xref: /qemu/net/net.c (revision 93e9d730)
1 /*
2  * QEMU System Emulator
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
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:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
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.
23  */
24 
25 #include "qemu/osdep.h"
26 
27 #include "net/net.h"
28 #include "clients.h"
29 #include "hub.h"
30 #include "hw/qdev-properties.h"
31 #include "net/slirp.h"
32 #include "net/eth.h"
33 #include "util.h"
34 
35 #include "monitor/monitor.h"
36 #include "qemu/help_option.h"
37 #include "qapi/qapi-commands-net.h"
38 #include "qapi/qapi-visit-net.h"
39 #include "qapi/qmp/qdict.h"
40 #include "qapi/qmp/qerror.h"
41 #include "qemu/error-report.h"
42 #include "qemu/sockets.h"
43 #include "qemu/cutils.h"
44 #include "qemu/config-file.h"
45 #include "qemu/ctype.h"
46 #include "qemu/id.h"
47 #include "qemu/iov.h"
48 #include "qemu/qemu-print.h"
49 #include "qemu/main-loop.h"
50 #include "qemu/option.h"
51 #include "qemu/keyval.h"
52 #include "qapi/error.h"
53 #include "qapi/opts-visitor.h"
54 #include "sysemu/runstate.h"
55 #include "net/colo-compare.h"
56 #include "net/filter.h"
57 #include "qapi/string-output-visitor.h"
58 #include "qapi/qobject-input-visitor.h"
59 
60 /* Net bridge is currently not supported for W32. */
61 #if !defined(_WIN32)
62 # define CONFIG_NET_BRIDGE
63 #endif
64 
65 static VMChangeStateEntry *net_change_state_entry;
66 NetClientStateList net_clients;
67 
68 typedef struct NetdevQueueEntry {
69     Netdev *nd;
70     Location loc;
71     QSIMPLEQ_ENTRY(NetdevQueueEntry) entry;
72 } NetdevQueueEntry;
73 
74 typedef QSIMPLEQ_HEAD(, NetdevQueueEntry) NetdevQueue;
75 
76 static NetdevQueue nd_queue = QSIMPLEQ_HEAD_INITIALIZER(nd_queue);
77 
78 /***********************************************************/
79 /* network device redirectors */
80 
81 int convert_host_port(struct sockaddr_in *saddr, const char *host,
82                       const char *port, Error **errp)
83 {
84     struct hostent *he;
85     const char *r;
86     long p;
87 
88     memset(saddr, 0, sizeof(*saddr));
89 
90     saddr->sin_family = AF_INET;
91     if (host[0] == '\0') {
92         saddr->sin_addr.s_addr = 0;
93     } else {
94         if (qemu_isdigit(host[0])) {
95             if (!inet_aton(host, &saddr->sin_addr)) {
96                 error_setg(errp, "host address '%s' is not a valid "
97                            "IPv4 address", host);
98                 return -1;
99             }
100         } else {
101             he = gethostbyname(host);
102             if (he == NULL) {
103                 error_setg(errp, "can't resolve host address '%s'", host);
104                 return -1;
105             }
106             saddr->sin_addr = *(struct in_addr *)he->h_addr;
107         }
108     }
109     if (qemu_strtol(port, &r, 0, &p) != 0) {
110         error_setg(errp, "port number '%s' is invalid", port);
111         return -1;
112     }
113     saddr->sin_port = htons(p);
114     return 0;
115 }
116 
117 int parse_host_port(struct sockaddr_in *saddr, const char *str,
118                     Error **errp)
119 {
120     gchar **substrings;
121     int ret;
122 
123     substrings = g_strsplit(str, ":", 2);
124     if (!substrings || !substrings[0] || !substrings[1]) {
125         error_setg(errp, "host address '%s' doesn't contain ':' "
126                    "separating host from port", str);
127         ret = -1;
128         goto out;
129     }
130 
131     ret = convert_host_port(saddr, substrings[0], substrings[1], errp);
132 
133 out:
134     g_strfreev(substrings);
135     return ret;
136 }
137 
138 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
139 {
140     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
141                            macaddr[0], macaddr[1], macaddr[2],
142                            macaddr[3], macaddr[4], macaddr[5]);
143 }
144 
145 void qemu_set_info_str(NetClientState *nc, const char *fmt, ...)
146 {
147     va_list ap;
148 
149     va_start(ap, fmt);
150     vsnprintf(nc->info_str, sizeof(nc->info_str), fmt, ap);
151     va_end(ap);
152 }
153 
154 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
155 {
156     qemu_set_info_str(nc, "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
157                       nc->model, macaddr[0], macaddr[1], macaddr[2],
158                       macaddr[3], macaddr[4], macaddr[5]);
159 }
160 
161 static int mac_table[256] = {0};
162 
163 static void qemu_macaddr_set_used(MACAddr *macaddr)
164 {
165     int index;
166 
167     for (index = 0x56; index < 0xFF; index++) {
168         if (macaddr->a[5] == index) {
169             mac_table[index]++;
170         }
171     }
172 }
173 
174 static void qemu_macaddr_set_free(MACAddr *macaddr)
175 {
176     int index;
177     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
178 
179     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
180         return;
181     }
182     for (index = 0x56; index < 0xFF; index++) {
183         if (macaddr->a[5] == index) {
184             mac_table[index]--;
185         }
186     }
187 }
188 
189 static int qemu_macaddr_get_free(void)
190 {
191     int index;
192 
193     for (index = 0x56; index < 0xFF; index++) {
194         if (mac_table[index] == 0) {
195             return index;
196         }
197     }
198 
199     return -1;
200 }
201 
202 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
203 {
204     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
205     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
206 
207     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
208         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
209             return;
210         } else {
211             qemu_macaddr_set_used(macaddr);
212             return;
213         }
214     }
215 
216     macaddr->a[0] = 0x52;
217     macaddr->a[1] = 0x54;
218     macaddr->a[2] = 0x00;
219     macaddr->a[3] = 0x12;
220     macaddr->a[4] = 0x34;
221     macaddr->a[5] = qemu_macaddr_get_free();
222     qemu_macaddr_set_used(macaddr);
223 }
224 
225 /**
226  * Generate a name for net client
227  *
228  * Only net clients created with the legacy -net option and NICs need this.
