1 /* 2 * QEMU Bridge Helper 3 * 4 * Copyright IBM, Corp. 2011 5 * 6 * Authors: 7 * Anthony Liguori <aliguori@us.ibm.com> 8 * Richa Marwaha <rmarwah@linux.vnet.ibm.com> 9 * Corey Bryant <coreyb@linux.vnet.ibm.com> 10 * 11 * This work is licensed under the terms of the GNU GPL, version 2. See 12 * the COPYING file in the top-level directory. 13 * 14 */ 15 16 #include "qemu/osdep.h" 17 18 #include <glib.h> 19 20 #include <sys/ioctl.h> 21 #include <sys/socket.h> 22 #include <sys/un.h> 23 #include <sys/prctl.h> 24 25 #include <net/if.h> 26 27 #include <linux/sockios.h> 28 29 #ifndef SIOCBRADDIF 30 #include <linux/if_bridge.h> 31 #endif 32 33 #include "qemu/queue.h" 34 35 #include "net/tap-linux.h" 36 37 #ifdef CONFIG_LIBCAP 38 #include <cap-ng.h> 39 #endif 40 41 #define DEFAULT_ACL_FILE CONFIG_QEMU_CONFDIR "/bridge.conf" 42 43 enum { 44 ACL_ALLOW = 0, 45 ACL_ALLOW_ALL, 46 ACL_DENY, 47 ACL_DENY_ALL, 48 }; 49 50 typedef struct ACLRule { 51 int type; 52 char iface[IFNAMSIZ]; 53 QSIMPLEQ_ENTRY(ACLRule) entry; 54 } ACLRule; 55 56 typedef QSIMPLEQ_HEAD(ACLList, ACLRule) ACLList; 57 58 static void usage(void) 59 { 60 fprintf(stderr, 61 "Usage: qemu-bridge-helper [--use-vnet] --br=bridge --fd=unixfd\n"); 62 } 63 64 static int parse_acl_file(const char *filename, ACLList *acl_list) 65 { 66 FILE *f; 67 char line[4096]; 68 ACLRule *acl_rule; 69 70 f = fopen(filename, "r"); 71 if (f == NULL) { 72 return -1; 73 } 74 75 while (fgets(line, sizeof(line), f) != NULL) { 76 char *ptr = line; 77 char *cmd, *arg, *argend; 78 79 while (isspace(*ptr)) { 80 ptr++; 81 } 82 83 /* skip comments and empty lines */ 84 if (*ptr == '#' || *ptr == 0) { 85 continue; 86 } 87 88 cmd = ptr; 89 arg = strchr(cmd, ' '); 90 if (arg == NULL) { 91 arg = strchr(cmd, '\t'); 92 } 93 94 if (arg == NULL) { 95 fprintf(stderr, "Invalid config line:\n %s\n", line); 96 fclose(f); 97 errno = EINVAL; 98 return -1; 99 } 100 101 *arg = 0; 102 arg++; 103 while (isspace(*arg)) { 104 arg++; 105 } 106 107 argend = arg + strlen(arg); 108 while (arg != argend && isspace(*(argend - 1))) { 109 argend--; 110 } 111 *argend = 0; 112 113 if (strcmp(cmd, "deny") == 0) { 114 acl_rule = g_malloc(sizeof(*acl_rule)); 115 if (strcmp(arg, "all") == 0) { 116 acl_rule->type = ACL_DENY_ALL; 117 } else { 118 acl_rule->type = ACL_DENY; 119 snprintf(acl_rule->iface, IFNAMSIZ, "%s", arg); 120 } 121 QSIMPLEQ_INSERT_TAIL(acl_list, acl_rule, entry); 122 } else if (strcmp(cmd, "allow") == 0) { 123 acl_rule = g_malloc(sizeof(*acl_rule)); 124 if (strcmp(arg, "all") == 0) { 125 acl_rule->type = ACL_ALLOW_ALL; 126 } else { 127 acl_rule->type = ACL_ALLOW; 128 snprintf(acl_rule->iface, IFNAMSIZ, "%s", arg); 129 } 130 QSIMPLEQ_INSERT_TAIL(acl_list, acl_rule, entry); 131 } else if (strcmp(cmd, "include") == 0) { 132 /* ignore errors */ 133 parse_acl_file(arg, acl_list); 134 } else { 135 fprintf(stderr, "Unknown command `%s'\n", cmd); 136 fclose(f); 137 errno = EINVAL; 138 return -1; 139 } 140 } 141 142 fclose(f); 143 144 return 0; 145 } 146 147 static bool has_vnet_hdr(int fd) 148 { 149 unsigned int features = 0; 150 151 if (ioctl(fd, TUNGETFEATURES, &features) == -1) { 152 return false; 153 } 154 155 if (!