/* SPDX-License-Identifier: GPL-2.0-only */
#include "m7a.h"

#include <sys/ipc.h>
#include <sys/sem.h>
#include <fcntl.h>
#include <ctype.h>
#include <sys/mount.h>
#include <dirent.h>
#include <net/if.h>
#include <linux/sockios.h>
#include <libmnl/libmnl.h>
#include <linux/if_link.h>
#include <linux/rtnetlink.h>
#include <sys/un.h>

void die(const char *fmt, ...)
{
	char *str;
	va_list argp;
	int ret;

	va_start(argp, fmt);
	ret = vasprintf(&str, fmt, argp);
	va_end(argp);
	if (ret < 0) { /* give up */
		EMERG_LOG("OOM\n");
		exit(EXIT_FAILURE);
	}
	m7a_log(LL_EMERG, "%s\n", str);
	exit(EXIT_FAILURE);
}

void die_errno(const char *fmt, ...)
{
	char *str;
	va_list argp;
	int ret, save_errno = errno;

	va_start(argp, fmt);
	ret = vasprintf(&str, fmt, argp);
	va_end(argp);
	if (ret < 0) {
		EMERG_LOG("OOM\n");
		exit(EXIT_FAILURE);
	}
	m7a_log(LL_EMERG, "%s: %s\n", str, strerror(save_errno));
	exit(EXIT_FAILURE);
}

void *xrealloc(void *p, size_t size)
{
	assert(size > 0);
	assert((p = realloc(p, size)));
	return p;
}

void *xmalloc(size_t size)
{
	return xrealloc(NULL, size);
}

void *xzmalloc(size_t size)
{
	void *p = xrealloc(NULL, size);
	memset(p, 0, size);
	return p;
}

void *xstrdup(const char *s)
{
	char *ret = strdup(s? s: "");

	assert(ret);
	return ret;
}

char *msg(const char *fmt, ...)
{
	char *m;
	size_t size = 100;

	m = xmalloc(size);
	while (1) {
		int n;
		va_list ap;

		/* Try to print in the allocated space. */
		va_start(ap, fmt);
		n = vsnprintf(m, size, fmt, ap);
		va_end(ap);
		/* If that worked, return the string. */
		if (n < size)
			return m;
		/* Else try again with more space. */
		size = n + 1; /* precisely what is needed */
		m = xrealloc(m, size);
	}
}

char *xstrcat(char *a, const char *b)
{
	char *tmp;

	if (!a)
		return xstrdup(b);
	if (!b)
		return a;
	tmp = msg("%s%s", a, b);
	free(a);
	return tmp;
}

void die_empty_arg(const char *opt)
{
	die("argument to --%s must not be empty", opt);
}

__attribute__ ((noreturn))
static void die_range(const char *opt)
{
	die("argument to --%s is out of range", opt);
}

void check_range(uint32_t val, uint32_t min, uint32_t max, const char *opt)
{
	if (val < min || val > max)
		die_range(opt);
}

bool fd2buf(int fd, const struct iovec *iov)
{
	ssize_t ret, nread = 0, max;
	char *buf = iov->iov_base;

	assert(iov->iov_len > 1);
	max = iov->iov_len - 1;
	for (;;) {
		ret = read(fd, buf + nread, max - nread);
		if (ret < 0) {
			if (errno == EAGAIN || errno == EINTR)
				continue;
			ERROR_LOG("read error: %s\n", strerror(errno));
			return false;
		}
		if (ret == 0) {
			buf[nread] = '\0';
			DEBUG_LOG("read %zd bytes\n", nread);
			return true;
		}
		nread += ret;
		if (nread >= max) {
			ERROR_LOG("cmd output truncated\n");
			return false;
		}
	}
}

bool xexec(char * const argv[])
{
	pid_t pid;
	unsigned n;

