migration/rdma: Clean up qemu_rdma_poll()'s return type
[qemu/kevin.git] / net / dgram.c
blob48f653bceb257f8a306d0ea39fb6133f679d320d
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 * Copyright (c) 2022 Red Hat, Inc.
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23 * THE SOFTWARE.
26 #include "qemu/osdep.h"
28 #include "net/net.h"
29 #include "clients.h"
30 #include "monitor/monitor.h"
31 #include "qapi/error.h"
32 #include "qemu/error-report.h"
33 #include "qemu/option.h"
34 #include "qemu/sockets.h"
35 #include "qemu/iov.h"
36 #include "qemu/main-loop.h"
37 #include "qemu/cutils.h"
39 typedef struct NetDgramState {
40 NetClientState nc;
41 int fd;
42 SocketReadState rs;
43 bool read_poll; /* waiting to receive data? */
44 bool write_poll; /* waiting to transmit data? */
45 /* contains destination iff connectionless */
46 struct sockaddr *dest_addr;
47 socklen_t dest_len;
48 } NetDgramState;
50 static void net_dgram_send(void *opaque);
51 static void net_dgram_writable(void *opaque);
53 static void net_dgram_update_fd_handler(NetDgramState *s)
55 qemu_set_fd_handler(s->fd,
56 s->read_poll ? net_dgram_send : NULL,
57 s->write_poll ? net_dgram_writable : NULL,
58 s);
61 static void net_dgram_read_poll(NetDgramState *s, bool enable)
63 s->read_poll = enable;
64 net_dgram_update_fd_handler(s);
67 static void net_dgram_write_poll(NetDgramState *s, bool enable)
69 s->write_poll = enable;
70 net_dgram_update_fd_handler(s);
73 static void net_dgram_writable(void *opaque)
75 NetDgramState *s = opaque;
77 net_dgram_write_poll(s, false);
79 qemu_flush_queued_packets(&s->nc);
82 static ssize_t net_dgram_receive(NetClientState *nc,
83 const uint8_t *buf, size_t size)
85 NetDgramState *s = DO_UPCAST(NetDgramState, nc, nc);
86 ssize_t ret;
88 do {
89 if (s->dest_addr) {
90 ret = sendto(s->fd, buf, size, 0, s->dest_addr, s->dest_len);
91 } else {
92 ret = send(s->fd, buf, size, 0);
94 } while (ret == -1 && errno == EINTR);
96 if (ret == -1 && errno == EAGAIN) {
97 net_dgram_write_poll(s, true);
98 return 0;
100 return ret;
103 static void net_dgram_send_completed(NetClientState *nc, ssize_t len)
105 NetDgramState *s = DO_UPCAST(NetDgramState, nc, nc);
107 if (!s->read_poll) {
108 net_dgram_read_poll(s, true);
112 static void net_dgram_rs_finalize(SocketReadState *rs)
114 NetDgramState *s = container_of(rs, NetDgramState, rs);
116 if (qemu_send_packet_async(&s->nc, rs->buf,
117 rs->packet_len,
118 net_dgram_send_completed) == 0) {
119 net_dgram_read_poll(s, false);
123 static void net_dgram_send(void *opaque)
125 NetDgramState *s = opaque;
126 int size;
128 size = recv(s->fd, s->rs.buf, sizeof(s->rs.buf), 0);
129 if (size < 0) {
130 return;
132 if (size == 0) {
133 /* end of connection */
134 net_dgram_read_poll(s, false);
135 net_dgram_write_poll(s, false);
136 return;
138 if (qemu_send_packet_async(&s->nc, s->rs.buf, size,
139 net_dgram_send_completed) == 0) {
140 net_dgram_read_poll(s, false);
144 static int net_dgram_mcast_create(struct sockaddr_in *mcastaddr,
145 struct in_addr *localaddr,
146 Error **errp)
148 struct ip_mreq imr;
149 int fd;
150 int val, ret;
151 #ifdef __OpenBSD__
152 unsigned char loop;
153 #else
154 int loop;
155 #endif
157 if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
158 error_setg(errp, "specified mcastaddr %s (0x%08x) "
159 "does not contain a multicast address",
160 inet_ntoa(mcastaddr->sin_addr),
161 (int)ntohl(mcastaddr->sin_addr.s_addr));
162 return -1;
165 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
166 if (fd < 0) {
167 error_setg_errno(errp, errno, "can't create datagram socket");
168 return -1;
172 * Allow multiple sockets to bind the same multicast ip and port by setting
173 * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set
174 * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR
175 * only on posix systems.
