dump: allow target to set the page size
[qemu/ar7.git] / net / socket.c
blobe8605d4ded9c946537ce7567d43fe86055310bf2
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 static void net_socket_update_fd_handler(NetSocketState *s)
56 qemu_set_fd_handler(s->fd,
57 s->read_poll ? s->send_fn : NULL,
58 s->write_poll ? net_socket_writable : NULL,
59 s);
62 static void net_socket_read_poll(NetSocketState *s, bool enable)
64 s->read_poll = enable;
65 net_socket_update_fd_handler(s);
68 static void net_socket_write_poll(NetSocketState *s, bool enable)
70 s->write_poll = enable;
71 net_socket_update_fd_handler(s);
74 static void net_socket_writable(void *opaque)
76 NetSocketState *s = opaque;
78 net_socket_write_poll(s, false);
80 qemu_flush_queued_packets(&s->nc);
83 static ssize_t net_socket_receive(NetClientState *nc, const uint8_t *buf, size_t size)
85 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
86 uint32_t len = htonl(size);
87 struct iovec iov[] = {
89 .iov_base = &len,
90 .iov_len = sizeof(len),
91 }, {
92 .iov_base = (void *)buf,
93 .iov_len = size,
96 size_t remaining;
97 ssize_t ret;
99 remaining = iov_size(iov, 2) - s->send_index;
100 ret = iov_send(s->fd, iov, 2, s->send_index, remaining);
102 if (ret == -1 && errno == EAGAIN) {
103 ret = 0; /* handled further down */
105 if (ret == -1) {
106 s->send_index = 0;
107 return -errno;
109 if (ret < (ssize_t)remaining) {
110 s->send_index += ret;
111 net_socket_write_poll(s, true);
112 return 0;
114 s->send_index = 0;
115 return size;
118 static ssize_t net_socket_receive_dgram(NetClientState *nc, const uint8_t *buf, size_t size)
120 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
121 ssize_t ret;
123 do {
124 ret = qemu_sendto(s->fd, buf, size, 0,
125 (struct sockaddr *)&s->dgram_dst,
126 sizeof(s->dgram_dst));
127 } while (ret == -1 && errno == EINTR);
129 if (ret == -1 && errno == EAGAIN) {
130 net_socket_write_poll(s, true);
131 return 0;
133 return ret;
136 static void net_socket_send_completed(NetClientState *nc, ssize_t len)
138 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
140 if (!s->read_poll) {
141 net_socket_read_poll(s, true);
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 s->index = 0;
215 s->state = 0;
216 if (qemu_send_packet_async(&s->nc, s->buf, s->packet_len,
217 net_socket_send_completed) == 0) {
218 net_socket_read_poll(s, false);
219 break;
222 break;
227 static void net_socket_send_dgram(void *opaque)
229 NetSocketState *s = opaque;
230 int size;
232 size = qemu_recv(s->fd, s->buf, sizeof(s->buf), 0);
233 if (size < 0)
234 return;
235 if (size == 0) {
236 /* end of connection */
237 net_socket_read_poll(s, false);
238 net_socket_write_poll(s, false);
239 return;
241 if (qemu_send_packet_async(&s->nc, s->buf, size,
242 net_socket_send_completed) == 0) {
243 net_socket_read_poll(s, false);
247 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr, struct in_addr *localaddr)
249 struct ip_mreq imr;
250 int fd;
251 int val, ret;
252 #ifdef __OpenBSD__
253 unsigned char loop;
254 #else
255 int loop;
256 #endif
258 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
259 fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) "
260 "does not contain a multicast address\n",
261 inet_ntoa(mcastaddr->sin_addr),
262 (int)ntohl(mcastaddr->sin_addr.s_addr));
263 return -1;
266 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
267 if (fd < 0) {
268 perror("socket(PF_INET, SOCK_DGRAM)");
269 return -1;
272 /* Allow multiple sockets to bind the same multicast ip and port by setting
273 * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set
274 * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR
275 * only on posix systems.
