LinuxEthernetTap.cpp 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483
  1. /*
  2. * ZeroTier One - Network Virtualization Everywhere
  3. * Copyright (C) 2011-2016 ZeroTier, Inc. https://www.zerotier.com/
  4. *
  5. * This program is free software: you can redistribute it and/or modify
  6. * it under the terms of the GNU General Public License as published by
  7. * the Free Software Foundation, either version 3 of the License, or
  8. * (at your option) any later version.
  9. *
  10. * This program is distributed in the hope that it will be useful,
  11. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13. * GNU General Public License for more details.
  14. *
  15. * You should have received a copy of the GNU General Public License
  16. * along with this program. If not, see <http://www.gnu.org/licenses/>.
  17. */
  18. #include <stdint.h>
  19. #include <stdio.h>
  20. #include <stdlib.h>
  21. #include <string.h>
  22. #include <unistd.h>
  23. #include <signal.h>
  24. #include <fcntl.h>
  25. #include <errno.h>
  26. #include <sys/types.h>
  27. #include <sys/stat.h>
  28. #include <sys/ioctl.h>
  29. #include <sys/wait.h>
  30. #include <sys/select.h>
  31. #include <netinet/in.h>
  32. #include <net/if_arp.h>
  33. #include <arpa/inet.h>
  34. #include <linux/if.h>
  35. #include <linux/if_tun.h>
  36. #include <linux/if_addr.h>
  37. #include <linux/if_ether.h>
  38. #include <ifaddrs.h>
  39. #include <algorithm>
  40. #include <utility>
  41. #include "../node/Constants.hpp"
  42. #include "../node/Utils.hpp"
  43. #include "../node/Mutex.hpp"
  44. #include "../node/Dictionary.hpp"
  45. #include "OSUtils.hpp"
  46. #include "LinuxEthernetTap.hpp"
  47. // ff:ff:ff:ff:ff:ff with no ADI
  48. static const ZeroTier::MulticastGroup _blindWildcardMulticastGroup(ZeroTier::MAC(0xff),0);
  49. namespace ZeroTier {
  50. static Mutex __tapCreateLock;
  51. LinuxEthernetTap::LinuxEthernetTap(
  52. const char *homePath,
  53. const MAC &mac,
  54. unsigned int mtu,
  55. unsigned int metric,
  56. uint64_t nwid,
  57. const char *friendlyName,
  58. void (*handler)(void *,uint64_t,const MAC &,const MAC &,unsigned int,unsigned int,const void *,unsigned int),
  59. void *arg) :
  60. _handler(handler),
  61. _arg(arg),
  62. _nwid(nwid),
  63. _homePath(homePath),
  64. _mtu(mtu),
  65. _fd(0),
  66. _enabled(true)
  67. {
  68. char procpath[128],nwids[32];
  69. struct stat sbuf;
  70. Utils::snprintf(nwids,sizeof(nwids),"%.16llx",nwid);
  71. Mutex::Lock _l(__tapCreateLock); // create only one tap at a time, globally
  72. if (mtu > 2800)
  73. throw std::runtime_error("max tap MTU is 2800");
  74. _fd = ::open("/dev/net/tun",O_RDWR);
  75. if (_fd <= 0) {
  76. _fd = ::open("/dev/tun",O_RDWR);
  77. if (_fd <= 0)
  78. throw std::runtime_error(std::string("could not open TUN/TAP device: ") + strerror(errno));
  79. }
  80. struct ifreq ifr;
  81. memset(&ifr,0,sizeof(ifr));
  82. // Try to recall our last device name, or pick an unused one if that fails.
