Deleted Added
full compact
if.c (182106) if.c (183381)
1/*-
2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
1/*-
2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
30 * $FreeBSD: head/sys/net/if.c 182106 2008-08-24 11:03:43Z bz $
30 * $FreeBSD: head/sys/net/if.c 183381 2008-09-26 14:19:52Z ed $
31 */
32
33#include "opt_compat.h"
34#include "opt_inet6.h"
35#include "opt_inet.h"
36#include "opt_mac.h"
37#include "opt_carp.h"
38
39#include <sys/param.h>
40#include <sys/types.h>
41#include <sys/conf.h>
42#include <sys/malloc.h>
43#include <sys/sbuf.h>
44#include <sys/bus.h>
45#include <sys/mbuf.h>
46#include <sys/systm.h>
47#include <sys/priv.h>
48#include <sys/proc.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/protosw.h>
52#include <sys/kernel.h>
53#include <sys/sockio.h>
54#include <sys/syslog.h>
55#include <sys/sysctl.h>
56#include <sys/taskqueue.h>
57#include <sys/domain.h>
58#include <sys/jail.h>
59#include <sys/vimage.h>
60#include <machine/stdarg.h>
61
62#include <net/if.h>
63#include <net/if_clone.h>
64#include <net/if_dl.h>
65#include <net/if_types.h>
66#include <net/if_var.h>
67#include <net/radix.h>
68#include <net/route.h>
69
70#if defined(INET) || defined(INET6)
71/*XXX*/
72#include <netinet/in.h>
73#include <netinet/in_var.h>
74#ifdef INET6
75#include <netinet6/in6_var.h>
76#include <netinet6/in6_ifattach.h>
77#endif
78#endif
79#ifdef INET
80#include <netinet/if_ether.h>
81#endif
82#ifdef DEV_CARP
83#include <netinet/ip_carp.h>
84#endif
85
86#include <security/mac/mac_framework.h>
87
88SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
89SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
90
91/* Log link state change events */
92static int log_link_state_change = 1;
93
94SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
95 &log_link_state_change, 0,
96 "log interface link state change events");
97
98void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
99void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
100void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
101
102struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
103
104/*
105 * XXX: Style; these should be sorted alphabetically, and unprototyped
106 * static functions should be prototyped. Currently they are sorted by
107 * declaration order.
108 */
109static void if_attachdomain(void *);
110static void if_attachdomain1(struct ifnet *);
111static int ifconf(u_long, caddr_t);
112static void if_freemulti(struct ifmultiaddr *);
113static void if_grow(void);
114static void if_init(void *);
115static void if_qflush(struct ifaltq *);
116static void if_route(struct ifnet *, int flag, int fam);
117static int if_setflag(struct ifnet *, int, int, int *, int);
118static void if_slowtimo(void *);
119static void if_unroute(struct ifnet *, int flag, int fam);
120static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
121static int if_rtdel(struct radix_node *, void *);
122static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
123static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
124static void if_start_deferred(void *context, int pending);
125static void do_link_state_change(void *, int);
126static int if_getgroup(struct ifgroupreq *, struct ifnet *);
127static int if_getgroupmembers(struct ifgroupreq *);
128#ifdef INET6
129/*
130 * XXX: declare here to avoid to include many inet6 related files..
131 * should be more generalized?
132 */
133extern void nd6_setmtu(struct ifnet *);
134#endif
135
136int if_index = 0;
137int ifqmaxlen = IFQ_MAXLEN;
138struct ifnethead ifnet; /* depend on static init XXX */
139struct ifgrouphead ifg_head;
140struct mtx ifnet_lock;
141static if_com_alloc_t *if_com_alloc[256];
142static if_com_free_t *if_com_free[256];
143
144static int if_indexlim = 8;
145static struct knlist ifklist;
146
147/*
148 * Table of ifnet/cdev by index. Locked with ifnet_lock.
149 */
150static struct ifindex_entry *ifindex_table = NULL;
151
152static void filt_netdetach(struct knote *kn);
153static int filt_netdev(struct knote *kn, long hint);
154
155static struct filterops netdev_filtops =
156 { 1, NULL, filt_netdetach, filt_netdev };
157
158/*
159 * System initialization
160 */
161SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
162SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_slowtimo, NULL);
163
164MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
165MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
166MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
167
168struct ifnet *
169ifnet_byindex(u_short idx)
170{
171 struct ifnet *ifp;
172
173 IFNET_RLOCK();
174 ifp = V_ifindex_table[idx].ife_ifnet;
175 IFNET_RUNLOCK();
176 return (ifp);
177}
178
179static void
180ifnet_setbyindex(u_short idx, struct ifnet *ifp)
181{
182
183 IFNET_WLOCK_ASSERT();
184
185 V_ifindex_table[idx].ife_ifnet = ifp;
186}
187
188struct ifaddr *
189ifaddr_byindex(u_short idx)
190{
191 struct ifaddr *ifa;
192
193 IFNET_RLOCK();
194 ifa = ifnet_byindex(idx)->if_addr;
195 IFNET_RUNLOCK();
196 return (ifa);
197}
198
199struct cdev *
200ifdev_byindex(u_short idx)
201{
202 struct cdev *cdev;
203
204 IFNET_RLOCK();
205 cdev = V_ifindex_table[idx].ife_dev;
206 IFNET_RUNLOCK();
207 return (cdev);
208}
209
210static void
211ifdev_setbyindex(u_short idx, struct cdev *cdev)
212{
213
214 IFNET_WLOCK();
215 V_ifindex_table[idx].ife_dev = cdev;
216 IFNET_WUNLOCK();
217}
218
219static d_open_t netopen;
220static d_close_t netclose;
221static d_ioctl_t netioctl;
222static d_kqfilter_t netkqfilter;
223
224static struct cdevsw net_cdevsw = {
225 .d_version = D_VERSION,
226 .d_flags = D_NEEDGIANT,
227 .d_open = netopen,
228 .d_close = netclose,
229 .d_ioctl = netioctl,
230 .d_name = "net",
231 .d_kqfilter = netkqfilter,
232};
233
234static int
235netopen(struct cdev *dev, int flag, int mode, struct thread *td)
236{
237 return (0);
238}
239
240static int
241netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
242{
243 return (0);
244}
245
246static int
247netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
248{
249 struct ifnet *ifp;
250 int error, idx;
251
252 /* only support interface specific ioctls */
253 if (IOCGROUP(cmd) != 'i')
254 return (EOPNOTSUPP);
255 idx = minor(dev);
256 if (idx == 0) {
257 /*
258 * special network device, not interface.
259 */
260 if (cmd == SIOCGIFCONF)
261 return (ifconf(cmd, data)); /* XXX remove cmd */
262#ifdef __amd64__
263 if (cmd == SIOCGIFCONF32)
264 return (ifconf(cmd, data)); /* XXX remove cmd */
265#endif
266 return (EOPNOTSUPP);
267 }
268
269 ifp = ifnet_byindex(idx);
270 if (ifp == NULL)
271 return (ENXIO);
272
273 error = ifhwioctl(cmd, ifp, data, td);
274 if (error == ENOIOCTL)
275 error = EOPNOTSUPP;
276 return (error);
277}
278
279static int
280netkqfilter(struct cdev *dev, struct knote *kn)
281{
282 struct knlist *klist;
283 struct ifnet *ifp;
284 int idx;
285
286 switch (kn->kn_filter) {
287 case EVFILT_NETDEV:
288 kn->kn_fop = &netdev_filtops;
289 break;
290 default:
291 return (EINVAL);
292 }
293
294 idx = minor(dev);
295 if (idx == 0) {
296 klist = &V_ifklist;
297 } else {
298 ifp = ifnet_byindex(idx);
299 if (ifp == NULL)
300 return (1);
301 klist = &ifp->if_klist;
302 }
303
304 kn->kn_hook = (caddr_t)klist;
305
306 knlist_add(klist, kn, 0);
307
308 return (0);
309}
310
311static void
312filt_netdetach(struct knote *kn)
313{
314 struct knlist *klist = (struct knlist *)kn->kn_hook;
315
316 knlist_remove(klist, kn, 0);
317}
318
319static int
320filt_netdev(struct knote *kn, long hint)
321{
322 struct knlist *klist = (struct knlist *)kn->kn_hook;
323
324 /*
325 * Currently NOTE_EXIT is abused to indicate device detach.
326 */
327 if (hint == NOTE_EXIT) {
328 kn->kn_data = NOTE_LINKINV;
329 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
330 knlist_remove_inevent(klist, kn);
331 return (1);
332 }
333 if (hint != 0)
334 kn->kn_data = hint; /* current status */
335 if (kn->kn_sfflags & hint)
336 kn->kn_fflags |= hint;
337 return (kn->kn_fflags != 0);
338}
339
340/*
341 * Network interface utility routines.
342 *
343 * Routines with ifa_ifwith* names take sockaddr *'s as
344 * parameters.
345 */
346
347/* ARGSUSED*/
348static void
349if_init(void *dummy __unused)
350{
351
352 IFNET_LOCK_INIT();
353 TAILQ_INIT(&V_ifnet);
354 TAILQ_INIT(&V_ifg_head);
355 knlist_init(&V_ifklist, NULL, NULL, NULL, NULL);
356 if_grow(); /* create initial table */
357 ifdev_setbyindex(0, make_dev(&net_cdevsw, 0, UID_ROOT, GID_WHEEL,
358 0600, "network"));
359 if_clone_init();
360}
361
362static void
363if_grow(void)
364{
365 u_int n;
366 struct ifindex_entry *e;
367
368 V_if_indexlim <<= 1;
369 n = V_if_indexlim * sizeof(*e);
370 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
371 if (V_ifindex_table != NULL) {
372 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
373 free((caddr_t)V_ifindex_table, M_IFNET);
374 }
375 V_ifindex_table = e;
376}
377
378/*
379 * Allocate a struct ifnet and an index for an interface. A layer 2
380 * common structure will also be allocated if an allocation routine is
381 * registered for the passed type.
382 */
383struct ifnet*
384if_alloc(u_char type)
385{
386 struct ifnet *ifp;
387
388 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
389
390 /*
391 * Try to find an empty slot below if_index. If we fail, take
392 * the next slot.
393 *
394 * XXX: should be locked!
395 */
396 for (ifp->if_index = 1; ifp->if_index <= V_if_index; ifp->if_index++) {
397 if (ifnet_byindex(ifp->if_index) == NULL)
398 break;
399 }
400 /* Catch if_index overflow. */
401 if (ifp->if_index < 1) {
402 free(ifp, M_IFNET);
403 return (NULL);
404 }
405 if (ifp->if_index > V_if_index)
406 V_if_index = ifp->if_index;
407 if (V_if_index >= V_if_indexlim)
408 if_grow();
409
410 ifp->if_type = type;
411
412 if (if_com_alloc[type] != NULL) {
413 ifp->if_l2com = if_com_alloc[type](type, ifp);
414 if (ifp->if_l2com == NULL) {
415 free(ifp, M_IFNET);
416 return (NULL);
417 }
418 }
419 IFNET_WLOCK();
420 ifnet_setbyindex(ifp->if_index, ifp);
421 IFNET_WUNLOCK();
422 IF_ADDR_LOCK_INIT(ifp);
423
424 return (ifp);
425}
426
427/*
428 * Free the struct ifnet, the associated index, and the layer 2 common
429 * structure if needed. All the work is done in if_free_type().
430 *
431 * Do not add code to this function! Add it to if_free_type().
432 */
433void
434if_free(struct ifnet *ifp)
435{
436
437 if_free_type(ifp, ifp->if_type);
438}
439
440/*
441 * Do the actual work of freeing a struct ifnet, associated index, and
442 * layer 2 common structure. This version should only be called by
443 * intefaces that switch their type after calling if_alloc().
444 */
445void
446if_free_type(struct ifnet *ifp, u_char type)
447{
448
449 if (ifp != ifnet_byindex(ifp->if_index)) {
450 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
451 __func__);
452 return;
453 }
454
455 IFNET_WLOCK();
456 ifnet_setbyindex(ifp->if_index, NULL);
457
458 /* XXX: should be locked with if_findindex() */
459 while (V_if_index > 0 && ifnet_byindex(V_if_index) == NULL)
460 V_if_index--;
461 IFNET_WUNLOCK();
462
463 if (if_com_free[type] != NULL)
464 if_com_free[type](ifp->if_l2com, type);
465
466 IF_ADDR_LOCK_DESTROY(ifp);
467 free(ifp, M_IFNET);
468};
469
470/*
471 * Perform generic interface initalization tasks and attach the interface
472 * to the list of "active" interfaces.
473 *
474 * XXX:
475 * - The decision to return void and thus require this function to
476 * succeed is questionable.
477 * - We do more initialization here then is probably a good idea.
478 * Some of this should probably move to if_alloc().
479 * - We should probably do more sanity checking. For instance we don't
480 * do anything to insure if_xname is unique or non-empty.
481 */
482void
483if_attach(struct ifnet *ifp)
484{
485 unsigned socksize, ifasize;
486 int namelen, masklen;
487 struct sockaddr_dl *sdl;
488 struct ifaddr *ifa;
489
490 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
491 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
492 ifp->if_xname);
493
494 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
495 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
496 IF_AFDATA_LOCK_INIT(ifp);
497 ifp->if_afdata_initialized = 0;
498
499 TAILQ_INIT(&ifp->if_addrhead);
500 TAILQ_INIT(&ifp->if_prefixhead);
501 TAILQ_INIT(&ifp->if_multiaddrs);
502 TAILQ_INIT(&ifp->if_groups);
503
504 if_addgroup(ifp, IFG_ALL);
505
506 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
507 getmicrotime(&ifp->if_lastchange);
508 ifp->if_data.ifi_epoch = time_uptime;
509 ifp->if_data.ifi_datalen = sizeof(struct if_data);
510
511#ifdef MAC
512 mac_ifnet_init(ifp);
513 mac_ifnet_create(ifp);
514#endif
515
516 ifdev_setbyindex(ifp->if_index, make_dev(&net_cdevsw,
31 */
32
33#include "opt_compat.h"
34#include "opt_inet6.h"
35#include "opt_inet.h"
36#include "opt_mac.h"
37#include "opt_carp.h"
38
39#include <sys/param.h>
40#include <sys/types.h>
41#include <sys/conf.h>
42#include <sys/malloc.h>
43#include <sys/sbuf.h>
44#include <sys/bus.h>
45#include <sys/mbuf.h>
46#include <sys/systm.h>
47#include <sys/priv.h>
48#include <sys/proc.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/protosw.h>
52#include <sys/kernel.h>
53#include <sys/sockio.h>
54#include <sys/syslog.h>
55#include <sys/sysctl.h>
56#include <sys/taskqueue.h>
57#include <sys/domain.h>
58#include <sys/jail.h>
59#include <sys/vimage.h>
60#include <machine/stdarg.h>
61
62#include <net/if.h>
63#include <net/if_clone.h>
64#include <net/if_dl.h>
65#include <net/if_types.h>
66#include <net/if_var.h>
67#include <net/radix.h>
68#include <net/route.h>
69
70#if defined(INET) || defined(INET6)
71/*XXX*/
72#include <netinet/in.h>
73#include <netinet/in_var.h>
74#ifdef INET6
75#include <netinet6/in6_var.h>
76#include <netinet6/in6_ifattach.h>
77#endif
78#endif
79#ifdef INET
80#include <netinet/if_ether.h>
81#endif
82#ifdef DEV_CARP
83#include <netinet/ip_carp.h>
84#endif
85
86#include <security/mac/mac_framework.h>
87
88SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
89SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
90
91/* Log link state change events */
92static int log_link_state_change = 1;
93
94SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
95 &log_link_state_change, 0,
96 "log interface link state change events");
97
98void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
99void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
100void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
101
102struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
103
104/*
105 * XXX: Style; these should be sorted alphabetically, and unprototyped
106 * static functions should be prototyped. Currently they are sorted by
107 * declaration order.
108 */
109static void if_attachdomain(void *);
110static void if_attachdomain1(struct ifnet *);
111static int ifconf(u_long, caddr_t);
112static void if_freemulti(struct ifmultiaddr *);
113static void if_grow(void);
114static void if_init(void *);
115static void if_qflush(struct ifaltq *);
116static void if_route(struct ifnet *, int flag, int fam);
117static int if_setflag(struct ifnet *, int, int, int *, int);
118static void if_slowtimo(void *);
119static void if_unroute(struct ifnet *, int flag, int fam);
120static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
121static int if_rtdel(struct radix_node *, void *);
122static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
123static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
124static void if_start_deferred(void *context, int pending);
125static void do_link_state_change(void *, int);
126static int if_getgroup(struct ifgroupreq *, struct ifnet *);
127static int if_getgroupmembers(struct ifgroupreq *);
128#ifdef INET6
129/*
130 * XXX: declare here to avoid to include many inet6 related files..
131 * should be more generalized?
132 */
133extern void nd6_setmtu(struct ifnet *);
134#endif
135
136int if_index = 0;
137int ifqmaxlen = IFQ_MAXLEN;
138struct ifnethead ifnet; /* depend on static init XXX */
139struct ifgrouphead ifg_head;
140struct mtx ifnet_lock;
141static if_com_alloc_t *if_com_alloc[256];
142static if_com_free_t *if_com_free[256];
143
144static int if_indexlim = 8;
145static struct knlist ifklist;
146
147/*
148 * Table of ifnet/cdev by index. Locked with ifnet_lock.
149 */
150static struct ifindex_entry *ifindex_table = NULL;
151
152static void filt_netdetach(struct knote *kn);
153static int filt_netdev(struct knote *kn, long hint);
154
155static struct filterops netdev_filtops =
156 { 1, NULL, filt_netdetach, filt_netdev };
157
158/*
159 * System initialization
160 */
161SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
162SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_slowtimo, NULL);
163
164MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
165MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
166MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
167
168struct ifnet *
169ifnet_byindex(u_short idx)
170{
171 struct ifnet *ifp;
172
173 IFNET_RLOCK();
174 ifp = V_ifindex_table[idx].ife_ifnet;
175 IFNET_RUNLOCK();
176 return (ifp);
177}
178
179static void
180ifnet_setbyindex(u_short idx, struct ifnet *ifp)
181{
182
183 IFNET_WLOCK_ASSERT();
184
185 V_ifindex_table[idx].ife_ifnet = ifp;
186}
187
188struct ifaddr *
189ifaddr_byindex(u_short idx)
190{
191 struct ifaddr *ifa;
192
193 IFNET_RLOCK();
194 ifa = ifnet_byindex(idx)->if_addr;
195 IFNET_RUNLOCK();
196 return (ifa);
197}
198
199struct cdev *
200ifdev_byindex(u_short idx)
201{
202 struct cdev *cdev;
203
204 IFNET_RLOCK();
205 cdev = V_ifindex_table[idx].ife_dev;
206 IFNET_RUNLOCK();
207 return (cdev);
208}
209
210static void
211ifdev_setbyindex(u_short idx, struct cdev *cdev)
212{
213
214 IFNET_WLOCK();
215 V_ifindex_table[idx].ife_dev = cdev;
216 IFNET_WUNLOCK();
217}
218
219static d_open_t netopen;
220static d_close_t netclose;
221static d_ioctl_t netioctl;
222static d_kqfilter_t netkqfilter;
223
224static struct cdevsw net_cdevsw = {
225 .d_version = D_VERSION,
226 .d_flags = D_NEEDGIANT,
227 .d_open = netopen,
228 .d_close = netclose,
229 .d_ioctl = netioctl,
230 .d_name = "net",
231 .d_kqfilter = netkqfilter,
232};
233
234static int
235netopen(struct cdev *dev, int flag, int mode, struct thread *td)
236{
237 return (0);
238}
239
240static int
241netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
242{
243 return (0);
244}
245
246static int
247netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
248{
249 struct ifnet *ifp;
250 int error, idx;
251
252 /* only support interface specific ioctls */
253 if (IOCGROUP(cmd) != 'i')
254 return (EOPNOTSUPP);
255 idx = minor(dev);
256 if (idx == 0) {
257 /*
258 * special network device, not interface.