277 val = 1;
278 ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
279 if (ret < 0) {
280 perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
281 goto fail;
284 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
285 if (ret < 0) {
286 perror("bind");
287 goto fail;
290 /* Add host to multicast group */
291 imr.imr_multiaddr = mcastaddr->sin_addr;
292 if (localaddr) {
293 imr.imr_interface = *localaddr;
294 } else {
295 imr.imr_interface.s_addr = htonl(INADDR_ANY);
298 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
299 &imr, sizeof(struct ip_mreq));
300 if (ret < 0) {
301 perror("setsockopt(IP_ADD_MEMBERSHIP)");
302 goto fail;
305 /* Force mcast msgs to loopback (eg. several QEMUs in same host */
306 loop = 1;
307 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
308 &loop, sizeof(loop));
309 if (ret < 0) {
310 perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)");
311 goto fail;
314 /* If a bind address is given, only send packets from that address */
315 if (localaddr != NULL) {
316 ret = qemu_setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF,
317 localaddr, sizeof(*localaddr));
318 if (ret < 0) {
319 perror("setsockopt(IP_MULTICAST_IF)");
320 goto fail;
324 qemu_set_nonblock(fd);
325 return fd;
326 fail:
327 if (fd >= 0)
328 closesocket(fd);
329 return -1;
332 static void net_socket_cleanup(NetClientState *nc)
334 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
335 if (s->fd != -1) {
336 net_socket_read_poll(s, false);
337 net_socket_write_poll(s, false);
338 close(s->fd);
339 s->fd = -1;
341 if (s->listen_fd != -1) {
342 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
343 closesocket(s->listen_fd);
344 s->listen_fd = -1;
348 static NetClientInfo net_dgram_socket_info = {
349 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
350 .size = sizeof(NetSocketState),
351 .receive = net_socket_receive_dgram,
352 .cleanup = net_socket_cleanup,
355 static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer,
356 const char *model,
357 const char *name,
358 int fd, int is_connected)
360 struct sockaddr_in saddr;
361 int newfd;
362 socklen_t saddr_len = sizeof(saddr);
363 NetClientState *nc;
364 NetSocketState *s;
366 /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
367 * Because this may be "shared" socket from a "master" process, datagrams would be recv()
368 * by ONLY ONE process: we must "clone" this dgram socket --jjo
371 if (is_connected) {
372 if (getsockname(fd, (struct sockaddr *) &saddr, &saddr_len) == 0) {
373 /* must be bound */
374 if (saddr.sin_addr.s_addr == 0) {
375 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, "
376 "cannot setup multicast dst addr\n", fd);
377 goto err;
379 /* clone dgram socket */
380 newfd = net_socket_mcast_create(&saddr, NULL);
381 if (newfd < 0) {
382 /* error already reported by net_socket_mcast_create() */
383 goto err;
385 /* clone newfd to fd, close newfd */
386 dup2(newfd, fd);
387 close(newfd);
389 } else {
390 fprintf(stderr,
391 "qemu: error: init_dgram: fd=%d failed getsockname(): %s\n",
392 fd, strerror(errno));
393 goto err;
397 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name);
399 s = DO_UPCAST(NetSocketState, nc, nc);
401 s->fd = fd;
402 s->listen_fd = -1;
403 s->send_fn = net_socket_send_dgram;
404 net_socket_read_poll(s, true);
406 /* mcast: save bound address as dst */
407 if (is_connected) {
408 s->dgram_dst = saddr;
409 snprintf(nc->info_str, sizeof(nc->info_str),
410 "socket: fd=%d (cloned mcast=%s:%d)",
411 fd, inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
412 } else {
413 snprintf(nc->info_str, sizeof(nc->info_str),
414 "socket: fd=%d", fd);
417 return s;
419 err:
420 closesocket(fd);
421 return NULL;
424 static void net_socket_connect(void *opaque)
426 NetSocketState *s = opaque;
427 s->send_fn = net_socket_send;
428 net_socket_read_poll(s, true);
431 static NetClientInfo net_socket_info = {
