chiark / gitweb /
export USEC_INITIALIZED= and take timestamp on message receive time
[elogind.git] / udev / udevd.c
index e4fa2b63e10a58044d6cf356139bef8095aaadf0..325ad034c2b2bc05554824acde20beadd39f7fca 100644 (file)
@@ -1,6 +1,8 @@
 /*
- * Copyright (C) 2004-2008 Kay Sievers <kay.sievers@vrfy.org>
+ * Copyright (C) 2004-2011 Kay Sievers <kay.sievers@vrfy.org>
  * Copyright (C) 2004 Chris Friesen <chris_friesen@sympatico.ca>
+ * Copyright (C) 2009 Canonical Ltd.
+ * Copyright (C) 2009 Scott James Remnant <scott@netsplit.com>
  *
  * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
 #include <errno.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdbool.h>
 #include <string.h>
 #include <ctype.h>
 #include <fcntl.h>
 #include <time.h>
 #include <getopt.h>
 #include <dirent.h>
-#include <sys/select.h>
+#include <sys/time.h>
+#include <sys/prctl.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <sys/signalfd.h>
+#include <sys/epoll.h>
 #include <sys/poll.h>
 #include <sys/wait.h>
 #include <sys/stat.h>
 #include <sys/ioctl.h>
-#ifdef HAVE_INOTIFY
 #include <sys/inotify.h>
-#endif
+#include <sys/utsname.h>
 
 #include "udev.h"
+#include "sd-daemon.h"
 
-#define UDEVD_PRIORITY                 -4
-#define UDEV_PRIORITY                  -2
-
-/* maximum limit of forked childs */
-#define UDEVD_MAX_CHILDS               256
-
-static int debug;
+static bool debug;
 
 static void log_fn(struct udev *udev, int priority,
                   const char *file, int line, const char *fn,
                   const char *format, va_list args)
 {
        if (debug) {
-               fprintf(stderr, "[%d] %s: ", (int) getpid(), fn);
-               vfprintf(stderr, format, args);
+               char buf[1024];
+               struct timespec ts;
+
+               vsnprintf(buf, sizeof(buf), format, args);
+               clock_gettime(CLOCK_MONOTONIC, &ts);
+               fprintf(stderr, "[%llu.%06u] [%u] %s: %s",
+                       (unsigned long long) ts.tv_sec, (unsigned int) ts.tv_nsec/1000,
+                       (int) getpid(), fn, buf);
        } else {
                vsyslog(priority, format, args);
        }
 }
 
-static void reap_sigchilds(void);
-
-static int debug_trace;
 static struct udev_rules *rules;
+static struct udev_queue_export *udev_queue_export;
 static struct udev_ctrl *udev_ctrl;
-static struct udev_monitor *kernel_monitor;
-static volatile int sigchilds_waiting;
-static volatile int udev_exit;
-static volatile int reload_config;
-static volatile int signal_received;
-static int run_exec_q;
-static int stop_exec_q;
-static int max_childs;
-static int childs;
-static struct udev_list_node event_list;
+static struct udev_monitor *monitor;
+static int worker_watch[2] = { -1, -1 };
+static int fd_signal = -1;
+static int fd_ep = -1;
+static int fd_inotify = -1;
+static bool stop_exec_queue;
+static bool reload_config;
+static int children;
+static int children_max;
+static int exec_delay;
+static sigset_t sigmask_orig;
+static UDEV_LIST(event_list);
+static UDEV_LIST(worker_list);
+static bool udev_exit;
 
 enum event_state {
+       EVENT_UNDEF,
        EVENT_QUEUED,
-       EVENT_FINISHED,
-       EVENT_FAILED,
+       EVENT_RUNNING,
 };
 
-static struct udev_event *node_to_event(struct udev_list_node *node)
+struct event {
+       struct udev_list_node node;
+       struct udev *udev;
+       struct udev_device *dev;
+       enum event_state state;
+       int exitcode;
+       unsigned long long int delaying_seqnum;
+       unsigned long long int seqnum;
+       const char *devpath;
+       size_t devpath_len;
+       const char *devpath_old;
+       dev_t devnum;
+       bool is_block;
+       int ifindex;
+};
+
+static struct event *node_to_event(struct udev_list_node *node)
 {
        char *event;
 
        event = (char *)node;
-       event -= offsetof(struct udev_event, node);
-       return (struct udev_event *)event;
+       event -= offsetof(struct event, node);
+       return (struct event *)event;
 }
 
-static void export_event_state(struct udev_event *event, enum event_state state)
-{
-       char filename[UTIL_PATH_SIZE];
-       char filename_failed[UTIL_PATH_SIZE];
-       size_t start;
-
-       /* location of queue file */
-       snprintf(filename, sizeof(filename), "%s/.udev/queue/%llu",
-                udev_get_dev_path(event->udev), udev_device_get_seqnum(event->dev));
-
-       /* location of failed file */
-       util_strlcpy(filename_failed, udev_get_dev_path(event->udev), sizeof(filename_failed));
-       util_strlcat(filename_failed, "/", sizeof(filename_failed));
-       start = util_strlcat(filename_failed, ".udev/failed/", sizeof(filename_failed));
-       util_strlcat(filename_failed, udev_device_get_devpath(event->dev), sizeof(filename_failed));
-       util_path_encode(&filename_failed[start], sizeof(filename_failed) - start);
-
-       switch (state) {
-       case EVENT_QUEUED:
-               if(unlink(filename_failed) == 0)
-                       util_delete_path(event->udev, filename_failed);
-               util_create_path(event->udev, filename);
-               udev_selinux_setfscreatecon(event->udev, filename, S_IFLNK);
-               symlink(udev_device_get_devpath(event->dev), filename);
-               udev_selinux_resetfscreatecon(event->udev);
-               break;
-       case EVENT_FINISHED:
-               if (udev_device_get_devpath_old(event->dev) != NULL) {
-                       /* "move" event - rename failed file to current name, do not delete failed */
-                       char filename_failed_old[UTIL_PATH_SIZE];
-
-                       util_strlcpy(filename_failed_old, udev_get_dev_path(event->udev), sizeof(filename_failed_old));
-                       util_strlcat(filename_failed_old, "/", sizeof(filename_failed_old));
-                       start = util_strlcat(filename_failed_old, ".udev/failed/", sizeof(filename_failed_old));
-                       util_strlcat(filename_failed_old, udev_device_get_devpath_old(event->dev), sizeof(filename_failed_old));
-                       util_path_encode(&filename_failed_old[start], sizeof(filename) - start);
-
-                       if (rename(filename_failed_old, filename_failed) == 0)
-                               info(event->udev, "renamed devpath, moved failed state of '%s' to %s'\n",
-                                    udev_device_get_devpath_old(event->dev), udev_device_get_devpath(event->dev));
-               } else {
-                       if (unlink(filename_failed) == 0)
-                               util_delete_path(event->udev, filename_failed);
-               }
+static void event_queue_cleanup(struct udev *udev, enum event_state type);
 
-               unlink(filename);
+enum worker_state {
+       WORKER_UNDEF,
+       WORKER_RUNNING,
+       WORKER_IDLE,
+       WORKER_KILLED,
+};
 
-               /* clean up possibly empty queue directory */
-               if (udev_list_is_empty(&event_list))
-                       util_delete_path(event->udev, filename);
-               break;
-       case EVENT_FAILED:
-               /* move failed event to the failed directory */
-               util_create_path(event->udev, filename_failed);
-               rename(filename, filename_failed);
-
-               /* clean up possibly empty queue directory */
-               if (udev_list_is_empty(&event_list))
-                       util_delete_path(event->udev, filename);
-               break;
-       }
+struct worker {
+       struct udev_list_node node;
+       struct udev *udev;
+       int refcount;
+       pid_t pid;
+       struct udev_monitor *monitor;
+       enum worker_state state;
+       struct event *event;
+};
+
+/* passed from worker to main process */
+struct worker_message {
+       pid_t pid;
+       int exitcode;
+};
 
-       return;
+static struct worker *node_to_worker(struct udev_list_node *node)
+{
+       char *worker;
+
+       worker = (char *)node;
+       worker -= offsetof(struct worker, node);
+       return (struct worker *)worker;
 }
 
-static void event_queue_delete(struct udev_event *event)
+static void event_queue_delete(struct event *event, bool export)
 {
        udev_list_node_remove(&event->node);
 
-       /* mark as failed, if "add" event returns non-zero */
-       if (event->exitstatus && strcmp(udev_device_get_action(event->dev), "add") == 0)
-               export_event_state(event, EVENT_FAILED);
-       else
-               export_event_state(event, EVENT_FINISHED);
-
+       if (export) {
+               /* mark as failed, if "add" event returns non-zero */
+               if (event->exitcode != 0 && strcmp(udev_device_get_action(event->dev), "remove") != 0)
+                       udev_queue_export_device_failed(udev_queue_export, event->dev);
+               else
+                       udev_queue_export_device_finished(udev_queue_export, event->dev);
+               info(event->udev, "seq %llu done with %i\n", udev_device_get_seqnum(event->dev), event->exitcode);
+       }
        udev_device_unref(event->dev);
-       udev_event_unref(event);
+       free(event);
 }
 
-static void asmlinkage event_sig_handler(int signum)
+static struct worker *worker_ref(struct worker *worker)
 {
-       if (signum == SIGALRM)
-               exit(1);
+       worker->refcount++;
+       return worker;
 }
 