259 */
260 if (cmd == SIOCGIFCONF)
261 return (ifconf(cmd, data)); /* XXX remove cmd */
262#ifdef __amd64__
263 if (cmd == SIOCGIFCONF32)
264 return (ifconf(cmd, data)); /* XXX remove cmd */
265#endif
266 return (EOPNOTSUPP);
267 }
268
269 ifp = ifnet_byindex(idx);
270 if (ifp == NULL)
271 return (ENXIO);
272
273 error = ifhwioctl(cmd, ifp, data, td);
274 if (error == ENOIOCTL)
275 error = EOPNOTSUPP;
276 return (error);
277}
278
279static int
280netkqfilter(struct cdev *dev, struct knote *kn)
281{
282 struct knlist *klist;
283 struct ifnet *ifp;
284 int idx;
285
286 switch (kn->kn_filter) {
287 case EVFILT_NETDEV:
288 kn->kn_fop = &netdev_filtops;
289 break;
290 default:
291 return (EINVAL);
292 }
293
294 idx = minor(dev);
295 if (idx == 0) {
296 klist = &V_ifklist;
297 } else {
298 ifp = ifnet_byindex(idx);
299 if (ifp == NULL)
300 return (1);
301 klist = &ifp->if_klist;
302 }
303
304 kn->kn_hook = (caddr_t)klist;
305
306 knlist_add(klist, kn, 0);
307
308 return (0);
309}
310
311static void
312filt_netdetach(struct knote *kn)
313{
314 struct knlist *klist = (struct knlist *)kn->kn_hook;
315
316 knlist_remove(klist, kn, 0);
317}
318
319static int
320filt_netdev(struct knote *kn, long hint)
321{
322 struct knlist *klist = (struct knlist *)kn->kn_hook;
323
324 /*
325 * Currently NOTE_EXIT is abused to indicate device detach.
326 */
327 if (hint == NOTE_EXIT) {
328 kn->kn_data = NOTE_LINKINV;
329 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
330 knlist_remove_inevent(klist, kn);
331 return (1);
332 }
333 if (hint != 0)
334 kn->kn_data = hint; /* current status */
335 if (kn->kn_sfflags & hint)
336 kn->kn_fflags |= hint;
337 return (kn->kn_fflags != 0);
338}
339
340/*
341 * Network interface utility routines.
342 *
343 * Routines with ifa_ifwith* names take sockaddr *'s as
344 * parameters.
345 */
346
347/* ARGSUSED*/
348static void
349if_init(void *dummy __unused)
350{
351
352 IFNET_LOCK_INIT();
353 TAILQ_INIT(&V_ifnet);
354 TAILQ_INIT(&V_ifg_head);
355 knlist_init(&V_ifklist, NULL, NULL, NULL, NULL);
356 if_grow(); /* create initial table */
357 ifdev_setbyindex(0, make_dev(&net_cdevsw, 0, UID_ROOT, GID_WHEEL,
358 0600, "network"));
359 if_clone_init();
360}
361
362static void
363if_grow(void)
364{
365 u_int n;
366 struct ifindex_entry *e;
367
368 V_if_indexlim <<= 1;
369 n = V_if_indexlim * sizeof(*e);
370 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
371 if (V_ifindex_table != NULL) {
372 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
373 free((caddr_t)V_ifindex_table, M_IFNET);
374 }
375 V_ifindex_table = e;
376}
377
378/*
379 * Allocate a struct ifnet and an index for an interface. A layer 2
380 * common structure will also be allocated if an allocation routine is
381 * registered for the passed type.
382 */
383struct ifnet*
384if_alloc(u_char type)
385{
386 struct ifnet *ifp;
387
388 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
389
390 /*
391 * Try to find an empty slot below if_index. If we fail, take
392 * the next slot.
393 *
394 * XXX: should be locked!
395 */
396 for (ifp->if_index = 1; ifp->if_index <= V_if_index; ifp->if_index++) {
397 if (ifnet_byindex(ifp->if_index) == NULL)
398 break;
399 }
400 /* Catch if_index overflow. */
401 if (ifp->if_index < 1) {
402 free(ifp, M_IFNET);
403 return (NULL);
404 }
405 if (ifp->if_index > V_if_index)
406 V_if_index = ifp->if_index;
407 if (V_if_index >= V_if_indexlim)
408 if_grow();
409
410 ifp->if_type = type;
411
412 if (if_com_alloc[type] != NULL) {
413 ifp->if_l2com = if_com_alloc[type](type, ifp);
414 if (ifp->if_l2com == NULL) {
415 free(ifp, M_IFNET);
416 return (NULL);
417 }
418 }
419 IFNET_WLOCK();
420 ifnet_setbyindex(ifp->if_index, ifp);
421 IFNET_WUNLOCK();
422 IF_ADDR_LOCK_INIT(ifp);
423
424 return (ifp);
425}
426
427/*
428 * Free the struct ifnet, the associated index, and the layer 2 common
429 * structure if needed. All the work is done in if_free_type().
430 *
431 * Do not add code to this function! Add it to if_free_type().
432 */
433void
434if_free(struct ifnet *ifp)
435{
436
437 if_free_type(ifp, ifp->if_type);
438}
439
440/*
441 * Do the actual work of freeing a struct ifnet, associated index, and
442 * layer 2 common structure. This version should only be called by
443 * intefaces that switch their type after calling if_alloc().
444 */
445void
446if_free_type(struct ifnet *ifp, u_char type)
447{
448
449 if (ifp != ifnet_byindex(ifp->if_index)) {
450 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
451 __func__);
452 return;
453 }
454
455 IFNET_WLOCK();
456 ifnet_setbyindex(ifp->if_index, NULL);
457
458 /* XXX: should be locked with if_findindex() */
459 while (V_if_index > 0 && ifnet_byindex(V_if_index) == NULL)
460 V_if_index--;
461 IFNET_WUNLOCK();
462
463 if (if_com_free[type] != NULL)
464 if_com_free[type](ifp->if_l2com, type);
465
466 IF_ADDR_LOCK_DESTROY(ifp);
467 free(ifp, M_IFNET);
468};
469
470/*
471 * Perform generic interface initalization tasks and attach the interface
472 * to the list of "active" interfaces.
473 *
474 * XXX:
475 * - The decision to return void and thus require this function to
476 * succeed is questionable.
477 * - We do more initialization here then is probably a good idea.
478 * Some of this should probably move to if_alloc().
479 * - We should probably do more sanity checking. For instance we don't
480 * do anything to insure if_xname is unique or non-empty.
481 */
482void
483if_attach(struct ifnet *ifp)
484{
485 unsigned socksize, ifasize;
486 int namelen, masklen;
487 struct sockaddr_dl *sdl;
488 struct ifaddr *ifa;
489
490 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
491 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
492 ifp->if_xname);
493
494 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
495 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
496 IF_AFDATA_LOCK_INIT(ifp);
497 ifp->if_afdata_initialized = 0;
498
499 TAILQ_INIT(&ifp->if_addrhead);
500 TAILQ_INIT(&ifp->if_prefixhead);
501 TAILQ_INIT(&ifp->if_multiaddrs);
502 TAILQ_INIT(&ifp->if_groups);
503
504 if_addgroup(ifp, IFG_ALL);
505
506 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
507 getmicrotime(&ifp->if_lastchange);
508 ifp->if_data.ifi_epoch = time_uptime;
509 ifp->if_data.ifi_datalen = sizeof(struct if_data);
510
511#ifdef MAC
512 mac_ifnet_init(ifp);
513 mac_ifnet_create(ifp);
514#endif
515
516 ifdev_setbyindex(ifp->if_index, make_dev(&net_cdevsw,
517 unit2minor(ifp->if_index), UID_ROOT, GID_WHEEL, 0600, "%s/%s",
517 ifp->if_index, UID_ROOT, GID_WHEEL, 0600, "%s/%s",
518 net_cdevsw.d_name, ifp->if_xname));
519 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
520 net_cdevsw.d_name, ifp->if_index);
521
522 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
523
524 /*
525 * create a Link Level name for this device
526 */
527 namelen = strlen(ifp->if_xname);
528 /*
529 * Always save enough space for any possiable name so we can do
530 * a rename in place later.
531 */
532 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
533 socksize = masklen + ifp->if_addrlen;
534 if (socksize < sizeof(*sdl))
535 socksize = sizeof(*sdl);
536 socksize = roundup2(socksize, sizeof(long));
537 ifasize = sizeof(*ifa) + 2 * socksize;
538 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
539 IFA_LOCK_INIT(ifa);
540 sdl = (struct sockaddr_dl *)(ifa + 1);
541 sdl->sdl_len = socksize;
542 sdl->sdl_family = AF_LINK;
543 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
544 sdl->sdl_nlen = namelen;
545 sdl->sdl_index = ifp->if_index;
546 sdl->sdl_type = ifp->if_type;
547 ifp->if_addr = ifa;
548 ifa->ifa_ifp = ifp;
549 ifa->ifa_rtrequest = link_rtrequest;
550 ifa->ifa_addr = (struct sockaddr *)sdl;
551 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
552 ifa->ifa_netmask = (struct sockaddr *)sdl;
553 sdl->sdl_len = masklen;
554 while (namelen != 0)
555 sdl->sdl_data[--namelen] = 0xff;
556 ifa->ifa_refcnt = 1;
557 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
558 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
559
560 /*
561 * XXX: why do we warn about this? We're correcting it and most
562 * drivers just set the value the way we do.
563 */
564 if (ifp->if_snd.ifq_maxlen == 0) {
565 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
566 ifp->if_snd.ifq_maxlen = ifqmaxlen;
567 }
568 ifp->if_snd.altq_type = 0;
569 ifp->if_snd.altq_disc = NULL;
570 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
571 ifp->if_snd.altq_tbr = NULL;
572 ifp->if_snd.altq_ifp = ifp;
573
574 IFNET_WLOCK();
575 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
576 IFNET_WUNLOCK();
577
578 if (domain_init_status >= 2)
579 if_attachdomain1(ifp);
580
581 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
582 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
583
584 /* Announce the interface. */
585 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
586
587 if (ifp->if_watchdog != NULL)
588 if_printf(ifp,
589 "WARNING: using obsoleted if_watchdog interface\n");
590 if (ifp->if_flags & IFF_NEEDSGIANT)
591 if_printf(ifp,
592 "WARNING: using obsoleted IFF_NEEDSGIANT flag\n");
593}
594
595static void
596if_attachdomain(void *dummy)
597{
598 struct ifnet *ifp;
599 int s;
600
601 s = splnet();
602 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
603 if_attachdomain1(ifp);
604 splx(s);
605}
606SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
607 if_attachdomain, NULL);
608
609static void
610if_attachdomain1(struct ifnet *ifp)
611{
612 struct domain *dp;
613 int s;
614
615 s = splnet();
616
617 /*
618 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
619 * cannot lock ifp->if_afdata initialization, entirely.
620 */
621 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
622 splx(s);
623 return;
624 }
625 if (ifp->if_afdata_initialized >= domain_init_status) {
626 IF_AFDATA_UNLOCK(ifp);
627 splx(s);
628 printf("if_attachdomain called more than once on %s\n",
629 ifp->if_xname);
630 return;
631 }
632 ifp->if_afdata_initialized = domain_init_status;
633 IF_AFDATA_UNLOCK(ifp);
634
635 /* address family dependent data region */
636 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
637 for (dp = domains; dp; dp = dp->dom_next) {
638 if (dp->dom_ifattach)
639 ifp->if_afdata[dp->dom_family] =
640 (*dp->dom_ifattach)(ifp);
641 }
642
643 splx(s);
644}
645
646/*
647 * Remove any unicast or broadcast network addresses from an interface.
648 */
649void
650if_purgeaddrs(struct ifnet *ifp)
651{
652 struct ifaddr *ifa, *next;
653
654 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
655 if (ifa->ifa_addr->sa_family == AF_LINK)
656 continue;
657#ifdef INET
658 /* XXX: Ugly!! ad hoc just for INET */
659 if (ifa->ifa_addr->sa_family == AF_INET) {
660 struct ifaliasreq ifr;
661
662 bzero(&ifr, sizeof(ifr));
663 ifr.ifra_addr = *ifa->ifa_addr;
664 if (ifa->ifa_dstaddr)
665 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
666 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
667 NULL) == 0)
668 continue;
669 }
670#endif /* INET */
671#ifdef INET6
672 if (ifa->ifa_addr->sa_family == AF_INET6) {
673 in6_purgeaddr(ifa);
674 /* ifp_addrhead is already updated */
675 continue;
676 }
677#endif /* INET6 */
678 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
679 IFAFREE(ifa);
680 }
681}
682
683/*
684 * Remove any multicast network addresses from an interface.
685 */
686void
687if_purgemaddrs(struct ifnet *ifp)
688{
689 struct ifmultiaddr *ifma;
690 struct ifmultiaddr *next;
691
692 IF_ADDR_LOCK(ifp);
693 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
694 if_delmulti_locked(ifp, ifma, 1);
695 IF_ADDR_UNLOCK(ifp);
696}
697
698/*
699 * Detach an interface, removing it from the
700 * list of "active" interfaces.
701 *
702 * XXXRW: There are some significant questions about event ordering, and
703 * how to prevent things from starting to use the interface during detach.
704 */
705void
706if_detach(struct ifnet *ifp)
707{
708 struct ifaddr *ifa;
709 struct radix_node_head *rnh;
710 int s;
711 int i;
712 struct domain *dp;
713 struct ifnet *iter;
714 int found = 0;
715
716 IFNET_WLOCK();
717 TAILQ_FOREACH(iter, &V_ifnet, if_link)
718 if (iter == ifp) {
719 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
720 found = 1;
721 break;
722 }
723 IFNET_WUNLOCK();
724 if (!found)
725 return;
726
727 /*
728 * Remove/wait for pending events.
729 */
730 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
731
732 /*
733 * Remove routes and flush queues.
734 */
735 s = splnet();
736 if_down(ifp);
737#ifdef ALTQ
738 if (ALTQ_IS_ENABLED(&ifp->if_snd))
739 altq_disable(&ifp->if_snd);
740 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
741 altq_detach(&ifp->if_snd);
742#endif
743
744 if_purgeaddrs(ifp);
745
746#ifdef INET
747 in_ifdetach(ifp);
748#endif
749
750#ifdef INET6
751 /*
752 * Remove all IPv6 kernel structs related to ifp. This should be done
753 * before removing routing entries below, since IPv6 interface direct
754 * routes are expected to be removed by the IPv6-specific kernel API.
755 * Otherwise, the kernel will detect some inconsistency and bark it.
756 */
757 in6_ifdetach(ifp);
758#endif
759 if_purgemaddrs(ifp);
760
761 /*
762 * Remove link ifaddr pointer and maybe decrement if_index.
763 * Clean up all addresses.
764 */
765 ifp->if_addr = NULL;
766 destroy_dev(ifdev_byindex(ifp->if_index));
767 ifdev_setbyindex(ifp->if_index, NULL);
768
769 /* We can now free link ifaddr. */
770 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
771 ifa = TAILQ_FIRST(&ifp->if_addrhead);
772 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
773 IFAFREE(ifa);
774 }
775
776 /*
777 * Delete all remaining routes using this interface
778 * Unfortuneatly the only way to do this is to slog through
779 * the entire routing table looking for routes which point
780 * to this interface...oh well...
781 */
782 for (i = 1; i <= AF_MAX; i++) {
783 int j;
784 for (j = 0; j < rt_numfibs; j++) {
785 if ((rnh = V_rt_tables[j][i]) == NULL)
786 continue;
787 RADIX_NODE_HEAD_LOCK(rnh);
788 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
789 RADIX_NODE_HEAD_UNLOCK(rnh);
790 }
791 }
792
793 /* Announce that the interface is gone. */
794 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
795 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
796 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
797
798 IF_AFDATA_LOCK(ifp);
799 for (dp = domains; dp; dp = dp->dom_next) {
800 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
801 (*dp->dom_ifdetach)(ifp,
802 ifp->if_afdata[dp->dom_family]);
803 }
804 IF_AFDATA_UNLOCK(ifp);
805
806#ifdef MAC
807 mac_ifnet_destroy(ifp);
808#endif /* MAC */
809 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
810 knlist_clear(&ifp->if_klist, 0);
811 knlist_destroy(&ifp->if_klist);
812 mtx_destroy(&ifp->if_snd.ifq_mtx);
813 IF_AFDATA_DESTROY(ifp);
814 splx(s);
815}
816
817/*
818 * Add a group to an interface
819 */
820int
821if_addgroup(struct ifnet *ifp, const char *groupname)
822{
823 struct ifg_list *ifgl;
824 struct ifg_group *ifg = NULL;
825 struct ifg_member *ifgm;
826
827 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
828 groupname[strlen(groupname) - 1] <= '9')
829 return (EINVAL);
830
831 IFNET_WLOCK();
832 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
833 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
834 IFNET_WUNLOCK();
835 return (EEXIST);
836 }
837
838 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
839 M_NOWAIT)) == NULL) {
840 IFNET_WUNLOCK();
841 return (ENOMEM);
842 }
843
844 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
845 M_TEMP, M_NOWAIT)) == NULL) {
846 free(ifgl, M_TEMP);
847 IFNET_WUNLOCK();
848 return (ENOMEM);
849 }
850
851 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
852 if (!strcmp(ifg->ifg_group, groupname))
853 break;
854
855 if (ifg == NULL) {
856 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
857 M_TEMP, M_NOWAIT)) == NULL) {
858 free(ifgl, M_TEMP);
859 free(ifgm, M_TEMP);
860 IFNET_WUNLOCK();
861 return (ENOMEM);
862 }
863 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
864 ifg->ifg_refcnt = 0;
865 TAILQ_INIT(&ifg->ifg_members);
866 EVENTHANDLER_INVOKE(group_attach_event, ifg);
867 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
868 }
869
870 ifg->ifg_refcnt++;
871 ifgl->ifgl_group = ifg;
872 ifgm->ifgm_ifp = ifp;
873
874 IF_ADDR_LOCK(ifp);
875 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
876 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
877 IF_ADDR_UNLOCK(ifp);
878
879 IFNET_WUNLOCK();
880
881 EVENTHANDLER_INVOKE(group_change_event, groupname);
882
883 return (0);
884}
885
886/*
887 * Remove a group from an interface
888 */
889int
890if_delgroup(struct ifnet *ifp, const char *groupname)
891{
892 struct ifg_list *ifgl;
893 struct ifg_member *ifgm;
894
895 IFNET_WLOCK();
896 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
897 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
898 break;
899 if (ifgl == NULL) {
900 IFNET_WUNLOCK();
901 return (ENOENT);
902 }
903
904 IF_ADDR_LOCK(ifp);
905 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
906 IF_ADDR_UNLOCK(ifp);
907
908 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
909 if (ifgm->ifgm_ifp == ifp)
910 break;
911
912 if (ifgm != NULL) {
913 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
914 free(ifgm, M_TEMP);
915 }
916
917 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
918 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
919 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
920 free(ifgl->ifgl_group, M_TEMP);
921 }
922 IFNET_WUNLOCK();
923
924 free(ifgl, M_TEMP);
925
926 EVENTHANDLER_INVOKE(group_change_event, groupname);
927
928 return (0);
929}
930
931/*
932 * Stores all groups from an interface in memory pointed
933 * to by data
934 */
935static int
936if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
937{
938 int len, error;
939 struct ifg_list *ifgl;
940 struct ifg_req ifgrq, *ifgp;
941 struct ifgroupreq *ifgr = data;
942
943 if (ifgr->ifgr_len == 0) {
944 IF_ADDR_LOCK(ifp);
945 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
946 ifgr->ifgr_len += sizeof(struct ifg_req);
947 IF_ADDR_UNLOCK(ifp);
948 return (0);
949 }
950
951 len = ifgr->ifgr_len;
952 ifgp = ifgr->ifgr_groups;
953 /* XXX: wire */
954 IF_ADDR_LOCK(ifp);
955 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
956 if (len < sizeof(ifgrq)) {
957 IF_ADDR_UNLOCK(ifp);
958 return (EINVAL);
959 }
960 bzero(&ifgrq, sizeof ifgrq);
961 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
962 sizeof(ifgrq.ifgrq_group));
963 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
964 IF_ADDR_UNLOCK(ifp);
965 return (error);
966 }
967 len -= sizeof(ifgrq);
968 ifgp++;
969 }
970 IF_ADDR_UNLOCK(ifp);
971
972 return (0);
973}
974
975/*
976 * Stores all members of a group in memory pointed to by data
977 */
978static int
979if_getgroupmembers(struct ifgroupreq *data)
980{
981 struct ifgroupreq *ifgr = data;
982 struct ifg_group *ifg;
983 struct ifg_member *ifgm;
984 struct ifg_req ifgrq, *ifgp;
985 int len, error;
986
987 IFNET_RLOCK();
988 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
989 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
990 break;
991 if (ifg == NULL) {
992 IFNET_RUNLOCK();
993 return (ENOENT);
994 }
995
996 if (ifgr->ifgr_len == 0) {
997 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
998 ifgr->ifgr_len += sizeof(ifgrq);
999 IFNET_RUNLOCK();
1000 return (0);
1001 }
1002
1003 len = ifgr->ifgr_len;
1004 ifgp = ifgr->ifgr_groups;
1005 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1006 if (len < sizeof(ifgrq)) {
1007 IFNET_RUNLOCK();
1008 return (EINVAL);
1009 }
1010 bzero(&ifgrq, sizeof ifgrq);
1011 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1012 sizeof(ifgrq.ifgrq_member));
1013 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1014 IFNET_RUNLOCK();
1015 return (error);
1016 }
1017 len -= sizeof(ifgrq);
1018 ifgp++;
1019 }
1020 IFNET_RUNLOCK();
1021
1022 return (0);
1023}
1024
1025/*
1026 * Delete Routes for a Network Interface
1027 *
1028 * Called for each routing entry via the rnh->rnh_walktree() call above
1029 * to delete all route entries referencing a detaching network interface.