  83. bool recalledDevice = false;
  84. std::string devmapbuf;
  85. Dictionary<8194> devmap;
  86. if (OSUtils::readFile((_homePath + ZT_PATH_SEPARATOR_S + "devicemap").c_str(),devmapbuf)) {
  87. devmap.load(devmapbuf.c_str());
  88. char desiredDevice[128];
  89. if (devmap.get(nwids,desiredDevice,sizeof(desiredDevice)) > 0) {
  90. Utils::scopy(ifr.ifr_name,sizeof(ifr.ifr_name),desiredDevice);
  91. Utils::snprintf(procpath,sizeof(procpath),"/proc/sys/net/ipv4/conf/%s",ifr.ifr_name);
  92. recalledDevice = (stat(procpath,&sbuf) != 0);
  93. }
  94. }
  95. if (!recalledDevice) {
  96. int devno = 0;
  97. do {
  98. #ifdef __SYNOLOGY__
  99. devno+=50; // Arbitrary number to prevent interface name conflicts
  100. Utils::snprintf(ifr.ifr_name,sizeof(ifr.ifr_name),"eth%d",devno++);
  101. #else
  102. Utils::snprintf(ifr.ifr_name,sizeof(ifr.ifr_name),"zt%d",devno++);
  103. #endif
  104. Utils::snprintf(procpath,sizeof(procpath),"/proc/sys/net/ipv4/conf/%s",ifr.ifr_name);
  105. } while (stat(procpath,&sbuf) == 0); // try zt#++ until we find one that does not exist
  106. }
  107. ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
  108. if (ioctl(_fd,TUNSETIFF,(void *)&ifr) < 0) {
  109. ::close(_fd);
  110. throw std::runtime_error("unable to configure TUN/TAP device for TAP operation");
  111. }
  112. _dev = ifr.ifr_name;
  113. ::ioctl(_fd,TUNSETPERSIST,0); // valgrind may generate a false alarm here
  114. // Open an arbitrary socket to talk to netlink
  115. int sock = socket(AF_INET,SOCK_DGRAM,0);
  116. if (sock <= 0) {
  117. ::close(_fd);
  118. throw std::runtime_error("unable to open netlink socket");
  119. }
  120. // Set MAC address
  121. ifr.ifr_ifru.ifru_hwaddr.sa_family = ARPHRD_ETHER;
  122. mac.copyTo(ifr.ifr_ifru.ifru_hwaddr.sa_data,6);
  123. if (ioctl(sock,SIOCSIFHWADDR,(void *)&ifr) < 0) {
  124. ::close(_fd);
  125. ::close(sock);
  126. throw std::runtime_error("unable to configure TAP hardware (MAC) address");
  127. return;
  128. }
  129. // Set MTU
  130. ifr.ifr_ifru.ifru_mtu = (int)mtu;
  131. if (ioctl(sock,SIOCSIFMTU,(void *)&ifr) < 0) {
  132. ::close(_fd);
  133. ::close(sock);
  134. throw std::runtime_error("unable to configure TAP MTU");
  135. }
  136. if (fcntl(_fd,F_SETFL,fcntl(_fd,F_GETFL) & ~O_NONBLOCK) == -1) {
  137. ::close(_fd);
  138. throw std::runtime_error("unable to set flags on file descriptor for TAP device");
  139. }
  140. /* Bring interface up */
  141. if (ioctl(sock,SIOCGIFFLAGS,(void *)&ifr) < 0) {
  142. ::close(_fd);
  143. ::close(sock);
  144. throw std::runtime_error("unable to get TAP interface flags");
  145. }
  146. ifr.ifr_flags |= IFF_UP;
  147. if (ioctl(sock,SIOCSIFFLAGS,(void *)&ifr) < 0) {
  148. ::close(_fd);
  149. ::close(sock);
  150. throw std::runtime_error("unable to set TAP interface flags");
  151. }
  152. ::close(sock);
  153. // Set close-on-exec so that devices cannot persist if we fork/exec for update
  154. ::fcntl(_fd,F_SETFD,fcntl(_fd,F_GETFD) | FD_CLOEXEC);
  155. (void)::pipe(_shutdownSignalPipe);
  156. devmap.erase(nwids);
  157. devmap.add(nwids,_dev.c_str());
  158. OSUtils::writeFile((_homePath + ZT_PATH_SEPARATOR_S + "devicemap").c_str(),(const void *)devmap.data(),devmap.sizeBytes());
  159. _thread = Thread::start(this);
