fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <stddef.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <signal.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <unistd.h>

6bedfcbb2970e06a4d3280c8fb62083d252ede73Lennart Poettering#include <errno.h>

8b43440b7ef4b81c69c31de7ff820dc07a780254Lennart Poettering#include <stdio.h>

07630cea1f3a845c09309f197ac7c4f11edd3b62Lennart Poettering#include <stdlib.h>

07630cea1f3a845c09309f197ac7c4f11edd3b62Lennart Poettering#include <stdbool.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <string.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <ctype.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <fcntl.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <time.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <getopt.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <dirent.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/file.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/time.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/prctl.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/socket.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/un.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/signalfd.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/epoll.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/mount.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/poll.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/wait.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/stat.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/ioctl.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/inotify.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include <sys/utsname.h>

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "udev.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "udev-util.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "sd-daemon.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "cgroup-util.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "dev-setup.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#include "fileio.h"

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic bool debug;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringvoid udev_main_log(struct udev *udev, int priority,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering const char *file, int line, const char *fn,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering const char *format, va_list args)

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering{

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering log_metav(priority, file, line, fn, format, args);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering}

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic struct udev_rules *rules;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic struct udev_ctrl *udev_ctrl;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic struct udev_monitor *monitor;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int worker_watch[2] = { -1, -1 };

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int fd_signal = -1;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int fd_ep = -1;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int fd_inotify = -1;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic bool stop_exec_queue;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic bool reload;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int children;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int children_max;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic int exec_delay;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic sigset_t sigmask_orig;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic UDEV_LIST(event_list);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic UDEV_LIST(worker_list);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic char *udev_cgroup;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic bool udev_exit;

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poetteringenum event_state {

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering EVENT_UNDEF,

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering EVENT_QUEUED,

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering EVENT_RUNNING,

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering};

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poetteringstruct event {

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev_list_node node;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev *udev;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev_device *dev;

765afd5c4dbc71940d6dd6007ecc3eaa5a0b2aa1Lennart Poettering enum event_state state;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering int exitcode;

85e070c2b3f66d838c11f8f008106b9e42c9afafMichal Sekletar unsigned long long int delaying_seqnum;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering unsigned long long int seqnum;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering const char *devpath;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering size_t devpath_len;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering const char *devpath_old;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering dev_t devnum;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering int ifindex;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering bool is_block;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#ifdef HAVE_FIRMWARE

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering bool nodelay;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering#endif

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering};

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic inline struct event *node_to_event(struct udev_list_node *node)

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering{

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering return container_of(node, struct event, node);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering}

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic void event_queue_cleanup(struct udev *udev, enum event_state type);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringenum worker_state {

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering WORKER_UNDEF,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering WORKER_RUNNING,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering WORKER_IDLE,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering WORKER_KILLED,

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering};

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstruct worker {

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev_list_node node;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev *udev;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering int refcount;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering pid_t pid;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct udev_monitor *monitor;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering enum worker_state state;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering struct event *event;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering usec_t event_start_usec;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering};

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstruct worker_message {

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering pid_t pid;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering int exitcode;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering};

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic inline struct worker *node_to_worker(struct udev_list_node *node)

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering{

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering return container_of(node, struct worker, node);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering}

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic void event_queue_delete(struct event *event)

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering{

12ca818ffddb77eb6a0fabe369a5bcbf6994ff8bLennart Poettering udev_list_node_remove(&event->node);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering udev_device_unref(event->dev);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering free(event);

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering}

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poetteringstatic struct worker *worker_ref(struct worker *worker)

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering{

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering worker->refcount++;

fc2f9534d07db2b185c02b1961428b53daf1986eLennart Poettering return worker;

}

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;

log_debug("worker [%u] cleaned up", 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;

/* 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 = new0(struct worker, 1);

if (worker == NULL) {

udev_monitor_unref(worker_monitor);

return;