229  */
230 static char *assign_name(NetClientState *nc1, const char *model)
231 {
232     NetClientState *nc;
233     int id = 0;
234 
235     QTAILQ_FOREACH(nc, &net_clients, next) {
236         if (nc == nc1) {
237             continue;
238         }
239         if (strcmp(nc->model, model) == 0) {
240             id++;
241         }
242     }
243 
244     return g_strdup_printf("%s.%d", model, id);
245 }
246 
247 static void qemu_net_client_destructor(NetClientState *nc)
248 {
249     g_free(nc);
250 }
251 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
252                                        unsigned flags,
253                                        const struct iovec *iov,
254                                        int iovcnt,
255                                        void *opaque);
256 
257 static void qemu_net_client_setup(NetClientState *nc,
258                                   NetClientInfo *info,
259                                   NetClientState *peer,
260                                   const char *model,
261                                   const char *name,
262                                   NetClientDestructor *destructor,
263                                   bool is_datapath)
264 {
265     nc->info = info;
266     nc->model = g_strdup(model);
267     if (name) {
268         nc->name = g_strdup(name);
269     } else {
270         nc->name = assign_name(nc, model);
271     }
272 
273     if (peer) {
274         assert(!peer->peer);
275         nc->peer = peer;
276         peer->peer = nc;
277     }
278     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
279 
280     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
281     nc->destructor = destructor;
282     nc->is_datapath = is_datapath;
283     QTAILQ_INIT(&nc->filters);
284 }
285 
286 NetClientState *qemu_new_net_client(NetClientInfo *info,
287                                     NetClientState *peer,
288                                     const char *model,
289                                     const char *name)
290 {
291     NetClientState *nc;
292 
293     assert(info->size >= sizeof(NetClientState));
294 
295     nc = g_malloc0(info->size);
296     qemu_net_client_setup(nc, info, peer, model, name,
297                           qemu_net_client_destructor, true);
298 
299     return nc;
300 }
301 
302 NetClientState *qemu_new_net_control_client(NetClientInfo *info,
303                                             NetClientState *peer,
304                                             const char *model,
305                                             const char *name)
306 {
307     NetClientState *nc;
308 
309     assert(info->size >= sizeof(NetClientState));
310 
311     nc = g_malloc0(info->size);
312     qemu_net_client_setup(nc, info, peer, model, name,
313                           qemu_net_client_destructor, false);
314 
315     return nc;
316 }
317 
318 NICState *qemu_new_nic(NetClientInfo *info,
319                        NICConf *conf,
320                        const char *model,
321                        const char *name,
322                        MemReentrancyGuard *reentrancy_guard,
323                        void *opaque)
324 {
325     NetClientState **peers = conf->peers.ncs;
326     NICState *nic;
327     int i, queues = MAX(1, conf->peers.queues);
328 
329     assert(info->type == NET_CLIENT_DRIVER_NIC);
330     assert(info->size >= sizeof(NICState));
331 
332     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
333     nic->ncs = (void *)nic + info->size;
334     nic->conf = conf;
335     nic->reentrancy_guard = reentrancy_guard,
336     nic->opaque = opaque;
337 
338     for (i = 0; i < queues; i++) {
339         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
340                               NULL, true);
341         nic->ncs[i].queue_index = i;
342     }
343 
344     return nic;
345 }
346 
347 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
348 {
349     return nic->ncs + queue_index;
350 }
351 
352 NetClientState *qemu_get_queue(NICState *nic)
353 {
354     return qemu_get_subqueue(nic, 0);
355 }
356 
357 NICState *qemu_get_nic(NetClientState *nc)
358 {
359     NetClientState *nc0 = nc - nc->queue_index;
360 
361     return (NICState *)((void *)nc0 - nc->info->size);
362 }
363 
364 void *qemu_get_nic_opaque(NetClientState *nc)
365 {
366     NICState *nic = qemu_get_nic(nc);
367 
368     return nic->opaque;
369 }
370 
371 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
372 {
373     assert(nc != NULL);
374     NetClientState *ncs = nc + queue_index;
375     return ncs->peer;
376 }
377 
378 static void qemu_cleanup_net_client(NetClientState *nc)
379 {
380     QTAILQ_REMOVE(&net_clients, nc, next);
381 
382     if (nc->info->cleanup) {
383         nc->info->cleanup(nc);
384     }
385 }
386 
387 static void qemu_free_net_client(NetClientState *nc)
388 {
389     if (nc->incoming_queue) {
390         qemu_del_net_queue(nc->incoming_queue);
391     }
392     if (nc->peer) {
393         nc->peer->peer = NULL;
394     }
395     g_free(nc->name);
396     g_free(nc->model);
397     if (nc->destructor) {
398         nc->destructor(nc);
399     }
400 }
401 
402 void qemu_del_net_client(NetClientState *nc)
403 {
404     NetClientState *ncs[MAX_QUEUE_NUM];
405     int queues, i;
406     NetFilterState *nf, *next;
407 
408     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
409 
410     /* If the NetClientState belongs to a multiqueue backend, we will change all
411      * other NetClientStates also.
412      */
413     queues = qemu_find_net_clients_except(nc->name, ncs,
414                                           NET_CLIENT_DRIVER_NIC,
415                                           MAX_QUEUE_NUM);
416     assert(queues != 0);
417 
418     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
419         object_unparent(OBJECT(nf));
420     }
421 
422     /* If there is a peer NIC, delete and cleanup client, but do not free. */
423     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
424         NICState *nic = qemu_get_nic(nc->peer);
425         if (nic->peer_deleted) {
426             return;
427         }
428         nic->peer_deleted = true;
429 
430         for (i = 0; i < queues; i++) {
431             ncs[i]->peer->link_down = true;
432         }
433 
434         if (nc->peer->info->link_status_changed) {
435             nc->peer->info->link_status_changed(nc->peer);
436         }
437 
438         for (i = 0; i < queues; i++) {
439             qemu_cleanup_net_client(ncs[i]);
440         }
441 
442         return;
443     }
444 
445     for (i = 0; i < queues; i++) {
446         qemu_cleanup_net_client(ncs[i]);
447         qemu_free_net_client(ncs[i]);
448     }
449 }
450 
451 void qemu_del_nic(NICState *nic)
452 {
453     int i, queues = MAX(nic->conf->peers.queues, 1);
454 
455     qemu_macaddr_set_free(&nic->conf->macaddr);
456 
457     for (i = 0; i < queues; i++) {
458         NetClientState *nc = qemu_get_subqueue(nic, i);
459         /* If this is a peer NIC and peer has already been deleted, free it now. */
460         if (nic->peer_deleted) {
461             qemu_free_net_client(nc->peer);
462         } else if (nc->peer) {
463             /* if there are RX packets pending, complete them */
464             qemu_purge_queued_packets(nc->peer);
465         }
466     }
467 
468     for (i = queues - 1; i >= 0; i--) {
469         NetClientState *nc = qemu_get_subqueue(nic, i);
470 
471         qemu_cleanup_net_client(nc);
472         qemu_free_net_client(nc);
473     }
474 
475     g_free(nic);
476 }
477 
478 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
479 {
480     NetClientState *nc;
481 
482     QTAILQ_FOREACH(nc, &net_clients, next) {
483         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
484             if (nc->queue_index == 0) {
485                 func(qemu_get_nic(nc), opaque);
486             }
487         }
