block: add accounting for merged requests
[qemu/cris-port.git] / net / socket.c
blob68a93cd7e30b523727f8fec035b851757372c946
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 = sizeof(saddr);
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 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;
402 snprintf(nc->info_str, sizeof(nc->info_str),
403 "socket: fd=%d (cloned mcast=%s:%d)",
404 fd, inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
405 } else {
406 snprintf(nc->info_str, sizeof(nc->info_str),
407 "socket: fd=%d", fd);
410 return s;
412 err:
413 closesocket(fd);
414 return NULL;
417 static void net_socket_connect(void *opaque)
419 NetSocketState *s = opaque;
420 s->send_fn = net_socket_send;
421 net_socket_read_poll(s, true);
424 static NetClientInfo net_socket_info = {
425 .type = NET_CLIENT_OPTIONS_KIND_SOCKET,
426 .size = sizeof(NetSocketState),
427 .receive = net_socket_receive,
428 .cleanup = net_socket_cleanup,
431 static NetSocketState *net_socket_fd_init_stream(NetClientState *peer,
432 const char *model,
433 const char *name,
434 int fd, int is_connected)
436 NetClientState *nc;
437 NetSocketState *s;
439 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
441 snprintf(nc->info_str, sizeof(nc->info_str), "socket: fd=%d", fd);
443 s = DO_UPCAST(NetSocketState, nc, nc);
445 s->fd = fd;
446 s->listen_fd = -1;
448 /* Disable Nagle algorithm on TCP sockets to reduce latency */
449 socket_set_nodelay(fd);
451 if (is_connected) {
452 net_socket_connect(s);
453 } else {
454 qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
456 return s;
459 static NetSocketState *net_socket_fd_init(NetClientState *peer,
460 const char *model, const char *name,
461 int fd, int is_connected)
463 int so_type = -1, optlen=sizeof(so_type);
465 if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
466 (socklen_t *)&optlen)< 0) {
467 fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n",
468 fd);
469 closesocket(fd);
470 return NULL;
472 switch(so_type) {
473 case SOCK_DGRAM:
474 return net_socket_fd_init_dgram(peer, model, name, fd, is_connected);
475 case SOCK_STREAM:
476 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
477 default:
478 /* who knows ... this could be a eg. a pty, do warn and continue as stream */
479 fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
480 return net_socket_fd_init_stream(peer, model, name, fd, is_connected);
482 return NULL;
485 static void net_socket_accept(void *opaque)
487 NetSocketState *s = opaque;
488 struct sockaddr_in saddr;
489 socklen_t len;
490 int fd;
492 for(;;) {
493 len = sizeof(saddr);
494 fd = qemu_accept(s->listen_fd, (struct sockaddr *)&saddr, &len);
495 if (fd < 0 && errno != EINTR) {
496 return;
497 } else if (fd >= 0) {
498 qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
499 break;
503 s->fd = fd;
504 s->nc.link_down = false;
505 net_socket_connect(s);
506 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
507 "socket: connection from %s:%d",
508 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
511 static int net_socket_listen_init(NetClientState *peer,
512 const char *model,
513 const char *name,
514 const char *host_str)
516 NetClientState *nc;
517 NetSocketState *s;
518 struct sockaddr_in saddr;
519 int fd, ret;
521 if (parse_host_port(&saddr, host_str) < 0)
522 return -1;
524 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
525 if (fd < 0) {
526 perror("socket");
527 return -1;
529 qemu_set_nonblock(fd);
531 socket_set_fast_reuse(fd);
533 ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
534 if (ret < 0) {
535 perror("bind");
536 closesocket(fd);
537 return -1;
539 ret = listen(fd, 0);
540 if (ret < 0) {
541 perror("listen");
542 closesocket(fd);
543 return -1;
546 nc = qemu_new_net_client(&net_socket_info, peer, model, name);
547 s = DO_UPCAST(NetSocketState, nc, nc);
548 s->fd = -1;
549 s->listen_fd = fd;
550 s->nc.link_down = true;
552 qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
553 return 0;
556 static int net_socket_connect_init(NetClientState *peer,
557 const char *model,
558 const char *name,
559 const char *host_str)
561 NetSocketState *s;
562 int fd, connected, ret, err;
563 struct sockaddr_in saddr;
565 if (parse_host_port(&saddr, host_str) < 0)
566 return -1;
