1 /* 2 * QEMU System Emulator 3 * 4 * Copyright (c) 2003-2008 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 #include "qemu/osdep.h" 25 26 #include "net/net.h" 27 #include "clients.h" 28 #include "monitor/monitor.h" 29 #include "qapi/error.h" 30 #include "qemu-common.h" 31 #include "qemu/error-report.h" 32 #include "qemu/option.h" 33 #include "qemu/sockets.h" 34 #include "qemu/iov.h" 35 #include "qemu/main-loop.h" 36 37 typedef struct NetSocketState { 38 NetClientState nc; 39 int listen_fd; 40 int fd; 41 SocketReadState rs; 42 unsigned int send_index; /* number of bytes sent (only SOCK_STREAM) */ 43 struct sockaddr_in dgram_dst; /* contains inet host and port destination iff connectionless (SOCK_DGRAM) */ 44 IOHandler *send_fn; /* differs between SOCK_STREAM/SOCK_DGRAM */ 45 bool read_poll; /* waiting to receive data? */ 46 bool write_poll; /* waiting to transmit data? */ 47 } NetSocketState; 48 49 static void net_socket_accept(void *opaque); 50 static void net_socket_writable(void *opaque); 51 52 static void net_socket_update_fd_handler(NetSocketState *s) 53 { 54 qemu_set_fd_handler(s->fd, 55 s->read_poll ? s->send_fn : NULL, 56 s->write_poll ? net_socket_writable : NULL, 57 s); 58 } 59 60 static void net_socket_read_poll(NetSocketState *s, bool enable) 61 { 62 s->read_poll = enable; 63 net_socket_update_fd_handler(s); 64 } 65 66 static void net_socket_write_poll(NetSocketState *s, bool enable) 67 { 68 s->write_poll = enable; 69 net_socket_update_fd_handler(s); 70 } 71 72 static void net_socket_writable(void *opaque) 73 { 74 NetSocketState *s = opaque; 75 76 net_socket_write_poll(s, false); 77 78 qemu_flush_queued_packets(&s->nc); 79 } 80 81 static ssize_t net_socket_receive(NetClientState *nc, const uint8_t *buf, size_t size) 82 { 83 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc); 84 uint32_t len = htonl(size); 85 struct iovec iov[] = { 86 { 87 .iov_base = &len, 88 .iov_len = sizeof(len), 89 }, { 90 .iov_base = (void *)buf, 91 .iov_len = size, 92 }, 93 }; 94 size_t remaining; 95 ssize_t ret; 96 97 remaining = iov_size(iov, 2) - s->send_index; 98 ret = iov_send(s->fd, iov, 2, s->send_index, remaining); 99 100 if (ret == -1 && errno == EAGAIN) { 101 ret = 0; /* handled further down */ 102 } 103 if (ret == -1) { 104 s->send_index = 0; 105 return -errno; 106 } 107 if (ret < (ssize_t)remaining) { 108 s->send_index += ret; 109 net_socket_write_poll(s, true); 110 return 0; 111 } 112 s->send_index = 0; 113 return size; 114 } 115 116 static ssize_t net_socket_receive_dgram(NetClientState *nc, const uint8_t *buf, size_t size) 117 { 118 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc); 119 ssize_t ret; 120 121 do { 122 ret = qemu_sendto(s->fd, buf, size, 0, 123 (struct sockaddr *)&s->dgram_dst, 124 sizeof(s->dgram_dst)); 125 } while (ret == -1 && errno == EINTR); 126 127 if (ret == -1 && errno == EAGAIN) { 128 net_socket_write_poll(s, true); 129 return 0; 130 } 131 return ret; 132 } 133 134 static void net_socket_send_completed(NetClientState *nc, ssize_t len) 135 { 136 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc); 137 138 if (!s->read_poll) { 139 net_socket_read_poll(s, true); 140 } 141 } 142 143 static void net_socket_rs_finalize(SocketReadState *rs) 144 { 145 NetSocketState *s = container_of(rs, NetSocketState, rs); 146 147 if (qemu_send_packet_async(&s->nc, rs->buf, 148 rs->packet_len, 149 net_socket_send_completed) == 0) { 150 net_socket_read_poll(s, false); 151 } 152 } 153 154 static void net_socket_send(void *opaque) 155 { 156 NetSocketState *s = opaque; 157 int size; 158 int ret; 159 uint8_t buf1[NET_BUFSIZE]; 160 const uint8_t *buf; 161 162 size = qemu_recv(s->fd, buf1, sizeof(buf1), 0); 163 if (size < 0) { 164 if (errno != EWOULDBLOCK) 165 goto eoc; 166 } else if (size == 0) { 167 /* end of connection */ 168 eoc: 169 net_socket_read_poll(s, false); 170 net_socket_write_poll(s, false); 171 if (s->listen_fd != -1) { 172 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s); 173 } 174 closesocket(s->fd); 175 176 s->fd = -1; 177 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false); 178 s->nc.link_down = true; 179 memset(s->nc.info_str, 0, sizeof(s->nc.info_str)); 180 181 return; 182 } 183 buf = buf1; 184 185 ret = net_fill_rstate(&s->rs, buf, size); 186 187 if (ret == -1) { 188 goto eoc; 189 } 190 } 191 192 static void net_socket_send_dgram(void *opaque) 193 { 194 NetSocketState *s = opaque; 195 int size; 196 197 size = qemu_recv(s->fd, s->rs.buf, sizeof(s->rs.buf), 0); 198 if (size < 0) 199 return; 200 if (size == 0) { 201 /* end of connection */ 202 net_socket_read_poll(s, false); 203 net_socket_write_poll(s, false); 204 return; 205 } 206 if (qemu_send_packet_async(&s->nc, s->rs.buf, size, 207 net_socket_send_completed) == 0) { 208 net_socket_read_poll(s, false); 209 } 210 } 211 212 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr, struct in_addr *localaddr) 213 { 214 struct ip_mreq imr; 215 int fd; 216 int val, ret; 217 #ifdef __OpenBSD__ 218 unsigned char loop; 219 #else 220 int loop; 221 #endif 222 223 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) { 224 fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) " 225 "does not contain a multicast address\n", 226 inet_ntoa(mcastaddr->sin_addr), 227 (int)ntohl(mcastaddr->sin_addr.s_addr)); 228 return -1; 229 230 } 231 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0); 232 if (fd < 0) { 233 perror("socket(PF_INET, SOCK_DGRAM)"); 234 return -1; 235 } 236 237 /* Allow multiple sockets to bind the same multicast ip and port by setting 238 * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set 239 * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR 240 * only on posix systems. 241 */ 242 val = 1; 243 ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val)); 244 if (ret < 0) { 245 perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)"); 246 goto fail; 247 } 248 249 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr)); 250 if (ret < 0) { 251 perror("bind"); 252 goto fail; 253 } 254 255 /* Add host to multicast group */ 256 imr.imr_multiaddr = mcastaddr->sin_addr; 257 if (localaddr) { 258 imr.imr_interface = *localaddr; 259 } else { 260 imr.imr_interface.s_addr = htonl(INADDR_ANY); 261 } 262 263 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP, 264 &imr, sizeof(struct ip_mreq)); 265 if (ret < 0) { 266 perror("setsockopt(IP_ADD_MEMBERSHIP)"); 267 goto fail; 268 } 269 270 /* Force mcast msgs to loopback (eg. several QEMUs in same host */ 271 loop = 1; 272 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP, 273 &loop, sizeof(loop)); 274 if (ret < 0) { 275 perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)"); 276 goto fail; 277 } 278 279 /* If a bind address is given, only send packets from that address */ 280 if (localaddr != NULL) { 281 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF, 282 localaddr, sizeof(*localaddr)); 283 if (ret < 0) { 284 perror("setsockopt(IP_MULTICAST_IF)"); 285 goto fail; 286 } 287 } 288 289 qemu_set_nonblock(fd); 290 return fd; 291 fail: 292 if (fd >= 0) 293 closesocket(fd); 294 return -1; 295 } 296 297 static void net_socket_cleanup(NetClientState *nc) 298 { 299 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc); 300 if (s->fd != -1) { 301 net_socket_read_poll(s, false); 302 net_socket_write_poll(s, false); 