488     }
489 }
490 
491 bool qemu_has_ufo(NetClientState *nc)
492 {
493     if (!nc || !nc->info->has_ufo) {
494         return false;
495     }
496 
497     return nc->info->has_ufo(nc);
498 }
499 
500 bool qemu_has_uso(NetClientState *nc)
501 {
502     if (!nc || !nc->info->has_uso) {
503         return false;
504     }
505 
506     return nc->info->has_uso(nc);
507 }
508 
509 bool qemu_has_vnet_hdr(NetClientState *nc)
510 {
511     if (!nc || !nc->info->has_vnet_hdr) {
512         return false;
513     }
514 
515     return nc->info->has_vnet_hdr(nc);
516 }
517 
518 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
519 {
520     if (!nc || !nc->info->has_vnet_hdr_len) {
521         return false;
522     }
523 
524     return nc->info->has_vnet_hdr_len(nc, len);
525 }
526 
527 bool qemu_get_using_vnet_hdr(NetClientState *nc)
528 {
529     if (!nc || !nc->info->get_using_vnet_hdr) {
530         return false;
531     }
532 
533     return nc->info->get_using_vnet_hdr(nc);
534 }
535 
536 void qemu_using_vnet_hdr(NetClientState *nc, bool enable)
537 {
538     if (!nc || !nc->info->using_vnet_hdr) {
539         return;
540     }
541 
542     nc->info->using_vnet_hdr(nc, enable);
543 }
544 
545 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
546                           int ecn, int ufo, int uso4, int uso6)
547 {
548     if (!nc || !nc->info->set_offload) {
549         return;
550     }
551 
552     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo, uso4, uso6);
553 }
554 
555 int qemu_get_vnet_hdr_len(NetClientState *nc)
556 {
557     if (!nc || !nc->info->get_vnet_hdr_len) {
558         return 0;
559     }
560 
561     return nc->info->get_vnet_hdr_len(nc);
562 }
563 
564 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
565 {
566     if (!nc || !nc->info->set_vnet_hdr_len) {
567         return;
568     }
569 
570     nc->vnet_hdr_len = len;
571     nc->info->set_vnet_hdr_len(nc, len);
572 }
573 
574 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
575 {
576 #if HOST_BIG_ENDIAN
577     if (!nc || !nc->info->set_vnet_le) {
578         return -ENOSYS;
579     }
580 
581     return nc->info->set_vnet_le(nc, is_le);
582 #else
583     return 0;
584 #endif
585 }
586 
587 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
588 {
589 #if HOST_BIG_ENDIAN
590     return 0;
591 #else
592     if (!nc || !nc->info->set_vnet_be) {
593         return -ENOSYS;
594     }
595 
596     return nc->info->set_vnet_be(nc, is_be);
597 #endif
598 }
599 
600 int qemu_can_receive_packet(NetClientState *nc)
601 {
602     if (nc->receive_disabled) {
603         return 0;
604     } else if (nc->info->can_receive &&
605                !nc->info->can_receive(nc)) {
606         return 0;
607     }
608     return 1;
609 }
610 
611 int qemu_can_send_packet(NetClientState *sender)
612 {
613     int vm_running = runstate_is_running();
614 
615     if (!vm_running) {
616         return 0;
617     }
618 
619     if (!sender->peer) {
620         return 1;
621     }
622 
623     return qemu_can_receive_packet(sender->peer);
624 }
625 
626 static ssize_t filter_receive_iov(NetClientState *nc,
627                                   NetFilterDirection direction,
628                                   NetClientState *sender,
629                                   unsigned flags,
630                                   const struct iovec *iov,
631                                   int iovcnt,
632                                   NetPacketSent *sent_cb)
633 {
634     ssize_t ret = 0;
635     NetFilterState *nf = NULL;
636 
637     if (direction == NET_FILTER_DIRECTION_TX) {
638         QTAILQ_FOREACH(nf, &nc->filters, next) {
639             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
640                                          iovcnt, sent_cb);
641             if (ret) {
642                 return ret;
643             }
644         }
645     } else {
646         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
647             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
648                                          iovcnt, sent_cb);
649             if (ret) {
650                 return ret;
651             }
652         }
653     }
654 
655     return ret;
656 }
657 
658 static ssize_t filter_receive(NetClientState *nc,
659                               NetFilterDirection direction,
660                               NetClientState *sender,
661                               unsigned flags,
662                               const uint8_t *data,
663                               size_t size,
664                               NetPacketSent *sent_cb)
665 {
666     struct iovec iov = {
667         .iov_base = (void *)data,
668         .iov_len = size
669     };
670 
671     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
672 }
673 
674 void qemu_purge_queued_packets(NetClientState *nc)
675 {
676     if (!nc->peer) {
677         return;
678     }
679 
680     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
681 }
682 
683 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
684 {
685     nc->receive_disabled = 0;
686 
687     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
688         if (net_hub_flush(nc->peer)) {
689             qemu_notify_event();
690         }
691     }
692     if (qemu_net_queue_flush(nc->incoming_queue)) {
693         /* We emptied the queue successfully, signal to the IO thread to repoll
694          * the file descriptor (for tap, for example).
695          */
696         qemu_notify_event();
697     } else if (purge) {
698         /* Unable to empty the queue, purge remaining packets */
699         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
700     }
701 }
702 
703 void qemu_flush_queued_packets(NetClientState *nc)
704 {
705     qemu_flush_or_purge_queued_packets(nc, false);
706 }
707 
708 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
709                                                  unsigned flags,
710                                                  const uint8_t *buf, int size,
711                                                  NetPacketSent *sent_cb)
712 {
713     NetQueue *queue;
714     int ret;
715 
716 #ifdef DEBUG_NET
717     printf("qemu_send_packet_async:\n");
718     qemu_hexdump(stdout, "net", buf, size);
719 #endif
720 
721     if (sender->link_down || !sender->peer) {
722         return size;
723     }
724 
725     /* Let filters handle the packet first */
726     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
727                          sender, flags, buf, size, sent_cb);
728     if (ret) {
729         return ret;
730     }
731 
732     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
733                          sender, flags, buf, size, sent_cb);
734     if (ret) {
735         return ret;
736     }
737 
738     queue = sender->peer->incoming_queue;
739 
740     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
741 }
742 
743 ssize_t qemu_send_packet_async(NetClientState *sender,
744                                const uint8_t *buf, int size,
745                                NetPacketSent *sent_cb)
746 {
747     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
748                                              buf, size, sent_cb);
749 }
750 
751 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
752 {
753     return qemu_send_packet_async(nc, buf, size, NULL);
754 }
755 
756 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
757 {
758     if (!qemu_can_receive_packet(nc)) {
759         return 0;
760     }
761 
762     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
763 }
764 
765 ssize_t qemu_receive_packet_iov(NetClientState *nc, const struct iovec *iov,
766                                 int iovcnt)
767 {
768     if (!qemu_can_receive_packet(nc)) {
769         return 0;
770     }
771 
772     return qemu_net_queue_receive_iov(nc->incoming_queue, iov, iovcnt);
773 }
774 
