exe = NULL;
-        dual_timestamp ts;
-        struct iovec iovec[15];
-        unsigned n = 0;
+                *exe = NULL, *cgroup = NULL;
+
         char idbuf[33];
         sd_id128_t id;
         int r;
         char *t;
-        int priority = LOG_USER | LOG_INFO;
-
-        dual_timestamp_get(&ts);
-
-        parse_syslog_priority((char**) &buf, &priority);
-        skip_syslog_date((char**) &buf);
-
-        if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
-                IOVEC_SET_STRING(iovec[n++], syslog_priority);
-
-        if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
-                IOVEC_SET_STRING(iovec[n++], syslog_facility);
+        uid_t loginuid = 0, realuid = 0;
+        JournalFile *f;
+        bool vacuumed = false;
 
-        message = strappend("MESSAGE=", buf);
-        if (message)
-                IOVEC_SET_STRING(iovec[n++], message);
+        assert(s);
+        assert(iovec);
+        assert(n > 0);
+        assert(n + 13 <= m);
 
         if (ucred) {
                 uint32_t session;
-                uid_t loginuid;
+                char *path;
 
-                if (asprintf(&pid, "PID=%lu", (unsigned long) ucred->pid) >= 0)
+                realuid = ucred->uid;
+
+                if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], pid);
 
-                if (asprintf(&uid, "UID=%lu", (unsigned long) ucred->uid) >= 0)
+                if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], uid);
 
-                if (asprintf(&gid, "GID=%lu", (unsigned long) ucred->gid) >= 0)
+                if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], gid);
 
                 r = get_process_comm(ucred->pid, &t);
                 if (r >= 0) {
-                        comm = strappend("COMM=", t);
+                        comm = strappend("_COMM=", t);
                         if (comm)
                                 IOVEC_SET_STRING(iovec[n++], comm);
                         free(t);
@@ -94,7 +431,7 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = get_process_exe(ucred->pid, &t);
                 if (r >= 0) {
-                        exe = strappend("EXE=", t);
+                        exe = strappend("_EXE=", t);
                         if (comm)
                                 IOVEC_SET_STRING(iovec[n++], exe);
                         free(t);
@@ -102,7 +439,7 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
                 if (r >= 0) {
-                        cmdline = strappend("CMDLINE=", t);
+                        cmdline = strappend("_CMDLINE=", t);
                         if (cmdline)
                                 IOVEC_SET_STRING(iovec[n++], cmdline);
                         free(t);
@@ -110,72 +447,833 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = audit_session_from_pid(ucred->pid, &session);
                 if (r >= 0)
-                        if (asprintf(&audit_session, "AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
+                        if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
                                 IOVEC_SET_STRING(iovec[n++], audit_session);
 