}

/* worker + event reference */

worker->refcount = 2;

worker->udev = udev;

pid = fork();

switch (pid) {

case 0: {

struct udev_device *dev = NULL;

int fd_monitor;

struct epoll_event ep_signal, ep_monitor;

sigset_t mask;

int rc = EXIT_SUCCESS;

/* take initial device from queue */

dev = event->dev;

event->dev = NULL;

free(worker);

worker_list_cleanup(udev);

event_queue_cleanup(udev, EVENT_UNDEF);

udev_monitor_unref(monitor);

udev_ctrl_unref(udev_ctrl);

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) {

log_error("error creating signalfd %m");

rc = 2;

goto out;

}

fd_ep = epoll_create1(EPOLL_CLOEXEC);

if (fd_ep < 0) {

log_error("error creating epoll fd: %m");

rc = 3;

goto out;

}

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);

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) {

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 fd_lock = -1;

int err = 0;

log_debug("seq %llu running", 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;

/*

/*

if (streq_ptr("block", udev_device_get_subsystem(dev)) &&

!startswith(udev_device_get_sysname(dev), "dm-")) {

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 */

udev_event_execute_rules(udev_event, rules, &sigmask_orig);

udev_event_execute_run(udev_event, &sigmask_orig);

/* apply/restore 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 */

memzero(&msg, sizeof(struct worker_message));

msg.exitcode = err;

msg.pid = getpid();

send(worker_watch[WRITE_END], &msg, sizeof(struct worker_message), 0);

log_debug("seq %llu processed with %i", 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, ELEMENTSOF(ev), -1);

if (fdcount < 0) {

if (errno == EINTR)

continue;

log_error("failed to poll: %m");

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);

break;

} 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);

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);

log_close();

exit(rc);

}

case -1:

udev_monitor_unref(worker_monitor);

event->state = EVENT_QUEUED;

free(worker);

log_error("fork of child failed: %m");

break;

default:

/* close monitor, but keep address around */

udev_monitor_disconnect(worker_monitor);

worker->monitor = worker_monitor;

worker->pid = pid;

worker->state = WORKER_RUNNING;

worker->event_start_usec = now(CLOCK_MONOTONIC);

worker->event = event;

event->state = EVENT_RUNNING;

udev_list_node_append(&worker->node, &worker_list);

children++;

log_debug("seq %llu forked new worker [%u]", udev_device_get_seqnum(event->dev), pid);

break;

}

}

static void event_run(struct event *event)

{

struct udev_list_node *loop;

udev_list_node_foreach(loop, &worker_list) {

struct worker *worker = node_to_worker(loop);

ssize_t count;

if (worker->state != WORKER_IDLE)

continue;

count = udev_monitor_send_device(monitor, worker->monitor, event->dev);

if (count < 0) {

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(CLOCK_MONOTONIC);

event->state = EVENT_RUNNING;

return;

}

if (children >= children_max) {

if (children_max > 1)

log_debug("maximum number (%i) of children reached", children);

return;

}

/* start new worker and pass initial device */

worker_new(event);

}

static int event_queue_insert(struct udev_device *dev)

{

struct event *event;

event = new0(struct event, 1);

if (event == NULL)

return -1;

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 = 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

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;

udev_list_node_append(&event->node, &event_list);

return 0;

}

static void worker_kill(struct udev *udev)

{

struct udev_list_node *loop;

udev_list_node_foreach(loop, &worker_list) {

struct worker *worker = node_to_worker(loop);

if (worker->state == WORKER_KILLED)

continue;

worker->state = WORKER_KILLED;

kill(worker->pid, SIGTERM);

}

}

static bool is_devpath_busy(struct event *event)

{

struct udev_list_node *loop;

size_t common;

/* check if queue contains events we depend on */

udev_list_node_foreach(loop, &event_list) {

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 (loop_event->seqnum < event->delaying_seqnum)

continue;

/* event we checked earlier still exists, no need to check again */

if (loop_event->seqnum == event->delaying_seqnum)

return true;

/* found ourself, no later event can block us */

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 (event->devpath_old != NULL && streq(loop_event->devpath, event->devpath_old)) {

event->delaying_seqnum = loop_event->seqnum;

return true;