775 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
776 {
777     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
778                                              buf, size, NULL);
779 }
780 
781 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
782                                int iovcnt, unsigned flags)
783 {
784     uint8_t *buf = NULL;
785     uint8_t *buffer;
786     size_t offset;
787     ssize_t ret;
788 
789     if (iovcnt == 1) {
790         buffer = iov[0].iov_base;
791         offset = iov[0].iov_len;
792     } else {
793         offset = iov_size(iov, iovcnt);
794         if (offset > NET_BUFSIZE) {
795             return -1;
796         }
797         buf = g_malloc(offset);
798         buffer = buf;
799         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
800     }
801 
802     if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
803         ret = nc->info->receive_raw(nc, buffer, offset);
804     } else {
805         ret = nc->info->receive(nc, buffer, offset);
806     }
807 
808     g_free(buf);
809     return ret;
810 }
811 
812 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
813                                        unsigned flags,
814                                        const struct iovec *iov,
815                                        int iovcnt,
816                                        void *opaque)
817 {
818     MemReentrancyGuard *owned_reentrancy_guard;
819     NetClientState *nc = opaque;
820     int ret;
821 
822 
823     if (nc->link_down) {
824         return iov_size(iov, iovcnt);
825     }
826 
827     if (nc->receive_disabled) {
828         return 0;
829     }
830 
831     if (nc->info->type != NET_CLIENT_DRIVER_NIC ||
832         qemu_get_nic(nc)->reentrancy_guard->engaged_in_io) {
833         owned_reentrancy_guard = NULL;
834     } else {
835         owned_reentrancy_guard = qemu_get_nic(nc)->reentrancy_guard;
836         owned_reentrancy_guard->engaged_in_io = true;
837     }
838 
839     if (nc->info->receive_iov && !(flags & QEMU_NET_PACKET_FLAG_RAW)) {
840         ret = nc->info->receive_iov(nc, iov, iovcnt);
841     } else {
842         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
843     }
844 
845     if (owned_reentrancy_guard) {
846         owned_reentrancy_guard->engaged_in_io = false;
847     }
848 
849     if (ret == 0) {
850         nc->receive_disabled = 1;
851     }
852 
853     return ret;
854 }
855 
856 ssize_t qemu_sendv_packet_async(NetClientState *sender,
857                                 const struct iovec *iov, int iovcnt,
858                                 NetPacketSent *sent_cb)
859 {
860     NetQueue *queue;
861     size_t size = iov_size(iov, iovcnt);
862     int ret;
863 
864     if (size > NET_BUFSIZE) {
865         return size;
866     }
867 
868     if (sender->link_down || !sender->peer) {
869         return size;
870     }
871 
872     /* Let filters handle the packet first */
873     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
874                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
875     if (ret) {
876         return ret;
877     }
878 
879     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
880                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
881     if (ret) {
882         return ret;
883     }
884 
885     queue = sender->peer->incoming_queue;
886 
887     return qemu_net_queue_send_iov(queue, sender,
888                                    QEMU_NET_PACKET_FLAG_NONE,
889                                    iov, iovcnt, sent_cb);
890 }
891 
892 ssize_t
893 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
894 {
895     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
896 }
897 
898 NetClientState *qemu_find_netdev(const char *id)
899 {
900     NetClientState *nc;
901 
902     QTAILQ_FOREACH(nc, &net_clients, next) {
903         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
904             continue;
905         if (!strcmp(nc->name, id)) {
906             return nc;
907         }
908     }
909 
910     return NULL;
911 }
912 
913 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
914                                  NetClientDriver type, int max)
915 {
916     NetClientState *nc;
917     int ret = 0;
918 
919     QTAILQ_FOREACH(nc, &net_clients, next) {
920         if (nc->info->type == type) {
921             continue;
922         }
923         if (!id || !strcmp(nc->name, id)) {
924             if (ret < max) {
925                 ncs[ret] = nc;
926             }
927             ret++;
928         }
929     }
930 
931     return ret;
932 }
933 
934 static int nic_get_free_idx(void)
935 {
936     int index;
937 
938     for (index = 0; index < MAX_NICS; index++)
939         if (!nd_table[index].used)
940             return index;
941     return -1;
942 }
943 
944 GPtrArray *qemu_get_nic_models(const char *device_type)
945 {
946     GPtrArray *nic_models = g_ptr_array_new();
947     GSList *list = object_class_get_list_sorted(device_type, false);
948 
949     while (list) {
950         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
951                                              TYPE_DEVICE);
952         GSList *next;
953         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
954             dc->user_creatable) {
955             const char *name = object_class_get_name(list->data);
956             /*
957              * A network device might also be something else than a NIC, see
958              * e.g. the "rocker" device. Thus we have to look for the "netdev"
959              * property, too. Unfortunately, some devices like virtio-net only
960              * create this property during instance_init, so we have to create
961              * a temporary instance here to be able to check it.
962              */
963             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
964             if (object_property_find(obj, "netdev")) {
965                 g_ptr_array_add(nic_models, (gpointer)name);
966             }
967             object_unref(obj);
968         }
969         next = list->next;
970         g_slist_free_1(list);
971         list = next;
972     }
973     g_ptr_array_add(nic_models, NULL);
974 
975     return nic_models;
976 }
977 
978 int qemu_show_nic_models(const char *arg, const char *const *models)
979 {
980     int i;
981 
982     if (!arg || !is_help_option(arg)) {
983         return 0;
984     }
985 
986     printf("Available NIC models:\n");
987     for (i = 0 ; models[i]; i++) {
988         printf("%s\n", models[i]);
989     }
990     return 1;
991 }
992 
993 void qemu_check_nic_model(NICInfo *nd, const char *model)
994 {
995     const char *models[2];
996 
997     models[0] = model;
998     models[1] = NULL;
999 
1000     if (qemu_show_nic_models(nd->model, models))
1001         exit(0);
1002     if (qemu_find_nic_model(nd, models, model) < 0)
1003         exit(1);
1004 }
1005 
1006 int qemu_find_nic_model(NICInfo *nd, const char * const *models,
1007                         const char *default_model)
1008 {
1009     int i;
1010 
1011     if (!nd->model)
1012         nd->model = g_strdup(default_model);
1013 
1014     for (i = 0 ; models[i]; i++) {
1015         if (strcmp(nd->model, models[i]) == 0)
1016             return i;
1017     }
1018 
1019     error_report("Unsupported NIC model: %s", nd->model);
1020     return -1;
1021 }
1022 
1023 static int net_init_nic(const Netdev *netdev, const char *name,
1024                         NetClientState *peer, Error **errp)
1025 {
1026     int idx;
1027     NICInfo *nd;
1028     const NetLegacyNicOptions *nic;
1029 
1030     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
1031     nic = &netdev->u.nic;
1032 
1033     idx = nic_get_free_idx();
1034     if (idx == -1 || nb_nics >= MAX_NICS) {
1035         error_setg(errp, "too many NICs");
1036         return -1;
1037     }
1038 
1039     nd = &nd_table[idx];
1040 