-                r = audit_loginuid_from_pid(ucred->pid, &loginuid);
-                if (r >= 0)
-                        if (asprintf(&audit_loginuid, "AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
-                                IOVEC_SET_STRING(iovec[n++], audit_loginuid);
+                r = audit_loginuid_from_pid(ucred->pid, &loginuid);
+                if (r >= 0)
+                        if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
+                                IOVEC_SET_STRING(iovec[n++], audit_loginuid);
+
+                path = shortened_cgroup_path(ucred->pid);
+                if (path) {
+                        cgroup = strappend("_SYSTEMD_CGROUP=", path);
+                        if (cgroup)
+                                IOVEC_SET_STRING(iovec[n++], cgroup);
+
+                        free(path);
+                }
+        }
+
+        if (tv) {
+                if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
+                             (unsigned long long) timeval_load(tv)) >= 0)
+                        IOVEC_SET_STRING(iovec[n++], source_time);
+        }
+
+        /* Note that strictly speaking storing the boot id here is
+         * redundant since the entry includes this in-line
+         * anyway. However, we need this indexed, too. */
+        r = sd_id128_get_boot(&id);
+        if (r >= 0)
+                if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
+                        IOVEC_SET_STRING(iovec[n++], boot_id);
+
+        r = sd_id128_get_machine(&id);
+        if (r >= 0)
+                if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
+                        IOVEC_SET_STRING(iovec[n++], machine_id);
+
+        t = gethostname_malloc();
+        if (t) {
+                hostname = strappend("_HOSTNAME=", t);
+                if (hostname)
+                        IOVEC_SET_STRING(iovec[n++], hostname);
+                free(t);
+        }
+
+        assert(n <= m);
+
+        server_flush_to_var(s);
+
+retry:
+        f = find_journal(s, realuid == 0 ? 0 : loginuid);
+        if (!f)
+                log_warning("Dropping message, as we can't find a place to store the data.");
+        else {
+                r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
+
+                if (r == -E2BIG && !vacuumed) {
+                        log_info("Allocation limit reached.");
+
+                        server_vacuum(s);
+                        vacuumed = true;
+
+                        log_info("Retrying write.");
+                        goto retry;
+                }
+
+                if (r < 0)
+                        log_error("Failed to write entry, ignoring: %s", strerror(-r));
+        }
+
+        free(pid);
+        free(uid);
+        free(gid);
+        free(comm);
+        free(exe);
+        free(cmdline);
+        free(source_time);
+        free(boot_id);
+        free(machine_id);
+        free(hostname);
+        free(audit_session);
+        free(audit_loginuid);
+        free(cgroup);
+}
+
+static void dispatch_message(Server *s,
+                             struct iovec *iovec, unsigned n, unsigned m,
+                             struct ucred *ucred,
+                             struct timeval *tv,
+                             int priority) {
+        int rl;
+        char *path, *c;
+
+        assert(s);
+        assert(iovec || n == 0);
+
+        if (n == 0)
+                return;
+
+        if (!ucred)
+                goto finish;
+
+        path = shortened_cgroup_path(ucred->pid);
+        if (!path)
+                goto finish;
+
+        /* example: /user/lennart/3/foobar
+         *          /system/dbus.service/foobar
+         *
+         * So let's cut of everything past the third /, since that is
+         * wher user directories start */
+
+        c = strchr(path, '/');
+        if (c) {
+                c = strchr(c+1, '/');
+                if (c) {
+                        c = strchr(c+1, '/');
+                        if (c)
+                                *c = 0;
+                }
+        }
+
+        rl = journal_rate_limit_test(s->rate_limit, path, priority, available_space(s));
+
+        if (rl == 0) {
+                free(path);
+                return;
+        }
+
+        if (rl > 1) {
+                int j = 0;
+                char suppress_message[LINE_MAX];
+                struct iovec suppress_iovec[15];
+
+                /* Write a suppression message if we suppressed something */
+
+                snprintf(suppress_message, sizeof(suppress_message), "MESSAGE=Suppressed %u messages from %s", rl - 1, path);
+                char_array_0(suppress_message);
+
+                IOVEC_SET_STRING(suppress_iovec[j++], "PRIORITY=5");
+                IOVEC_SET_STRING(suppress_iovec[j++], suppress_message);
+
+                dispatch_message_real(s, suppress_iovec, j, ELEMENTSOF(suppress_iovec), NULL, NULL);
+        }
+
+        free(path);
+
+finish:
+        dispatch_message_real(s, iovec, n, m, ucred, tv);
+}
+
+static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
+        char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL;
+        struct iovec iovec[16];
+        unsigned n = 0;
+        int priority = LOG_USER | LOG_INFO;
+
+        assert(s);
+        assert(buf);
+
+        parse_syslog_priority((char**) &buf, &priority);
+        skip_syslog_date((char**) &buf);
+
+        if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
+                IOVEC_SET_STRING(iovec[n++], syslog_priority);
+
+        if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
+                IOVEC_SET_STRING(iovec[n++], syslog_facility);
+
+        message = strappend("MESSAGE=", buf);
+        if (message)
+                IOVEC_SET_STRING(iovec[n++], message);
+
+        dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority & LOG_PRIMASK);
+
+        free(message);
+        free(syslog_facility);
+        free(syslog_priority);
+}
+
+static bool valid_user_field(const char *p, size_t l) {
