xref: /openbmc/qemu/net/net.c (revision 409be85c2f54223673dde89940948508189941c8)
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 "qobject/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 "system/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 #include "standard-headers/linux/virtio_net.h"
60 
61 /* Net bridge is currently not supported for W32. */
62 #if !defined(_WIN32)
63 # define CONFIG_NET_BRIDGE
64 #endif
65 
66 static VMChangeStateEntry *net_change_state_entry;
67 NetClientStateList net_clients;
68 
69 typedef struct NetdevQueueEntry {
70     Netdev *nd;
71     Location loc;
72     QSIMPLEQ_ENTRY(NetdevQueueEntry) entry;
73 } NetdevQueueEntry;
74 
75 typedef QSIMPLEQ_HEAD(, NetdevQueueEntry) NetdevQueue;
76 
77 static NetdevQueue nd_queue = QSIMPLEQ_HEAD_INITIALIZER(nd_queue);
78 
79 static GHashTable *nic_model_help;
80 
81 static int nb_nics;
82 static NICInfo nd_table[MAX_NICS];
83 
84 /***********************************************************/
85 /* network device redirectors */
86 
87 int convert_host_port(struct sockaddr_in *saddr, const char *host,
88                       const char *port, Error **errp)
89 {
90     struct hostent *he;
91     const char *r;
92     long p;
93 
94     memset(saddr, 0, sizeof(*saddr));
95 
96     saddr->sin_family = AF_INET;
97     if (host[0] == '\0') {
98         saddr->sin_addr.s_addr = 0;
99     } else {
100         if (qemu_isdigit(host[0])) {
101             if (!inet_aton(host, &saddr->sin_addr)) {
102                 error_setg(errp, "host address '%s' is not a valid "
103                            "IPv4 address", host);
104                 return -1;
105             }
106         } else {
107             he = gethostbyname(host);
108             if (he == NULL) {
109                 error_setg(errp, "can't resolve host address '%s'", host);
110                 return -1;
111             }
112             saddr->sin_addr = *(struct in_addr *)he->h_addr;
113         }
114     }
115     if (qemu_strtol(port, &r, 0, &p) != 0) {
116         error_setg(errp, "port number '%s' is invalid", port);
117         return -1;
118     }
119     saddr->sin_port = htons(p);
120     return 0;
121 }
122 
123 int parse_host_port(struct sockaddr_in *saddr, const char *str,
124                     Error **errp)
125 {
126     gchar **substrings;
127     int ret;
128 
129     substrings = g_strsplit(str, ":", 2);
130     if (!substrings || !substrings[0] || !substrings[1]) {
131         error_setg(errp, "host address '%s' doesn't contain ':' "
132                    "separating host from port", str);
133         ret = -1;
134         goto out;
135     }
136 
137     ret = convert_host_port(saddr, substrings[0], substrings[1], errp);
138 
139 out:
140     g_strfreev(substrings);
141     return ret;
142 }
143 
144 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
145 {
146     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
147                            macaddr[0], macaddr[1], macaddr[2],
148                            macaddr[3], macaddr[4], macaddr[5]);
149 }
150 
151 void qemu_set_info_str(NetClientState *nc, const char *fmt, ...)
152 {
153     va_list ap;
154 
155     va_start(ap, fmt);
156     vsnprintf(nc->info_str, sizeof(nc->info_str), fmt, ap);
157     va_end(ap);
158 }
159 
160 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
161 {
162     qemu_set_info_str(nc, "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
163                       nc->model, macaddr[0], macaddr[1], macaddr[2],
164                       macaddr[3], macaddr[4], macaddr[5]);
165 }
166 
167 static int mac_table[256] = {0};
168 
169 static void qemu_macaddr_set_used(MACAddr *macaddr)
170 {
171     int index;
172 
173     for (index = 0x56; index < 0xFF; index++) {
174         if (macaddr->a[5] == index) {
175             mac_table[index]++;
176         }
177     }
178 }
179 
180 static void qemu_macaddr_set_free(MACAddr *macaddr)
181 {
182     int index;
183     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
184 
185     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
186         return;
187     }
188     for (index = 0x56; index < 0xFF; index++) {
189         if (macaddr->a[5] == index) {
190             mac_table[index]--;
191         }
192     }
193 }
194 
195 static int qemu_macaddr_get_free(void)
196 {
197     int index;
198 
199     for (index = 0x56; index < 0xFF; index++) {
200         if (mac_table[index] == 0) {
201             return index;
202         }
203     }
204 
205     return -1;
206 }
207 
208 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
209 {
210     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
211     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
212 
213     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
214         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
215             return;
216         } else {
217             qemu_macaddr_set_used(macaddr);
218             return;
219         }
220     }
221 
222     macaddr->a[0] = 0x52;
223     macaddr->a[1] = 0x54;
224     macaddr->a[2] = 0x00;
225     macaddr->a[3] = 0x12;
226     macaddr->a[4] = 0x34;
227     macaddr->a[5] = qemu_macaddr_get_free();
228     qemu_macaddr_set_used(macaddr);
229 }
230 
231 /**
232  * Generate a name for net client
233  *
234  * Only net clients created with the legacy -net option and NICs need this.
235  */
236 static char *assign_name(NetClientState *nc1, const char *model)
237 {
238     NetClientState *nc;
239     int id = 0;
240 
241     QTAILQ_FOREACH(nc, &net_clients, next) {
242         if (nc == nc1) {
243             continue;
244         }
245         if (strcmp(nc->model, model) == 0) {
246             id++;
247         }
248     }
249 
250     return g_strdup_printf("%s.%d", model, id);
251 }
252 
253 static void qemu_net_client_destructor(NetClientState *nc)
254 {
255     g_free(nc);
256 }
257 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
258                                        unsigned flags,
259                                        const struct iovec *iov,
260                                        int iovcnt,
261                                        void *opaque);
262 
263 static void qemu_net_client_setup(NetClientState *nc,
264                                   NetClientInfo *info,
265                                   NetClientState *peer,
266                                   const char *model,
267                                   const char *name,
268                                   NetClientDestructor *destructor,
269                                   bool is_datapath)
270 {
271     nc->info = info;
272     nc->model = g_strdup(model);
273     if (name) {
274         nc->name = g_strdup(name);
275     } else {
276         nc->name = assign_name(nc, model);
277     }
278 
279     if (peer) {
280         assert(!peer->peer);
281         nc->peer = peer;
282         peer->peer = nc;
283     }
284     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
285 
286     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
287     nc->destructor = destructor;
288     nc->is_datapath = is_datapath;
289     QTAILQ_INIT(&nc->filters);
290 }
291 
292 NetClientState *qemu_new_net_client(NetClientInfo *info,
293                                     NetClientState *peer,
294                                     const char *model,
295                                     const char *name)
296 {
297     NetClientState *nc;
298 
299     assert(info->size >= sizeof(NetClientState));
300 
301     nc = g_malloc0(info->size);
302     qemu_net_client_setup(nc, info, peer, model, name,
303                           qemu_net_client_destructor, true);
304 
305     return nc;
306 }
307 
308 NetClientState *qemu_new_net_control_client(NetClientInfo *info,
309                                             NetClientState *peer,
310                                             const char *model,
311                                             const char *name)
312 {
313     NetClientState *nc;
314 
315     assert(info->size >= sizeof(NetClientState));
316 
317     nc = g_malloc0(info->size);
318     qemu_net_client_setup(nc, info, peer, model, name,
319                           qemu_net_client_destructor, false);
320 
321     return nc;
322 }
323 
324 NICState *qemu_new_nic(NetClientInfo *info,
325                        NICConf *conf,
326                        const char *model,
327                        const char *name,
328                        MemReentrancyGuard *reentrancy_guard,