-static void event_fork(struct udev_event *event)
+static void worker_cleanup(struct worker *worker)
 {
+       udev_list_node_remove(&worker->node);
+       udev_monitor_unref(worker->monitor);
+       children--;
+       free(worker);
+}
+
+static void worker_unref(struct worker *worker)
+{
+       worker->refcount--;
+       if (worker->refcount > 0)
+               return;
+       info(worker->udev, "worker [%u] cleaned up\n", worker->pid);
+       worker_cleanup(worker);
+}
+
+static void worker_list_cleanup(struct udev *udev)
+{
+       struct udev_list_node *loop, *tmp;
+
+       udev_list_node_foreach_safe(loop, tmp, &worker_list) {
+               struct worker *worker = node_to_worker(loop);
+
+               worker_cleanup(worker);
+       }
+}
+
+static void worker_new(struct event *event)
+{
+       struct udev *udev = event->udev;
+       struct worker *worker;
+       struct udev_monitor *worker_monitor;
        pid_t pid;
-       struct sigaction act;
-       int err;
-
-       if (debug_trace) {
-               event->trace = 1;
-               fprintf(stderr, "fork %s (%llu)\n",
-                      udev_device_get_syspath(event->dev),
-                      udev_device_get_seqnum(event->dev));
+
+       /* listen for new events */
+       worker_monitor = udev_monitor_new_from_netlink(udev, NULL);
+       if (worker_monitor == NULL)
+               return;
+       /* allow the main daemon netlink address to send devices to the worker */
+       udev_monitor_allow_unicast_sender(worker_monitor, monitor);
+       udev_monitor_enable_receiving(worker_monitor);
+
+       worker = calloc(1, sizeof(struct worker));
+       if (worker == NULL) {
+               udev_monitor_unref(worker_monitor);
+               return;
        }
+       /* worker + event reference */
+       worker->refcount = 2;
+       worker->udev = udev;
 
        pid = fork();
        switch (pid) {
-       case 0:
-               /* child */
-               udev_monitor_unref(kernel_monitor);
+       case 0: {
+               struct udev_device *dev = NULL;
+               int fd_monitor;
+               struct epoll_event ep_signal, ep_monitor;
+               sigset_t mask;
+               int rc = EXIT_SUCCESS;
+
+               /* move initial device from queue */
+               dev = event->dev;
+               event->dev = NULL;
+
+               free(worker);
+               worker_list_cleanup(udev);
+               event_queue_cleanup(udev, EVENT_UNDEF);
+               udev_queue_export_unref(udev_queue_export);
+               udev_monitor_unref(monitor);
                udev_ctrl_unref(udev_ctrl);
-               logging_close();
-               logging_init("udevd-event");
-               setpriority(PRIO_PROCESS, 0, UDEV_PRIORITY);
-
-               /* set signal handlers */
-               memset(&act, 0x00, sizeof(act));
-               act.sa_handler = (void (*)(int)) event_sig_handler;
-               sigemptyset (&act.sa_mask);
-               act.sa_flags = 0;
-               sigaction(SIGALRM, &act, NULL);
-
-               /* reset to default */
-               act.sa_handler = SIG_DFL;
-               sigaction(SIGINT, &act, NULL);
-               sigaction(SIGTERM, &act, NULL);
-               sigaction(SIGCHLD, &act, NULL);
-               sigaction(SIGHUP, &act, NULL);
-
-               /* set timeout to prevent hanging processes */
-               alarm(UDEV_EVENT_TIMEOUT);
-
-               /* apply rules, create node, symlinks */
-               err = udev_event_execute_rules(event, rules);
-
-               /* rules may change/disable the timeout */
-               if (udev_device_get_event_timeout(event->dev) >= 0)
-                       alarm(udev_device_get_event_timeout(event->dev));
-
-               /* execute RUN= */
-               if (err == 0 && !event->ignore_device && udev_get_run(event->udev))
-                       udev_event_execute_run(event);
-
-               /* apply/restore inotify watch */
-               if (err == 0 && event->inotify_watch && inotify_fd != -1 &&
-                   major(udev_device_get_devnum(event->dev)) != 0 &&
-                   strcmp(udev_device_get_action(event->dev), "remove") != 0) {
-                       info(event->udev, "device will be watched for changes\n");
-                       udev_watch_begin(event->udev, event->dev);
-               } else {
-                       udev_watch_clear(event->udev, event->dev);
+               close(fd_signal);
+               close(fd_ep);
+               close(worker_watch[READ_END]);
+
+               sigfillset(&mask);
+               fd_signal = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
+               if (fd_signal < 0) {
+                       err(udev, "error creating signalfd %m\n");
+                       rc = 2;
+                       goto out;
                }
 
-               info(event->udev, "seq %llu exit with %i\n", udev_device_get_seqnum(event->dev), err);
-               logging_close();
-               if (err != 0)
-                       exit(1);
-               exit(0);
+               fd_ep = epoll_create1(EPOLL_CLOEXEC);
+               if (fd_ep < 0) {
+                       err(udev, "error creating epoll fd: %m\n");
+                       rc = 3;
+                       goto out;
+               }
+
+               memset(&ep_signal, 0, sizeof(struct epoll_event));
+               ep_signal.events = EPOLLIN;
+               ep_signal.data.fd = fd_signal;
+
+               fd_monitor = udev_monitor_get_fd(worker_monitor);
+               memset(&ep_monitor, 0, sizeof(struct epoll_event));
+               ep_monitor.events = EPOLLIN;
+               ep_monitor.data.fd = fd_monitor;
+
+               if (epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_signal, &ep_signal) < 0 ||
+                   epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_monitor, &ep_monitor) < 0) {
+                       err(udev, "fail to add fds to epoll: %m\n");
+                       rc = 4;
+                       goto out;
+               }
+
+               /* request TERM signal if parent exits */
+               prctl(PR_SET_PDEATHSIG, SIGTERM);
+
+               for (;;) {
+                       struct udev_event *udev_event;
+                       struct worker_message msg;
+                       int failed = 0;
+                       int err;
+
+                       info(udev, "seq %llu running\n", udev_device_get_seqnum(dev));
+                       udev_event = udev_event_new(dev);
+                       if (udev_event == NULL) {
+                               rc = 5;
+                               goto out;
+                       }
+
+                       /* needed for SIGCHLD/SIGTERM in spawn() */
+                       udev_event->fd_signal = fd_signal;
+
+                       if (exec_delay > 0)
+                               udev_event->exec_delay = exec_delay;
+
+                       /* apply rules, create node, symlinks */
+                       err = udev_event_execute_rules(udev_event, rules, &sigmask_orig);
+
+                       if (err == 0)
+                               failed = udev_event_execute_run(udev_event, &sigmask_orig);
+
+                       /* apply/restore inotify watch */
+                       if (err == 0 && udev_event->inotify_watch) {
+                               udev_watch_begin(udev, dev);
+                               udev_device_update_db(dev);
+                       }
+
+                       /* send processed event back to libudev listeners */
+                       udev_monitor_send_device(worker_monitor, NULL, dev);
+
+                       /* send udevd the result of the event execution */
+                       memset(&msg, 0, sizeof(struct worker_message));
+                       if (err != 0)
+                               msg.exitcode = err;
+                       else if (failed != 0)
+                               msg.exitcode = failed;
+                       msg.pid = getpid();
+                       send(worker_watch[WRITE_END], &msg, sizeof(struct worker_message), 0);
+
+                       info(udev, "seq %llu processed with %i\n", udev_device_get_seqnum(dev), err);
+
+                       udev_device_unref(dev);
+                       dev = NULL;
+
+                       if (udev_event->sigterm) {
+                               udev_event_unref(udev_event);
+                               goto out;
+                       }
+
+                       udev_event_unref(udev_event);
+
+                       /* wait for more device messages from main udevd, or term signal */
+                       while (dev == NULL) {
+                               struct epoll_event ev[4];
+                               int fdcount;
+                               int i;
+
+                               fdcount = epoll_wait(fd_ep, ev, ARRAY_SIZE(ev), -1);
+                               if (fdcount < 0) {
+                                       if (errno == EINTR)
+                                               continue;
+                                       err = -errno;
+                                       err(udev, "failed to poll: %m\n");
+                                       goto out;
+                               }
+
+                               for (i = 0; i < fdcount; i++) {
+                                       if (ev[i].data.fd == fd_monitor && ev[i].events & EPOLLIN) {
+                                               dev = udev_monitor_receive_device(worker_monitor);
+                                       } else if (ev[i].data.fd == fd_signal && ev[i].events & EPOLLIN) {
+                                               struct signalfd_siginfo fdsi;
+                                               ssize_t size;
+
+                                               size = read(fd_signal, &fdsi, sizeof(struct signalfd_siginfo));
+                                               if (size != sizeof(struct signalfd_siginfo))
+                                                       continue;
+                                               switch (fdsi.ssi_signo) {
+                                               case SIGTERM:
+                                                       goto out;
+                                               }
+                                       }
+                               }
+                       }
+               }
+out:
+               udev_device_unref(dev);
+               if (fd_signal >= 0)
+                       close(fd_signal);
+               if (fd_ep >= 0)
+                       close(fd_ep);
+               close(fd_inotify);
+               close(worker_watch[WRITE_END]);
+               udev_rules_unref(rules);
+               udev_monitor_unref(worker_monitor);
+               udev_unref(udev);
+               udev_log_close();
+               exit(rc);
+       }
        case -1:
-               err(event->udev, "fork of child failed: %m\n");
-               event_queue_delete(event);
+               udev_monitor_unref(worker_monitor);
+               event->state = EVENT_QUEUED;
+               free(worker);
+               err(udev, "fork of child failed: %m\n");
                break;
        default:
-               /* get SIGCHLD in main loop */
-               info(event->udev, "seq %llu forked, pid [%d], '%s' '%s', %ld seconds old\n",
-                    udev_device_get_seqnum(event->dev),
-                    pid,
-                    udev_device_get_action(event->dev),
-                    udev_device_get_subsystem(event->dev),
-                    time(NULL) - event->queue_time);
-               event->pid = pid;
-               childs++;
+               /* close monitor, but keep address around */
+               udev_monitor_disconnect(worker_monitor);
+               worker->monitor = worker_monitor;
+               worker->pid = pid;
+               worker->state = WORKER_RUNNING;
+               worker->event = event;
+               event->state = EVENT_RUNNING;
+               udev_list_node_append(&worker->node, &worker_list);
+               children++;
+               info(udev, "seq %llu forked new worker [%u]\n", udev_device_get_seqnum(event->dev), pid);
+               break;
        }
 }
 
-static void event_queue_insert(struct udev_event *event)
+static void event_run(struct event *event, bool force)
 {
-       char filename[UTIL_PATH_SIZE];
-       int fd;
-
-       event->queue_time = time(NULL);
+       struct udev_list_node *loop;
 
-       export_event_state(event, EVENT_QUEUED);
-       info(event->udev, "seq %llu queued, '%s' '%s'\n", udev_device_get_seqnum(event->dev),
-            udev_device_get_action(event->dev), udev_device_get_subsystem(event->dev));
+       udev_list_node_foreach(loop, &worker_list) {
+               struct worker *worker = node_to_worker(loop);
+               ssize_t count;
 
-       util_strlcpy(filename, udev_get_dev_path(event->udev), sizeof(filename));
-       util_strlcat(filename, "/.udev/uevent_seqnum", sizeof(filename));
-       fd = open(filename, O_WRONLY|O_TRUNC|O_CREAT, 0644);
-       if (fd >= 0) {
-               char str[32];
-               int len;
+               if (worker->state != WORKER_IDLE)
+                       continue;
 
-               len = sprintf(str, "%llu\n", udev_device_get_seqnum(event->dev));
-               write(fd, str, len);
-               close(fd);
+               count = udev_monitor_send_device(monitor, worker->monitor, event->dev);
+               if (count < 0) {
+                       err(event->udev, "worker [%u] did not accept message %zi (%m), kill it\n", worker->pid, count);
+                       kill(worker->pid, SIGKILL);
+                       worker->state = WORKER_KILLED;
+                       continue;
+               }
+               worker_ref(worker);
+               worker->event = event;
+               worker->state = WORKER_RUNNING;
+               event->state = EVENT_RUNNING;
+               return;
        }
 