1030 *
1031 * Arguments:
1032 * rn pointer to node in the routing table
1033 * arg argument passed to rnh->rnh_walktree() - detaching interface
1034 *
1035 * Returns:
1036 * 0 successful
1037 * errno failed - reason indicated
1038 *
1039 */
1040static int
1041if_rtdel(struct radix_node *rn, void *arg)
1042{
1043 struct rtentry *rt = (struct rtentry *)rn;
1044 struct ifnet *ifp = arg;
1045 int err;
1046
1047 if (rt->rt_ifp == ifp) {
1048
1049 /*
1050 * Protect (sorta) against walktree recursion problems
1051 * with cloned routes
1052 */
1053 if ((rt->rt_flags & RTF_UP) == 0)
1054 return (0);
1055
1056 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1057 rt_mask(rt), rt->rt_flags,
1058 (struct rtentry **) NULL, rt->rt_fibnum);
1059 if (err) {
1060 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1061 }
1062 }
1063
1064 return (0);
1065}
1066
1067/*
1068 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1069 * structs used to represent other address families, it is necessary
1070 * to perform a different comparison.
1071 */
1072
1073#define sa_equal(a1, a2) \
1074 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1075
1076#define sa_dl_equal(a1, a2) \
1077 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1078 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1079 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1080 LLADDR((struct sockaddr_dl *)(a2)), \
1081 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1082
1083/*
1084 * Locate an interface based on a complete address.
1085 */
1086/*ARGSUSED*/
1087struct ifaddr *
1088ifa_ifwithaddr(struct sockaddr *addr)
1089{
1090 struct ifnet *ifp;
1091 struct ifaddr *ifa;
1092
1093 IFNET_RLOCK();
1094 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
1095 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1096 if (ifa->ifa_addr->sa_family != addr->sa_family)
1097 continue;
1098 if (sa_equal(addr, ifa->ifa_addr))
1099 goto done;
1100 /* IP6 doesn't have broadcast */
1101 if ((ifp->if_flags & IFF_BROADCAST) &&
1102 ifa->ifa_broadaddr &&
1103 ifa->ifa_broadaddr->sa_len != 0 &&
1104 sa_equal(ifa->ifa_broadaddr, addr))
1105 goto done;
1106 }
1107 ifa = NULL;
1108done:
1109 IFNET_RUNLOCK();
1110 return (ifa);
1111}
1112
1113/*
1114 * Locate an interface based on the broadcast address.
1115 */
1116/* ARGSUSED */
1117struct ifaddr *
1118ifa_ifwithbroadaddr(struct sockaddr *addr)
1119{
1120 struct ifnet *ifp;
1121 struct ifaddr *ifa;
1122
1123 IFNET_RLOCK();
1124 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
1125 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1126 if (ifa->ifa_addr->sa_family != addr->sa_family)
1127 continue;
1128 if ((ifp->if_flags & IFF_BROADCAST) &&
1129 ifa->ifa_broadaddr &&
1130 ifa->ifa_broadaddr->sa_len != 0 &&
1131 sa_equal(ifa->ifa_broadaddr, addr))
1132 goto done;
1133 }
1134 ifa = NULL;
1135done:
1136 IFNET_RUNLOCK();
1137 return (ifa);
1138}
1139
1140/*
1141 * Locate the point to point interface with a given destination address.
1142 */
1143/*ARGSUSED*/
1144struct ifaddr *
1145ifa_ifwithdstaddr(struct sockaddr *addr)
1146{
1147 struct ifnet *ifp;
1148 struct ifaddr *ifa;
1149
1150 IFNET_RLOCK();
1151 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1152 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1153 continue;
1154 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1155 if (ifa->ifa_addr->sa_family != addr->sa_family)
1156 continue;
1157 if (ifa->ifa_dstaddr != NULL &&
1158 sa_equal(addr, ifa->ifa_dstaddr))
1159 goto done;
1160 }
1161 }
1162 ifa = NULL;
1163done:
1164 IFNET_RUNLOCK();
1165 return (ifa);
1166}
1167
1168/*
1169 * Find an interface on a specific network. If many, choice
1170 * is most specific found.
1171 */
1172struct ifaddr *
1173ifa_ifwithnet(struct sockaddr *addr)
1174{
1175 struct ifnet *ifp;
1176 struct ifaddr *ifa;
1177 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
1178 u_int af = addr->sa_family;
1179 char *addr_data = addr->sa_data, *cplim;
1180
1181 /*
1182 * AF_LINK addresses can be looked up directly by their index number,
1183 * so do that if we can.
1184 */
1185 if (af == AF_LINK) {
1186 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1187 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1188 return (ifaddr_byindex(sdl->sdl_index));
1189 }
1190
1191 /*
1192 * Scan though each interface, looking for ones that have
1193 * addresses in this address family.
1194 */
1195 IFNET_RLOCK();
1196 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1197 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1198 char *cp, *cp2, *cp3;
1199
1200 if (ifa->ifa_addr->sa_family != af)
1201next: continue;
1202 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
1203 /*
1204 * This is a bit broken as it doesn't
1205 * take into account that the remote end may
1206 * be a single node in the network we are
1207 * looking for.
1208 * The trouble is that we don't know the
1209 * netmask for the remote end.
1210 */
1211 if (ifa->ifa_dstaddr != NULL &&
1212 sa_equal(addr, ifa->ifa_dstaddr))
1213 goto done;
1214 } else {
1215 /*
1216 * if we have a special address handler,
1217 * then use it instead of the generic one.
1218 */
1219 if (ifa->ifa_claim_addr) {
1220 if ((*ifa->ifa_claim_addr)(ifa, addr))
1221 goto done;
1222 continue;
1223 }
1224
1225 /*
1226 * Scan all the bits in the ifa's address.
1227 * If a bit dissagrees with what we are
1228 * looking for, mask it with the netmask
1229 * to see if it really matters.
1230 * (A byte at a time)
1231 */
1232 if (ifa->ifa_netmask == 0)
1233 continue;
1234 cp = addr_data;
1235 cp2 = ifa->ifa_addr->sa_data;
1236 cp3 = ifa->ifa_netmask->sa_data;
1237 cplim = ifa->ifa_netmask->sa_len
1238 + (char *)ifa->ifa_netmask;
1239 while (cp3 < cplim)
1240 if ((*cp++ ^ *cp2++) & *cp3++)
1241 goto next; /* next address! */
1242 /*
1243 * If the netmask of what we just found
1244 * is more specific than what we had before
1245 * (if we had one) then remember the new one
1246 * before continuing to search
1247 * for an even better one.
1248 */
1249 if (ifa_maybe == 0 ||
1250 rn_refines((caddr_t)ifa->ifa_netmask,
1251 (caddr_t)ifa_maybe->ifa_netmask))
1252 ifa_maybe = ifa;
1253 }
1254 }
1255 }
1256 ifa = ifa_maybe;
1257done:
1258 IFNET_RUNLOCK();
1259 return (ifa);
1260}
1261
1262/*
1263 * Find an interface address specific to an interface best matching
1264 * a given address.
1265 */
1266struct ifaddr *
1267ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1268{
1269 struct ifaddr *ifa;
1270 char *cp, *cp2, *cp3;
1271 char *cplim;
1272 struct ifaddr *ifa_maybe = 0;
1273 u_int af = addr->sa_family;
1274
1275 if (af >= AF_MAX)
1276 return (0);
1277 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1278 if (ifa->ifa_addr->sa_family != af)
1279 continue;
1280 if (ifa_maybe == 0)
1281 ifa_maybe = ifa;
1282 if (ifa->ifa_netmask == 0) {
1283 if (sa_equal(addr, ifa->ifa_addr) ||
1284 (ifa->ifa_dstaddr &&
1285 sa_equal(addr, ifa->ifa_dstaddr)))
1286 goto done;
1287 continue;
1288 }
1289 if (ifp->if_flags & IFF_POINTOPOINT) {
1290 if (sa_equal(addr, ifa->ifa_dstaddr))
1291 goto done;
1292 } else {
1293 cp = addr->sa_data;
1294 cp2 = ifa->ifa_addr->sa_data;
1295 cp3 = ifa->ifa_netmask->sa_data;
1296 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1297 for (; cp3 < cplim; cp3++)
1298 if ((*cp++ ^ *cp2++) & *cp3)
1299 break;
1300 if (cp3 == cplim)
1301 goto done;
1302 }
1303 }
1304 ifa = ifa_maybe;
1305done:
1306 return (ifa);
1307}
1308
1309#include <net/route.h>
1310
1311/*
1312 * Default action when installing a route with a Link Level gateway.
1313 * Lookup an appropriate real ifa to point to.
1314 * This should be moved to /sys/net/link.c eventually.
1315 */
1316static void
1317link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1318{
1319 struct ifaddr *ifa, *oifa;
1320 struct sockaddr *dst;
1321 struct ifnet *ifp;
1322
1323 RT_LOCK_ASSERT(rt);
1324
1325 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1326 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1327 return;
1328 ifa = ifaof_ifpforaddr(dst, ifp);
1329 if (ifa) {
1330 IFAREF(ifa); /* XXX */
1331 oifa = rt->rt_ifa;
1332 rt->rt_ifa = ifa;
1333 IFAFREE(oifa);
1334 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1335 ifa->ifa_rtrequest(cmd, rt, info);
1336 }
1337}
1338
1339/*
1340 * Mark an interface down and notify protocols of
1341 * the transition.
1342 * NOTE: must be called at splnet or eqivalent.
1343 */
1344static void
1345if_unroute(struct ifnet *ifp, int flag, int fam)
1346{
1347 struct ifaddr *ifa;
1348
1349 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1350
1351 ifp->if_flags &= ~flag;
1352 getmicrotime(&ifp->if_lastchange);
1353 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1354 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1355 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1356 if_qflush(&ifp->if_snd);
1357#ifdef DEV_CARP
1358 if (ifp->if_carp)
1359 carp_carpdev_state(ifp->if_carp);
1360#endif
1361 rt_ifmsg(ifp);
1362}
1363
1364/*
1365 * Mark an interface up and notify protocols of
1366 * the transition.
1367 * NOTE: must be called at splnet or eqivalent.
1368 */
1369static void
1370if_route(struct ifnet *ifp, int flag, int fam)
1371{
1372 struct ifaddr *ifa;
1373
1374 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1375
1376 ifp->if_flags |= flag;
1377 getmicrotime(&ifp->if_lastchange);
1378 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1379 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1380 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1381#ifdef DEV_CARP
1382 if (ifp->if_carp)
1383 carp_carpdev_state(ifp->if_carp);
1384#endif
1385 rt_ifmsg(ifp);
1386#ifdef INET6
1387 in6_if_up(ifp);
1388#endif
1389}
1390
1391void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1392void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
1393
1394/*
1395 * Handle a change in the interface link state. To avoid LORs
1396 * between driver lock and upper layer locks, as well as possible
1397 * recursions, we post event to taskqueue, and all job
1398 * is done in static do_link_state_change().
1399 */
1400void
1401if_link_state_change(struct ifnet *ifp, int link_state)
1402{
1403 /* Return if state hasn't changed. */
1404 if (ifp->if_link_state == link_state)
1405 return;
1406
1407 ifp->if_link_state = link_state;
1408
1409 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1410}
1411
1412static void
1413do_link_state_change(void *arg, int pending)
1414{
1415 struct ifnet *ifp = (struct ifnet *)arg;
1416 int link_state = ifp->if_link_state;
1417 int link;
1418
1419 /* Notify that the link state has changed. */
1420 rt_ifmsg(ifp);
1421 if (link_state == LINK_STATE_UP)
1422 link = NOTE_LINKUP;
1423 else if (link_state == LINK_STATE_DOWN)
1424 link = NOTE_LINKDOWN;
1425 else
1426 link = NOTE_LINKINV;
1427 KNOTE_UNLOCKED(&ifp->if_klist, link);
1428 if (ifp->if_vlantrunk != NULL)
1429 (*vlan_link_state_p)(ifp, link);
1430
1431 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1432 IFP2AC(ifp)->ac_netgraph != NULL)
1433 (*ng_ether_link_state_p)(ifp, link_state);
1434#ifdef DEV_CARP
1435 if (ifp->if_carp)
1436 carp_carpdev_state(ifp->if_carp);
1437#endif
1438 if (ifp->if_bridge) {
1439 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1440 (*bstp_linkstate_p)(ifp, link_state);
1441 }
1442 if (ifp->if_lagg) {
1443 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1444 (*lagg_linkstate_p)(ifp, link_state);
1445 }
1446
1447 devctl_notify("IFNET", ifp->if_xname,
1448 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
1449 if (pending > 1)
1450 if_printf(ifp, "%d link states coalesced\n", pending);
1451 if (log_link_state_change)
1452 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1453 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1454}
1455
1456/*
1457 * Mark an interface down and notify protocols of
1458 * the transition.
1459 * NOTE: must be called at splnet or eqivalent.
1460 */
1461void
1462if_down(struct ifnet *ifp)
1463{
1464
1465 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1466}
1467
1468/*
1469 * Mark an interface up and notify protocols of
1470 * the transition.
1471 * NOTE: must be called at splnet or eqivalent.
1472 */
1473void
1474if_up(struct ifnet *ifp)
1475{
1476
1477 if_route(ifp, IFF_UP, AF_UNSPEC);
1478}
1479
1480/*
1481 * Flush an interface queue.
1482 */
1483static void
1484if_qflush(struct ifaltq *ifq)
1485{
1486 struct mbuf *m, *n;
1487
1488 IFQ_LOCK(ifq);
1489#ifdef ALTQ
1490 if (ALTQ_IS_ENABLED(ifq))
1491 ALTQ_PURGE(ifq);
1492#endif
1493 n = ifq->ifq_head;
1494 while ((m = n) != 0) {
1495 n = m->m_act;
1496 m_freem(m);
1497 }
1498 ifq->ifq_head = 0;
1499 ifq->ifq_tail = 0;
1500 ifq->ifq_len = 0;
1501 IFQ_UNLOCK(ifq);
1502}
1503
1504/*
1505 * Handle interface watchdog timer routines. Called
1506 * from softclock, we decrement timers (if set) and
1507 * call the appropriate interface routine on expiration.
1508 *
1509 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1510 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab
1511 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1512 */
1513static void
1514if_slowtimo(void *arg)
1515{
1516 struct ifnet *ifp;
1517 int s = splimp();
1518
1519 IFNET_RLOCK();
1520 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1521 if (ifp->if_timer == 0 || --ifp->if_timer)
1522 continue;
1523 if (ifp->if_watchdog)
1524 (*ifp->if_watchdog)(ifp);
1525 }
1526 IFNET_RUNLOCK();
1527 splx(s);
1528 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1529}
1530
1531/*
1532 * Map interface name to
1533 * interface structure pointer.
1534 */
1535struct ifnet *
1536ifunit(const char *name)
1537{
1538 struct ifnet *ifp;
1539
1540 IFNET_RLOCK();
1541 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1542 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1543 break;
1544 }
1545 IFNET_RUNLOCK();
1546 return (ifp);
1547}
1548
1549/*
1550 * Hardware specific interface ioctls.
1551 */
1552static int
1553ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1554{
1555 struct ifreq *ifr;
1556 struct ifstat *ifs;
1557 int error = 0;
1558 int new_flags, temp_flags;
1559 size_t namelen, onamelen;
1560 char new_name[IFNAMSIZ];
1561 struct ifaddr *ifa;
1562 struct sockaddr_dl *sdl;
1563
1564 ifr = (struct ifreq *)data;
1565 switch (cmd) {
1566 case SIOCGIFINDEX:
1567 ifr->ifr_index = ifp->if_index;
1568 break;
1569
1570 case SIOCGIFFLAGS:
1571 temp_flags = ifp->if_flags | ifp->if_drv_flags;
1572 ifr->ifr_flags = temp_flags & 0xffff;
1573 ifr->ifr_flagshigh = temp_flags >> 16;
1574 break;
1575
1576 case SIOCGIFCAP:
1577 ifr->ifr_reqcap = ifp->if_capabilities;
1578 ifr->ifr_curcap = ifp->if_capenable;
1579 break;
1580
1581#ifdef MAC
1582 case SIOCGIFMAC:
1583 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
1584 break;
1585#endif
1586
1587 case SIOCGIFMETRIC:
1588 ifr->ifr_metric = ifp->if_metric;
1589 break;
1590
1591 case SIOCGIFMTU:
1592 ifr->ifr_mtu = ifp->if_mtu;
1593 break;
1594
1595 case SIOCGIFPHYS:
1596 ifr->ifr_phys = ifp->if_physical;
1597 break;
1598
1599 case SIOCSIFFLAGS:
1600 error = priv_check(td, PRIV_NET_SETIFFLAGS);
1601 if (error)
1602 return (error);
1603 /*
1604 * Currently, no driver owned flags pass the IFF_CANTCHANGE
1605 * check, so we don't need special handling here yet.
1606 */
1607 new_flags = (ifr->ifr_flags & 0xffff) |
1608 (ifr->ifr_flagshigh << 16);
1609 if (ifp->if_flags & IFF_SMART) {
1610 /* Smart drivers twiddle their own routes */
1611 } else if (ifp->if_flags & IFF_UP &&
1612 (new_flags & IFF_UP) == 0) {
1613 int s = splimp();
1614 if_down(ifp);
1615 splx(s);
1616 } else if (new_flags & IFF_UP &&
1617 (ifp->if_flags & IFF_UP) == 0) {
1618 int s = splimp();
1619 if_up(ifp);
1620 splx(s);
1621 }
1622 /* See if permanently promiscuous mode bit is about to flip */
1623 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
1624 if (new_flags & IFF_PPROMISC)
1625 ifp->if_flags |= IFF_PROMISC;
1626 else if (ifp->if_pcount == 0)
1627 ifp->if_flags &= ~IFF_PROMISC;
1628 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
1629 ifp->if_xname,
1630 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
1631 }
1632 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1633 (new_flags &~ IFF_CANTCHANGE);
1634 if (ifp->if_ioctl) {
1635 IFF_LOCKGIANT(ifp);
1636 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1637 IFF_UNLOCKGIANT(ifp);
1638 }
1639 getmicrotime(&ifp->if_lastchange);
1640 break;
1641
1642 case SIOCSIFCAP:
1643 error = priv_check(td, PRIV_NET_SETIFCAP);
1644 if (error)
1645 return (error);
1646 if (ifp->if_ioctl == NULL)
1647 return (EOPNOTSUPP);
1648 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1649 return (EINVAL);
1650 IFF_LOCKGIANT(ifp);
1651 error = (*ifp->if_ioctl)(ifp, cmd, data);
1652 IFF_UNLOCKGIANT(ifp);
1653 if (error == 0)
1654 getmicrotime(&ifp->if_lastchange);
1655 break;
1656
1657#ifdef MAC
1658 case SIOCSIFMAC:
1659 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
1660 break;
1661#endif
1662
1663 case SIOCSIFNAME:
1664 error = priv_check(td, PRIV_NET_SETIFNAME);
1665 if (error)
1666 return (error);
1667 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1668 if (error != 0)
1669 return (error);
1670 if (new_name[0] == '\0')
1671 return (EINVAL);
1672 if (ifunit(new_name) != NULL)
1673 return (EEXIST);
1674
1675 /* Announce the departure of the interface. */
1676 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1677 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1678
1679 log(LOG_INFO, "%s: changing name to '%s'\n",
1680 ifp->if_xname, new_name);
1681
1682 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1683 ifa = ifp->if_addr;
1684 IFA_LOCK(ifa);
1685 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1686 namelen = strlen(new_name);
1687 onamelen = sdl->sdl_nlen;
1688 /*
1689 * Move the address if needed. This is safe because we
1690 * allocate space for a name of length IFNAMSIZ when we
1691 * create this in if_attach().