	for (n = 0; argv[n]; n++)
		DEBUG_LOG("argv[%u]=%s\n", n, argv[n]);
	if ((pid = fork()) < 0)
		die_errno("fork");
	if (pid > 0) { /* parent */
		int wstatus;
		if (waitpid(pid, &wstatus, WUNTRACED) < 0)
			die_errno("waitp");
		/*
		 * If the user shuts down the container from an interactive
		 * com_enter() shell, the nsenter process receives SIGSTOP
		 * because the container shuts down the controlling tty,
		 * causing the shell to hang. Avoid this by letting the process
		 * continue once.
		 */
		if (WIFSTOPPED(wstatus))
			kill(pid, SIGCONT);
		if (!WIFEXITED(wstatus))
			return false;
		if (WEXITSTATUS(wstatus) != EXIT_SUCCESS)
			return false;
		return true;
	}
	execvp(argv[0], argv);
	EMERG_LOG("execvp error: %s\n", strerror(errno));
	_exit(EXIT_FAILURE);
}

void valid_fd012(void)
{
	/* Ensure that file descriptors 0, 1, and 2 are valid. */
	while (1) {
		int fd = open("/dev/null", O_RDWR);
		if (fd < 0)
			die_errno("open");
		if (fd > 2) {
			close(fd);
			break;
		}
	}
}

void check_name(const char *arg)
{
	size_t m, len;
	char c;

	len = strlen(arg);
	if (len == 0)
		die("empty name");
	if (len > 32)
		die("name too long: %s", arg);
	for (m = 0; m < len; m++) {
		c = arg[m];
		if (!isascii(c))
			goto invalid;
		if (!isalnum(c) && c != '-')
			goto invalid;
	}
	return;
invalid:
	die("invalid character '%c' in name %s", c, arg);
}

/* allocates two new strings that should be freed by the caller */
void parse_compound_arg(const char *arg, const char *opt, char **name, char **val)
{
	char *copy, *p;

	if (arg[0] == '\0')
		die_empty_arg(opt);
	copy = xstrdup(arg);
	p = strchr(copy, ':');
	if (!p)
		die("could not parse argument to --%s", opt);
	*p = '\0';
	check_name(copy);
	*name = copy;
	p++;
	*val = xstrdup(p);
}

char *parse_cgroup_acl(const char *arg)
{
	if (!strncmp(arg, "allow ", 6))
		return msg("a%s", arg + 6);
	if (!strncmp(arg, "deny ", 5))
		return msg("d%s", arg + 5);
	die("invalid cgroup access specifier: %s", arg);
}

void parse_ifspec(const char *arg, char **bridge, uint8_t *hwaddr)
{
	const char *colon = strchr(arg, ':');
	size_t len;
	unsigned n, x[6];

	if (colon) {
		len = colon - arg;
		*bridge = xmalloc(len + 1);
		memcpy(*bridge, arg, len);
		(*bridge)[len] = '\0';
	} else
		*bridge = xstrdup(arg);
	check_name(*bridge);
	if (!colon) {
		memset(hwaddr, 0, 6);
		return;
	}
	if (sscanf(colon + 1, "%02x:%02x:%02x:%02x:%02x:%02x",
		x, x + 1, x + 2, x + 3, x + 4, x + 5) != 6)
		die("invalid hwaddress for ifspec %s", arg);
	if (colon[1 + 6 * 2 + 5] != '\0')
		die("trailing garbage at the end of ifspec %s", arg);
	for (n = 0; n < 6; n++)
		hwaddr[n] = x[n];
}

uint32_t atou32(const char *str, const char *opt)
{
	char *endptr;
	long long tmp;

	errno = 0; /* To distinguish success/failure after call */
	tmp = strtoll(str, &endptr, 10);
	if (errno == ERANGE && (tmp == LLONG_MAX || tmp == LLONG_MIN))
		die_range(opt);
	if (tmp < 0 || tmp > (uint32_t)-1)
		die_range(opt);
	/*
	 * If there were no digits at all, strtoll() stores the original value
	 * of str in *endptr.
	 */
	if (endptr == str)
		die_empty_arg(opt);
	/*
	 * The implementation may also set errno and return 0 in case no
	 * conversion was performed.
	 */
	if (errno != 0 && tmp == 0)
		die_empty_arg(opt);
	if (*endptr != '\0') /* Further characters after number */
		die("--%s: trailing characters after number", opt);
	return tmp;
}

bool remove_subdirs_recursively(const char *path)
{
	DIR *d = opendir(path);
	struct dirent *entry;
	int dfd;
	struct stat stat;