(features & IFF_VNET_HDR)) { 156 return false; 157 } 158 159 return true; 160 } 161 162 static void prep_ifreq(struct ifreq *ifr, const char *ifname) 163 { 164 memset(ifr, 0, sizeof(*ifr)); 165 snprintf(ifr->ifr_name, IFNAMSIZ, "%s", ifname); 166 } 167 168 static int send_fd(int c, int fd) 169 { 170 char msgbuf[CMSG_SPACE(sizeof(fd))]; 171 struct msghdr msg = { 172 .msg_control = msgbuf, 173 .msg_controllen = sizeof(msgbuf), 174 }; 175 struct cmsghdr *cmsg; 176 struct iovec iov; 177 char req[1] = { 0x00 }; 178 179 cmsg = CMSG_FIRSTHDR(&msg); 180 cmsg->cmsg_level = SOL_SOCKET; 181 cmsg->cmsg_type = SCM_RIGHTS; 182 cmsg->cmsg_len = CMSG_LEN(sizeof(fd)); 183 msg.msg_controllen = cmsg->cmsg_len; 184 185 iov.iov_base = req; 186 iov.iov_len = sizeof(req); 187 188 msg.msg_iov = &iov; 189 msg.msg_iovlen = 1; 190 memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd)); 191 192 return sendmsg(c, &msg, 0); 193 } 194 195 #ifdef CONFIG_LIBCAP 196 static int drop_privileges(void) 197 { 198 /* clear all capabilities */ 199 capng_clear(CAPNG_SELECT_BOTH); 200 201 if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED, 202 CAP_NET_ADMIN) < 0) { 203 return -1; 204 } 205 206 /* change to calling user's real uid and gid, retaining supplemental 207 * groups and CAP_NET_ADMIN */ 208 if (capng_change_id(getuid(), getgid(), CAPNG_CLEAR_BOUNDING)) { 209 return -1; 210 } 211 212 return 0; 213 } 214 #endif 215 216 int main(int argc, char **argv) 217 { 218 struct ifreq ifr; 219 #ifndef SIOCBRADDIF 220 unsigned long ifargs[4]; 221 #endif 222 int ifindex; 223 int fd = -1, ctlfd = -1, unixfd = -1; 224 int use_vnet = 0; 225 int mtu; 226 const char *bridge = NULL; 227 char iface[IFNAMSIZ]; 228 int index; 229 ACLRule *acl_rule; 230 ACLList acl_list; 231 int access_allowed, access_denied; 232 int ret = EXIT_SUCCESS; 233 234 #ifdef CONFIG_LIBCAP 235 /* if we're run from an suid binary, immediately drop privileges preserving 236 * cap_net_admin */ 237 if (geteuid() == 0 && getuid() != geteuid()) { 238 if (drop_privileges() == -1) { 239 fprintf(stderr, "failed to drop privileges\n"); 240 return 1; 241 } 242 } 243 #endif 244 245 /* parse arguments */ 246 for (index = 1; index < argc; index++) { 247 if (strcmp(argv[index], "--use-vnet") == 0) { 248 use_vnet = 1; 249 } else if (strncmp(argv[index], "--br=", 5) == 0) { 250 bridge = &argv[index][5]; 251 } else if (strncmp(argv[index], "--fd=", 5) == 0) { 252 unixfd = atoi(&argv[index][5]); 253 } else { 254 usage(); 255 return EXIT_FAILURE; 256 } 257 } 258 259 if (bridge == NULL || unixfd == -1) { 260 usage(); 261 return EXIT_FAILURE; 262 } 263 264 /* parse default acl file */ 265 QSIMPLEQ_INIT(&acl_list); 266 if (parse_acl_file(DEFAULT_ACL_FILE, &acl_list) == -1) { 267 fprintf(stderr, "failed to parse default acl file `%s'\n", 268 DEFAULT_ACL_FILE); 269 ret = EXIT_FAILURE; 270 goto cleanup; 271 } 272 273 /* validate bridge against acl -- default policy is to deny 274 * according acl policy if we have a deny and allow both 275 * then deny should always win over allow 276 */ 277 access_allowed = 0; 278 access_denied = 0; 279 QSIMPLEQ_FOREACH(acl_rule, &acl_list, entry) { 280 switch (acl_rule->type) { 281 case ACL_ALLOW_ALL: 282 access_allowed = 1; 283 break; 284 case ACL_ALLOW: 285 if (strcmp(bridge, acl_rule->iface) == 0) { 286 access_allowed = 1; 287 } 288 break; 289 case ACL_DENY_ALL: 290 access_denied = 1; 291 break; 292 case ACL_DENY: 293 if (strcmp(bridge, acl_rule->iface) == 0) { 294 access_denied = 1; 295 } 296 break; 297 } 298 } 299 300 if ((access_allowed == 0) || (access_denied == 1)) { 301 fprintf(stderr, "access denied by acl file\n"); 302 ret = EXIT_FAILURE; 303 goto cleanup; 304 } 305 306 /* open a socket to use to control the network interfaces */ 307 ctlfd = socket(AF_INET, SOCK_STREAM, 0); 308 if (ctlfd == -1) { 309 fprintf(stderr, "failed to open control socket: %s\n", strerror(errno)); 310 ret = EXIT_FAILURE; 311 goto cleanup; 312 } 313 314 /* open