1692 */
1693 if (namelen != onamelen) {
1694 bcopy(sdl->sdl_data + onamelen,
1695 sdl->sdl_data + namelen, sdl->sdl_alen);
1696 }
1697 bcopy(new_name, sdl->sdl_data, namelen);
1698 sdl->sdl_nlen = namelen;
1699 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1700 bzero(sdl->sdl_data, onamelen);
1701 while (namelen != 0)
1702 sdl->sdl_data[--namelen] = 0xff;
1703 IFA_UNLOCK(ifa);
1704
1705 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1706 /* Announce the return of the interface. */
1707 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1708 break;
1709
1710 case SIOCSIFMETRIC:
1711 error = priv_check(td, PRIV_NET_SETIFMETRIC);
1712 if (error)
1713 return (error);
1714 ifp->if_metric = ifr->ifr_metric;
1715 getmicrotime(&ifp->if_lastchange);
1716 break;
1717
1718 case SIOCSIFPHYS:
1719 error = priv_check(td, PRIV_NET_SETIFPHYS);
1720 if (error)
1721 return (error);
1722 if (ifp->if_ioctl == NULL)
1723 return (EOPNOTSUPP);
1724 IFF_LOCKGIANT(ifp);
1725 error = (*ifp->if_ioctl)(ifp, cmd, data);
1726 IFF_UNLOCKGIANT(ifp);
1727 if (error == 0)
1728 getmicrotime(&ifp->if_lastchange);
1729 break;
1730
1731 case SIOCSIFMTU:
1732 {
1733 u_long oldmtu = ifp->if_mtu;
1734
1735 error = priv_check(td, PRIV_NET_SETIFMTU);
1736 if (error)
1737 return (error);
1738 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1739 return (EINVAL);
1740 if (ifp->if_ioctl == NULL)
1741 return (EOPNOTSUPP);
1742 IFF_LOCKGIANT(ifp);
1743 error = (*ifp->if_ioctl)(ifp, cmd, data);
1744 IFF_UNLOCKGIANT(ifp);
1745 if (error == 0) {
1746 getmicrotime(&ifp->if_lastchange);
1747 rt_ifmsg(ifp);
1748 }
1749 /*
1750 * If the link MTU changed, do network layer specific procedure.
1751 */
1752 if (ifp->if_mtu != oldmtu) {
1753#ifdef INET6
1754 nd6_setmtu(ifp);
1755#endif
1756 }
1757 break;
1758 }
1759
1760 case SIOCADDMULTI:
1761 case SIOCDELMULTI:
1762 if (cmd == SIOCADDMULTI)
1763 error = priv_check(td, PRIV_NET_ADDMULTI);
1764 else
1765 error = priv_check(td, PRIV_NET_DELMULTI);
1766 if (error)
1767 return (error);
1768
1769 /* Don't allow group membership on non-multicast interfaces. */
1770 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1771 return (EOPNOTSUPP);
1772
1773 /* Don't let users screw up protocols' entries. */
1774 if (ifr->ifr_addr.sa_family != AF_LINK)
1775 return (EINVAL);
1776
1777 if (cmd == SIOCADDMULTI) {
1778 struct ifmultiaddr *ifma;
1779
1780 /*
1781 * Userland is only permitted to join groups once
1782 * via the if_addmulti() KPI, because it cannot hold
1783 * struct ifmultiaddr * between calls. It may also
1784 * lose a race while we check if the membership
1785 * already exists.
1786 */
1787 IF_ADDR_LOCK(ifp);
1788 ifma = if_findmulti(ifp, &ifr->ifr_addr);
1789 IF_ADDR_UNLOCK(ifp);
1790 if (ifma != NULL)
1791 error = EADDRINUSE;
1792 else
1793 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1794 } else {
1795 error = if_delmulti(ifp, &ifr->ifr_addr);
1796 }
1797 if (error == 0)
1798 getmicrotime(&ifp->if_lastchange);
1799 break;
1800
1801 case SIOCSIFPHYADDR:
1802 case SIOCDIFPHYADDR:
1803#ifdef INET6
1804 case SIOCSIFPHYADDR_IN6:
1805#endif
1806 case SIOCSLIFPHYADDR:
1807 case SIOCSIFMEDIA:
1808 case SIOCSIFGENERIC:
1809 error = priv_check(td, PRIV_NET_HWIOCTL);
1810 if (error)
1811 return (error);
1812 if (ifp->if_ioctl == NULL)
1813 return (EOPNOTSUPP);
1814 IFF_LOCKGIANT(ifp);
1815 error = (*ifp->if_ioctl)(ifp, cmd, data);
1816 IFF_UNLOCKGIANT(ifp);
1817 if (error == 0)
1818 getmicrotime(&ifp->if_lastchange);
1819 break;
1820
1821 case SIOCGIFSTATUS:
1822 ifs = (struct ifstat *)data;
1823 ifs->ascii[0] = '\0';
1824
1825 case SIOCGIFPSRCADDR:
1826 case SIOCGIFPDSTADDR:
1827 case SIOCGLIFPHYADDR:
1828 case SIOCGIFMEDIA:
1829 case SIOCGIFGENERIC:
1830 if (ifp->if_ioctl == NULL)
1831 return (EOPNOTSUPP);
1832 IFF_LOCKGIANT(ifp);
1833 error = (*ifp->if_ioctl)(ifp, cmd, data);
1834 IFF_UNLOCKGIANT(ifp);
1835 break;
1836
1837 case SIOCSIFLLADDR:
1838 error = priv_check(td, PRIV_NET_SETLLADDR);
1839 if (error)
1840 return (error);
1841 error = if_setlladdr(ifp,
1842 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1843 break;
1844
1845 case SIOCAIFGROUP:
1846 {
1847 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1848
1849 error = priv_check(td, PRIV_NET_ADDIFGROUP);
1850 if (error)
1851 return (error);
1852 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
1853 return (error);
1854 break;
1855 }
1856
1857 case SIOCGIFGROUP:
1858 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
1859 return (error);
1860 break;
1861
1862 case SIOCDIFGROUP:
1863 {
1864 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1865
1866 error = priv_check(td, PRIV_NET_DELIFGROUP);
1867 if (error)
1868 return (error);
1869 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
1870 return (error);
1871 break;
1872 }
1873
1874 default:
1875 error = ENOIOCTL;
1876 break;
1877 }
1878 return (error);
1879}
1880
1881/*
1882 * Interface ioctls.
1883 */
1884int
1885ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1886{
1887 struct ifnet *ifp;
1888 struct ifreq *ifr;
1889 int error;
1890 int oif_flags;
1891
1892 switch (cmd) {
1893 case SIOCGIFCONF:
1894 case OSIOCGIFCONF:
1895#ifdef __amd64__
1896 case SIOCGIFCONF32:
1897#endif
1898 return (ifconf(cmd, data));
1899 }
1900 ifr = (struct ifreq *)data;
1901
1902 switch (cmd) {
1903 case SIOCIFCREATE:
1904 case SIOCIFCREATE2:
1905 error = priv_check(td, PRIV_NET_IFCREATE);
1906 if (error)
1907 return (error);
1908 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
1909 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
1910 case SIOCIFDESTROY:
1911 error = priv_check(td, PRIV_NET_IFDESTROY);
1912 if (error)
1913 return (error);
1914 return if_clone_destroy(ifr->ifr_name);
1915
1916 case SIOCIFGCLONERS:
1917 return (if_clone_list((struct if_clonereq *)data));
1918 case SIOCGIFGMEMB:
1919 return (if_getgroupmembers((struct ifgroupreq *)data));
1920 }
1921
1922 ifp = ifunit(ifr->ifr_name);
1923 if (ifp == 0)
1924 return (ENXIO);
1925
1926 error = ifhwioctl(cmd, ifp, data, td);
1927 if (error != ENOIOCTL)
1928 return (error);
1929
1930 oif_flags = ifp->if_flags;
1931 if (so->so_proto == 0)
1932 return (EOPNOTSUPP);
1933#ifndef COMPAT_43
1934 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1935 data,
1936 ifp, td));
1937#else
1938 {
1939 int ocmd = cmd;
1940
1941 switch (cmd) {
1942
1943 case SIOCSIFDSTADDR:
1944 case SIOCSIFADDR:
1945 case SIOCSIFBRDADDR:
1946 case SIOCSIFNETMASK:
1947#if BYTE_ORDER != BIG_ENDIAN
1948 if (ifr->ifr_addr.sa_family == 0 &&
1949 ifr->ifr_addr.sa_len < 16) {
1950 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1951 ifr->ifr_addr.sa_len = 16;
1952 }
1953#else
1954 if (ifr->ifr_addr.sa_len == 0)
1955 ifr->ifr_addr.sa_len = 16;
1956#endif
1957 break;
1958
1959 case OSIOCGIFADDR:
1960 cmd = SIOCGIFADDR;
1961 break;
1962
1963 case OSIOCGIFDSTADDR:
1964 cmd = SIOCGIFDSTADDR;
1965 break;
1966
1967 case OSIOCGIFBRDADDR:
1968 cmd = SIOCGIFBRDADDR;
1969 break;
1970
1971 case OSIOCGIFNETMASK:
1972 cmd = SIOCGIFNETMASK;
1973 }
1974 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
1975 cmd,
1976 data,
1977 ifp, td));
1978 switch (ocmd) {
1979
1980 case OSIOCGIFADDR:
1981 case OSIOCGIFDSTADDR:
1982 case OSIOCGIFBRDADDR:
1983 case OSIOCGIFNETMASK:
1984 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1985
1986 }
1987 }
1988#endif /* COMPAT_43 */
1989
1990 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1991#ifdef INET6
1992 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1993 if (ifp->if_flags & IFF_UP) {
1994 int s = splimp();
1995 in6_if_up(ifp);
1996 splx(s);
1997 }
1998#endif
1999 }
2000 return (error);
2001}
2002
2003/*
2004 * The code common to handling reference counted flags,
2005 * e.g., in ifpromisc() and if_allmulti().
2006 * The "pflag" argument can specify a permanent mode flag to check,
2007 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2008 *
2009 * Only to be used on stack-owned flags, not driver-owned flags.
2010 */
2011static int
2012if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2013{
2014 struct ifreq ifr;
2015 int error;
2016 int oldflags, oldcount;
2017
2018 /* Sanity checks to catch programming errors */
2019 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2020 ("%s: setting driver-owned flag %d", __func__, flag));
2021
2022 if (onswitch)
2023 KASSERT(*refcount >= 0,
2024 ("%s: increment negative refcount %d for flag %d",
2025 __func__, *refcount, flag));
2026 else
2027 KASSERT(*refcount > 0,
2028 ("%s: decrement non-positive refcount %d for flag %d",
2029 __func__, *refcount, flag));
2030
2031 /* In case this mode is permanent, just touch refcount */
2032 if (ifp->if_flags & pflag) {
2033 *refcount += onswitch ? 1 : -1;
2034 return (0);
2035 }
2036
2037 /* Save ifnet parameters for if_ioctl() may fail */
2038 oldcount = *refcount;
2039 oldflags = ifp->if_flags;
2040
2041 /*
2042 * See if we aren't the only and touching refcount is enough.
2043 * Actually toggle interface flag if we are the first or last.
2044 */
2045 if (onswitch) {
2046 if ((*refcount)++)
2047 return (0);
2048 ifp->if_flags |= flag;
2049 } else {
2050 if (--(*refcount))
2051 return (0);
2052 ifp->if_flags &= ~flag;
2053 }
2054
2055 /* Call down the driver since we've changed interface flags */
2056 if (ifp->if_ioctl == NULL) {
2057 error = EOPNOTSUPP;
2058 goto recover;
2059 }
2060 ifr.ifr_flags = ifp->if_flags & 0xffff;
2061 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2062 IFF_LOCKGIANT(ifp);
2063 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2064 IFF_UNLOCKGIANT(ifp);
2065 if (error)
2066 goto recover;
2067 /* Notify userland that interface flags have changed */
2068 rt_ifmsg(ifp);
2069 return (0);
2070
2071recover:
2072 /* Recover after driver error */
2073 *refcount = oldcount;
2074 ifp->if_flags = oldflags;
2075 return (error);
2076}
2077
2078/*
2079 * Set/clear promiscuous mode on interface ifp based on the truth value
2080 * of pswitch. The calls are reference counted so that only the first
2081 * "on" request actually has an effect, as does the final "off" request.
2082 * Results are undefined if the "off" and "on" requests are not matched.
2083 */
2084int
2085ifpromisc(struct ifnet *ifp, int pswitch)
2086{
2087 int error;
2088 int oldflags = ifp->if_flags;
2089
2090 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2091 &ifp->if_pcount, pswitch);
2092 /* If promiscuous mode status has changed, log a message */
2093 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2094 log(LOG_INFO, "%s: promiscuous mode %s\n",
2095 ifp->if_xname,
2096 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2097 return (error);
2098}
2099
2100/*
2101 * Return interface configuration
2102 * of system. List may be used
2103 * in later ioctl's (above) to get
2104 * other information.
2105 */
2106/*ARGSUSED*/
2107static int
2108ifconf(u_long cmd, caddr_t data)
2109{
2110 struct ifconf *ifc = (struct ifconf *)data;
2111#ifdef __amd64__
2112 struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2113 struct ifconf ifc_swab;
2114#endif
2115 struct ifnet *ifp;
2116 struct ifaddr *ifa;
2117 struct ifreq ifr;
2118 struct sbuf *sb;
2119 int error, full = 0, valid_len, max_len;
2120
2121#ifdef __amd64__
2122 if (cmd == SIOCGIFCONF32) {
2123 ifc_swab.ifc_len = ifc32->ifc_len;
2124 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2125 ifc = &ifc_swab;
2126 }
2127#endif
2128 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2129 max_len = MAXPHYS - 1;
2130
2131 /* Prevent hostile input from being able to crash the system */
2132 if (ifc->ifc_len <= 0)
2133 return (EINVAL);
2134
2135again:
2136 if (ifc->ifc_len <= max_len) {
2137 max_len = ifc->ifc_len;
2138 full = 1;
2139 }
2140 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2141 max_len = 0;
2142 valid_len = 0;
2143
2144 IFNET_RLOCK(); /* could sleep XXX */
2145 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2146 int addrs;
2147
2148 /*
2149 * Zero the ifr_name buffer to make sure we don't
2150 * disclose the contents of the stack.
2151 */
2152 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2153
2154 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2155 >= sizeof(ifr.ifr_name)) {
2156 sbuf_delete(sb);
2157 IFNET_RUNLOCK();
2158 return (ENAMETOOLONG);
2159 }
2160
2161 addrs = 0;
2162 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2163 struct sockaddr *sa = ifa->ifa_addr;
2164
2165 if (jailed(curthread->td_ucred) &&
2166 prison_if(curthread->td_ucred, sa))
2167 continue;
2168 addrs++;
2169#ifdef COMPAT_43
2170 if (cmd == OSIOCGIFCONF) {
2171 struct osockaddr *osa =
2172 (struct osockaddr *)&ifr.ifr_addr;
2173 ifr.ifr_addr = *sa;
2174 osa->sa_family = sa->sa_family;
2175 sbuf_bcat(sb, &ifr, sizeof(ifr));
2176 max_len += sizeof(ifr);
2177 } else
2178#endif
2179 if (sa->sa_len <= sizeof(*sa)) {
2180 ifr.ifr_addr = *sa;
2181 sbuf_bcat(sb, &ifr, sizeof(ifr));
2182 max_len += sizeof(ifr);
2183 } else {
2184 sbuf_bcat(sb, &ifr,
2185 offsetof(struct ifreq, ifr_addr));
2186 max_len += offsetof(struct ifreq, ifr_addr);
2187 sbuf_bcat(sb, sa, sa->sa_len);
2188 max_len += sa->sa_len;
2189 }
2190
2191 if (!sbuf_overflowed(sb))
2192 valid_len = sbuf_len(sb);
2193 }
2194 if (addrs == 0) {
2195 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2196 sbuf_bcat(sb, &ifr, sizeof(ifr));
2197 max_len += sizeof(ifr);
2198
2199 if (!sbuf_overflowed(sb))
2200 valid_len = sbuf_len(sb);
2201 }
2202 }
2203 IFNET_RUNLOCK();
2204
2205 /*
2206 * If we didn't allocate enough space (uncommon), try again. If
2207 * we have already allocated as much space as we are allowed,
2208 * return what we've got.
2209 */
2210 if (valid_len != max_len && !full) {
2211 sbuf_delete(sb);
2212 goto again;
2213 }
2214
2215 ifc->ifc_len = valid_len;
2216#ifdef __amd64__
2217 if (cmd == SIOCGIFCONF32)
2218 ifc32->ifc_len = valid_len;
2219#endif
2220 sbuf_finish(sb);
2221 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2222 sbuf_delete(sb);
2223 return (error);
2224}
2225
2226/*
2227 * Just like ifpromisc(), but for all-multicast-reception mode.
2228 */
2229int
2230if_allmulti(struct ifnet *ifp, int onswitch)
2231{
2232
2233 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2234}
2235
2236struct ifmultiaddr *
2237if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2238{
2239 struct ifmultiaddr *ifma;
2240
2241 IF_ADDR_LOCK_ASSERT(ifp);
2242
2243 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2244 if (sa->sa_family == AF_LINK) {
2245 if (sa_dl_equal(ifma->ifma_addr, sa))
2246 break;
2247 } else {
2248 if (sa_equal(ifma->ifma_addr, sa))
2249 break;
2250 }
2251 }
2252
2253 return ifma;
2254}
2255
2256/*
2257 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2258 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2259 * the ifnet multicast address list here, so the caller must do that and
2260 * other setup work (such as notifying the device driver). The reference
2261 * count is initialized to 1.
2262 */
2263static struct ifmultiaddr *
2264if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2265 int mflags)
2266{
2267 struct ifmultiaddr *ifma;
2268 struct sockaddr *dupsa;
2269
2270 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
2271 M_ZERO);
2272 if (ifma == NULL)
2273 return (NULL);
2274
2275 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
2276 if (dupsa == NULL) {
2277 FREE(ifma, M_IFMADDR);
2278 return (NULL);
2279 }
2280 bcopy(sa, dupsa, sa->sa_len);
2281 ifma->ifma_addr = dupsa;
2282
2283 ifma->ifma_ifp = ifp;
2284 ifma->ifma_refcount = 1;
2285 ifma->ifma_protospec = NULL;
2286
2287 if (llsa == NULL) {
2288 ifma->ifma_lladdr = NULL;
2289 return (ifma);
2290 }
2291
2292 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
2293 if (dupsa == NULL) {
2294 FREE(ifma->ifma_addr, M_IFMADDR);
2295 FREE(ifma, M_IFMADDR);
2296 return (NULL);
2297 }
2298 bcopy(llsa, dupsa, llsa->sa_len);
2299 ifma->ifma_lladdr = dupsa;
2300
2301 return (ifma);
2302}
2303
2304/*
2305 * if_freemulti: free ifmultiaddr structure and possibly attached related
2306 * addresses. The caller is responsible for implementing reference
2307 * counting, notifying the driver, handling routing messages, and releasing
2308 * any dependent link layer state.
2309 */
2310static void
2311if_freemulti(struct ifmultiaddr *ifma)
2312{
2313
2314 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2315 ifma->ifma_refcount));
2316 KASSERT(ifma->ifma_protospec == NULL,
2317 ("if_freemulti: protospec not NULL"));
2318
2319 if (ifma->ifma_lladdr != NULL)
2320 FREE(ifma->ifma_lladdr, M_IFMADDR);
2321 FREE(ifma->ifma_addr, M_IFMADDR);
2322 FREE(ifma, M_IFMADDR);
2323}
2324
2325/*
2326 * Register an additional multicast address with a network interface.
2327 *
2328 * - If the address is already present, bump the reference count on the
2329 * address and return.
2330 * - If the address is not link-layer, look up a link layer address.
2331 * - Allocate address structures for one or both addresses, and attach to the
2332 * multicast address list on the interface. If automatically adding a link
2333 * layer address, the protocol address will own a reference to the link
2334 * layer address, to be freed when it is freed.
2335 * - Notify the network device driver of an addition to the multicast address
2336 * list.
2337 *
2338 * 'sa' points to caller-owned memory with the desired multicast address.
2339 *
2340 * 'retifma' will be used to return a pointer to the resulting multicast
2341 * address reference, if desired.
2342 */
2343int
2344if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2345 struct ifmultiaddr **retifma)
2346{
2347 struct ifmultiaddr *ifma, *ll_ifma;
2348 struct sockaddr *llsa;
2349 int error;
2350
2351 /*
2352 * If the address is already present, return a new reference to it;
2353 * otherwise, allocate storage and set up a new address.
2354 */
2355 IF_ADDR_LOCK(ifp);
2356 ifma = if_findmulti(ifp, sa);
2357 if (ifma != NULL) {
2358 ifma->ifma_refcount++;
2359 if (retifma != NULL)
2360 *retifma = ifma;
2361 IF_ADDR_UNLOCK(ifp);
2362 return (0);
2363 }
2364
2365 /*
2366 * The address isn't already present; resolve the protocol address
2367 * into a link layer address, and then look that up, bump its
2368 * refcount or allocate an ifma for that also. If 'llsa' was
2369 * returned, we will need to free it later.