	if (!d) {
		ERROR_LOG("opendir %s: %m\n", path);
		return false;
	}
	dfd = dirfd(d);
	assert(dfd >= 0);
	while ((entry = readdir(d))) {
		char *subpath;
		if (!strcmp(entry->d_name, "."))
			continue;
		if (!strcmp(entry->d_name, ".."))
			continue;
		if (fstatat(dfd, entry->d_name, &stat, 0) == -1) {
			WARNING_LOG("%s/%s: %m", path, entry->d_name);
			continue;
		}
		if (!S_ISDIR(stat.st_mode))
			continue;
		subpath = msg("%s/%s", path, entry->d_name);
		remove_subdirs_recursively(subpath);
		DEBUG_LOG("removing %s\n", subpath);
		if (rmdir(subpath) < 0) {
			ERROR_LOG("rmdir %s: %m\n", subpath);
			return false;
		}
		free(subpath);
	}
	closedir(d);
	return true;
}

void daemonize(const char *logfile)
{
	pid_t pid;
	int nullfd, logfd;

	if ((pid = fork()) < 0)
		die_errno("fork");
	if (pid) /* parent exits */
		exit(EXIT_SUCCESS);
	valid_fd012();
	/* become session leader */
	if (setsid() < 0)
		die_errno("setsid");
	if ((nullfd = open("/dev/null", O_RDWR)) < 0)
		die_errno("open /dev/null");
	logfile = logfile? logfile : "/dev/null";
	if ((logfd = open(logfile, O_WRONLY | O_APPEND | O_CREAT, 0666)) < 0)
		die_errno("open %s", logfile);
	NOTICE_LOG("subsequent log messages go to %s\n", logfile);
	if (dup2(nullfd, STDIN_FILENO) < 0)
		die_errno("dup2");
	close(nullfd);
	if (dup2(logfd, STDOUT_FILENO) < 0)
		die_errno("dup2");
	if (dup2(logfd, STDERR_FILENO) < 0)
		die_errno("dup2");
	close(logfd);
	if (chdir("/") < 0)
		die_errno("chdir");
}

static int super_dull_hash(const char *input)
{
	const uint8_t *x = (typeof(x))input;
	const unsigned p1 = 16777619, p2 = 2971215073;
	unsigned n, m, h, result = 0;

	for (n = 0; n < 4; n++) {
		h = p1 * (x[0] + n);
		for (m = 1; x[m] != 0; m++)
			h = p2 * (h ^ x[m]);
		result = (result << 8) | (h % 256);
	}
	return result >> 1;
}

/**
 * We use a semaphore set with two semaphores. The first semaphore is modified
 * in all locking related functions while the second semaphore is modified only
 * in try_lock() and aquire_lock(). This allows us to obtain the PID of the
 * lock holder by querying the PID that last performed an operation on the
 * second semaphore. This is achieved by passing GETPID as the control
 * operation to semctl().
 */

static bool get_lock(const char *string, pid_t *pid, bool wait)
{
	int semid, ret;
	struct sembuf sops[4];
	key_t key = super_dull_hash(string);
	bool success;
	short sem_flg = SEM_UNDO;

	if (!wait)
		sem_flg |= IPC_NOWAIT;
	ret = semget(key, 2, IPC_CREAT | 0600);
	if (ret < 0) {
		ERROR_LOG("semget: %m\n");
		return false;
	}
	semid = ret;
	DEBUG_LOG("key: 0x%0x, semid: %d\n", (unsigned)key, semid);
	ret = semctl(semid, 1, GETPID);
	if (ret < 0)
		return false;
	if (pid)
		*pid = ret;
	sops[0].sem_num = 0;
	sops[0].sem_op = 0;
	sops[0].sem_flg = sem_flg;

	sops[1].sem_num = 0;
	sops[1].sem_op = 1;
	sops[1].sem_flg = sem_flg;

	sops[2].sem_num = 1;
	sops[2].sem_op = 0;
	sops[2].sem_flg = sem_flg;

	sops[3].sem_num = 1;
	sops[3].sem_op = 1;
	sops[3].sem_flg = sem_flg;

	success = semop(semid, sops, 4) >= 0;
	if (!success)
		INFO_LOG("semop: %m\n");
	return success;
}