the tap device */ 315 fd = open("/dev/net/tun", O_RDWR); 316 if (fd == -1) { 317 fprintf(stderr, "failed to open /dev/net/tun: %s\n", strerror(errno)); 318 ret = EXIT_FAILURE; 319 goto cleanup; 320 } 321 322 /* request a tap device, disable PI, and add vnet header support if 323 * requested and it's available. */ 324 prep_ifreq(&ifr, "tap%d"); 325 ifr.ifr_flags = IFF_TAP|IFF_NO_PI; 326 if (use_vnet && has_vnet_hdr(fd)) { 327 ifr.ifr_flags |= IFF_VNET_HDR; 328 } 329 330 if (ioctl(fd, TUNSETIFF, &ifr) == -1) { 331 fprintf(stderr, "failed to create tun device: %s\n", strerror(errno)); 332 ret = EXIT_FAILURE; 333 goto cleanup; 334 } 335 336 /* save tap device name */ 337 snprintf(iface, sizeof(iface), "%s", ifr.ifr_name); 338 339 /* get the mtu of the bridge */ 340 prep_ifreq(&ifr, bridge); 341 if (ioctl(ctlfd, SIOCGIFMTU, &ifr) == -1) { 342 fprintf(stderr, "failed to get mtu of bridge `%s': %s\n", 343 bridge, strerror(errno)); 344 ret = EXIT_FAILURE; 345 goto cleanup; 346 } 347 348 /* save mtu */ 349 mtu = ifr.ifr_mtu; 350 351 /* set the mtu of the interface based on the bridge */ 352 prep_ifreq(&ifr, iface); 353 ifr.ifr_mtu = mtu; 354 if (ioctl(ctlfd, SIOCSIFMTU, &ifr) == -1) { 355 fprintf(stderr, "failed to set mtu of device `%s' to %d: %s\n", 356 iface, mtu, strerror(errno)); 357 ret = EXIT_FAILURE; 358 goto cleanup; 359 } 360 361 /* Linux uses the lowest enslaved MAC address as the MAC address of 362 * the bridge. Set MAC address to a high value so that it doesn't 363 * affect the MAC address of the bridge. 364 */ 365 if (ioctl(ctlfd, SIOCGIFHWADDR, &ifr) < 0) { 366 fprintf(stderr, "failed to get MAC address of device `%s': %s\n", 367 iface, strerror(errno)); 368 ret = EXIT_FAILURE; 369 goto cleanup; 370 } 371 ifr.ifr_hwaddr.sa_data[0] = 0xFE; 372 if (ioctl(ctlfd, SIOCSIFHWADDR, &ifr) < 0) { 373 fprintf(stderr, "failed to set MAC address of device `%s': %s\n", 374 iface, strerror(errno)); 375 ret = EXIT_FAILURE; 376 goto cleanup; 377 } 378 379 /* add the interface to the bridge */ 380 prep_ifreq(&ifr, bridge); 381 ifindex = if_nametoindex(iface); 382 #ifndef SIOCBRADDIF 383 ifargs[0] = BRCTL_ADD_IF; 384 ifargs[1] = ifindex; 385 ifargs[2] = 0; 386 ifargs[3] = 0; 387 ifr.ifr_data = (void *)ifargs; 388 ret = ioctl(ctlfd, SIOCDEVPRIVATE, &ifr); 389 #else 390 ifr.ifr_ifindex = ifindex; 391 ret = ioctl(ctlfd, SIOCBRADDIF, &ifr); 392 #endif 393 if (ret == -1) { 394 fprintf(stderr, "failed to add interface `%s' to bridge `%s': %s\n", 395 iface, bridge, strerror(errno)); 396 ret = EXIT_FAILURE; 397 goto cleanup; 398 } 399 400 /* bring the interface up */ 401 prep_ifreq(&ifr, iface); 402 if (ioctl(ctlfd, SIOCGIFFLAGS, &ifr) == -1) { 403 fprintf(stderr, "failed to get interface flags for `%s': %s\n", 404 iface, strerror(errno)); 405 ret = EXIT_FAILURE; 406 goto cleanup; 407 } 408 409 ifr.ifr_flags |= IFF_UP; 410 if (ioctl(ctlfd, SIOCSIFFLAGS, &ifr) == -1) { 411 fprintf(stderr, "failed to bring up interface `%s': %s\n", 412 iface, strerror(errno)); 413 ret = EXIT_FAILURE; 414 goto cleanup; 415 } 416 417 /* write fd to the domain socket */ 418 if (send_fd(unixfd, fd) == -1) { 419 fprintf(stderr, "failed to write fd to unix socket: %s\n", 420 strerror(errno)); 421 ret = EXIT_FAILURE; 422 goto cleanup; 423 } 424 425 /* ... */ 426 427 /* profit! */ 428 429 cleanup: 430 if (fd >= 0) { 431 close(fd); 432 } 433 if (ctlfd >= 0) { 434 close(ctlfd); 435 } 436 while ((acl_rule = QSIMPLEQ_FIRST(&acl_list)) != NULL) { 437 QSIMPLEQ_REMOVE_HEAD(&acl_list, entry); 438 g_free(acl_rule); 439 } 440 441 return ret; 442 } 443