}

/* 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;

}

#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;

return true;

}

/* child device event found */

if (loop_event->devpath[common] == '/') {

event->delaying_seqnum = loop_event->seqnum;

return true;

}

/* no matching device */

continue;

}

return false;

}

static void event_queue_start(struct udev *udev)

{

struct udev_list_node *loop;

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))

continue;

event_run(event);

}

}

static void event_queue_cleanup(struct udev *udev, enum event_state match_type)

{

struct udev_list_node *loop, *tmp;

udev_list_node_foreach_safe(loop, tmp, &event_list) {

struct event *event = node_to_event(loop);

if (match_type != EVENT_UNDEF && match_type != event->state)

continue;

event_queue_delete(event);

}

}

static void worker_returned(int fd_worker)

{

for (;;) {

struct worker_message msg;

ssize_t size;

struct udev_list_node *loop;

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;

/* worker returned */

if (worker->event) {

worker->event->exitcode = msg.exitcode;

event_queue_delete(worker->event);

worker->event = NULL;

}

if (worker->state != WORKER_KILLED)

worker->state = WORKER_IDLE;

worker_unref(worker);

break;

}

}

}

static struct udev_ctrl_connection *handle_ctrl_msg(struct udev_ctrl *uctrl)

{

struct udev *udev = udev_ctrl_get_udev(uctrl);

struct udev_ctrl_connection *ctrl_conn;

struct udev_ctrl_msg *ctrl_msg = NULL;

const char *str;

int i;

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)

goto out;

i = udev_ctrl_get_set_log_level(ctrl_msg);

if (i >= 0) {

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);

}

if (udev_ctrl_get_stop_exec_queue(ctrl_msg) > 0) {

log_debug("udevd message (STOP_EXEC_QUEUE) received");

stop_exec_queue = true;

}

if (udev_ctrl_get_start_exec_queue(ctrl_msg) > 0) {

log_debug("udevd message (START_EXEC_QUEUE) received");

stop_exec_queue = false;

}

if (udev_ctrl_get_reload(ctrl_msg) > 0) {

log_debug("udevd message (RELOAD) received");

reload = true;

}

str = udev_ctrl_get_set_env(ctrl_msg);

if (str != NULL) {

char *key;

key = strdup(str);

if (key != NULL) {

char *val;

val = strchr(key, '=');

if (val != NULL) {

val[0] = '\0';

val = &val[1];

if (val[0] == '\0') {

log_debug("udevd message (ENV) received, unset '%s'", key);

udev_add_property(udev, key, NULL);

} else {

log_debug("udevd message (ENV) received, set '%s=%s'", key, val);

udev_add_property(udev, key, val);

}

} else {

log_error("wrong key format '%s'", key);

}

free(key);

}

worker_kill(udev);

}

i = udev_ctrl_get_set_children_max(ctrl_msg);

if (i >= 0) {

log_debug("udevd message (SET_MAX_CHILDREN) received, children_max=%i", i);

children_max = i;

}

if (udev_ctrl_get_ping(ctrl_msg) > 0)

log_debug("udevd message (SYNC) received");

if (udev_ctrl_get_exit(ctrl_msg) > 0) {

log_debug("udevd message (EXIT) received");

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);

}

static int synthesize_change(struct udev_device *dev) {

char filename[UTIL_PATH_SIZE];

int r;

if (streq_ptr("block", udev_device_get_subsystem(dev)) &&

streq_ptr("disk", udev_device_get_devtype(dev)) &&

!startswith(udev_device_get_sysname(dev), "dm-")) {

bool part_table_read = false;

bool has_partitions = false;

int fd;

struct udev *udev = udev_device_get_udev(dev);

_cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL;

struct udev_list_entry *item;

/*

fd = open(udev_device_get_devnode(dev), O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK);

if (fd >= 0) {

r = flock(fd, LOCK_EX|LOCK_NB);

if (r >= 0)

r = ioctl(fd, BLKRRPART, 0);

close(fd);

if (r >= 0)

part_table_read = true;

