mmu-hash*: Clean up real address calculation
[qemu/agraf.git] / net / socket.c
blob396dc8c0b100ec1a81a2e30d004cfc7cab22e36b
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"
35 typedef struct NetSocketState {
36 NetClientState nc;
37 int listen_fd;
38 int fd;
39 int state; /* 0 = getting length, 1 = getting data */
40 unsigned int index;
41 unsigned int packet_len;
42 unsigned int send_index; /* number of bytes sent (only SOCK_STREAM) */
43 uint8_t buf[4096];
44 struct sockaddr_in dgram_dst; /* contains inet host and port destination iff connectionless (SOCK_DGRAM) */
45 IOHandler *send_fn; /* differs between SOCK_STREAM/SOCK_DGRAM */
46 bool read_poll; /* waiting to receive data? */
47 bool write_poll; /* waiting to transmit data? */
48 } NetSocketState;
50 static void net_socket_accept(void *opaque);
51 static void net_socket_writable(void *opaque);
53 /* Only read packets from socket when peer can receive them */
54 static int net_socket_can_send(void *opaque)
56 NetSocketState *s = opaque;
58 return qemu_can_send_packet(&s->nc);
61 static void net_socket_update_fd_handler(NetSocketState *s)
63 qemu_set_fd_handler2(s->fd,
64 s->read_poll ? net_socket_can_send : NULL,
65 s->read_poll ? s->send_fn : NULL,
66 s->write_poll ? net_socket_writable : NULL,
67 s);
70 static void net_socket_read_poll(NetSocketState *s, bool enable)
72 s->read_poll = enable;
73 net_socket_update_fd_handler(s);
76 static void net_socket_write_poll(NetSocketState *s, bool enable)
78 s->write_poll = enable;
79 net_socket_update_fd_handler(s);
82 static void net_socket_writable(void *opaque)
84 NetSocketState *s = opaque;
86 net_socket_write_poll(s, false);
88 qemu_flush_queued_packets(&s->nc);
91 static ssize_t net_socket_receive(NetClientState *nc, const uint8_t *buf, size_t size)
93 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
94 uint32_t len = htonl(size);
95 struct iovec iov[] = {
97 .iov_base = &len,
98 .iov_len = sizeof(len),
99 }, {
100 .iov_base = (void *)buf,
101 .iov_len = size,
104 size_t remaining;
105 ssize_t ret;
107 remaining = iov_size(iov, 2) - s->send_index;
108 ret = iov_send(s->fd, iov, 2, s->send_index, remaining);
110 if (ret == -1 && errno == EAGAIN) {
111 ret = 0; /* handled further down */
113 if (ret == -1) {
114 s->send_index = 0;
115 return -errno;
117 if (ret < (ssize_t)remaining) {
118 s->send_index += ret;
119 net_socket_write_poll(s, true);
120 return 0;
122 s->send_index = 0;
123 return size;
126 static ssize_t net_socket_receive_dgram(NetClientState *nc, const uint8_t *buf, size_t size)
128 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
129 ssize_t ret;
131 do {
132 ret = qemu_sendto(s->fd, buf, size, 0,
133 (struct sockaddr *)&s->dgram_dst,
134 sizeof(s->dgram_dst));
135 } while (ret == -1 && errno == EINTR);
137 if (ret == -1 && errno == EAGAIN) {
138 net_socket_write_poll(s, true);
139 return 0;
141 return ret;
144 static void net_socket_send(void *opaque)
146 NetSocketState *s = opaque;
147 int size, err;
148 unsigned l;
149 uint8_t buf1[4096];
150 const uint8_t *buf;
152 size = qemu_recv(s->fd, buf1, sizeof(buf1), 0);
153 if (size < 0) {
154 err = socket_error();
155 if (err != EWOULDBLOCK)
156 goto eoc;
157 } else if (size == 0) {
158 /* end of connection */
159 eoc:
160 net_socket_read_poll(s, false);
161 net_socket_write_poll(s, false);
162 if (s->listen_fd != -1) {
163 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
165 closesocket(s->fd);
167 s->fd = -1;
168 s->state = 0;
169 s->index = 0;
170 s->packet_len = 0;
171 s->nc.link_down = true;
172 memset(s->buf, 0, sizeof(s->buf));
173 memset(s->nc.info_str, 0, sizeof(s->nc.info_str));
175 return;
177 buf = buf1;
178 while (size > 0) {
179 /* reassemble a packet from the network */
180 switch(s->state) {
181 case 0:
182 l = 4 - s->index;
183 if (l > size)
184 l = size;
185 memcpy(s->buf + s->index, buf, l);
186 buf += l;
187 size -= l;
188 s->index += l;
189 if (s->index == 4) {
190 /* got length */
191 s->packet_len = ntohl(*(uint32_t *)s->buf);
192 s->index = 0;
193 s->state = 1;
195 break;
196 case 1:
197 l = s->packet_len - s->index;
198 if (l > size)
199 l = size;
200 if (s->index + l <= sizeof(s->buf)) {
201 memcpy(s->buf + s->index, buf, l);
202 } else {
203 fprintf(stderr, "serious error: oversized packet received,"
