/*
- * Copyright (C) 2004-2011 Kay Sievers <kay.sievers@vrfy.org>
+ * Copyright (C) 2004-2012 Kay Sievers <kay@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>
#include <time.h>
#include <getopt.h>
#include <dirent.h>
+#include <sys/file.h>
#include <sys/time.h>
#include <sys/prctl.h>
#include <sys/socket.h>
#include "udev.h"
#include "sd-daemon.h"
+#include "cgroup-util.h"
+#include "dev-setup.h"
+#include "fileio.h"
static bool debug;
const char *file, int line, const char *fn,
const char *format, va_list args)
{
- if (debug) {
- 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);
- }
+ log_metav(priority, file, line, fn, format, args);
}
static struct udev_rules *rules;
-static struct udev_queue_export *udev_queue_export;
static struct udev_ctrl *udev_ctrl;
static struct udev_monitor *monitor;
static int worker_watch[2] = { -1, -1 };
static sigset_t sigmask_orig;
static UDEV_LIST(event_list);
static UDEV_LIST(worker_list);
+static char *udev_cgroup;
static bool udev_exit;
enum event_state {
size_t devpath_len;
const char *devpath_old;
dev_t devnum;
- bool is_block;
int ifindex;
+ bool is_block;
+#ifdef HAVE_FIRMWARE
+ bool nodelay;
+#endif
};
-static struct event *node_to_event(struct udev_list_node *node)
+static inline struct event *node_to_event(struct udev_list_node *node)
{
- char *event;
-
- event = (char *)node;
- event -= offsetof(struct event, node);
- return (struct event *)event;
+ return container_of(node, struct event, node);
}
static void event_queue_cleanup(struct udev *udev, enum event_state type);
struct udev_monitor *monitor;
enum worker_state state;
struct event *event;
- unsigned long long event_start_usec;
+ usec_t event_start_usec;
};
/* passed from worker to main process */
int exitcode;
};
-static struct worker *node_to_worker(struct udev_list_node *node)
+static inline struct worker *node_to_worker(struct udev_list_node *node)
{
- char *worker;
-
- worker = (char *)node;
- worker -= offsetof(struct worker, node);
- return (struct worker *)worker;
+ return container_of(node, struct worker, node);
}
-static void event_queue_delete(struct event *event, bool export)
+static void event_queue_delete(struct event *event)
{
udev_list_node_remove(&event->node);
-
- if (export) {
- 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);
free(event);
}
worker->refcount--;
if (worker->refcount > 0)
return;
- info(worker->udev, "worker [%u] cleaned up\n", worker->pid);
+ log_debug("worker [%u] cleaned up", worker->pid);
worker_cleanup(worker);
}
udev_monitor_allow_unicast_sender(worker_monitor, monitor);
udev_monitor_enable_receiving(worker_monitor);
- worker = calloc(1, sizeof(struct worker));
+ worker = new0(struct worker, 1);
if (worker == NULL) {
udev_monitor_unref(worker_monitor);
return;
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);
close(fd_signal);
sigfillset(&mask);
fd_signal = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
if (fd_signal < 0) {
- err(udev, "error creating signalfd %m\n");
+ log_error("error creating signalfd %m");
rc = 2;
goto out;
}
fd_ep = epoll_create1(EPOLL_CLOEXEC);
if (fd_ep < 0) {
- err(udev, "error creating epoll fd: %m\n");
+ log_error("error creating epoll fd: %m");
rc = 3;
goto out;
}
- memset(&ep_signal, 0, sizeof(struct epoll_event));
+ memzero(&ep_signal, 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));
+ memzero(&ep_monitor, 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");
+ log_error("fail to add fds to epoll: %m");
rc = 4;
goto out;
}
/* request TERM signal if parent exits */
prctl(PR_SET_PDEATHSIG, SIGTERM);
+ /* reset OOM score, we only protect the main daemon */
+ write_string_file("/proc/self/oom_score_adj", "0");
+
for (;;) {
struct udev_event *udev_event;
struct worker_message msg;
- int err;
+ int fd_lock = -1;
+ int err = 0;
- info(udev, "seq %llu running\n", udev_device_get_seqnum(dev));
+ log_debug("seq %llu running", udev_device_get_seqnum(dev));
udev_event = udev_event_new(dev);
if (udev_event == NULL) {
rc = 5;
if (exec_delay > 0)
udev_event->exec_delay = exec_delay;
+ /*
+ * Take a "read lock" on the device node; this establishes
+ * a concept of device "ownership" to serialize device
+ * access. External processes holding a "write lock" will
+ * cause udev to skip the event handling; in the case udev
+ * acquired the lock, the external process will block until
+ * udev has finished its event handling.
