xref: /openbmc/qemu/net/net.c (revision e69b2c67)
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 #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 {
386     QTAILQ_REMOVE(&net_clients, nc, next);
387 
388     if (nc->info->cleanup) {
389         nc->info->cleanup(nc);
390     }
391 }
392 
393 static void qemu_free_net_client(NetClientState *nc)
394 {
395     if (nc->incoming_queue) {
396         qemu_del_net_queue(nc->incoming_queue);
397     }
398     if (nc->peer) {
399         nc->peer->peer = NULL;
400     }
401     g_free(nc->name);
402     g_free(nc->model);
403     if (nc->destructor) {
404         nc->destructor(nc);
405     }
406 }
407 
408 void qemu_del_net_client(NetClientState *nc)
409 {
410     NetClientState *ncs[MAX_QUEUE_NUM];
411     int queues, i;
412     NetFilterState *nf, *next;
413 
414     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
415 
416     /* If the NetClientState belongs to a multiqueue backend, we will change all
417      * other NetClientStates also.
418      */
419     queues = qemu_find_net_clients_except(nc->name, ncs,
420                                           NET_CLIENT_DRIVER_NIC,
421                                           MAX_QUEUE_NUM);
422     assert(queues != 0);
423 
424     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
425         object_unparent(OBJECT(nf));
426     }
427 
428     /* If there is a peer NIC, delete and cleanup client, but do not free. */
429     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
430         NICState *nic = qemu_get_nic(nc->peer);
431         if (nic->peer_deleted) {
432             return;
433         }
434         nic->peer_deleted = true;
435 
436         for (i = 0; i < queues; i++) {
437             ncs[i]->peer->link_down = true;
438         }
439 
440         if (nc->peer->info->link_status_changed) {
441             nc->peer->info->link_status_changed(nc->peer);
442         }
443 
444         for (i = 0; i < queues; i++) {
445             qemu_cleanup_net_client(ncs[i]);
446         }
447 
448         return;
449     }
450 
451     for (i = 0; i < queues; i++) {
452         qemu_cleanup_net_client(ncs[i]);
453         qemu_free_net_client(ncs[i]);
454     }
455 }
456 
457 void qemu_del_nic(NICState *nic)
458 {
459     int i, queues = MAX(nic->conf->peers.queues, 1);
460 
461     qemu_macaddr_set_free(&nic->conf->macaddr);
462 
463     for (i = 0; i < queues; i++) {
464         NetClientState *nc = qemu_get_subqueue(nic, i);
465         /* If this is a peer NIC and peer has already been deleted, free it now. */
466         if (nic->peer_deleted) {
467             qemu_free_net_client(nc->peer);
468         } else if (nc->peer) {
469             /* if there are RX packets pending, complete them */
470             qemu_purge_queued_packets(nc->peer);
471         }
472     }
473 
474     for (i = queues - 1; i >= 0; i--) {
475         NetClientState *nc = qemu_get_subqueue(nic, i);
476 
477         qemu_cleanup_net_client(nc);
478         qemu_free_net_client(nc);
479     }
480 
481     g_free(nic);
482 }
483 
484 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
485 {
486     NetClientState *nc;
487 
488     QTAILQ_FOREACH(nc, &net_clients, next) {
489         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
490             if (nc->queue_index == 0) {
491                 func(qemu_get_nic(nc), opaque);
492             }
493         }
494     }
495 }
496 
497 bool qemu_has_ufo(NetClientState *nc)
498 {
499     if (!nc || !nc->info->has_ufo) {
500         return false;
501     }
502 
503     return nc->info->has_ufo(nc);
504 }
505 
506 bool qemu_has_uso(NetClientState *nc)
507 {
508     if (!nc || !nc->info->has_uso) {
509         return false;
510     }
511 
512     return nc->info->has_uso(nc);
513 }
514 
515 bool qemu_has_vnet_hdr(NetClientState *nc)
516 {
517     if (!nc || !nc->info->has_vnet_hdr) {
518         return false;
519     }
520 
521     return nc->info->has_vnet_hdr(nc);
522 }
523 
524 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
525 {
526     if (!nc || !nc->info->has_vnet_hdr_len) {
527         return false;
528     }
529 
530     return nc->info->has_vnet_hdr_len(nc, len);
531 }
532 
533 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
534                           int ecn, int ufo, int uso4, int uso6)
535 {
536     if (!nc || !nc->info->set_offload) {
537         return;
538     }
539 
540     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo, uso4, uso6);
541 }
542 
543 int qemu_get_vnet_hdr_len(NetClientState *nc)
544 {
545     if (!nc) {
546         return 0;
547     }
548 
549     return nc->vnet_hdr_len;
550 }
551 
552 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
553 {
554     if (!nc || !nc->info->set_vnet_hdr_len) {
555         return;
556     }
557 
558     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
559            len == sizeof(struct virtio_net_hdr) ||
560            len == sizeof(struct virtio_net_hdr_v1_hash));
561 
562     nc->vnet_hdr_len = len;
563     nc->info->set_vnet_hdr_len(nc, len);
564 }
565 
566 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
567 {
568 #if HOST_BIG_ENDIAN
569     if (!nc || !nc->info->set_vnet_le) {
570         return -ENOSYS;
571     }
572 
573     return nc->info->set_vnet_le(nc, is_le);
574 #else
575     return 0;
576 #endif
577 }
578 
579 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
580 {
581 #if HOST_BIG_ENDIAN
582     return 0;
583 #else
584     if (!nc || !nc->info->set_vnet_be) {
585         return -ENOSYS;
586     }
587 
588     return nc->info->set_vnet_be(nc, is_be);
589 #endif
590 }
591 
592 int qemu_can_receive_packet(NetClientState *nc)
593 {
594     if (nc->receive_disabled) {
595         return 0;
596     } else if (nc->info->can_receive &&
597                !nc->info->can_receive(nc)) {
598         return 0;
599     }
600     return 1;
601 }
602 
603 int qemu_can_send_packet(NetClientState *sender)
604 {
605     int vm_running = runstate_is_running();
606 
607     if (!vm_running) {
608         return 0;
609     }
610 
611     if (!sender->peer) {
612         return 1;
613     }
614 
615     return qemu_can_receive_packet(sender->peer);
616 }
617 
618 static ssize_t filter_receive_iov(NetClientState *nc,
619                                   NetFilterDirection direction,
620                                   NetClientState *sender,
621                                   unsigned flags,
622                                   const struct iovec *iov,
623                                   int iovcnt,
624                                   NetPacketSent *sent_cb)
625 {
626     ssize_t ret = 0;
627     NetFilterState *nf = NULL;
628 
629     if (direction == NET_FILTER_DIRECTION_TX) {
630         QTAILQ_FOREACH(nf, &nc->filters, next) {
631             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
632                                          iovcnt, sent_cb);
633             if (ret) {
634                 return ret;
635             }
636         }
637     } else {
638         QTAILQ_FOREACH_REVERSE(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     }
646 
647     return ret;
648 }
649 
650 static ssize_t filter_receive(NetClientState *nc,
651                               NetFilterDirection direction,
652                               NetClientState *sender,
653                               unsigned flags,
654                               const uint8_t *data,
655                               size_t size,
656                               NetPacketSent *sent_cb)
657 {
658     struct iovec iov = {
659         .iov_base = (void *)data,
660         .iov_len = size
661     };
662 
663     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
664 }
665 
666 void qemu_purge_queued_packets(NetClientState *nc)
667 {
668     if (!nc->peer) {
669         return;
670     }
671 
672     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
673 }
674 
675 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
676 {
677     nc->receive_disabled = 0;
678 
679     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
680         if (net_hub_flush(nc->peer)) {
681             qemu_notify_event();
682         }
683     }
684     if (qemu_net_queue_flush(nc->incoming_queue)) {
685         /* We emptied the queue successfully, signal to the IO thread to repoll
686          * the file descriptor (for tap, for example).