432 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
433 .size = sizeof(NetSocketState),
434 .receive = net_socket_receive,
435 .cleanup = net_socket_cleanup,
438 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer,
439 const char *model,
440 const char *name,
441 int fd, int is_connected)
443 NetClientState *nc;
444 NetSocketState *s;
446 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
448 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd);
450 s = DO_UPCAST(NetSocketState, nc, nc);
452 s->fd = fd;
453 s->listen_fd = -1;
455 /* Disable Nagle algorithm on TCP sockets to reduce latency */
456 socket_set_nodelay(fd);
458 if (is_connected) {
459 net_socket_connect(s);
460 } else {
461 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
463 return s;
466 static NetSocketState *net_socket_fd_init(NetClientState *peer,
467 const char *model, const char *name,
468 int fd, int is_connected)
470 int so_type = -1, optlen=sizeof(so_type);
472 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
473 (socklen_t *)&optlen)< 0) {
474 fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n",
475 fd);
476 closesocket(fd);
477 return NULL;
479 switch(so_type) {
480 case SOCK_DGRAM:
481 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected);
482 case SOCK_STREAM:
483 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
484 default:
485 /* who knows ... this could be a eg. a pty, do warn and continue as stream */
486 fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
487 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
489 return NULL;
492 static void net_socket_accept(void *opaque)
494 NetSocketState *s = opaque;
495 struct sockaddr_in saddr;
496 socklen_t len;
497 int fd;
499 for(;;) {
500 len = sizeof(saddr);
501 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len);
502 if (fd < 0 && errno != EINTR) {
503 return;
504 } else if (fd >= 0) {
505 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
506 break;
510 s->fd = fd;
511 s->nc.link_down = false;
512 net_socket_connect(s);
513 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
514 "socket: connection from %s:%d",
515 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
518 static int net_socket_listen_init(NetClientState *peer,
519 const char *model,
520 const char *name,
521 const char *host_str)
523 NetClientState *nc;
524 NetSocketState *s;
525 struct sockaddr_in saddr;
526 int fd, ret;
528 if (parse_host_port(&saddr, host_str) < 0)
529 return -1;
531 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
532 if (fd < 0) {
533 perror("socket");
534 return -1;
536 qemu_set_nonblock(fd);
538 socket_set_fast_reuse(fd);
540 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
541 if (ret < 0) {
542 perror("bind");
543 closesocket(fd);
544 return -1;
546 ret = listen(fd, 0);
547 if (ret < 0) {
548 perror("listen");
549 closesocket(fd);
550 return -1;
553 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
554 s = DO_UPCAST(NetSocketState, nc, nc);
555 s->fd = -1;
556 s->listen_fd = fd;
557 s->nc.link_down = true;
559 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
560 return 0;
563 static int net_socket_connect_init(NetClientState *peer,
564 const char *model,
565 const char *name,
566 const char *host_str)
568 NetSocketState *s;
569 int fd, connected, ret, err;
570 struct sockaddr_in saddr;
572 if (parse_host_port(&saddr, host_str) < 0)
573 return -1;
575 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
576 if (fd < 0) {
577 perror("socket");
578 return -1;
580 qemu_set_nonblock(fd);
582 connected = 0;
583 for(;;) {
584 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
585 if (ret < 0) {
586 err = socket_error();
587 if (err == EINTR || err == EWOULDBLOCK) {
588 } else if (err == EINPROGRESS) {
589 break;
590 #ifdef _WIN32
591 } else if (err == WSAEALREADY || err == WSAEINVAL) {
592 break;
593 #endif
594 } else {