204 "connection terminated.\n");
205 s->state = 0;
206 goto eoc;
209 s->index += l;
210 buf += l;
211 size -= l;
212 if (s->index >= s->packet_len) {
213 qemu_send_packet(&s->nc, s->buf, s->packet_len);
214 s->index = 0;
215 s->state = 0;
217 break;
222 static void net_socket_send_dgram(void *opaque)
224 NetSocketState *s = opaque;
225 int size;
227 size = qemu_recv(s->fd, s->buf, sizeof(s->buf), 0);
228 if (size < 0)
229 return;
230 if (size == 0) {
231 /* end of connection */
232 net_socket_read_poll(s, false);
233 net_socket_write_poll(s, false);
234 return;
236 qemu_send_packet(&s->nc, s->buf, size);
239 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr, struct in_addr *localaddr)
241 struct ip_mreq imr;
242 int fd;
243 int val, ret;
244 #ifdef __OpenBSD__
245 unsigned char loop;
246 #else
247 int loop;
248 #endif
250 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
251 fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) "
252 "does not contain a multicast address\n",
253 inet_ntoa(mcastaddr->sin_addr),
254 (int)ntohl(mcastaddr->sin_addr.s_addr));
255 return -1;
258 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
259 if (fd < 0) {
260 perror("socket(PF_INET, SOCK_DGRAM)");
261 return -1;
264 val = 1;
265 ret=setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
266 (const char *)&val, sizeof(val));
267 if (ret < 0) {
268 perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
269 goto fail;
272 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
273 if (ret < 0) {
274 perror("bind");
275 goto fail;
278 /* Add host to multicast group */
279 imr.imr_multiaddr = mcastaddr->sin_addr;
280 if (localaddr) {
281 imr.imr_interface = *localaddr;
282 } else {
283 imr.imr_interface.s_addr = htonl(INADDR_ANY);
286 ret = setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
287 (const char *)&imr, sizeof(struct ip_mreq));
288 if (ret < 0) {
289 perror("setsockopt(IP_ADD_MEMBERSHIP)");
290 goto fail;
293 /* Force mcast msgs to loopback (eg. several QEMUs in same host */
294 loop = 1;
295 ret=setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
296 (const char *)&loop, sizeof(loop));
297 if (ret < 0) {
298 perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)");
299 goto fail;
302 /* If a bind address is given, only send packets from that address */
303 if (localaddr != NULL) {
304 ret = setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF,
305 (const char *)localaddr, sizeof(*localaddr));
306 if (ret < 0) {
307 perror("setsockopt(IP_MULTICAST_IF)");
308 goto fail;
312 socket_set_nonblock(fd);
313 return fd;
314 fail:
315 if (fd >= 0)
316 closesocket(fd);
317 return -1;
320 static void net_socket_cleanup(NetClientState *nc)
322 NetSocketState *s = DO_UPCAST(NetSocketState, nc, nc);
323 if (s->fd != -1) {
324 net_socket_read_poll(s, false);
325 net_socket_write_poll(s, false);
326 close(s->fd);
327 s->fd = -1;
329 if (s->listen_fd != -1) {
330 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
331 closesocket(s->listen_fd);
332 s->listen_fd = -1;
336 static NetClientInfo net_dgram_socket_info = {
337 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
338 .size = sizeof(NetSocketState),
339 .receive = net_socket_receive_dgram,
340 .cleanup = net_socket_cleanup,
343 static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer,
344 const char *model,
345 const char *name,
346 int fd, int is_connected)
348 struct sockaddr_in saddr;
349 int newfd;
350 socklen_t saddr_len;
351 NetClientState *nc;
352 NetSocketState *s;
354 /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
355 * Because this may be "shared" socket from a "master" process, datagrams would be recv()
356 * by ONLY ONE process: we must "clone" this dgram socket --jjo
359 if (is_connected) {
360 if (getsockname(fd, (struct sockaddr *) &saddr, &saddr_len) == 0) {
361 /* must be bound */
362 if (saddr.sin_addr.s_addr == 0) {
363 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, "
364 "cannot setup multicast dst addr\n", fd);
365 goto err;
367 /* clone dgram socket */
368 newfd = net_socket_mcast_create(&saddr, NULL);
369 if (newfd < 0) {
370 /* error already reported by net_socket_mcast_create() */
371 goto err;
373 /* clone newfd to fd, close newfd */
374 dup2(newfd, fd);
375 close(newfd);
377 } else {
378 fprintf(stderr,
