ZeroTierOne/osdep/NetBSDEthernetTap.cpp

/*
 * Copyright (c)2019 ZeroTier, Inc.
 *
 * Use of this software is governed by the Business Source License included
 * in the LICENSE.TXT file in the project's root directory.
 *
 * Change Date: 2026-01-01
 *
 * On the date above, in accordance with the Business Source License, use
 * of this software will be governed by version 2.0 of the Apache License.
 */
/****/

#include "NetBSDEthernetTap.hpp"

#include "../node/Constants.hpp"
#include "../node/Mutex.hpp"
#include "../node/Utils.hpp"
#include "OSUtils.hpp"
#include "freebsd_getifmaddrs.h"

#include <algorithm>
#include <arpa/inet.h>
#include <errno.h>
#include <fcntl.h>
#include <ifaddrs.h>
#include <iostream>
#include <map>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/route.h>
#include <netinet/in.h>
#include <set>
#include <signal.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <string>
#include <sys/cdefs.h>
#include <sys/ioctl.h>
#include <sys/param.h>
#include <sys/select.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/sysctl.h>
#include <sys/types.h>
#include <sys/uio.h>
#include <sys/wait.h>
#include <unistd.h>
#include <utility>

#include "../node/Constants.hpp"
#include "../node/Utils.hpp"
#include "../node/Mutex.hpp"
#include "OSUtils.hpp"
#include "NetBSDEthernetTap.hpp"
#include "../node/Metrics.hpp"

#include <iostream>
using namespace std;
#define ZT_BASE32_CHARS "0123456789abcdefghijklmnopqrstuv"

// ff:ff:ff:ff:ff:ff with no ADI
static const ZeroTier::MulticastGroup _blindWildcardMulticastGroup(ZeroTier::MAC(0xff), 0);

namespace ZeroTier {

NetBSDEthernetTap::NetBSDEthernetTap(
	const char* homePath,
	const MAC& mac,
	unsigned int mtu,
	unsigned int metric,
	uint64_t nwid,
	const char* friendlyName,
	void (*handler)(void*, uint64_t, const MAC&, const MAC&, unsigned int, unsigned int, const void*, unsigned int),
	void* arg)
	: _handler(handler)
	, _arg(arg)
	, _nwid(nwid)
	, _mtu(mtu)
	, _metric(metric)
	, _fd(0)
	, _enabled(true)
{
	static Mutex globalTapCreateLock;
	char devpath[64], ethaddr[64], mtustr[32], metstr[32], tmpdevname[32];
	struct stat stattmp;

	Mutex::Lock _gl(globalTapCreateLock);

	if (mtu > 2800)
		throw std::runtime_error("max tap MTU is 2800");

	// we can create /dev/tap*
	std::vector<std::string> devFiles(OSUtils::listDirectory("/dev"));
	for (int i = 9993; i < (9993 + 128); ++i) {
		Utils::snprintf(tmpdevname, sizeof(tmpdevname), "tap%d", i);
		Utils::snprintf(devpath, sizeof(devpath), "/dev/%s", tmpdevname);
		if (std::find(devFiles.begin(), devFiles.end(), std::string(tmpdevname)) == devFiles.end()) {
			long cpid = (long)vfork();
			if (cpid == 0) {
				::execl("/sbin/ifconfig", "/sbin/ifconfig", tmpdevname, "create", (const char*)0);
				::_exit(-1);
			}
			else if (cpid > 0) {
				int exitcode = -1;
				::waitpid(cpid, &exitcode, 0);
			}
			else
				throw std::runtime_error("fork() failed");

			cpid = (long)vfork();
			if (cpid == 0) {
				string tmp;
				sprintf((char*)tmp.c_str(), "%d", i);
				string minor = tmp.c_str();
				::execl("/sbin/mknod", "/sbin/mknod", devpath, "c", "169", minor.c_str(), (const char*)0);
				// http://ftp.netbsd.org/pub/NetBSD/NetBSD-current/src/sys/conf/majors
				// major 169 => tap
				::_exit(-1);
			}
			else if (cpid > 0) {
				int exitcode = -1;
				::waitpid(cpid, &exitcode, 0);
			}
			else
				throw std::runtime_error("fork() failed");

			cerr << "created device " << devpath << endl;