1041     memset(nd, 0, sizeof(*nd));
1042 
1043     if (nic->netdev) {
1044         nd->netdev = qemu_find_netdev(nic->netdev);
1045         if (!nd->netdev) {
1046             error_setg(errp, "netdev '%s' not found", nic->netdev);
1047             return -1;
1048         }
1049     } else {
1050         assert(peer);
1051         nd->netdev = peer;
1052     }
1053     nd->name = g_strdup(name);
1054     if (nic->model) {
1055         nd->model = g_strdup(nic->model);
1056     }
1057     if (nic->addr) {
1058         nd->devaddr = g_strdup(nic->addr);
1059     }
1060 
1061     if (nic->macaddr &&
1062         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1063         error_setg(errp, "invalid syntax for ethernet address");
1064         return -1;
1065     }
1066     if (nic->macaddr &&
1067         is_multicast_ether_addr(nd->macaddr.a)) {
1068         error_setg(errp,
1069                    "NIC cannot have multicast MAC address (odd 1st byte)");
1070         return -1;
1071     }
1072     qemu_macaddr_default_if_unset(&nd->macaddr);
1073 
1074     if (nic->has_vectors) {
1075         if (nic->vectors > 0x7ffffff) {
1076             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1077             return -1;
1078         }
1079         nd->nvectors = nic->vectors;
1080     } else {
1081         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1082     }
1083 
1084     nd->used = 1;
1085     nb_nics++;
1086 
1087     return idx;
1088 }
1089 
1090 NICInfo *qemu_find_nic_info(const char *typename, bool match_default,
1091                             const char *alias)
1092 {
1093     NICInfo *nd;
1094     int i;
1095 
1096     for (i = 0; i < nb_nics; i++) {
1097         nd = &nd_table[i];
1098 
1099         if (!nd->used || nd->instantiated) {
1100             continue;
1101         }
1102 
1103         if ((match_default && !nd->model) || !g_strcmp0(nd->model, typename)
1104             || (alias && !g_strcmp0(nd->model, alias))) {
1105             return nd;
1106         }
1107     }
1108     return NULL;
1109 }
1110 
1111 
1112 /* "I have created a device. Please configure it if you can" */
1113 bool qemu_configure_nic_device(DeviceState *dev, bool match_default,
1114                                const char *alias)
1115 {
1116     NICInfo *nd = qemu_find_nic_info(object_get_typename(OBJECT(dev)),
1117                                      match_default, alias);
1118 
1119     if (nd) {
1120         qdev_set_nic_properties(dev, nd);
1121         return true;
1122     }
1123     return false;
1124 }
1125 
1126 /* "Please create a device, if you have a configuration for it" */
1127 DeviceState *qemu_create_nic_device(const char *typename, bool match_default,
1128                                     const char *alias)
1129 {
1130     NICInfo *nd = qemu_find_nic_info(typename, match_default, alias);
1131     DeviceState *dev;
1132 
1133     if (!nd) {
1134         return NULL;
1135     }
1136 
1137     dev = qdev_new(typename);
1138     qdev_set_nic_properties(dev, nd);
1139     return dev;
1140 }
1141 
1142 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1143     const Netdev *netdev,
1144     const char *name,
1145     NetClientState *peer, Error **errp) = {
1146         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1147 #ifdef CONFIG_SLIRP
1148         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1149 #endif
1150         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1151         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1152         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1153         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1154 #ifdef CONFIG_VDE
1155         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1156 #endif
1157 #ifdef CONFIG_NETMAP
1158         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1159 #endif
1160 #ifdef CONFIG_AF_XDP
1161         [NET_CLIENT_DRIVER_AF_XDP]    = net_init_af_xdp,
1162 #endif
1163 #ifdef CONFIG_NET_BRIDGE
1164         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1165 #endif
1166         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1167 #ifdef CONFIG_VHOST_NET_USER
1168         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1169 #endif
1170 #ifdef CONFIG_VHOST_NET_VDPA
1171         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1172 #endif
1173 #ifdef CONFIG_L2TPV3
1174         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1175 #endif
1176 #ifdef CONFIG_VMNET
1177         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1178         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1179         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1180 #endif /* CONFIG_VMNET */
1181 };
1182 
1183 
1184 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1185 {
1186     NetClientState *peer = NULL;
1187     NetClientState *nc;
1188 
1189     if (is_netdev) {
1190         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1191             !net_client_init_fun[netdev->type]) {
1192             error_setg(errp, "network backend '%s' is not compiled into this binary",
1193                        NetClientDriver_str(netdev->type));
1194             return -1;
1195         }
1196     } else {
1197         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1198             return 0; /* nothing to do */
1199         }
1200         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1201             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1202                        NetClientDriver_str(netdev->type));
1203             return -1;
1204         }
1205 
1206         if (!net_client_init_fun[netdev->type]) {
1207             error_setg(errp, "network backend '%s' is not compiled into this binary",
1208                        NetClientDriver_str(netdev->type));
1209             return -1;
1210         }
1211 
1212         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1213         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1214             !netdev->u.nic.netdev) {
1215             peer = net_hub_add_port(0, NULL, NULL);
1216         }
1217     }
1218 
1219     nc = qemu_find_netdev(netdev->id);
1220     if (nc) {
1221         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1222         return -1;
1223     }
1224 
1225     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1226         /* FIXME drop when all init functions store an Error */
1227         if (errp && !*errp) {
1228             error_setg(errp, "Device '%s' could not be initialized",
1229                        NetClientDriver_str(netdev->type));
1230         }
1231         return -1;
1232     }
1233 
1234     if (is_netdev) {
1235         nc = qemu_find_netdev(netdev->id);
1236         assert(nc);
1237         nc->is_netdev = true;
1238     }
1239 
1240     return 0;
1241 }
1242 
1243 void show_netdevs(void)
1244 {
1245     int idx;
1246     const char *available_netdevs[] = {
1247         "socket",
1248         "stream",
1249         "dgram",
1250         "hubport",
1251         "tap",
1252 #ifdef CONFIG_SLIRP
1253         "user",
1254 #endif
1255 #ifdef CONFIG_L2TPV3
1256         "l2tpv3",
1257 #endif
1258 #ifdef CONFIG_VDE
1259         "vde",
1260 #endif
1261 #ifdef CONFIG_NET_BRIDGE
1262         "bridge",
1263 #endif
1264 #ifdef CONFIG_NETMAP
1265         "netmap",
1266 #endif
1267 #ifdef CONFIG_AF_XDP
1268         "af-xdp",
1269 #endif
1270 #ifdef CONFIG_POSIX
1271         "vhost-user",
1272 #endif
1273 #ifdef CONFIG_VHOST_VDPA
1274         "vhost-vdpa",
1275 #endif
1276 #ifdef CONFIG_VMNET
1277         "vmnet-host",
1278         "vmnet-shared",
1279         "vmnet-bridged",
1280 #endif
1281     };
1282 
1283     qemu_printf("Available netdev backend types:\n");
1284     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1285         qemu_printf("%s\n", available_netdevs[idx]);
1286     }
1287 }
1288 