303 close(s->fd); 304 s->fd = -1; 305 } 306 if (s->listen_fd != -1) { 307 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL); 308 closesocket(s->listen_fd); 309 s->listen_fd = -1; 310 } 311 } 312 313 static NetClientInfo net_dgram_socket_info = { 314 .type = NET_CLIENT_DRIVER_SOCKET, 315 .size = sizeof(NetSocketState), 316 .receive = net_socket_receive_dgram, 317 .cleanup = net_socket_cleanup, 318 }; 319 320 static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer, 321 const char *model, 322 const char *name, 323 int fd, int is_connected, 324 const char *mcast) 325 { 326 struct sockaddr_in saddr; 327 int newfd; 328 NetClientState *nc; 329 NetSocketState *s; 330 331 /* fd passed: multicast: "learn" dgram_dst address from bound address and save it 332 * Because this may be "shared" socket from a "master" process, datagrams would be recv() 333 * by ONLY ONE process: we must "clone" this dgram socket --jjo 334 */ 335 336 if (is_connected && mcast != NULL) { 337 if (parse_host_port(&saddr, mcast) < 0) { 338 fprintf(stderr, 339 "qemu: error: init_dgram: fd=%d failed parse_host_port()\n", 340 fd); 341 goto err; 342 } 343 /* must be bound */ 344 if (saddr.sin_addr.s_addr == 0) { 345 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, " 346 "cannot setup multicast dst addr\n", fd); 347 goto err; 348 } 349 /* clone dgram socket */ 350 newfd = net_socket_mcast_create(&saddr, NULL); 351 if (newfd < 0) { 352 /* error already reported by net_socket_mcast_create() */ 353 goto err; 354 } 355 /* clone newfd to fd, close newfd */ 356 dup2(newfd, fd); 357 close(newfd); 358 359 } 360 361 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name); 362 363 s = DO_UPCAST(NetSocketState, nc, nc); 364 365 s->fd = fd; 366 s->listen_fd = -1; 367 s->send_fn = net_socket_send_dgram; 368 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false); 369 net_socket_read_poll(s, true); 370 371 /* mcast: save bound address as dst */ 372 if (is_connected) { 373 s->dgram_dst = saddr; 374 snprintf(nc->info_str, sizeof(nc->info_str), 375 "socket: fd=%d (cloned mcast=%s:%d)", 376 fd, inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port)); 377 } else { 378 snprintf(nc->info_str, sizeof(nc->info_str), 379 "socket: fd=%d", fd); 380 } 381 382 return s; 383 384 err: 385 closesocket(fd); 386 return NULL; 387 } 388 389 static void net_socket_connect(void *opaque) 390 { 391 NetSocketState *s = opaque; 392 s->send_fn = net_socket_send; 393 net_socket_read_poll(s, true); 394 } 395 396 static NetClientInfo net_socket_info = { 397 .type = NET_CLIENT_DRIVER_SOCKET, 398 .size = sizeof(NetSocketState), 399 .receive = net_socket_receive, 400 .cleanup = net_socket_cleanup, 401 }; 402 403 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer, 404 const char *model, 405 const char *name, 406 int fd, int is_connected) 407 { 408 NetClientState *nc; 409 NetSocketState *s; 410 411 nc = qemu_new_net_client(&net_socket_info, peer, model, name); 412 413 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd); 414 415 s = DO_UPCAST(NetSocketState, nc, nc); 416 417 s->fd = fd; 418 s->listen_fd = -1; 419 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false); 420 421 /* Disable Nagle algorithm on TCP sockets to reduce latency */ 422 socket_set_nodelay(fd); 423 424 if (is_connected) { 425 net_socket_connect(s); 426 } else { 427 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s); 428 } 429 return s; 430 } 431 432 static NetSocketState *net_socket_fd_init(NetClientState *peer, 433 const char *model, const char *name, 434 int fd, int is_connected, const char *mc) 435 { 436 int so_type = -1, optlen=sizeof(so_type); 437 