-
-       udev_list_node_append(&event->node, &event_list);
-       run_exec_q = 1;
-
-       /* run all events with a timeout set immediately */
-       if (udev_device_get_timeout(event->dev) > 0) {
-               event_fork(event);
+       if (!force && children >= children_max) {
+               if (children_max > 1)
+                       info(event->udev, "maximum number (%i) of children reached\n", children);
                return;
        }
+
+       /* start new worker and pass initial device */
+       worker_new(event);
 }
 
-static int mem_size_mb(void)
+static int event_queue_insert(struct udev_device *dev)
 {
-       FILE* f;
-       char buf[4096];
-       long int memsize = -1;
+       struct event *event;
 
-       f = fopen("/proc/meminfo", "r");
-       if (f == NULL)
+       event = calloc(1, sizeof(struct event));
+       if (event == NULL)
                return -1;
 
-       while (fgets(buf, sizeof(buf), f) != NULL) {
-               long int value;
+       event->udev = udev_device_get_udev(dev);
+       event->dev = dev;
+       event->seqnum = udev_device_get_seqnum(dev);
+       event->devpath = udev_device_get_devpath(dev);
+       event->devpath_len = strlen(event->devpath);
+       event->devpath_old = udev_device_get_devpath_old(dev);
+       event->devnum = udev_device_get_devnum(dev);
+       event->is_block = (strcmp("block", udev_device_get_subsystem(dev)) == 0);
+       event->ifindex = udev_device_get_ifindex(dev);
+
+       udev_queue_export_device_queued(udev_queue_export, dev);
+       info(event->udev, "seq %llu queued, '%s' '%s'\n", udev_device_get_seqnum(dev),
+            udev_device_get_action(dev), udev_device_get_subsystem(dev));
+
+       event->state = EVENT_QUEUED;
+       udev_list_node_append(&event->node, &event_list);
 
-               if (sscanf(buf, "MemTotal: %ld kB", &value) == 1) {
-                       memsize = value / 1024;
-                       break;
-               }
+       /* run all events with a timeout set immediately */
+       if (udev_device_get_timeout(dev) > 0) {
+               event_run(event, true);
+               return 0;
        }
 
-       fclose(f);
-       return memsize;
+       return 0;
 }
 
-static int compare_devpath(const char *running, const char *waiting)
+static void worker_kill(struct udev *udev, int retain)
 {
-       int i = 0;
+       struct udev_list_node *loop;
+       int max;
 
-       while (running[i] != '\0' && running[i] == waiting[i])
-               i++;
+       if (children <= retain)
+               return;
 
-       /* identical device event found */
-       if (running[i] == '\0' && waiting[i] == '\0')
-               return 1;
+       max = children - retain;
 
-       /* parent device event found */
-       if (running[i] == '\0' && waiting[i] == '/')
-               return 2;
+       udev_list_node_foreach(loop, &worker_list) {
+               struct worker *worker = node_to_worker(loop);
 
-       /* child device event found */
-       if (running[i] == '/' && waiting[i] == '\0')
-               return 3;
+               if (max-- <= 0)
+                       break;
 
-       /* no matching event */
-       return 0;
+               if (worker->state == WORKER_KILLED)
+                       continue;
+
+               worker->state = WORKER_KILLED;
+               kill(worker->pid, SIGTERM);
+       }
 }
 
-/* lookup event for identical, parent, child, or physical device */
-static int devpath_busy(struct udev_event *event)
+/* lookup event for identical, parent, child device */
+static bool is_devpath_busy(struct event *event)
 {
        struct udev_list_node *loop;
+       size_t common;
 
-       if (event->delaying_seqnum > 0) {
-       }
        /* check if queue contains events we depend on */
        udev_list_node_foreach(loop, &event_list) {
-               struct udev_event *loop_event = node_to_event(loop);
+               struct event *loop_event = node_to_event(loop);
 
                /* we already found a later event, earlier can not block us, no need to check again */
-               if (udev_device_get_seqnum(loop_event->dev) < event->delaying_seqnum)
+               if (loop_event->seqnum < event->delaying_seqnum)
                        continue;
 
                /* event we checked earlier still exists, no need to check again */
-               if (udev_device_get_seqnum(loop_event->dev) == event->delaying_seqnum)
-                       return 2;
+               if (loop_event->seqnum == event->delaying_seqnum)
+                       return true;
 
                /* found ourself, no later event can block us */
-               if (udev_device_get_seqnum(loop_event->dev) >= udev_device_get_seqnum(event->dev))
+               if (loop_event->seqnum >= event->seqnum)
                        break;
 
+               /* check major/minor */
+               if (major(event->devnum) != 0 && event->devnum == loop_event->devnum && event->is_block == loop_event->is_block)
+                       return true;
+
+               /* check network device ifindex */
+               if (event->ifindex != 0 && event->ifindex == loop_event->ifindex)
+                       return true;
+
                /* check our old name */
-               if (udev_device_get_devpath_old(event->dev) != NULL)
-                       if (strcmp(udev_device_get_devpath(loop_event->dev), udev_device_get_devpath_old(event->dev)) == 0) {
-                               event->delaying_seqnum = udev_device_get_seqnum(loop_event->dev);
-                               return 3;
-                       }
+               if (event->devpath_old != NULL && strcmp(loop_event->devpath, event->devpath_old) == 0) {
+                       event->delaying_seqnum = loop_event->seqnum;
+                       return true;
+               }
 
-               /* check identical, parent, or child device event */
-               if (compare_devpath(udev_device_get_devpath(loop_event->dev), udev_device_get_devpath(event->dev)) != 0) {
-                       dbg(event->udev, "%llu, device event still pending %llu (%s)\n",
-                           udev_device_get_seqnum(event->dev),
-                           udev_device_get_seqnum(loop_event->dev),
-                           udev_device_get_devpath(loop_event->dev));
-                       event->delaying_seqnum = udev_device_get_seqnum(loop_event->dev);
-                       return 4;
+               /* compare devpath */
+               common = MIN(loop_event->devpath_len, event->devpath_len);
+
+               /* one devpath is contained in the other? */
+               if (memcmp(loop_event->devpath, event->devpath, common) != 0)
+                       continue;
+
+               /* identical device event found */
+               if (loop_event->devpath_len == event->devpath_len) {
+                       /* devices names might have changed/swapped in the meantime */
+                       if (major(event->devnum) != 0 && (event->devnum != loop_event->devnum || event->is_block != loop_event->is_block))
+                               continue;
+                       if (event->ifindex != 0 && event->ifindex != loop_event->ifindex)
+                               continue;
+                       event->delaying_seqnum = loop_event->seqnum;
+                       return true;
                }
 
-               /* check for our major:minor number */
-               if (major(udev_device_get_devnum(event->dev)) > 0 &&
-                   udev_device_get_devnum(loop_event->dev) == udev_device_get_devnum(event->dev) &&
-                   strcmp(udev_device_get_subsystem(event->dev), udev_device_get_subsystem(loop_event->dev)) == 0) {
-                       dbg(event->udev, "%llu, device event still pending %llu (%d:%d)\n",
-                           udev_device_get_seqnum(event->dev),
-                           udev_device_get_seqnum(loop_event->dev),
-                           major(udev_device_get_devnum(loop_event->dev)), minor(udev_device_get_devnum(loop_event->dev)));
-                       event->delaying_seqnum = udev_device_get_seqnum(loop_event->dev);
-                       return 5;
+               /* parent device event found */
+               if (event->devpath[common] == '/') {
+                       event->delaying_seqnum = loop_event->seqnum;
+                       return true;
                }
 
-               /* check physical device event (special case of parent) */
-               if (udev_device_get_physdevpath(event->dev) != NULL &&
-                   strcmp(udev_device_get_action(event->dev), "add") == 0)
-                       if (compare_devpath(udev_device_get_devpath(loop_event->dev),
-                                           udev_device_get_physdevpath(event->dev)) != 0) {
-                               dbg(event->udev, "%llu, physical device event still pending %llu (%s)\n",
-                                   udev_device_get_seqnum(event->dev),
-                                   udev_device_get_seqnum(loop_event->dev),
-                                   udev_device_get_devpath(loop_event->dev));
-                               event->delaying_seqnum = udev_device_get_seqnum(loop_event->dev);
-                               return 6;
-                       }
+               /* child device event found */
+               if (loop_event->devpath[common] == '/') {
+                       event->delaying_seqnum = loop_event->seqnum;
+                       return true;
+               }
+
+               /* no matching device */
+               continue;
        }
-       return 0;
+
+       return false;
 }
 
-/* serializes events for the identical and parent and child devices */
-static void event_queue_manager(struct udev *udev)
+static void event_queue_start(struct udev *udev)
 {
        struct udev_list_node *loop;
-       struct udev_list_node *tmp;
 
-start_over:
-       if (udev_list_is_empty(&event_list)) {
-               if (childs > 0) {
-                       err(udev, "event list empty, but childs count is %i", childs);
-                       childs = 0;
+       udev_list_node_foreach(loop, &event_list) {
+               struct event *event = node_to_event(loop);
+
+               if (event->state != EVENT_QUEUED)
+                       continue;
+
+               /* do not start event if parent or child event is still running */
+               if (is_devpath_busy(event)) {
+                       dbg(udev, "delay seq %llu (%s)\n", event->seqnum, event->devpath);
+                       continue;
                }
-               return;
+
+               event_run(event, false);
        }
+}
 
-       udev_list_node_foreach_safe(loop, tmp, &event_list) {
-               struct udev_event *loop_event = node_to_event(loop);
+static void event_queue_cleanup(struct udev *udev, enum event_state match_type)
+{
+       struct udev_list_node *loop, *tmp;
 
-               if (childs >= max_childs) {
-                       info(udev, "maximum number (%i) of childs reached\n", childs);
-                       break;
-               }
+       udev_list_node_foreach_safe(loop, tmp, &event_list) {
+               struct event *event = node_to_event(loop);
 
-               if (loop_event->pid != 0)
+               if (match_type != EVENT_UNDEF && match_type != event->state)
                        continue;
 
-               /* do not start event if parent or child event is still running */
-               if (devpath_busy(loop_event) != 0) {
-                       dbg(udev, "delay seq %llu (%s)\n",
-                           udev_device_get_seqnum(loop_event->dev),
-                           udev_device_get_devpath(loop_event->dev));
-                       continue;
-               }
+               event_queue_delete(event, false);
+       }
+}
+
+static void worker_returned(int fd_worker)
+{
+       for (;;) {
+               struct worker_message msg;
+               ssize_t size;
+               struct udev_list_node *loop;
 
