vhost-user-blk: set config ops before vhost-user init
[qemu/ar7.git] / net / socket.c
blob6917fbcbf582cf3606abe168a55918f262250bcc
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 "qemu/osdep.h"
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"
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;
49 static void net_socket_accept(void *opaque);
50 static void net_socket_writable(void *opaque);
52 static void net_socket_update_fd_handler(NetSocketState *s)
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);
60 static void net_socket_read_poll(NetSocketState *s, bool enable)
62 s->read_poll = enable;
63 net_socket_update_fd_handler(s);
66 static void net_socket_write_poll(NetSocketState *s, bool enable)
68 s->write_poll = enable;
69 net_socket_update_fd_handler(s);
72 static void net_socket_writable(void *opaque)
74 NetSocketState *s = opaque;
76 net_socket_write_poll(s, false);
78 qemu_flush_queued_packets(&s->nc);
81 static ssize_t net_socket_receive(NetClientState *nc, const uint8_t *buf, size_t size)
83 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
84 uint32_t len = htonl(size);
85 struct iovec iov[] = {
87 .iov_base = &len,
88 .iov_len = sizeof(len),
89 }, {
90 .iov_base = (void *)buf,
91 .iov_len = size,
94 size_t remaining;
95 ssize_t ret;
97 remaining = iov_size(iov, 2) - s->send_index;
98 ret = iov_send(s->fd, iov, 2, s->send_index, remaining);
100 if (ret == -1 && errno == EAGAIN) {
101 ret = 0; /* handled further down */
103 if (ret == -1) {
104 s->send_index = 0;
105 return -errno;
107 if (ret < (ssize_t)remaining) {
108 s->send_index += ret;
109 net_socket_write_poll(s, true);
110 return 0;
112 s->send_index = 0;
113 return size;
116 static ssize_t net_socket_receive_dgram(NetClientState *nc, const uint8_t *buf, size_t size)
118 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
119 ssize_t ret;
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);
127 if (ret == -1 && errno == EAGAIN) {
128 net_socket_write_poll(s, true);
129 return 0;
131 return ret;
134 static void net_socket_send_completed(NetClientState *nc, ssize_t len)
136 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
138 if (!s->read_poll) {
139 net_socket_read_poll(s, true);
143 static void net_socket_rs_finalize(SocketReadState *rs)
145 NetSocketState *s = container_of(rs, NetSocketState, rs);
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);
154 static void net_socket_send(void *opaque)
156 NetSocketState *s = opaque;
157 int size;
158 int ret;
159 uint8_t buf1[NET_BUFSIZE];
160 const uint8_t *buf;
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);
174 closesocket(s->fd);
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));
181 return;
183 buf = buf1;
185 ret = net_fill_rstate(&s->rs, buf, size);
187 if (ret == -1) {
188 goto eoc;
192 static void net_socket_send_dgram(void *opaque)
194 NetSocketState *s = opaque;
195 int size;
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;
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);
212 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr,
213 struct in_addr *localaddr,
214 Error **errp)
216 struct ip_mreq imr;
217 int fd;
218 int val, ret;
219 #ifdef __OpenBSD__
220 unsigned char loop;
221 #else
222 int loop;
223 #endif
225 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
226 error_setg(errp, "specified mcastaddr %s (0x%08x) "
227 "does not contain a multicast address",
228 inet_ntoa(mcastaddr->sin_addr),
229 (int)ntohl(mcastaddr->sin_addr.s_addr));
230 return -1;
233 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
234 if (fd < 0) {
235 error_setg_errno(errp, errno, "can't create datagram socket");
236 return -1;
239 /* Allow multiple sockets to bind the same multicast ip and port by setting
240 * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set
241 * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR
242 * only on posix systems.