bool try_lock(const char *string, pid_t *pid)
{
	return get_lock(string, pid, false /* don't wait */);
}

bool acquire_lock(const char *string)
{
	return get_lock(string, NULL /* don't need pid */, true /* do wait */);
}

bool release_lock(const char *string)
{
	int semid, ret;
	struct sembuf sops[2];
	key_t key = super_dull_hash(string);
	bool success;

	ret = semget(key, 2, IPC_CREAT | 0600);
	if (ret < 0) {
		ERROR_LOG("semget: %m\n");
		return false;
	}
	semid = ret;
	DEBUG_LOG("key: 0x%0x, semid: %d\n", (unsigned)key, semid);
	sops[0].sem_num = 0;
	sops[0].sem_op = -1;
	sops[0].sem_flg = SEM_UNDO;
	sops[1].sem_num = 1;
	sops[1].sem_op = -1;
	sops[1].sem_flg = SEM_UNDO;
	success = semop(semid, sops, 2) >= 0;
	if (!success)
		INFO_LOG("semop: %m\n");
	return success;
}

bool is_locked(const char *string, pid_t *pid)
{
	int ret, semid;
	struct sembuf sops = {
		.sem_num = 0,
		.sem_op = 0,
		.sem_flg = SEM_UNDO | IPC_NOWAIT
	};
	key_t key = super_dull_hash(string);

	if (pid)
		*pid = 0;
	ret = semget(key, 2, 0);
	if (ret < 0)
		return false;
	semid = ret;
	DEBUG_LOG("key: 0x%0x, semid: %d\n", (unsigned)key, semid);
	if (semop(semid, &sops, 1) >= 0)
		return false;
	ret = semctl(semid, 1, GETPID);
	if (ret < 0)
		return false;
	if (pid)
		*pid = ret;
	return true;
}

bool attach_to_bridge(const char *iface, const char *bridge)
{
	int fd, idx;
	struct ifreq ifr;
	bool success;

	INFO_LOG("adding interface %s to bridge %s\n", iface, bridge);
	if (!(idx = if_nametoindex(iface))) {
		ERROR_LOG("no index for %s: %m\n", iface);
		return false;
	}
	if ((fd = socket(AF_INET, SOCK_STREAM, 0)) < 0) {
		ERROR_LOG("socket: %m\n");
		return false;
	}
	strncpy(ifr.ifr_name, bridge, IFNAMSIZ - 1);
	ifr.ifr_name[IFNAMSIZ - 1] = '\0';
	ifr.ifr_ifindex = idx;
	success = ioctl(fd, SIOCBRADDIF, &ifr) == 0;
	if (!success)
		ERROR_LOG("interface %s, bridge %s: ioctl SIOCBRADDIF: %m\n",
			iface, bridge);
	close(fd);
	return success;
}


#define NLMSG_TAIL(nmsg) \
	((struct rtattr *) (((void *) (nmsg)) + NLMSG_ALIGN((nmsg)->nlmsg_len)))

static void addattr_l(struct nlmsghdr *nlh, int type, const void *data,
		int alen)
{
	int len = RTA_LENGTH(alen);
	struct rtattr *rta;

	rta = NLMSG_TAIL(nlh);
	rta->rta_type = type;
	rta->rta_len = len;
	if (alen > 0)
		memcpy(RTA_DATA(rta), data, alen);
	nlh->nlmsg_len = NLMSG_ALIGN(nlh->nlmsg_len) + RTA_ALIGN(len);
}

static struct rtattr *addattr_nest(struct nlmsghdr *n, int type)
{
	struct rtattr *nest = NLMSG_TAIL(n);
	addattr_l(n, type, NULL, 0);
	return nest;
}

static void end_nest(struct nlmsghdr *nlh, struct rtattr *attr)
{
	attr->rta_len = (void *)NLMSG_TAIL(nlh) - (void *)attr;
}

static struct mnl_socket *get_and_bind_netlink_socket(void)
{
	struct mnl_socket *nl = mnl_socket_open(NETLINK_ROUTE);

	if (!nl) {
		ERROR_LOG("mnl_socket_open error\n");
		return NULL;
	}
	if (mnl_socket_bind(nl, 0, MNL_SOCKET_AUTOPID) < 0) {
		ERROR_LOG("mnl_socket_bind\n");
		mnl_socket_close(nl);
		return NULL;
	}
	return nl;
}

static struct nlmsghdr *prepare_netlink_msg_header(char *buf)
{
	struct nlmsghdr *nlh = mnl_nlmsg_put_header(buf);
	nlh->nlmsg_flags = NLM_F_REQUEST;
	nlh->nlmsg_seq = time(NULL);
	return nlh;
}

bool rename_interface(const char *before, const char *after)
{
	int idx;
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;
	bool success;