-               event_fork(loop_event);
-               dbg(udev, "moved seq %llu to running list\n", udev_device_get_seqnum(loop_event->dev));
+               size = recv(fd_worker, &msg, sizeof(struct worker_message), MSG_DONTWAIT);
+               if (size != sizeof(struct worker_message))
+                       break;
+
+               /* lookup worker who sent the signal */
+               udev_list_node_foreach(loop, &worker_list) {
+                       struct worker *worker = node_to_worker(loop);
+
+                       if (worker->pid != msg.pid)
+                               continue;
 
-               /* retry if events finished in the meantime */
-               if (sigchilds_waiting) {
-                       sigchilds_waiting = 0;
-                       reap_sigchilds();
-                       goto start_over;
+                       /* worker returned */
+                       worker->event->exitcode = msg.exitcode;
+                       event_queue_delete(worker->event, true);
+                       worker->event = NULL;
+                       if (worker->state != WORKER_KILLED)
+                               worker->state = WORKER_IDLE;
+                       worker_unref(worker);
+                       break;
                }
        }
 }
 
 /* receive the udevd message from userspace */
-static void handle_ctrl_msg(struct udev_ctrl *uctrl)
+static struct udev_ctrl_connection *handle_ctrl_msg(struct udev_ctrl *uctrl)
 {
        struct udev *udev = udev_ctrl_get_udev(uctrl);
-       struct udev_ctrl_msg *ctrl_msg;
+       struct udev_ctrl_connection *ctrl_conn;
+       struct udev_ctrl_msg *ctrl_msg = NULL;
        const char *str;
        int i;
 
-       ctrl_msg = udev_ctrl_receive_msg(uctrl);
+       ctrl_conn = udev_ctrl_get_connection(uctrl);
+       if (ctrl_conn == NULL)
+               goto out;
+
+       ctrl_msg = udev_ctrl_receive_msg(ctrl_conn);
        if (ctrl_msg == NULL)
-               return;
+               goto out;
 
        i = udev_ctrl_get_set_log_level(ctrl_msg);
        if (i >= 0) {
                info(udev, "udevd message (SET_LOG_PRIORITY) received, log_priority=%i\n", i);
                udev_set_log_priority(udev, i);
+               worker_kill(udev, 0);
        }
 
        if (udev_ctrl_get_stop_exec_queue(ctrl_msg) > 0) {
                info(udev, "udevd message (STOP_EXEC_QUEUE) received\n");
-               stop_exec_q = 1;
+               stop_exec_queue = true;
        }
 
        if (udev_ctrl_get_start_exec_queue(ctrl_msg) > 0) {
                info(udev, "udevd message (START_EXEC_QUEUE) received\n");
-               stop_exec_q = 0;
-               event_queue_manager(udev);
+               stop_exec_queue = false;
        }
 
        if (udev_ctrl_get_reload_rules(ctrl_msg) > 0) {
                info(udev, "udevd message (RELOAD_RULES) received\n");
-               reload_config = 1;
+               reload_config = true;
        }
 
        str = udev_ctrl_get_set_env(ctrl_msg);
@@ -510,14 +697,27 @@ static void handle_ctrl_msg(struct udev_ctrl *uctrl)
                        }
                        free(key);
                }
+               worker_kill(udev, 0);
        }
 
-       i = udev_ctrl_get_set_max_childs(ctrl_msg);
+       i = udev_ctrl_get_set_children_max(ctrl_msg);
        if (i >= 0) {
-               info(udev, "udevd message (SET_MAX_CHILDS) received, max_childs=%i\n", i);
-               max_childs = i;
+               info(udev, "udevd message (SET_MAX_CHILDREN) received, children_max=%i\n", i);
+               children_max = i;
        }
+
+       if (udev_ctrl_get_ping(ctrl_msg) > 0)
+               info(udev, "udevd message (SYNC) received\n");
+
+       if (udev_ctrl_get_exit(ctrl_msg) > 0) {
+               info(udev, "udevd message (EXIT) received\n");
+               udev_exit = true;
+               /* keep reference to block the client until we exit */
+               udev_ctrl_connection_ref(ctrl_conn);
+       }
+out:
        udev_ctrl_msg_unref(ctrl_msg);
+       return udev_ctrl_connection_unref(ctrl_conn);
 }
 
 /* read inotify messages */
@@ -527,219 +727,531 @@ static int handle_inotify(struct udev *udev)
        char *buf;
        struct inotify_event *ev;
 
-       if ((ioctl(inotify_fd, FIONREAD, &nbytes) < 0) || (nbytes <= 0))
+       if ((ioctl(fd_inotify, FIONREAD, &nbytes) < 0) || (nbytes <= 0))
                return 0;
 
        buf = malloc(nbytes);
        if (buf == NULL) {
-               err(udev, "error getting buffer for inotify, disable watching\n");
-               close(inotify_fd);
-               inotify_fd = -1;
-               return 0;
+               err(udev, "error getting buffer for inotify\n");
+               return -1;
        }
 
-       read(inotify_fd, buf, nbytes);
+       nbytes = read(fd_inotify, buf, nbytes);
 
-       for (pos = 0, ev = (struct inotify_event *)(buf + pos); pos < nbytes; pos += sizeof(struct inotify_event) + ev->len) {
-               const char *syspath;
+       for (pos = 0; pos < nbytes; pos += sizeof(struct inotify_event) + ev->len) {
+               struct udev_device *dev;
 
-               dbg(udev, "inotify event: %x for %d (%s)\n", ev->mask, ev->wd, ev->len ? ev->name : "*");
+               ev = (struct inotify_event *)(buf + pos);
+               if (ev->len) {
+                       const char *s;
 
-               syspath = udev_watch_lookup(udev, ev->wd);
-               if (syspath != NULL) {
-                       dbg(udev, "inotify event: %x for %s\n", ev->mask, syspath);
+                       info(udev, "inotify event: %x for %s\n", ev->mask, ev->name);
+                       s = strstr(ev->name, ".rules");
+                       if (s == NULL)
+                               continue;
+                       if (strlen(s) != strlen(".rules"))
+                               continue;
+                       reload_config = true;
+                       continue;
+               }
+
+               dev = udev_watch_lookup(udev, ev->wd);
+               if (dev != NULL) {
+                       info(udev, "inotify event: %x for %s\n", ev->mask, udev_device_get_devnode(dev));
                        if (ev->mask & IN_CLOSE_WRITE) {
                                char filename[UTIL_PATH_SIZE];
                                int fd;
 
-                               info(udev, "device %s closed, synthesising 'change'\n", syspath);
-                               util_strlcpy(filename, syspath, sizeof(filename));
-                               util_strlcat(filename, "/uevent", sizeof(filename));
+                               info(udev, "device %s closed, synthesising 'change'\n", udev_device_get_devnode(dev));
+                               util_strscpyl(filename, sizeof(filename), udev_device_get_syspath(dev), "/uevent", NULL);
                                fd = open(filename, O_WRONLY);
-                               if (fd < 0 || write(fd, "change", 6) < 0)
-                                       info(udev, "error writing uevent: %m\n");
-                               close(fd);
+                               if (fd >= 0) {
+                                       if (write(fd, "change", 6) < 0)
+                                               info(udev, "error writing uevent: %m\n");
+                                       close(fd);
+                               }
                        }
                        if (ev->mask & IN_IGNORED)
-                               udev_watch_end(udev, ev->wd);
-               } else {
-                       reload_config = 1;
+                               udev_watch_end(udev, dev);
+
+                       udev_device_unref(dev);
                }
+
        }
 
-       free (buf);
+       free(buf);
        return 0;
 }
 
-static void asmlinkage sig_handler(int signum)
+static void handle_signal(struct udev *udev, int signo)
 {
-       switch (signum) {
-               case SIGINT:
-               case SIGTERM:
-                       udev_exit = 1;
-                       break;
-               case SIGCHLD:
-                       /* set flag, then write to pipe if needed */
-                       sigchilds_waiting = 1;
-                       break;
-               case SIGHUP:
-                       reload_config = 1;
-                       break;
+       switch (signo) {
+       case SIGINT:
+       case SIGTERM:
+               udev_exit = true;
+               break;
+       case SIGCHLD:
+               for (;;) {
+                       pid_t pid;
+                       int status;
+                       struct udev_list_node *loop, *tmp;
+
+                       pid = waitpid(-1, &status, WNOHANG);
+                       if (pid <= 0)
+                               break;
+
+                       udev_list_node_foreach_safe(loop, tmp, &worker_list) {
+                               struct worker *worker = node_to_worker(loop);
+
+                               if (worker->pid != pid)
+                                       continue;
+                               info(udev, "worker [%u] exit\n", pid);
+
+                               if (WIFEXITED(status)) {
+                                       if (WEXITSTATUS(status) != 0)
+                                               err(udev, "worker [%u] exit with return code %i\n", pid, WEXITSTATUS(status));
+                               } else if (WIFSIGNALED(status)) {
+                                       err(udev, "worker [%u] terminated by signal %i (%s)\n",
+                                           pid, WTERMSIG(status), strsignal(WTERMSIG(status)));
+                               } else if (WIFSTOPPED(status)) {
+                                       err(udev, "worker [%u] stopped\n", pid);
+                               } else if (WIFCONTINUED(status)) {
+                                       err(udev, "worker [%u] continued\n", pid);
+                               } else {
+                                       err(udev, "worker [%u] exit with status 0x%04x\n", pid, status);
+                               }
+
+                               if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+                                       if (worker->event != NULL) {
+                                               err(udev, "worker [%u] failed while handling '%s'\n",
+                                                   pid, worker->event->devpath);
+                                               worker->event->exitcode = -32;
+                                               event_queue_delete(worker->event, true);
+                                               /* drop reference taken for state 'running' */
+                                               worker_unref(worker);
+                                       }
+                               }
+                               worker_unref(worker);
+                               break;
+                       }
+               }
+               break;
+       case SIGHUP:
+               reload_config = true;
+               break;
+       }
+}
+
+static void static_dev_create_from_modules(struct udev *udev)
+{
+       struct utsname kernel;
+       char modules[UTIL_PATH_SIZE];
+       char buf[4096];
+       FILE *f;
+
+       uname(&kernel);
+       util_strscpyl(modules, sizeof(modules), "/lib/modules/", kernel.release, "/modules.devname", NULL);
+       f = fopen(modules, "r");
+       if (f == NULL)
+               return;
+
+       while (fgets(buf, sizeof(buf), f) != NULL) {
+               char *s;
+               const char *modname;
+               const char *devname;
+               const char *devno;
+               int maj, min;
+               char type;
+               mode_t mode;
+               char filename[UTIL_PATH_SIZE];
+
+               if (buf[0] == '#')
+                       continue;
+
+               modname = buf;
+               s = strchr(modname, ' ');
+               if (s == NULL)
+                       continue;
+               s[0] = '\0';
+
+               devname = &s[1];
+               s = strchr(devname, ' ');
+               if (s == NULL)
+                       continue;
+               s[0] = '\0';
+
+               devno = &s[1];
+               s = strchr(devno, ' ');
+               if (s == NULL)
+                       s = strchr(devno, '\n');
+               if (s != NULL)
+                       s[0] = '\0';
+               if (sscanf(devno, "%c%u:%u", &type, &maj, &min) != 3)
+                       continue;
+
+               /* set sticky bit, so we do not remove the node on module unload */
+               if (type == 'c')
+                       mode = 01600|S_IFCHR;
+               else if (type == 'b')
+                       mode = 01600|S_IFBLK;
+               else
+                       continue;
+
+               util_strscpyl(filename, sizeof(filename), udev_get_dev_path(udev), "/", devname, NULL);
+               util_create_path_selinux(udev, filename);
+               udev_selinux_setfscreatecon(udev, filename, mode);
+               info(udev, "mknod '%s' %c%u:%u\n", filename, type, maj, min);
+               if (mknod(filename, mode, makedev(maj, min)) < 0 && errno == EEXIST)
+                       utimensat(AT_FDCWD, filename, NULL, 0);
+               udev_selinux_resetfscreatecon(udev);
        }
 