1289 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1290 {
1291     gchar **substrings = NULL;
1292     Netdev *object = NULL;
1293     int ret = -1;
1294     Visitor *v = opts_visitor_new(opts);
1295 
1296     /* Parse convenience option format ipv6-net=fec0::0[/64] */
1297     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1298 
1299     if (ip6_net) {
1300         char *prefix_addr;
1301         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1302 
1303         substrings = g_strsplit(ip6_net, "/", 2);
1304         if (!substrings || !substrings[0]) {
1305             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1306                        "a valid IPv6 prefix");
1307             goto out;
1308         }
1309 
1310         prefix_addr = substrings[0];
1311 
1312         /* Handle user-specified prefix length. */
1313         if (substrings[1] &&
1314             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1315         {
1316             error_setg(errp,
1317                        "parameter 'ipv6-net' expects a number after '/'");
1318             goto out;
1319         }
1320 
1321         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1322         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1323                             &error_abort);
1324         qemu_opt_unset(opts, "ipv6-net");
1325     }
1326 
1327     /* Create an ID for -net if the user did not specify one */
1328     if (!is_netdev && !qemu_opts_id(opts)) {
1329         qemu_opts_set_id(opts, id_generate(ID_NET));
1330     }
1331 
1332     if (visit_type_Netdev(v, NULL, &object, errp)) {
1333         ret = net_client_init1(object, is_netdev, errp);
1334     }
1335 
1336     qapi_free_Netdev(object);
1337 
1338 out:
1339     g_strfreev(substrings);
1340     visit_free(v);
1341     return ret;
1342 }
1343 
1344 void netdev_add(QemuOpts *opts, Error **errp)
1345 {
1346     net_client_init(opts, true, errp);
1347 }
1348 
1349 void qmp_netdev_add(Netdev *netdev, Error **errp)
1350 {
1351     if (!id_wellformed(netdev->id)) {
1352         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1353         return;
1354     }
1355 
1356     net_client_init1(netdev, true, errp);
1357 }
1358 
1359 void qmp_netdev_del(const char *id, Error **errp)
1360 {
1361     NetClientState *nc;
1362     QemuOpts *opts;
1363 
1364     nc = qemu_find_netdev(id);
1365     if (!nc) {
1366         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1367                   "Device '%s' not found", id);
1368         return;
1369     }
1370 
1371     if (!nc->is_netdev) {
1372         error_setg(errp, "Device '%s' is not a netdev", id);
1373         return;
1374     }
1375 
1376     qemu_del_net_client(nc);
1377 
1378     /*
1379      * Wart: we need to delete the QemuOpts associated with netdevs
1380      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1381      * HMP netdev_add.
1382      */
1383     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1384     if (opts) {
1385         qemu_opts_del(opts);
1386     }
1387 }
1388 
1389 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1390 {
1391     char *str;
1392     ObjectProperty *prop;
1393     ObjectPropertyIterator iter;
1394     Visitor *v;
1395 
1396     /* generate info str */
1397     object_property_iter_init(&iter, OBJECT(nf));
1398     while ((prop = object_property_iter_next(&iter))) {
1399         if (!strcmp(prop->name, "type")) {
1400             continue;
1401         }
1402         v = string_output_visitor_new(false, &str);
1403         object_property_get(OBJECT(nf), prop->name, v, NULL);
1404         visit_complete(v, &str);
1405         visit_free(v);
1406         monitor_printf(mon, ",%s=%s", prop->name, str);
1407         g_free(str);
1408     }
1409     monitor_printf(mon, "\n");
1410 }
1411 
1412 void print_net_client(Monitor *mon, NetClientState *nc)
1413 {
1414     NetFilterState *nf;
1415 
1416     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1417                    nc->queue_index,
1418                    NetClientDriver_str(nc->info->type),
1419                    nc->info_str);
1420     if (!QTAILQ_EMPTY(&nc->filters)) {
1421         monitor_printf(mon, "filters:\n");
1422     }
1423     QTAILQ_FOREACH(nf, &nc->filters, next) {
1424         monitor_printf(mon, "  - %s: type=%s",
1425                        object_get_canonical_path_component(OBJECT(nf)),
1426                        object_get_typename(OBJECT(nf)));
1427         netfilter_print_info(mon, nf);
1428     }
1429 }
1430 
1431 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1432 {
1433     NetClientState *nc;
1434     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1435 
1436     QTAILQ_FOREACH(nc, &net_clients, next) {
1437         RxFilterInfo *info;
1438 
1439         if (name && strcmp(nc->name, name) != 0) {
1440             continue;
1441         }
1442 
1443         /* only query rx-filter information of NIC */
1444         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1445             if (name) {
1446                 error_setg(errp, "net client(%s) isn't a NIC", name);
1447                 assert(!filter_list);
1448                 return NULL;
1449             }
1450             continue;
1451         }
1452 
1453         /* only query information on queue 0 since the info is per nic,
1454          * not per queue
1455          */
1456         if (nc->queue_index != 0)
1457             continue;
1458 
1459         if (nc->info->query_rx_filter) {
1460             info = nc->info->query_rx_filter(nc);
1461             QAPI_LIST_APPEND(tail, info);
1462         } else if (name) {
1463             error_setg(errp, "net client(%s) doesn't support"
1464                        " rx-filter querying", name);
1465             assert(!filter_list);
1466             return NULL;
1467         }
1468 
1469         if (name) {
1470             break;
1471         }
1472     }
1473 
1474     if (filter_list == NULL && name) {
1475         error_setg(errp, "invalid net client name: %s", name);
1476     }
1477 
1478     return filter_list;
1479 }
1480 
1481 void colo_notify_filters_event(int event, Error **errp)
1482 {
1483     NetClientState *nc;
1484     NetFilterState *nf;
1485     NetFilterClass *nfc = NULL;
1486     Error *local_err = NULL;
1487 
1488     QTAILQ_FOREACH(nc, &net_clients, next) {
1489         QTAILQ_FOREACH(nf, &nc->filters, next) {
1490             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1491             nfc->handle_event(nf, event, &local_err);
1492             if (local_err) {
1493                 error_propagate(errp, local_err);
1494                 return;
1495             }
1496         }
1497     }
1498 }
1499 
1500 void qmp_set_link(const char *name, bool up, Error **errp)
1501 {
1502     NetClientState *ncs[MAX_QUEUE_NUM];
1503     NetClientState *nc;
1504     int queues, i;
1505 
1506     queues = qemu_find_net_clients_except(name, ncs,
1507                                           NET_CLIENT_DRIVER__MAX,
1508                                           MAX_QUEUE_NUM);
1509 
1510     if (queues == 0) {
1511         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1512                   "Device '%s' not found", name);
1513         return;
1514     }
1515     nc = ncs[0];
1516 
1517     for (i = 0; i < queues; i++) {
1518         ncs[i]->link_down = !up;
1519     }
1520 
1521     if (nc->info->link_status_changed) {
1522         nc->info->link_status_changed(nc);
1523     }
1524 
1525     if (nc->peer) {
1526         /* Change peer link only if the peer is NIC and then notify peer.
1527          * If the peer is a HUBPORT or a backend, we do not change the
1528          * link status.
1529          *
1530          * This behavior is compatible with qemu hubs where there could be
1531          * multiple clients that can still communicate with each other in
1532          * disconnected mode. For now maintain this compatibility.