+ */
+
+ /*
+ * <kabi_> since we make check - device seems unused - we try
+ * ioctl to deactivate - and device is found to be opened
+ * <kay> sure, you try to take a write lock
+ * <kay> if you get it udev is out
+ * <kay> if you can't get it, udev is busy
+ * <kabi_> we cannot deactivate openned device (as it is in-use)
+ * <kay> maybe we should just exclude dm from that thing entirely
+ * <kabi_> IMHO this sounds like a good plan for this moment
+ */
+ if (streq_ptr("block", udev_device_get_subsystem(dev)) &&
+ !startswith("dm-", udev_device_get_sysname(dev))) {
+ struct udev_device *d = dev;
+
+ if (streq_ptr("partition", udev_device_get_devtype(d)))
+ d = udev_device_get_parent(d);
+
+ if (d) {
+ fd_lock = open(udev_device_get_devnode(d), O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK);
+ if (fd_lock >= 0 && flock(fd_lock, LOCK_SH|LOCK_NB) < 0) {
+ log_debug("Unable to flock(%s), skipping event handling: %m", udev_device_get_devnode(d));
+ err = -EWOULDBLOCK;
+ fd_lock = safe_close(fd_lock);
+ goto skip;
+ }
+ }
+ }
+
/* apply rules, create node, symlinks */
- err = udev_event_execute_rules(udev_event, rules, &sigmask_orig);
+ udev_event_execute_rules(udev_event, rules, &sigmask_orig);
- if (err == 0)
- udev_event_execute_run(udev_event, &sigmask_orig);
+ udev_event_execute_run(udev_event, &sigmask_orig);
/* apply/restore inotify watch */
- if (err == 0 && udev_event->inotify_watch) {
+ if (udev_event->inotify_watch) {
udev_watch_begin(udev, dev);
udev_device_update_db(dev);
}
+ safe_close(fd_lock);
+
/* send processed event back to libudev listeners */
udev_monitor_send_device(worker_monitor, NULL, dev);
+skip:
/* send udevd the result of the event execution */
- memset(&msg, 0, sizeof(struct worker_message));
- if (err != 0)
- msg.exitcode = err;
+ memzero(&msg, sizeof(struct worker_message));
+ msg.exitcode = err;
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);
+ log_debug("seq %llu processed with %i", udev_device_get_seqnum(dev), err);
udev_device_unref(dev);
dev = NULL;
int fdcount;
int i;
- fdcount = epoll_wait(fd_ep, ev, ARRAY_SIZE(ev), -1);
+ fdcount = epoll_wait(fd_ep, ev, ELEMENTSOF(ev), -1);
if (fdcount < 0) {
if (errno == EINTR)
continue;
- err = -errno;
- err(udev, "failed to poll: %m\n");
+ log_error("failed to poll: %m");
goto out;
}
}
out:
udev_device_unref(dev);
- if (fd_signal >= 0)
- close(fd_signal);
- if (fd_ep >= 0)
- close(fd_ep);
+ safe_close(fd_signal);
+ safe_close(fd_ep);
close(fd_inotify);
close(worker_watch[WRITE_END]);
udev_rules_unref(rules);
udev_builtin_exit(udev);
udev_monitor_unref(worker_monitor);
udev_unref(udev);
- udev_log_close();
+ log_close();
exit(rc);
}
case -1:
udev_monitor_unref(worker_monitor);
event->state = EVENT_QUEUED;
free(worker);
- err(udev, "fork of child failed: %m\n");
+ log_error("fork of child failed: %m");
break;
default:
/* close monitor, but keep address around */
worker->monitor = worker_monitor;
worker->pid = pid;
worker->state = WORKER_RUNNING;
- worker->event_start_usec = now_usec();
+ worker->event_start_usec = now(CLOCK_MONOTONIC);
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);
+ log_debug("seq %llu forked new worker [%u]", udev_device_get_seqnum(event->dev), pid);
break;
}
}
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);
+ log_error("worker [%u] did not accept message %zi (%m), kill it", worker->pid, count);
kill(worker->pid, SIGKILL);
worker->state = WORKER_KILLED;
continue;
worker_ref(worker);
worker->event = event;