329                        void *opaque)
330 {
331     NetClientState **peers = conf->peers.ncs;
332     NICState *nic;
333     int i, queues = MAX(1, conf->peers.queues);
334 
335     assert(info->type == NET_CLIENT_DRIVER_NIC);
336     assert(info->size >= sizeof(NICState));
337 
338     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
339     nic->ncs = (void *)nic + info->size;
340     nic->conf = conf;
341     nic->reentrancy_guard = reentrancy_guard,
342     nic->opaque = opaque;
343 
344     for (i = 0; i < queues; i++) {
345         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
346                               NULL, true);
347         nic->ncs[i].queue_index = i;
348     }
349 
350     return nic;
351 }
352 
353 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
354 {
355     return nic->ncs + queue_index;
356 }
357 
358 NetClientState *qemu_get_queue(NICState *nic)
359 {
360     return qemu_get_subqueue(nic, 0);
361 }
362 
363 NICState *qemu_get_nic(NetClientState *nc)
364 {
365     NetClientState *nc0 = nc - nc->queue_index;
366 
367     return (NICState *)((void *)nc0 - nc->info->size);
368 }
369 
370 void *qemu_get_nic_opaque(NetClientState *nc)
371 {
372     NICState *nic = qemu_get_nic(nc);
373 
374     return nic->opaque;
375 }
376 
377 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
378 {
379     assert(nc != NULL);
380     NetClientState *ncs = nc + queue_index;
381     return ncs->peer;
382 }
383 
384 static void qemu_cleanup_net_client(NetClientState *nc,
385                                     bool remove_from_net_clients)
386 {
387     if (remove_from_net_clients) {
388         QTAILQ_REMOVE(&net_clients, nc, next);
389     }
390 
391     if (nc->info->cleanup) {
392         nc->info->cleanup(nc);
393     }
394 }
395 
396 static void qemu_free_net_client(NetClientState *nc)
397 {
398     if (nc->incoming_queue) {
399         qemu_del_net_queue(nc->incoming_queue);
400     }
401     if (nc->peer) {
402         nc->peer->peer = NULL;
403     }
404     g_free(nc->name);
405     g_free(nc->model);
406     if (nc->destructor) {
407         nc->destructor(nc);
408     }
409 }
410 
411 void qemu_del_net_client(NetClientState *nc)
412 {
413     NetClientState *ncs[MAX_QUEUE_NUM];
414     int queues, i;
415     NetFilterState *nf, *next;
416 
417     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
418 
419     /* If the NetClientState belongs to a multiqueue backend, we will change all
420      * other NetClientStates also.
421      */
422     queues = qemu_find_net_clients_except(nc->name, ncs,
423                                           NET_CLIENT_DRIVER_NIC,
424                                           MAX_QUEUE_NUM);
425     assert(queues != 0);
426 
427     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
428         object_unparent(OBJECT(nf));
429     }
430 
431     /*
432      * If there is a peer NIC, transfer ownership to it.  Delete the client
433      * from net_client list but do not cleanup nor free.  This way NIC can
434      * still access to members of the backend.
435      *
436      * The cleanup and free will be done when the NIC is free.
437      */
438     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
439         NICState *nic = qemu_get_nic(nc->peer);
440         if (nic->peer_deleted) {
441             return;
442         }
443         nic->peer_deleted = true;
444 
445         for (i = 0; i < queues; i++) {
446             ncs[i]->peer->link_down = true;
447             QTAILQ_REMOVE(&net_clients, ncs[i], next);
448         }
449 
450         if (nc->peer->info->link_status_changed) {
451             nc->peer->info->link_status_changed(nc->peer);
452         }
453 
454         return;
455     }
456 
457     for (i = 0; i < queues; i++) {
458         qemu_cleanup_net_client(ncs[i], true);
459         qemu_free_net_client(ncs[i]);
460     }
461 }
462 
463 void qemu_del_nic(NICState *nic)
464 {
465     int i, queues = MAX(nic->conf->peers.queues, 1);
466 
467     qemu_macaddr_set_free(&nic->conf->macaddr);
468 
469     for (i = 0; i < queues; i++) {
470         NetClientState *nc = qemu_get_subqueue(nic, i);
471         /*
472          * If this is a peer NIC and peer has already been deleted, clean it up
473          * and free it now.
474          */
475         if (nic->peer_deleted) {
476             qemu_cleanup_net_client(nc->peer, false);
477             qemu_free_net_client(nc->peer);
478         } else if (nc->peer) {
479             /* if there are RX packets pending, complete them */
480             qemu_purge_queued_packets(nc->peer);
481         }
482     }
483 
484     for (i = queues - 1; i >= 0; i--) {
485         NetClientState *nc = qemu_get_subqueue(nic, i);
486 
487         qemu_cleanup_net_client(nc, true);
488         qemu_free_net_client(nc);
489     }
490 
491     g_free(nic);
492 }
493 
494 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
495 {
496     NetClientState *nc;
497 
498     QTAILQ_FOREACH(nc, &net_clients, next) {
499         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
500             if (nc->queue_index == 0) {
501                 func(qemu_get_nic(nc), opaque);
502             }
503         }
504     }
505 }
506 
507 bool qemu_has_ufo(NetClientState *nc)
508 {
509     if (!nc || !nc->info->has_ufo) {
510         return false;
511     }
512 
513     return nc->info->has_ufo(nc);
514 }
515 
516 bool qemu_has_uso(NetClientState *nc)
517 {
518     if (!nc || !nc->info->has_uso) {
519         return false;
520     }
521 
522     return nc->info->has_uso(nc);
523 }
524 
525 bool qemu_has_tunnel(NetClientState *nc)
526 {
527     if (!nc || !nc->info->has_tunnel) {
528         return false;
529     }
530 
531     return nc->info->has_tunnel(nc);
532 }
533 
534 bool qemu_has_vnet_hdr(NetClientState *nc)
535 {
536     if (!nc || !nc->info->has_vnet_hdr) {
537         return false;
538     }
539 
540     return nc->info->has_vnet_hdr(nc);
541 }
542 
543 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
544 {
545     if (!nc || !nc->info->has_vnet_hdr_len) {
546         return false;
547     }
548 
549     return nc->info->has_vnet_hdr_len(nc, len);
550 }
551 
552 void qemu_set_offload(NetClientState *nc, const NetOffloads *ol)
553 {
554     if (!nc || !nc->info->set_offload) {
555         return;
556     }
557 
558     nc->info->set_offload(nc, ol);
559 }
560 
561 int qemu_get_vnet_hdr_len(NetClientState *nc)
562 {
563     if (!nc) {
564         return 0;
565     }
566 
567     return nc->vnet_hdr_len;
568 }
569 
570 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
571 {
572     if (!nc || !nc->info->set_vnet_hdr_len) {
573         return;
574     }
575 
576     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
577            len == sizeof(struct virtio_net_hdr) ||
578            len == sizeof(struct virtio_net_hdr_v1_hash) ||
579            len == sizeof(struct virtio_net_hdr_v1_hash_tunnel));
580 
581     nc->vnet_hdr_len = len;
582     nc->info->set_vnet_hdr_len(nc, len);
583 }
584 
585 bool qemu_get_vnet_hash_supported_types(NetClientState *nc, uint32_t *types)
586 {
587     if (!nc || !nc->info->get_vnet_hash_supported_types) {
588         return false;
589     }
590 
591     return nc->info->get_vnet_hash_supported_types(nc, types);
592 }
593 
594 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
595 {
596 #if HOST_BIG_ENDIAN
597     if (!nc || !nc->info->set_vnet_le) {
598         return -ENOSYS;
599     }
600 
601     return nc->info->set_vnet_le(nc, is_le);
602 #else
603     return 0;
604 #endif
605 }
606 
607 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
608 {
609 #if HOST_BIG_ENDIAN
610     return 0;
611 #else
612     if (!nc || !nc->info->set_vnet_be) {
613         return -ENOSYS;
614     }
615 
616     return nc->info->set_vnet_be(nc, is_be);
617 #endif
618 }
619 
620 int qemu_can_receive_packet(NetClientState *nc)
621 {
622     if (nc->receive_disabled) {
623         return 0;
624     } else if (nc->info->can_receive &&
625                !nc->info->can_receive(nc)) {
626         return 0;
627     }
628     return 1;
629 }
630 
631 int qemu_can_send_packet(NetClientState *sender)
632 {
633     int vm_running = runstate_is_running();
634 
635     if (!vm_running) {
636         return 0;
637     }
638 
639     if (!sender->peer) {
640         return 1;
641     }
642 
643     return qemu_can_receive_packet(sender->peer);
644 }
645 
646 static ssize_t filter_receive_iov(NetClientState *nc,