+        const char *a;
+
+        /* We kinda enforce POSIX syntax recommendations for
+           environment variables here, but make a couple of additional
+           requirements.
+
+           http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
+
+        /* No empty field names */
+        if (l <= 0)
+                return false;
+
+        /* Don't allow names longer than 64 chars */
+        if (l > 64)
+                return false;
+
+        /* Variables starting with an underscore are protected */
+        if (p[0] == '_')
+                return false;
+
+        /* Don't allow digits as first character */
+        if (p[0] >= '0' && p[0] <= '9')
+                return false;
+
+        /* Only allow A-Z0-9 and '_' */
+        for (a = p; a < p + l; a++)
+                if (!((*a >= 'A' && *a <= 'Z') ||
+                      (*a >= '0' && *a <= '9') ||
+                      *a == '_'))
+                        return false;
+
+        return true;
+}
+
+static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) {
+        struct iovec *iovec = NULL;
+        unsigned n = 0, m = 0, j;
+        const char *p;
+        size_t remaining;
+        int priority = LOG_INFO;
+
+        assert(s);
+        assert(buffer || n == 0);
+
+        p = buffer;
+        remaining = buffer_size;
+
+        while (remaining > 0) {
+                const char *e, *q;
+
+                e = memchr(p, '\n', remaining);
+
+                if (!e) {
+                        /* Trailing noise, let's ignore it, and flush what we collected */
+                        log_debug("Received message with trailing noise, ignoring.");
+                        break;
+                }
+
+                if (e == p) {
+                        /* Entry separator */
+                        dispatch_message(s, iovec, n, m, ucred, tv, priority);
+                        n = 0;
+                        priority = LOG_INFO;
+
+                        p++;
+                        remaining--;
+                        continue;
+                }
+
+                if (*p == '.' || *p == '#') {
+                        /* Ignore control commands for now, and
+                         * comments too. */
+                        remaining -= (e - p) + 1;
+                        p = e + 1;
+                        continue;
+                }
+
+                /* A property follows */
+
+                if (n+13 >= m) {
+                        struct iovec *c;
+                        unsigned u;
+
+                        u = MAX((n+13U) * 2U, 4U);
+                        c = realloc(iovec, u * sizeof(struct iovec));
+                        if (!c) {
+                                log_error("Out of memory");
+                                break;
+                        }
+
+                        iovec = c;
+                        m = u;
+                }
+
+                q = memchr(p, '=', e - p);
+                if (q) {
+                        if (valid_user_field(p, q - p)) {
+                                /* If the field name starts with an
+                                 * underscore, skip the variable,
+                                 * since that indidates a trusted
+                                 * field */
+                                iovec[n].iov_base = (char*) p;
+                                iovec[n].iov_len = e - p;
+                                n++;
+
+                                /* We need to determine the priority
+                                 * of this entry for the rate limiting
+                                 * logic */
+                                if (e - p == 10 &&
+                                    memcmp(p, "PRIORITY=", 10) == 0 &&
+                                    p[10] >= '0' &&
+                                    p[10] <= '9')
+                                        priority = p[10] - '0';
+                        }
+
+                        remaining -= (e - p) + 1;
+                        p = e + 1;
+                        continue;
+                } else {
+                        uint64_t l;
+                        char *k;
+
+                        if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
+                                log_debug("Failed to parse message, ignoring.");
+                                break;
+                        }
+
+                        memcpy(&l, e + 1, sizeof(uint64_t));
+                        l = le64toh(l);
+
+                        if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
+                            e[1+sizeof(uint64_t)+l] != '\n') {
+                                log_debug("Failed to parse message, ignoring.");
+                                break;
+                        }
+
+                        k = malloc((e - p) + 1 + l);
+                        if (!k) {
+                                log_error("Out of memory");
+                                break;
+                        }
+
+                        memcpy(k, p, e - p);
+                        k[e - p] = '=';
+                        memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
+
+                        if (valid_user_field(p, e - p)) {
+                                iovec[n].iov_base = k;
+                                iovec[n].iov_len = (e - p) + 1 + l;
+                                n++;
+                        } else
+                                free(k);
+
+                        remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
+                        p = e + 1 + sizeof(uint64_t) + l + 1;
+                }
+        }
+
+        dispatch_message(s, iovec, n, m, ucred, tv, priority);
+
+        for (j = 0; j < n; j++)
+                if (iovec[j].iov_base < buffer ||
+                    (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
+                        free(iovec[j].iov_base);
+}
+
+static int stdout_stream_log(StdoutStream *s, const char *p, size_t l) {