worker->state = WORKER_RUNNING;
- worker->event_start_usec = now_usec();
+ worker->event_start_usec = now(CLOCK_MONOTONIC);
event->state = EVENT_RUNNING;
return;
}
if (children >= children_max) {
if (children_max > 1)
- info(event->udev, "maximum number (%i) of children reached\n", children);
+ log_debug("maximum number (%i) of children reached", children);
return;
}
{
struct event *event;
- event = calloc(1, sizeof(struct event));
+ event = new0(struct event, 1);
if (event == NULL)
return -1;
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->is_block = streq("block", udev_device_get_subsystem(dev));
event->ifindex = udev_device_get_ifindex(dev);
+#ifdef HAVE_FIRMWARE
+ if (streq(udev_device_get_subsystem(dev), "firmware"))
+ event->nodelay = true;
+#endif
- udev_queue_export_device_queued(udev_queue_export, dev);
- info(event->udev, "seq %llu queued, '%s' '%s'\n", udev_device_get_seqnum(dev),
+ log_debug("seq %llu queued, '%s' '%s'", udev_device_get_seqnum(dev),
udev_device_get_action(dev), udev_device_get_subsystem(dev));
event->state = EVENT_QUEUED;
return 0;
}
-static void worker_kill(struct udev *udev, int retain)
+static void worker_kill(struct udev *udev)
{
struct udev_list_node *loop;
- int max;
-
- if (children <= retain)
- return;
-
- max = children - retain;
udev_list_node_foreach(loop, &worker_list) {
struct worker *worker = node_to_worker(loop);
- if (max-- <= 0)
- break;
-
if (worker->state == WORKER_KILLED)
continue;
return true;
/* check our old name */
- if (event->devpath_old != NULL && strcmp(loop_event->devpath, event->devpath_old) == 0) {
+ if (event->devpath_old != NULL && streq(loop_event->devpath, event->devpath_old)) {
event->delaying_seqnum = loop_event->seqnum;
return true;
}
return true;
}
+#ifdef HAVE_FIRMWARE
+ /* allow to bypass the dependency tracking */
+ if (event->nodelay)
+ continue;
+#endif
+
/* parent device event found */
if (event->devpath[common] == '/') {
event->delaying_seqnum = loop_event->seqnum;
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);
+ if (is_devpath_busy(event))
continue;
- }
event_run(event);
}
if (match_type != EVENT_UNDEF && match_type != event->state)
continue;
- event_queue_delete(event, false);
+ event_queue_delete(event);
}
}
/* worker returned */
if (worker->event) {
worker->event->exitcode = msg.exitcode;
- event_queue_delete(worker->event, true);
+ event_queue_delete(worker->event);
worker->event = NULL;
}
if (worker->state != WORKER_KILLED)
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);
+ log_debug("udevd message (SET_LOG_LEVEL) received, log_priority=%i", i);
+ log_set_max_level(i);
udev_set_log_priority(udev, i);
- worker_kill(udev, 0);
+ worker_kill(udev);
}
if (udev_ctrl_get_stop_exec_queue(ctrl_msg) > 0) {
- info(udev, "udevd message (STOP_EXEC_QUEUE) received\n");
+ log_debug("udevd message (STOP_EXEC_QUEUE) received");
stop_exec_queue = true;
}
if (udev_ctrl_get_start_exec_queue(ctrl_msg) > 0) {
- info(udev, "udevd message (START_EXEC_QUEUE) received\n");
+ log_debug("udevd message (START_EXEC_QUEUE) received");
stop_exec_queue = false;
}
if (udev_ctrl_get_reload(ctrl_msg) > 0) {
- info(udev, "udevd message (RELOAD) received\n");
+ log_debug("udevd message (RELOAD) received");
reload = true;
}
val[0] = '\0';
val = &val[1];
if (val[0] == '\0') {
- info(udev, "udevd message (ENV) received, unset '%s'\n", key);
+ log_debug("udevd message (ENV) received, unset '%s'", key);
udev_add_property(udev, key, NULL);
} else {