	INFO_LOG("%s -> %s\n", before, after);
	if (!(idx = if_nametoindex(before))) {
		ERROR_LOG("no index for %s\n", before);
		return false;
	}
	if (!(nl = get_and_bind_netlink_socket()))
		return false;

	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_NEWLINK;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	ifm->ifi_index = idx;
	addattr_l(nlh, IFLA_IFNAME, after, strlen(after) + 1);
	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("mnl_socket_sendto failed\n");
		success = false;
		goto close;
	}
	success = true;
close:
	mnl_socket_close(nl);
	return success;
}

void pretty_print_hwaddr(const uint8_t *hwaddr, char *result)
{
	sprintf(result, "%02x:%02x:%02x:%02x:%02x:%02x", hwaddr[0], hwaddr[1],
		hwaddr[2], hwaddr[3], hwaddr[4], hwaddr[5]);
}

bool set_hwaddr(const char *iface, const uint8_t *hwaddr)
{
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;
	bool success;
	char pretty_hwaddr[18];

	pretty_print_hwaddr(hwaddr, pretty_hwaddr);
	INFO_LOG("hardware address of %s: %s\n", iface, pretty_hwaddr);
	if (!(nl = get_and_bind_netlink_socket()))
		return false;

	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_NEWLINK;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	addattr_l(nlh, IFLA_ADDRESS, hwaddr, 6);
	addattr_l(nlh, IFLA_IFNAME, iface, strlen(iface) + 1);
	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("%s: mnl_socket_sendto failed\n", iface);
		success = false;
		goto close;
	}
	success = true;
close:
	mnl_socket_close(nl);
	return success;
}

bool link_del(const char *iface)
{
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;
	bool success;

	INFO_LOG("removing interface %s\n", iface);
	if (!(nl = get_and_bind_netlink_socket()))
		return false;

	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_DELLINK;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	ifm->ifi_change = IFF_UP;
	ifm->ifi_flags = IFF_UP;
	addattr_l(nlh, IFLA_IFNAME, iface, strlen(iface) + 1);
	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("%s: mnl_socket_sendto failed\n", iface);
		success = false;
		goto close;
	}
	success = true;
close:
	mnl_socket_close(nl);
	return success;
}

bool link_up(const char *iface)
{
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;
	bool success;

	INFO_LOG("activating interface %s\n", iface);
	if (!(nl = get_and_bind_netlink_socket()))
		return false;
	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_NEWLINK;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	ifm->ifi_change = IFF_UP;
	ifm->ifi_flags = IFF_UP;
	addattr_l(nlh, IFLA_IFNAME, iface, strlen(iface) + 1);
	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("%s: mnl_socket_sendto failed\n", iface);
		success = false;
		goto close;
	}
	success = true;
close:
	mnl_socket_close(nl);
	return success;
}

#ifndef VETH_INFO_PEER
#define VETH_INFO_PEER 1
#endif

bool create_veth_device_pair(const char *name, char *peer)
{
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct rtattr *n1, *n2, *n3;
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;
	bool success;

	INFO_LOG("new pair: %s <-> %s\n", name, peer);
	if (!(nl = get_and_bind_netlink_socket()))
		return false;