244 val = 1;
245 ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
246 if (ret < 0) {
247 error_setg_errno(errp, errno,
248 "can't set socket option SO_REUSEADDR");
249 goto fail;
252 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
253 if (ret < 0) {
254 error_setg_errno(errp, errno, "can't bind ip=%s to socket",
255 inet_ntoa(mcastaddr->sin_addr));
256 goto fail;
259 /* Add host to multicast group */
260 imr.imr_multiaddr = mcastaddr->sin_addr;
261 if (localaddr) {
262 imr.imr_interface = *localaddr;
263 } else {
264 imr.imr_interface.s_addr = htonl(INADDR_ANY);
267 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
268 &imr, sizeof(struct ip_mreq));
269 if (ret < 0) {
270 error_setg_errno(errp, errno,
271 "can't add socket to multicast group %s",
272 inet_ntoa(imr.imr_multiaddr));
273 goto fail;
276 /* Force mcast msgs to loopback (eg. several QEMUs in same host */
277 loop = 1;
278 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
279 &loop, sizeof(loop));
280 if (ret < 0) {
281 error_setg_errno(errp, errno,
282 "can't force multicast message to loopback");
283 goto fail;
286 /* If a bind address is given, only send packets from that address */
287 if (localaddr != NULL) {
288 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF,
289 localaddr, sizeof(*localaddr));
290 if (ret < 0) {
291 error_setg_errno(errp, errno,
292 "can't set the default network send interface");
293 goto fail;
297 qemu_set_nonblock(fd);
298 return fd;
299 fail:
300 if (fd >= 0)
301 closesocket(fd);
302 return -1;
305 static void net_socket_cleanup(NetClientState *nc)
307 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
308 if (s->fd != -1) {
309 net_socket_read_poll(s, false);
310 net_socket_write_poll(s, false);
311 close(s->fd);
312 s->fd = -1;
314 if (s->listen_fd != -1) {
315 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
316 closesocket(s->listen_fd);
317 s->listen_fd = -1;
321 static NetClientInfo net_dgram_socket_info = {
322 .type = NET_CLIENT_DRIVER_SOCKET,
323 .size = sizeof(NetSocketState),
324 .receive = net_socket_receive_dgram,
325 .cleanup = net_socket_cleanup,
328 static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer,
329 const char *model,
330 const char *name,
331 int fd, int is_connected,
332 const char *mcast,
333 Error **errp)
335 struct sockaddr_in saddr;
336 int newfd;
337 NetClientState *nc;
338 NetSocketState *s;
340 /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
341 * Because this may be "shared" socket from a "master" process, datagrams would be recv()
342 * by ONLY ONE process: we must "clone" this dgram socket --jjo
345 if (is_connected && mcast != NULL) {
346 if (parse_host_port(&saddr, mcast, errp) < 0) {
347 goto err;
349 /* must be bound */
350 if (saddr.sin_addr.s_addr == 0) {
351 error_setg(errp, "can't setup multicast destination address");
352 goto err;
354 /* clone dgram socket */
355 newfd = net_socket_mcast_create(&saddr, NULL, errp);
356 if (newfd < 0) {
357 goto err;
359 /* clone newfd to fd, close newfd */
360 dup2(newfd, fd);
361 close(newfd);
365 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name);
367 s = DO_UPCAST(NetSocketState, nc, nc);
369 s->fd = fd;
370 s->listen_fd = -1;
371 s->send_fn = net_socket_send_dgram;
372 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false);
373 net_socket_read_poll(s, true);
375 /* mcast: save bound address as dst */
376 if (is_connected && mcast != NULL) {
377 s->dgram_dst = saddr;
378 snprintf(nc->info_str, sizeof(nc->info_str),
379 "socket: fd=%d (cloned mcast=%s:%d)",
380 fd, inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
381 } else {
382 snprintf(nc->info_str, sizeof(nc->info_str),
383 "socket: fd=%d", fd);
386 return s;
388 err:
389 closesocket(fd);
390 return NULL;
393 static void net_socket_connect(void *opaque)
395 NetSocketState *s = opaque;
396 s->send_fn = net_socket_send;
397 net_socket_read_poll(s, true);
400 static NetClientInfo net_socket_info = {
401 .type = NET_CLIENT_DRIVER_SOCKET,
402 .size = sizeof(NetSocketState),
403 .receive = net_socket_receive,
404 .cleanup = net_socket_cleanup,