647                                   NetFilterDirection direction,
648                                   NetClientState *sender,
649                                   unsigned flags,
650                                   const struct iovec *iov,
651                                   int iovcnt,
652                                   NetPacketSent *sent_cb)
653 {
654     ssize_t ret = 0;
655     NetFilterState *nf = NULL;
656 
657     if (direction == NET_FILTER_DIRECTION_TX) {
658         QTAILQ_FOREACH(nf, &nc->filters, next) {
659             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
660                                          iovcnt, sent_cb);
661             if (ret) {
662                 return ret;
663             }
664         }
665     } else {
666         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
667             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
668                                          iovcnt, sent_cb);
669             if (ret) {
670                 return ret;
671             }
672         }
673     }
674 
675     return ret;
676 }
677 
678 static ssize_t filter_receive(NetClientState *nc,
679                               NetFilterDirection direction,
680                               NetClientState *sender,
681                               unsigned flags,
682                               const uint8_t *data,
683                               size_t size,
684                               NetPacketSent *sent_cb)
685 {
686     struct iovec iov = {
687         .iov_base = (void *)data,
688         .iov_len = size
689     };
690 
691     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
692 }
693 
694 void qemu_purge_queued_packets(NetClientState *nc)
695 {
696     if (!nc->peer) {
697         return;
698     }
699 
700     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
701 }
702 
703 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
704 {
705     nc->receive_disabled = 0;
706 
707     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
708         if (net_hub_flush(nc->peer)) {
709             qemu_notify_event();
710         }
711     }
712     if (qemu_net_queue_flush(nc->incoming_queue)) {
713         /* We emptied the queue successfully, signal to the IO thread to repoll
714          * the file descriptor (for tap, for example).
715          */
716         qemu_notify_event();
717     } else if (purge) {
718         /* Unable to empty the queue, purge remaining packets */
719         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
720     }
721 }
722 
723 void qemu_flush_queued_packets(NetClientState *nc)
724 {
725     qemu_flush_or_purge_queued_packets(nc, false);
726 }
727 
728 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
729                                                  unsigned flags,
730                                                  const uint8_t *buf, int size,
731                                                  NetPacketSent *sent_cb)
732 {
733     NetQueue *queue;
734     int ret;
735 
736 #ifdef DEBUG_NET
737     printf("qemu_send_packet_async:\n");
738     qemu_hexdump(stdout, "net", buf, size);
739 #endif
740 
741     if (sender->link_down || !sender->peer) {
742         return size;
743     }
744 
745     /* Let filters handle the packet first */
746     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
747                          sender, flags, buf, size, sent_cb);
748     if (ret) {
749         return ret;
750     }
751 
752     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
753                          sender, flags, buf, size, sent_cb);
754     if (ret) {
755         return ret;
756     }
757 
758     queue = sender->peer->incoming_queue;
759 
760     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
761 }
762 
763 ssize_t qemu_send_packet_async(NetClientState *sender,
764                                const uint8_t *buf, int size,
765                                NetPacketSent *sent_cb)
766 {
767     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
768                                              buf, size, sent_cb);
769 }
770 
771 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
772 {
773     return qemu_send_packet_async(nc, buf, size, NULL);
774 }
775 
776 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
777 {
778     uint8_t min_pkt[ETH_ZLEN];
779     size_t min_pktsz = sizeof(min_pkt);
780 
781     if (!qemu_can_receive_packet(nc)) {
782         return 0;
783     }
784 
785     if (net_peer_needs_padding(nc)) {
786         if (eth_pad_short_frame(min_pkt, &min_pktsz, buf, size)) {
787             buf = min_pkt;
788             size = min_pktsz;
789         }
790     }
791 
792     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
793 }
794 
795 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
796 {
797     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
798                                              buf, size, NULL);
799 }
800 
801 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
802                                int iovcnt, unsigned flags)
803 {
804     uint8_t *buf = NULL;
805     uint8_t *buffer;
806     size_t offset;
807     ssize_t ret;
808 
809     if (iovcnt == 1) {
810         buffer = iov[0].iov_base;
811         offset = iov[0].iov_len;
812     } else {
813         offset = iov_size(iov, iovcnt);
814         if (offset > NET_BUFSIZE) {
815             return -1;
816         }
817         buf = g_malloc(offset);
818         buffer = buf;
819         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
820     }
821 
822     ret = nc->info->receive(nc, buffer, offset);
823 
824     g_free(buf);
825     return ret;
826 }
827 
828 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
829                                        unsigned flags,
830                                        const struct iovec *iov,
831                                        int iovcnt,
832                                        void *opaque)
833 {
834     MemReentrancyGuard *owned_reentrancy_guard;
835     NetClientState *nc = opaque;
836     int ret;
837     struct virtio_net_hdr_v1_hash vnet_hdr = { };
838     g_autofree struct iovec *iov_copy = NULL;
839 
840 
841     if (nc->link_down) {
842         return iov_size(iov, iovcnt);
843     }
844 
845     if (nc->receive_disabled) {
846         return 0;
847     }
848 
849     if (nc->info->type != NET_CLIENT_DRIVER_NIC ||
850         qemu_get_nic(nc)->reentrancy_guard->engaged_in_io) {
851         owned_reentrancy_guard = NULL;
852     } else {
853         owned_reentrancy_guard = qemu_get_nic(nc)->reentrancy_guard;
854         owned_reentrancy_guard->engaged_in_io = true;
855     }
856 
857     if ((flags & QEMU_NET_PACKET_FLAG_RAW) && nc->vnet_hdr_len) {
858         iov_copy = g_new(struct iovec, iovcnt + 1);
859         iov_copy[0].iov_base = &vnet_hdr;
860         iov_copy[0].iov_len =  nc->vnet_hdr_len;
861         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
862         iov = iov_copy;
863         iovcnt++;
864     }
865 
866     if (nc->info->receive_iov) {
867         ret = nc->info->receive_iov(nc, iov, iovcnt);
868     } else {
869         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
870     }
871 
872     if (owned_reentrancy_guard) {
873         owned_reentrancy_guard->engaged_in_io = false;
874     }
875 
876     if (ret == 0) {
877         nc->receive_disabled = 1;
878     }
879 
880     return ret;
881 }
882 
883 ssize_t qemu_sendv_packet_async(NetClientState *sender,
884                                 const struct iovec *iov, int iovcnt,
885                                 NetPacketSent *sent_cb)
886 {
887     NetQueue *queue;
888     size_t size = iov_size(iov, iovcnt);
889     int ret;
890 
891     if (size > NET_BUFSIZE) {
892         return size;
893     }
894 
895     if (sender->link_down || !sender->peer) {
896         return size;
897     }
898 
899     /* Let filters handle the packet first */
900     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
901                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
902     if (ret) {
903         return ret;
904     }
905 
906     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
907                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
908     if (ret) {
909         return ret;
910     }
911 
912     queue = sender->peer->incoming_queue;
913 
914     return qemu_net_queue_send_iov(queue, sender,
915                                    QEMU_NET_PACKET_FLAG_NONE,