			_dev = tmpdevname;
			_fd = ::open(devpath, O_RDWR);
			if (! stat(devpath, &stattmp)) {
				if (_fd > 0)
					break;
				else
					throw std::runtime_error("unable to open created tap device ");
			}
			else {
				throw std::runtime_error("cannot find /dev node for newly created tap device");
			}
		}
	}

	if (_fd <= 0)
		throw std::runtime_error("unable to open TAP device or no more devices available");

	if (fcntl(_fd, F_SETFL, fcntl(_fd, F_GETFL) & ~O_NONBLOCK) == -1) {
		::close(_fd);
		throw std::runtime_error("unable to set flags on file descriptor for TAP device");
	}

	// Configure MAC address and MTU, bring interface up
	Utils::snprintf(ethaddr, sizeof(ethaddr), "%.2x:%.2x:%.2x:%.2x:%.2x:%.2x", (int)mac[0], (int)mac[1], (int)mac[2], (int)mac[3], (int)mac[4], (int)mac[5]);
	Utils::snprintf(mtustr, sizeof(mtustr), "%u", _mtu);
	Utils::snprintf(metstr, sizeof(metstr), "%u", _metric);
	long cpid = (long)vfork();
	if (cpid == 0) {
		::execl("/sbin/ifconfig", "/sbin/ifconfig", _dev.c_str(), "link", ethaddr, "mtu", mtustr, "metric", metstr, "up", (const char*)0);
		::_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		::waitpid(cpid, &exitcode, 0);
		if (exitcode) {
			::close(_fd);
			throw std::runtime_error("ifconfig failure setting link-layer address and activating tap interface");
		}
	}

	// ifconfig link seems to be different from address
	// https://wiki.netbsd.org/tutorials/faking_a_mac_address/
	cpid = (long)vfork();
	if (cpid == 0) {
		string cmdline = "net.link.tap." + string(_dev);
		cmdline += "=" + string(ethaddr);
		::execl("/sbin/sysctl", "/sbin/sysctl", "-w", cmdline.c_str(), (const char*)0);
		::_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		::waitpid(cpid, &exitcode, 0);
		if (exitcode) {
			::close(_fd);
			throw std::runtime_error("sysctl failure setting link-layer address and activating tap interface");
		}
	}

	// Set close-on-exec so that devices cannot persist if we fork/exec for update
	fcntl(_fd, F_SETFD, fcntl(_fd, F_GETFD) | FD_CLOEXEC);

	::pipe(_shutdownSignalPipe);

	_thread = Thread::start(this);
}

NetBSDEthernetTap::~NetBSDEthernetTap()
{
	::write(_shutdownSignalPipe[1], "\0", 1);	// causes thread to exit
	Thread::join(_thread);
	::close(_fd);
	::close(_shutdownSignalPipe[0]);
	::close(_shutdownSignalPipe[1]);

	long cpid = (long)vfork();
	if (cpid == 0) {
		::execl("/sbin/ifconfig", "/sbin/ifconfig", _dev.c_str(), "destroy", (const char*)0);
		::_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		::waitpid(cpid, &exitcode, 0);
	}

	cpid = (long)vfork();
	if (cpid == 0) {
		string tmp = "/dev/";
		tmp += _dev.c_str();
		::execl("/bin/rm", "/bin/rm", tmp.c_str(), (const char*)0);
		::_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		::waitpid(cpid, &exitcode, 0);
	}
	else
		throw std::runtime_error("fork() failed");
}

void NetBSDEthernetTap::setEnabled(bool en)
{
	_enabled = en;
}

bool NetBSDEthernetTap::enabled() const
{
	return _enabled;
}

static bool ___removeIp(const std::string& _dev, const InetAddress& ip)
{
	long cpid = (long)vfork();
	if (cpid == 0) {
		execl("/sbin/ifconfig", "/sbin/ifconfig", _dev.c_str(), "inet", ip.toIpString().c_str(), "-alias", (const char*)0);
		_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		waitpid(cpid, &exitcode, 0);
		return (exitcode == 0);
	}
	return false;	// never reached, make compiler shut up about return value
}

bool NetBSDEthernetTap::addIp(const InetAddress& ip)
{
	if (! ip)
		return false;

	std::vector<InetAddress> allIps(ips());
	if (std::find(allIps.begin(), allIps.end(), ip) != allIps.end())
		return true;   // IP/netmask already assigned