407 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer,
408 const char *model,
409 const char *name,
410 int fd, int is_connected)
412 NetClientState *nc;
413 NetSocketState *s;
415 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
417 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd);
419 s = DO_UPCAST(NetSocketState, nc, nc);
421 s->fd = fd;
422 s->listen_fd = -1;
423 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false);
425 /* Disable Nagle algorithm on TCP sockets to reduce latency */
426 socket_set_nodelay(fd);
428 if (is_connected) {
429 net_socket_connect(s);
430 } else {
431 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
433 return s;
436 static NetSocketState *net_socket_fd_init(NetClientState *peer,
437 const char *model, const char *name,
438 int fd, int is_connected,
439 const char *mc, Error **errp)
441 int so_type = -1, optlen=sizeof(so_type);
443 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
444 (socklen_t *)&optlen)< 0) {
445 error_setg(errp, "can't get socket option SO_TYPE");
446 closesocket(fd);
447 return NULL;
449 switch(so_type) {
450 case SOCK_DGRAM:
451 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected,
452 mc, errp);
453 case SOCK_STREAM:
454 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
455 default:
456 error_report("socket type=%d for fd=%d must be either"
457 " SOCK_DGRAM or SOCK_STREAM", so_type, fd);
458 closesocket(fd);
460 return NULL;
463 static void net_socket_accept(void *opaque)
465 NetSocketState *s = opaque;
466 struct sockaddr_in saddr;
467 socklen_t len;
468 int fd;
470 for(;;) {
471 len = sizeof(saddr);
472 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len);
473 if (fd < 0 && errno != EINTR) {
474 return;
475 } else if (fd >= 0) {
476 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
477 break;
481 s->fd = fd;
482 s->nc.link_down = false;
483 net_socket_connect(s);
484 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
485 "socket: connection from %s:%d",
486 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
489 static int net_socket_listen_init(NetClientState *peer,
490 const char *model,
491 const char *name,
492 const char *host_str,
493 Error **errp)
495 NetClientState *nc;
496 NetSocketState *s;
497 struct sockaddr_in saddr;
498 int fd, ret;
500 if (parse_host_port(&saddr, host_str, errp) < 0) {
501 return -1;
504 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
505 if (fd < 0) {
506 error_setg_errno(errp, errno, "can't create stream socket");
507 return -1;
509 qemu_set_nonblock(fd);
511 socket_set_fast_reuse(fd);
513 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
514 if (ret < 0) {
515 error_setg_errno(errp, errno, "can't bind ip=%s to socket",
516 inet_ntoa(saddr.sin_addr));
517 closesocket(fd);
518 return -1;
520 ret = listen(fd, 0);
521 if (ret < 0) {
522 error_setg_errno(errp, errno, "can't listen on socket");
523 closesocket(fd);
524 return -1;
527 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
528 s = DO_UPCAST(NetSocketState, nc, nc);
529 s->fd = -1;
530 s->listen_fd = fd;
531 s->nc.link_down = true;
532 net_socket_rs_init(&s->rs, net_socket_rs_finalize, false);
534 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
535 return 0;
538 static int net_socket_connect_init(NetClientState *peer,
539 const char *model,
540 const char *name,
541 const char *host_str,
542 Error **errp)
544 NetSocketState *s;
545 int fd, connected, ret;
546 struct sockaddr_in saddr;
548 if (parse_host_port(&saddr, host_str, errp) < 0) {
549 return -1;
552 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
553 if (fd < 0) {
554 error_setg_errno(errp, errno, "can't create stream socket");
555 return -1;
557 qemu_set_nonblock(fd);
559 connected = 0;
560 for(;;) {
561 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
562 if (ret < 0) {
563 if (errno == EINTR || errno == EWOULDBLOCK) {
564 /* continue */
565 } else if (errno == EINPROGRESS ||
566 errno == EALREADY ||
567 errno == EINVAL) {
568 break;
569 } else {
570 error_setg_errno(errp, errno, "can't connect socket");
571 closesocket(fd);
572 return -1;
574 } else {
575 connected = 1;