-       signal_received = 1;
+       fclose(f);
 }
 
-static void udev_done(int pid, int exitstatus)
+static int copy_dev_dir(struct udev *udev, DIR *dir_from, DIR *dir_to, int maxdepth)
 {
-       struct udev_list_node *loop;
+       struct dirent *dent;
 
-       /* find event associated with pid and delete it */
-       udev_list_node_foreach(loop, &event_list) {
-               struct udev_event *loop_event = node_to_event(loop);
-
-               if (loop_event->pid == pid) {
-                       info(loop_event->udev, "seq %llu cleanup, pid [%d], status %i, %ld seconds old\n",
-                            udev_device_get_seqnum(loop_event->dev), loop_event->pid,
-                            exitstatus, time(NULL) - loop_event->queue_time);
-                       loop_event->exitstatus = exitstatus;
-                       if (debug_trace)
-                               fprintf(stderr, "exit %s (%llu)\n",
-                                      udev_device_get_syspath(loop_event->dev),
-                                      udev_device_get_seqnum(loop_event->dev));
-                       event_queue_delete(loop_event);
-                       childs--;
-
-                       /* there may be dependent events waiting */
-                       run_exec_q = 1;
-                       return;
+       for (dent = readdir(dir_from); dent != NULL; dent = readdir(dir_from)) {
+               struct stat stats;
+
+               if (dent->d_name[0] == '.')
+                       continue;
+               if (fstatat(dirfd(dir_from), dent->d_name, &stats, AT_SYMLINK_NOFOLLOW) != 0)
+                       continue;
+
+               if (S_ISBLK(stats.st_mode) || S_ISCHR(stats.st_mode)) {
+                       udev_selinux_setfscreateconat(udev, dirfd(dir_to), dent->d_name, stats.st_mode & 0777);
+                       if (mknodat(dirfd(dir_to), dent->d_name, stats.st_mode, stats.st_rdev) == 0) {
+                               fchmodat(dirfd(dir_to), dent->d_name, stats.st_mode & 0777, 0);
+                               fchownat(dirfd(dir_to), dent->d_name, stats.st_uid, stats.st_gid, 0);
+                       } else {
+                               utimensat(dirfd(dir_to), dent->d_name, NULL, 0);
+                       }
+                       udev_selinux_resetfscreatecon(udev);
+               } else if (S_ISLNK(stats.st_mode)) {
+                       char target[UTIL_PATH_SIZE];
+                       ssize_t len;
+
+                       len = readlinkat(dirfd(dir_from), dent->d_name, target, sizeof(target));
+                       if (len <= 0 || len == (ssize_t)sizeof(target))
+                               continue;
+                       target[len] = '\0';
+                       udev_selinux_setfscreateconat(udev, dirfd(dir_to), dent->d_name, S_IFLNK);
+                       if (symlinkat(target, dirfd(dir_to), dent->d_name) < 0 && errno == EEXIST)
+                               utimensat(dirfd(dir_to), dent->d_name, NULL, AT_SYMLINK_NOFOLLOW);
+                       udev_selinux_resetfscreatecon(udev);
+               } else if (S_ISDIR(stats.st_mode)) {
+                       DIR *dir2_from, *dir2_to;
+
+                       if (maxdepth == 0)
+                               continue;
+
+                       udev_selinux_setfscreateconat(udev, dirfd(dir_to), dent->d_name, S_IFDIR|0755);
+                       mkdirat(dirfd(dir_to), dent->d_name, 0755);
+                       udev_selinux_resetfscreatecon(udev);
+
+                       dir2_to = fdopendir(openat(dirfd(dir_to), dent->d_name, O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC));
+                       if (dir2_to == NULL)
+                               continue;
+
+                       dir2_from = fdopendir(openat(dirfd(dir_from), dent->d_name, O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC));
+                       if (dir2_from == NULL) {
+                               closedir(dir2_to);
+                               continue;
+                       }
+
+                       copy_dev_dir(udev, dir2_from, dir2_to, maxdepth-1);
+
+                       closedir(dir2_to);
+                       closedir(dir2_from);
                }
        }
+
+       return 0;
 }
 
-static void reap_sigchilds(void)
+static void static_dev_create_links(struct udev *udev, DIR *dir)
 {
-       pid_t pid;
-       int status;
+       struct stdlinks {
+               const char *link;
+               const char *target;
+       };
+       static const struct stdlinks stdlinks[] = {
+               { "core", "/proc/kcore" },
+               { "fd", "/proc/self/fd" },
+               { "stdin", "/proc/self/fd/0" },
+               { "stdout", "/proc/self/fd/1" },
+               { "stderr", "/proc/self/fd/2" },
+       };
+       unsigned int i;
 
-       while (1) {
-               pid = waitpid(-1, &status, WNOHANG);
-               if (pid <= 0)
-                       break;
-               if (WIFEXITED(status))
-                       status = WEXITSTATUS(status);
-               else if (WIFSIGNALED(status))
-                       status = WTERMSIG(status) + 128;
-               else
-                       status = 0;
-               udev_done(pid, status);
+       for (i = 0; i < ARRAY_SIZE(stdlinks); i++) {
+               struct stat sb;
+
+               if (stat(stdlinks[i].target, &sb) == 0) {
+                       udev_selinux_setfscreateconat(udev, dirfd(dir), stdlinks[i].link, S_IFLNK);
+                       if (symlinkat(stdlinks[i].target, dirfd(dir), stdlinks[i].link) < 0 && errno == EEXIST)
+                               utimensat(dirfd(dir), stdlinks[i].link, NULL, AT_SYMLINK_NOFOLLOW);
+                       udev_selinux_resetfscreatecon(udev);
+               }
        }
 }
 
-static void cleanup_queue_dir(struct udev *udev)
+static void static_dev_create_from_devices(struct udev *udev, DIR *dir)
+{
+       DIR *dir_from;
+
+       dir_from = opendir(LIBEXECDIR "/devices");
+       if (dir_from == NULL)
+               return;
+       copy_dev_dir(udev, dir_from, dir, 8);
+       closedir(dir_from);
+}
+
+static void static_dev_create(struct udev *udev)
 {
-       char dirname[UTIL_PATH_SIZE];
-       char filename[UTIL_PATH_SIZE];
        DIR *dir;
 
-       util_strlcpy(filename, udev_get_dev_path(udev), sizeof(filename));
-       util_strlcat(filename, "/.udev/uevent_seqnum", sizeof(filename));
-       unlink(filename);
+       dir = opendir(udev_get_dev_path(udev));
+       if (dir == NULL)
+               return;
 
-       util_strlcpy(dirname, udev_get_dev_path(udev), sizeof(dirname));
-       util_strlcat(dirname, "/.udev/queue", sizeof(dirname));
-       dir = opendir(dirname);
-       if (dir != NULL) {
-               while (1) {
-                       struct dirent *dent;
+       static_dev_create_links(udev, dir);
+       static_dev_create_from_devices(udev, dir);
 
-                       dent = readdir(dir);
-                       if (dent == NULL || dent->d_name[0] == '\0')
-                               break;
-                       if (dent->d_name[0] == '.')
-                               continue;
-                       util_strlcpy(filename, dirname, sizeof(filename));
-                       util_strlcat(filename, "/", sizeof(filename));
-                       util_strlcat(filename, dent->d_name, sizeof(filename));
-                       unlink(filename);
+       closedir(dir);
+}
+
+static int mem_size_mb(void)
+{
+       FILE *f;
+       char buf[4096];
+       long int memsize = -1;
+
+       f = fopen("/proc/meminfo", "r");
+       if (f == NULL)
+               return -1;
+
+       while (fgets(buf, sizeof(buf), f) != NULL) {
+               long int value;
+
+               if (sscanf(buf, "MemTotal: %ld kB", &value) == 1) {
+                       memsize = value / 1024;
+                       break;
                }
-               closedir(dir);
-               rmdir(dirname);
        }
+
+       fclose(f);
+       return memsize;
 }
 