568 fd = qemu_socket(PF_INET, SOCK_STREAM, 0);
569 if (fd < 0) {
570 perror("socket");
571 return -1;
573 qemu_set_nonblock(fd);
575 connected = 0;
576 for(;;) {
577 ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
578 if (ret < 0) {
579 err = socket_error();
580 if (err == EINTR || err == EWOULDBLOCK) {
581 } else if (err == EINPROGRESS) {
582 break;
583 #ifdef _WIN32
584 } else if (err == WSAEALREADY || err == WSAEINVAL) {
585 break;
586 #endif
587 } else {
588 perror("connect");
589 closesocket(fd);
590 return -1;
592 } else {
593 connected = 1;
594 break;
597 s = net_socket_fd_init(peer, model, name, fd, connected);
598 if (!s)
599 return -1;
600 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
601 "socket: connect to %s:%d",
602 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
603 return 0;
606 static int net_socket_mcast_init(NetClientState *peer,
607 const char *model,
608 const char *name,
609 const char *host_str,
610 const char *localaddr_str)
612 NetSocketState *s;
613 int fd;
614 struct sockaddr_in saddr;
615 struct in_addr localaddr, *param_localaddr;
617 if (parse_host_port(&saddr, host_str) < 0)
618 return -1;
620 if (localaddr_str != NULL) {
621 if (inet_aton(localaddr_str, &localaddr) == 0)
622 return -1;
623 param_localaddr = &localaddr;
624 } else {
625 param_localaddr = NULL;
628 fd = net_socket_mcast_create(&saddr, param_localaddr);
629 if (fd < 0)
630 return -1;
632 s = net_socket_fd_init(peer, model, name, fd, 0);
633 if (!s)
634 return -1;
636 s->dgram_dst = saddr;
638 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
639 "socket: mcast=%s:%d",
640 inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
641 return 0;
645 static int net_socket_udp_init(NetClientState *peer,
646 const char *model,
647 const char *name,
648 const char *rhost,
649 const char *lhost)
651 NetSocketState *s;
652 int fd, ret;
653 struct sockaddr_in laddr, raddr;
655 if (parse_host_port(&laddr, lhost) < 0) {
656 return -1;
659 if (parse_host_port(&raddr, rhost) < 0) {
660 return -1;
663 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
664 if (fd < 0) {
665 perror("socket(PF_INET, SOCK_DGRAM)");
666 return -1;
669 ret = socket_set_fast_reuse(fd);
670 if (ret < 0) {
671 closesocket(fd);
672 return -1;
674 ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
675 if (ret < 0) {
676 perror("bind");
677 closesocket(fd);
678 return -1;
680 qemu_set_nonblock(fd);
682 s = net_socket_fd_init(peer, model, name, fd, 0);
683 if (!s) {
684 return -1;
687 s->dgram_dst = raddr;
689 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
690 "socket: udp=%s:%d",
691 inet_ntoa(raddr.sin_addr), ntohs(raddr.sin_port));
692 return 0;
695 int net_init_socket(const NetClientOptions *opts, const char *name,
696 NetClientState *peer)
698 const NetdevSocketOptions *sock;
700 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_SOCKET);
701 sock = opts->socket;
703 if (sock->has_fd + sock->has_listen + sock->has_connect + sock->has_mcast +
704 sock->has_udp != 1) {
705 error_report("exactly one of fd=, listen=, connect=, mcast= or udp="
706 " is required");
707 return -1;
710 if (sock->has_localaddr && !sock->has_mcast && !sock->has_udp) {
711 error_report("localaddr= is only valid with mcast= or udp=");
712 return -1;
715 if (sock->has_fd) {
716 int fd;
718 fd = monitor_handle_fd_param(cur_mon, sock->fd);
719 if (fd == -1) {
720 return -1;
722 qemu_set_nonblock(fd);
723 if (!net_socket_fd_init(peer, "socket", name, fd, 1)) {
724 return -1;
726 return 0;
729 if (sock->has_listen) {
730 if (net_socket_listen_init(peer, "socket", name, sock->listen) == -1) {
731 return -1;
733 return 0;
736 if (sock->has_connect) {
737 if (net_socket_connect_init(peer, "socket", name, sock->connect) ==
738 -1) {
739 return -1;
741 return 0;
744 if (sock->has_mcast) {
745 /* if sock->localaddr is missing, it has been initialized to "all bits
746 * zero" */
747 if (net_socket_mcast_init(peer, "socket", name, sock->mcast,
748 sock->localaddr) == -1) {
749 return -1;
751 return 0;
754 assert(sock->has_udp);
755 if (!sock->has_localaddr) {
756 error_report("localaddr= is mandatory with udp=");
757 return -1;
759 if (net_socket_udp_init(peer, "socket", name, sock->udp, sock->localaddr) ==
760 -1) {
761 return -1;
763 return 0;