+        struct iovec iovec[15];
+        char *message = NULL, *syslog_priority = NULL;
+        unsigned n = 0;
+        size_t tag_len;
+        int priority;
+
+        assert(s);
+        assert(p);
+
+        priority = s->priority;
+
+        if (s->priority_prefix &&
+            l > 3 &&
+            p[0] == '<' &&
+            p[1] >= '0' && p[1] <= '7' &&
+            p[2] == '>') {
+
+                priority = p[1] - '0';
+                p += 3;
+                l -= 3;
+        }
+
+        if (l <= 0)
+                return 0;
+
+        if (asprintf(&syslog_priority, "PRIORITY=%i", priority) >= 0)
+                IOVEC_SET_STRING(iovec[n++], syslog_priority);
+
+        tag_len = s->tag ? strlen(s->tag) + 2: 0;
+        message = malloc(8 + tag_len + l);
+        if (message) {
+                memcpy(message, "MESSAGE=", 8);
+
+                if (s->tag) {
+                        memcpy(message+8, s->tag, tag_len-2);
+                        memcpy(message+8+tag_len-2, ": ", 2);
+                }
+
+                memcpy(message+8+tag_len, p, l);
+                iovec[n].iov_base = message;
+                iovec[n].iov_len = 8+tag_len+l;
+                n++;
+        }
+
+        dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority);
+
+        if (s->tee_console) {
+                int console;
+
+                console = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
+                if (console >= 0) {
+                        n = 0;
+                        if (s->tag) {
+                                IOVEC_SET_STRING(iovec[n++], s->tag);
+                                IOVEC_SET_STRING(iovec[n++], ": ");
+                        }
+
+                        iovec[n].iov_base = (void*) p;
+                        iovec[n].iov_len = l;
+                        n++;
+
+                        IOVEC_SET_STRING(iovec[n++], (char*) "\n");
+
+                        writev(console, iovec, n);
+                }
+        }
+
+        free(message);
+        free(syslog_priority);
+
+        return 0;
+}
+
+static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) {
+        assert(s);
+        assert(p);
+
+        while (l > 0 && strchr(WHITESPACE, *p)) {
+                l--;
+                p++;
+        }
+
+        while (l > 0 && strchr(WHITESPACE, *(p+l-1)))
+                l--;
+
+        switch (s->state) {
+
+        case STDOUT_STREAM_TAG:
+
+                if (l > 0) {
+                        s->tag = strndup(p, l);
+                        if (!s->tag) {
+                                log_error("Out of memory");
+                                return -EINVAL;
+                        }
+                }
+
+                s->state = STDOUT_STREAM_PRIORITY;
+                return 0;
+
+        case STDOUT_STREAM_PRIORITY:
+                if (l != 1 || *p < '0' || *p > '7') {
+                        log_warning("Failed to parse log priority line.");
+                        return -EINVAL;
+                }
+
+                s->priority = *p - '0';
+                s->state = STDOUT_STREAM_PRIORITY_PREFIX;
+                return 0;
+
+        case STDOUT_STREAM_PRIORITY_PREFIX:
+                if (l != 1 || *p < '0' || *p > '1') {
+                        log_warning("Failed to parse priority prefix line.");
+                        return -EINVAL;
+                }
+
+                s->priority_prefix = *p - '0';
+                s->state = STDOUT_STREAM_TEE_CONSOLE;
+                return 0;
+
+        case STDOUT_STREAM_TEE_CONSOLE:
+                if (l != 1 || *p < '0' || *p > '1') {
+                        log_warning("Failed to parse tee to console line.");
+                        return -EINVAL;
+                }
+
+                s->tee_console = *p - '0';
+                s->state = STDOUT_STREAM_RUNNING;
+                return 0;
+
+        case STDOUT_STREAM_RUNNING:
+                return stdout_stream_log(s, p, l);
+        }
+
+        assert_not_reached("Unknown stream state");
+}
+
+static int stdout_stream_scan(StdoutStream *s, bool force_flush) {
+        char *p;
+        size_t remaining;
+        int r;
+
+        assert(s);
+
+        p = s->buffer;
+        remaining = s->length;
+        for (;;) {
+                char *end;
+                size_t skip;
+
+                end = memchr(p, '\n', remaining);
+                if (!end) {
+                        if (remaining >= LINE_MAX) {
+                                end = p + LINE_MAX;
+                                skip = LINE_MAX;
+                        } else
+                                break;
+                } else
+                        skip = end - p + 1;
+
+                r = stdout_stream_line(s, p, end - p);
+                if (r < 0)
+                        return r;
+
+                remaining -= skip;
+                p += skip;
+        }
+
+        if (force_flush && remaining > 0) {
+                r = stdout_stream_line(s, p, remaining);
+                if (r < 0)
+                        return r;
+
+                p += remaining;
+                remaining = 0;
+        }
+
+        if (p > s->buffer) {
+                memmove(s->buffer, p, remaining);
+                s->length = remaining;
+        }
+
+        return 0;
+}
+
+static int stdout_stream_process(StdoutStream *s) {
+        ssize_t l;
+        int r;
+
+        assert(s);
+
+        l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length);
+        if (l < 0) {
+
+                if (errno == EAGAIN)
+                        return 0;
+
+                log_warning("Failed to read from stream: %m");
+                return -errno;
+        }
+
+        if (l == 0) {
+                r = stdout_stream_scan(s, true);
+                if (r < 0)
+                        return r;
+
+                return 0;
+        }
+
+        s->length += l;
+        r = stdout_stream_scan(s, false);
+        if (r < 0)
+                return r;
+
+        return 1;
+
+}
+