  160. }
  161. LinuxEthernetTap::~LinuxEthernetTap()
  162. {
  163. (void)::write(_shutdownSignalPipe[1],"\0",1); // causes thread to exit
  164. Thread::join(_thread);
  165. ::close(_fd);
  166. ::close(_shutdownSignalPipe[0]);
  167. ::close(_shutdownSignalPipe[1]);
  168. }
  169. void LinuxEthernetTap::setEnabled(bool en)
  170. {
  171. _enabled = en;
  172. }
  173. bool LinuxEthernetTap::enabled() const
  174. {
  175. return _enabled;
  176. }
  177. static bool ___removeIp(const std::string &_dev,const InetAddress &ip)
  178. {
  179. long cpid = (long)vfork();
  180. if (cpid == 0) {
  181. OSUtils::redirectUnixOutputs("/dev/null",(const char *)0);
  182. setenv("PATH", "/sbin:/bin:/usr/sbin:/usr/bin", 1);
  183. ::execlp("ip","ip","addr","del",ip.toString().c_str(),"dev",_dev.c_str(),(const char *)0);
  184. ::_exit(-1);
  185. } else {
  186. int exitcode = -1;
  187. ::waitpid(cpid,&exitcode,0);
  188. return (exitcode == 0);
  189. }
  190. }
  191. #ifdef __SYNOLOGY__
  192. bool LinuxEthernetTap::addIpSyn(std::vector<InetAddress> ips)
  193. {
  194. // Here we fill out interface config (ifcfg-dev) to prevent it from being killed
  195. std::string filepath = "/etc/sysconfig/network-scripts/ifcfg-"+_dev;
  196. std::string cfg_contents = "DEVICE="+_dev+"\nBOOTPROTO=static";
  197. int ip4=0,ip6=0,ip4_tot=0,ip6_tot=0;
  198. long cpid = (long)vfork();
  199. if (cpid == 0) {
  200. OSUtils::redirectUnixOutputs("/dev/null",(const char *)0);
  201. setenv("PATH", "/sbin:/bin:/usr/sbin:/usr/bin", 1);
  202. // We must know if there is at least (one) of each protocol version so we
  203. // can properly enumerate address/netmask combinations in the ifcfg-dev file
  204. for(int i=0; i<(int)ips.size(); i++) {
  205. if (ips[i].isV4())
  206. ip4_tot++;
  207. else
  208. ip6_tot++;
  209. }
  210. // Assemble and write contents of ifcfg-dev file
  211. for(int i=0; i<(int)ips.size(); i++) {
  212. if (ips[i].isV4()) {
  213. std::string numstr4 = ip4_tot > 1 ? std::to_string(ip4) : "";
  214. cfg_contents += "\nIPADDR"+numstr4+"="+ips[i].toIpString()
  215. + "\nNETMASK"+numstr4+"="+ips[i].netmask().toIpString()+"\n";
  216. ip4++;
  217. }
  218. else {
  219. std::string numstr6 = ip6_tot > 1 ? std::to_string(ip6) : "";
  220. cfg_contents += "\nIPV6ADDR"+numstr6+"="+ips[i].toIpString()
  221. + "\nNETMASK"+numstr6+"="+ips[i].netmask().toIpString()+"\n";
  222. ip6++;
  223. }
  224. }
  225. OSUtils::writeFile(filepath.c_str(), cfg_contents.c_str(), cfg_contents.length());
  226. // Finaly, add IPs
  227. for(int i=0; i<(int)ips.size(); i++){
  228. if (ips[i].isV4())
  229. ::execlp("ip","ip","addr","add",ips[i].toString().c_str(),"broadcast",ips[i].broadcast().toIpString().c_str(),"dev",_dev.c_str(),(const char *)0);
  230. else
  231. ::execlp("ip","ip","addr","add",ips[i].toString().c_str(),"dev",_dev.c_str(),(const char *)0);
  232. }
  233. ::_exit(-1);
  234. } else if (cpid > 0) {
  235. int exitcode = -1;
  236. ::waitpid(cpid,&exitcode,0);
  237. return (exitcode == 0);
  238. }
  239. return true;
  240. }
  241. #endif // __SYNOLOGY__
  242. bool LinuxEthernetTap::addIp(const InetAddress &ip)
  243. {
  244. if (!ip)
  245. return false;
  246. std::vector<InetAddress> allIps(ips());