438 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type, 439 (socklen_t *)&optlen)< 0) { 440 fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n", 441 fd); 442 closesocket(fd); 443 return NULL; 444 } 445 switch(so_type) { 446 case SOCK_DGRAM: 447 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected, mc); 448 case SOCK_STREAM: 449 return net_socket_fd_init_stream(peer, model, name, fd, is_connected); 450 default: 451 /* who knows ... this could be a eg. a pty, do warn and continue as stream */ 452 fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd); 453 return net_socket_fd_init_stream(peer, model, name, fd, is_connected); 454 } 455 return NULL; 456 } 457 458 static void net_socket_accept(void *opaque) 459 { 460 NetSocketState *s = opaque; 461 struct sockaddr_in saddr; 462 socklen_t len; 463 int fd; 464 465 for(;;) { 466 len = sizeof(saddr); 467 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len); 468 if (fd < 0 && errno != EINTR) { 469 return; 470 } else if (fd >= 0) { 471 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL); 472 break; 473 } 474 } 475 476 s->fd = fd; 477 s->nc.link_down = false; 478 net_socket_connect(s); 479 snprintf(s->nc.info_str, sizeof(s->nc.info_str), 480 "socket: connection from %s:%d", 481 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port)); 482 } 483 484 static int net_socket_listen_init(NetClientState *peer, 485 const char *model, 486 const char *name, 487 const char *host_str) 488 { 489 NetClientState *nc; 490 NetSocketState *s; 491 struct sockaddr_in saddr; 492 int fd, ret; 493 494 if (parse_host_port(&saddr, host_str) < 0) 495 return -1; 496 497 fd = qemu_socket(PF_INET, SOCK_STREAM, 0); 498 if (fd < 0) { 499 perror("socket"); 500 return -1; 501 } 502 qemu_set_nonblock(fd); 503 504 socket_set_fast_reuse(fd); 505 506 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr)); 507 if (ret < 0) { 508 perror("bind"); 509 closesocket(fd); 510 return -1; 511 } 512 ret = listen(fd, 0); 513 if (ret < 0) { 514 perror("listen"); 515 closesocket(fd); 516 return -1; 517 } 518 519 nc = qemu_new_net_client(&net_socket_info, peer, model, name); 520 s = DO_UPCAST(NetSocketState, nc, nc); 521 s->fd = -1; 522 s->listen_fd = fd; 523 s->nc.link_down = true; 524 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false); 525 526 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s); 527 return 0; 528 } 529 530 static int net_socket_connect_init(NetClientState *peer, 531 const char *model, 532 const char *name, 533 const char *host_str) 534 { 535 NetSocketState *s; 536 int fd, connected, ret; 537 struct sockaddr_in saddr; 538 539 if (parse_host_port(&saddr, host_str) < 0) 540 return -1; 541 542 fd = qemu_socket(PF_INET, SOCK_STREAM, 0); 543 if (fd < 0) { 544 perror("socket"); 545 return -1; 546 } 547 qemu_set_nonblock(fd); 548 549 connected = 0; 550 for(;;) { 551 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr)); 552 if (ret < 0) { 553 if (errno == EINTR || errno == EWOULDBLOCK) { 554 /* continue */ 555 } else if (errno == EINPROGRESS || 556 errno == EALREADY || 557 errno == EINVAL) { 558 break; 559 } else { 560 perror("connect"); 561 closesocket(fd); 562 return -1; 563 } 564 } else { 565 connected = 1; 566 break; 567 } 568 } 569 s = net_socket_fd_init(peer, model, name, fd, connected, NULL); 570 if (!s) 571 return -1; 572 snprintf(s->nc.info_str, sizeof(s->nc.info_str), 573 "socket: connect to %s:%d", 574 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port)); 575 return 0; 576 } 577 578 static int net_socket_mcast_init(NetClientState *peer, 579 const char *model, 580 const char *name, 581 const char *host_str, 582 const char *localaddr_str) 583 { 584 NetSocketState *s; 585 int fd; 586 struct sockaddr_in saddr; 587 struct in_addr localaddr, *param_localaddr; 588 589 if (parse_host_port(&saddr, host_str) < 0) 590 return -1; 591 592 if (localaddr_str != NULL) { 593 if (inet_aton(localaddr_str, &localaddr) == 0) 594 return -1; 595 param_localaddr = &localaddr; 596 } else { 597 param_localaddr = NULL; 598 } 599 600 fd = net_socket_mcast_create(&saddr, param_localaddr); 601 if (fd < 0) 602 return -1; 603 604 s = net_socket_fd_init(peer, model, name, fd, 0, NULL); 605 if (!s) 606 return -1; 607 608 s->dgram_dst = saddr; 609 610 snprintf(s->nc.info_str, sizeof(s->nc.info_str), 611 "socket: mcast=%s:%d", 612 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port)); 613 return 0; 614 615 } 616 617 static int net_socket_udp_init(NetClientState *peer, 618 const char *model, 619 const char *name, 620 const char *rhost, 621 const char *lhost) 622 { 623 NetSocketState *s; 624 int fd, ret; 625 struct sockaddr_in laddr, raddr; 626 627 if (parse_host_port(&laddr, lhost) < 0) { 628 return -1; 629 } 630 631 if (parse_host_port(&raddr, rhost) < 0) { 632 return -1; 633 } 634 635 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0); 636 if (fd < 0) { 637 perror("socket(PF_INET, SOCK_DGRAM)"); 638 return -1; 639 } 640 641 ret = socket_set_fast_reuse(fd); 642 if (ret < 0) { 643 closesocket(fd); 644 return -1; 645 } 646 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr)); 647 if (ret < 0) { 648 perror("bind"); 649 closesocket(fd); 650 return -1; 651 } 652 qemu_set_nonblock(fd); 653 654 s = net_socket_fd_init(peer, model, name, fd, 0, NULL); 655 if (!s) { 656 return -1; 657 } 658 659 s->dgram_dst = raddr; 660 661 snprintf(s->nc.info_str, sizeof(s->nc.info_str), 662 "socket: udp=%s:%d", 663 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port)); 664 return 0; 665 } 666 667 int net_init_socket(const Netdev *netdev, const char *name, 668 NetClientState *peer, Error **errp) 669 { 670 /* FIXME error_setg(errp, ...) on failure */ 671 Error *err = NULL; 672 const NetdevSocketOptions *sock; 673 674 assert(netdev->type == NET_CLIENT_DRIVER_SOCKET); 675 sock = &netdev->u.socket; 676 677 if (sock->has_listen + sock->has_connect + sock->has_mcast + 678 sock->has_udp > 1) { 679 error_report("exactly one of listen=, connect=, mcast= or udp=" 680 " is required"); 681 return -1; 682 } 683 684 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) { 685 error_report("localaddr= is only valid with mcast= or udp="); 686 return -1; 687 } 688 689 if (sock->has_fd) { 690 int fd; 691 692 fd = monitor_fd_param(cur_mon, sock->fd, &err); 693 if (fd == -1) { 694 error_report_err(err); 695 return -1; 696 } 697 qemu_set_nonblock(fd); 698 if (!net_socket_fd_init(peer, "socket", name, fd, 1, sock->mcast)) { 699 return -1; 700 } 701 return 0; 702 } 703 704 if (sock->has_listen) { 705 if (net_socket_listen_init(peer, "socket", name, sock->listen) == -1) { 706 return -1; 707 } 708 return 0; 709 } 710 711 if (sock->has_connect) { 712 if (net_socket_connect_init(peer, "socket", name, sock->connect) == 713 -1) { 714 return -1; 715 } 716 return 0; 717 } 718 719 if (sock->has_mcast) { 720 /* if sock->localaddr is missing, it has been initialized to "all bits 721 * zero" */ 722 if (net_socket_mcast_init(peer, "socket", name, sock->mcast, 723 sock->localaddr) == -1) { 724 return -1; 725 } 726 return 0; 727 } 728 729 assert(sock->has_udp); 730 if (!sock->has_localaddr) { 731 error_report("localaddr= is mandatory with udp="); 732 return -1; 733 } 734 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr) == 735 -1) { 736 return -1; 737 } 738 return 0; 739 } 740