- info(udev, "udevd message (ENV) received, set '%s=%s'\n", key, val);
+ log_debug("udevd message (ENV) received, set '%s=%s'", key, val);
udev_add_property(udev, key, val);
}
} else {
- err(udev, "wrong key format '%s'\n", key);
+ log_error("wrong key format '%s'", key);
}
free(key);
}
- worker_kill(udev, 0);
+ worker_kill(udev);
}
i = udev_ctrl_get_set_children_max(ctrl_msg);
if (i >= 0) {
- info(udev, "udevd message (SET_MAX_CHILDREN) received, children_max=%i\n", i);
+ log_debug("udevd message (SET_MAX_CHILDREN) received, children_max=%i", i);
children_max = i;
}
if (udev_ctrl_get_ping(ctrl_msg) > 0)
- info(udev, "udevd message (SYNC) received\n");
+ log_debug("udevd message (SYNC) received");
if (udev_ctrl_get_exit(ctrl_msg) > 0) {
- info(udev, "udevd message (EXIT) received\n");
+ log_debug("udevd message (EXIT) received");
udev_exit = true;
/* keep reference to block the client until we exit */
udev_ctrl_connection_ref(ctrl_conn);
buf = malloc(nbytes);
if (buf == NULL) {
- err(udev, "error getting buffer for inotify\n");
+ log_error("error getting buffer for inotify");
return -1;
}
ev = (struct inotify_event *)(buf + pos);
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));
+ log_debug("inotify event: %x for %s", 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", udev_device_get_devnode(dev));
- util_strscpyl(filename, sizeof(filename), udev_device_get_syspath(dev), "/uevent", NULL);
- fd = open(filename, O_WRONLY);
+ log_debug("device %s closed, synthesising 'change'", udev_device_get_devnode(dev));
+ strscpyl(filename, sizeof(filename), udev_device_get_syspath(dev), "/uevent", NULL);
+ fd = open(filename, O_WRONLY|O_CLOEXEC);
if (fd >= 0) {
if (write(fd, "change", 6) < 0)
- info(udev, "error writing uevent: %m\n");
+ log_debug("error writing uevent: %m");
close(fd);
}
}
if (worker->pid != pid)
continue;
- info(udev, "worker [%u] exit\n", pid);
+ log_debug("worker [%u] exit", pid);
if (WIFEXITED(status)) {
if (WEXITSTATUS(status) != 0)
- err(udev, "worker [%u] exit with return code %i\n", pid, WEXITSTATUS(status));
+ log_error("worker [%u] exit with return code %i",
+ pid, WEXITSTATUS(status));
} else if (WIFSIGNALED(status)) {
- err(udev, "worker [%u] terminated by signal %i (%s)\n",
- pid, WTERMSIG(status), strsignal(WTERMSIG(status)));
+ log_error("worker [%u] terminated by signal %i (%s)",
+ pid, WTERMSIG(status), strsignal(WTERMSIG(status)));
} else if (WIFSTOPPED(status)) {
- err(udev, "worker [%u] stopped\n", pid);
+ log_error("worker [%u] stopped", pid);
} else if (WIFCONTINUED(status)) {
- err(udev, "worker [%u] continued\n", pid);
+ log_error("worker [%u] continued", pid);
} else {
- err(udev, "worker [%u] exit with status 0x%04x\n", pid, status);
+ log_error("worker [%u] exit with status 0x%04x", pid, status);
}
if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
if (worker->event) {
- err(udev, "worker [%u] failed while handling '%s'\n",
- pid, worker->event->devpath);
+ log_error("worker [%u] failed while handling '%s'",
+ pid, worker->event->devpath);
worker->event->exitcode = -32;
- event_queue_delete(worker->event, true);
+ event_queue_delete(worker->event);
+
/* drop reference taken for state 'running' */
worker_unref(worker);
}
}
}
-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;
-
- if (type == 'c')
- mode = S_IFCHR;
- else if (type == 'b')
- mode = 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);
- }
-
- fclose(f);
-}
-