595 perror("connect");
596 closesocket(fd);
597 return -1;
599 } else {
600 connected = 1;
601 break;
604 s = net_socket_fd_init(peer, model, name, fd, connected);
605 if (!s)
606 return -1;
607 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
608 "socket: connect to %s:%d",
609 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
610 return 0;
613 static int net_socket_mcast_init(NetClientState *peer,
614 const char *model,
615 const char *name,
616 const char *host_str,
617 const char *localaddr_str)
619 NetSocketState *s;
620 int fd;
621 struct sockaddr_in saddr;
622 struct in_addr localaddr, *param_localaddr;
624 if (parse_host_port(&saddr, host_str) < 0)
625 return -1;
627 if (localaddr_str != NULL) {
628 if (inet_aton(localaddr_str, &localaddr) == 0)
629 return -1;
630 param_localaddr = &localaddr;
631 } else {
632 param_localaddr = NULL;
635 fd = net_socket_mcast_create(&saddr, param_localaddr);
636 if (fd < 0)
637 return -1;
639 s = net_socket_fd_init(peer, model, name, fd, 0);
640 if (!s)
641 return -1;
643 s->dgram_dst = saddr;
645 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
646 "socket: mcast=%s:%d",
647 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
648 return 0;
652 static int net_socket_udp_init(NetClientState *peer,
653 const char *model,
654 const char *name,
655 const char *rhost,
656 const char *lhost)
658 NetSocketState *s;
659 int fd, ret;
660 struct sockaddr_in laddr, raddr;
662 if (parse_host_port(&laddr, lhost) < 0) {
663 return -1;
666 if (parse_host_port(&raddr, rhost) < 0) {
667 return -1;
670 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
671 if (fd < 0) {
672 perror("socket(PF_INET, SOCK_DGRAM)");
673 return -1;
676 ret = socket_set_fast_reuse(fd);
677 if (ret < 0) {
678 closesocket(fd);
679 return -1;
681 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
682 if (ret < 0) {
683 perror("bind");
684 closesocket(fd);
685 return -1;
687 qemu_set_nonblock(fd);
689 s = net_socket_fd_init(peer, model, name, fd, 0);
690 if (!s) {
691 return -1;
694 s->dgram_dst = raddr;
696 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
697 "socket: udp=%s:%d",
698 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port));
699 return 0;
702 int net_init_socket(const NetClientOptions *opts, const char *name,
703 NetClientState *peer, Error **errp)
705 /* FIXME error_setg(errp, ...) on failure */
706 Error *err = NULL;
707 const NetdevSocketOptions *sock;
709 assert(opts->type == NET_CLIENT_OPTIONS_KIND_SOCKET);
710 sock = opts->u.socket;
712 if (sock->has_fd + sock->has_listen + sock->has_connect + sock->has_mcast +
713 sock->has_udp != 1) {
714 error_report("exactly one of fd=, listen=, connect=, mcast= or udp="
715 " is required");
716 return -1;
719 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) {
720 error_report("localaddr= is only valid with mcast= or udp=");
721 return -1;
724 if (sock->has_fd) {
725 int fd;
727 fd = monitor_fd_param(cur_mon, sock->fd, &err);
728 if (fd == -1) {
729 error_report_err(err);
730 return -1;
732 qemu_set_nonblock(fd);
733 if (!net_socket_fd_init(peer, "socket", name, fd, 1)) {
734 return -1;
736 return 0;
739 if (sock->has_listen) {
740 if (net_socket_listen_init(peer, "socket", name, sock->listen) == -1) {
741 return -1;
743 return 0;
746 if (sock->has_connect) {
747 if (net_socket_connect_init(peer, "socket", name, sock->connect) ==
748 -1) {
749 return -1;
751 return 0;
754 if (sock->has_mcast) {
755 /* if sock->localaddr is missing, it has been initialized to "all bits
756 * zero" */
757 if (net_socket_mcast_init(peer, "socket", name, sock->mcast,
758 sock->localaddr) == -1) {
759 return -1;
761 return 0;
764 assert(sock->has_udp);
765 if (!sock->has_localaddr) {
766 error_report("localaddr= is mandatory with udp=");
767 return -1;
769 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr) ==
770 -1) {
771 return -1;
773 return 0;