	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_NEWLINK;
	nlh->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	n1 = addattr_nest(nlh, IFLA_LINKINFO);
	addattr_l(nlh, IFLA_INFO_KIND, "veth", 5);
	n2 = addattr_nest(nlh, IFLA_INFO_DATA);
	n3 = addattr_nest(nlh, VETH_INFO_PEER);
	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	addattr_l(nlh, IFLA_IFNAME, peer, strlen(peer) + 1);
	end_nest(nlh, n3);
	end_nest(nlh, n2);
	end_nest(nlh, n1);
	addattr_l(nlh, IFLA_IFNAME, name, strlen(name) + 1);
	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("%s: mnl_socket_sendto\n", name);
		success = false;
		goto close;
	}
	success = true;
close:
	mnl_socket_close(nl);
	return success;
}

bool set_netns(const char *iface, pid_t pid)
{
	struct mnl_socket *nl;
	char buf[MNL_SOCKET_BUFFER_SIZE];
	struct nlmsghdr *nlh;
	struct ifinfomsg *ifm;

	INFO_LOG("changing net namespace of interface %s to pid %d\n",
		iface, (int)pid);
	if (!(nl = get_and_bind_netlink_socket()))
		return false;

	nlh = prepare_netlink_msg_header(buf);
	nlh->nlmsg_type	= RTM_NEWLINK;

	ifm = mnl_nlmsg_put_extra_header(nlh, sizeof(*ifm));
	ifm->ifi_family = AF_UNSPEC;
	ifm->ifi_change = 0;
	ifm->ifi_flags = 0;
	addattr_l(nlh, IFLA_NET_NS_PID, &pid, sizeof(pid));
	mnl_attr_put_str(nlh, IFLA_IFNAME, iface);

	if (mnl_socket_sendto(nl, nlh, nlh->nlmsg_len) < 0) {
		ERROR_LOG("%s: mnl_socket_sendto failed\n", iface);
		return false;
	}
	mnl_socket_close(nl);
	return true;
}

#ifndef UNIX_PATH_MAX
#define UNIX_PATH_MAX (sizeof(((struct sockaddr_un *)0)->sun_path))
#endif

static bool init_unix_socket(const char *socket_path, int *socketfd,
		struct sockaddr_un *sau)
{
	int fd;

	*socketfd = -1;
	if (strlen(socket_path) + 1 >= UNIX_PATH_MAX) {
		ERROR_LOG("socket path to long: %s\n", socket_path);
		return false;
	}
	memset(sau, 0, sizeof(struct sockaddr_un));
	sau->sun_family = PF_UNIX;
	sau->sun_path[0] = '\0'; /* use the abstract socket namespace */
	strcpy(sau->sun_path + 1, socket_path);
	fd = socket(PF_UNIX, SOCK_STREAM, 0);
	if (fd < 0) {
		ERROR_LOG("socket: %m\n");
		return false;
	}
	*socketfd = fd;
	return true;
}

bool listen_on_unix_socket(const char *socket_path, int *result)
{
	struct sockaddr_un sau;
	int fd, flags;
	bool success = false;

	if (!init_unix_socket(socket_path, &fd, &sau))
		return false;
	flags = fcntl(fd, F_GETFL);
	if (flags < 0) {
		ERROR_LOG("fcntl (F_GETFL): %m\n");
		goto fail;
	}
	flags = fcntl(fd, F_SETFL, ((long)flags) | O_NONBLOCK);
	if (flags < 0) {
		ERROR_LOG("fcntl (F_SETFL): %m\n");
		goto fail;
	}
	if (bind(fd, (struct sockaddr *)&sau, sizeof(sau)) < 0) {
		ERROR_LOG("bind: %m\n");
		goto fail;
	}
	if (listen(fd , 5) < 0) {
		ERROR_LOG("listen: %m\n");
		goto fail;
	}
	*result = fd;
	NOTICE_LOG("listening on fd %d\n", fd);
	return true;
fail:
	close(fd);
	return success;
}
/*
 * Send a buffer and the credentials of the current process to a socket.
 *
 * buf must be zero-terminated.
 * return the return value of the underlying call to sendmsg().
 */
static bool send_cred_buffer(int sock, char *buf)
{
	char control[255] __attribute__((__aligned__(8)));
	struct msghdr msg;
	struct cmsghdr *cmsg;
	static struct iovec iov;
	struct ucred c;