379 "qemu: error: init_dgram: fd=%d failed getsockname(): %s\n",
380 fd, strerror(errno));
381 goto err;
385 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name);
387 snprintf(nc->info_str, sizeof(nc->info_str),
388 "socket: fd=%d (%s mcast=%s:%d)",
389 fd, is_connected ? "cloned" : "",
390 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
392 s = DO_UPCAST(NetSocketState, nc, nc);
394 s->fd = fd;
395 s->listen_fd = -1;
396 s->send_fn = net_socket_send_dgram;
397 net_socket_read_poll(s, true);
399 /* mcast: save bound address as dst */
400 if (is_connected) {
401 s->dgram_dst = saddr;
404 return s;
406 err:
407 closesocket(fd);
408 return NULL;
411 static void net_socket_connect(void *opaque)
413 NetSocketState *s = opaque;
414 s->send_fn = net_socket_send;
415 net_socket_read_poll(s, true);
418 static NetClientInfo net_socket_info = {
419 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
420 .size = sizeof(NetSocketState),
421 .receive = net_socket_receive,
422 .cleanup = net_socket_cleanup,
425 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer,
426 const char *model,
427 const char *name,
428 int fd, int is_connected)
430 NetClientState *nc;
431 NetSocketState *s;
433 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
435 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd);
437 s = DO_UPCAST(NetSocketState, nc, nc);
439 s->fd = fd;
440 s->listen_fd = -1;
442 if (is_connected) {
443 net_socket_connect(s);
444 } else {
445 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
447 return s;
450 static NetSocketState *net_socket_fd_init(NetClientState *peer,
451 const char *model, const char *name,
452 int fd, int is_connected)
454 int so_type = -1, optlen=sizeof(so_type);
456 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
457 (socklen_t *)&optlen)< 0) {
458 fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n",
459 fd);
460 closesocket(fd);
461 return NULL;
463 switch(so_type) {
464 case SOCK_DGRAM:
465 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected);
466 case SOCK_STREAM:
467 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
468 default:
469 /* who knows ... this could be a eg. a pty, do warn and continue as stream */
470 fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
471 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
473 return NULL;
476 static void net_socket_accept(void *opaque)
478 NetSocketState *s = opaque;
479 struct sockaddr_in saddr;
480 socklen_t len;
481 int fd;
483 for(;;) {
484 len = sizeof(saddr);
485 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len);
486 if (fd < 0 && errno != EINTR) {
487 return;
488 } else if (fd >= 0) {
489 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
490 break;
494 s->fd = fd;
495 s->nc.link_down = false;
496 net_socket_connect(s);
497 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
498 "socket: connection from %s:%d",
499 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
502 static int net_socket_listen_init(NetClientState *peer,
503 const char *model,
504 const char *name,
505 const char *host_str)
507 NetClientState *nc;
508 NetSocketState *s;
509 struct sockaddr_in saddr;
510 int fd, val, ret;
512 if (parse_host_port(&saddr, host_str) < 0)
513 return -1;
515 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
516 if (fd < 0) {
517 perror("socket");
518 return -1;
520 socket_set_nonblock(fd);
522 /* allow fast reuse */
523 val = 1;
524 setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const char *)&val, sizeof(val));
526 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
527 if (ret < 0) {
528 perror("bind");
529 closesocket(fd);
530 return -1;
532 ret = listen(fd, 0);
533 if (ret < 0) {
534 perror("listen");
535 closesocket(fd);
536 return -1;
539 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
540 s = DO_UPCAST(NetSocketState, nc, nc);
541 s->fd = -1;
542 s->listen_fd = fd;
543 s->nc.link_down = true;
545 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
546 return 0;
549 static int net_socket_connect_init(NetClientState *peer,
550 const char *model,
551 const char *name,
552 const char *host_str)
554 NetSocketState *s;
555 int fd, connected, ret, err;
556 struct sockaddr_in saddr;
558 if (parse_host_port(&saddr, host_str) < 0)
559 return -1;
561 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