177 val = 1;
178 ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
179 if (ret < 0) {
180 error_setg_errno(errp, errno, "can't set socket option SO_REUSEADDR");
181 goto fail;
184 ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
185 if (ret < 0) {
186 error_setg_errno(errp, errno, "can't bind ip=%s to socket",
187 inet_ntoa(mcastaddr->sin_addr));
188 goto fail;
191 /* Add host to multicast group */
192 imr.imr_multiaddr = mcastaddr->sin_addr;
193 if (localaddr) {
194 imr.imr_interface = *localaddr;
195 } else {
196 imr.imr_interface.s_addr = htonl(INADDR_ANY);
199 ret = setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
200 &imr, sizeof(struct ip_mreq));
201 if (ret < 0) {
202 error_setg_errno(errp, errno,
203 "can't add socket to multicast group %s",
204 inet_ntoa(imr.imr_multiaddr));
205 goto fail;
208 /* Force mcast msgs to loopback (eg. several QEMUs in same host */
209 loop = 1;
210 ret = setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
211 &loop, sizeof(loop));
212 if (ret < 0) {
213 error_setg_errno(errp, errno,
214 "can't force multicast message to loopback");
215 goto fail;
218 /* If a bind address is given, only send packets from that address */
219 if (localaddr != NULL) {
220 ret = setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF,
221 localaddr, sizeof(*localaddr));
222 if (ret < 0) {
223 error_setg_errno(errp, errno,
224 "can't set the default network send interface");
225 goto fail;
229 qemu_socket_set_nonblock(fd);
230 return fd;
231 fail:
232 if (fd >= 0) {
233 close(fd);
235 return -1;
238 static void net_dgram_cleanup(NetClientState *nc)
240 NetDgramState *s = DO_UPCAST(NetDgramState, nc, nc);
241 if (s->fd != -1) {
242 net_dgram_read_poll(s, false);
243 net_dgram_write_poll(s, false);
244 close(s->fd);
245 s->fd = -1;
247 g_free(s->dest_addr);
248 s->dest_addr = NULL;
249 s->dest_len = 0;
252 static NetClientInfo net_dgram_socket_info = {
253 .type = NET_CLIENT_DRIVER_DGRAM,
254 .size = sizeof(NetDgramState),
255 .receive = net_dgram_receive,
256 .cleanup = net_dgram_cleanup,
259 static NetDgramState *net_dgram_fd_init(NetClientState *peer,
260 const char *model,
261 const char *name,
262 int fd,
263 Error **errp)
265 NetClientState *nc;
266 NetDgramState *s;
268 nc = qemu_new_net_client(&net_dgram_socket_info, peer, model, name);
270 s = DO_UPCAST(NetDgramState, nc, nc);
272 s->fd = fd;
273 net_socket_rs_init(&s->rs, net_dgram_rs_finalize, false);
274 net_dgram_read_poll(s, true);
276 return s;
279 static int net_dgram_mcast_init(NetClientState *peer,
280 const char *model,
281 const char *name,
282 SocketAddress *remote,
283 SocketAddress *local,
284 Error **errp)
286 NetDgramState *s;
287 int fd, ret;
288 struct sockaddr_in *saddr;
290 if (remote->type != SOCKET_ADDRESS_TYPE_INET) {
291 error_setg(errp, "multicast only support inet type");
292 return -1;
295 saddr = g_new(struct sockaddr_in, 1);
296 if (convert_host_port(saddr, remote->u.inet.host, remote->u.inet.port,
297 errp) < 0) {
298 g_free(saddr);
299 return -1;
302 if (!local) {
303 fd = net_dgram_mcast_create(saddr, NULL, errp);
304 if (fd < 0) {
305 g_free(saddr);
306 return -1;
308 } else {
309 switch (local->type) {
310 case SOCKET_ADDRESS_TYPE_INET: {
311 struct in_addr localaddr;
313 if (inet_aton(local->u.inet.host, &localaddr) == 0) {
314 g_free(saddr);
315 error_setg(errp, "localaddr '%s' is not a valid IPv4 address",
316 local->u.inet.host);
317 return -1;
320 fd = net_dgram_mcast_create(saddr, &localaddr, errp);
321 if (fd < 0) {
322 g_free(saddr);
323 return -1;
325 break;
327 case SOCKET_ADDRESS_TYPE_FD: {
328 int newfd;