1533          */
1534         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1535             for (i = 0; i < queues; i++) {
1536                 ncs[i]->peer->link_down = !up;
1537             }
1538         }
1539         if (nc->peer->info->link_status_changed) {
1540             nc->peer->info->link_status_changed(nc->peer);
1541         }
1542     }
1543 }
1544 
1545 static void net_vm_change_state_handler(void *opaque, bool running,
1546                                         RunState state)
1547 {
1548     NetClientState *nc;
1549     NetClientState *tmp;
1550 
1551     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1552         if (running) {
1553             /* Flush queued packets and wake up backends. */
1554             if (nc->peer && qemu_can_send_packet(nc)) {
1555                 qemu_flush_queued_packets(nc->peer);
1556             }
1557         } else {
1558             /* Complete all queued packets, to guarantee we don't modify
1559              * state later when VM is not running.
1560              */
1561             qemu_flush_or_purge_queued_packets(nc, true);
1562         }
1563     }
1564 }
1565 
1566 void net_cleanup(void)
1567 {
1568     NetClientState *nc, **p = &QTAILQ_FIRST(&net_clients);
1569 
1570     /*cleanup colo compare module for COLO*/
1571     colo_compare_cleanup();
1572 
1573     /*
1574      * Walk the net_clients list and remove the netdevs but *not* any
1575      * NET_CLIENT_DRIVER_NIC entries. The latter are owned by the device
1576      * model which created them, and in some cases (e.g. xen-net-device)
1577      * the device itself may do cleanup at exit and will be upset if we
1578      * just delete its NIC from underneath it.
1579      *
1580      * Since qemu_del_net_client() may delete multiple entries, using
1581      * QTAILQ_FOREACH_SAFE() is not safe here. The only safe pointer
1582      * to keep as a bookmark is a NET_CLIENT_DRIVER_NIC entry, so keep
1583      * 'p' pointing to either the head of the list, or the 'next' field
1584      * of the latest NET_CLIENT_DRIVER_NIC, and operate on *p as we walk
1585      * the list.
1586      *
1587      * The 'nc' variable isn't part of the list traversal; it's purely
1588      * for convenience as too much '(*p)->' has a tendency to make the
1589      * readers' eyes bleed.
1590      */
1591     while (*p) {
1592         nc = *p;
1593         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1594             /* Skip NET_CLIENT_DRIVER_NIC entries */
1595             p = &QTAILQ_NEXT(nc, next);
1596         } else {
1597             qemu_del_net_client(nc);
1598         }
1599     }
1600 
1601     qemu_del_vm_change_state_handler(net_change_state_entry);
1602 }
1603 
1604 void net_check_clients(void)
1605 {
1606     NetClientState *nc;
1607     int i;
1608 
1609     net_hub_check_clients();
1610 
1611     QTAILQ_FOREACH(nc, &net_clients, next) {
1612         if (!nc->peer) {
1613             warn_report("%s %s has no peer",
1614                         nc->info->type == NET_CLIENT_DRIVER_NIC
1615                         ? "nic" : "netdev",
1616                         nc->name);
1617         }
1618     }
1619 
1620     /* Check that all NICs requested via -net nic actually got created.
1621      * NICs created via -device don't need to be checked here because
1622      * they are always instantiated.
1623      */
1624     for (i = 0; i < MAX_NICS; i++) {
1625         NICInfo *nd = &nd_table[i];
1626         if (nd->used && !nd->instantiated) {
1627             warn_report("requested NIC (%s, model %s) "
1628                         "was not created (not supported by this machine?)",
1629                         nd->name ? nd->name : "anonymous",
1630                         nd->model ? nd->model : "unspecified");
1631         }
1632     }
1633 }
1634 
1635 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1636 {
1637     return net_client_init(opts, false, errp);
1638 }
1639 
1640 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1641 {
1642     const char *type = qemu_opt_get(opts, "type");
1643 
1644     if (type && is_help_option(type)) {
1645         show_netdevs();
1646         exit(0);
1647     }
1648     return net_client_init(opts, true, errp);
1649 }
1650 
1651 /* For the convenience "--nic" parameter */
1652 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1653 {
1654     char *mac, *nd_id;
1655     int idx, ret;
1656     NICInfo *ni;
1657     const char *type;
1658 
1659     type = qemu_opt_get(opts, "type");
1660     if (type) {
1661         if (g_str_equal(type, "none")) {
1662             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1663         }
1664         if (is_help_option(type)) {
1665             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1666             show_netdevs();
1667             printf("\n");
1668             qemu_show_nic_models(type, (const char **)nic_models->pdata);
1669             g_ptr_array_free(nic_models, true);
1670             exit(0);
1671         }
1672     }
1673 
1674     idx = nic_get_free_idx();
1675     if (idx == -1 || nb_nics >= MAX_NICS) {
1676         error_setg(errp, "no more on-board/default NIC slots available");
1677         return -1;
1678     }
1679 
1680     if (!type) {
1681         qemu_opt_set(opts, "type", "user", &error_abort);
1682     }
1683 
1684     ni = &nd_table[idx];
1685     memset(ni, 0, sizeof(*ni));
1686     ni->model = qemu_opt_get_del(opts, "model");
1687 
1688     /* Create an ID if the user did not specify one */
1689     nd_id = g_strdup(qemu_opts_id(opts));
1690     if (!nd_id) {
1691         nd_id = id_generate(ID_NET);
1692         qemu_opts_set_id(opts, nd_id);
1693     }
1694 
1695     /* Handle MAC address */
1696     mac = qemu_opt_get_del(opts, "mac");
1697     if (mac) {
1698         ret = net_parse_macaddr(ni->macaddr.a, mac);
1699         g_free(mac);
1700         if (ret) {
1701             error_setg(errp, "invalid syntax for ethernet address");
1702             goto out;
1703         }
1704         if (is_multicast_ether_addr(ni->macaddr.a)) {
1705             error_setg(errp, "NIC cannot have multicast MAC address");
1706             ret = -1;
1707             goto out;
1708         }
1709     }
1710     qemu_macaddr_default_if_unset(&ni->macaddr);
1711 
1712     ret = net_client_init(opts, true, errp);
1713     if (ret == 0) {
1714         ni->netdev = qemu_find_netdev(nd_id);
1715         ni->used = true;
1716         nb_nics++;
1717     }
1718 
1719 out:
1720     g_free(nd_id);
1721     return ret;
1722 }
1723 
1724 static void netdev_init_modern(void)
1725 {
1726     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1727         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1728 
1729         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1730         loc_push_restore(&nd->loc);
1731         net_client_init1(nd->nd, true, &error_fatal);
1732         loc_pop(&nd->loc);
1733         qapi_free_Netdev(nd->nd);
1734         g_free(nd);
1735     }
1736 }
1737 
1738 void net_init_clients(void)
1739 {
1740     net_change_state_entry =
1741         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1742 
1743     QTAILQ_INIT(&net_clients);
1744 
1745     netdev_init_modern();
1746 
1747     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1748                       &error_fatal);
1749 
1750     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1751                       &error_fatal);
1752 
1753     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1754                       &error_fatal);
1755 }
1756 
1757 /*
1758  * Does this -netdev argument use modern rather than traditional syntax?