687          */
688         qemu_notify_event();
689     } else if (purge) {
690         /* Unable to empty the queue, purge remaining packets */
691         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
692     }
693 }
694 
695 void qemu_flush_queued_packets(NetClientState *nc)
696 {
697     qemu_flush_or_purge_queued_packets(nc, false);
698 }
699 
700 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
701                                                  unsigned flags,
702                                                  const uint8_t *buf, int size,
703                                                  NetPacketSent *sent_cb)
704 {
705     NetQueue *queue;
706     int ret;
707 
708 #ifdef DEBUG_NET
709     printf("qemu_send_packet_async:\n");
710     qemu_hexdump(stdout, "net", buf, size);
711 #endif
712 
713     if (sender->link_down || !sender->peer) {
714         return size;
715     }
716 
717     /* Let filters handle the packet first */
718     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
719                          sender, flags, buf, size, sent_cb);
720     if (ret) {
721         return ret;
722     }
723 
724     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
725                          sender, flags, buf, size, sent_cb);
726     if (ret) {
727         return ret;
728     }
729 
730     queue = sender->peer->incoming_queue;
731 
732     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
733 }
734 
735 ssize_t qemu_send_packet_async(NetClientState *sender,
736                                const uint8_t *buf, int size,
737                                NetPacketSent *sent_cb)
738 {
739     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
740                                              buf, size, sent_cb);
741 }
742 
743 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
744 {
745     return qemu_send_packet_async(nc, buf, size, NULL);
746 }
747 
748 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
749 {
750     if (!qemu_can_receive_packet(nc)) {
751         return 0;
752     }
753 
754     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
755 }
756 
757 ssize_t qemu_receive_packet_iov(NetClientState *nc, const struct iovec *iov,
758                                 int iovcnt)
759 {
760     if (!qemu_can_receive_packet(nc)) {
761         return 0;
762     }
763 
764     return qemu_net_queue_receive_iov(nc->incoming_queue, iov, iovcnt);
765 }
766 
767 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
768 {
769     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
770                                              buf, size, NULL);
771 }
772 
773 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
774                                int iovcnt, unsigned flags)
775 {
776     uint8_t *buf = NULL;
777     uint8_t *buffer;
778     size_t offset;
779     ssize_t ret;
780 
781     if (iovcnt == 1) {
782         buffer = iov[0].iov_base;
783         offset = iov[0].iov_len;
784     } else {
785         offset = iov_size(iov, iovcnt);
786         if (offset > NET_BUFSIZE) {
787             return -1;
788         }
789         buf = g_malloc(offset);
790         buffer = buf;
791         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
792     }
793 
794     ret = nc->info->receive(nc, buffer, offset);
795 
796     g_free(buf);
797     return ret;
798 }
799 
800 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
801                                        unsigned flags,
802                                        const struct iovec *iov,
803                                        int iovcnt,
804                                        void *opaque)
805 {
806     MemReentrancyGuard *owned_reentrancy_guard;
807     NetClientState *nc = opaque;
808     int ret;
809     struct virtio_net_hdr_v1_hash vnet_hdr = { };
810     g_autofree struct iovec *iov_copy = NULL;
811 
812 
813     if (nc->link_down) {
814         return iov_size(iov, iovcnt);
815     }
816 
817     if (nc->receive_disabled) {
818         return 0;
819     }
820 
821     if (nc->info->type != NET_CLIENT_DRIVER_NIC ||
822         qemu_get_nic(nc)->reentrancy_guard->engaged_in_io) {
823         owned_reentrancy_guard = NULL;
824     } else {
825         owned_reentrancy_guard = qemu_get_nic(nc)->reentrancy_guard;
826         owned_reentrancy_guard->engaged_in_io = true;
827     }
828 
829     if ((flags & QEMU_NET_PACKET_FLAG_RAW) && nc->vnet_hdr_len) {
830         iov_copy = g_new(struct iovec, iovcnt + 1);
831         iov_copy[0].iov_base = &vnet_hdr;
832         iov_copy[0].iov_len =  nc->vnet_hdr_len;
833         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
834         iov = iov_copy;
835     }
836 
837     if (nc->info->receive_iov) {
838         ret = nc->info->receive_iov(nc, iov, iovcnt);
839     } else {
840         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
841     }
842 
843     if (owned_reentrancy_guard) {
844         owned_reentrancy_guard->engaged_in_io = false;
845     }
846 
847     if (ret == 0) {
848         nc->receive_disabled = 1;
849     }
850 
851     return ret;
852 }
853 
854 ssize_t qemu_sendv_packet_async(NetClientState *sender,
855                                 const struct iovec *iov, int iovcnt,
856                                 NetPacketSent *sent_cb)
857 {
858     NetQueue *queue;
859     size_t size = iov_size(iov, iovcnt);
860     int ret;
861 
862     if (size > NET_BUFSIZE) {
863         return size;
864     }
865 
866     if (sender->link_down || !sender->peer) {
867         return size;
868     }
869 
870     /* Let filters handle the packet first */
871     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
872                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
873     if (ret) {
874         return ret;
875     }
876 
877     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
878                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
879     if (ret) {
880         return ret;
881     }
882 
883     queue = sender->peer->incoming_queue;
884 
885     return qemu_net_queue_send_iov(queue, sender,
886                                    QEMU_NET_PACKET_FLAG_NONE,
887                                    iov, iovcnt, sent_cb);
888 }
889 
890 ssize_t
891 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
892 {