+static void stdout_stream_free(StdoutStream *s) {
+        assert(s);
+
+        if (s->server) {
+                assert(s->server->n_stdout_streams > 0);
+                s->server->n_stdout_streams --;
+                LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s);
+        }
+
+        if (s->fd >= 0) {
+                if (s->server)
+                        epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL);
+
+                close_nointr_nofail(s->fd);
+        }
+
+        free(s->tag);
+        free(s);
+}
+
+static int stdout_stream_new(Server *s) {
+        StdoutStream *stream;
+        int fd, r;
+        socklen_t len;
+        struct epoll_event ev;
+
+        assert(s);
+
+        fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC);
+        if (fd < 0) {
+                if (errno == EAGAIN)
+                        return 0;
+
+                log_error("Failed to accept stdout connection: %m");
+                return -errno;
+        }
+
+        if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) {
+                log_warning("Too many stdout streams, refusing connection.");
+                close_nointr_nofail(fd);
+                return 0;
+        }
+
+        stream = new0(StdoutStream, 1);
+        if (!stream) {
+                log_error("Out of memory.");
+                close_nointr_nofail(fd);
+                return -ENOMEM;
+        }
+
+        stream->fd = fd;
+
+        len = sizeof(stream->ucred);
+        if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) {
+                log_error("Failed to determine peer credentials: %m");
+                r = -errno;
+                goto fail;
+        }
+
+        if (shutdown(fd, SHUT_WR) < 0) {
+                log_error("Failed to shutdown writing side of socket: %m");
+                r = -errno;
+                goto fail;
+        }
+
+        zero(ev);
+        ev.data.ptr = stream;
+        ev.events = EPOLLIN;
+        if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) {
+                log_error("Failed to add stream to event loop: %m");
+                r = -errno;
+                goto fail;
+        }
+
+        stream->server = s;
+        LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream);
+        s->n_stdout_streams ++;
+
+        return 0;
+
+fail:
+        stdout_stream_free(stream);
+        return r;
+}
+
+static int system_journal_open(Server *s) {
+        int r;
+        char *fn;
+        sd_id128_t machine;
+        char ids[33];
+
+        r = sd_id128_get_machine(&machine);
+        if (r < 0)
+                return r;
+
+        sd_id128_to_string(machine, ids);
+
+        if (!s->system_journal) {
+
+                /* First try to create the machine path, but not the prefix */
+                fn = strappend("/var/log/journal/", ids);
+                if (!fn)
+                        return -ENOMEM;
+                (void) mkdir(fn, 0755);
+                free(fn);
+
+                /* The create the system journal file */
+                fn = join("/var/log/journal/", ids, "/system.journal", NULL);
+                if (!fn)
+                        return -ENOMEM;
+
+                r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
+                free(fn);
+
+                if (r >= 0) {
+                        s->system_journal->metrics = s->metrics;
+                        s->system_journal->compress = s->compress;
+
+                        fix_perms(s->system_journal, 0);
+                } else if (r < 0) {
+
+                        if (r == -ENOENT)
+                                r = 0;
+                        else {
+                                log_error("Failed to open system journal: %s", strerror(-r));
+                                return r;
+                        }
+                }
+        }
+
+        if (!s->runtime_journal) {
+
+                fn = join("/run/log/journal/", ids, "/system.journal", NULL);
+                if (!fn)
+                        return -ENOMEM;
+
+                if (s->system_journal) {
+
+                        /* Try to open the runtime journal, but only
+                         * 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);
+                        free(fn);
+
+                        if (r < 0) {
+
+                                if (r == -ENOENT)
+                                        r = 0;
+                                else {
+                                        log_error("Failed to open runtime journal: %s", strerror(-r));
+                                        return r;
+                                }
+                        }
+
+                } else {
+
+                        /* OK, we really need the runtime journal, so create
+                         * it if necessary. */
+
+                        (void) mkdir_parents(fn, 0755);
+                        r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
+                        free(fn);
+
+                        if (r < 0) {
+                                log_error("Failed to open runtime journal: %s", strerror(-r));
+                                return r;
+                        }
+                }
+
+                if (s->runtime_journal) {
+                        s->runtime_journal->metrics = s->metrics;
+                        s->runtime_journal->compress = s->compress;
+
+                        fix_perms(s->runtime_journal, 0);
+                }
+        }
+
+        return r;
+}
+
+static int server_flush_to_var(Server *s) {
+        char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
+        Object *o = NULL;
+        int r;
+        sd_id128_t machine;
+        sd_journal *j;
+
+        assert(s);
+
+        system_journal_open(s);
+
+        if (!s->system_journal || !s->runtime_journal)
+                return 0;
+
+        r = sd_id128_get_machine(&machine);
+        if (r < 0) {
+                log_error("Failed to get machine id: %s", strerror(-r));
+                return r;
         }
 