916                                    iov, iovcnt, sent_cb);
917 }
918 
919 ssize_t
920 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
921 {
922     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
923 }
924 
925 NetClientState *qemu_find_netdev(const char *id)
926 {
927     NetClientState *nc;
928 
929     QTAILQ_FOREACH(nc, &net_clients, next) {
930         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
931             continue;
932         if (!strcmp(nc->name, id)) {
933             return nc;
934         }
935     }
936 
937     return NULL;
938 }
939 
940 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
941                                  NetClientDriver type, int max)
942 {
943     NetClientState *nc;
944     int ret = 0;
945 
946     QTAILQ_FOREACH(nc, &net_clients, next) {
947         if (nc->info->type == type) {
948             continue;
949         }
950         if (!id || !strcmp(nc->name, id)) {
951             if (ret < max) {
952                 ncs[ret] = nc;
953             }
954             ret++;
955         }
956     }
957 
958     return ret;
959 }
960 
961 static int nic_get_free_idx(void)
962 {
963     int index;
964 
965     for (index = 0; index < MAX_NICS; index++)
966         if (!nd_table[index].used)
967             return index;
968     return -1;
969 }
970 
971 GPtrArray *qemu_get_nic_models(const char *device_type)
972 {
973     GPtrArray *nic_models = g_ptr_array_new();
974     GSList *list = object_class_get_list_sorted(device_type, false);
975 
976     while (list) {
977         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
978                                              TYPE_DEVICE);
979         GSList *next;
980         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
981             dc->user_creatable) {
982             const char *name = object_class_get_name(list->data);
983             /*
984              * A network device might also be something else than a NIC, see
985              * e.g. the "rocker" device. Thus we have to look for the "netdev"
986              * property, too. Unfortunately, some devices like virtio-net only
987              * create this property during instance_init, so we have to create
988              * a temporary instance here to be able to check it.
989              */
990             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
991             if (object_property_find(obj, "netdev")) {
992                 g_ptr_array_add(nic_models, (gpointer)name);
993             }
994             object_unref(obj);
995         }
996         next = list->next;
997         g_slist_free_1(list);
998         list = next;
999     }
1000     g_ptr_array_add(nic_models, NULL);
1001 
1002     return nic_models;
1003 }
1004 
1005 static int net_init_nic(const Netdev *netdev, const char *name,
1006                         NetClientState *peer, Error **errp)
1007 {
1008     int idx;
1009     NICInfo *nd;
1010     const NetLegacyNicOptions *nic;
1011 
1012     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
1013     nic = &netdev->u.nic;
1014 
1015     idx = nic_get_free_idx();
1016     if (idx == -1 || nb_nics >= MAX_NICS) {
1017         error_setg(errp, "too many NICs");
1018         return -1;
1019     }
1020 
1021     nd = &nd_table[idx];
1022 
1023     memset(nd, 0, sizeof(*nd));
1024 
1025     if (nic->netdev) {
1026         nd->netdev = qemu_find_netdev(nic->netdev);
1027         if (!nd->netdev) {
1028             error_setg(errp, "netdev '%s' not found", nic->netdev);
1029             return -1;
1030         }
1031     } else {
1032         assert(peer);
1033         nd->netdev = peer;
1034     }
1035     nd->name = g_strdup(name);
1036     if (nic->model) {
1037         nd->model = g_strdup(nic->model);
1038     }
1039     if (nic->addr) {
1040         nd->devaddr = g_strdup(nic->addr);
1041     }
1042 
1043     if (nic->macaddr &&
1044         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1045         error_setg(errp, "invalid syntax for ethernet address");
1046         return -1;
1047     }
1048     if (nic->macaddr &&
1049         is_multicast_ether_addr(nd->macaddr.a)) {
1050         error_setg(errp,
1051                    "NIC cannot have multicast MAC address (odd 1st byte)");
1052         return -1;
1053     }
1054     qemu_macaddr_default_if_unset(&nd->macaddr);
1055 
1056     if (nic->has_vectors) {
1057         if (nic->vectors > 0x7ffffff) {
1058             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1059             return -1;
1060         }
1061         nd->nvectors = nic->vectors;
1062     } else {
1063         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1064     }
1065 
1066     nd->used = 1;
1067     nb_nics++;
1068 
1069     return idx;
1070 }
1071 
1072 static gboolean add_nic_result(gpointer key, gpointer value, gpointer user_data)
1073 {
1074     GPtrArray *results = user_data;
1075     GPtrArray *alias_list = value;
1076     const char *model = key;
1077     char *result;
1078 
1079     if (!alias_list) {
1080         result = g_strdup(model);
1081     } else {
1082         GString *result_str = g_string_new(model);
1083         int i;
1084 
1085         g_string_append(result_str, " (aka ");
1086         for (i = 0; i < alias_list->len; i++) {
1087             if (i) {
1088                 g_string_append(result_str, ", ");
1089             }
1090             g_string_append(result_str, alias_list->pdata[i]);
1091         }
1092         g_string_append(result_str, ")");
1093         result = result_str->str;
1094         g_string_free(result_str, false);
1095         g_ptr_array_unref(alias_list);
1096     }
1097     g_ptr_array_add(results, result);
1098     return true;
1099 }
1100 
1101 static int model_cmp(char **a, char **b)
1102 {
1103     return strcmp(*a, *b);
1104 }
1105 
1106 static void show_nic_models(void)
1107 {
1108     GPtrArray *results = g_ptr_array_new();
1109     int i;
1110 
1111     g_hash_table_foreach_remove(nic_model_help, add_nic_result, results);
1112     g_ptr_array_sort(results, (GCompareFunc)model_cmp);
1113 
1114     printf("Available NIC models for this configuration:\n");
1115     for (i = 0 ; i < results->len; i++) {
1116         printf("%s\n", (char *)results->pdata[i]);
1117     }
1118     g_hash_table_unref(nic_model_help);
1119     nic_model_help = NULL;
1120 }
1121 
1122 static void add_nic_model_help(const char *model, const char *alias)
1123 {
1124     GPtrArray *alias_list = NULL;
1125 
1126     if (g_hash_table_lookup_extended(nic_model_help, model, NULL,
1127                                      (gpointer *)&alias_list)) {
1128         /* Already exists, no alias to add: return */
1129         if (!alias) {
1130             return;
1131         }
1132         if (alias_list) {
1133             /* Check if this alias is already in the list. Add if not. */
1134             if (!g_ptr_array_find_with_equal_func(alias_list, alias,
1135                                                   g_str_equal, NULL)) {
1136                 g_ptr_array_add(alias_list, g_strdup(alias));
1137             }
1138             return;
1139         }
1140     }
1141     /* Either this model wasn't in the list already, or a first alias added */
1142     if (alias) {
1143         alias_list = g_ptr_array_new();
1144         g_ptr_array_set_free_func(alias_list, g_free);
1145         g_ptr_array_add(alias_list, g_strdup(alias));
1146     }
1147     g_hash_table_replace(nic_model_help, g_strdup(model), alias_list);
1148 }
1149 
1150 NICInfo *qemu_find_nic_info(const char *typename, bool match_default,
1151                             const char *alias)
1152 {
1153     NICInfo *nd;
1154     int i;
1155 
1156     if (nic_model_help) {
1157         add_nic_model_help(typename, alias);
1158     }
1159 
1160     for (i = 0; i < nb_nics; i++) {
1161         nd = &nd_table[i];
1162 
1163         if (!nd->used || nd->instantiated) {
1164             continue;
1165         }
1166 
1167         if ((match_default && !nd->model) || !g_strcmp0(nd->model, typename)
1168             || (alias && !g_strcmp0(nd->model, alias))) {
1169             return nd;
1170         }
1171     }
1172     return NULL;
1173 }
1174 
1175 static bool is_nic_model_help_option(const char *model)
1176 {
1177     if (model && is_help_option(model)) {
1178         /*
1179          * Trigger the help output by instantiating the hash table which
1180          * will gather tha available models as they get registered.