}

/* search for partitions */

e = udev_enumerate_new(udev);

if (!e)

return -ENOMEM;

r = udev_enumerate_add_match_parent(e, dev);

if (r < 0)

return r;

r = udev_enumerate_add_match_subsystem(e, "block");

if (r < 0)

return r;

r = udev_enumerate_scan_devices(e);

udev_list_entry_foreach(item, udev_enumerate_get_list_entry(e)) {

_cleanup_udev_device_unref_ struct udev_device *d = NULL;

d = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));

if (!d)

continue;

if (!streq_ptr("partition", udev_device_get_devtype(d)))

continue;

has_partitions = true;

break;

}

/*

if (part_table_read && has_partitions)

return 0;

/*

log_debug("device %s closed, synthesising 'change'", udev_device_get_devnode(dev));

strscpyl(filename, sizeof(filename), udev_device_get_syspath(dev), "/uevent", NULL);

write_string_file(filename, "change");

udev_list_entry_foreach(item, udev_enumerate_get_list_entry(e)) {

_cleanup_udev_device_unref_ struct udev_device *d = NULL;

d = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));

if (!d)

continue;

if (!streq_ptr("partition", udev_device_get_devtype(d)))

continue;

log_debug("device %s closed, synthesising partition '%s' 'change'",

udev_device_get_devnode(dev), udev_device_get_devnode(d));

strscpyl(filename, sizeof(filename), udev_device_get_syspath(d), "/uevent", NULL);

write_string_file(filename, "change");

}

return 0;

}

log_debug("device %s closed, synthesising 'change'", udev_device_get_devnode(dev));

strscpyl(filename, sizeof(filename), udev_device_get_syspath(dev), "/uevent", NULL);

write_string_file(filename, "change");

return 0;

}

static int handle_inotify(struct udev *udev)

{

int nbytes, pos;

char *buf;

struct inotify_event *ev;

int r;

r = ioctl(fd_inotify, FIONREAD, &nbytes);

if (r < 0 || nbytes <= 0)

return -errno;

buf = malloc(nbytes);

if (!buf) {

log_error("error getting buffer for inotify");

return -ENOMEM;

}

nbytes = read(fd_inotify, buf, nbytes);

for (pos = 0; pos < nbytes; pos += sizeof(struct inotify_event) + ev->len) {

struct udev_device *dev;

ev = (struct inotify_event *)(buf + pos);

dev = udev_watch_lookup(udev, ev->wd);

if (!dev)

continue;

log_debug("inotify event: %x for %s", ev->mask, udev_device_get_devnode(dev));

if (ev->mask & IN_CLOSE_WRITE)

synthesize_change(dev);

else if (ev->mask & IN_IGNORED)

udev_watch_end(udev, dev);

udev_device_unref(dev);

}

free(buf);

return 0;

}

static void handle_signal(struct udev *udev, int signo)

{

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;

log_debug("worker [%u] exit", pid);

if (WIFEXITED(status)) {

if (WEXITSTATUS(status) != 0)

log_error("worker [%u] exit with return code %i",

pid, WEXITSTATUS(status));

} else if (WIFSIGNALED(status)) {

log_error("worker [%u] terminated by signal %i (%s)",

pid, WTERMSIG(status), strsignal(WTERMSIG(status)));

} else if (WIFSTOPPED(status)) {

log_error("worker [%u] stopped", pid);

} else if (WIFCONTINUED(status)) {

log_error("worker [%u] continued", pid);

} else {

log_error("worker [%u] exit with status 0x%04x", pid, status);

}

if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {

if (worker->event) {

log_error("worker [%u] failed while handling '%s'",

pid, worker->event->devpath);

worker->event->exitcode = -32;

event_queue_delete(worker->event);

/* drop reference taken for state 'running' */

worker_unref(worker);

}

}

worker_unref(worker);

break;

}

}

break;

case SIGHUP:

reload = true;

break;

}

}

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;

log_debug("ctrl=%i netlink=%i", ctrl, netlink);

*rctrl = ctrl;

*rnetlink = netlink;

return 0;

}

static void kernel_cmdline_options(struct udev *udev)

{

_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;