330 fd = monitor_fd_param(monitor_cur(), local->u.fd.str, errp);
331 if (fd == -1) {
332 g_free(saddr);
333 return -1;
335 ret = qemu_socket_try_set_nonblock(fd);
336 if (ret < 0) {
337 g_free(saddr);
338 error_setg_errno(errp, -ret, "%s: Can't use file descriptor %d",
339 name, fd);
340 return -1;
344 * fd passed: multicast: "learn" dest_addr address from bound
345 * address and save it. Because this may be "shared" socket from a
346 * "master" process, datagrams would be recv() by ONLY ONE process:
347 * we must "clone" this dgram socket --jjo
350 saddr = g_new(struct sockaddr_in, 1);
352 if (convert_host_port(saddr, local->u.inet.host, local->u.inet.port,
353 errp) < 0) {
354 g_free(saddr);
355 close(fd);
356 return -1;
359 /* must be bound */
360 if (saddr->sin_addr.s_addr == 0) {
361 error_setg(errp, "can't setup multicast destination address");
362 g_free(saddr);
363 close(fd);
364 return -1;
366 /* clone dgram socket */
367 newfd = net_dgram_mcast_create(saddr, NULL, errp);
368 if (newfd < 0) {
369 g_free(saddr);
370 close(fd);
371 return -1;
373 /* clone newfd to fd, close newfd */
374 dup2(newfd, fd);
375 close(newfd);
376 break;
378 default:
379 g_free(saddr);
380 error_setg(errp, "only support inet or fd type for local");
381 return -1;
385 s = net_dgram_fd_init(peer, model, name, fd, errp);
386 if (!s) {
387 g_free(saddr);
388 return -1;
391 g_assert(s->dest_addr == NULL);
392 s->dest_addr = (struct sockaddr *)saddr;
393 s->dest_len = sizeof(*saddr);
395 if (!local) {
396 qemu_set_info_str(&s->nc, "mcast=%s:%d",
397 inet_ntoa(saddr->sin_addr),
398 ntohs(saddr->sin_port));
399 } else {
400 switch (local->type) {
401 case SOCKET_ADDRESS_TYPE_INET:
402 qemu_set_info_str(&s->nc, "mcast=%s:%d",
403 inet_ntoa(saddr->sin_addr),
404 ntohs(saddr->sin_port));
405 break;
406 case SOCKET_ADDRESS_TYPE_FD:
407 qemu_set_info_str(&s->nc, "fd=%d (cloned mcast=%s:%d)",
408 fd, inet_ntoa(saddr->sin_addr),
409 ntohs(saddr->sin_port));
410 break;
411 default:
412 g_assert_not_reached();
416 return 0;
421 int net_init_dgram(const Netdev *netdev, const char *name,
422 NetClientState *peer, Error **errp)
424 NetDgramState *s;
425 int fd, ret;
426 SocketAddress *remote, *local;
427 struct sockaddr *dest_addr;
428 struct sockaddr_in laddr_in, raddr_in;
429 struct sockaddr_un laddr_un, raddr_un;
430 socklen_t dest_len;
432 assert(netdev->type == NET_CLIENT_DRIVER_DGRAM);
434 remote = netdev->u.dgram.remote;
435 local = netdev->u.dgram.local;
437 /* detect multicast address */
438 if (remote && remote->type == SOCKET_ADDRESS_TYPE_INET) {
439 struct sockaddr_in mcastaddr;
441 if (convert_host_port(&mcastaddr, remote->u.inet.host,
442 remote->u.inet.port, errp) < 0) {
443 return -1;
446 if (IN_MULTICAST(ntohl(mcastaddr.sin_addr.s_addr))) {
447 return net_dgram_mcast_init(peer, "dram", name, remote, local,
448 errp);
452 /* unicast address */
453 if (!local) {
454 error_setg(errp, "dgram requires local= parameter");
455 return -1;
458 if (remote) {
459 if (local->type == SOCKET_ADDRESS_TYPE_FD) {
460 error_setg(errp, "don't set remote with local.fd");
461 return -1;
463 if (remote->type != local->type) {
464 error_setg(errp, "remote and local types must be the same");
465 return -1;
467 } else {
468 if (local->type != SOCKET_ADDRESS_TYPE_FD) {
469 error_setg(errp,
470 "type=inet or type=unix requires remote parameter");
471 return -1;
475 switch (local->type) {
476 case SOCKET_ADDRESS_TYPE_INET:
477 if (convert_host_port(&laddr_in, local->u.inet.host, local->u.inet.port,
478 errp) < 0) {
479 return -1;
482 if (convert_host_port(&raddr_in, remote->u.inet.host,
483 remote->u.inet.port, errp) < 0) {