	// Remove and reconfigure if address is the same but netmask is different
	for (std::vector<InetAddress>::iterator i(allIps.begin()); i != allIps.end(); ++i) {
		if ((i->ipsEqual(ip)) && (i->netmaskBits() != ip.netmaskBits())) {
			if (___removeIp(_dev, *i))
				break;
		}
	}

	long cpid = (long)vfork();
	if (cpid == 0) {
		::execl("/sbin/ifconfig", "/sbin/ifconfig", _dev.c_str(), ip.isV4() ? "inet" : "inet6", ip.toString().c_str(), "alias", (const char*)0);
		::_exit(-1);
	}
	else if (cpid > 0) {
		int exitcode = -1;
		::waitpid(cpid, &exitcode, 0);
		return (exitcode == 0);
	}
	return false;
}

bool NetBSDEthernetTap::removeIp(const InetAddress& ip)
{
	if (! ip)
		return false;
	std::vector<InetAddress> allIps(ips());
	if (std::find(allIps.begin(), allIps.end(), ip) != allIps.end()) {
		if (___removeIp(_dev, ip))
			return true;
	}
	return false;
}

std::vector<InetAddress> NetBSDEthernetTap::ips() const
{
	struct ifaddrs* ifa = (struct ifaddrs*)0;
	if (getifaddrs(&ifa))
		return std::vector<InetAddress>();

	std::vector<InetAddress> r;

	struct ifaddrs* p = ifa;
	while (p) {
		if ((! strcmp(p->ifa_name, _dev.c_str())) && (p->ifa_addr) && (p->ifa_netmask) && (p->ifa_addr->sa_family == p->ifa_netmask->sa_family)) {
			switch (p->ifa_addr->sa_family) {
				case AF_INET: {
					struct sockaddr_in* sin = (struct sockaddr_in*)p->ifa_addr;
					struct sockaddr_in* nm = (struct sockaddr_in*)p->ifa_netmask;
					r.push_back(InetAddress(&(sin->sin_addr.s_addr), 4, Utils::countBits((uint32_t)nm->sin_addr.s_addr)));
				} break;
				case AF_INET6: {
					struct sockaddr_in6* sin = (struct sockaddr_in6*)p->ifa_addr;
					struct sockaddr_in6* nm = (struct sockaddr_in6*)p->ifa_netmask;
					uint32_t b[4];
					memcpy(b, nm->sin6_addr.s6_addr, sizeof(b));
					r.push_back(InetAddress(sin->sin6_addr.s6_addr, 16, Utils::countBits(b[0]) + Utils::countBits(b[1]) + Utils::countBits(b[2]) + Utils::countBits(b[3])));
				} break;
			}
		}
		p = p->ifa_next;
	}

	if (ifa)
		freeifaddrs(ifa);

	std::sort(r.begin(), r.end());
	std::unique(r.begin(), r.end());

	return r;
}

void NetBSDEthernetTap::put(const MAC& from, const MAC& to, unsigned int etherType, const void* data, unsigned int len)
{
	// VL2 frame size histogram
	Metrics::vl2_frame_size_hist.Observe(len);

	if (len > this->_mtu) {
		Metrics::vl2_would_fragment_or_drop_rx++;
	}
	char putBuf[4096];
	if ((_fd > 0) && (len <= _mtu) && (_enabled)) {
		to.copyTo(putBuf, 6);
		from.copyTo(putBuf + 6, 6);
		*((uint16_t*)(putBuf + 12)) = htons((uint16_t)etherType);
		memcpy(putBuf + 14, data, len);
		len += 14;
		::write(_fd, putBuf, len);
	}
}

std::string NetBSDEthernetTap::deviceName() const
{
	return _dev;
}

void NetBSDEthernetTap::setFriendlyName(const char* friendlyName)
{
}

void NetBSDEthernetTap::scanMulticastGroups(std::vector<MulticastGroup>& added, std::vector<MulticastGroup>& removed)
{
	std::vector<MulticastGroup> newGroups;

	struct ifmaddrs* ifmap = (struct ifmaddrs*)0;
	if (! getifmaddrs(&ifmap)) {
		struct ifmaddrs* p = ifmap;
		while (p) {
			if (p->ifma_addr->sa_family == AF_LINK) {
				struct sockaddr_dl* in = (struct sockaddr_dl*)p->ifma_name;
				struct sockaddr_dl* la = (struct sockaddr_dl*)p->ifma_addr;
				if ((la->sdl_alen == 6) && (in->sdl_nlen <= _dev.length()) && (! memcmp(_dev.data(), in->sdl_data, in->sdl_nlen)))
					newGroups.push_back(MulticastGroup(MAC(la->sdl_data + la->sdl_nlen, 6), 0));
			}
			p = p->ifma_next;
		}
		freeifmaddrs(ifmap);
	}