2370 */
2371 llsa = NULL;
2372 ll_ifma = NULL;
2373 if (ifp->if_resolvemulti != NULL) {
2374 error = ifp->if_resolvemulti(ifp, &llsa, sa);
2375 if (error)
2376 goto unlock_out;
2377 }
2378
2379 /*
2380 * Allocate the new address. Don't hook it up yet, as we may also
2381 * need to allocate a link layer multicast address.
2382 */
2383 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2384 if (ifma == NULL) {
2385 error = ENOMEM;
2386 goto free_llsa_out;
2387 }
2388
2389 /*
2390 * If a link layer address is found, we'll need to see if it's
2391 * already present in the address list, or allocate is as well.
2392 * When this block finishes, the link layer address will be on the
2393 * list.
2394 */
2395 if (llsa != NULL) {
2396 ll_ifma = if_findmulti(ifp, llsa);
2397 if (ll_ifma == NULL) {
2398 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2399 if (ll_ifma == NULL) {
2400 --ifma->ifma_refcount;
2401 if_freemulti(ifma);
2402 error = ENOMEM;
2403 goto free_llsa_out;
2404 }
2405 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2406 ifma_link);
2407 } else
2408 ll_ifma->ifma_refcount++;
2409 ifma->ifma_llifma = ll_ifma;
2410 }
2411
2412 /*
2413 * We now have a new multicast address, ifma, and possibly a new or
2414 * referenced link layer address. Add the primary address to the
2415 * ifnet address list.
2416 */
2417 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2418
2419 if (retifma != NULL)
2420 *retifma = ifma;
2421
2422 /*
2423 * Must generate the message while holding the lock so that 'ifma'
2424 * pointer is still valid.
2425 */
2426 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2427 IF_ADDR_UNLOCK(ifp);
2428
2429 /*
2430 * We are certain we have added something, so call down to the
2431 * interface to let them know about it.
2432 */
2433 if (ifp->if_ioctl != NULL) {
2434 IFF_LOCKGIANT(ifp);
2435 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2436 IFF_UNLOCKGIANT(ifp);
2437 }
2438
2439 if (llsa != NULL)
2440 FREE(llsa, M_IFMADDR);
2441
2442 return (0);
2443
2444free_llsa_out:
2445 if (llsa != NULL)
2446 FREE(llsa, M_IFMADDR);
2447
2448unlock_out:
2449 IF_ADDR_UNLOCK(ifp);
2450 return (error);
2451}
2452
2453/*
2454 * Delete a multicast group membership by network-layer group address.
2455 *
2456 * Returns ENOENT if the entry could not be found. If ifp no longer
2457 * exists, results are undefined. This entry point should only be used
2458 * from subsystems which do appropriate locking to hold ifp for the
2459 * duration of the call.
2460 * Network-layer protocol domains must use if_delmulti_ifma().
2461 */
2462int
2463if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2464{
2465 struct ifmultiaddr *ifma;
2466 int lastref;
2467#ifdef INVARIANTS
2468 struct ifnet *oifp;
2469
2470 IFNET_RLOCK();
2471 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2472 if (ifp == oifp)
2473 break;
2474 if (ifp != oifp)
2475 ifp = NULL;
2476 IFNET_RUNLOCK();
2477
2478 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
2479#endif
2480 if (ifp == NULL)
2481 return (ENOENT);
2482
2483 IF_ADDR_LOCK(ifp);
2484 lastref = 0;
2485 ifma = if_findmulti(ifp, sa);
2486 if (ifma != NULL)
2487 lastref = if_delmulti_locked(ifp, ifma, 0);
2488 IF_ADDR_UNLOCK(ifp);
2489
2490 if (ifma == NULL)
2491 return (ENOENT);
2492
2493 if (lastref && ifp->if_ioctl != NULL) {
2494 IFF_LOCKGIANT(ifp);
2495 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2496 IFF_UNLOCKGIANT(ifp);
2497 }
2498
2499 return (0);
2500}
2501
2502/*
2503 * Delete a multicast group membership by group membership pointer.
2504 * Network-layer protocol domains must use this routine.
2505 *
2506 * It is safe to call this routine if the ifp disappeared. Callers should
2507 * hold IFF_LOCKGIANT() to avoid a LOR in case the hardware needs to be
2508 * reconfigured.
2509 */
2510void
2511if_delmulti_ifma(struct ifmultiaddr *ifma)
2512{
2513 struct ifnet *ifp;
2514 int lastref;
2515
2516 ifp = ifma->ifma_ifp;
2517#ifdef DIAGNOSTIC
2518 if (ifp == NULL) {
2519 printf("%s: ifma_ifp seems to be detached\n", __func__);
2520 } else {
2521 struct ifnet *oifp;
2522
2523 IFNET_RLOCK();
2524 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2525 if (ifp == oifp)
2526 break;
2527 if (ifp != oifp) {
2528 printf("%s: ifnet %p disappeared\n", __func__, ifp);
2529 ifp = NULL;
2530 }
2531 IFNET_RUNLOCK();
2532 }
2533#endif
2534 /*
2535 * If and only if the ifnet instance exists: Acquire the address lock.
2536 */
2537 if (ifp != NULL)
2538 IF_ADDR_LOCK(ifp);
2539
2540 lastref = if_delmulti_locked(ifp, ifma, 0);
2541
2542 if (ifp != NULL) {
2543 /*
2544 * If and only if the ifnet instance exists:
2545 * Release the address lock.
2546 * If the group was left: update the hardware hash filter.
2547 */
2548 IF_ADDR_UNLOCK(ifp);
2549 if (lastref && ifp->if_ioctl != NULL) {
2550 IFF_LOCKGIANT(ifp);
2551 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2552 IFF_UNLOCKGIANT(ifp);
2553 }
2554 }
2555}
2556
2557/*
2558 * Perform deletion of network-layer and/or link-layer multicast address.
2559 *
2560 * Return 0 if the reference count was decremented.
2561 * Return 1 if the final reference was released, indicating that the
2562 * hardware hash filter should be reprogrammed.
2563 */
2564static int
2565if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
2566{
2567 struct ifmultiaddr *ll_ifma;
2568
2569 if (ifp != NULL && ifma->ifma_ifp != NULL) {
2570 KASSERT(ifma->ifma_ifp == ifp,
2571 ("%s: inconsistent ifp %p", __func__, ifp));
2572 IF_ADDR_LOCK_ASSERT(ifp);
2573 }
2574
2575 ifp = ifma->ifma_ifp;
2576
2577 /*
2578 * If the ifnet is detaching, null out references to ifnet,
2579 * so that upper protocol layers will notice, and not attempt
2580 * to obtain locks for an ifnet which no longer exists. The
2581 * routing socket announcement must happen before the ifnet
2582 * instance is detached from the system.
2583 */
2584 if (detaching) {
2585#ifdef DIAGNOSTIC
2586 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
2587#endif
2588 /*
2589 * ifp may already be nulled out if we are being reentered
2590 * to delete the ll_ifma.
2591 */
2592 if (ifp != NULL) {
2593 rt_newmaddrmsg(RTM_DELMADDR, ifma);
2594 ifma->ifma_ifp = NULL;
2595 }
2596 }
2597
2598 if (--ifma->ifma_refcount > 0)
2599 return 0;
2600
2601 /*
2602 * If this ifma is a network-layer ifma, a link-layer ifma may
2603 * have been associated with it. Release it first if so.
2604 */
2605 ll_ifma = ifma->ifma_llifma;
2606 if (ll_ifma != NULL) {
2607 KASSERT(ifma->ifma_lladdr != NULL,
2608 ("%s: llifma w/o lladdr", __func__));
2609 if (detaching)
2610 ll_ifma->ifma_ifp = NULL; /* XXX */
2611 if (--ll_ifma->ifma_refcount == 0) {
2612 if (ifp != NULL) {
2613 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
2614 ifma_link);
2615 }
2616 if_freemulti(ll_ifma);
2617 }
2618 }
2619
2620 if (ifp != NULL)
2621 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
2622
2623 if_freemulti(ifma);
2624
2625 /*
2626 * The last reference to this instance of struct ifmultiaddr
2627 * was released; the hardware should be notified of this change.
2628 */
2629 return 1;
2630}
2631
2632/*
2633 * Set the link layer address on an interface.
2634 *
2635 * At this time we only support certain types of interfaces,
2636 * and we don't allow the length of the address to change.
2637 */
2638int
2639if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
2640{
2641 struct sockaddr_dl *sdl;
2642 struct ifaddr *ifa;
2643 struct ifreq ifr;
2644
2645 ifa = ifp->if_addr;
2646 if (ifa == NULL)
2647 return (EINVAL);
2648 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2649 if (sdl == NULL)
2650 return (EINVAL);
2651 if (len != sdl->sdl_alen) /* don't allow length to change */
2652 return (EINVAL);
2653 switch (ifp->if_type) {
2654 case IFT_ETHER:
2655 case IFT_FDDI:
2656 case IFT_XETHER:
2657 case IFT_ISO88025:
2658 case IFT_L2VLAN:
2659 case IFT_BRIDGE:
2660 case IFT_ARCNET:
2661 case IFT_IEEE8023ADLAG:
2662 bcopy(lladdr, LLADDR(sdl), len);
2663 break;
2664 default:
2665 return (ENODEV);
2666 }
2667 /*
2668 * If the interface is already up, we need
2669 * to re-init it in order to reprogram its
2670 * address filter.
2671 */
2672 if ((ifp->if_flags & IFF_UP) != 0) {
2673 if (ifp->if_ioctl) {
2674 IFF_LOCKGIANT(ifp);
2675 ifp->if_flags &= ~IFF_UP;
2676 ifr.ifr_flags = ifp->if_flags & 0xffff;
2677 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2678 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2679 ifp->if_flags |= IFF_UP;
2680 ifr.ifr_flags = ifp->if_flags & 0xffff;
2681 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2682 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2683 IFF_UNLOCKGIANT(ifp);
2684 }
2685#ifdef INET
2686 /*
2687 * Also send gratuitous ARPs to notify other nodes about
2688 * the address change.
2689 */
2690 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2691 if (ifa->ifa_addr->sa_family == AF_INET)
2692 arp_ifinit(ifp, ifa);
2693 }
2694#endif
2695 }
2696 return (0);
2697}
2698
2699/*
2700 * The name argument must be a pointer to storage which will last as
2701 * long as the interface does. For physical devices, the result of
2702 * device_get_name(dev) is a good choice and for pseudo-devices a
2703 * static string works well.
2704 */
2705void
2706if_initname(struct ifnet *ifp, const char *name, int unit)
2707{
2708 ifp->if_dname = name;
2709 ifp->if_dunit = unit;
2710 if (unit != IF_DUNIT_NONE)
2711 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
2712 else
2713 strlcpy(ifp->if_xname, name, IFNAMSIZ);
2714}
2715
2716int
2717if_printf(struct ifnet *ifp, const char * fmt, ...)
2718{
2719 va_list ap;
2720 int retval;
2721
2722 retval = printf("%s: ", ifp->if_xname);
2723 va_start(ap, fmt);
2724 retval += vprintf(fmt, ap);
2725 va_end(ap);
2726 return (retval);
2727}
2728
2729/*
2730 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
2731 * be called without Giant. However, we often can't acquire the Giant lock
2732 * at those points; instead, we run it via a task queue that holds Giant via
2733 * if_start_deferred.
2734 *
2735 * XXXRW: We need to make sure that the ifnet isn't fully detached until any
2736 * outstanding if_start_deferred() tasks that will run after the free. This
2737 * probably means waiting in if_detach().
2738 */
2739void
2740if_start(struct ifnet *ifp)
2741{
2742
2743 if (ifp->if_flags & IFF_NEEDSGIANT) {
2744 if (mtx_owned(&Giant))
2745 (*(ifp)->if_start)(ifp);
2746 else
2747 taskqueue_enqueue(taskqueue_swi_giant,
2748 &ifp->if_starttask);
2749 } else
2750 (*(ifp)->if_start)(ifp);
2751}
2752
2753static void
2754if_start_deferred(void *context, int pending)
2755{
2756 struct ifnet *ifp;
2757
2758 GIANT_REQUIRED;
2759
2760 ifp = context;
2761 (ifp->if_start)(ifp);
2762}
2763
2764int
2765if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
2766{
2767 int active = 0;
2768
2769 IF_LOCK(ifq);
2770 if (_IF_QFULL(ifq)) {
2771 _IF_DROP(ifq);
2772 IF_UNLOCK(ifq);
2773 m_freem(m);
2774 return (0);
2775 }
2776 if (ifp != NULL) {
2777 ifp->if_obytes += m->m_pkthdr.len + adjust;
2778 if (m->m_flags & (M_BCAST|M_MCAST))
2779 ifp->if_omcasts++;
2780 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
2781 }
2782 _IF_ENQUEUE(ifq, m);
2783 IF_UNLOCK(ifq);
2784 if (ifp != NULL && !active)
2785 if_start(ifp);
2786 return (1);
2787}
2788
2789void
2790if_register_com_alloc(u_char type,
2791 if_com_alloc_t *a, if_com_free_t *f)
2792{
2793
2794 KASSERT(if_com_alloc[type] == NULL,
2795 ("if_register_com_alloc: %d already registered", type));
2796 KASSERT(if_com_free[type] == NULL,
2797 ("if_register_com_alloc: %d free already registered", type));
2798
2799 if_com_alloc[type] = a;
2800 if_com_free[type] = f;
2801}
2802
2803void
2804if_deregister_com_alloc(u_char type)
2805{
2806
2807 KASSERT(if_com_alloc[type] != NULL,
2808 ("if_deregister_com_alloc: %d not registered", type));
2809 KASSERT(if_com_free[type] != NULL,
2810 ("if_deregister_com_alloc: %d free not registered", type));
2811 if_com_alloc[type] = NULL;
2812 if_com_free[type] = NULL;
2813}
518 net_cdevsw.d_name, ifp->if_xname));
519 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
520 net_cdevsw.d_name, ifp->if_index);
521
522 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
523
524 /*
525 * create a Link Level name for this device
526 */
527 namelen = strlen(ifp->if_xname);
528 /*
529 * Always save enough space for any possiable name so we can do
530 * a rename in place later.
531 */
532 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
533 socksize = masklen + ifp->if_addrlen;
534 if (socksize < sizeof(*sdl))
535 socksize = sizeof(*sdl);
536 socksize = roundup2(socksize, sizeof(long));
537 ifasize = sizeof(*ifa) + 2 * socksize;
538 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
539 IFA_LOCK_INIT(ifa);
540 sdl = (struct sockaddr_dl *)(ifa + 1);
541 sdl->sdl_len = socksize;
542 sdl->sdl_family = AF_LINK;
543 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
544 sdl->sdl_nlen = namelen;
545 sdl->sdl_index = ifp->if_index;
546 sdl->sdl_type = ifp->if_type;
547 ifp->if_addr = ifa;
548 ifa->ifa_ifp = ifp;
549 ifa->ifa_rtrequest = link_rtrequest;
550 ifa->ifa_addr = (struct sockaddr *)sdl;
551 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
552 ifa->ifa_netmask = (struct sockaddr *)sdl;
553 sdl->sdl_len = masklen;
554 while (namelen != 0)
555 sdl->sdl_data[--namelen] = 0xff;
556 ifa->ifa_refcnt = 1;
557 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
558 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
559
560 /*
561 * XXX: why do we warn about this? We're correcting it and most
562 * drivers just set the value the way we do.
563 */
564 if (ifp->if_snd.ifq_maxlen == 0) {
565 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
566 ifp->if_snd.ifq_maxlen = ifqmaxlen;
567 }
568 ifp->if_snd.altq_type = 0;
569 ifp->if_snd.altq_disc = NULL;
570 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
571 ifp->if_snd.altq_tbr = NULL;
572 ifp->if_snd.altq_ifp = ifp;
573
574 IFNET_WLOCK();
575 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
576 IFNET_WUNLOCK();
577
578 if (domain_init_status >= 2)
579 if_attachdomain1(ifp);
580
581 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
582 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
583
584 /* Announce the interface. */
585 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
586
587 if (ifp->if_watchdog != NULL)
588 if_printf(ifp,
589 "WARNING: using obsoleted if_watchdog interface\n");
590 if (ifp->if_flags & IFF_NEEDSGIANT)
591 if_printf(ifp,
592 "WARNING: using obsoleted IFF_NEEDSGIANT flag\n");
593}
594
595static void
596if_attachdomain(void *dummy)
597{
598 struct ifnet *ifp;
599 int s;
600
601 s = splnet();
602 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
603 if_attachdomain1(ifp);
604 splx(s);
605}
606SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
607 if_attachdomain, NULL);
608
609static void
610if_attachdomain1(struct ifnet *ifp)
611{
612 struct domain *dp;
613 int s;
614
615 s = splnet();
616
617 /*
618 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
619 * cannot lock ifp->if_afdata initialization, entirely.
620 */
621 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
622 splx(s);
623 return;
624 }
625 if (ifp->if_afdata_initialized >= domain_init_status) {
626 IF_AFDATA_UNLOCK(ifp);
627 splx(s);
628 printf("if_attachdomain called more than once on %s\n",
629 ifp->if_xname);
630 return;
631 }
632 ifp->if_afdata_initialized = domain_init_status;
633 IF_AFDATA_UNLOCK(ifp);
634
635 /* address family dependent data region */
636 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
637 for (dp = domains; dp; dp = dp->dom_next) {
638 if (dp->dom_ifattach)
639 ifp->if_afdata[dp->dom_family] =
640 (*dp->dom_ifattach)(ifp);
641 }
642
643 splx(s);
644}
645
646/*
647 * Remove any unicast or broadcast network addresses from an interface.
648 */
649void
650if_purgeaddrs(struct ifnet *ifp)
651{
652 struct ifaddr *ifa, *next;
653
654 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
655 if (ifa->ifa_addr->sa_family == AF_LINK)
656 continue;
657#ifdef INET
658 /* XXX: Ugly!! ad hoc just for INET */
659 if (ifa->ifa_addr->sa_family == AF_INET) {
660 struct ifaliasreq ifr;
661
662 bzero(&ifr, sizeof(ifr));
663 ifr.ifra_addr = *ifa->ifa_addr;
664 if (ifa->ifa_dstaddr)
665 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
666 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
667 NULL) == 0)
668 continue;
669 }
670#endif /* INET */
671#ifdef INET6
672 if (ifa->ifa_addr->sa_family == AF_INET6) {
673 in6_purgeaddr(ifa);
674 /* ifp_addrhead is already updated */
675 continue;
676 }
677#endif /* INET6 */
678 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
679 IFAFREE(ifa);
680 }
681}
682
683/*
684 * Remove any multicast network addresses from an interface.
685 */
686void
687if_purgemaddrs(struct ifnet *ifp)
688{
689 struct ifmultiaddr *ifma;
690 struct ifmultiaddr *next;
691
692 IF_ADDR_LOCK(ifp);
693 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
694 if_delmulti_locked(ifp, ifma, 1);
695 IF_ADDR_UNLOCK(ifp);
696}
697
698/*
699 * Detach an interface, removing it from the
700 * list of "active" interfaces.
701 *
702 * XXXRW: There are some significant questions about event ordering, and
703 * how to prevent things from starting to use the interface during detach.
704 */
705void
706if_detach(struct ifnet *ifp)
707{
708 struct ifaddr *ifa;
709 struct radix_node_head *rnh;
710 int s;
711 int i;
712 struct domain *dp;
713 struct ifnet *iter;
714 int found = 0;
715
716 IFNET_WLOCK();
717 TAILQ_FOREACH(iter, &V_ifnet, if_link)
718 if (iter == ifp) {
719 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
720 found = 1;
721 break;
722 }
723 IFNET_WUNLOCK();
724 if (!found)
725 return;
726
727 /*
728 * Remove/wait for pending events.
729 */
730 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
731
732 /*
733 * Remove routes and flush queues.
734 */
735 s = splnet();
736 if_down(ifp);
737#ifdef ALTQ
738 if (ALTQ_IS_ENABLED(&ifp->if_snd))
739 altq_disable(&ifp->if_snd);
740 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
741 altq_detach(&ifp->if_snd);
742#endif
743
744 if_purgeaddrs(ifp);
745
746#ifdef INET
747 in_ifdetach(ifp);
748#endif
749
750#ifdef INET6
751 /*
752 * Remove all IPv6 kernel structs related to ifp. This should be done
753 * before removing routing entries below, since IPv6 interface direct
754 * routes are expected to be removed by the IPv6-specific kernel API.
755 * Otherwise, the kernel will detect some inconsistency and bark it.