-        if (tv) {
-                if (asprintf(&source_time, "SOURCE_REALTIME_TIMESTAMP=%llu",
-                             (unsigned long long) timeval_load(tv)) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], source_time);
+        r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);
+        if (r < 0) {
+                log_error("Failed to read runtime journal: %s", strerror(-r));
+                return r;
         }
 
-        r = sd_id128_get_boot(&id);
-        if (r >= 0)
-                if (asprintf(&boot_id, "BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], boot_id);
+        SD_JOURNAL_FOREACH(j) {
+                JournalFile *f;
 
-        r = sd_id128_get_machine(&id);
-        if (r >= 0)
-                if (asprintf(&machine_id, "MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
-                        IOVEC_SET_STRING(iovec[n++], machine_id);
+                f = j->current_file;
+                assert(f && f->current_offset > 0);
 
-        t = gethostname_malloc();
-        if (t) {
-                hostname = strappend("HOSTNAME=", t);
-                if (hostname)
-                        IOVEC_SET_STRING(iovec[n++], hostname);
-                free(t);
+                r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
+                if (r < 0) {
+                        log_error("Can't read entry: %s", strerror(-r));
+                        goto finish;
+                }
+
+                r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
+                if (r == -E2BIG) {
+                        log_info("Allocation limit reached.");
+
+                        journal_file_post_change(s->system_journal);
+                        server_vacuum(s);
+
+                        r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
+                }
+
+                if (r < 0) {
+                        log_error("Can't write entry: %s", strerror(-r));
+                        goto finish;
+                }
         }
 
-        r = journal_file_append_entry(s->system_journal, &ts, iovec, n, NULL, NULL);
-        if (r < 0)
-                log_error("Failed to write entry: %s", strerror(-r));
+finish:
+        journal_file_post_change(s->system_journal);
 
+        journal_file_close(s->runtime_journal);
+        s->runtime_journal = NULL;
 
-        free(message);
-        free(pid);
-        free(uid);
-        free(gid);
-        free(comm);
-        free(cmdline);
-        free(source_time);
-        free(boot_id);
-        free(machine_id);
-        free(hostname);
-        free(audit_session);
-        free(audit_loginuid);
-        free(syslog_facility);
-        free(syslog_priority);
+        if (r >= 0) {
+                sd_id128_to_string(machine, path + 17);
+                rm_rf(path, false, true, false);
+        }
+
+        return r;
 }
 
 static int process_event(Server *s, struct epoll_event *ev) {
         assert(s);
 
-        if (ev->events != EPOLLIN) {
-                log_info("Got invalid event from epoll.");
-                return -EIO;
-        }
-
         if (ev->data.fd == s->signal_fd) {
                 struct signalfd_siginfo sfsi;
                 ssize_t n;
 
+                if (ev->events != EPOLLIN) {
+                        log_info("Got invalid event from epoll.");
+                        return -EIO;
+                }
+
                 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
                 if (n != sizeof(sfsi)) {
 
@@ -188,12 +1286,23 @@ static int process_event(Server *s, struct epoll_event *ev) {
                         return -errno;
                 }
 
+                if (sfsi.ssi_signo == SIGUSR1) {
+                        server_flush_to_var(s);
+                        return 0;
+                }
+
                 log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
                 return 0;
 
-        } else {
+        } else if (ev->data.fd == s->native_fd ||
+                   ev->data.fd == s->syslog_fd) {
+
+                if (ev->events != EPOLLIN) {
+                        log_info("Got invalid event from epoll.");
+                        return -EIO;
+                }
+
                 for (;;) {
-                        char buf[LINE_MAX+1];
                         struct msghdr msghdr;
                         struct iovec iovec;
                         struct ucred *ucred = NULL;
@@ -205,11 +1314,35 @@ static int process_event(Server *s, struct epoll_event *ev) {
                                             CMSG_SPACE(sizeof(struct timeval))];
                         } control;
                         ssize_t n;
-                        char *e;
+                        int v;
+
+                        if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
+                                log_error("SIOCINQ failed: %m");
+                                return -errno;
+                        }
+
+                        if (v <= 0)
+                                return 1;
+
+                        if (s->buffer_size < (size_t) v) {
+                                void *b;
+                                size_t l;
+
+                                l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
+                                b = realloc(s->buffer, l+1);
+
+                                if (!b) {
+                                        log_error("Couldn't increase buffer.");
+                                        return -ENOMEM;
+                                }
+
+                                s->buffer_size = l;
+                                s->buffer = b;
+                        }
 
                         zero(iovec);
-                        iovec.iov_base = buf;
-                        iovec.iov_len = sizeof(buf)-1;
+                        iovec.iov_base = s->buffer;
+                        iovec.iov_len = s->buffer_size;
 
                         zero(control);
                         zero(msghdr);
@@ -240,64 +1373,67 @@ static int process_event(Server *s, struct epoll_event *ev) {
                                         tv = (struct timeval*) CMSG_DATA(cmsg);
                         }
 
-                        e = memchr(buf, '\n', n);
-                        if (e)
-                                *e = 0;
-                        else
-                                buf[n] = 0;
+                        if (ev->data.fd == s->syslog_fd) {
+                                char *e;
+
+                                e = memchr(s->buffer, '\n', n);
+                                if (e)
+                                        *e = 0;
+                                else
+                                        s->buffer[n] = 0;
 
-                        process_message(s, strstrip(buf), ucred, tv);
+                                process_syslog_message(s, strstrip(s->buffer), ucred, tv);
+                        } else
+                                process_native_message(s, s->buffer, n, ucred, tv);
                 }
-        }
 
-        return 1;
-}
+                return 1;
 
+        } else if (ev->data.fd == s->stdout_fd) {
 
-static int open_system_journal(JournalFile **f) {
-        int r;
+                if (ev->events != EPOLLIN) {
+                        log_info("Got invalid event from epoll.");
+                        return -EIO;
+                }
 