  247. if (std::binary_search(allIps.begin(),allIps.end(),ip))
  248. return true;
  249. // Remove and reconfigure if address is the same but netmask is different
  250. for(std::vector<InetAddress>::iterator i(allIps.begin());i!=allIps.end();++i) {
  251. if (i->ipsEqual(ip))
  252. ___removeIp(_dev,*i);
  253. }
  254. long cpid = (long)vfork();
  255. if (cpid == 0) {
  256. OSUtils::redirectUnixOutputs("/dev/null",(const char *)0);
  257. setenv("PATH", "/sbin:/bin:/usr/sbin:/usr/bin", 1);
  258. if (ip.isV4()) {
  259. ::execlp("ip","ip","addr","add",ip.toString().c_str(),"broadcast",ip.broadcast().toIpString().c_str(),"dev",_dev.c_str(),(const char *)0);
  260. } else {
  261. ::execlp("ip","ip","addr","add",ip.toString().c_str(),"dev",_dev.c_str(),(const char *)0);
  262. }
  263. ::_exit(-1);
  264. } else if (cpid > 0) {
  265. int exitcode = -1;
  266. ::waitpid(cpid,&exitcode,0);
  267. return (exitcode == 0);
  268. }
  269. return false;
  270. }
  271. bool LinuxEthernetTap::removeIp(const InetAddress &ip)
  272. {
  273. if (!ip)
  274. return true;
  275. std::vector<InetAddress> allIps(ips());
  276. if (std::find(allIps.begin(),allIps.end(),ip) != allIps.end()) {
  277. if (___removeIp(_dev,ip))
  278. return true;
  279. }
  280. return false;
  281. }
  282. std::vector<InetAddress> LinuxEthernetTap::ips() const
  283. {
  284. struct ifaddrs *ifa = (struct ifaddrs *)0;
  285. if (getifaddrs(&ifa))
  286. return std::vector<InetAddress>();
  287. std::vector<InetAddress> r;
  288. struct ifaddrs *p = ifa;
  289. while (p) {
  290. if ((!strcmp(p->ifa_name,_dev.c_str()))&&(p->ifa_addr)&&(p->ifa_netmask)&&(p->ifa_addr->sa_family == p->ifa_netmask->sa_family)) {
  291. switch(p->ifa_addr->sa_family) {
  292. case AF_INET: {
  293. struct sockaddr_in *sin = (struct sockaddr_in *)p->ifa_addr;
  294. struct sockaddr_in *nm = (struct sockaddr_in *)p->ifa_netmask;
  295. r.push_back(InetAddress(&(sin->sin_addr.s_addr),4,Utils::countBits((uint32_t)nm->sin_addr.s_addr)));
  296. } break;
  297. case AF_INET6: {
  298. struct sockaddr_in6 *sin = (struct sockaddr_in6 *)p->ifa_addr;
  299. struct sockaddr_in6 *nm = (struct sockaddr_in6 *)p->ifa_netmask;
  300. uint32_t b[4];
  301. memcpy(b,nm->sin6_addr.s6_addr,sizeof(b));
  302. 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])));
  303. } break;
  304. }
  305. }
  306. p = p->ifa_next;
  307. }
  308. if (ifa)
  309. freeifaddrs(ifa);
  310. std::sort(r.begin(),r.end());
  311. r.erase(std::unique(r.begin(),r.end()),r.end());
  312. return r;
  313. }
  314. void LinuxEthernetTap::put(const MAC &from,const MAC &to,unsigned int etherType,const void *data,unsigned int len)
  315. {
  316. char putBuf[8194];
  317. if ((_fd > 0)&&(len <= _mtu)&&(_enabled)) {
  318. to.copyTo(putBuf,6);
  319. from.copyTo(putBuf + 6,6);
  320. *((uint16_t *)(putBuf + 12)) = htons((uint16_t)etherType);
  321. memcpy(putBuf + 14,data,len);
  322. len += 14;
  323. (void)::write(_fd,putBuf,len);
  324. }
  325. }
  326. std::string LinuxEthernetTap::deviceName() const
  327. {
  328. return _dev;
  329. }
  330. void LinuxEthernetTap::setFriendlyName(const char *friendlyName)
  331. {
  332. }
  333. void LinuxEthernetTap::scanMulticastGroups(std::vector<MulticastGroup> &added,std::vector<MulticastGroup> &removed)