756 */
757 in6_ifdetach(ifp);
758#endif
759 if_purgemaddrs(ifp);
760
761 /*
762 * Remove link ifaddr pointer and maybe decrement if_index.
763 * Clean up all addresses.
764 */
765 ifp->if_addr = NULL;
766 destroy_dev(ifdev_byindex(ifp->if_index));
767 ifdev_setbyindex(ifp->if_index, NULL);
768
769 /* We can now free link ifaddr. */
770 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
771 ifa = TAILQ_FIRST(&ifp->if_addrhead);
772 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
773 IFAFREE(ifa);
774 }
775
776 /*
777 * Delete all remaining routes using this interface
778 * Unfortuneatly the only way to do this is to slog through
779 * the entire routing table looking for routes which point
780 * to this interface...oh well...
781 */
782 for (i = 1; i <= AF_MAX; i++) {
783 int j;
784 for (j = 0; j < rt_numfibs; j++) {
785 if ((rnh = V_rt_tables[j][i]) == NULL)
786 continue;
787 RADIX_NODE_HEAD_LOCK(rnh);
788 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
789 RADIX_NODE_HEAD_UNLOCK(rnh);
790 }
791 }
792
793 /* Announce that the interface is gone. */
794 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
795 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
796 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
797
798 IF_AFDATA_LOCK(ifp);
799 for (dp = domains; dp; dp = dp->dom_next) {
800 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
801 (*dp->dom_ifdetach)(ifp,
802 ifp->if_afdata[dp->dom_family]);
803 }
804 IF_AFDATA_UNLOCK(ifp);
805
806#ifdef MAC
807 mac_ifnet_destroy(ifp);
808#endif /* MAC */
809 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
810 knlist_clear(&ifp->if_klist, 0);
811 knlist_destroy(&ifp->if_klist);
812 mtx_destroy(&ifp->if_snd.ifq_mtx);
813 IF_AFDATA_DESTROY(ifp);
814 splx(s);
815}
816
817/*
818 * Add a group to an interface
819 */
820int
821if_addgroup(struct ifnet *ifp, const char *groupname)
822{
823 struct ifg_list *ifgl;
824 struct ifg_group *ifg = NULL;
825 struct ifg_member *ifgm;
826
827 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
828 groupname[strlen(groupname) - 1] <= '9')
829 return (EINVAL);
830
831 IFNET_WLOCK();
832 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
833 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
834 IFNET_WUNLOCK();
835 return (EEXIST);
836 }
837
838 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
839 M_NOWAIT)) == NULL) {
840 IFNET_WUNLOCK();
841 return (ENOMEM);
842 }
843
844 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
845 M_TEMP, M_NOWAIT)) == NULL) {
846 free(ifgl, M_TEMP);
847 IFNET_WUNLOCK();
848 return (ENOMEM);
849 }
850
851 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
852 if (!strcmp(ifg->ifg_group, groupname))
853 break;
854
855 if (ifg == NULL) {
856 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
857 M_TEMP, M_NOWAIT)) == NULL) {
858 free(ifgl, M_TEMP);
859 free(ifgm, M_TEMP);
860 IFNET_WUNLOCK();
861 return (ENOMEM);
862 }
863 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
864 ifg->ifg_refcnt = 0;
865 TAILQ_INIT(&ifg->ifg_members);
866 EVENTHANDLER_INVOKE(group_attach_event, ifg);
867 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
868 }
869
870 ifg->ifg_refcnt++;
871 ifgl->ifgl_group = ifg;
872 ifgm->ifgm_ifp = ifp;
873
874 IF_ADDR_LOCK(ifp);
875 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
876 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
877 IF_ADDR_UNLOCK(ifp);
878
879 IFNET_WUNLOCK();
880
881 EVENTHANDLER_INVOKE(group_change_event, groupname);
882
883 return (0);
884}
885
886/*
887 * Remove a group from an interface
888 */
889int
890if_delgroup(struct ifnet *ifp, const char *groupname)
891{
892 struct ifg_list *ifgl;
893 struct ifg_member *ifgm;
894
895 IFNET_WLOCK();
896 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
897 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
898 break;
899 if (ifgl == NULL) {
900 IFNET_WUNLOCK();
901 return (ENOENT);
902 }
903
904 IF_ADDR_LOCK(ifp);
905 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
906 IF_ADDR_UNLOCK(ifp);
907
908 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
909 if (ifgm->ifgm_ifp == ifp)
910 break;
911
912 if (ifgm != NULL) {
913 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
914 free(ifgm, M_TEMP);
915 }
916
917 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
918 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
919 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
920 free(ifgl->ifgl_group, M_TEMP);
921 }
922 IFNET_WUNLOCK();
923
924 free(ifgl, M_TEMP);
925
926 EVENTHANDLER_INVOKE(group_change_event, groupname);
927
928 return (0);
929}
930
931/*
932 * Stores all groups from an interface in memory pointed
933 * to by data
934 */
935static int
936if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
937{
938 int len, error;
939 struct ifg_list *ifgl;
940 struct ifg_req ifgrq, *ifgp;
941 struct ifgroupreq *ifgr = data;
942
943 if (ifgr->ifgr_len == 0) {
944 IF_ADDR_LOCK(ifp);
945 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
946 ifgr->ifgr_len += sizeof(struct ifg_req);
947 IF_ADDR_UNLOCK(ifp);
948 return (0);
949 }
950
951 len = ifgr->ifgr_len;
952 ifgp = ifgr->ifgr_groups;
953 /* XXX: wire */
954 IF_ADDR_LOCK(ifp);
955 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
956 if (len < sizeof(ifgrq)) {
957 IF_ADDR_UNLOCK(ifp);
958 return (EINVAL);
959 }
960 bzero(&ifgrq, sizeof ifgrq);
961 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
962 sizeof(ifgrq.ifgrq_group));
963 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
964 IF_ADDR_UNLOCK(ifp);
965 return (error);
966 }
967 len -= sizeof(ifgrq);
968 ifgp++;
969 }
970 IF_ADDR_UNLOCK(ifp);
971
972 return (0);
973}
974
975/*
976 * Stores all members of a group in memory pointed to by data
977 */
978static int
979if_getgroupmembers(struct ifgroupreq *data)
980{
981 struct ifgroupreq *ifgr = data;
982 struct ifg_group *ifg;
983 struct ifg_member *ifgm;
984 struct ifg_req ifgrq, *ifgp;
985 int len, error;
986
987 IFNET_RLOCK();
988 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
989 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
990 break;
991 if (ifg == NULL) {
992 IFNET_RUNLOCK();
993 return (ENOENT);
994 }
995
996 if (ifgr->ifgr_len == 0) {
997 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
998 ifgr->ifgr_len += sizeof(ifgrq);
999 IFNET_RUNLOCK();
1000 return (0);
1001 }
1002
1003 len = ifgr->ifgr_len;
1004 ifgp = ifgr->ifgr_groups;
1005 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1006 if (len < sizeof(ifgrq)) {
1007 IFNET_RUNLOCK();
1008 return (EINVAL);
1009 }
1010 bzero(&ifgrq, sizeof ifgrq);
1011 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1012 sizeof(ifgrq.ifgrq_member));
1013 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1014 IFNET_RUNLOCK();
1015 return (error);
1016 }
1017 len -= sizeof(ifgrq);
1018 ifgp++;
1019 }
1020 IFNET_RUNLOCK();
1021
1022 return (0);
1023}
1024
1025/*
1026 * Delete Routes for a Network Interface
1027 *
1028 * Called for each routing entry via the rnh->rnh_walktree() call above
1029 * to delete all route entries referencing a detaching network interface.
1030 *
1031 * Arguments:
1032 * rn pointer to node in the routing table
1033 * arg argument passed to rnh->rnh_walktree() - detaching interface
1034 *
1035 * Returns:
1036 * 0 successful
1037 * errno failed - reason indicated
1038 *
1039 */
1040static int
1041if_rtdel(struct radix_node *rn, void *arg)
1042{
1043 struct rtentry *rt = (struct rtentry *)rn;
1044 struct ifnet *ifp = arg;
1045 int err;
1046
1047 if (rt->rt_ifp == ifp) {
1048
1049 /*
1050 * Protect (sorta) against walktree recursion problems
1051 * with cloned routes
1052 */
1053 if ((rt->rt_flags & RTF_UP) == 0)
1054 return (0);
1055
1056 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1057 rt_mask(rt), rt->rt_flags,
1058 (struct rtentry **) NULL, rt->rt_fibnum);
1059 if (err) {
1060 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1061 }
1062 }
1063
1064 return (0);
1065}
1066
1067/*
1068 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1069 * structs used to represent other address families, it is necessary
1070 * to perform a different comparison.
1071 */
1072
1073#define sa_equal(a1, a2) \
1074 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1075
1076#define sa_dl_equal(a1, a2) \
1077 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1078 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1079 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1080 LLADDR((struct sockaddr_dl *)(a2)), \
1081 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1082
1083/*
1084 * Locate an interface based on a complete address.
1085 */
1086/*ARGSUSED*/
1087struct ifaddr *
1088ifa_ifwithaddr(struct sockaddr *addr)
1089{
1090 struct ifnet *ifp;
1091 struct ifaddr *ifa;
1092
1093 IFNET_RLOCK();
1094 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
1095 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1096 if (ifa->ifa_addr->sa_family != addr->sa_family)
1097 continue;
1098 if (sa_equal(addr, ifa->ifa_addr))
1099 goto done;
1100 /* IP6 doesn't have broadcast */
1101 if ((ifp->if_flags & IFF_BROADCAST) &&
1102 ifa->ifa_broadaddr &&
1103 ifa->ifa_broadaddr->sa_len != 0 &&
1104 sa_equal(ifa->ifa_broadaddr, addr))
1105 goto done;
1106 }
1107 ifa = NULL;
1108done:
1109 IFNET_RUNLOCK();
1110 return (ifa);
1111}
1112
1113/*
1114 * Locate an interface based on the broadcast address.
1115 */
1116/* ARGSUSED */
1117struct ifaddr *
1118ifa_ifwithbroadaddr(struct sockaddr *addr)
1119{
1120 struct ifnet *ifp;
1121 struct ifaddr *ifa;
1122
1123 IFNET_RLOCK();
1124 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
1125 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1126 if (ifa->ifa_addr->sa_family != addr->sa_family)
1127 continue;
1128 if ((ifp->if_flags & IFF_BROADCAST) &&
1129 ifa->ifa_broadaddr &&
1130 ifa->ifa_broadaddr->sa_len != 0 &&
1131 sa_equal(ifa->ifa_broadaddr, addr))
1132 goto done;
1133 }
1134 ifa = NULL;
1135done:
1136 IFNET_RUNLOCK();
1137 return (ifa);
1138}
1139
1140/*
1141 * Locate the point to point interface with a given destination address.
1142 */
1143/*ARGSUSED*/
1144struct ifaddr *
1145ifa_ifwithdstaddr(struct sockaddr *addr)
1146{
1147 struct ifnet *ifp;
1148 struct ifaddr *ifa;
1149
1150 IFNET_RLOCK();
1151 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1152 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1153 continue;
1154 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1155 if (ifa->ifa_addr->sa_family != addr->sa_family)
1156 continue;
1157 if (ifa->ifa_dstaddr != NULL &&
1158 sa_equal(addr, ifa->ifa_dstaddr))
1159 goto done;
1160 }
1161 }
1162 ifa = NULL;
1163done:
1164 IFNET_RUNLOCK();
1165 return (ifa);
1166}
1167
1168/*
1169 * Find an interface on a specific network. If many, choice
1170 * is most specific found.
1171 */
1172struct ifaddr *
1173ifa_ifwithnet(struct sockaddr *addr)
1174{
1175 struct ifnet *ifp;
1176 struct ifaddr *ifa;
1177 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
1178 u_int af = addr->sa_family;
1179 char *addr_data = addr->sa_data, *cplim;
1180
1181 /*
1182 * AF_LINK addresses can be looked up directly by their index number,
1183 * so do that if we can.
1184 */
1185 if (af == AF_LINK) {
1186 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1187 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1188 return (ifaddr_byindex(sdl->sdl_index));
1189 }
1190
1191 /*
1192 * Scan though each interface, looking for ones that have
1193 * addresses in this address family.
1194 */
1195 IFNET_RLOCK();
1196 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1197 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1198 char *cp, *cp2, *cp3;
1199
1200 if (ifa->ifa_addr->sa_family != af)
1201next: continue;
1202 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
1203 /*
1204 * This is a bit broken as it doesn't
1205 * take into account that the remote end may
1206 * be a single node in the network we are
1207 * looking for.
1208 * The trouble is that we don't know the
1209 * netmask for the remote end.
1210 */
1211 if (ifa->ifa_dstaddr != NULL &&
1212 sa_equal(addr, ifa->ifa_dstaddr))
1213 goto done;
1214 } else {
1215 /*
1216 * if we have a special address handler,
1217 * then use it instead of the generic one.
1218 */
1219 if (ifa->ifa_claim_addr) {
1220 if ((*ifa->ifa_claim_addr)(ifa, addr))
1221 goto done;
1222 continue;
1223 }
1224
1225 /*
1226 * Scan all the bits in the ifa's address.
1227 * If a bit dissagrees with what we are
1228 * looking for, mask it with the netmask
1229 * to see if it really matters.
1230 * (A byte at a time)
1231 */
1232 if (ifa->ifa_netmask == 0)
1233 continue;
1234 cp = addr_data;
1235 cp2 = ifa->ifa_addr->sa_data;
1236 cp3 = ifa->ifa_netmask->sa_data;
1237 cplim = ifa->ifa_netmask->sa_len
1238 + (char *)ifa->ifa_netmask;
1239 while (cp3 < cplim)
1240 if ((*cp++ ^ *cp2++) & *cp3++)
1241 goto next; /* next address! */
1242 /*
1243 * If the netmask of what we just found
1244 * is more specific than what we had before
1245 * (if we had one) then remember the new one
1246 * before continuing to search
1247 * for an even better one.
1248 */
1249 if (ifa_maybe == 0 ||
1250 rn_refines((caddr_t)ifa->ifa_netmask,
1251 (caddr_t)ifa_maybe->ifa_netmask))
1252 ifa_maybe = ifa;
1253 }
1254 }
1255 }
1256 ifa = ifa_maybe;
1257done:
1258 IFNET_RUNLOCK();
1259 return (ifa);
1260}
1261
1262/*
1263 * Find an interface address specific to an interface best matching
1264 * a given address.
1265 */
1266struct ifaddr *
1267ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1268{
1269 struct ifaddr *ifa;
1270 char *cp, *cp2, *cp3;
1271 char *cplim;
1272 struct ifaddr *ifa_maybe = 0;
1273 u_int af = addr->sa_family;
1274
1275 if (af >= AF_MAX)
1276 return (0);
1277 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1278 if (ifa->ifa_addr->sa_family != af)
1279 continue;
1280 if (ifa_maybe == 0)
1281 ifa_maybe = ifa;
1282 if (ifa->ifa_netmask == 0) {
1283 if (sa_equal(addr, ifa->ifa_addr) ||
1284 (ifa->ifa_dstaddr &&
1285 sa_equal(addr, ifa->ifa_dstaddr)))
1286 goto done;
1287 continue;
1288 }
1289 if (ifp->if_flags & IFF_POINTOPOINT) {
1290 if (sa_equal(addr, ifa->ifa_dstaddr))
1291 goto done;
1292 } else {
1293 cp = addr->sa_data;
1294 cp2 = ifa->ifa_addr->sa_data;
1295 cp3 = ifa->ifa_netmask->sa_data;
1296 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1297 for (; cp3 < cplim; cp3++)
1298 if ((*cp++ ^ *cp2++) & *cp3)
1299 break;
1300 if (cp3 == cplim)
1301 goto done;
1302 }
1303 }
1304 ifa = ifa_maybe;
1305done:
1306 return (ifa);
1307}
1308
1309#include <net/route.h>
1310
1311/*
1312 * Default action when installing a route with a Link Level gateway.
1313 * Lookup an appropriate real ifa to point to.
1314 * This should be moved to /sys/net/link.c eventually.
1315 */
1316static void
1317link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1318{
1319 struct ifaddr *ifa, *oifa;
1320 struct sockaddr *dst;
1321 struct ifnet *ifp;
1322
1323 RT_LOCK_ASSERT(rt);
1324
1325 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1326 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1327 return;
1328 ifa = ifaof_ifpforaddr(dst, ifp);
1329 if (ifa) {
1330 IFAREF(ifa); /* XXX */
1331 oifa = rt->rt_ifa;
1332 rt->rt_ifa = ifa;
1333 IFAFREE(oifa);
1334 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1335 ifa->ifa_rtrequest(cmd, rt, info);
1336 }
1337}
1338
1339/*
1340 * Mark an interface down and notify protocols of
1341 * the transition.
1342 * NOTE: must be called at splnet or eqivalent.
1343 */
1344static void
1345if_unroute(struct ifnet *ifp, int flag, int fam)
1346{
1347 struct ifaddr *ifa;
1348
1349 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1350
1351 ifp->if_flags &= ~flag;
1352 getmicrotime(&ifp->if_lastchange);
1353 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1354 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1355 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1356 if_qflush(&ifp->if_snd);
1357#ifdef DEV_CARP
1358 if (ifp->if_carp)
1359 carp_carpdev_state(ifp->if_carp);
1360#endif
1361 rt_ifmsg(ifp);
1362}
1363
1364/*
1365 * Mark an interface up and notify protocols of
1366 * the transition.
1367 * NOTE: must be called at splnet or eqivalent.
1368 */
1369static void
1370if_route(struct ifnet *ifp, int flag, int fam)
1371{
1372 struct ifaddr *ifa;
1373
1374 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1375
1376 ifp->if_flags |= flag;
1377 getmicrotime(&ifp->if_lastchange);
1378 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1379 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1380 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1381#ifdef DEV_CARP
1382 if (ifp->if_carp)
1383 carp_carpdev_state(ifp->if_carp);
1384#endif
1385 rt_ifmsg(ifp);
1386#ifdef INET6
1387 in6_if_up(ifp);
1388#endif
1389}
1390
1391void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1392void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
1393
1394/*
1395 * Handle a change in the interface link state. To avoid LORs
1396 * between driver lock and upper layer locks, as well as possible
1397 * recursions, we post event to taskqueue, and all job
1398 * is done in static do_link_state_change().
1399 */
1400void
1401if_link_state_change(struct ifnet *ifp, int link_state)
1402{
1403 /* Return if state hasn't changed. */
1404 if (ifp->if_link_state == link_state)
1405 return;
1406
1407 ifp->if_link_state = link_state;
1408
1409 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1410}
1411
1412static void
1413do_link_state_change(void *arg, int pending)
1414{
1415 struct ifnet *ifp = (struct ifnet *)arg;
1416 int link_state = ifp->if_link_state;
1417 int link;
1418
1419 /* Notify that the link state has changed. */
1420 rt_ifmsg(ifp);
1421 if (link_state == LINK_STATE_UP)
1422 link = NOTE_LINKUP;
1423 else if (link_state == LINK_STATE_DOWN)
1424 link = NOTE_LINKDOWN;
1425 else
1426 link = NOTE_LINKINV;
1427 KNOTE_UNLOCKED(&ifp->if_klist, link);
1428 if (ifp->if_vlantrunk != NULL)
1429 (*vlan_link_state_p)(ifp, link);
1430
1431 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1432 IFP2AC(ifp)->ac_netgraph != NULL)
1433 (*ng_ether_link_state_p)(ifp, link_state);
1434#ifdef DEV_CARP
1435 if (ifp->if_carp)
1436 carp_carpdev_state(ifp->if_carp);
1437#endif
1438 if (ifp->if_bridge) {
1439 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1440 (*bstp_linkstate_p)(ifp, link_state);
1441 }
1442 if (ifp->if_lagg) {
1443 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1444 (*lagg_linkstate_p)(ifp, link_state);
1445 }
1446
1447 devctl_notify("IFNET", ifp->if_xname,
1448 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
1449 if (pending > 1)
1450 if_printf(ifp, "%d link states coalesced\n", pending);
1451 if (log_link_state_change)
1452 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1453 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1454}
1455
1456/*
1457 * Mark an interface down and notify protocols of
1458 * the transition.
1459 * NOTE: must be called at splnet or eqivalent.
1460 */
1461void
1462if_down(struct ifnet *ifp)
1463{
1464
1465 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1466}
1467
1468/*
1469 * Mark an interface up and notify protocols of
1470 * the transition.
1471 * NOTE: must be called at splnet or eqivalent.
1472 */
1473void
1474if_up(struct ifnet *ifp)
1475{
1476
1477 if_route(ifp, IFF_UP, AF_UNSPEC);
1478}
1479
1480/*
1481 * Flush an interface queue.