-        r = journal_file_open(NULL, "/var/log/journal/system.journal", O_RDWR|O_CREAT, 0644, f);
-        if (r == -ENOENT) {
-                mkdir_p("/run/log/journal", 0755);
+                stdout_stream_new(s);
+                return 1;
 
-                r = journal_file_open(NULL, "/run/log/journal/system.journal", O_RDWR|O_CREAT, 0644, f);
-        }
+        } else {
+                StdoutStream *stream;
 
-        return r;
-}
+                if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
+                        log_info("Got invalid event from epoll.");
+                        return -EIO;
+                }
 
-static int server_init(Server *s) {
-        int n, one, r;
-        struct epoll_event ev;
-        sigset_t mask;
+                /* If it is none of the well-known fds, it must be an
+                 * stdout stream fd. Note that this is a bit ugly here
+                 * (since we rely that none of the well-known fds
+                 * could be interpreted as pointer), but nonetheless
+                 * safe, since the well-known fds would never get an
+                 * fd > 4096, i.e. beyond the first memory page */
 
-        assert(s);
+                stream = ev->data.ptr;
 
-        zero(*s);
-        s->syslog_fd = s->signal_fd = -1;
+                if (stdout_stream_process(stream) <= 0)
+                        stdout_stream_free(stream);
 
-        s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
-        if (s->epoll_fd < 0) {
-                log_error("Failed to create epoll object: %m");
-                return -errno;
+                return 1;
         }
 
-        n = sd_listen_fds(true);
-        if (n < 0) {
-                log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
-                return n;
-        }
+        log_error("Unknown event.");
+        return 0;
+}
 
-        if (n > 1) {
-                log_error("Too many file descriptors passed.");
-                return -EINVAL;
-        }
+static int open_syslog_socket(Server *s) {
+        union sockaddr_union sa;
+        int one, r;
+        struct epoll_event ev;
 
-        if (n == 1)
-                s->syslog_fd = SD_LISTEN_FDS_START;
-        else {
-                union sockaddr_union sa;
+        assert(s);
+
+        if (s->syslog_fd < 0) {
 
                 s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
                 if (s->syslog_fd < 0) {
@@ -311,7 +1447,7 @@ static int server_init(Server *s) {
 
                 unlink(sa.un.sun_path);
 
-                r = bind(s->syslog_fd, &sa.sa, sizeof(sa.un));
+                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;
@@ -338,24 +1474,122 @@ static int server_init(Server *s) {
         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 server fd to epoll object: %m");
+                log_error("Failed to add syslog server fd to epoll object: %m");
                 return -errno;
         }
 
-        s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
-        if (!s->user_journals) {
-                log_error("Out of memory.");
-                return -ENOMEM;
+        return 0;
+}
+
+static int open_native_socket(Server*s) {
+        union sockaddr_union sa;
+        int one, r;
+        struct epoll_event ev;
+
+        assert(s);
+
+        if (s->native_fd < 0) {
+
+                s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
+                if (s->native_fd < 0) {
+                        log_error("socket() failed: %m");
+                        return -errno;
+                }
+
+                zero(sa);
+                sa.un.sun_family = AF_UNIX;
+                strncpy(sa.un.sun_path, "/run/systemd/journal", sizeof(sa.un.sun_path));
+
+                unlink(sa.un.sun_path);
+
+                r = bind(s->native_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);
         }
 
-        r = open_system_journal(&s->system_journal);
+        one = 1;
+        r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
         if (r < 0) {
-                log_error("Failed to open journal: %s", strerror(-r));
-                return r;
+                log_error("SO_PASSCRED failed: %m");
+                return -errno;
+        }
+
+        one = 1;
+        r = setsockopt(s->native_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->native_fd;
+        if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->native_fd, &ev) < 0) {
+                log_error("Failed to add native server fd to epoll object: %m");
+                return -errno;
+        }
+
+        return 0;
+}
+
+static int open_stdout_socket(Server *s) {
+        union sockaddr_union sa;
+        int r;
+        struct epoll_event ev;
+
+        assert(s);
+
+        if (s->stdout_fd < 0) {
+
+                s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0);
+                if (s->stdout_fd < 0) {
+                        log_error("socket() failed: %m");
+                        return -errno;
+                }
+
+                zero(sa);
+                sa.un.sun_family = AF_UNIX;
+                strncpy(sa.un.sun_path, "/run/systemd/stdout", sizeof(sa.un.sun_path));
+
+                unlink(sa.un.sun_path);
+
+                r = bind(s->stdout_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);
+
+                if (listen(s->stdout_fd, SOMAXCONN) < 0) {
+                        log_error("liste() failed: %m");
+                        return -errno;
+                }
         }
 