  334. {
  335. char *ptr,*ptr2;
  336. unsigned char mac[6];
  337. std::vector<MulticastGroup> newGroups;
  338. int fd = ::open("/proc/net/dev_mcast",O_RDONLY);
  339. if (fd > 0) {
  340. char buf[131072];
  341. int n = (int)::read(fd,buf,sizeof(buf));
  342. if ((n > 0)&&(n < (int)sizeof(buf))) {
  343. buf[n] = (char)0;
  344. for(char *l=strtok_r(buf,"\r\n",&ptr);(l);l=strtok_r((char *)0,"\r\n",&ptr)) {
  345. int fno = 0;
  346. char *devname = (char *)0;
  347. char *mcastmac = (char *)0;
  348. for(char *f=strtok_r(l," \t",&ptr2);(f);f=strtok_r((char *)0," \t",&ptr2)) {
  349. if (fno == 1)
  350. devname = f;
  351. else if (fno == 4)
  352. mcastmac = f;
  353. ++fno;
  354. }
  355. if ((devname)&&(!strcmp(devname,_dev.c_str()))&&(mcastmac)&&(Utils::unhex(mcastmac,mac,6) == 6))
  356. newGroups.push_back(MulticastGroup(MAC(mac,6),0));
  357. }
  358. }
  359. ::close(fd);
  360. }
  361. std::vector<InetAddress> allIps(ips());
  362. for(std::vector<InetAddress>::iterator ip(allIps.begin());ip!=allIps.end();++ip)
  363. newGroups.push_back(MulticastGroup::deriveMulticastGroupForAddressResolution(*ip));
  364. std::sort(newGroups.begin(),newGroups.end());
  365. newGroups.erase(std::unique(newGroups.begin(),newGroups.end()),newGroups.end());
  366. for(std::vector<MulticastGroup>::iterator m(newGroups.begin());m!=newGroups.end();++m) {
  367. if (!std::binary_search(_multicastGroups.begin(),_multicastGroups.end(),*m))
  368. added.push_back(*m);
  369. }
  370. for(std::vector<MulticastGroup>::iterator m(_multicastGroups.begin());m!=_multicastGroups.end();++m) {
  371. if (!std::binary_search(newGroups.begin(),newGroups.end(),*m))
  372. removed.push_back(*m);
  373. }
  374. _multicastGroups.swap(newGroups);
  375. }
  376. void LinuxEthernetTap::threadMain()
  377. throw()
  378. {
  379. fd_set readfds,nullfds;
  380. MAC to,from;
  381. int n,nfds,r;
  382. char getBuf[8194];
  383. Thread::sleep(500);
  384. FD_ZERO(&readfds);
  385. FD_ZERO(&nullfds);
  386. nfds = (int)std::max(_shutdownSignalPipe[0],_fd) + 1;
  387. r = 0;
  388. for(;;) {
  389. FD_SET(_shutdownSignalPipe[0],&readfds);
  390. FD_SET(_fd,&readfds);
  391. select(nfds,&readfds,&nullfds,&nullfds,(struct timeval *)0);
  392. if (FD_ISSET(_shutdownSignalPipe[0],&readfds)) // writes to shutdown pipe terminate thread
  393. break;
  394. if (FD_ISSET(_fd,&readfds)) {
  395. n = (int)::read(_fd,getBuf + r,sizeof(getBuf) - r);
  396. if (n < 0) {
  397. if ((errno != EINTR)&&(errno != ETIMEDOUT))
  398. break;
  399. } else {
  400. // Some tap drivers like to send the ethernet frame and the
  401. // payload in two chunks, so handle that by accumulating
  402. // data until we have at least a frame.
  403. r += n;
  404. if (r > 14) {
  405. if (r > ((int)_mtu + 14)) // sanity check for weird TAP behavior on some platforms
  406. r = _mtu + 14;
  407. if (_enabled) {
  408. to.setTo(getBuf,6);
  409. from.setTo(getBuf + 6,6);
  410. unsigned int etherType = ntohs(((const uint16_t *)getBuf)[6]);
  411. // TODO: VLAN support
  412. _handler(_arg,_nwid,from,to,etherType,0,(const void *)(getBuf + 14),r - 14);
  413. }
  414. r = 0;
  415. }
  416. }
  417. }
  418. }
  419. }
  420. } // namespace ZeroTier