-static void export_initial_seqnum(struct udev *udev)
+static int convert_db(struct udev *udev)
 {
        char filename[UTIL_PATH_SIZE];
-       int fd;
-       char seqnum[32];
-       ssize_t len = 0;
-
-       util_strlcpy(filename, udev_get_sys_path(udev), sizeof(filename));
-       util_strlcat(filename, "/kernel/uevent_seqnum", sizeof(filename));
-       fd = open(filename, O_RDONLY);
-       if (fd >= 0) {
-               len = read(fd, seqnum, sizeof(seqnum)-1);
-               close(fd);
+       FILE *f;
+       struct udev_enumerate *udev_enumerate;
+       struct udev_list_entry *list_entry;
+
+       /* current database */
+       util_strscpyl(filename, sizeof(filename), udev_get_run_path(udev), "/data", NULL);
+       if (access(filename, F_OK) >= 0)
+               return 0;
+
+       /* make sure we do not get here again */
+       util_create_path(udev, filename);
+       mkdir(filename, 0755);
+
+       /* old database */
+       util_strscpyl(filename, sizeof(filename), udev_get_dev_path(udev), "/.udev/db", NULL);
+       if (access(filename, F_OK) < 0)
+               return 0;
+
+       f = fopen("/dev/kmsg", "w");
+       if (f != NULL) {
+               fprintf(f, "<30>udevd[%u]: converting old udev database\n", getpid());
+               fclose(f);
        }
-       if (len <= 0) {
-               strcpy(seqnum, "0\n");
-               len = 3;
+
+       udev_enumerate = udev_enumerate_new(udev);
+       if (udev_enumerate == NULL)
+               return -1;
+       udev_enumerate_scan_devices(udev_enumerate);
+       udev_list_entry_foreach(list_entry, udev_enumerate_get_list_entry(udev_enumerate)) {
+               struct udev_device *device;
+
+               device = udev_device_new_from_syspath(udev, udev_list_entry_get_name(list_entry));
+               if (device == NULL)
+                       continue;
+
+               /* try to find the old database for devices without a current one */
+               if (udev_device_read_db(device, NULL) < 0) {
+                       bool have_db;
+                       const char *id;
+                       struct stat stats;
+                       char devpath[UTIL_PATH_SIZE];
+                       char from[UTIL_PATH_SIZE];
+
+                       have_db = false;
+
+                       /* find database in old location */
+                       id = udev_device_get_id_filename(device);
+                       util_strscpyl(from, sizeof(from), udev_get_dev_path(udev), "/.udev/db/", id, NULL);
+                       if (lstat(from, &stats) == 0) {
+                               if (!have_db) {
+                                       udev_device_read_db(device, from);
+                                       have_db = true;
+                               }
+                               unlink(from);
+                       }
+
+                       /* find old database with $subsys:$sysname name */
+                       util_strscpyl(from, sizeof(from), udev_get_dev_path(udev),
+                                    "/.udev/db/", udev_device_get_subsystem(device), ":",
+                                    udev_device_get_sysname(device), NULL);
+                       if (lstat(from, &stats) == 0) {
+                               if (!have_db) {
+                                       udev_device_read_db(device, from);
+                                       have_db = true;
+                               }
+                               unlink(from);
+                       }
+
+                       /* find old database with the encoded devpath name */
+                       util_path_encode(udev_device_get_devpath(device), devpath, sizeof(devpath));
+                       util_strscpyl(from, sizeof(from), udev_get_dev_path(udev), "/.udev/db/", devpath, NULL);
+                       if (lstat(from, &stats) == 0) {
+                               if (!have_db) {
+                                       udev_device_read_db(device, from);
+                                       have_db = true;
+                               }
+                               unlink(from);
+                       }
+
+                       /* write out new database */
+                       if (have_db)
+                               udev_device_update_db(device);
+               }
+               udev_device_unref(device);
        }
-       util_strlcpy(filename, udev_get_dev_path(udev), sizeof(filename));
-       util_strlcat(filename, "/.udev/uevent_seqnum", sizeof(filename));
-       util_create_path(udev, filename);
-       fd = open(filename, O_WRONLY|O_TRUNC|O_CREAT, 0644);
-       if (fd >= 0) {
-               write(fd, seqnum, len);
-               close(fd);
+       udev_enumerate_unref(udev_enumerate);
+       return 0;
+}
+
+static int systemd_fds(struct udev *udev, int *rctrl, int *rnetlink)
+{
+       int ctrl = -1, netlink = -1;
+       int fd, n;
+
+       n = sd_listen_fds(true);
+       if (n <= 0)
+               return -1;
+
+       for (fd = SD_LISTEN_FDS_START; fd < n + SD_LISTEN_FDS_START; fd++) {
+               if (sd_is_socket(fd, AF_LOCAL, SOCK_SEQPACKET, -1)) {
+                       if (ctrl >= 0)
+                               return -1;
+                       ctrl = fd;
+                       continue;
+               }
+
+               if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1)) {
+                       if (netlink >= 0)
+                               return -1;
+                       netlink = fd;
+                       continue;
+               }
+
+               return -1;
        }
+
+       if (ctrl < 0 || netlink < 0)
+               return -1;
+
+       info(udev, "ctrl=%i netlink=%i\n", ctrl, netlink);
+       *rctrl = ctrl;
+       *rnetlink = netlink;
+       return 0;
 }
 
 int main(int argc, char *argv[])
 {
        struct udev *udev;
-       int fd;
-       struct sigaction act;
-       const char *value;
-       int daemonize = 0;
+       FILE *f;
+       sigset_t mask;
+       int daemonize = false;
        int resolve_names = 1;
        static const struct option options[] = {
                { "daemon", no_argument, NULL, 'd' },
-               { "debug-trace", no_argument, NULL, 't' },
                { "debug", no_argument, NULL, 'D' },
+               { "children-max", required_argument, NULL, 'c' },
+               { "exec-delay", required_argument, NULL, 'e' },
+               { "resolve-names", required_argument, NULL, 'N' },
                { "help", no_argument, NULL, 'h' },
                { "version", no_argument, NULL, 'V' },
-               { "resolve-names", required_argument, NULL, 'N' },
                {}
        };
+       int fd_ctrl = -1;
+       int fd_netlink = -1;
+       int fd_worker = -1;
+       struct epoll_event ep_ctrl, ep_inotify, ep_signal, ep_netlink, ep_worker;
+       struct udev_ctrl_connection *ctrl_conn = NULL;
        int rc = 1;
 
        udev = udev_new();
        if (udev == NULL)
                goto exit;
 
-       logging_init("udevd");
+       udev_log_init("udevd");
        udev_set_log_fn(udev, log_fn);
        info(udev, "version %s\n", VERSION);
        udev_selinux_init(udev);
 
-       while (1) {
+       /* make sure, that our runtime dir exists and is writable */
+       if (utimensat(AT_FDCWD, udev_get_run_config_path(udev), NULL, 0) < 0) {
+               /* try to create our own subdirectory, do not create parent directories */
+               mkdir(udev_get_run_config_path(udev), 0755);
+
+               if (utimensat(AT_FDCWD, udev_get_run_config_path(udev), NULL, 0) >= 0) {
+                       /* directory seems writable now */
+                       udev_set_run_path(udev, udev_get_run_config_path(udev));
+               } else {
+                       /* fall back to /dev/.udev */
+                       char filename[UTIL_PATH_SIZE];
+
+                       util_strscpyl(filename, sizeof(filename), udev_get_dev_path(udev), "/.udev", NULL);
+                       if (udev_set_run_path(udev, filename) == NULL)
+                               goto exit;
+                       mkdir(udev_get_run_path(udev), 0755);
+                       err(udev, "error: runtime directory '%s' not writable, for now falling back to '%s'",
+                           udev_get_run_config_path(udev), udev_get_run_path(udev));
+               }
+       }
+       /* relabel runtime dir only if it resides below /dev */
+       if (strncmp(udev_get_run_path(udev), udev_get_dev_path(udev), strlen(udev_get_dev_path(udev))) == 0)
+               udev_selinux_lsetfilecon(udev, udev_get_run_path(udev), 0755);
+       info(udev, "runtime dir '%s'\n", udev_get_run_path(udev));
+
+       for (;;) {
                int option;
 
-               option = getopt_long(argc, argv, "dDthV", options, NULL);
+               option = getopt_long(argc, argv, "c:deDtN:hV", options, NULL);
                if (option == -1)
                        break;
 
                switch (option) {
                case 'd':
-                       daemonize = 1;
+                       daemonize = true;
+                       break;
+               case 'c':
+                       children_max = strtoul(optarg, NULL, 0);
                        break;
-               case 't':
-                       debug_trace = 1;
+               case 'e':
+                       exec_delay = strtoul(optarg, NULL, 0);
                        break;
                case 'D':
-                       debug = 1;
+                       debug = true;
                        if (udev_get_log_priority(udev) < LOG_INFO)
                                udev_set_log_priority(udev, LOG_INFO);
                        break;
@@ -757,8 +1269,15 @@ int main(int argc, char *argv[])
                        }
                        break;
                case 'h':
-                       printf("Usage: udevd [--help] [--daemon] [--debug-trace] [--debug] "
-                              "[--resolve-names=early|late|never] [--version]\n");
+                       printf("Usage: udevd OPTIONS\n"
+                              "  --daemon\n"
+                              "  --debug\n"
+                              "  --children-max=<maximum number of workers>\n"
+                              "  --exec-delay=<seconds to wait before executing RUN=>\n"
+                              "  --resolve-names=early|late|never\n" 
+                              "  --version\n"
+                              "  --help\n"
+                              "\n");
                        goto exit;
                case 'V':
                        printf("%s\n", VERSION);
@@ -768,270 +1287,447 @@ int main(int argc, char *argv[])
                }
        }
 
+       /*
+        * read the kernel commandline, in case we need to get into debug mode
+        *   udev.log-priority=<level>              syslog priority
+        *   udev.children-max=<number of workers>  events are fully serialized if set to 1
+        *
+        */
+       f = fopen("/proc/cmdline", "r");
+       if (f != NULL) {
+               char cmdline[4096];
+
+               if (fgets(cmdline, sizeof(cmdline), f) != NULL) {
+                       char *pos;
+
+                       pos = strstr(cmdline, "udev.log-priority=");
+                       if (pos != NULL) {
+                               pos += strlen("udev.log-priority=");
+                               udev_set_log_priority(udev, util_log_priority(pos));
+                       }
+
+                       pos = strstr(cmdline, "udev.children-max=");
+                       if (pos != NULL) {
+                               pos += strlen("udev.children-max=");
+                               children_max = strtoul(pos, NULL, 0);
+                       }
+
+                       pos = strstr(cmdline, "udev.exec-delay=");
+                       if (pos != NULL) {
+                               pos += strlen("udev.exec-delay=");
+                               exec_delay = strtoul(pos, NULL, 0);
+                       }
+               }
+               fclose(f);
+       }
+
        if (getuid() != 0) {
                fprintf(stderr, "root privileges required\n");
                err(udev, "root privileges required\n");
                goto exit;
        }
 