-static int copy_dev_dir(struct udev *udev, DIR *dir_from, DIR *dir_to, int maxdepth)
-{
- struct dirent *dent;
-
- 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 static_dev_create_links(struct udev *udev, DIR *dir)
-{
- 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;
-
- 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 static_dev_create_from_devices(struct udev *udev, DIR *dir)
-{
- DIR *dir_from;
-
- dir_from = opendir(UDEVLIBEXECDIR "/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)
-{
- DIR *dir;
-
- dir = opendir(udev_get_dev_path(udev));
- if (dir == NULL)
- return;
-
- static_dev_create_links(udev, dir);
- static_dev_create_from_devices(udev, dir);
-
- 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;
- }
- }
-
- fclose(f);
- return memsize;
-}
-
-static int convert_db(struct udev *udev)
-{
- char filename[UTIL_PATH_SIZE];
- 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);
- }
-
- 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);
- }
- udev_enumerate_unref(udev_enumerate);
- return 0;
-}
-
static int systemd_fds(struct udev *udev, int *rctrl, int *rnetlink)
{
int ctrl = -1, netlink = -1;
if (ctrl < 0 || netlink < 0)
return -1;
- info(udev, "ctrl=%i netlink=%i\n", ctrl, netlink);
+ log_debug("ctrl=%i netlink=%i", ctrl, netlink);
*rctrl = ctrl;
*rnetlink = netlink;
return 0;
}
-static bool check_rules_timestamp(struct udev *udev)
+/*
+ * 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
+ * udev.exec-delay=<number of seconds> delay execution of every executed program
+ */
+static void kernel_cmdline_options(struct udev *udev)
{
- char **p;
- unsigned long long *stamp_usec;
- int i, n;
- bool changed = false;
-
- n = udev_get_rules_path(udev, &p, &stamp_usec);
- for (i = 0; i < n; i++) {
- struct stat stats;
+ _cleanup_free_ char *line = NULL;
+ char *w, *state;
+ size_t l;
+ int r;
+
+ r = proc_cmdline(&line);
+ if (r < 0)
+ log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));
+ if (r <= 0)
+ return;
- if (stat(p[i], &stats) < 0)
- continue;
+ FOREACH_WORD_QUOTED(w, l, line, state) {
+ char *s, *opt;
- if (stamp_usec[i] == ts_usec(&stats.st_mtim))
- continue;
+ s = strndup(w, l);
+ if (!s)
+ break;
- /* first check */
- if (stamp_usec[i] != 0) {
- info(udev, "reload - timestamp of '%s' changed\n", p[i]);
- changed = true;
+ /* accept the same options for the initrd, prefixed with "rd." */
+ if (in_initrd() && startswith(s, "rd."))
+ opt = s + 3;
+ else
+ opt = s;
+
+ if (startswith(opt, "udev.log-priority=")) {
+ int prio;
+
+ prio = util_log_priority(opt + 18);
+ log_set_max_level(prio);
+ udev_set_log_priority(udev, prio);
+ } else if (startswith(opt, "udev.children-max=")) {
+ children_max = strtoul(opt + 18, NULL, 0);
+ } else if (startswith(opt, "udev.exec-delay=")) {
+ exec_delay = strtoul(opt + 16, NULL, 0);
}
- /* update timestamp */
- stamp_usec[i] = ts_usec(&stats.st_mtim);
+ free(s);
}
-
- return changed;
}
int main(int argc, char *argv[])
{
struct udev *udev;
- FILE *f;
sigset_t mask;
int daemonize = false;
int resolve_names = 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;
- char **s;
int rc = 1;
udev = udev_new();
if (udev == NULL)
goto exit;
- udev_log_init("udevd");
+ log_set_target(LOG_TARGET_AUTO);
+ log_parse_environment();
+ log_open();
+
udev_set_log_fn(udev, udev_main_log);
- info(udev, "version %s\n", VERSION);
- udev_selinux_init(udev);
+ log_set_max_level(udev_get_log_priority(udev));
+
+ log_debug("version %s", VERSION);
+ label_init("/dev");
for (;;) {
int option;