893     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
894 }
895 
896 NetClientState *qemu_find_netdev(const char *id)
897 {
898     NetClientState *nc;
899 
900     QTAILQ_FOREACH(nc, &net_clients, next) {
901         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
902             continue;
903         if (!strcmp(nc->name, id)) {
904             return nc;
905         }
906     }
907 
908     return NULL;
909 }
910 
911 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
912                                  NetClientDriver type, int max)
913 {
914     NetClientState *nc;
915     int ret = 0;
916 
917     QTAILQ_FOREACH(nc, &net_clients, next) {
918         if (nc->info->type == type) {
919             continue;
920         }
921         if (!id || !strcmp(nc->name, id)) {
922             if (ret < max) {
923                 ncs[ret] = nc;
924             }
925             ret++;
926         }
927     }
928 
929     return ret;
930 }
931 
932 static int nic_get_free_idx(void)
933 {
934     int index;
935 
936     for (index = 0; index < MAX_NICS; index++)
937         if (!nd_table[index].used)
938             return index;
939     return -1;
940 }
941 
942 GPtrArray *qemu_get_nic_models(const char *device_type)
943 {
944     GPtrArray *nic_models = g_ptr_array_new();
945     GSList *list = object_class_get_list_sorted(device_type, false);
946 
947     while (list) {
948         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
949                                              TYPE_DEVICE);
950         GSList *next;
951         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
952             dc->user_creatable) {
953             const char *name = object_class_get_name(list->data);
954             /*
955              * A network device might also be something else than a NIC, see
956              * e.g. the "rocker" device. Thus we have to look for the "netdev"
957              * property, too. Unfortunately, some devices like virtio-net only
958              * create this property during instance_init, so we have to create
959              * a temporary instance here to be able to check it.
960              */
961             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
962             if (object_property_find(obj, "netdev")) {
963                 g_ptr_array_add(nic_models, (gpointer)name);
964             }
965             object_unref(obj);
966         }
967         next = list->next;
968         g_slist_free_1(list);
969         list = next;
970     }
971     g_ptr_array_add(nic_models, NULL);
972 
973     return nic_models;
974 }
975 
976 static int net_init_nic(const Netdev *netdev, const char *name,
977                         NetClientState *peer, Error **errp)
978 {
979     int idx;
980     NICInfo *nd;
981     const NetLegacyNicOptions *nic;
982 
983     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
984     nic = &netdev->u.nic;
985 
986     idx = nic_get_free_idx();
987     if (idx == -1 || nb_nics >= MAX_NICS) {
988         error_setg(errp, "too many NICs");
989         return -1;
990     }
991 
992     nd = &nd_table[idx];
993 
994     memset(nd, 0, sizeof(*nd));
995 
996     if (nic->netdev) {
997         nd->netdev = qemu_find_netdev(nic->netdev);
998         if (!nd->netdev) {
999             error_setg(errp, "netdev '%s' not found", nic->netdev);
1000             return -1;
1001         }
1002     } else {
1003         assert(peer);
1004         nd->netdev = peer;
1005     }
1006     nd->name = g_strdup(name);
1007     if (nic->model) {
1008         nd->model = g_strdup(nic->model);
1009     }
1010     if (nic->addr) {
1011         nd->devaddr = g_strdup(nic->addr);
1012     }
1013 
1014     if (nic->macaddr &&
1015         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1016         error_setg(errp, "invalid syntax for ethernet address");
1017         return -1;
1018     }
1019     if (nic->macaddr &&
1020         is_multicast_ether_addr(nd->macaddr.a)) {
1021         error_setg(errp,
1022                    "NIC cannot have multicast MAC address (odd 1st byte)");
1023         return -1;
1024     }
1025     qemu_macaddr_default_if_unset(&nd->macaddr);
1026 
1027     if (nic->has_vectors) {
1028         if (nic->vectors > 0x7ffffff) {
1029             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1030             return -1;
1031         }
1032         nd->nvectors = nic->vectors;
1033     } else {
1034         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1035     }
1036 
1037     nd->used = 1;
1038     nb_nics++;
1039 
1040     return idx;
1041 }
1042 
1043 static gboolean add_nic_result(gpointer key, gpointer value, gpointer user_data)
1044 {
1045     GPtrArray *results = user_data;
1046     GPtrArray *alias_list = value;
1047     const char *model = key;
1048     char *result;
1049 
1050     if (!alias_list) {
1051         result = g_strdup(model);
1052     } else {
1053         GString *result_str = g_string_new(model);
1054         int i;
1055 
1056         g_string_append(result_str, " (aka ");
1057         for (i = 0; i < alias_list->len; i++) {
1058             if (i) {
1059                 g_string_append(result_str, ", ");
1060             }
1061             g_string_append(result_str, alias_list->pdata[i]);
1062         }
1063         g_string_append(result_str, ")");
1064         result = result_str->str;
1065         g_string_free(result_str, false);
1066         g_ptr_array_unref(alias_list);
1067     }
1068     g_ptr_array_add(results, result);
1069     return true;
1070 }
1071 
1072 static int model_cmp(char **a, char **b)
1073 {
1074     return strcmp(*a, *b);
1075 }
1076 
1077 static void show_nic_models(void)
1078 {
1079     GPtrArray *results = g_ptr_array_new();
1080     int i;
1081 
1082     g_hash_table_foreach_remove(nic_model_help, add_nic_result, results);
1083     g_ptr_array_sort(results, (GCompareFunc)model_cmp);
1084 
1085     printf("Available NIC models for this configuration:\n");
1086     for (i = 0 ; i < results->len; i++) {
1087         printf("%s\n", (char *)results->pdata[i]);
1088     }
1089     g_hash_table_unref(nic_model_help);
1090     nic_model_help = NULL;
1091 }
1092 
1093 static void add_nic_model_help(const char *model, const char *alias)
1094 {
1095     GPtrArray *alias_list = NULL;
1096 
1097     if (g_hash_table_lookup_extended(nic_model_help, model, NULL,