576 break;
579 s = net_socket_fd_init(peer, model, name, fd, connected, NULL, errp);
580 if (!s) {
581 return -1;
584 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
585 "socket: connect to %s:%d",
586 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
587 return 0;
590 static int net_socket_mcast_init(NetClientState *peer,
591 const char *model,
592 const char *name,
593 const char *host_str,
594 const char *localaddr_str,
595 Error **errp)
597 NetSocketState *s;
598 int fd;
599 struct sockaddr_in saddr;
600 struct in_addr localaddr, *param_localaddr;
602 if (parse_host_port(&saddr, host_str, errp) < 0) {
603 return -1;
606 if (localaddr_str != NULL) {
607 if (inet_aton(localaddr_str, &localaddr) == 0) {
608 error_setg(errp, "localaddr '%s' is not a valid IPv4 address",
609 localaddr_str);
610 return -1;
612 param_localaddr = &localaddr;
613 } else {
614 param_localaddr = NULL;
617 fd = net_socket_mcast_create(&saddr, param_localaddr, errp);
618 if (fd < 0) {
619 return -1;
622 s = net_socket_fd_init(peer, model, name, fd, 0, NULL, errp);
623 if (!s) {
624 return -1;
627 s->dgram_dst = saddr;
629 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
630 "socket: mcast=%s:%d",
631 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
632 return 0;
636 static int net_socket_udp_init(NetClientState *peer,
637 const char *model,
638 const char *name,
639 const char *rhost,
640 const char *lhost,
641 Error **errp)
643 NetSocketState *s;
644 int fd, ret;
645 struct sockaddr_in laddr, raddr;
647 if (parse_host_port(&laddr, lhost, errp) < 0) {
648 return -1;
651 if (parse_host_port(&raddr, rhost, errp) < 0) {
652 return -1;
655 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
656 if (fd < 0) {
657 error_setg_errno(errp, errno, "can't create datagram socket");
658 return -1;
661 ret = socket_set_fast_reuse(fd);
662 if (ret < 0) {
663 error_setg_errno(errp, errno,
664 "can't set socket option SO_REUSEADDR");
665 closesocket(fd);
666 return -1;
668 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
669 if (ret < 0) {
670 error_setg_errno(errp, errno, "can't bind ip=%s to socket",
671 inet_ntoa(laddr.sin_addr));
672 closesocket(fd);
673 return -1;
675 qemu_set_nonblock(fd);
677 s = net_socket_fd_init(peer, model, name, fd, 0, NULL, errp);
678 if (!s) {
679 return -1;
682 s->dgram_dst = raddr;
684 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
685 "socket: udp=%s:%d",
686 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port));
687 return 0;
690 int net_init_socket(const Netdev *netdev, const char *name,
691 NetClientState *peer, Error **errp)
693 const NetdevSocketOptions *sock;
695 assert(netdev->type == NET_CLIENT_DRIVER_SOCKET);
696 sock = &netdev->u.socket;
698 if (sock->has_fd + sock->has_listen + sock->has_connect + sock->has_mcast +
699 sock->has_udp != 1) {
700 error_setg(errp, "exactly one of listen=, connect=, mcast= or udp="
701 " is required");
702 return -1;
705 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) {
706 error_setg(errp, "localaddr= is only valid with mcast= or udp=");
707 return -1;
710 if (sock->has_fd) {
711 int fd;
713 fd = monitor_fd_param(cur_mon, sock->fd, errp);
714 if (fd == -1) {
715 return -1;
717 qemu_set_nonblock(fd);
718 if (!net_socket_fd_init(peer, "socket", name, fd, 1, sock->mcast,
719 errp)) {
720 return -1;
722 return 0;
725 if (sock->has_listen) {
726 if (net_socket_listen_init(peer, "socket", name, sock->listen, errp)
727 < 0) {
728 return -1;
730 return 0;
733 if (sock->has_connect) {
734 if (net_socket_connect_init(peer, "socket", name, sock->connect, errp)
735 < 0) {
736 return -1;
738 return 0;
741 if (sock->has_mcast) {
742 /* if sock->localaddr is missing, it has been initialized to "all bits
743 * zero" */
744 if (net_socket_mcast_init(peer, "socket", name, sock->mcast,
745 sock->localaddr, errp) < 0) {
746 return -1;
748 return 0;
751 assert(sock->has_udp);
752 if (!sock->has_localaddr) {
753 error_setg(errp, "localaddr= is mandatory with udp=");
754 return -1;
756 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr,
757 errp) < 0) {
758 return -1;
760 return 0;