- option = getopt_long(argc, argv, "c:deDtN:hV", options, NULL);
+ option = getopt_long(argc, argv, "c:de:DtN:hV", options, NULL);
if (option == -1)
break;
break;
case 'D':
debug = true;
- if (udev_get_log_priority(udev) < LOG_INFO)
- udev_set_log_priority(udev, LOG_INFO);
+ log_set_max_level(LOG_DEBUG);
+ udev_set_log_priority(udev, LOG_DEBUG);
break;
case 'N':
- if (strcmp (optarg, "early") == 0) {
+ if (streq(optarg, "early")) {
resolve_names = 1;
- } else if (strcmp (optarg, "late") == 0) {
+ } else if (streq(optarg, "late")) {
resolve_names = 0;
- } else if (strcmp (optarg, "never") == 0) {
+ } else if (streq(optarg, "never")) {
resolve_names = -1;
} else {
fprintf(stderr, "resolve-names must be early, late or never\n");
- err(udev, "resolve-names must be early, late or never\n");
+ log_error("resolve-names must be early, late or never");
goto exit;
}
break;
}
}
- /*
- * 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);
- }
+ kernel_cmdline_options(udev);
if (getuid() != 0) {
fprintf(stderr, "root privileges required\n");
- err(udev, "root privileges required\n");
+ log_error("root privileges required");
goto exit;
}
chdir("/");
umask(022);
- /* /run/udev */
- mkdir(udev_get_run_path(udev), 0755);
+ mkdir("/run/udev", 0755);
- /* create standard links, copy static nodes, create nodes from modules */
- static_dev_create(udev);
- static_dev_create_from_modules(udev);
+ dev_setup(NULL);
/* before opening new files, make sure std{in,out,err} fds are in a sane state */
if (daemonize) {
close(fd);
} else {
fprintf(stderr, "cannot open /dev/null\n");
- err(udev, "cannot open /dev/null\n");
+ log_error("cannot open /dev/null");
}
}
if (systemd_fds(udev, &fd_ctrl, &fd_netlink) >= 0) {
- /* get control and netlink socket from from systemd */
+ /* get control and netlink socket from systemd */
udev_ctrl = udev_ctrl_new_from_fd(udev, fd_ctrl);
if (udev_ctrl == NULL) {
- err(udev, "error taking over udev control socket");
+ log_error("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");
+ log_error("error taking over netlink socket");
rc = 3;
goto exit;
}
+
+ /* get our own cgroup, we regularly kill everything udev has left behind */
+ if (cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &udev_cgroup) < 0)
+ udev_cgroup = NULL;
} else {
/* open control and netlink socket */
udev_ctrl = udev_ctrl_new(udev);
if (udev_ctrl == NULL) {
fprintf(stderr, "error initializing udev control socket");
- err(udev, "error initializing udev control socket");
+ log_error("error initializing udev control socket");
rc = 1;
goto exit;
}
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");
+ log_error("error initializing netlink socket");
rc = 3;
goto exit;
}
if (udev_monitor_enable_receiving(monitor) < 0) {
fprintf(stderr, "error binding netlink socket\n");
- err(udev, "error binding netlink socket\n");
+ log_error("error binding netlink socket");
rc = 3;
goto exit;
}
if (udev_ctrl_enable_receiving(udev_ctrl) < 0) {
fprintf(stderr, "error binding udev control socket\n");
- err(udev, "error binding udev control socket\n");
+ log_error("error binding udev control socket");
rc = 1;
goto exit;
}
- udev_monitor_set_receive_buffer_size(monitor, 128*1024*1024);
-
- /* 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_monitor_set_receive_buffer_size(monitor, 128 * 1024 * 1024);
if (daemonize) {
pid_t pid;
- int fd;
pid = fork();
switch (pid) {
case 0:
break;
case -1:
- err(udev, "fork of daemon failed: %m\n");
