block: vpc - use block layer ops in vpc_create, instead of posix calls
[qemu.git] / net / socket.c
blobfb21e20a54caacb8e612f118966bfc7e77f954da
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 #include "config-host.h"
26 #include "net/net.h"
27 #include "clients.h"
28 #include "monitor/monitor.h"
29 #include "qemu-common.h"
30 #include "qemu/error-report.h"
31 #include "qemu/option.h"
32 #include "qemu/sockets.h"
33 #include "qemu/iov.h"
34 #include "qemu/main-loop.h"
36 typedef struct NetSocketState {
37 NetClientState nc;
38 int listen_fd;
39 int fd;
40 int state; /* 0 = getting length, 1 = getting data */
41 unsigned int index;
42 unsigned int packet_len;
43 unsigned int send_index; /* number of bytes sent (only SOCK_STREAM) */
44 uint8_t buf[NET_BUFSIZE];
45 struct sockaddr_in dgram_dst; /* contains inet host and port destination iff connectionless (SOCK_DGRAM) */
46 IOHandler *send_fn; /* differs between SOCK_STREAM/SOCK_DGRAM */
47 bool read_poll; /* waiting to receive data? */
48 bool write_poll; /* waiting to transmit data? */
49 } NetSocketState;
51 static void net_socket_accept(void *opaque);
52 static void net_socket_writable(void *opaque);
54 /* Only read packets from socket when peer can receive them */
55 static int net_socket_can_send(void *opaque)
57 NetSocketState *s = opaque;
59 return qemu_can_send_packet(&s->nc);
62 static void net_socket_update_fd_handler(NetSocketState *s)
64 qemu_set_fd_handler2(s->fd,
65 s->read_poll ? net_socket_can_send : NULL,
66 s->read_poll ? s->send_fn : NULL,
67 s->write_poll ? net_socket_writable : NULL,
68 s);
71 static void net_socket_read_poll(NetSocketState *s, bool enable)
73 s->read_poll = enable;
74 net_socket_update_fd_handler(s);
77 static void net_socket_write_poll(NetSocketState *s, bool enable)
79 s->write_poll = enable;
80 net_socket_update_fd_handler(s);
83 static void net_socket_writable(void *opaque)
85 NetSocketState *s = opaque;
87 net_socket_write_poll(s, false);
89 qemu_flush_queued_packets(&s->nc);
92 static ssize_t net_socket_receive(NetClientState *nc, const uint8_t *buf, size_t size)
94 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
95 uint32_t len = htonl(size);
96 struct iovec iov[] = {
98 .iov_base = &len,
99 .iov_len = sizeof(len),
100 }, {
101 .iov_base = (void *)buf,
102 .iov_len = size,
105 size_t remaining;
106 ssize_t ret;
108 remaining = iov_size(iov, 2) - s->send_index;
109 ret = iov_send(s->fd, iov, 2, s->send_index, remaining);
111 if (ret == -1 && errno == EAGAIN) {
112 ret = 0; /* handled further down */
114 if (ret == -1) {
115 s->send_index = 0;
116 return -errno;
118 if (ret < (ssize_t)remaining) {
119 s->send_index += ret;
120 net_socket_write_poll(s, true);
121 return 0;
123 s->send_index = 0;
124 return size;
127 static ssize_t net_socket_receive_dgram(NetClientState *nc, const uint8_t *buf, size_t size)
129 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
130 ssize_t ret;
132 do {
133 ret = qemu_sendto(s->fd, buf, size, 0,
134 (struct sockaddr *)&s->dgram_dst,
135 sizeof(s->dgram_dst));
136 } while (ret == -1 && errno == EINTR);
138 if (ret == -1 && errno == EAGAIN) {
139 net_socket_write_poll(s, true);
140 return 0;
142 return ret;
145 static void net_socket_send(void *opaque)
147 NetSocketState *s = opaque;
148 int size, err;
149 unsigned l;
150 uint8_t buf1[NET_BUFSIZE];
151 const uint8_t *buf;
153 size = qemu_recv(s->fd, buf1, sizeof(buf1), 0);
154 if (size < 0) {
155 err = socket_error();
156 if (err != EWOULDBLOCK)
157 goto eoc;
158 } else if (size == 0) {
159 /* end of connection */
160 eoc:
161 net_socket_read_poll(s, false);
162 net_socket_write_poll(s, false);
163 if (s->listen_fd != -1) {
164 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
166 closesocket(s->fd);
168 s->fd = -1;
169 s->state = 0;
170 s->index = 0;
171 s->packet_len = 0;
172 s->nc.link_down = true;
173 memset(s->buf, 0, sizeof(s->buf));
174 memset(s->nc.info_str, 0, sizeof(s->nc.info_str));
176 return;
178 buf = buf1;
179 while (size > 0) {
180 /* reassemble a packet from the network */
181 switch(s->state) {
182 case 0:
183 l = 4 - s->index;
184 if (l > size)
185 l = size;
186 memcpy(s->buf + s->index, buf, l);
187 buf += l;
188 size -= l;
189 s->index += l;
190 if (s->index == 4) {
191 /* got length */
192 s->packet_len = ntohl(*(uint32_t *)s->buf);
193 s->index = 0;
194 s->state = 1;
196 break;
197 case 1:
198 l = s->packet_len - s->index;
199 if (l > size)
200 l = size;
201 if (s->index + l <= sizeof(s->buf)) {
202 memcpy(s->buf + s->index, buf, l);
203 } else {
204 fprintf(stderr, "serious error: oversized packet received,"
205 "connection terminated.\n");
206 s->state = 0;
207 goto eoc;
210 s->index += l;
211 buf += l;
212 size -= l;
213 if (s->index >= s->packet_len) {
214 qemu_send_packet(&s->nc, s->buf, s->packet_len);
215 s->index = 0;
216 s->state = 0;
218 break;
223 static void net_socket_send_dgram(void *opaque)
225 NetSocketState *s = opaque;
226 int size;
228 size = qemu_recv(s->fd, s->buf, sizeof(s->buf), 0);
229 if (size < 0)
230 return;
231 if (size == 0) {
232 /* end of connection */
233 net_socket_read_poll(s, false);
234 net_socket_write_poll(s, false);
235 return;
237 qemu_send_packet(&s->nc, s->buf, size);
240 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr, struct in_addr *localaddr)
242 struct ip_mreq imr;
243 int fd;
244 int val, ret;
245 #ifdef __OpenBSD__
246 unsigned char loop;
247 #else
248 int loop;
249 #endif
251 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
252 fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) "
253 "does not contain a multicast address\n",
254 inet_ntoa(mcastaddr->sin_addr),
255 (int)ntohl(mcastaddr->sin_addr.s_addr));
256 return -1;
259 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
260 if (fd < 0) {
261 perror("socket(PF_INET, SOCK_DGRAM)");
262 return -1;
265 /* Allow multiple sockets to bind the same multicast ip and port by setting
266 * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set
267 * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR
268 * only on posix systems.
270 val = 1;
271 ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
272 if (ret < 0) {
273 perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
274 goto fail;
277 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
278 if (ret < 0) {
279 perror("bind");
280 goto fail;
283 /* Add host to multicast group */
284 imr.imr_multiaddr = mcastaddr->sin_addr;
285 if (localaddr) {
286 imr.imr_interface = *localaddr;
287 } else {
288 imr.imr_interface.s_addr = htonl(INADDR_ANY);
291 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
292 &imr, sizeof(struct ip_mreq));
293 if (ret < 0) {
294 perror("setsockopt(IP_ADD_MEMBERSHIP)");
295 goto fail;
298 /* Force mcast msgs to loopback (eg. several QEMUs in same host */
299 loop = 1;
300 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
301 &loop, sizeof(loop));
302 if (ret < 0) {
303 perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)");
304 goto fail;
307 /* If a bind address is given, only send packets from that address */
308 if (localaddr != NULL) {
309 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF,
310 localaddr, sizeof(*localaddr));
311 if (ret < 0) {
312 perror("setsockopt(IP_MULTICAST_IF)");
313 goto fail;
317 qemu_set_nonblock(fd);
318 return fd;
319 fail:
320 if (fd >= 0)
321 closesocket(fd);
322 return -1;
325 static void net_socket_cleanup(NetClientState *nc)
327 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
328 if (s->fd != -1) {
329 net_socket_read_poll(s, false);
330 net_socket_write_poll(s, false);
331 close(s->fd);
332 s->fd = -1;
334 if (s->listen_fd != -1) {
335 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
336 closesocket(s->listen_fd);
337 s->listen_fd = -1;
341 static NetClientInfo net_dgram_socket_info = {
342 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
343 .size = sizeof(NetSocketState),
344 .receive = net_socket_receive_dgram,
345 .cleanup = net_socket_cleanup,
348 static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer,
349 const char *model,
350 const char *name,
351 int fd, int is_connected)
353 struct sockaddr_in saddr;
354 int newfd;
355 socklen_t saddr_len;
356 NetClientState *nc;
357 NetSocketState *s;
359 /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
360 * Because this may be "shared" socket from a "master" process, datagrams would be recv()
361 * by ONLY ONE process: we must "clone" this dgram socket --jjo
364 if (is_connected) {
365 if (getsockname(fd, (struct sockaddr *) &saddr, &saddr_len) == 0) {
366 /* must be bound */
367 if (saddr.sin_addr.s_addr == 0) {
368 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, "
369 "cannot setup multicast dst addr\n", fd);
370 goto err;
372 /* clone dgram socket */