562 if (fd < 0) {
563 perror("socket");
564 return -1;
566 socket_set_nonblock(fd);
568 connected = 0;
569 for(;;) {
570 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
571 if (ret < 0) {
572 err = socket_error();
573 if (err == EINTR || err == EWOULDBLOCK) {
574 } else if (err == EINPROGRESS) {
575 break;
576 #ifdef _WIN32
577 } else if (err == WSAEALREADY || err == WSAEINVAL) {
578 break;
579 #endif
580 } else {
581 perror("connect");
582 closesocket(fd);
583 return -1;
585 } else {
586 connected = 1;
587 break;
590 s = net_socket_fd_init(peer, model, name, fd, connected);
591 if (!s)
592 return -1;
593 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
594 "socket: connect to %s:%d",
595 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
596 return 0;
599 static int net_socket_mcast_init(NetClientState *peer,
600 const char *model,
601 const char *name,
602 const char *host_str,
603 const char *localaddr_str)
605 NetSocketState *s;
606 int fd;
607 struct sockaddr_in saddr;
608 struct in_addr localaddr, *param_localaddr;
610 if (parse_host_port(&saddr, host_str) < 0)
611 return -1;
613 if (localaddr_str != NULL) {
614 if (inet_aton(localaddr_str, &localaddr) == 0)
615 return -1;
616 param_localaddr = &localaddr;
617 } else {
618 param_localaddr = NULL;
621 fd = net_socket_mcast_create(&saddr, param_localaddr);
622 if (fd < 0)
623 return -1;
625 s = net_socket_fd_init(peer, model, name, fd, 0);
626 if (!s)
627 return -1;
629 s->dgram_dst = saddr;
631 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
632 "socket: mcast=%s:%d",
633 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
634 return 0;
638 static int net_socket_udp_init(NetClientState *peer,
639 const char *model,
640 const char *name,
641 const char *rhost,
642 const char *lhost)
644 NetSocketState *s;
645 int fd, val, ret;
646 struct sockaddr_in laddr, raddr;
648 if (parse_host_port(&laddr, lhost) < 0) {
649 return -1;
652 if (parse_host_port(&raddr, rhost) < 0) {
653 return -1;
656 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
657 if (fd < 0) {
658 perror("socket(PF_INET, SOCK_DGRAM)");
659 return -1;
661 val = 1;
662 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
663 (const char *)&val, sizeof(val));
664 if (ret < 0) {
665 perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
666 closesocket(fd);
667 return -1;
669 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
670 if (ret < 0) {
671 perror("bind");
672 closesocket(fd);
673 return -1;
676 s = net_socket_fd_init(peer, model, name, fd, 0);
677 if (!s) {
678 return -1;
681 s->dgram_dst = raddr;
683 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
684 "socket: udp=%s:%d",
685 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port));
686 return 0;
689 int net_init_socket(const NetClientOptions *opts, const char *name,
690 NetClientState *peer)
692 const NetdevSocketOptions *sock;
694 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_SOCKET);
695 sock = opts->socket;
697 if (sock->has_fd + sock->has_listen + sock->has_connect + sock->has_mcast +
698 sock->has_udp != 1) {
699 error_report("exactly one of fd=, listen=, connect=, mcast= or udp="
700 " is required");
701 return -1;
704 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) {
705 error_report("localaddr= is only valid with mcast= or udp=");
706 return -1;
709 if (sock->has_fd) {
710 int fd;
712 fd = monitor_handle_fd_param(cur_mon, sock->fd);
713 if (fd == -1 || !net_socket_fd_init(peer, "socket", name, fd, 1)) {
714 return -1;
716 return 0;
719 if (sock->has_listen) {
720 if (net_socket_listen_init(peer, "socket", name, sock->listen) == -1) {
721 return -1;
723 return 0;
726 if (sock->has_connect) {
727 if (net_socket_connect_init(peer, "socket", name, sock->connect) ==
728 -1) {
729 return -1;
731 return 0;
734 if (sock->has_mcast) {
735 /* if sock->localaddr is missing, it has been initialized to "all bits
736 * zero" */
737 if (net_socket_mcast_init(peer, "socket", name, sock->mcast,
738 sock->localaddr) == -1) {
739 return -1;
741 return 0;
744 assert(sock->has_udp);
745 if (!sock->has_localaddr) {
746 error_report("localaddr= is mandatory with udp=");
747 return -1;
749 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr) ==
750 -1) {
751 return -1;
753 return 0;