+ log_error("fork of daemon failed: %m");
rc = 4;
goto exit;
default:
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);
- }
+ write_string_file("/proc/self/oom_score_adj", "-1000");
} else {
sd_notify(1, "READY=1");
}
- f = fopen("/dev/kmsg", "w");
- if (f != NULL) {
- fprintf(f, "<30>udevd[%u]: starting version " VERSION "\n", getpid());
- fclose(f);
- }
+ print_kmsg("starting version " VERSION "\n");
if (!debug) {
int fd;
fd_inotify = udev_watch_init(udev);
if (fd_inotify < 0) {
fprintf(stderr, "error initializing inotify\n");
- err(udev, "error initializing inotify\n");
+ log_error("error initializing inotify");
rc = 4;
goto exit;
}
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");
+ log_error("error creating signalfd");
rc = 5;
goto exit;
}
/* 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");
+ log_error("error creating socketpair");
rc = 6;
goto exit;
}
rules = udev_rules_new(udev, resolve_names);
if (rules == NULL) {
- err(udev, "error reading rules\n");
+ log_error("error reading rules");
goto exit;
}
- memset(&ep_ctrl, 0, sizeof(struct epoll_event));
+ memzero(&ep_ctrl, sizeof(struct epoll_event));
ep_ctrl.events = EPOLLIN;
ep_ctrl.data.fd = fd_ctrl;
- memset(&ep_inotify, 0, sizeof(struct epoll_event));
+ memzero(&ep_inotify, sizeof(struct epoll_event));
ep_inotify.events = EPOLLIN;
ep_inotify.data.fd = fd_inotify;
- memset(&ep_signal, 0, sizeof(struct epoll_event));
+ memzero(&ep_signal, sizeof(struct epoll_event));
ep_signal.events = EPOLLIN;
ep_signal.data.fd = fd_signal;
- memset(&ep_netlink, 0, sizeof(struct epoll_event));
+ memzero(&ep_netlink, sizeof(struct epoll_event));
ep_netlink.events = EPOLLIN;
ep_netlink.data.fd = fd_netlink;
- memset(&ep_worker, 0, sizeof(struct epoll_event));
+ memzero(&ep_worker, 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");
+ log_error("error creating epoll fd: %m");
goto exit;
}
if (epoll_ctl(fd_ep, EPOLL_CTL_ADD, fd_ctrl, &ep_ctrl) < 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");
+ log_error("fail to add fds to epoll: %m");
goto exit;
}
- /* if needed, convert old database from earlier udev version */
- convert_db(udev);
-
if (children_max <= 0) {
- int memsize = mem_size_mb();
+ cpu_set_t cpu_set;
- /* set value depending on the amount of RAM */
- if (memsize > 0)
- children_max = 128 + (memsize / 8);
- else
- children_max = 128;
+ children_max = 8;
+
+ if (sched_getaffinity(0, sizeof (cpu_set), &cpu_set) == 0) {
+ children_max += CPU_COUNT(&cpu_set) * 2;
+ }
}
- info(udev, "set children_max to %u\n", children_max);
+ log_debug("set children_max to %u", children_max);
- udev_rules_apply_static_dev_perms(rules);
+ rc = udev_rules_apply_static_dev_perms(rules);
+ if (rc < 0)
+ log_error("failed to apply permissions on static device nodes - %s", strerror(-rc));
udev_list_node_init(&event_list);
udev_list_node_init(&worker_list);
for (;;) {
- static unsigned long long last_usec;
+ static usec_t last_usec;
struct epoll_event ev[8];
int fdcount;
int timeout;
/* discard queued events and kill workers */
event_queue_cleanup(udev, EVENT_QUEUED);
- worker_kill(udev, 0);
+ worker_kill(udev);
/* exit after all has cleaned up */
- if (udev_list_node_is_empty(&event_list) && udev_list_node_is_empty(&worker_list))
+ if (udev_list_node_is_empty(&event_list) && children == 0)
break;
/* timeout at exit for workers to finish */
- timeout = 30 * 1000;
- } else if (udev_list_node_is_empty(&event_list) && children <= 2) {
+ timeout = 30 * MSEC_PER_SEC;