373 newfd = net_socket_mcast_create(&saddr, NULL);
374 if (newfd < 0) {
375 /* error already reported by net_socket_mcast_create() */
376 goto err;
378 /* clone newfd to fd, close newfd */
379 dup2(newfd, fd);
380 close(newfd);
382 } else {
383 fprintf(stderr,
384 "qemu: error: init_dgram: fd=%d failed getsockname(): %s\n",
385 fd, strerror(errno));
386 goto err;
390 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name);
392 snprintf(nc->info_str, sizeof(nc->info_str),
393 "socket: fd=%d (%s mcast=%s:%d)",
394 fd, is_connected ? "cloned" : "",
395 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
397 s = DO_UPCAST(NetSocketState, nc, nc);
399 s->fd = fd;
400 s->listen_fd = -1;
401 s->send_fn = net_socket_send_dgram;
402 net_socket_read_poll(s, true);
404 /* mcast: save bound address as dst */
405 if (is_connected) {
406 s->dgram_dst = saddr;
409 return s;
411 err:
412 closesocket(fd);
413 return NULL;
416 static void net_socket_connect(void *opaque)
418 NetSocketState *s = opaque;
419 s->send_fn = net_socket_send;
420 net_socket_read_poll(s, true);
423 static NetClientInfo net_socket_info = {
424 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
425 .size = sizeof(NetSocketState),
426 .receive = net_socket_receive,
427 .cleanup = net_socket_cleanup,
430 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer,
431 const char *model,
432 const char *name,
433 int fd, int is_connected)
435 NetClientState *nc;
436 NetSocketState *s;
438 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
440 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd);
442 s = DO_UPCAST(NetSocketState, nc, nc);
444 s->fd = fd;
445 s->listen_fd = -1;
447 /* Disable Nagle algorithm on TCP sockets to reduce latency */
448 socket_set_nodelay(fd);
450 if (is_connected) {
451 net_socket_connect(s);
452 } else {
453 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
455 return s;
458 static NetSocketState *net_socket_fd_init(NetClientState *peer,
459 const char *model, const char *name,
460 int fd, int is_connected)
462 int so_type = -1, optlen=sizeof(so_type);
464 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
465 (socklen_t *)&optlen)< 0) {
466 fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n",
467 fd);
468 closesocket(fd);
469 return NULL;
471 switch(so_type) {
472 case SOCK_DGRAM:
473 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected);
474 case SOCK_STREAM:
475 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
476 default:
477 /* who knows ... this could be a eg. a pty, do warn and continue as stream */
478 fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
479 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
481 return NULL;
484 static void net_socket_accept(void *opaque)
486 NetSocketState *s = opaque;
487 struct sockaddr_in saddr;
488 socklen_t len;
489 int fd;
491 for(;;) {
492 len = sizeof(saddr);
493 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len);
494 if (fd < 0 && errno != EINTR) {
495 return;
496 } else if (fd >= 0) {
497 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
498 break;
502 s->fd = fd;
503 s->nc.link_down = false;
504 net_socket_connect(s);
505 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
506 "socket: connection from %s:%d",
507 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
510 static int net_socket_listen_init(NetClientState *peer,
511 const char *model,
512 const char *name,
513 const char *host_str)
515 NetClientState *nc;
516 NetSocketState *s;
517 struct sockaddr_in saddr;
518 int fd, ret;
520 if (parse_host_port(&saddr, host_str) < 0)
521 return -1;
523 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
524 if (fd < 0) {
525 perror("socket");
526 return -1;
528 qemu_set_nonblock(fd);
530 socket_set_fast_reuse(fd);
532 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
533 if (ret < 0) {
534 perror("bind");
535 closesocket(fd);
536 return -1;
538 ret = listen(fd, 0);
539 if (ret < 0) {
540 perror("listen");
541 closesocket(fd);
542 return -1;
545 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
546 s = DO_UPCAST(NetSocketState, nc, nc);
547 s->fd = -1;
548 s->listen_fd = fd;
549 s->nc.link_down = true;
551 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
552 return 0;
555 static int net_socket_connect_init(NetClientState *peer,
556 const char *model,
557 const char *name,
558 const char *host_str)
560 NetSocketState *s;