	/* Response data */
	iov.iov_base = buf;
	iov.iov_len = strlen(buf) + 1;
	c.pid = getpid();
	c.uid = getuid();
	c.gid = getgid();
	/* compose the message */
	memset(&msg, 0, sizeof(msg));
	msg.msg_iov = &iov;
	msg.msg_iovlen = 1;
	msg.msg_control = control;
	msg.msg_controllen = sizeof(control);
	/* attach the ucred struct */
	cmsg = CMSG_FIRSTHDR(&msg);
	cmsg->cmsg_level = SOL_SOCKET;
	cmsg->cmsg_type = SCM_CREDENTIALS;
	cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
	*(struct ucred *)CMSG_DATA(cmsg) = c;
	msg.msg_controllen = cmsg->cmsg_len;
	if (sendmsg(sock, &msg, 0) < 0) {
		ERROR_LOG("sendmsg: %m\n");
		return false;
	}
	return true;
}

static void dispose_fds(int *fds, unsigned num)
{
	int i;

	for (i = 0; i < num; i++)
		close(fds[i]);
}

/* Receive a buffer and the Unix credentials of the sending process. */
bool recv_cred_buffer(int socketfd, char *buf, size_t size,
		int *clientfd, uid_t *uid)
{
	char control[255] __attribute__((__aligned__(8)));
	struct msghdr msg;
	struct cmsghdr *cmsg;
	struct iovec iov;
	int yes = 1, cfd, ret;
	struct ucred cred;
	struct sockaddr_un sau;
	socklen_t sizeof_sau = sizeof(sau);

	ret = accept(socketfd, (struct sockaddr *)&sau, &sizeof_sau);
	if (ret < 0) {
		ERROR_LOG("accept: %m\n");
		return false;
	}
	cfd = ret;
	setsockopt(cfd, SOL_SOCKET, SO_PASSCRED, &yes, sizeof(int));
	memset(&msg, 0, sizeof(msg));
	iov.iov_base = buf;
	iov.iov_len = size;
	msg.msg_iov = &iov;
	msg.msg_iovlen = 1;
	msg.msg_control = control;
	msg.msg_controllen = sizeof(control);
	if (recvmsg(cfd, &msg, 0) < 0) {
		ERROR_LOG("recvmsg: %m\n");
		goto fail;
	}
	cmsg = CMSG_FIRSTHDR(&msg);
	while (cmsg) {
		if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type
				== SCM_CREDENTIALS) {
			memcpy(&cred, CMSG_DATA(cmsg), sizeof(struct ucred));
			*uid = cred.uid;
			*clientfd = cfd;
			return true;
		} else
			if (cmsg->cmsg_level == SOL_SOCKET
					&& cmsg->cmsg_type == SCM_RIGHTS) {
				dispose_fds((int *)CMSG_DATA(cmsg),
					(cmsg->cmsg_len - CMSG_LEN(0))
					/ sizeof(int));
			}
		cmsg = CMSG_NXTHDR(&msg, cmsg);
	}
fail:
	close(*clientfd);
	*clientfd = -1;
	return false;
}

bool pass_fd(int passfd, int socketfd)
{
	struct msghdr msg = {.msg_iov = NULL};
	struct cmsghdr *cmsg;
	char control[255] __attribute__((__aligned__(8)));
	struct iovec iov;
	char buf[] = "\0OK";

	iov.iov_base = buf;
	iov.iov_len  = sizeof(buf);

	msg.msg_iov = &iov;
	msg.msg_iovlen = 1;

	msg.msg_control = control;
	msg.msg_controllen = sizeof(control);

	cmsg = CMSG_FIRSTHDR(&msg);
	cmsg->cmsg_level = SOL_SOCKET;
	cmsg->cmsg_type = SCM_RIGHTS;
	cmsg->cmsg_len = CMSG_LEN(sizeof(int));
	*(int *)CMSG_DATA(cmsg) = passfd;

	/* Sum of the length of all control messages in the buffer */
	msg.msg_controllen = cmsg->cmsg_len;
	DEBUG_LOG("passing %s and fd %d\n", buf, passfd);
	if (sendmsg(socketfd, &msg, 0) < 0) {
		ERROR_LOG("sendmsg: %m\n");
		return false;
	}
	return true;
}

static bool recv_fd(int socketfd, int *recvfd)
{
	char control[255] __attribute__((__aligned__(8)));
	struct msghdr msg = {.msg_iov = NULL};
	struct cmsghdr *cmsg;
	struct iovec iov;
	char buf[100];
	ssize_t sz = sizeof(buf), ssz;