-       /* make sure std{in,out,err} fd's are in a sane state */
-       fd = open("/dev/null", O_RDWR);
-       if (fd < 0) {
-               fprintf(stderr, "cannot open /dev/null\n");
-               err(udev, "cannot open /dev/null\n");
+       /* set umask before creating any file/directory */
+       chdir("/");
+       umask(022);
+
+       /* create standard links, copy static nodes, create nodes from modules */
+       static_dev_create(udev);
+       static_dev_create_from_modules(udev);
+
+       /* before opening new files, make sure std{in,out,err} fds are in a sane state */
+       if (daemonize) {
+               int fd;
+
+               fd = open("/dev/null", O_RDWR);
+               if (fd >= 0) {
+                       if (write(STDOUT_FILENO, 0, 0) < 0)
+                               dup2(fd, STDOUT_FILENO);
+                       if (write(STDERR_FILENO, 0, 0) < 0)
+                               dup2(fd, STDERR_FILENO);
+                       if (fd > STDERR_FILENO)
+                               close(fd);
+               } else {
+                       fprintf(stderr, "cannot open /dev/null\n");
+                       err(udev, "cannot open /dev/null\n");
+               }
        }
-       if (write(STDOUT_FILENO, 0, 0) < 0)
-               dup2(fd, STDOUT_FILENO);
-       if (write(STDERR_FILENO, 0, 0) < 0)
-               dup2(fd, STDERR_FILENO);
-
-       /* init control socket, bind() ensures, that only one udevd instance is running */
-       udev_ctrl = udev_ctrl_new_from_socket(udev, UDEV_CTRL_SOCK_PATH);
-       if (udev_ctrl == NULL) {
-               fprintf(stderr, "error initializing control socket");
-               err(udev, "error initializing udevd socket");
-               rc = 1;
+
+       if (systemd_fds(udev, &fd_ctrl, &fd_netlink) >= 0) {
+               /* get control and netlink socket from from systemd */
+               udev_ctrl = udev_ctrl_new_from_socket_fd(udev, UDEV_CTRL_SOCK_PATH, fd_ctrl);
+               if (udev_ctrl == NULL) {
+                       err(udev, "error taking over udev control socket");
+                       rc = 1;
+                       goto exit;
+               }
+
+               monitor = udev_monitor_new_from_netlink_fd(udev, "kernel", fd_netlink);
+               if (monitor == NULL) {
+                       err(udev, "error taking over netlink socket\n");
+                       rc = 3;
+                       goto exit;
+               }
+       } else {
+               /* open control and netlink socket */
+               udev_ctrl = udev_ctrl_new_from_socket(udev, UDEV_CTRL_SOCK_PATH);
+               if (udev_ctrl == NULL) {
+                       fprintf(stderr, "error initializing udev control socket");
+                       err(udev, "error initializing udev control socket");
+                       rc = 1;
+                       goto exit;
+               }
+               fd_ctrl = udev_ctrl_get_fd(udev_ctrl);
+
+               monitor = udev_monitor_new_from_netlink(udev, "kernel");
+               if (monitor == NULL) {
+                       fprintf(stderr, "error initializing netlink socket\n");
+                       err(udev, "error initializing netlink socket\n");
+                       rc = 3;
+                       goto exit;
+               }
+               fd_netlink = udev_monitor_get_fd(monitor);
+       }
+
+       if (udev_monitor_enable_receiving(monitor) < 0) {
+               fprintf(stderr, "error binding netlink socket\n");
+               err(udev, "error binding netlink socket\n");
+               rc = 3;
                goto exit;
        }
 
        if (udev_ctrl_enable_receiving(udev_ctrl) < 0) {
-               fprintf(stderr, "error binding control socket, seems udevd is already running\n");
-               err(udev, "error binding control socket, seems udevd is already running\n");
+               fprintf(stderr, "error binding udev control socket\n");
+               err(udev, "error binding udev control socket\n");
                rc = 1;
                goto exit;
        }
 
-       kernel_monitor = udev_monitor_new_from_netlink(udev);
-       if (kernel_monitor == NULL || udev_monitor_enable_receiving(kernel_monitor) < 0) {
-               fprintf(stderr, "error initializing netlink socket\n");
-               err(udev, "error initializing netlink socket\n");
-               rc = 3;
-               goto exit;
-       }
-       udev_monitor_set_receive_buffer_size(kernel_monitor, 128*1024*1024);
+       udev_monitor_set_receive_buffer_size(monitor, 128*1024*1024);
 
-       rules = udev_rules_new(udev, resolve_names);
-       if (rules == NULL) {
-               err(udev, "error reading rules\n");
+       /* create queue file before signalling 'ready', to make sure we block 'settle' */
+       udev_queue_export = udev_queue_export_new(udev);
+       if (udev_queue_export == NULL) {
+               err(udev, "error creating queue file\n");
                goto exit;
        }
-       udev_list_init(&event_list);
-       cleanup_queue_dir(udev);
-       export_initial_seqnum(udev);
 
        if (daemonize) {
                pid_t pid;
+               int fd;
 
                pid = fork();
                switch (pid) {
                case 0:
-                       dbg(udev, "daemonized fork running\n");
                        break;
                case -1:
                        err(udev, "fork of daemon failed: %m\n");
                        rc = 4;
                        goto exit;
                default:
-                       dbg(udev, "child [%u] running, parent exits\n", pid);
-                       rc = 0;
-                       goto exit;
+                       rc = EXIT_SUCCESS;
+                       goto exit_keep_queue;
+               }
+
+               setsid();
+
+               fd = open("/proc/self/oom_score_adj", O_RDWR);
+               if (fd < 0) {
+                       /* Fallback to old interface */
+                       fd = open("/proc/self/oom_adj", O_RDWR);
+                       if (fd < 0) {
+                               err(udev, "error disabling OOM: %m\n");
+                       } else {
+                               /* OOM_DISABLE == -17 */
+                               write(fd, "-17", 3);
+                               close(fd);
+                       }
+               } else {
+                       write(fd, "-1000", 5);
+                       close(fd);
                }
+       } else {
+               sd_notify(1, "READY=1");
        }
 
-       /* redirect std{out,err} */
-       if (!debug && !debug_trace) {
-               dup2(fd, STDIN_FILENO);
-               dup2(fd, STDOUT_FILENO);
-               dup2(fd, STDERR_FILENO);
+       f = fopen("/dev/kmsg", "w");
+       if (f != NULL) {
+               fprintf(f, "<30>udevd[%u]: starting version " VERSION "\n", getpid());
+               fclose(f);
        }
-       if (fd > STDERR_FILENO)
-               close(fd);
 
-       /* set scheduling priority for the daemon */
-       setpriority(PRIO_PROCESS, 0, UDEVD_PRIORITY);
+       if (!debug) {
+               int fd;
 
-       chdir("/");
-       umask(022);
-       setsid();
-
-       /* OOM_DISABLE == -17 */
-       fd = open("/proc/self/oom_adj", O_RDWR);
-       if (fd < 0)
-               err(udev, "error disabling OOM: %m\n");
-       else {
-               write(fd, "-17", 3);
-               close(fd);
+               fd = open("/dev/null", O_RDWR);
+               if (fd >= 0) {
+                       dup2(fd, STDIN_FILENO);
+                       dup2(fd, STDOUT_FILENO);
+                       dup2(fd, STDERR_FILENO);
+                       close(fd);
+               }
        }
 
-       fd = open("/dev/kmsg", O_WRONLY);
-       if (fd > 0) {
-               const char *ver_str = "<6>udev: starting version " VERSION "\n";
-               char path[UTIL_PATH_SIZE];
+       fd_inotify = udev_watch_init(udev);
+       if (fd_inotify < 0) {
+               fprintf(stderr, "error initializing inotify\n");
+               err(udev, "error initializing inotify\n");
+               rc = 4;
+               goto exit;
+       }
+
+       if (udev_get_rules_path(udev) != NULL) {
+               inotify_add_watch(fd_inotify, udev_get_rules_path(udev),
+                                 IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
+       } else {
+               char filename[UTIL_PATH_SIZE];
                struct stat statbuf;
 
-               write(fd, ver_str, strlen(ver_str));
-               util_strlcpy(path, udev_get_sys_path(udev), sizeof(path));
-               util_strlcat(path, "/class/mem/null", sizeof(path));
-               if (lstat(path, &statbuf) == 0) {
-                       if (S_ISDIR(statbuf.st_mode)) {
-                               const char *depr_str =
-                                       "<6>udev: deprecated sysfs layout; update the kernel or "
-                                       "disable CONFIG_SYSFS_DEPRECATED; some udev features will "
-                                       "not work correctly\n";
-
-                               write(fd, depr_str, strlen(depr_str));
-                       }
+               inotify_add_watch(fd_inotify, LIBEXECDIR "/rules.d",
+                                 IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
+               inotify_add_watch(fd_inotify, SYSCONFDIR "/udev/rules.d",
+                                 IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
+
+               /* watch dynamic rules directory */
+               util_strscpyl(filename, sizeof(filename), udev_get_run_path(udev), "/rules.d", NULL);
+               if (stat(filename, &statbuf) != 0) {
+                       util_create_path(udev, filename);
+                       mkdir(filename, 0755);
                }
-               close(fd);
+               inotify_add_watch(fd_inotify, filename,
+                                 IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
+       }
+       udev_watch_restore(udev);
+
+       /* block and listen to all signals on signalfd */
+       sigfillset(&mask);
+       sigprocmask(SIG_SETMASK, &mask, &sigmask_orig);
+       fd_signal = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
+       if (fd_signal < 0) {
+               fprintf(stderr, "error creating signalfd\n");
+               err(udev, "error creating signalfd\n");
+               rc = 5;
+               goto exit;
        }
 
-       /* set signal handlers */
-       memset(&act, 0x00, sizeof(struct sigaction));
-       act.sa_handler = (void (*)(int)) sig_handler;
-       sigemptyset(&act.sa_mask);
-       act.sa_flags = SA_RESTART;
-       sigaction(SIGINT, &act, NULL);
-       sigaction(SIGTERM, &act, NULL);
-       sigaction(SIGCHLD, &act, NULL);
-       sigaction(SIGHUP, &act, NULL);
-
-       /* watch rules directory */
-       udev_watch_init(udev);
-       if (inotify_fd >= 0) {
-               if (udev_get_rules_path(udev) != NULL) {
-                       inotify_add_watch(inotify_fd, udev_get_rules_path(udev),
-                                         IN_CREATE | IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
-               } else {
-                       char filename[UTIL_PATH_SIZE];
+       /* unnamed socket from workers to the main daemon */
+       if (socketpair(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0, worker_watch) < 0) {
+               fprintf(stderr, "error creating socketpair\n");
+               err(udev, "error creating socketpair\n");
+               rc = 6;
+               goto exit;
+       }
+       fd_worker = worker_watch[READ_END];
 
-                       inotify_add_watch(inotify_fd, UDEV_PREFIX "/lib/udev/rules.d",
-                                         IN_CREATE | IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
-                       inotify_add_watch(inotify_fd, SYSCONFDIR "/udev/rules.d",
-                                         IN_CREATE | IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
+       rules = udev_rules_new(udev, resolve_names);
+       if (rules == NULL) {
+               err(udev, "error reading rules\n");
+               goto exit;
+       }
 