561 int fd, connected, ret, err;
562 struct sockaddr_in saddr;
564 if (parse_host_port(&saddr, host_str) < 0)
565 return -1;
567 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
568 if (fd < 0) {
569 perror("socket");
570 return -1;
572 qemu_set_nonblock(fd);
574 connected = 0;
575 for(;;) {
576 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
577 if (ret < 0) {
578 err = socket_error();
579 if (err == EINTR || err == EWOULDBLOCK) {
580 } else if (err == EINPROGRESS) {
581 break;
582 #ifdef _WIN32
583 } else if (err == WSAEALREADY || err == WSAEINVAL) {
584 break;
585 #endif
586 } else {
587 perror("connect");
588 closesocket(fd);
589 return -1;
591 } else {
592 connected = 1;
593 break;
596 s = net_socket_fd_init(peer, model, name, fd, connected);
597 if (!s)
598 return -1;
599 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
600 "socket: connect to %s:%d",
601 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
602 return 0;
605 static int net_socket_mcast_init(NetClientState *peer,
606 const char *model,
607 const char *name,
608 const char *host_str,
609 const char *localaddr_str)
611 NetSocketState *s;
612 int fd;
613 struct sockaddr_in saddr;
614 struct in_addr localaddr, *param_localaddr;
616 if (parse_host_port(&saddr, host_str) < 0)
617 return -1;
619 if (localaddr_str != NULL) {
620 if (inet_aton(localaddr_str, &localaddr) == 0)
621 return -1;
622 param_localaddr = &localaddr;
623 } else {
624 param_localaddr = NULL;
627 fd = net_socket_mcast_create(&saddr, param_localaddr);
628 if (fd < 0)
629 return -1;
631 s = net_socket_fd_init(peer, model, name, fd, 0);
632 if (!s)
633 return -1;
635 s->dgram_dst = saddr;
637 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
638 "socket: mcast=%s:%d",
639 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
640 return 0;
644 static int net_socket_udp_init(NetClientState *peer,
645 const char *model,
646 const char *name,
647 const char *rhost,
648 const char *lhost)
650 NetSocketState *s;
651 int fd, ret;
652 struct sockaddr_in laddr, raddr;
654 if (parse_host_port(&laddr, lhost) < 0) {
655 return -1;
658 if (parse_host_port(&raddr, rhost) < 0) {
659 return -1;
662 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
663 if (fd < 0) {
664 perror("socket(PF_INET, SOCK_DGRAM)");
665 return -1;
668 ret = socket_set_fast_reuse(fd);
669 if (ret < 0) {
670 closesocket(fd);
671 return -1;
673 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
674 if (ret < 0) {
675 perror("bind");
676 closesocket(fd);
677 return -1;
679 qemu_set_nonblock(fd);
681 s = net_socket_fd_init(peer, model, name, fd, 0);
682 if (!s) {
683 return -1;
686 s->dgram_dst = raddr;
688 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
689 "socket: udp=%s:%d",
690 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port));
691 return 0;
694 int net_init_socket(const NetClientOptions *opts, const char *name,
695 NetClientState *peer)
697 const NetdevSocketOptions *sock;
699 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_SOCKET);
700 sock = opts->socket;
702 if (sock->has_fd + sock->has_listen + sock->has_connect + sock->has_mcast +
703 sock->has_udp != 1) {
704 error_report("exactly one of fd=, listen=, connect=, mcast= or udp="
705 " is required");
706 return -1;
709 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) {
710 error_report("localaddr= is only valid with mcast= or udp=");
711 return -1;
714 if (sock->has_fd) {
715 int fd;
717 fd = monitor_handle_fd_param(cur_mon, sock->fd);
718 if (fd == -1) {
719 return -1;
721 qemu_set_nonblock(fd);
722 if (!net_socket_fd_init(peer, "socket", name, fd, 1)) {
723 return -1;
725 return 0;
728 if (sock->has_listen) {
729 if (net_socket_listen_init(peer, "socket", name, sock->listen) == -1) {
730 return -1;
732 return 0;
735 if (sock->has_connect) {
736 if (net_socket_connect_init(peer, "socket", name, sock->connect) ==
737 -1) {
738 return -1;
740 return 0;
743 if (sock->has_mcast) {
744 /* if sock->localaddr is missing, it has been initialized to "all bits
745 * zero" */
746 if (net_socket_mcast_init(peer, "socket", name, sock->mcast,
747 sock->localaddr) == -1) {
748 return -1;
750 return 0;
753 assert(sock->has_udp);
754 if (!sock->has_localaddr) {
755 error_report("localaddr= is mandatory with udp=");
756 return -1;
758 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr) ==
759 -1) {
760 return -1;
762 return 0;