	*recvfd = -1;
	iov.iov_base = buf;
	iov.iov_len = sz - 1;
	msg.msg_iov = &iov;
	msg.msg_iovlen = 1;
	msg.msg_control = control;
	msg.msg_controllen = sizeof(control);
	memset(buf, 0, sz);
	ssz = recvmsg(socketfd, &msg, 0);
	if (ssz < 0) {
		ERROR_LOG("recvmsg: %m\n");
		return false;
	}
	buf[ssz] = '\0';
	INFO_LOG("server response: %u (%s)\n", (unsigned)buf[0], buf + 1);
	for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
		if (cmsg->cmsg_level != SOL_SOCKET
			|| cmsg->cmsg_type != SCM_RIGHTS)
			continue;
		if ((cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int) != 1)
			continue;
		*recvfd = *(int *)CMSG_DATA(cmsg);
		return true;
	}
	return false;
}

int request_fd(const char *socket_path, char *msg, int *result)
{
	struct sockaddr_un sau;
	int socketfd, receivefd;

	if (!init_unix_socket(socket_path, &socketfd, &sau))
		die("could not init socket");
	if (connect(socketfd, (struct sockaddr *)&sau, sizeof(sau)) < 0)
		die_errno("connect");
	if (!send_cred_buffer(socketfd, msg))
		die("could not send cred buffer");
	if (!recv_fd(socketfd, &receivefd))
		die("did not receive tty fd");
	NOTICE_LOG("received fd %d\n", receivefd);
	*result = receivefd;
	return socketfd;
}

bool request_int(const char *socket_path, char *msg, int *result)
{
	struct sockaddr_un sau;
	int socketfd;
	bool success = false;
	char buf[100];
	ssize_t ssz;

	*result = -1;
	if (!init_unix_socket(socket_path, &socketfd, &sau))
		return false;
	if (connect(socketfd, (struct sockaddr *)&sau, sizeof(sau)) < 0) {
		ERROR_LOG("connect: %m\n");
		goto close;
	}
	if (!send_cred_buffer(socketfd, msg)) {
		ERROR_LOG("could not send cred msg \"%s\"\n", msg);
		goto close;
	}
	ssz = read(socketfd, buf, sizeof(buf) - 1);
	if (ssz < 0) {
		ERROR_LOG("did not receive integer: %m\n");
		goto close;
	}
	if (buf[0] != 0) {
		ERROR_LOG("did not receive integer: %s\n", buf + 1);
		goto close;
	}
	if (ssz != sizeof(int) + 1) {
		ERROR_LOG("protocol mismatch, server msg: %s\n", buf + 1);
		goto close;
	}
	memcpy(result, buf + 1, sizeof(int));
	DEBUG_LOG("received integer: %d\n", *result);
	success = true;
close:
	close(socketfd);
	return success;
}

int signal_pipe[2];

static void signal_handler(int signum)
{
	uint8_t u = signum;
	int save_errno = errno;
	assert(signum > 0 && signum < 256);
	if (write(signal_pipe[1], &u, 1) < 0)
		ERROR_LOG("write to signal pipe: %m\n");
	errno = save_errno;
}

void init_signal_handling(void)
{
	struct sigaction act;

	if (pipe(signal_pipe) < 0)
		die_errno("signal pipe");
	act.sa_handler = signal_handler;
	sigemptyset(&act.sa_mask);
	act.sa_flags = SA_RESTART;
	if (sigaction(SIGINT, &act, NULL) < 0)
		die_errno("sigaction");
	if (sigaction(SIGTERM, &act, NULL) < 0)
		die_errno("sigaction");
	if (sigaction(SIGCHLD, &act, NULL) < 0)
		die_errno("sigaction");
	if (sigaction(SIGUSR1, &act, NULL) < 0)
		die_errno("sigaction");
}

int next_signal(void)
{
	uint8_t u = 0;
again:
	if (read(signal_pipe[0], &u, 1) < 0) {
		if (errno != EINTR)
			die_errno("read");
		goto again;
	}
	DEBUG_LOG("process %d received signal %u\n", getpid(), u);
	return u;
}