1181          */
1182         if (!nic_model_help) {
1183             nic_model_help = g_hash_table_new_full(g_str_hash, g_str_equal,
1184                                                    g_free, NULL);
1185         }
1186         return true;
1187     }
1188     return false;
1189 }
1190 
1191 /* "I have created a device. Please configure it if you can" */
1192 bool qemu_configure_nic_device(DeviceState *dev, bool match_default,
1193                                const char *alias)
1194 {
1195     NICInfo *nd = qemu_find_nic_info(object_get_typename(OBJECT(dev)),
1196                                      match_default, alias);
1197 
1198     if (nd) {
1199         qdev_set_nic_properties(dev, nd);
1200         return true;
1201     }
1202     return false;
1203 }
1204 
1205 /* "Please create a device, if you have a configuration for it" */
1206 DeviceState *qemu_create_nic_device(const char *typename, bool match_default,
1207                                     const char *alias)
1208 {
1209     NICInfo *nd = qemu_find_nic_info(typename, match_default, alias);
1210     DeviceState *dev;
1211 
1212     if (!nd) {
1213         return NULL;
1214     }
1215 
1216     dev = qdev_new(typename);
1217     qdev_set_nic_properties(dev, nd);
1218     return dev;
1219 }
1220 
1221 void qemu_create_nic_bus_devices(BusState *bus, const char *parent_type,
1222                                  const char *default_model,
1223                                  const char *alias, const char *alias_target)
1224 {
1225     GPtrArray *nic_models = qemu_get_nic_models(parent_type);
1226     const char *model;
1227     DeviceState *dev;
1228     NICInfo *nd;
1229     int i;
1230 
1231     if (nic_model_help) {
1232         if (alias_target) {
1233             add_nic_model_help(alias_target, alias);
1234         }
1235         for (i = 0; i < nic_models->len - 1; i++) {
1236             add_nic_model_help(nic_models->pdata[i], NULL);
1237         }
1238     }
1239 
1240     /* Drop the NULL terminator which would make g_str_equal() unhappy */
1241     nic_models->len--;
1242 
1243     for (i = 0; i < nb_nics; i++) {
1244         nd = &nd_table[i];
1245 
1246         if (!nd->used || nd->instantiated) {
1247             continue;
1248         }
1249 
1250         model = nd->model ? nd->model : default_model;
1251         if (!model) {
1252             continue;
1253         }
1254 
1255         /* Each bus type is allowed *one* substitution */
1256         if (g_str_equal(model, alias)) {
1257             model = alias_target;
1258         }
1259 
1260         if (!g_ptr_array_find_with_equal_func(nic_models, model,
1261                                               g_str_equal, NULL)) {
1262             /* This NIC does not live on this bus. */
1263             continue;
1264         }
1265 
1266         dev = qdev_new(model);
1267         qdev_set_nic_properties(dev, nd);
1268         qdev_realize_and_unref(dev, bus, &error_fatal);
1269     }
1270 
1271     g_ptr_array_free(nic_models, true);
1272 }
1273 
1274 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1275     const Netdev *netdev,
1276     const char *name,
1277     NetClientState *peer, Error **errp) = {
1278         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1279 #ifdef CONFIG_PASST
1280         [NET_CLIENT_DRIVER_PASST]     = net_init_passt,
1281 #endif
1282 #ifdef CONFIG_SLIRP
1283         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1284 #endif
1285         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1286         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1287         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1288         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1289 #ifdef CONFIG_VDE
1290         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1291 #endif
1292 #ifdef CONFIG_NETMAP
1293         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1294 #endif
1295 #ifdef CONFIG_AF_XDP
1296         [NET_CLIENT_DRIVER_AF_XDP]    = net_init_af_xdp,
1297 #endif
1298 #ifdef CONFIG_NET_BRIDGE
1299         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1300 #endif
1301         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1302 #ifdef CONFIG_VHOST_NET_USER
1303         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1304 #endif
1305 #ifdef CONFIG_VHOST_NET_VDPA
1306         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1307 #endif
1308 #ifdef CONFIG_L2TPV3
1309         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1310 #endif
1311 #ifdef CONFIG_VMNET
1312         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1313         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1314         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1315 #endif /* CONFIG_VMNET */
1316 };
1317 
1318 
1319 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1320 {
1321     NetClientState *peer = NULL;
1322     NetClientState *nc;
1323 
1324     if (is_netdev) {
1325         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1326             !net_client_init_fun[netdev->type]) {
1327             error_setg(errp, "network backend '%s' is not compiled into this binary",
1328                        NetClientDriver_str(netdev->type));
1329             return -1;
1330         }
1331     } else {
1332         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1333             return 0; /* nothing to do */
1334         }
1335         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1336             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1337                        NetClientDriver_str(netdev->type));
1338             return -1;
1339         }
1340 
1341         if (!net_client_init_fun[netdev->type]) {
1342             error_setg(errp, "network backend '%s' is not compiled into this binary",
1343                        NetClientDriver_str(netdev->type));
1344             return -1;
1345         }
1346 
1347         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1348         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1349             !netdev->u.nic.netdev) {
1350             peer = net_hub_add_port(0, NULL, NULL);
1351         }
1352     }
1353 
1354     nc = qemu_find_netdev(netdev->id);
1355     if (nc) {
1356         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1357         return -1;
1358     }
1359 
1360     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1361         /* FIXME drop when all init functions store an Error */
1362         if (errp && !*errp) {
1363             error_setg(errp, "Device '%s' could not be initialized",
1364                        NetClientDriver_str(netdev->type));
1365         }
1366         return -1;
1367     }
1368 
1369     if (is_netdev) {
1370         nc = qemu_find_netdev(netdev->id);
1371         assert(nc);
1372         nc->is_netdev = true;
1373     }
1374 
1375     return 0;
1376 }
1377 
1378 void show_netdevs(void)
1379 {
1380     int idx;
1381     const char *available_netdevs[] = {
1382         "socket",
1383         "stream",
1384         "dgram",
1385         "hubport",
1386         "tap",
1387         "passt",
1388 #ifdef CONFIG_SLIRP
1389         "user",
1390 #endif
1391 #ifdef CONFIG_L2TPV3
1392         "l2tpv3",
1393 #endif
1394 #ifdef CONFIG_VDE
1395         "vde",
1396 #endif
1397 #ifdef CONFIG_NET_BRIDGE
1398         "bridge",
1399 #endif
1400 #ifdef CONFIG_NETMAP
1401         "netmap",
1402 #endif
1403 #ifdef CONFIG_AF_XDP
1404         "af-xdp",
1405 #endif
1406 #ifdef CONFIG_POSIX
1407         "vhost-user",
1408 #endif
1409 #ifdef CONFIG_VHOST_VDPA
1410         "vhost-vdpa",
1411 #endif
1412 #ifdef CONFIG_VMNET
1413         "vmnet-host",
1414         "vmnet-shared",
1415         "vmnet-bridged",
1416 #endif