1098                                      (gpointer *)&alias_list)) {
1099         /* Already exists, no alias to add: return */
1100         if (!alias) {
1101             return;
1102         }
1103         if (alias_list) {
1104             /* Check if this alias is already in the list. Add if not. */
1105             if (!g_ptr_array_find_with_equal_func(alias_list, alias,
1106                                                   g_str_equal, NULL)) {
1107                 g_ptr_array_add(alias_list, g_strdup(alias));
1108             }
1109             return;
1110         }
1111     }
1112     /* Either this model wasn't in the list already, or a first alias added */
1113     if (alias) {
1114         alias_list = g_ptr_array_new();
1115         g_ptr_array_set_free_func(alias_list, g_free);
1116         g_ptr_array_add(alias_list, g_strdup(alias));
1117     }
1118     g_hash_table_replace(nic_model_help, g_strdup(model), alias_list);
1119 }
1120 
1121 NICInfo *qemu_find_nic_info(const char *typename, bool match_default,
1122                             const char *alias)
1123 {
1124     NICInfo *nd;
1125     int i;
1126 
1127     if (nic_model_help) {
1128         add_nic_model_help(typename, alias);
1129     }
1130 
1131     for (i = 0; i < nb_nics; i++) {
1132         nd = &nd_table[i];
1133 
1134         if (!nd->used || nd->instantiated) {
1135             continue;
1136         }
1137 
1138         if ((match_default && !nd->model) || !g_strcmp0(nd->model, typename)
1139             || (alias && !g_strcmp0(nd->model, alias))) {
1140             return nd;
1141         }
1142     }
1143     return NULL;
1144 }
1145 
1146 static bool is_nic_model_help_option(const char *model)
1147 {
1148     if (model && is_help_option(model)) {
1149         /*
1150          * Trigger the help output by instantiating the hash table which
1151          * will gather tha available models as they get registered.
1152          */
1153         if (!nic_model_help) {
1154             nic_model_help = g_hash_table_new_full(g_str_hash, g_str_equal,
1155                                                    g_free, NULL);
1156         }
1157         return true;
1158     }
1159     return false;
1160 }
1161 
1162 /* "I have created a device. Please configure it if you can" */
1163 bool qemu_configure_nic_device(DeviceState *dev, bool match_default,
1164                                const char *alias)
1165 {
1166     NICInfo *nd = qemu_find_nic_info(object_get_typename(OBJECT(dev)),
1167                                      match_default, alias);
1168 
1169     if (nd) {
1170         qdev_set_nic_properties(dev, nd);
1171         return true;
1172     }
1173     return false;
1174 }
1175 
1176 /* "Please create a device, if you have a configuration for it" */
1177 DeviceState *qemu_create_nic_device(const char *typename, bool match_default,
1178                                     const char *alias)
1179 {
1180     NICInfo *nd = qemu_find_nic_info(typename, match_default, alias);
1181     DeviceState *dev;
1182 
1183     if (!nd) {
1184         return NULL;
1185     }
1186 
1187     dev = qdev_new(typename);
1188     qdev_set_nic_properties(dev, nd);
1189     return dev;
1190 }
1191 
1192 void qemu_create_nic_bus_devices(BusState *bus, const char *parent_type,
1193                                  const char *default_model,
1194                                  const char *alias, const char *alias_target)
1195 {
1196     GPtrArray *nic_models = qemu_get_nic_models(parent_type);
1197     const char *model;
1198     DeviceState *dev;
1199     NICInfo *nd;
1200     int i;
1201 
1202     if (nic_model_help) {
1203         if (alias_target) {
1204             add_nic_model_help(alias_target, alias);
1205         }
1206         for (i = 0; i < nic_models->len - 1; i++) {
1207             add_nic_model_help(nic_models->pdata[i], NULL);
1208         }
1209     }
1210 
1211     /* Drop the NULL terminator which would make g_str_equal() unhappy */
1212     nic_models->len--;
1213 
1214     for (i = 0; i < nb_nics; i++) {
1215         nd = &nd_table[i];
1216 
1217         if (!nd->used || nd->instantiated) {
1218             continue;
1219         }
1220 
1221         model = nd->model ? nd->model : default_model;
1222         if (!model) {
1223             continue;
1224         }
1225 
1226         /* Each bus type is allowed *one* substitution */
1227         if (g_str_equal(model, alias)) {
1228             model = alias_target;
1229         }
1230 
1231         if (!g_ptr_array_find_with_equal_func(nic_models, model,
1232                                               g_str_equal, NULL)) {
1233             /* This NIC does not live on this bus. */
1234             continue;
1235         }
1236 
1237         dev = qdev_new(model);
1238         qdev_set_nic_properties(dev, nd);
1239         qdev_realize_and_unref(dev, bus, &error_fatal);
1240     }
1241 
1242     g_ptr_array_free(nic_models, true);
1243 }
1244 
1245 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1246     const Netdev *netdev,
1247     const char *name,
1248     NetClientState *peer, Error **errp) = {
1249         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1250 #ifdef CONFIG_SLIRP
1251         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1252 #endif
1253         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1254         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1255         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1256         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1257 #ifdef CONFIG_VDE
1258         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1259 #endif
1260 #ifdef CONFIG_NETMAP
1261         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1262 #endif
1263 #ifdef CONFIG_AF_XDP
1264         [NET_CLIENT_DRIVER_AF_XDP]    = net_init_af_xdp,
1265 #endif
1266 #ifdef CONFIG_NET_BRIDGE
1267         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1268 #endif
1269         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1270 #ifdef CONFIG_VHOST_NET_USER
1271         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1272 #endif
1273 #ifdef CONFIG_VHOST_NET_VDPA
1274         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1275 #endif
1276 #ifdef CONFIG_L2TPV3
1277         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1278 #endif