1482 */
1483static void
1484if_qflush(struct ifaltq *ifq)
1485{
1486 struct mbuf *m, *n;
1487
1488 IFQ_LOCK(ifq);
1489#ifdef ALTQ
1490 if (ALTQ_IS_ENABLED(ifq))
1491 ALTQ_PURGE(ifq);
1492#endif
1493 n = ifq->ifq_head;
1494 while ((m = n) != 0) {
1495 n = m->m_act;
1496 m_freem(m);
1497 }
1498 ifq->ifq_head = 0;
1499 ifq->ifq_tail = 0;
1500 ifq->ifq_len = 0;
1501 IFQ_UNLOCK(ifq);
1502}
1503
1504/*
1505 * Handle interface watchdog timer routines. Called
1506 * from softclock, we decrement timers (if set) and
1507 * call the appropriate interface routine on expiration.
1508 *
1509 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1510 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab
1511 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1512 */
1513static void
1514if_slowtimo(void *arg)
1515{
1516 struct ifnet *ifp;
1517 int s = splimp();
1518
1519 IFNET_RLOCK();
1520 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1521 if (ifp->if_timer == 0 || --ifp->if_timer)
1522 continue;
1523 if (ifp->if_watchdog)
1524 (*ifp->if_watchdog)(ifp);
1525 }
1526 IFNET_RUNLOCK();
1527 splx(s);
1528 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1529}
1530
1531/*
1532 * Map interface name to
1533 * interface structure pointer.
1534 */
1535struct ifnet *
1536ifunit(const char *name)
1537{
1538 struct ifnet *ifp;
1539
1540 IFNET_RLOCK();
1541 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1542 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1543 break;
1544 }
1545 IFNET_RUNLOCK();
1546 return (ifp);
1547}
1548
1549/*
1550 * Hardware specific interface ioctls.
1551 */
1552static int
1553ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1554{
1555 struct ifreq *ifr;
1556 struct ifstat *ifs;
1557 int error = 0;
1558 int new_flags, temp_flags;
1559 size_t namelen, onamelen;
1560 char new_name[IFNAMSIZ];
1561 struct ifaddr *ifa;
1562 struct sockaddr_dl *sdl;
1563
1564 ifr = (struct ifreq *)data;
1565 switch (cmd) {
1566 case SIOCGIFINDEX:
1567 ifr->ifr_index = ifp->if_index;
1568 break;
1569
1570 case SIOCGIFFLAGS:
1571 temp_flags = ifp->if_flags | ifp->if_drv_flags;
1572 ifr->ifr_flags = temp_flags & 0xffff;
1573 ifr->ifr_flagshigh = temp_flags >> 16;
1574 break;
1575
1576 case SIOCGIFCAP:
1577 ifr->ifr_reqcap = ifp->if_capabilities;
1578 ifr->ifr_curcap = ifp->if_capenable;
1579 break;
1580
1581#ifdef MAC
1582 case SIOCGIFMAC:
1583 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
1584 break;
1585#endif
1586
1587 case SIOCGIFMETRIC:
1588 ifr->ifr_metric = ifp->if_metric;
1589 break;
1590
1591 case SIOCGIFMTU:
1592 ifr->ifr_mtu = ifp->if_mtu;
1593 break;
1594
1595 case SIOCGIFPHYS:
1596 ifr->ifr_phys = ifp->if_physical;
1597 break;
1598
1599 case SIOCSIFFLAGS:
1600 error = priv_check(td, PRIV_NET_SETIFFLAGS);
1601 if (error)
1602 return (error);
1603 /*
1604 * Currently, no driver owned flags pass the IFF_CANTCHANGE
1605 * check, so we don't need special handling here yet.
1606 */
1607 new_flags = (ifr->ifr_flags & 0xffff) |
1608 (ifr->ifr_flagshigh << 16);
1609 if (ifp->if_flags & IFF_SMART) {
1610 /* Smart drivers twiddle their own routes */
1611 } else if (ifp->if_flags & IFF_UP &&
1612 (new_flags & IFF_UP) == 0) {
1613 int s = splimp();
1614 if_down(ifp);
1615 splx(s);
1616 } else if (new_flags & IFF_UP &&
1617 (ifp->if_flags & IFF_UP) == 0) {
1618 int s = splimp();
1619 if_up(ifp);
1620 splx(s);
1621 }
1622 /* See if permanently promiscuous mode bit is about to flip */
1623 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
1624 if (new_flags & IFF_PPROMISC)
1625 ifp->if_flags |= IFF_PROMISC;
1626 else if (ifp->if_pcount == 0)
1627 ifp->if_flags &= ~IFF_PROMISC;
1628 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
1629 ifp->if_xname,
1630 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
1631 }
1632 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1633 (new_flags &~ IFF_CANTCHANGE);
1634 if (ifp->if_ioctl) {
1635 IFF_LOCKGIANT(ifp);
1636 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1637 IFF_UNLOCKGIANT(ifp);
1638 }
1639 getmicrotime(&ifp->if_lastchange);
1640 break;
1641
1642 case SIOCSIFCAP:
1643 error = priv_check(td, PRIV_NET_SETIFCAP);
1644 if (error)
1645 return (error);
1646 if (ifp->if_ioctl == NULL)
1647 return (EOPNOTSUPP);
1648 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1649 return (EINVAL);
1650 IFF_LOCKGIANT(ifp);
1651 error = (*ifp->if_ioctl)(ifp, cmd, data);
1652 IFF_UNLOCKGIANT(ifp);
1653 if (error == 0)
1654 getmicrotime(&ifp->if_lastchange);
1655 break;
1656
1657#ifdef MAC
1658 case SIOCSIFMAC:
1659 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
1660 break;
1661#endif
1662
1663 case SIOCSIFNAME:
1664 error = priv_check(td, PRIV_NET_SETIFNAME);
1665 if (error)
1666 return (error);
1667 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1668 if (error != 0)
1669 return (error);
1670 if (new_name[0] == '\0')
1671 return (EINVAL);
1672 if (ifunit(new_name) != NULL)
1673 return (EEXIST);
1674
1675 /* Announce the departure of the interface. */
1676 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1677 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1678
1679 log(LOG_INFO, "%s: changing name to '%s'\n",
1680 ifp->if_xname, new_name);
1681
1682 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1683 ifa = ifp->if_addr;
1684 IFA_LOCK(ifa);
1685 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1686 namelen = strlen(new_name);
1687 onamelen = sdl->sdl_nlen;
1688 /*
1689 * Move the address if needed. This is safe because we
1690 * allocate space for a name of length IFNAMSIZ when we
1691 * create this in if_attach().
1692 */
1693 if (namelen != onamelen) {
1694 bcopy(sdl->sdl_data + onamelen,
1695 sdl->sdl_data + namelen, sdl->sdl_alen);
1696 }
1697 bcopy(new_name, sdl->sdl_data, namelen);
1698 sdl->sdl_nlen = namelen;
1699 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1700 bzero(sdl->sdl_data, onamelen);
1701 while (namelen != 0)
1702 sdl->sdl_data[--namelen] = 0xff;
1703 IFA_UNLOCK(ifa);
1704
1705 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1706 /* Announce the return of the interface. */
1707 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1708 break;
1709
1710 case SIOCSIFMETRIC:
1711 error = priv_check(td, PRIV_NET_SETIFMETRIC);
1712 if (error)
1713 return (error);
1714 ifp->if_metric = ifr->ifr_metric;
1715 getmicrotime(&ifp->if_lastchange);
1716 break;
1717
1718 case SIOCSIFPHYS:
1719 error = priv_check(td, PRIV_NET_SETIFPHYS);
1720 if (error)
1721 return (error);
1722 if (ifp->if_ioctl == NULL)
1723 return (EOPNOTSUPP);
1724 IFF_LOCKGIANT(ifp);
1725 error = (*ifp->if_ioctl)(ifp, cmd, data);
1726 IFF_UNLOCKGIANT(ifp);
1727 if (error == 0)
1728 getmicrotime(&ifp->if_lastchange);
1729 break;
1730
1731 case SIOCSIFMTU:
1732 {
1733 u_long oldmtu = ifp->if_mtu;
1734
1735 error = priv_check(td, PRIV_NET_SETIFMTU);
1736 if (error)
1737 return (error);
1738 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1739 return (EINVAL);
1740 if (ifp->if_ioctl == NULL)
1741 return (EOPNOTSUPP);
1742 IFF_LOCKGIANT(ifp);
1743 error = (*ifp->if_ioctl)(ifp, cmd, data);
1744 IFF_UNLOCKGIANT(ifp);
1745 if (error == 0) {
1746 getmicrotime(&ifp->if_lastchange);
1747 rt_ifmsg(ifp);
1748 }
1749 /*
1750 * If the link MTU changed, do network layer specific procedure.
1751 */
1752 if (ifp->if_mtu != oldmtu) {
1753#ifdef INET6
1754 nd6_setmtu(ifp);
1755#endif
1756 }
1757 break;
1758 }
1759
1760 case SIOCADDMULTI:
1761 case SIOCDELMULTI:
1762 if (cmd == SIOCADDMULTI)
1763 error = priv_check(td, PRIV_NET_ADDMULTI);
1764 else
1765 error = priv_check(td, PRIV_NET_DELMULTI);
1766 if (error)
1767 return (error);
1768
1769 /* Don't allow group membership on non-multicast interfaces. */
1770 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1771 return (EOPNOTSUPP);
1772
1773 /* Don't let users screw up protocols' entries. */
1774 if (ifr->ifr_addr.sa_family != AF_LINK)
1775 return (EINVAL);
1776
1777 if (cmd == SIOCADDMULTI) {
1778 struct ifmultiaddr *ifma;
1779
1780 /*
1781 * Userland is only permitted to join groups once
1782 * via the if_addmulti() KPI, because it cannot hold
1783 * struct ifmultiaddr * between calls. It may also
1784 * lose a race while we check if the membership
1785 * already exists.
1786 */
1787 IF_ADDR_LOCK(ifp);
1788 ifma = if_findmulti(ifp, &ifr->ifr_addr);
1789 IF_ADDR_UNLOCK(ifp);
1790 if (ifma != NULL)
1791 error = EADDRINUSE;
1792 else
1793 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1794 } else {
1795 error = if_delmulti(ifp, &ifr->ifr_addr);
1796 }
1797 if (error == 0)
1798 getmicrotime(&ifp->if_lastchange);
1799 break;
1800
1801 case SIOCSIFPHYADDR:
1802 case SIOCDIFPHYADDR:
1803#ifdef INET6
1804 case SIOCSIFPHYADDR_IN6:
1805#endif
1806 case SIOCSLIFPHYADDR:
1807 case SIOCSIFMEDIA:
1808 case SIOCSIFGENERIC:
1809 error = priv_check(td, PRIV_NET_HWIOCTL);
1810 if (error)
1811 return (error);
1812 if (ifp->if_ioctl == NULL)
1813 return (EOPNOTSUPP);
1814 IFF_LOCKGIANT(ifp);
1815 error = (*ifp->if_ioctl)(ifp, cmd, data);
1816 IFF_UNLOCKGIANT(ifp);
1817 if (error == 0)
1818 getmicrotime(&ifp->if_lastchange);
1819 break;
1820
1821 case SIOCGIFSTATUS:
1822 ifs = (struct ifstat *)data;
1823 ifs->ascii[0] = '\0';
1824
1825 case SIOCGIFPSRCADDR:
1826 case SIOCGIFPDSTADDR:
1827 case SIOCGLIFPHYADDR:
1828 case SIOCGIFMEDIA:
1829 case SIOCGIFGENERIC:
1830 if (ifp->if_ioctl == NULL)
1831 return (EOPNOTSUPP);
1832 IFF_LOCKGIANT(ifp);
1833 error = (*ifp->if_ioctl)(ifp, cmd, data);
1834 IFF_UNLOCKGIANT(ifp);
1835 break;
1836
1837 case SIOCSIFLLADDR:
1838 error = priv_check(td, PRIV_NET_SETLLADDR);
1839 if (error)
1840 return (error);
1841 error = if_setlladdr(ifp,
1842 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1843 break;
1844
1845 case SIOCAIFGROUP:
1846 {
1847 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1848
1849 error = priv_check(td, PRIV_NET_ADDIFGROUP);
1850 if (error)
1851 return (error);
1852 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
1853 return (error);
1854 break;
1855 }
1856
1857 case SIOCGIFGROUP:
1858 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
1859 return (error);
1860 break;
1861
1862 case SIOCDIFGROUP:
1863 {
1864 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1865
1866 error = priv_check(td, PRIV_NET_DELIFGROUP);
1867 if (error)
1868 return (error);
1869 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
1870 return (error);
1871 break;
1872 }
1873
1874 default:
1875 error = ENOIOCTL;
1876 break;
1877 }
1878 return (error);
1879}
1880
1881/*
1882 * Interface ioctls.
1883 */
1884int
1885ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1886{
1887 struct ifnet *ifp;
1888 struct ifreq *ifr;
1889 int error;
1890 int oif_flags;
1891
1892 switch (cmd) {
1893 case SIOCGIFCONF:
1894 case OSIOCGIFCONF:
1895#ifdef __amd64__
1896 case SIOCGIFCONF32:
1897#endif
1898 return (ifconf(cmd, data));
1899 }
1900 ifr = (struct ifreq *)data;
1901
1902 switch (cmd) {
1903 case SIOCIFCREATE:
1904 case SIOCIFCREATE2:
1905 error = priv_check(td, PRIV_NET_IFCREATE);
1906 if (error)
1907 return (error);
1908 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
1909 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
1910 case SIOCIFDESTROY:
1911 error = priv_check(td, PRIV_NET_IFDESTROY);
1912 if (error)
1913 return (error);
1914 return if_clone_destroy(ifr->ifr_name);
1915
1916 case SIOCIFGCLONERS:
1917 return (if_clone_list((struct if_clonereq *)data));
1918 case SIOCGIFGMEMB:
1919 return (if_getgroupmembers((struct ifgroupreq *)data));
1920 }
1921
1922 ifp = ifunit(ifr->ifr_name);
1923 if (ifp == 0)
1924 return (ENXIO);
1925
1926 error = ifhwioctl(cmd, ifp, data, td);
1927 if (error != ENOIOCTL)
1928 return (error);
1929
1930 oif_flags = ifp->if_flags;
1931 if (so->so_proto == 0)
1932 return (EOPNOTSUPP);
1933#ifndef COMPAT_43
1934 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1935 data,
1936 ifp, td));
1937#else
1938 {
1939 int ocmd = cmd;
1940
1941 switch (cmd) {
1942
1943 case SIOCSIFDSTADDR:
1944 case SIOCSIFADDR:
1945 case SIOCSIFBRDADDR:
1946 case SIOCSIFNETMASK:
1947#if BYTE_ORDER != BIG_ENDIAN
1948 if (ifr->ifr_addr.sa_family == 0 &&
1949 ifr->ifr_addr.sa_len < 16) {
1950 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1951 ifr->ifr_addr.sa_len = 16;
1952 }
1953#else
1954 if (ifr->ifr_addr.sa_len == 0)
1955 ifr->ifr_addr.sa_len = 16;
1956#endif
1957 break;
1958
1959 case OSIOCGIFADDR:
1960 cmd = SIOCGIFADDR;
1961 break;
1962
1963 case OSIOCGIFDSTADDR:
1964 cmd = SIOCGIFDSTADDR;
1965 break;
1966
1967 case OSIOCGIFBRDADDR:
1968 cmd = SIOCGIFBRDADDR;
1969 break;
1970
1971 case OSIOCGIFNETMASK:
1972 cmd = SIOCGIFNETMASK;
1973 }
1974 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
1975 cmd,
1976 data,
1977 ifp, td));
1978 switch (ocmd) {
1979
1980 case OSIOCGIFADDR:
1981 case OSIOCGIFDSTADDR:
1982 case OSIOCGIFBRDADDR:
1983 case OSIOCGIFNETMASK:
1984 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1985
1986 }
1987 }
1988#endif /* COMPAT_43 */
1989
1990 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1991#ifdef INET6
1992 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1993 if (ifp->if_flags & IFF_UP) {
1994 int s = splimp();
1995 in6_if_up(ifp);
1996 splx(s);
1997 }
1998#endif
1999 }
2000 return (error);
2001}
2002
2003/*
2004 * The code common to handling reference counted flags,
2005 * e.g., in ifpromisc() and if_allmulti().
2006 * The "pflag" argument can specify a permanent mode flag to check,
2007 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2008 *
2009 * Only to be used on stack-owned flags, not driver-owned flags.
2010 */
2011static int
2012if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2013{
2014 struct ifreq ifr;
2015 int error;
2016 int oldflags, oldcount;
2017
2018 /* Sanity checks to catch programming errors */
2019 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2020 ("%s: setting driver-owned flag %d", __func__, flag));
2021
2022 if (onswitch)
2023 KASSERT(*refcount >= 0,
2024 ("%s: increment negative refcount %d for flag %d",
2025 __func__, *refcount, flag));
2026 else
2027 KASSERT(*refcount > 0,
2028 ("%s: decrement non-positive refcount %d for flag %d",
2029 __func__, *refcount, flag));
2030
2031 /* In case this mode is permanent, just touch refcount */
2032 if (ifp->if_flags & pflag) {
2033 *refcount += onswitch ? 1 : -1;
2034 return (0);
2035 }
2036
2037 /* Save ifnet parameters for if_ioctl() may fail */
2038 oldcount = *refcount;
2039 oldflags = ifp->if_flags;
2040
2041 /*
2042 * See if we aren't the only and touching refcount is enough.
2043 * Actually toggle interface flag if we are the first or last.
2044 */
2045 if (onswitch) {
2046 if ((*refcount)++)
2047 return (0);
2048 ifp->if_flags |= flag;
2049 } else {
2050 if (--(*refcount))
2051 return (0);
2052 ifp->if_flags &= ~flag;
2053 }
2054
2055 /* Call down the driver since we've changed interface flags */
2056 if (ifp->if_ioctl == NULL) {
2057 error = EOPNOTSUPP;
2058 goto recover;
2059 }
2060 ifr.ifr_flags = ifp->if_flags & 0xffff;
2061 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2062 IFF_LOCKGIANT(ifp);
2063 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2064 IFF_UNLOCKGIANT(ifp);
2065 if (error)
2066 goto recover;
2067 /* Notify userland that interface flags have changed */
2068 rt_ifmsg(ifp);
2069 return (0);
2070
2071recover:
2072 /* Recover after driver error */
2073 *refcount = oldcount;
2074 ifp->if_flags = oldflags;
2075 return (error);
2076}
2077
2078/*
2079 * Set/clear promiscuous mode on interface ifp based on the truth value
2080 * of pswitch. The calls are reference counted so that only the first
2081 * "on" request actually has an effect, as does the final "off" request.
2082 * Results are undefined if the "off" and "on" requests are not matched.
2083 */
2084int
2085ifpromisc(struct ifnet *ifp, int pswitch)
2086{
2087 int error;
2088 int oldflags = ifp->if_flags;
2089
2090 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2091 &ifp->if_pcount, pswitch);
2092 /* If promiscuous mode status has changed, log a message */
2093 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2094 log(LOG_INFO, "%s: promiscuous mode %s\n",
2095 ifp->if_xname,
2096 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2097 return (error);
2098}
2099
2100/*
2101 * Return interface configuration
2102 * of system. List may be used
2103 * in later ioctl's (above) to get
2104 * other information.
2105 */
2106/*ARGSUSED*/
2107static int
2108ifconf(u_long cmd, caddr_t data)
2109{
2110 struct ifconf *ifc = (struct ifconf *)data;
2111#ifdef __amd64__
2112 struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2113 struct ifconf ifc_swab;
2114#endif
2115 struct ifnet *ifp;
2116 struct ifaddr *ifa;
2117 struct ifreq ifr;
2118 struct sbuf *sb;
2119 int error, full = 0, valid_len, max_len;
2120
2121#ifdef __amd64__
2122 if (cmd == SIOCGIFCONF32) {
2123 ifc_swab.ifc_len = ifc32->ifc_len;
2124 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2125 ifc = &ifc_swab;
2126 }
2127#endif
2128 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2129 max_len = MAXPHYS - 1;
2130
2131 /* Prevent hostile input from being able to crash the system */
2132 if (ifc->ifc_len <= 0)
2133 return (EINVAL);
2134
2135again:
2136 if (ifc->ifc_len <= max_len) {
2137 max_len = ifc->ifc_len;
2138 full = 1;
2139 }
2140 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2141 max_len = 0;
2142 valid_len = 0;
2143
2144 IFNET_RLOCK(); /* could sleep XXX */
2145 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2146 int addrs;
2147
2148 /*
2149 * Zero the ifr_name buffer to make sure we don't
2150 * disclose the contents of the stack.