+        zero(ev);
+        ev.events = EPOLLIN;
+        ev.data.fd = s->stdout_fd;
+        if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->stdout_fd, &ev) < 0) {
+                log_error("Failed to add stdout server fd to epoll object: %m");
+                return -errno;
+        }
+
+        return 0;
+}
+
+static int open_signalfd(Server *s) {
+        sigset_t mask;
+        struct epoll_event ev;
+
+        assert(s);
+
         assert_se(sigemptyset(&mask) == 0);
-        sigset_add_many(&mask, SIGINT, SIGTERM, -1);
+        sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, -1);
         assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
 
         s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
@@ -376,13 +1610,112 @@ static int server_init(Server *s) {
         return 0;
 }
 
+static int server_init(Server *s) {
+        int n, r, fd;
+
+        assert(s);
+
+        zero(*s);
+        s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = -1;
+        s->metrics.max_size = DEFAULT_MAX_SIZE;
+        s->metrics.min_size = DEFAULT_MIN_SIZE;
+        s->metrics.keep_free = DEFAULT_KEEP_FREE;
+        s->max_use = DEFAULT_MAX_USE;
+        s->compress = true;
+
+        s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
+        if (!s->user_journals) {
+                log_error("Out of memory.");
+                return -ENOMEM;
+        }
+
+        s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
+        if (s->epoll_fd < 0) {
+                log_error("Failed to create epoll object: %m");
+                return -errno;
+        }
+
+        n = sd_listen_fds(true);
+        if (n < 0) {
+                log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
+                return n;
+        }
+
+        for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
+
+                if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/native", 0) > 0) {
+
+                        if (s->native_fd >= 0) {
+                                log_error("Too many native sockets passed.");
+                                return -EINVAL;
+                        }
+
+                        s->native_fd = fd;
+
+                } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/stdout", 0) > 0) {
+
+                        if (s->stdout_fd >= 0) {
+                                log_error("Too many stdout sockets passed.");
+                                return -EINVAL;
+                        }
+
+                        s->stdout_fd = fd;
+
+                } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
+
+                        if (s->syslog_fd >= 0) {
+                                log_error("Too many /dev/log sockets passed.");
+                                return -EINVAL;
+                        }
+
+                        s->syslog_fd = fd;
+
+                } else {
+                        log_error("Unknown socket passed.");
+                        return -EINVAL;
+                }
+        }
+
+        r = open_syslog_socket(s);
+        if (r < 0)
+                return r;
+
+        r = open_native_socket(s);
+        if (r < 0)
+                return r;
+
+        r = open_stdout_socket(s);
+        if (r < 0)
+                return r;
+
+        r = system_journal_open(s);
+        if (r < 0)
+                return r;
+
+        r = open_signalfd(s);
+        if (r < 0)
+                return r;
+
+        s->rate_limit = journal_rate_limit_new(DEFAULT_RATE_LIMIT_INTERVAL, DEFAULT_RATE_LIMIT_BURST);
+        if (!s->rate_limit)
+                return -ENOMEM;
+
+        return 0;
+}
+
 static void server_done(Server *s) {
         JournalFile *f;
         assert(s);
 
+        while (s->stdout_streams)
+                stdout_stream_free(s->stdout_streams);
+
         if (s->system_journal)
                 journal_file_close(s->system_journal);
 
+        if (s->runtime_journal)
+                journal_file_close(s->runtime_journal);
+
         while ((f = hashmap_steal_first(s->user_journals)))
                 journal_file_close(f);
 
@@ -396,6 +1729,15 @@ static void server_done(Server *s) {
 
         if (s->syslog_fd >= 0)
                 close_nointr_nofail(s->syslog_fd);
+
+        if (s->native_fd >= 0)
+                close_nointr_nofail(s->native_fd);
+
+        if (s->stdout_fd >= 0)
+                close_nointr_nofail(s->stdout_fd);
+
+        if (s->rate_limit)
+                journal_rate_limit_free(s->rate_limit);
 }
 
 int main(int argc, char *argv[]) {
@@ -412,7 +1754,8 @@ int main(int argc, char *argv[]) {
                 return EXIT_FAILURE;
         }
 
-        log_set_target(LOG_TARGET_AUTO);
+        log_set_target(LOG_TARGET_CONSOLE);
+        log_set_max_level(LOG_DEBUG);
         log_parse_environment();
         log_open();
 
@@ -426,7 +1769,10 @@ int main(int argc, char *argv[]) {
 
         sd_notify(false,
                   "READY=1\n"
-                  "STATUS=Processing messages...");
+                  "STATUS=Processing requests...");
+
+        server_vacuum(&server);
+        server_flush_to_var(&server);
 
         for (;;) {
                 struct epoll_event event;
@@ -450,6 +1796,8 @@ int main(int argc, char *argv[]) {
                         break;
         }
 
+        log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
+
 finish:
         sd_notify(false,
                   "STATUS=Shutting down...");