FOREACH_WORD_QUOTED(w, l, line, state) {

char *s, *opt;

s = strndup(w, l);

if (!s)

break;

/* 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);

}

free(s);

}

}

int main(int argc, char *argv[])

{

struct udev *udev;

sigset_t mask;

int daemonize = false;

int resolve_names = 1;

static const struct option options[] = {

{ "daemon", no_argument, NULL, 'd' },

{ "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' },

{}

};

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;

log_set_target(LOG_TARGET_AUTO);

log_parse_environment();

log_open();

udev_set_log_fn(udev, udev_main_log);

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:de:DtN:hV", options, NULL);

if (option == -1)

break;

switch (option) {

case 'd':

daemonize = true;

break;

case 'c':

children_max = strtoul(optarg, NULL, 0);

break;

case 'e':

exec_delay = strtoul(optarg, NULL, 0);

break;

case 'D':

debug = true;

log_set_max_level(LOG_DEBUG);

udev_set_log_priority(udev, LOG_DEBUG);

break;

case 'N':

if (streq(optarg, "early")) {

resolve_names = 1;

} else if (streq(optarg, "late")) {

resolve_names = 0;

} else if (streq(optarg, "never")) {

resolve_names = -1;

} else {

fprintf(stderr, "resolve-names must be early, late or never\n");

log_error("resolve-names must be early, late or never");

goto exit;

}

break;

case 'h':

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);

goto exit;

default:

goto exit;

}

}

kernel_cmdline_options(udev);

if (getuid() != 0) {

fprintf(stderr, "root privileges required\n");

log_error("root privileges required");

goto exit;

}

/* set umask before creating any file/directory */

chdir("/");

umask(022);

mkdir("/run/udev", 0755);

dev_setup(NULL);

/* 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");

log_error("cannot open /dev/null");

}

}

if (systemd_fds(udev, &fd_ctrl, &fd_netlink) >= 0) {

/* get control and netlink socket from systemd */

udev_ctrl = udev_ctrl_new_from_fd(udev, fd_ctrl);

if (udev_ctrl == NULL) {

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) {

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");

log_error("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");

log_error("error initializing netlink socket");

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");

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");

log_error("error binding udev control socket");

rc = 1;

goto exit;

}

udev_monitor_set_receive_buffer_size(monitor, 128 * 1024 * 1024);

if (daemonize) {

pid_t pid;

pid = fork();

switch (pid) {

case 0:

break;

case -1:

log_error("fork of daemon failed: %m");

rc = 4;

goto exit;

default:

rc = EXIT_SUCCESS;

goto exit_daemonize;

}

setsid();

write_string_file("/proc/self/oom_score_adj", "-1000");

} else {

sd_notify(1, "READY=1");

}

print_kmsg("starting version " VERSION "\n");

if (!debug) {

int 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_inotify = udev_watch_init(udev);

if (fd_inotify < 0) {

fprintf(stderr, "error initializing inotify\n");

log_error("error initializing inotify");

rc = 4;

goto exit;

}

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");

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");

log_error("error creating socketpair");

rc = 6;

goto exit;

}

fd_worker = worker_watch[READ_END];

udev_builtin_init(udev);

rules = udev_rules_new(udev, resolve_names);

if (rules == NULL) {

log_error("error reading rules");

goto exit;

}

memzero(&ep_ctrl, sizeof(struct epoll_event));

ep_ctrl.events = EPOLLIN;

ep_ctrl.data.fd = fd_ctrl;

memzero(&ep_inotify, sizeof(struct epoll_event));

ep_inotify.events = EPOLLIN;

ep_inotify.data.fd = fd_inotify;

memzero(&ep_signal, sizeof(struct epoll_event));

ep_signal.events = EPOLLIN;

ep_signal.data.fd = fd_signal;

memzero(&ep_netlink, sizeof(struct epoll_event));

ep_netlink.events = EPOLLIN;

ep_netlink.data.fd = fd_netlink;

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) {

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_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) {

log_error("fail to add fds to epoll: %m");

goto exit;

}

if (children_max <= 0) {

cpu_set_t cpu_set;