2151 */
2152 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2153
2154 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2155 >= sizeof(ifr.ifr_name)) {
2156 sbuf_delete(sb);
2157 IFNET_RUNLOCK();
2158 return (ENAMETOOLONG);
2159 }
2160
2161 addrs = 0;
2162 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2163 struct sockaddr *sa = ifa->ifa_addr;
2164
2165 if (jailed(curthread->td_ucred) &&
2166 prison_if(curthread->td_ucred, sa))
2167 continue;
2168 addrs++;
2169#ifdef COMPAT_43
2170 if (cmd == OSIOCGIFCONF) {
2171 struct osockaddr *osa =
2172 (struct osockaddr *)&ifr.ifr_addr;
2173 ifr.ifr_addr = *sa;
2174 osa->sa_family = sa->sa_family;
2175 sbuf_bcat(sb, &ifr, sizeof(ifr));
2176 max_len += sizeof(ifr);
2177 } else
2178#endif
2179 if (sa->sa_len <= sizeof(*sa)) {
2180 ifr.ifr_addr = *sa;
2181 sbuf_bcat(sb, &ifr, sizeof(ifr));
2182 max_len += sizeof(ifr);
2183 } else {
2184 sbuf_bcat(sb, &ifr,
2185 offsetof(struct ifreq, ifr_addr));
2186 max_len += offsetof(struct ifreq, ifr_addr);
2187 sbuf_bcat(sb, sa, sa->sa_len);
2188 max_len += sa->sa_len;
2189 }
2190
2191 if (!sbuf_overflowed(sb))
2192 valid_len = sbuf_len(sb);
2193 }
2194 if (addrs == 0) {
2195 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2196 sbuf_bcat(sb, &ifr, sizeof(ifr));
2197 max_len += sizeof(ifr);
2198
2199 if (!sbuf_overflowed(sb))
2200 valid_len = sbuf_len(sb);
2201 }
2202 }
2203 IFNET_RUNLOCK();
2204
2205 /*
2206 * If we didn't allocate enough space (uncommon), try again. If
2207 * we have already allocated as much space as we are allowed,
2208 * return what we've got.
2209 */
2210 if (valid_len != max_len && !full) {
2211 sbuf_delete(sb);
2212 goto again;
2213 }
2214
2215 ifc->ifc_len = valid_len;
2216#ifdef __amd64__
2217 if (cmd == SIOCGIFCONF32)
2218 ifc32->ifc_len = valid_len;
2219#endif
2220 sbuf_finish(sb);
2221 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2222 sbuf_delete(sb);
2223 return (error);
2224}
2225
2226/*
2227 * Just like ifpromisc(), but for all-multicast-reception mode.
2228 */
2229int
2230if_allmulti(struct ifnet *ifp, int onswitch)
2231{
2232
2233 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2234}
2235
2236struct ifmultiaddr *
2237if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2238{
2239 struct ifmultiaddr *ifma;
2240
2241 IF_ADDR_LOCK_ASSERT(ifp);
2242
2243 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2244 if (sa->sa_family == AF_LINK) {
2245 if (sa_dl_equal(ifma->ifma_addr, sa))
2246 break;
2247 } else {
2248 if (sa_equal(ifma->ifma_addr, sa))
2249 break;
2250 }
2251 }
2252
2253 return ifma;
2254}
2255
2256/*
2257 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2258 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2259 * the ifnet multicast address list here, so the caller must do that and
2260 * other setup work (such as notifying the device driver). The reference
2261 * count is initialized to 1.
2262 */
2263static struct ifmultiaddr *
2264if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2265 int mflags)
2266{
2267 struct ifmultiaddr *ifma;
2268 struct sockaddr *dupsa;
2269
2270 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
2271 M_ZERO);
2272 if (ifma == NULL)
2273 return (NULL);
2274
2275 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
2276 if (dupsa == NULL) {
2277 FREE(ifma, M_IFMADDR);
2278 return (NULL);
2279 }
2280 bcopy(sa, dupsa, sa->sa_len);
2281 ifma->ifma_addr = dupsa;
2282
2283 ifma->ifma_ifp = ifp;
2284 ifma->ifma_refcount = 1;
2285 ifma->ifma_protospec = NULL;
2286
2287 if (llsa == NULL) {
2288 ifma->ifma_lladdr = NULL;
2289 return (ifma);
2290 }
2291
2292 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
2293 if (dupsa == NULL) {
2294 FREE(ifma->ifma_addr, M_IFMADDR);
2295 FREE(ifma, M_IFMADDR);
2296 return (NULL);
2297 }
2298 bcopy(llsa, dupsa, llsa->sa_len);
2299 ifma->ifma_lladdr = dupsa;
2300
2301 return (ifma);
2302}
2303
2304/*
2305 * if_freemulti: free ifmultiaddr structure and possibly attached related
2306 * addresses. The caller is responsible for implementing reference
2307 * counting, notifying the driver, handling routing messages, and releasing
2308 * any dependent link layer state.
2309 */
2310static void
2311if_freemulti(struct ifmultiaddr *ifma)
2312{
2313
2314 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2315 ifma->ifma_refcount));
2316 KASSERT(ifma->ifma_protospec == NULL,
2317 ("if_freemulti: protospec not NULL"));
2318
2319 if (ifma->ifma_lladdr != NULL)
2320 FREE(ifma->ifma_lladdr, M_IFMADDR);
2321 FREE(ifma->ifma_addr, M_IFMADDR);
2322 FREE(ifma, M_IFMADDR);
2323}
2324
2325/*
2326 * Register an additional multicast address with a network interface.
2327 *
2328 * - If the address is already present, bump the reference count on the
2329 * address and return.
2330 * - If the address is not link-layer, look up a link layer address.
2331 * - Allocate address structures for one or both addresses, and attach to the
2332 * multicast address list on the interface. If automatically adding a link
2333 * layer address, the protocol address will own a reference to the link
2334 * layer address, to be freed when it is freed.
2335 * - Notify the network device driver of an addition to the multicast address
2336 * list.
2337 *
2338 * 'sa' points to caller-owned memory with the desired multicast address.
2339 *
2340 * 'retifma' will be used to return a pointer to the resulting multicast
2341 * address reference, if desired.
2342 */
2343int
2344if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2345 struct ifmultiaddr **retifma)
2346{
2347 struct ifmultiaddr *ifma, *ll_ifma;
2348 struct sockaddr *llsa;
2349 int error;
2350
2351 /*
2352 * If the address is already present, return a new reference to it;
2353 * otherwise, allocate storage and set up a new address.
2354 */
2355 IF_ADDR_LOCK(ifp);
2356 ifma = if_findmulti(ifp, sa);
2357 if (ifma != NULL) {
2358 ifma->ifma_refcount++;
2359 if (retifma != NULL)
2360 *retifma = ifma;
2361 IF_ADDR_UNLOCK(ifp);
2362 return (0);
2363 }
2364
2365 /*
2366 * The address isn't already present; resolve the protocol address
2367 * into a link layer address, and then look that up, bump its
2368 * refcount or allocate an ifma for that also. If 'llsa' was
2369 * returned, we will need to free it later.
2370 */
2371 llsa = NULL;
2372 ll_ifma = NULL;
2373 if (ifp->if_resolvemulti != NULL) {
2374 error = ifp->if_resolvemulti(ifp, &llsa, sa);
2375 if (error)
2376 goto unlock_out;
2377 }
2378
2379 /*
2380 * Allocate the new address. Don't hook it up yet, as we may also
2381 * need to allocate a link layer multicast address.
2382 */
2383 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2384 if (ifma == NULL) {
2385 error = ENOMEM;
2386 goto free_llsa_out;
2387 }
2388
2389 /*
2390 * If a link layer address is found, we'll need to see if it's
2391 * already present in the address list, or allocate is as well.
2392 * When this block finishes, the link layer address will be on the
2393 * list.
2394 */
2395 if (llsa != NULL) {
2396 ll_ifma = if_findmulti(ifp, llsa);
2397 if (ll_ifma == NULL) {
2398 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2399 if (ll_ifma == NULL) {
2400 --ifma->ifma_refcount;
2401 if_freemulti(ifma);
2402 error = ENOMEM;
2403 goto free_llsa_out;
2404 }
2405 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2406 ifma_link);
2407 } else
2408 ll_ifma->ifma_refcount++;
2409 ifma->ifma_llifma = ll_ifma;
2410 }
2411
2412 /*
2413 * We now have a new multicast address, ifma, and possibly a new or
2414 * referenced link layer address. Add the primary address to the
2415 * ifnet address list.
2416 */
2417 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2418
2419 if (retifma != NULL)
2420 *retifma = ifma;
2421
2422 /*
2423 * Must generate the message while holding the lock so that 'ifma'
2424 * pointer is still valid.
2425 */
2426 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2427 IF_ADDR_UNLOCK(ifp);
2428
2429 /*
2430 * We are certain we have added something, so call down to the
2431 * interface to let them know about it.
2432 */
2433 if (ifp->if_ioctl != NULL) {
2434 IFF_LOCKGIANT(ifp);
2435 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2436 IFF_UNLOCKGIANT(ifp);
2437 }
2438
2439 if (llsa != NULL)
2440 FREE(llsa, M_IFMADDR);
2441
2442 return (0);
2443
2444free_llsa_out:
2445 if (llsa != NULL)
2446 FREE(llsa, M_IFMADDR);
2447
2448unlock_out:
2449 IF_ADDR_UNLOCK(ifp);
2450 return (error);
2451}
2452
2453/*
2454 * Delete a multicast group membership by network-layer group address.
2455 *
2456 * Returns ENOENT if the entry could not be found. If ifp no longer
2457 * exists, results are undefined. This entry point should only be used
2458 * from subsystems which do appropriate locking to hold ifp for the
2459 * duration of the call.
2460 * Network-layer protocol domains must use if_delmulti_ifma().
2461 */
2462int
2463if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2464{
2465 struct ifmultiaddr *ifma;
2466 int lastref;
2467#ifdef INVARIANTS
2468 struct ifnet *oifp;
2469
2470 IFNET_RLOCK();
2471 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2472 if (ifp == oifp)
2473 break;
2474 if (ifp != oifp)
2475 ifp = NULL;
2476 IFNET_RUNLOCK();
2477
2478 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
2479#endif
2480 if (ifp == NULL)
2481 return (ENOENT);
2482
2483 IF_ADDR_LOCK(ifp);
2484 lastref = 0;
2485 ifma = if_findmulti(ifp, sa);
2486 if (ifma != NULL)
2487 lastref = if_delmulti_locked(ifp, ifma, 0);
2488 IF_ADDR_UNLOCK(ifp);
2489
2490 if (ifma == NULL)
2491 return (ENOENT);
2492
2493 if (lastref && ifp->if_ioctl != NULL) {
2494 IFF_LOCKGIANT(ifp);
2495 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2496 IFF_UNLOCKGIANT(ifp);
2497 }
2498
2499 return (0);
2500}
2501
2502/*
2503 * Delete a multicast group membership by group membership pointer.
2504 * Network-layer protocol domains must use this routine.
2505 *
2506 * It is safe to call this routine if the ifp disappeared. Callers should
2507 * hold IFF_LOCKGIANT() to avoid a LOR in case the hardware needs to be
2508 * reconfigured.
2509 */
2510void
2511if_delmulti_ifma(struct ifmultiaddr *ifma)
2512{
2513 struct ifnet *ifp;
2514 int lastref;
2515
2516 ifp = ifma->ifma_ifp;
2517#ifdef DIAGNOSTIC
2518 if (ifp == NULL) {
2519 printf("%s: ifma_ifp seems to be detached\n", __func__);
2520 } else {
2521 struct ifnet *oifp;
2522
2523 IFNET_RLOCK();
2524 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2525 if (ifp == oifp)
2526 break;
2527 if (ifp != oifp) {
2528 printf("%s: ifnet %p disappeared\n", __func__, ifp);
2529 ifp = NULL;
2530 }
2531 IFNET_RUNLOCK();
2532 }
2533#endif
2534 /*
2535 * If and only if the ifnet instance exists: Acquire the address lock.
2536 */
2537 if (ifp != NULL)
2538 IF_ADDR_LOCK(ifp);
2539
2540 lastref = if_delmulti_locked(ifp, ifma, 0);
2541
2542 if (ifp != NULL) {
2543 /*
2544 * If and only if the ifnet instance exists:
2545 * Release the address lock.
2546 * If the group was left: update the hardware hash filter.
2547 */
2548 IF_ADDR_UNLOCK(ifp);
2549 if (lastref && ifp->if_ioctl != NULL) {
2550 IFF_LOCKGIANT(ifp);
2551 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2552 IFF_UNLOCKGIANT(ifp);
2553 }
2554 }
2555}
2556
2557/*
2558 * Perform deletion of network-layer and/or link-layer multicast address.
2559 *
2560 * Return 0 if the reference count was decremented.
2561 * Return 1 if the final reference was released, indicating that the
2562 * hardware hash filter should be reprogrammed.
2563 */
2564static int
2565if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
2566{
2567 struct ifmultiaddr *ll_ifma;
2568
2569 if (ifp != NULL && ifma->ifma_ifp != NULL) {
2570 KASSERT(ifma->ifma_ifp == ifp,
2571 ("%s: inconsistent ifp %p", __func__, ifp));
2572 IF_ADDR_LOCK_ASSERT(ifp);
2573 }
2574
2575 ifp = ifma->ifma_ifp;
2576
2577 /*
2578 * If the ifnet is detaching, null out references to ifnet,
2579 * so that upper protocol layers will notice, and not attempt
2580 * to obtain locks for an ifnet which no longer exists. The
2581 * routing socket announcement must happen before the ifnet
2582 * instance is detached from the system.
2583 */
2584 if (detaching) {
2585#ifdef DIAGNOSTIC
2586 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
2587#endif
2588 /*
2589 * ifp may already be nulled out if we are being reentered
2590 * to delete the ll_ifma.
2591 */
2592 if (ifp != NULL) {
2593 rt_newmaddrmsg(RTM_DELMADDR, ifma);
2594 ifma->ifma_ifp = NULL;
2595 }
2596 }
2597
2598 if (--ifma->ifma_refcount > 0)
2599 return 0;
2600
2601 /*
2602 * If this ifma is a network-layer ifma, a link-layer ifma may
2603 * have been associated with it. Release it first if so.
2604 */
2605 ll_ifma = ifma->ifma_llifma;
2606 if (ll_ifma != NULL) {
2607 KASSERT(ifma->ifma_lladdr != NULL,
2608 ("%s: llifma w/o lladdr", __func__));
2609 if (detaching)
2610 ll_ifma->ifma_ifp = NULL; /* XXX */
2611 if (--ll_ifma->ifma_refcount == 0) {
2612 if (ifp != NULL) {
2613 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
2614 ifma_link);
2615 }
2616 if_freemulti(ll_ifma);
2617 }
2618 }
2619
2620 if (ifp != NULL)
2621 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
2622
2623 if_freemulti(ifma);
2624
2625 /*
2626 * The last reference to this instance of struct ifmultiaddr
2627 * was released; the hardware should be notified of this change.
2628 */
2629 return 1;
2630}
2631
2632/*
2633 * Set the link layer address on an interface.
2634 *
2635 * At this time we only support certain types of interfaces,
2636 * and we don't allow the length of the address to change.
2637 */
2638int
2639if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
2640{
2641 struct sockaddr_dl *sdl;
2642 struct ifaddr *ifa;
2643 struct ifreq ifr;
2644
2645 ifa = ifp->if_addr;
2646 if (ifa == NULL)
2647 return (EINVAL);
2648 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2649 if (sdl == NULL)
2650 return (EINVAL);
2651 if (len != sdl->sdl_alen) /* don't allow length to change */
2652 return (EINVAL);
2653 switch (ifp->if_type) {
2654 case IFT_ETHER:
2655 case IFT_FDDI:
2656 case IFT_XETHER:
2657 case IFT_ISO88025:
2658 case IFT_L2VLAN:
2659 case IFT_BRIDGE:
2660 case IFT_ARCNET:
2661 case IFT_IEEE8023ADLAG:
2662 bcopy(lladdr, LLADDR(sdl), len);
2663 break;
2664 default:
2665 return (ENODEV);
2666 }
2667 /*
2668 * If the interface is already up, we need
2669 * to re-init it in order to reprogram its
2670 * address filter.
2671 */
2672 if ((ifp->if_flags & IFF_UP) != 0) {
2673 if (ifp->if_ioctl) {
2674 IFF_LOCKGIANT(ifp);
2675 ifp->if_flags &= ~IFF_UP;
2676 ifr.ifr_flags = ifp->if_flags & 0xffff;
2677 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2678 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2679 ifp->if_flags |= IFF_UP;
2680 ifr.ifr_flags = ifp->if_flags & 0xffff;
2681 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2682 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2683 IFF_UNLOCKGIANT(ifp);
2684 }
2685#ifdef INET
2686 /*
2687 * Also send gratuitous ARPs to notify other nodes about
2688 * the address change.
2689 */
2690 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2691 if (ifa->ifa_addr->sa_family == AF_INET)
2692 arp_ifinit(ifp, ifa);
2693 }
2694#endif
2695 }
2696 return (0);
2697}
2698
2699/*
2700 * The name argument must be a pointer to storage which will last as
2701 * long as the interface does. For physical devices, the result of
2702 * device_get_name(dev) is a good choice and for pseudo-devices a
2703 * static string works well.
2704 */
2705void
2706if_initname(struct ifnet *ifp, const char *name, int unit)
2707{
2708 ifp->if_dname = name;
2709 ifp->if_dunit = unit;
2710 if (unit != IF_DUNIT_NONE)
2711 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
2712 else
2713 strlcpy(ifp->if_xname, name, IFNAMSIZ);
2714}
2715
2716int
2717if_printf(struct ifnet *ifp, const char * fmt, ...)
2718{
2719 va_list ap;
2720 int retval;
2721
2722 retval = printf("%s: ", ifp->if_xname);
2723 va_start(ap, fmt);
2724 retval += vprintf(fmt, ap);
2725 va_end(ap);
2726 return (retval);
2727}
2728
2729/*
2730 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
2731 * be called without Giant. However, we often can't acquire the Giant lock
2732 * at those points; instead, we run it via a task queue that holds Giant via
2733 * if_start_deferred.
2734 *
2735 * XXXRW: We need to make sure that the ifnet isn't fully detached until any
2736 * outstanding if_start_deferred() tasks that will run after the free. This
2737 * probably means waiting in if_detach().
2738 */
2739void
2740if_start(struct ifnet *ifp)
2741{
2742
2743 if (ifp->if_flags & IFF_NEEDSGIANT) {
2744 if (mtx_owned(&Giant))
2745 (*(ifp)->if_start)(ifp);
2746 else
2747 taskqueue_enqueue(taskqueue_swi_giant,
2748 &ifp->if_starttask);
2749 } else
2750 (*(ifp)->if_start)(ifp);
2751}
2752
2753static void
2754if_start_deferred(void *context, int pending)
2755{
2756 struct ifnet *ifp;
2757
2758 GIANT_REQUIRED;
2759
2760 ifp = context;
2761 (ifp->if_start)(ifp);
2762}
2763
2764int
2765if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
2766{
2767 int active = 0;
2768
2769 IF_LOCK(ifq);
2770 if (_IF_QFULL(ifq)) {
2771 _IF_DROP(ifq);
2772 IF_UNLOCK(ifq);
2773 m_freem(m);
2774 return (0);
2775 }
2776 if (ifp != NULL) {
2777 ifp->if_obytes += m->m_pkthdr.len + adjust;
2778 if (m->m_flags & (M_BCAST|M_MCAST))
2779 ifp->if_omcasts++;
2780 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
2781 }
2782 _IF_ENQUEUE(ifq, m);
2783 IF_UNLOCK(ifq);
2784 if (ifp != NULL && !active)
2785 if_start(ifp);
2786 return (1);
2787}
2788
2789void
2790if_register_com_alloc(u_char type,
2791 if_com_alloc_t *a, if_com_free_t *f)
2792{
2793
2794 KASSERT(if_com_alloc[type] == NULL,
2795 ("if_register_com_alloc: %d already registered", type));
2796 KASSERT(if_com_free[type] == NULL,
2797 ("if_register_com_alloc: %d free already registered", type));
2798
2799 if_com_alloc[type] = a;
2800 if_com_free[type] = f;
2801}
2802
2803void
2804if_deregister_com_alloc(u_char type)
2805{
2806
2807 KASSERT(if_com_alloc[type] != NULL,
2808 ("if_deregister_com_alloc: %d not registered", type));
2809 KASSERT(if_com_free[type] != NULL,
2810 ("if_deregister_com_alloc: %d free not registered", type));
2811 if_com_alloc[type] = NULL;
2812 if_com_free[type] = NULL;
2813}