1279 #ifdef CONFIG_VMNET
1280         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1281         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1282         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1283 #endif /* CONFIG_VMNET */
1284 };
1285 
1286 
1287 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1288 {
1289     NetClientState *peer = NULL;
1290     NetClientState *nc;
1291 
1292     if (is_netdev) {
1293         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1294             !net_client_init_fun[netdev->type]) {
1295             error_setg(errp, "network backend '%s' is not compiled into this binary",
1296                        NetClientDriver_str(netdev->type));
1297             return -1;
1298         }
1299     } else {
1300         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1301             return 0; /* nothing to do */
1302         }
1303         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1304             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1305                        NetClientDriver_str(netdev->type));
1306             return -1;
1307         }
1308 
1309         if (!net_client_init_fun[netdev->type]) {
1310             error_setg(errp, "network backend '%s' is not compiled into this binary",
1311                        NetClientDriver_str(netdev->type));
1312             return -1;
1313         }
1314 
1315         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1316         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1317             !netdev->u.nic.netdev) {
1318             peer = net_hub_add_port(0, NULL, NULL);
1319         }
1320     }
1321 
1322     nc = qemu_find_netdev(netdev->id);
1323     if (nc) {
1324         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1325         return -1;
1326     }
1327 
1328     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1329         /* FIXME drop when all init functions store an Error */
1330         if (errp && !*errp) {
1331             error_setg(errp, "Device '%s' could not be initialized",
1332                        NetClientDriver_str(netdev->type));
1333         }
1334         return -1;
1335     }
1336 
1337     if (is_netdev) {
1338         nc = qemu_find_netdev(netdev->id);
1339         assert(nc);
1340         nc->is_netdev = true;
1341     }
1342 
1343     return 0;
1344 }
1345 
1346 void show_netdevs(void)
1347 {
1348     int idx;
1349     const char *available_netdevs[] = {
1350         "socket",
1351         "stream",
1352         "dgram",
1353         "hubport",
1354         "tap",
1355 #ifdef CONFIG_SLIRP
1356         "user",
1357 #endif
1358 #ifdef CONFIG_L2TPV3
1359         "l2tpv3",
1360 #endif
1361 #ifdef CONFIG_VDE
1362         "vde",
1363 #endif
1364 #ifdef CONFIG_NET_BRIDGE
1365         "bridge",
1366 #endif
1367 #ifdef CONFIG_NETMAP
1368         "netmap",
1369 #endif
1370 #ifdef CONFIG_AF_XDP
1371         "af-xdp",
1372 #endif
1373 #ifdef CONFIG_POSIX
1374         "vhost-user",
1375 #endif
1376 #ifdef CONFIG_VHOST_VDPA
1377         "vhost-vdpa",
1378 #endif
1379 #ifdef CONFIG_VMNET
1380         "vmnet-host",
1381         "vmnet-shared",
1382         "vmnet-bridged",
1383 #endif
1384     };
1385 
1386     qemu_printf("Available netdev backend types:\n");
1387     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1388         qemu_printf("%s\n", available_netdevs[idx]);
1389     }
1390 }
1391 
1392 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1393 {
1394     gchar **substrings = NULL;
1395     Netdev *object = NULL;
1396     int ret = -1;
1397     Visitor *v = opts_visitor_new(opts);
1398 
1399     /* Parse convenience option format ipv6-net=fec0::0[/64] */
1400     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1401 
1402     if (ip6_net) {
1403         char *prefix_addr;
1404         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1405 
1406         substrings = g_strsplit(ip6_net, "/", 2);
1407         if (!substrings || !substrings[0]) {
1408             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1409                        "a valid IPv6 prefix");
1410             goto out;
1411         }
1412 
1413         prefix_addr = substrings[0];
1414 
1415         /* Handle user-specified prefix length. */
1416         if (substrings[1] &&
1417             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1418         {
1419             error_setg(errp,
1420                        "parameter 'ipv6-net' expects a number after '/'");
1421             goto out;
1422         }
1423 
1424         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1425         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1426                             &error_abort);
1427         qemu_opt_unset(opts, "ipv6-net");
1428     }
1429 
1430     /* Create an ID for -net if the user did not specify one */
1431     if (!is_netdev && !qemu_opts_id(opts)) {
1432         qemu_opts_set_id(opts, id_generate(ID_NET));
1433     }
1434 
1435     if (visit_type_Netdev(v, NULL, &object, errp)) {
1436         ret = net_client_init1(object, is_netdev, errp);
1437     }
1438 
1439     qapi_free_Netdev(object);
1440 
1441 out:
1442     g_strfreev(substrings);
1443     visit_free(v);
1444     return ret;
1445 }
1446 
1447 void netdev_add(QemuOpts *opts, Error **errp)
1448 {
1449     net_client_init(opts, true, errp);
1450 }
1451 
1452 void qmp_netdev_add(Netdev *netdev, Error **errp)
1453 {
1454     if (!id_wellformed(netdev->id)) {
1455         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1456         return;
1457     }
1458 
1459     net_client_init1(netdev, true, errp);
1460 }
1461 
1462 void qmp_netdev_del(const char *id, Error **errp)
1463 {
1464     NetClientState *nc;
1465     QemuOpts *opts;
1466 
1467     nc = qemu_find_netdev(id);
1468     if (!nc) {
1469         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1470                   "Device '%s' not found", id);
1471         return;
1472     }
1473 
1474     if (!nc->is_netdev) {
1475         error_setg(errp, "Device '%s' is not a netdev", id);
1476         return;
1477     }
1478 
1479     qemu_del_net_client(nc);
1480 
1481     /*
1482      * Wart: we need to delete the QemuOpts associated with netdevs
1483      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1484      * HMP netdev_add.