484 return -1;
487 fd = qemu_socket(PF_INET, SOCK_DGRAM, 0);
488 if (fd < 0) {
489 error_setg_errno(errp, errno, "can't create datagram socket");
490 return -1;
493 ret = socket_set_fast_reuse(fd);
494 if (ret < 0) {
495 error_setg_errno(errp, errno,
496 "can't set socket option SO_REUSEADDR");
497 close(fd);
498 return -1;
500 ret = bind(fd, (struct sockaddr *)&laddr_in, sizeof(laddr_in));
501 if (ret < 0) {
502 error_setg_errno(errp, errno, "can't bind ip=%s to socket",
503 inet_ntoa(laddr_in.sin_addr));
504 close(fd);
505 return -1;
507 qemu_socket_set_nonblock(fd);
509 dest_len = sizeof(raddr_in);
510 dest_addr = g_malloc(dest_len);
511 memcpy(dest_addr, &raddr_in, dest_len);
512 break;
513 case SOCKET_ADDRESS_TYPE_UNIX:
514 ret = unlink(local->u.q_unix.path);
515 if (ret < 0 && errno != ENOENT) {
516 error_setg_errno(errp, errno, "failed to unlink socket %s",
517 local->u.q_unix.path);
518 return -1;
521 laddr_un.sun_family = PF_UNIX;
522 ret = snprintf(laddr_un.sun_path, sizeof(laddr_un.sun_path), "%s",
523 local->u.q_unix.path);
524 if (ret < 0 || ret >= sizeof(laddr_un.sun_path)) {
525 error_setg(errp, "UNIX socket path '%s' is too long",
526 local->u.q_unix.path);
527 error_append_hint(errp, "Path must be less than %zu bytes\n",
528 sizeof(laddr_un.sun_path));
531 raddr_un.sun_family = PF_UNIX;
532 ret = snprintf(raddr_un.sun_path, sizeof(raddr_un.sun_path), "%s",
533 remote->u.q_unix.path);
534 if (ret < 0 || ret >= sizeof(raddr_un.sun_path)) {
535 error_setg(errp, "UNIX socket path '%s' is too long",
536 remote->u.q_unix.path);
537 error_append_hint(errp, "Path must be less than %zu bytes\n",
538 sizeof(raddr_un.sun_path));
541 fd = qemu_socket(PF_UNIX, SOCK_DGRAM, 0);
542 if (fd < 0) {
543 error_setg_errno(errp, errno, "can't create datagram socket");
544 return -1;
547 ret = bind(fd, (struct sockaddr *)&laddr_un, sizeof(laddr_un));
548 if (ret < 0) {
549 error_setg_errno(errp, errno, "can't bind unix=%s to socket",
550 laddr_un.sun_path);
551 close(fd);
552 return -1;
554 qemu_socket_set_nonblock(fd);
556 dest_len = sizeof(raddr_un);
557 dest_addr = g_malloc(dest_len);
558 memcpy(dest_addr, &raddr_un, dest_len);
559 break;
560 case SOCKET_ADDRESS_TYPE_FD:
561 fd = monitor_fd_param(monitor_cur(), local->u.fd.str, errp);
562 if (fd == -1) {
563 return -1;
565 ret = qemu_socket_try_set_nonblock(fd);
566 if (ret < 0) {
567 error_setg_errno(errp, -ret, "%s: Can't use file descriptor %d",
568 name, fd);
569 return -1;
571 dest_addr = NULL;
572 dest_len = 0;
573 break;
574 default:
575 error_setg(errp, "only support inet or fd type for local");
576 return -1;
579 s = net_dgram_fd_init(peer, "dgram", name, fd, errp);
580 if (!s) {
581 return -1;
584 if (remote) {
585 g_assert(s->dest_addr == NULL);
586 s->dest_addr = dest_addr;
587 s->dest_len = dest_len;
590 switch (local->type) {
591 case SOCKET_ADDRESS_TYPE_INET:
592 qemu_set_info_str(&s->nc, "udp=%s:%d/%s:%d",
593 inet_ntoa(laddr_in.sin_addr),
594 ntohs(laddr_in.sin_port),
595 inet_ntoa(raddr_in.sin_addr),
596 ntohs(raddr_in.sin_port));
597 break;
598 case SOCKET_ADDRESS_TYPE_UNIX:
599 qemu_set_info_str(&s->nc, "udp=%s:%s",
600 laddr_un.sun_path, raddr_un.sun_path);
601 break;
602 case SOCKET_ADDRESS_TYPE_FD: {
603 SocketAddress *sa;
604 SocketAddressType sa_type;
606 sa = socket_local_address(fd, errp);
607 if (sa) {
608 sa_type = sa->type;
609 qapi_free_SocketAddress(sa);
611 qemu_set_info_str(&s->nc, "fd=%d %s", fd,
612 SocketAddressType_str(sa_type));
613 } else {
614 qemu_set_info_str(&s->nc, "fd=%d", fd);
616 break;
618 default:
619 g_assert_not_reached();
622 return 0;