1759  * Modern syntax is to be parsed with netdev_parse_modern().
1760  * Traditional syntax is to be parsed with net_client_parse().
1761  */
1762 bool netdev_is_modern(const char *optstr)
1763 {
1764     QemuOpts *opts;
1765     bool is_modern;
1766     const char *type;
1767     static QemuOptsList dummy_opts = {
1768         .name = "netdev",
1769         .implied_opt_name = "type",
1770         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1771         .desc = { { } },
1772     };
1773 
1774     if (optstr[0] == '{') {
1775         /* This is JSON, which means it's modern syntax */
1776         return true;
1777     }
1778 
1779     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1780     qemu_opts_do_parse(opts, optstr, dummy_opts.implied_opt_name,
1781                        &error_abort);
1782     type = qemu_opt_get(opts, "type");
1783     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1784 
1785     qemu_opts_reset(&dummy_opts);
1786 
1787     return is_modern;
1788 }
1789 
1790 /*
1791  * netdev_parse_modern() uses modern, more expressive syntax than
1792  * net_client_parse(), but supports only the -netdev option.
1793  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1794  * appends to @qemu_netdev_opts.
1795  */
1796 void netdev_parse_modern(const char *optstr)
1797 {
1798     Visitor *v;
1799     NetdevQueueEntry *nd;
1800 
1801     v = qobject_input_visitor_new_str(optstr, "type", &error_fatal);
1802     nd = g_new(NetdevQueueEntry, 1);
1803     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1804     visit_free(v);
1805     loc_save(&nd->loc);
1806 
1807     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1808 }
1809 
1810 void net_client_parse(QemuOptsList *opts_list, const char *optstr)
1811 {
1812     if (!qemu_opts_parse_noisily(opts_list, optstr, true)) {
1813         exit(1);
1814     }
1815 }
1816 
1817 /* From FreeBSD */
1818 /* XXX: optimize */
1819 uint32_t net_crc32(const uint8_t *p, int len)
1820 {
1821     uint32_t crc;
1822     int carry, i, j;
1823     uint8_t b;
1824 
1825     crc = 0xffffffff;
1826     for (i = 0; i < len; i++) {
1827         b = *p++;
1828         for (j = 0; j < 8; j++) {
1829             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1830             crc <<= 1;
1831             b >>= 1;
1832             if (carry) {
1833                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1834             }
1835         }
1836     }
1837 
1838     return crc;
1839 }
1840 
1841 uint32_t net_crc32_le(const uint8_t *p, int len)
1842 {
1843     uint32_t crc;
1844     int carry, i, j;
1845     uint8_t b;
1846 
1847     crc = 0xffffffff;
1848     for (i = 0; i < len; i++) {
1849         b = *p++;
1850         for (j = 0; j < 8; j++) {
1851             carry = (crc & 0x1) ^ (b & 0x01);
1852             crc >>= 1;
1853             b >>= 1;
1854             if (carry) {
1855                 crc ^= POLYNOMIAL_LE;
1856             }
1857         }
1858     }
1859 
1860     return crc;
1861 }
1862 
1863 QemuOptsList qemu_netdev_opts = {
1864     .name = "netdev",
1865     .implied_opt_name = "type",
1866     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1867     .desc = {
1868         /*
1869          * no elements => accept any params
1870          * validation will happen later
1871          */
1872         { /* end of list */ }
1873     },
1874 };
1875 
1876 QemuOptsList qemu_nic_opts = {
1877     .name = "nic",
1878     .implied_opt_name = "type",
1879     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1880     .desc = {
1881         /*
1882          * no elements => accept any params
1883          * validation will happen later
1884          */
1885         { /* end of list */ }
1886     },
1887 };
1888 
1889 QemuOptsList qemu_net_opts = {
1890     .name = "net",
1891     .implied_opt_name = "type",
1892     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1893     .desc = {
1894         /*
1895          * no elements => accept any params
1896          * validation will happen later
1897          */
1898         { /* end of list */ }
1899     },
1900 };
1901 
1902 void net_socket_rs_init(SocketReadState *rs,
1903                         SocketReadStateFinalize *finalize,
1904                         bool vnet_hdr)
1905 {
1906     rs->state = 0;
1907     rs->vnet_hdr = vnet_hdr;
1908     rs->index = 0;
1909     rs->packet_len = 0;
1910     rs->vnet_hdr_len = 0;
1911     memset(rs->buf, 0, sizeof(rs->buf));
1912     rs->finalize = finalize;
1913 }
1914 
1915 /*
1916  * Returns
1917  * 0: success
1918  * -1: error occurs
1919  */
1920 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
1921 {
1922     unsigned int l;
1923 
1924     while (size > 0) {
1925         /* Reassemble a packet from the network.
1926          * 0 = getting length.
1927          * 1 = getting vnet header length.
1928          * 2 = getting data.
1929          */
1930         switch (rs->state) {
1931         case 0:
1932             l = 4 - rs->index;
1933             if (l > size) {
1934                 l = size;
1935             }
1936             memcpy(rs->buf + rs->index, buf, l);
1937             buf += l;
1938             size -= l;
1939             rs->index += l;
1940             if (rs->index == 4) {
1941                 /* got length */
1942                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
1943                 rs->index = 0;
1944                 if (rs->vnet_hdr) {
1945                     rs->state = 1;
1946                 } else {
1947                     rs->state = 2;
1948                     rs->vnet_hdr_len = 0;
1949                 }
1950             }
1951             break;
1952         case 1:
1953             l = 4 - rs->index;
1954             if (l > size) {
1955                 l = size;
1956             }
1957             memcpy(rs->buf + rs->index, buf, l);
1958             buf += l;
1959             size -= l;
1960             rs->index += l;
1961             if (rs->index == 4) {
1962                 /* got vnet header length */
1963                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
1964                 rs->index = 0;
1965                 rs->state = 2;
1966             }
1967             break;
1968         case 2:
1969             l = rs->packet_len - rs->index;
1970             if (l > size) {
1971                 l = size;
1972             }
1973             if (rs->index + l <= sizeof(rs->buf)) {
1974                 memcpy(rs->buf + rs->index, buf, l);
1975             } else {
1976                 fprintf(stderr, "serious error: oversized packet received,"
1977                     "connection terminated.\n");
1978                 rs->index = rs->state = 0;
1979                 return -1;
1980             }
1981 
1982             rs->index += l;
1983             buf += l;
1984             size -= l;
1985             if (rs->index >= rs->packet_len) {
1986                 rs->index = 0;
1987                 rs->state = 0;
1988                 assert(rs->finalize);
1989                 rs->finalize(rs);
1990             }
1991             break;
1992         }
1993     }
1994 
1995     assert(size == 0);
1996     return 0;
1997 }
1998