1485      */
1486     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1487     if (opts) {
1488         qemu_opts_del(opts);
1489     }
1490 }
1491 
1492 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1493 {
1494     char *str;
1495     ObjectProperty *prop;
1496     ObjectPropertyIterator iter;
1497     Visitor *v;
1498 
1499     /* generate info str */
1500     object_property_iter_init(&iter, OBJECT(nf));
1501     while ((prop = object_property_iter_next(&iter))) {
1502         if (!strcmp(prop->name, "type")) {
1503             continue;
1504         }
1505         v = string_output_visitor_new(false, &str);
1506         object_property_get(OBJECT(nf), prop->name, v, NULL);
1507         visit_complete(v, &str);
1508         visit_free(v);
1509         monitor_printf(mon, ",%s=%s", prop->name, str);
1510         g_free(str);
1511     }
1512     monitor_printf(mon, "\n");
1513 }
1514 
1515 void print_net_client(Monitor *mon, NetClientState *nc)
1516 {
1517     NetFilterState *nf;
1518 
1519     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1520                    nc->queue_index,
1521                    NetClientDriver_str(nc->info->type),
1522                    nc->info_str);
1523     if (!QTAILQ_EMPTY(&nc->filters)) {
1524         monitor_printf(mon, "filters:\n");
1525     }
1526     QTAILQ_FOREACH(nf, &nc->filters, next) {
1527         monitor_printf(mon, "  - %s: type=%s",
1528                        object_get_canonical_path_component(OBJECT(nf)),
1529                        object_get_typename(OBJECT(nf)));
1530         netfilter_print_info(mon, nf);
1531     }
1532 }
1533 
1534 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1535 {
1536     NetClientState *nc;
1537     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1538 
1539     QTAILQ_FOREACH(nc, &net_clients, next) {
1540         RxFilterInfo *info;
1541 
1542         if (name && strcmp(nc->name, name) != 0) {
1543             continue;
1544         }
1545 
1546         /* only query rx-filter information of NIC */
1547         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1548             if (name) {
1549                 error_setg(errp, "net client(%s) isn't a NIC", name);
1550                 assert(!filter_list);
1551                 return NULL;
1552             }
1553             continue;
1554         }
1555 
1556         /* only query information on queue 0 since the info is per nic,
1557          * not per queue
1558          */
1559         if (nc->queue_index != 0)
1560             continue;
1561 
1562         if (nc->info->query_rx_filter) {
1563             info = nc->info->query_rx_filter(nc);
1564             QAPI_LIST_APPEND(tail, info);
1565         } else if (name) {
1566             error_setg(errp, "net client(%s) doesn't support"
1567                        " rx-filter querying", name);
1568             assert(!filter_list);
1569             return NULL;
1570         }
1571 
1572         if (name) {
1573             break;
1574         }
1575     }
1576 
1577     if (filter_list == NULL && name) {
1578         error_setg(errp, "invalid net client name: %s", name);
1579     }
1580 
1581     return filter_list;
1582 }
1583 
1584 void colo_notify_filters_event(int event, Error **errp)
1585 {
1586     NetClientState *nc;
1587     NetFilterState *nf;
1588     NetFilterClass *nfc = NULL;
1589     Error *local_err = NULL;
1590 
1591     QTAILQ_FOREACH(nc, &net_clients, next) {
1592         QTAILQ_FOREACH(nf, &nc->filters, next) {
1593             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1594             nfc->handle_event(nf, event, &local_err);
1595             if (local_err) {
1596                 error_propagate(errp, local_err);
1597                 return;
1598             }
1599         }
1600     }
1601 }
1602 
1603 void qmp_set_link(const char *name, bool up, Error **errp)
1604 {
1605     NetClientState *ncs[MAX_QUEUE_NUM];
1606     NetClientState *nc;
1607     int queues, i;
1608 
1609     queues = qemu_find_net_clients_except(name, ncs,
1610                                           NET_CLIENT_DRIVER__MAX,
1611                                           MAX_QUEUE_NUM);
1612 
1613     if (queues == 0) {
1614         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1615                   "Device '%s' not found", name);
1616         return;
1617     }
1618     nc = ncs[0];
1619 
1620     for (i = 0; i < queues; i++) {
1621         ncs[i]->link_down = !up;
1622     }
1623 
1624     if (nc->info->link_status_changed) {
1625         nc->info->link_status_changed(nc);
1626     }
1627 
1628     if (nc->peer) {
1629         /* Change peer link only if the peer is NIC and then notify peer.
1630          * If the peer is a HUBPORT or a backend, we do not change the
1631          * link status.
1632          *
1633          * This behavior is compatible with qemu hubs where there could be
1634          * multiple clients that can still communicate with each other in
1635          * disconnected mode. For now maintain this compatibility.
1636          */
1637         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1638             for (i = 0; i < queues; i++) {
1639                 ncs[i]->peer->link_down = !up;
1640             }
1641         }
1642         if (nc->peer->info->link_status_changed) {
1643             nc->peer->info->link_status_changed(nc->peer);
1644         }
1645     }
1646 }
1647 
1648 static void net_vm_change_state_handler(void *opaque, bool running,
1649                                         RunState state)
1650 {
1651     NetClientState *nc;
1652     NetClientState *tmp;
1653 
1654     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1655         if (running) {
1656             /* Flush queued packets and wake up backends. */
1657             if (nc->peer && qemu_can_send_packet(nc)) {
1658                 qemu_flush_queued_packets(nc->peer);
1659             }
1660         } else {
1661             /* Complete all queued packets, to guarantee we don't modify
1662              * state later when VM is not running.
1663              */
1664             qemu_flush_or_purge_queued_packets(nc, true);
1665         }
1666     }
1667 }
1668 
1669 void net_cleanup(void)
1670 {
1671     NetClientState *nc, **p = &QTAILQ_FIRST(&net_clients);
1672 
1673     /*cleanup colo compare module for COLO*/
1674     colo_compare_cleanup();
1675 
1676     /*
1677      * Walk the net_clients list and remove the netdevs but *not* any
1678      * NET_CLIENT_DRIVER_NIC entries. The latter are owned by the device
1679      * model which created them, and in some cases (e.g. xen-net-device)
1680      * the device itself may do cleanup at exit and will be upset if we
1681      * just delete its NIC from underneath it.
1682      *
1683      * Since qemu_del_net_client() may delete multiple entries, using
1684      * QTAILQ_FOREACH_SAFE() is not safe here. The only safe pointer
1685      * to keep as a bookmark is a NET_CLIENT_DRIVER_NIC entry, so keep
1686      * 'p' pointing to either the head of the list, or the 'next' field
1687      * of the latest NET_CLIENT_DRIVER_NIC, and operate on *p as we walk
1688      * the list.