1417     };
1418 
1419     qemu_printf("Available netdev backend types:\n");
1420     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1421         qemu_printf("%s\n", available_netdevs[idx]);
1422     }
1423 }
1424 
1425 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1426 {
1427     gchar **substrings = NULL;
1428     Netdev *object = NULL;
1429     int ret = -1;
1430     Visitor *v = opts_visitor_new(opts);
1431 
1432     /* Parse convenience option format ipv6-net=fec0::0[/64] */
1433     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1434 
1435     if (ip6_net) {
1436         char *prefix_addr;
1437         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1438 
1439         substrings = g_strsplit(ip6_net, "/", 2);
1440         if (!substrings || !substrings[0]) {
1441             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1442                        "a valid IPv6 prefix");
1443             goto out;
1444         }
1445 
1446         prefix_addr = substrings[0];
1447 
1448         /* Handle user-specified prefix length. */
1449         if (substrings[1] &&
1450             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1451         {
1452             error_setg(errp,
1453                        "parameter 'ipv6-net' expects a number after '/'");
1454             goto out;
1455         }
1456 
1457         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1458         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1459                             &error_abort);
1460         qemu_opt_unset(opts, "ipv6-net");
1461     }
1462 
1463     /* Create an ID for -net if the user did not specify one */
1464     if (!is_netdev && !qemu_opts_id(opts)) {
1465         qemu_opts_set_id(opts, id_generate(ID_NET));
1466     }
1467 
1468     if (visit_type_Netdev(v, NULL, &object, errp)) {
1469         ret = net_client_init1(object, is_netdev, errp);
1470     }
1471 
1472     qapi_free_Netdev(object);
1473 
1474 out:
1475     g_strfreev(substrings);
1476     visit_free(v);
1477     return ret;
1478 }
1479 
1480 void netdev_add(QemuOpts *opts, Error **errp)
1481 {
1482     net_client_init(opts, true, errp);
1483 }
1484 
1485 void qmp_netdev_add(Netdev *netdev, Error **errp)
1486 {
1487     if (!id_wellformed(netdev->id)) {
1488         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1489         return;
1490     }
1491 
1492     net_client_init1(netdev, true, errp);
1493 }
1494 
1495 void qmp_netdev_del(const char *id, Error **errp)
1496 {
1497     NetClientState *nc;
1498     QemuOpts *opts;
1499 
1500     nc = qemu_find_netdev(id);
1501     if (!nc) {
1502         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1503                   "Device '%s' not found", id);
1504         return;
1505     }
1506 
1507     if (!nc->is_netdev) {
1508         error_setg(errp, "Device '%s' is not a netdev", id);
1509         return;
1510     }
1511 
1512     qemu_del_net_client(nc);
1513 
1514     /*
1515      * Wart: we need to delete the QemuOpts associated with netdevs
1516      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1517      * HMP netdev_add.
1518      */
1519     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1520     if (opts) {
1521         qemu_opts_del(opts);
1522     }
1523 }
1524 
1525 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1526 {
1527     char *str;
1528     ObjectProperty *prop;
1529     ObjectPropertyIterator iter;
1530     Visitor *v;
1531 
1532     /* generate info str */
1533     object_property_iter_init(&iter, OBJECT(nf));
1534     while ((prop = object_property_iter_next(&iter))) {
1535         if (!strcmp(prop->name, "type")) {
1536             continue;
1537         }
1538         v = string_output_visitor_new(false, &str);
1539         object_property_get(OBJECT(nf), prop->name, v, NULL);
1540         visit_complete(v, &str);
1541         visit_free(v);
1542         monitor_printf(mon, ",%s=%s", prop->name, str);
1543         g_free(str);
1544     }
1545     monitor_printf(mon, "\n");
1546 }
1547 
1548 void print_net_client(Monitor *mon, NetClientState *nc)
1549 {
1550     NetFilterState *nf;
1551 
1552     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1553                    nc->queue_index,
1554                    NetClientDriver_str(nc->info->type),
1555                    nc->info_str);
1556     if (!QTAILQ_EMPTY(&nc->filters)) {
1557         monitor_printf(mon, "filters:\n");
1558     }
1559     QTAILQ_FOREACH(nf, &nc->filters, next) {
1560         monitor_printf(mon, "  - %s: type=%s",
1561                        object_get_canonical_path_component(OBJECT(nf)),
1562                        object_get_typename(OBJECT(nf)));
1563         netfilter_print_info(mon, nf);
1564     }
1565 }
1566 
1567 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1568 {
1569     NetClientState *nc;
1570     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1571 
1572     QTAILQ_FOREACH(nc, &net_clients, next) {
1573         RxFilterInfo *info;
1574 
1575         if (name && strcmp(nc->name, name) != 0) {
1576             continue;
1577         }
1578 
1579         /* only query rx-filter information of NIC */
1580         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1581             if (name) {
1582                 error_setg(errp, "net client(%s) isn't a NIC", name);
1583                 assert(!filter_list);
1584                 return NULL;
1585             }
1586             continue;
1587         }
1588 
1589         /* only query information on queue 0 since the info is per nic,
1590          * not per queue
1591          */
1592         if (nc->queue_index != 0)
1593             continue;
1594 
1595         if (nc->info->query_rx_filter) {
1596             info = nc->info->query_rx_filter(nc);
1597             QAPI_LIST_APPEND(tail, info);
1598         } else if (name) {
1599             error_setg(errp, "net client(%s) doesn't support"
1600                        " rx-filter querying", name);
1601             assert(!filter_list);
1602             return NULL;
1603         }
1604 
1605         if (name) {
1606             break;
1607         }
1608     }
1609 
1610     if (filter_list == NULL && name) {
1611         error_setg(errp, "invalid net client name: %s", name);
1612     }
1613 
1614     return filter_list;
1615 }
1616 
1617 void colo_notify_filters_event(int event, Error **errp)
1618 {
1619     NetClientState *nc;
1620     NetFilterState *nf;
1621     NetFilterClass *nfc = NULL;
1622     Error *local_err = NULL;
1623 
1624     QTAILQ_FOREACH(nc, &net_clients, next) {
1625         QTAILQ_FOREACH(nf, &nc->filters, next) {
1626             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1627             nfc->handle_event(nf, event, &local_err);
1628             if (local_err) {
1629                 error_propagate(errp, local_err);
1630                 return;
1631             }
1632         }
1633     }
1634 }
1635 
1636 void net_client_set_link(NetClientState **ncs, int queues, bool up)
1637 {
1638     NetClientState *nc;
1639     int i;
1640 
1641     nc = ncs[0];
1642 
1643     for (i = 0; i < queues; i++) {
1644         ncs[i]->link_down = !up;
1645     }
1646 
1647     if (nc->info->link_status_changed) {
1648         nc->info->link_status_changed(nc);
1649     }
1650 
1651     if (nc->peer) {
1652         /* Change peer link only if the peer is NIC and then notify peer.
1653          * If the peer is a HUBPORT or a backend, we do not change the
1654          * link status.
1655          *
1656          * This behavior is compatible with qemu hubs where there could be
1657          * multiple clients that can still communicate with each other in
1658          * disconnected mode. For now maintain this compatibility.