+ } else if (udev_list_node_is_empty(&event_list) && children == 0) {
/* we are idle */
timeout = -1;
+
+ /* cleanup possible left-over processes in our cgroup */
+ if (udev_cgroup)
+ cg_kill(SYSTEMD_CGROUP_CONTROLLER, udev_cgroup, SIGKILL, false, true, NULL);
} else {
/* kill idle or hanging workers */
- timeout = 3 * 1000;
+ timeout = 3 * MSEC_PER_SEC;
+ }
+
+ /* tell settle that we are busy or idle */
+ if (!udev_list_node_is_empty(&event_list)) {
+ int fd;
+
+ fd = open("/run/udev/queue", O_WRONLY|O_CREAT|O_CLOEXEC|O_TRUNC|O_NOFOLLOW, 0444);
+ if (fd >= 0)
+ close(fd);
+ } else {
+ unlink("/run/udev/queue");
}
- fdcount = epoll_wait(fd_ep, ev, ARRAY_SIZE(ev), timeout);
+
+ fdcount = epoll_wait(fd_ep, ev, ELEMENTSOF(ev), timeout);
if (fdcount < 0)
continue;
/* timeout */
if (udev_exit) {
- err(udev, "timeout, giving up waiting for workers to finish\n");
+ log_error("timeout, giving up waiting for workers to finish");
break;
}
/* kill idle workers */
if (udev_list_node_is_empty(&event_list)) {
- info(udev, "cleanup idle workers\n");
- worker_kill(udev, 2);
+ log_debug("cleanup idle workers");
+ worker_kill(udev);
}
/* check for hanging events */
if (worker->state != WORKER_RUNNING)
continue;
- if ((now_usec() - worker->event_start_usec) > 30 * 1000 * 1000) {
- err(udev, "worker [%u] timeout, kill it\n", worker->pid,
+ if ((now(CLOCK_MONOTONIC) - worker->event_start_usec) > 30 * USEC_PER_SEC) {
+ log_error("worker [%u] %s timeout; kill it", worker->pid,
worker->event ? worker->event->devpath : "<idle>");
kill(worker->pid, SIGKILL);
worker->state = WORKER_KILLED;
+
/* drop reference taken for state 'running' */
worker_unref(worker);
if (worker->event) {
- err(udev, "seq %llu '%s' killed\n",
- udev_device_get_seqnum(worker->event->dev), worker->event->devpath);
+ log_error("seq %llu '%s' killed", udev_device_get_seqnum(worker->event->dev), worker->event->devpath);
worker->event->exitcode = -64;
- event_queue_delete(worker->event, true);
+ event_queue_delete(worker->event);
worker->event = NULL;
}
}
}
/* check for changed config, every 3 seconds at most */
- if ((now_usec() - last_usec) > 3 * 1000 * 1000) {
- if (check_rules_timestamp(udev))
+ if ((now(CLOCK_MONOTONIC) - last_usec) > 3 * USEC_PER_SEC) {
+ if (udev_rules_check_timestamp(rules))
reload = true;
if (udev_builtin_validate(udev))
reload = true;
- last_usec = now_usec();
+ last_usec = now(CLOCK_MONOTONIC);
}
/* reload requested, HUP signal received, rules changed, builtin changed */
if (reload) {
- worker_kill(udev, 0);
+ worker_kill(udev);
rules = udev_rules_unref(rules);
udev_builtin_exit(udev);
- reload = 0;
+ reload = false;
}
/* event has finished */
dev = udev_monitor_receive_device(monitor);
if (dev != NULL) {
- udev_device_set_usec_initialized(dev, now_usec());
+ udev_device_set_usec_initialized(dev, now(CLOCK_MONOTONIC));
if (event_queue_insert(dev) < 0)
udev_device_unref(dev);
}
/* start new events */
if (!udev_list_node_is_empty(&event_list) && !udev_exit && !stop_exec_queue) {
+ udev_builtin_init(udev);
if (rules == NULL)
rules = udev_rules_new(udev, resolve_names);
if (rules != NULL)
rc = EXIT_SUCCESS;
exit:
- udev_queue_export_cleanup(udev_queue_export);
udev_ctrl_cleanup(udev_ctrl);
+ unlink("/run/udev/queue");
exit_daemonize:
if (fd_ep >= 0)
close(fd_ep);
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);
- udev_selinux_exit(udev);
+ label_finish();
udev_unref(udev);
- udev_log_close();
+ log_close();
return rc;
}
~ianmdlvl / elogind.git / blobdiff