1689      *
1690      * The 'nc' variable isn't part of the list traversal; it's purely
1691      * for convenience as too much '(*p)->' has a tendency to make the
1692      * readers' eyes bleed.
1693      */
1694     while (*p) {
1695         nc = *p;
1696         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1697             /* Skip NET_CLIENT_DRIVER_NIC entries */
1698             p = &QTAILQ_NEXT(nc, next);
1699         } else {
1700             qemu_del_net_client(nc);
1701         }
1702     }
1703 
1704     qemu_del_vm_change_state_handler(net_change_state_entry);
1705 }
1706 
1707 void net_check_clients(void)
1708 {
1709     NetClientState *nc;
1710     int i;
1711 
1712     if (nic_model_help) {
1713         show_nic_models();
1714         exit(0);
1715     }
1716     net_hub_check_clients();
1717 
1718     QTAILQ_FOREACH(nc, &net_clients, next) {
1719         if (!nc->peer) {
1720             warn_report("%s %s has no peer",
1721                         nc->info->type == NET_CLIENT_DRIVER_NIC
1722                         ? "nic" : "netdev",
1723                         nc->name);
1724         }
1725     }
1726 
1727     /* Check that all NICs requested via -net nic actually got created.
1728      * NICs created via -device don't need to be checked here because
1729      * they are always instantiated.
1730      */
1731     for (i = 0; i < MAX_NICS; i++) {
1732         NICInfo *nd = &nd_table[i];
1733         if (nd->used && !nd->instantiated) {
1734             warn_report("requested NIC (%s, model %s) "
1735                         "was not created (not supported by this machine?)",
1736                         nd->name ? nd->name : "anonymous",
1737                         nd->model ? nd->model : "unspecified");
1738         }
1739     }
1740 }
1741 
1742 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1743 {
1744     const char *model = qemu_opt_get(opts, "model");
1745 
1746     if (is_nic_model_help_option(model)) {
1747         return 0;
1748     }
1749 
1750     return net_client_init(opts, false, errp);
1751 }
1752 
1753 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1754 {
1755     const char *type = qemu_opt_get(opts, "type");
1756 
1757     if (type && is_help_option(type)) {
1758         show_netdevs();
1759         exit(0);
1760     }
1761     return net_client_init(opts, true, errp);
1762 }
1763 
1764 /* For the convenience "--nic" parameter */
1765 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1766 {
1767     char *mac, *nd_id;
1768     int idx, ret;
1769     NICInfo *ni;
1770     const char *type;
1771 
1772     type = qemu_opt_get(opts, "type");
1773     if (type) {
1774         if (g_str_equal(type, "none")) {
1775             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1776         }
1777         if (is_help_option(type)) {
1778             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1779             int i;
1780             show_netdevs();
1781             printf("\n");
1782             printf("Available NIC models "
1783                    "(use -nic model=help for a filtered list):\n");
1784             for (i = 0 ; nic_models->pdata[i]; i++) {
1785                 printf("%s\n", (char *)nic_models->pdata[i]);
1786             }
1787             g_ptr_array_free(nic_models, true);
1788             exit(0);
1789         }
1790     }
1791 
1792     idx = nic_get_free_idx();
1793     if (idx == -1 || nb_nics >= MAX_NICS) {
1794         error_setg(errp, "no more on-board/default NIC slots available");
1795         return -1;
1796     }
1797 
1798     if (!type) {
1799         qemu_opt_set(opts, "type", "user", &error_abort);
1800     }
1801 
1802     ni = &nd_table[idx];
1803     memset(ni, 0, sizeof(*ni));
1804     ni->model = qemu_opt_get_del(opts, "model");
1805 
1806     if (is_nic_model_help_option(ni->model)) {
1807         return 0;
1808     }
1809 
1810     /* Create an ID if the user did not specify one */
1811     nd_id = g_strdup(qemu_opts_id(opts));
1812     if (!nd_id) {
1813         nd_id = id_generate(ID_NET);
1814         qemu_opts_set_id(opts, nd_id);
1815     }
1816 
1817     /* Handle MAC address */
1818     mac = qemu_opt_get_del(opts, "mac");
1819     if (mac) {
1820         ret = net_parse_macaddr(ni->macaddr.a, mac);
1821         g_free(mac);
1822         if (ret) {
1823             error_setg(errp, "invalid syntax for ethernet address");
1824             goto out;
1825         }
1826         if (is_multicast_ether_addr(ni->macaddr.a)) {
1827             error_setg(errp, "NIC cannot have multicast MAC address");
1828             ret = -1;
1829             goto out;
1830         }
1831     }
1832     qemu_macaddr_default_if_unset(&ni->macaddr);
1833 
1834     ret = net_client_init(opts, true, errp);
1835     if (ret == 0) {
1836         ni->netdev = qemu_find_netdev(nd_id);
1837         ni->used = true;
1838         nb_nics++;
1839     }
1840 
1841 out:
1842     g_free(nd_id);
1843     return ret;
1844 }
1845 
1846 static void netdev_init_modern(void)
1847 {
1848     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1849         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1850 
1851         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1852         loc_push_restore(&nd->loc);
1853         net_client_init1(nd->nd, true, &error_fatal);
1854         loc_pop(&nd->loc);
1855         qapi_free_Netdev(nd->nd);
1856         g_free(nd);
1857     }
1858 }
1859 
1860 void net_init_clients(void)
1861 {
1862     net_change_state_entry =
1863         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1864 
1865     QTAILQ_INIT(&net_clients);
1866 
1867     netdev_init_modern();
1868 
1869     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1870                       &error_fatal);
1871 
1872     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1873                       &error_fatal);
1874 
1875     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1876                       &error_fatal);
1877 }
1878 
1879 /*
1880  * Does this -netdev argument use modern rather than traditional syntax?
1881  * Modern syntax is to be parsed with netdev_parse_modern().
1882  * Traditional syntax is to be parsed with net_client_parse().