1659          */
1660         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1661             for (i = 0; i < queues; i++) {
1662                 ncs[i]->peer->link_down = !up;
1663             }
1664         }
1665         if (nc->peer->info->link_status_changed) {
1666             nc->peer->info->link_status_changed(nc->peer);
1667         }
1668     }
1669 }
1670 
1671 void qmp_set_link(const char *name, bool up, Error **errp)
1672 {
1673     NetClientState *ncs[MAX_QUEUE_NUM];
1674     int queues;
1675 
1676     queues = qemu_find_net_clients_except(name, ncs,
1677                                           NET_CLIENT_DRIVER__MAX,
1678                                           MAX_QUEUE_NUM);
1679 
1680     if (queues == 0) {
1681         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1682                   "Device '%s' not found", name);
1683         return;
1684     }
1685 
1686     net_client_set_link(ncs, queues, up);
1687 }
1688 
1689 static void net_vm_change_state_handler(void *opaque, bool running,
1690                                         RunState state)
1691 {
1692     NetClientState *nc;
1693     NetClientState *tmp;
1694 
1695     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1696         if (running) {
1697             /* Flush queued packets and wake up backends. */
1698             if (nc->peer && qemu_can_send_packet(nc)) {
1699                 qemu_flush_queued_packets(nc->peer);
1700             }
1701         } else {
1702             /* Complete all queued packets, to guarantee we don't modify
1703              * state later when VM is not running.
1704              */
1705             qemu_flush_or_purge_queued_packets(nc, true);
1706         }
1707     }
1708 }
1709 
1710 void net_cleanup(void)
1711 {
1712     NetClientState *nc, **p = &QTAILQ_FIRST(&net_clients);
1713 
1714     /*cleanup colo compare module for COLO*/
1715     colo_compare_cleanup();
1716 
1717     /*
1718      * Walk the net_clients list and remove the netdevs but *not* any
1719      * NET_CLIENT_DRIVER_NIC entries. The latter are owned by the device
1720      * model which created them, and in some cases (e.g. xen-net-device)
1721      * the device itself may do cleanup at exit and will be upset if we
1722      * just delete its NIC from underneath it.
1723      *
1724      * Since qemu_del_net_client() may delete multiple entries, using
1725      * QTAILQ_FOREACH_SAFE() is not safe here. The only safe pointer
1726      * to keep as a bookmark is a NET_CLIENT_DRIVER_NIC entry, so keep
1727      * 'p' pointing to either the head of the list, or the 'next' field
1728      * of the latest NET_CLIENT_DRIVER_NIC, and operate on *p as we walk
1729      * the list.
1730      *
1731      * However, the NIC may have peers that trust to be clean beyond this
1732      * point.  For example, if they have been removed with device_del.
1733      *
1734      * The 'nc' variable isn't part of the list traversal; it's purely
1735      * for convenience as too much '(*p)->' has a tendency to make the
1736      * readers' eyes bleed.
1737      */
1738     while (*p) {
1739         nc = *p;
1740         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1741             NICState *nic = qemu_get_nic(nc);
1742 
1743             if (nic->peer_deleted) {
1744                 int queues = MAX(nic->conf->peers.queues, 1);
1745 
1746                 for (int i = 0; i < queues; i++) {
1747                     nc = qemu_get_subqueue(nic, i);
1748                     qemu_cleanup_net_client(nc->peer, false);
1749                 }
1750             }
1751 
1752             /* Skip NET_CLIENT_DRIVER_NIC entries */
1753             p = &QTAILQ_NEXT(nc, next);
1754         } else {
1755             qemu_del_net_client(nc);
1756         }
1757     }
1758 
1759     qemu_del_vm_change_state_handler(net_change_state_entry);
1760 }
1761 
1762 void net_check_clients(void)
1763 {
1764     NetClientState *nc;
1765     int i;
1766 
1767     if (nic_model_help) {
1768         show_nic_models();
1769         exit(0);
1770     }
1771     net_hub_check_clients();
1772 
1773     QTAILQ_FOREACH(nc, &net_clients, next) {
1774         if (!nc->peer) {
1775             warn_report("%s %s has no peer",
1776                         nc->info->type == NET_CLIENT_DRIVER_NIC
1777                         ? "nic" : "netdev",
1778                         nc->name);
1779         }
1780     }
1781 
1782     /* Check that all NICs requested via -net nic actually got created.
1783      * NICs created via -device don't need to be checked here because
1784      * they are always instantiated.
1785      */
1786     for (i = 0; i < MAX_NICS; i++) {
1787         NICInfo *nd = &nd_table[i];
1788         if (nd->used && !nd->instantiated) {
1789             warn_report("requested NIC (%s, model %s) "
1790                         "was not created (not supported by this machine?)",
1791                         nd->name ? nd->name : "anonymous",
1792                         nd->model ? nd->model : "unspecified");
1793         }
1794     }
1795 }
1796 
1797 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1798 {
1799     const char *model = qemu_opt_get(opts, "model");
1800 
1801     if (is_nic_model_help_option(model)) {
1802         return 0;
1803     }
1804 
1805     return net_client_init(opts, false, errp);
1806 }
1807 
1808 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1809 {
1810     const char *type = qemu_opt_get(opts, "type");
1811 
1812     if (type && is_help_option(type)) {
1813         show_netdevs();
1814         exit(0);
1815     }
1816     return net_client_init(opts, true, errp);
1817 }
1818 
1819 /* For the convenience "--nic" parameter */
1820 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1821 {
1822     char *mac, *nd_id;
1823     int idx, ret;
1824     NICInfo *ni;
1825     const char *type;
1826 
1827     type = qemu_opt_get(opts, "type");
1828     if (type) {
1829         if (g_str_equal(type, "none")) {
1830             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1831         }
1832         if (is_help_option(type)) {
1833             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1834             int i;
1835             show_netdevs();
1836             printf("\n");
1837             printf("Available NIC models "
1838                    "(use -nic model=help for a filtered list):\n");
1839             for (i = 0 ; nic_models->pdata[i]; i++) {
1840                 printf("%s\n", (char *)nic_models->pdata[i]);
1841             }
1842             g_ptr_array_free(nic_models, true);
1843             exit(0);
1844         }
1845     }
1846 
1847     idx = nic_get_free_idx();
1848     if (idx == -1 || nb_nics >= MAX_NICS) {
1849         error_setg(errp, "no more on-board/default NIC slots available");
1850         return -1;
1851     }
1852 
1853     if (!type) {
1854         qemu_opt_set(opts, "type", "user", &error_abort);
1855     }
1856 
1857     ni = &nd_table[idx];
1858     memset(ni, 0, sizeof(*ni));
1859     ni->model = qemu_opt_get_del(opts, "model");
1860 
1861     if (is_nic_model_help_option(ni->model)) {
1862         return 0;
1863     }
1864 
1865     /* Create an ID if the user did not specify one */
1866     nd_id = g_strdup(qemu_opts_id(opts));
1867     if (!nd_id) {
1868         nd_id = id_generate(ID_NET);
1869         qemu_opts_set_id(opts, nd_id);
1870     }
1871 
1872     /* Handle MAC address */
1873     mac = qemu_opt_get_del(opts, "mac");
1874     if (mac) {
1875         ret = net_parse_macaddr(ni->macaddr.a, mac);
1876         g_free(mac);
1877         if (ret) {
1878             error_setg(errp, "invalid syntax for ethernet address");
1879             goto out;
1880         }
1881         if (is_multicast_ether_addr(ni->macaddr.a)) {
1882             error_setg(errp, "NIC cannot have multicast MAC address");
1883             ret = -1;
1884             goto out;
1885         }
1886     }
1887     qemu_macaddr_default_if_unset(&ni->macaddr);
1888 
1889     ret = net_client_init(opts, true, errp);
1890     if (ret == 0) {
1891         ni->netdev = qemu_find_netdev(nd_id);
1892         ni->used = true;
1893         nb_nics++;
1894     }
1895 
1896 out:
1897     g_free(nd_id);
1898     return ret;
1899 }
1900 
1901 static void netdev_init_modern(void)
1902 {
1903     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1904         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1905 
1906         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1907         loc_push_restore(&nd->loc);
1908         net_client_init1(nd->nd, true, &error_fatal);
1909         loc_pop(&nd->loc);
1910         qapi_free_Netdev(nd->nd);
1911         g_free(nd);
1912     }
1913 }
1914 
1915 void net_init_clients(void)
1916 {
1917     net_change_state_entry =
1918         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1919 
1920     QTAILQ_INIT(&net_clients);
1921 
1922     netdev_init_modern();
1923 
1924     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1925                       &error_fatal);
1926 
1927     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1928                       &error_fatal);
1929 
1930     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1931                       &error_fatal);
1932 }
1933 
1934 /*
1935  * Does this -netdev argument use modern rather than traditional syntax?