-                       /* watch dynamic rules directory */
-                       util_strlcpy(filename, udev_get_dev_path(udev), sizeof(filename));
-                       util_strlcat(filename, "/.udev/rules.d", sizeof(filename));
-                       inotify_add_watch(inotify_fd, filename,
-                                         IN_CREATE | IN_DELETE | IN_MOVE | IN_CLOSE_WRITE);
-               }
+       memset(&ep_ctrl, 0, sizeof(struct epoll_event));
+       ep_ctrl.events = EPOLLIN;
+       ep_ctrl.data.fd = fd_ctrl;
+
+       memset(&ep_inotify, 0, sizeof(struct epoll_event));
+       ep_inotify.events = EPOLLIN;
+       ep_inotify.data.fd = fd_inotify;
 
-               udev_watch_restore(udev);
+       memset(&ep_signal, 0, sizeof(struct epoll_event));
+       ep_signal.events = EPOLLIN;
+       ep_signal.data.fd = fd_signal;
+
+       memset(&ep_netlink, 0, sizeof(struct epoll_event));
+       ep_netlink.events = EPOLLIN;
+       ep_netlink.data.fd = fd_netlink;
+
+       memset(&ep_worker, 0, sizeof(struct epoll_event));
+       ep_worker.events = EPOLLIN;
+       ep_worker.data.fd = fd_worker;
+
+       fd_ep = epoll_create1(EPOLL_CLOEXEC);
+       if (fd_ep < 0) {
+               err(udev, "error creating epoll fd: %m\n");
+               goto exit;
+       }
+       if (epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_ctrl, &ep_ctrl) < 0 ||
+           epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_inotify, &ep_inotify) < 0 ||
+           epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_signal, &ep_signal) < 0 ||
+           epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_netlink, &ep_netlink) < 0 ||
+           epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_worker, &ep_worker) < 0) {
+               err(udev, "fail to add fds to epoll: %m\n");
+               goto exit;
        }
 
-       /* in trace mode run one event after the other */
-       if (debug_trace) {
-               max_childs = 1;
-       } else {
+       /* if needed, convert old database from earlier udev version */
+       convert_db(udev);
+
+       if (children_max <= 0) {
                int memsize = mem_size_mb();
+
+               /* set value depending on the amount of RAM */
                if (memsize > 0)
-                       max_childs = 128 + (memsize / 4);
+                       children_max = 128 + (memsize / 8);
                else
-                       max_childs = UDEVD_MAX_CHILDS;
+                       children_max = 128;
        }
-       /* possibly overwrite maximum limit of executed events */
-       value = getenv("UDEVD_MAX_CHILDS");
-       if (value)
-               max_childs = strtoul(value, NULL, 10);
-       info(udev, "initialize max_childs to %u\n", max_childs);
-
-       while (!udev_exit) {
-               sigset_t blocked_mask, orig_mask;
-               struct pollfd pfd[4];
-               struct pollfd *ctrl_poll, *monitor_poll, *inotify_poll = NULL;
-               int nfds = 0;
+       info(udev, "set children_max to %u\n", children_max);
+
+       udev_rules_apply_static_dev_perms(rules);
+
+       udev_list_node_init(&event_list);
+       udev_list_node_init(&worker_list);
+
+       for (;;) {
+               struct epoll_event ev[8];
                int fdcount;
+               int timeout;
+               bool is_worker, is_signal, is_inotify, is_netlink, is_ctrl;
+               int i;
+
+               if (udev_exit) {
+                       /* close sources of new events and discard buffered events */
+                       if (fd_ctrl >= 0) {
+                               epoll_ctl(fd_ep, EPOLL_CTL_DEL, fd_ctrl, NULL);
+                               fd_ctrl = -1;
+                       }
+                       if (monitor != NULL) {
+                               epoll_ctl(fd_ep, EPOLL_CTL_DEL, fd_netlink, NULL);
+                               udev_monitor_unref(monitor);
+                               monitor = NULL;
+                       }
+                       if (fd_inotify >= 0) {
+                               epoll_ctl(fd_ep, EPOLL_CTL_DEL, fd_inotify, NULL);
+                               close(fd_inotify);
+                               fd_inotify = -1;
+                       }
 
-               sigfillset(&blocked_mask);
-               sigprocmask(SIG_SETMASK, &blocked_mask, &orig_mask);
-               if (signal_received) {
-                       sigprocmask(SIG_SETMASK, &orig_mask, NULL);
-                       goto handle_signals;
-               }
+                       /* discard queued events and kill workers */
+                       event_queue_cleanup(udev, EVENT_QUEUED);
+                       worker_kill(udev, 0);
 
-               ctrl_poll = &pfd[nfds++];
-               ctrl_poll->fd = udev_ctrl_get_fd(udev_ctrl);
-               ctrl_poll->events = POLLIN;
+                       /* exit after all has cleaned up */
+                       if (udev_list_node_is_empty(&event_list) && udev_list_node_is_empty(&worker_list))
+                               break;
 
-               monitor_poll = &pfd[nfds++];
-               monitor_poll->fd = udev_monitor_get_fd(kernel_monitor);
-               monitor_poll->events = POLLIN;
+                       /* timeout at exit for workers to finish */
+                       timeout = 60 * 1000;
+               } else if (udev_list_node_is_empty(&event_list) && children > 2) {
+                       /* set timeout to kill idle workers */
+                       timeout = 3 * 1000;
+               } else {
+                       timeout = -1;
+               }
+               fdcount = epoll_wait(fd_ep, ev, ARRAY_SIZE(ev), timeout);
+               if (fdcount < 0)
+                       continue;
 
-               if (inotify_fd >= 0) {
-                       inotify_poll = &pfd[nfds++];
-                       inotify_poll->fd = inotify_fd;
-                       inotify_poll->events = POLLIN;
+               if (fdcount == 0) {
+                       if (udev_exit) {
+                               info(udev, "timeout, giving up waiting for workers to finish\n");
+                               break;
+                       }
+
+                       /* timeout - kill idle workers */
+                       worker_kill(udev, 2);
                }
 
-               fdcount = ppoll(pfd, nfds, NULL, &orig_mask);
-               sigprocmask(SIG_SETMASK, &orig_mask, NULL);
-               if (fdcount < 0) {
-                       if (errno == EINTR)
-                               goto handle_signals;
-                       err(udev, "error in select: %m\n");
-                       continue;
+               is_worker = is_signal = is_inotify = is_netlink = is_ctrl = false;
+               for (i = 0; i < fdcount; i++) {
+                       if (ev[i].data.fd == fd_worker && ev[i].events & EPOLLIN)
+                               is_worker = true;
+                       else if (ev[i].data.fd == fd_netlink && ev[i].events & EPOLLIN)
+                               is_netlink = true;
+                       else if (ev[i].data.fd == fd_signal && ev[i].events & EPOLLIN)
+                               is_signal = true;
+                       else if (ev[i].data.fd == fd_inotify && ev[i].events & EPOLLIN)
+                               is_inotify = true;
+                       else if (ev[i].data.fd == fd_ctrl && ev[i].events & EPOLLIN)
+                               is_ctrl = true;
                }
 
-               /* get control message */
-               if (ctrl_poll->revents & POLLIN)
-                       handle_ctrl_msg(udev_ctrl);
+               /* event has finished */
+               if (is_worker)
+                       worker_returned(fd_worker);
 
-               /* get kernel uevent */
-               if (monitor_poll->revents & POLLIN) {
+               if (is_netlink) {
                        struct udev_device *dev;
 
-                       dev = udev_monitor_receive_device(kernel_monitor);
-                       if (dev != NULL) {
-                               struct udev_event *event;
-
-                               event = udev_event_new(dev);
-                               if (event != NULL)
-                                       event_queue_insert(event);
-                               else
+                       dev = udev_monitor_receive_device(monitor);
+                       if (dev != NULL)
+                               udev_device_set_usec_initialized(dev, now_usec());
+                               if (event_queue_insert(dev) < 0)
                                        udev_device_unref(dev);
-                       }
                }
 
-               /* rules directory inotify watch */
-               if (inotify_poll && (inotify_poll->revents & POLLIN))
+               /* start new events */
+               if (!udev_list_node_is_empty(&event_list) && !udev_exit && !stop_exec_queue)
+                       event_queue_start(udev);
+
+               if (is_signal) {
+                       struct signalfd_siginfo fdsi;
+                       ssize_t size;
+
+                       size = read(fd_signal, &fdsi, sizeof(struct signalfd_siginfo));
+                       if (size == sizeof(struct signalfd_siginfo))
+                               handle_signal(udev, fdsi.ssi_signo);
+               }
+
+               /* we are shutting down, the events below are not handled anymore */
+               if (udev_exit)
+                       continue;
+
+               /* device node and rules directory inotify watch */
+               if (is_inotify)
                        handle_inotify(udev);
 
-handle_signals:
-               signal_received = 0;
+               /*
+                * This needs to be after the inotify handling, to make sure,
+                * that the ping is send back after the possibly generated
+                * "change" events by the inotify device node watch.
+                *
+                * A single time we may receive a client connection which we need to
+                * keep open to block the client. It will be closed right before we
+                * exit.
+                */
+               if (is_ctrl)
+                       ctrl_conn = handle_ctrl_msg(udev_ctrl);
 
                /* rules changed, set by inotify or a HUP signal */
                if (reload_config) {
                        struct udev_rules *rules_new;
 
-                       reload_config = 0;
+                       worker_kill(udev, 0);
                        rules_new = udev_rules_new(udev, resolve_names);
                        if (rules_new != NULL) {
                                udev_rules_unref(rules);
                                rules = rules_new;
                        }
-               }
-
-               if (sigchilds_waiting) {
-                       sigchilds_waiting = 0;
-                       reap_sigchilds();
-               }
-
-               if (run_exec_q) {
-                       run_exec_q = 0;
-                       if (!stop_exec_q)
-                               event_queue_manager(udev);
+                       reload_config = 0;
                }
        }
-       cleanup_queue_dir(udev);
-       rc = 0;
+
+       rc = EXIT_SUCCESS;
 exit:
+       udev_queue_export_cleanup(udev_queue_export);
+exit_keep_queue:
+       if (fd_ep >= 0)
+               close(fd_ep);
+       worker_list_cleanup(udev);
+       event_queue_cleanup(udev, EVENT_UNDEF);
        udev_rules_unref(rules);
+       if (fd_signal >= 0)
+               close(fd_signal);
+       if (worker_watch[READ_END] >= 0)
+               close(worker_watch[READ_END]);
+       if (worker_watch[WRITE_END] >= 0)
+               close(worker_watch[WRITE_END]);
+       udev_monitor_unref(monitor);
+       udev_queue_export_unref(udev_queue_export);
+       udev_ctrl_connection_unref(ctrl_conn);
        udev_ctrl_unref(udev_ctrl);
-       if (inotify_fd >= 0)
-               close(inotify_fd);
-       udev_monitor_unref(kernel_monitor);
        udev_selinux_exit(udev);
        udev_unref(udev);
-       logging_close();
+       udev_log_close();
        return rc;
 }