	std::vector<InetAddress> allIps(ips());
	for (std::vector<InetAddress>::iterator ip(allIps.begin()); ip != allIps.end(); ++ip)
		newGroups.push_back(MulticastGroup::deriveMulticastGroupForAddressResolution(*ip));

	std::sort(newGroups.begin(), newGroups.end());
	std::unique(newGroups.begin(), newGroups.end());

	for (std::vector<MulticastGroup>::iterator m(newGroups.begin()); m != newGroups.end(); ++m) {
		if (! std::binary_search(_multicastGroups.begin(), _multicastGroups.end(), *m))
			added.push_back(*m);
	}
	for (std::vector<MulticastGroup>::iterator m(_multicastGroups.begin()); m != _multicastGroups.end(); ++m) {
		if (! std::binary_search(newGroups.begin(), newGroups.end(), *m))
			removed.push_back(*m);
	}

	_multicastGroups.swap(newGroups);
}

/*
bool NetBSDEthernetTap::updateMulticastGroups(std::set<MulticastGroup> &groups)
{
	std::set<MulticastGroup> newGroups;
	struct ifmaddrs *ifmap = (struct ifmaddrs *)0;
	if (!getifmaddrs(&ifmap)) {
		struct ifmaddrs *p = ifmap;
		while (p) {
			if (p->ifma_addr->sa_family == AF_LINK) {
				struct sockaddr_dl *in = (struct sockaddr_dl *)p->ifma_name;
				struct sockaddr_dl *la = (struct sockaddr_dl *)p->ifma_addr;
				if ((la->sdl_alen == 6)&&(in->sdl_nlen <= _dev.length())&&(!memcmp(_dev.data(),in->sdl_data,in->sdl_nlen)))
					newGroups.insert(MulticastGroup(MAC(la->sdl_data + la->sdl_nlen,6),0));
			}
			p = p->ifma_next;
		}
		freeifmaddrs(ifmap);
	}

	{
		std::set<InetAddress> allIps(ips());
		for(std::set<InetAddress>::const_iterator i(allIps.begin());i!=allIps.end();++i)
			newGroups.insert(MulticastGroup::deriveMulticastGroupForAddressResolution(*i));
	}

	bool changed = false;

	for(std::set<MulticastGroup>::iterator mg(newGroups.begin());mg!=newGroups.end();++mg) {
		if (!groups.count(*mg)) {
			groups.insert(*mg);
			changed = true;
		}
	}
	for(std::set<MulticastGroup>::iterator mg(groups.begin());mg!=groups.end();) {
		if ((!newGroups.count(*mg))&&(*mg != _blindWildcardMulticastGroup)) {
			groups.erase(mg++);
			changed = true;
		} else ++mg;
	}

	return changed;
}
*/

void NetBSDEthernetTap::threadMain() throw()
{
	fd_set readfds, nullfds;
	MAC to, from;
	int n, nfds, r;
	char getBuf[8194];

	// Wait for a moment after startup -- wait for Network to finish
	// constructing itself.
	Thread::sleep(500);

	FD_ZERO(&readfds);
	FD_ZERO(&nullfds);
	nfds = (int)std::max(_shutdownSignalPipe[0], _fd) + 1;

	r = 0;
	for (;;) {
		FD_SET(_shutdownSignalPipe[0], &readfds);
		FD_SET(_fd, &readfds);
		select(nfds, &readfds, &nullfds, &nullfds, (struct timeval*)0);

		if (FD_ISSET(_shutdownSignalPipe[0], &readfds))	  // writes to shutdown pipe terminate thread
			break;

		if (FD_ISSET(_fd, &readfds)) {
			n = (int)::read(_fd, getBuf + r, sizeof(getBuf) - r);
			if (n < 0) {
				if ((errno != EINTR) && (errno != ETIMEDOUT))
					break;
			}
			else {
				// Some tap drivers like to send the ethernet frame and the
				// payload in two chunks, so handle that by accumulating
				// data until we have at least a frame.
				r += n;
				if (r > 14) {
					if (r > ((int)_mtu + 14))	// sanity check for weird TAP behavior on some platforms
						r = _mtu + 14;

					if (_enabled) {
						to.setTo(getBuf, 6);
						from.setTo(getBuf + 6, 6);
						unsigned int etherType = ntohs(((const uint16_t*)getBuf)[6]);
						// TODO: VLAN support
						_handler(_arg, _nwid, from, to, etherType, 0, (const void*)(getBuf + 14), r - 14);
					}

					r = 0;
				}
			}
		}
	}
}

}	// namespace ZeroTier