1936  * Modern syntax is to be parsed with netdev_parse_modern().
1937  * Traditional syntax is to be parsed with net_client_parse().
1938  */
1939 bool netdev_is_modern(const char *optstr)
1940 {
1941     QemuOpts *opts;
1942     bool is_modern;
1943     const char *type;
1944     static QemuOptsList dummy_opts = {
1945         .name = "netdev",
1946         .implied_opt_name = "type",
1947         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1948         .desc = { { } },
1949     };
1950 
1951     if (optstr[0] == '{') {
1952         /* This is JSON, which means it's modern syntax */
1953         return true;
1954     }
1955 
1956     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1957     qemu_opts_do_parse(opts, optstr, dummy_opts.implied_opt_name,
1958                        &error_abort);
1959     type = qemu_opt_get(opts, "type");
1960     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1961 
1962     qemu_opts_reset(&dummy_opts);
1963 
1964     return is_modern;
1965 }
1966 
1967 /*
1968  * netdev_parse_modern() uses modern, more expressive syntax than
1969  * net_client_parse(), but supports only the -netdev option.
1970  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1971  * appends to @qemu_netdev_opts.
1972  */
1973 void netdev_parse_modern(const char *optstr)
1974 {
1975     Visitor *v;
1976     NetdevQueueEntry *nd;
1977 
1978     v = qobject_input_visitor_new_str(optstr, "type", &error_fatal);
1979     nd = g_new(NetdevQueueEntry, 1);
1980     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1981     visit_free(v);
1982     loc_save(&nd->loc);
1983 
1984     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1985 }
1986 
1987 void net_client_parse(QemuOptsList *opts_list, const char *optstr)
1988 {
1989     if (!qemu_opts_parse_noisily(opts_list, optstr, true)) {
1990         exit(1);
1991     }
1992 }
1993 
1994 /* From FreeBSD */
1995 /* XXX: optimize */
1996 uint32_t net_crc32(const uint8_t *p, int len)
1997 {
1998     uint32_t crc;
1999     int carry, i, j;
2000     uint8_t b;
2001 
2002     crc = 0xffffffff;
2003     for (i = 0; i < len; i++) {
2004         b = *p++;
2005         for (j = 0; j < 8; j++) {
2006             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
2007             crc <<= 1;
2008             b >>= 1;
2009             if (carry) {
2010                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
2011             }
2012         }
2013     }
2014 
2015     return crc;
2016 }
2017 
2018 uint32_t net_crc32_le(const uint8_t *p, int len)
2019 {
2020     uint32_t crc;
2021     int carry, i, j;
2022     uint8_t b;
2023 
2024     crc = 0xffffffff;
2025     for (i = 0; i < len; i++) {
2026         b = *p++;
2027         for (j = 0; j < 8; j++) {
2028             carry = (crc & 0x1) ^ (b & 0x01);
2029             crc >>= 1;
2030             b >>= 1;
2031             if (carry) {
2032                 crc ^= POLYNOMIAL_LE;
2033             }
2034         }
2035     }
2036 
2037     return crc;
2038 }
2039 
2040 QemuOptsList qemu_netdev_opts = {
2041     .name = "netdev",
2042     .implied_opt_name = "type",
2043     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
2044     .desc = {
2045         /*
2046          * no elements => accept any params
2047          * validation will happen later
2048          */
2049         { /* end of list */ }
2050     },
2051 };
2052 
2053 QemuOptsList qemu_nic_opts = {
2054     .name = "nic",
2055     .implied_opt_name = "type",
2056     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
2057     .desc = {
2058         /*
2059          * no elements => accept any params
2060          * validation will happen later
2061          */
2062         { /* end of list */ }
2063     },
2064 };
2065 
2066 QemuOptsList qemu_net_opts = {
2067     .name = "net",
2068     .implied_opt_name = "type",
2069     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
2070     .desc = {
2071         /*
2072          * no elements => accept any params
2073          * validation will happen later
2074          */
2075         { /* end of list */ }
2076     },
2077 };
2078 
2079 void net_socket_rs_init(SocketReadState *rs,
2080                         SocketReadStateFinalize *finalize,
2081                         bool vnet_hdr)
2082 {
2083     rs->state = 0;
2084     rs->vnet_hdr = vnet_hdr;
2085     rs->index = 0;
2086     rs->packet_len = 0;
2087     rs->vnet_hdr_len = 0;
2088     memset(rs->buf, 0, sizeof(rs->buf));
2089     rs->finalize = finalize;
2090 }
2091 
2092 /*
2093  * Returns
2094  * 0: success
2095  * -1: error occurs
2096  */
2097 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
2098 {
2099     unsigned int l;
2100 
2101     while (size > 0) {
2102         /* Reassemble a packet from the network.
2103          * 0 = getting length.
2104          * 1 = getting vnet header length.
2105          * 2 = getting data.
2106          */
2107         switch (rs->state) {
2108         case 0:
2109             l = 4 - rs->index;
2110             if (l > size) {
2111                 l = size;
2112             }
2113             memcpy(rs->buf + rs->index, buf, l);
2114             buf += l;
2115             size -= l;
2116             rs->index += l;
2117             if (rs->index == 4) {
2118                 /* got length */
2119                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
2120                 rs->index = 0;
2121                 if (rs->vnet_hdr) {
2122                     rs->state = 1;
2123                 } else {
2124                     rs->state = 2;
2125                     rs->vnet_hdr_len = 0;
2126                 }
2127             }
2128             break;
2129         case 1:
2130             l = 4 - rs->index;
2131             if (l > size) {
2132                 l = size;
2133             }
2134             memcpy(rs->buf + rs->index, buf, l);
2135             buf += l;
2136             size -= l;
2137             rs->index += l;
2138             if (rs->index == 4) {
2139                 /* got vnet header length */
2140                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
2141                 rs->index = 0;
2142                 rs->state = 2;
2143             }
2144             break;
2145         case 2:
2146             l = rs->packet_len - rs->index;
2147             if (l > size) {
2148                 l = size;
2149             }
2150             if (rs->index + l <= sizeof(rs->buf)) {
2151                 memcpy(rs->buf + rs->index, buf, l);
2152             } else {
2153                 fprintf(stderr, "serious error: oversized packet received,"
2154                     "connection terminated.\n");
2155                 rs->index = rs->state = 0;
2156                 return -1;
2157             }
2158 
2159             rs->index += l;
2160             buf += l;
2161             size -= l;
2162             if (rs->index >= rs->packet_len) {
2163                 rs->index = 0;
2164                 rs->state = 0;
2165                 assert(rs->finalize);
2166                 rs->finalize(rs);
2167             }
2168             break;
2169         }
2170     }
2171 
2172     assert(size == 0);
2173     return 0;
2174 }
2175