1883  */
1884 bool netdev_is_modern(const char *optstr)
1885 {
1886     QemuOpts *opts;
1887     bool is_modern;
1888     const char *type;
1889     static QemuOptsList dummy_opts = {
1890         .name = "netdev",
1891         .implied_opt_name = "type",
1892         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1893         .desc = { { } },
1894     };
1895 
1896     if (optstr[0] == '{') {
1897         /* This is JSON, which means it's modern syntax */
1898         return true;
1899     }
1900 
1901     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1902     qemu_opts_do_parse(opts, optstr, dummy_opts.implied_opt_name,
1903                        &error_abort);
1904     type = qemu_opt_get(opts, "type");
1905     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1906 
1907     qemu_opts_reset(&dummy_opts);
1908 
1909     return is_modern;
1910 }
1911 
1912 /*
1913  * netdev_parse_modern() uses modern, more expressive syntax than
1914  * net_client_parse(), but supports only the -netdev option.
1915  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1916  * appends to @qemu_netdev_opts.
1917  */
1918 void netdev_parse_modern(const char *optstr)
1919 {
1920     Visitor *v;
1921     NetdevQueueEntry *nd;
1922 
1923     v = qobject_input_visitor_new_str(optstr, "type", &error_fatal);
1924     nd = g_new(NetdevQueueEntry, 1);
1925     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1926     visit_free(v);
1927     loc_save(&nd->loc);
1928 
1929     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1930 }
1931 
1932 void net_client_parse(QemuOptsList *opts_list, const char *optstr)
1933 {
1934     if (!qemu_opts_parse_noisily(opts_list, optstr, true)) {
1935         exit(1);
1936     }
1937 }
1938 
1939 /* From FreeBSD */
1940 /* XXX: optimize */
1941 uint32_t net_crc32(const uint8_t *p, int len)
1942 {
1943     uint32_t crc;
1944     int carry, i, j;
1945     uint8_t b;
1946 
1947     crc = 0xffffffff;
1948     for (i = 0; i < len; i++) {
1949         b = *p++;
1950         for (j = 0; j < 8; j++) {
1951             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1952             crc <<= 1;
1953             b >>= 1;
1954             if (carry) {
1955                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1956             }
1957         }
1958     }
1959 
1960     return crc;
1961 }
1962 
1963 uint32_t net_crc32_le(const uint8_t *p, int len)
1964 {
1965     uint32_t crc;
1966     int carry, i, j;
1967     uint8_t b;
1968 
1969     crc = 0xffffffff;
1970     for (i = 0; i < len; i++) {
1971         b = *p++;
1972         for (j = 0; j < 8; j++) {
1973             carry = (crc & 0x1) ^ (b & 0x01);
1974             crc >>= 1;
1975             b >>= 1;
1976             if (carry) {
1977                 crc ^= POLYNOMIAL_LE;
1978             }
1979         }
1980     }
1981 
1982     return crc;
1983 }
1984 
1985 QemuOptsList qemu_netdev_opts = {
1986     .name = "netdev",
1987     .implied_opt_name = "type",
1988     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1989     .desc = {
1990         /*
1991          * no elements => accept any params
1992          * validation will happen later
1993          */
1994         { /* end of list */ }
1995     },
1996 };
1997 
1998 QemuOptsList qemu_nic_opts = {
1999     .name = "nic",
2000     .implied_opt_name = "type",
2001     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
2002     .desc = {
2003         /*
2004          * no elements => accept any params
2005          * validation will happen later
2006          */
2007         { /* end of list */ }
2008     },
2009 };
2010 
2011 QemuOptsList qemu_net_opts = {
2012     .name = "net",
2013     .implied_opt_name = "type",
2014     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
2015     .desc = {
2016         /*
2017          * no elements => accept any params
2018          * validation will happen later
2019          */
2020         { /* end of list */ }
2021     },
2022 };
2023 
2024 void net_socket_rs_init(SocketReadState *rs,
2025                         SocketReadStateFinalize *finalize,
2026                         bool vnet_hdr)
2027 {
2028     rs->state = 0;
2029     rs->vnet_hdr = vnet_hdr;
2030     rs->index = 0;
2031     rs->packet_len = 0;
2032     rs->vnet_hdr_len = 0;
2033     memset(rs->buf, 0, sizeof(rs->buf));
2034     rs->finalize = finalize;
2035 }
2036 
2037 /*
2038  * Returns
2039  * 0: success
2040  * -1: error occurs
2041  */
2042 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
2043 {
2044     unsigned int l;
2045 
2046     while (size > 0) {
2047         /* Reassemble a packet from the network.
2048          * 0 = getting length.
2049          * 1 = getting vnet header length.
2050          * 2 = getting data.
2051          */
2052         switch (rs->state) {
2053         case 0:
2054             l = 4 - rs->index;
2055             if (l > size) {
2056                 l = size;
2057             }
2058             memcpy(rs->buf + rs->index, buf, l);
2059             buf += l;
2060             size -= l;
2061             rs->index += l;
2062             if (rs->index == 4) {
2063                 /* got length */
2064                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
2065                 rs->index = 0;
2066                 if (rs->vnet_hdr) {
2067                     rs->state = 1;
2068                 } else {
2069                     rs->state = 2;
2070                     rs->vnet_hdr_len = 0;
2071                 }
2072             }
2073             break;
2074         case 1:
2075             l = 4 - rs->index;
2076             if (l > size) {
2077                 l = size;
2078             }
2079             memcpy(rs->buf + rs->index, buf, l);
2080             buf += l;
2081             size -= l;
2082             rs->index += l;
2083             if (rs->index == 4) {
2084                 /* got vnet header length */
2085                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
2086                 rs->index = 0;
2087                 rs->state = 2;
2088             }
2089             break;
2090         case 2:
2091             l = rs->packet_len - rs->index;
2092             if (l > size) {
2093                 l = size;
2094             }
2095             if (rs->index + l <= sizeof(rs->buf)) {
2096                 memcpy(rs->buf + rs->index, buf, l);
2097             } else {
2098                 fprintf(stderr, "serious error: oversized packet received,"
2099                     "connection terminated.\n");
2100                 rs->index = rs->state = 0;
2101                 return -1;
2102             }
2103 
2104             rs->index += l;
2105             buf += l;
2106             size -= l;
2107             if (rs->index >= rs->packet_len) {
2108                 rs->index = 0;
2109                 rs->state = 0;
2110                 assert(rs->finalize);
2111                 rs->finalize(rs);
2112             }
2113             break;
2114         }
2115     }
2116 
2117     assert(size == 0);
2118     return 0;
2119 }
2120