FreeBSD/Linux Kernel Cross Reference
sys/net/if.c
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 * $FreeBSD$
31 */
32
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65
66 #include <net/if.h>
67 #include <net/if_arp.h>
68 #include <net/if_clone.h>
69 #include <net/if_dl.h>
70 #include <net/if_types.h>
71 #include <net/if_var.h>
72 #include <net/radix.h>
73 #include <net/route.h>
74 #include <net/vnet.h>
75
76 #if defined(INET) || defined(INET6)
77 /*XXX*/
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_carp.h>
81 #ifdef INET6
82 #include <netinet6/in6_var.h>
83 #include <netinet6/in6_ifattach.h>
84 #endif
85 #endif
86 #ifdef INET
87 #include <netinet/if_ether.h>
88 #endif
89
90 #include <security/mac/mac_framework.h>
91
92 #ifdef COMPAT_FREEBSD32
93 #include <sys/mount.h>
94 #include <compat/freebsd32/freebsd32.h>
95 #endif
96
97 struct ifindex_entry {
98 struct ifnet *ife_ifnet;
99 };
100
101 static int slowtimo_started;
102
103 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
104 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
105
106 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
107 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
108 &ifqmaxlen, 0, "max send queue size");
109
110 /* Log link state change events */
111 static int log_link_state_change = 1;
112
113 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
114 &log_link_state_change, 0,
115 "log interface link state change events");
116
117 /* Interface description */
118 static unsigned int ifdescr_maxlen = 1024;
119 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
120 &ifdescr_maxlen, 0,
121 "administrative maximum length for interface description");
122
123 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
124
125 /* global sx for non-critical path ifdescr */
126 static struct sx ifdescr_sx;
127 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
128
129 void (*bridge_linkstate_p)(struct ifnet *ifp);
130 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
131 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
132 /* These are external hooks for CARP. */
133 void (*carp_linkstate_p)(struct ifnet *ifp);
134 #if defined(INET) || defined(INET6)
135 struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
136 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
137 struct sockaddr *sa, struct rtentry *rt);
138 #endif
139 #ifdef INET
140 int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
141 u_int8_t **);
142 #endif
143 #ifdef INET6
144 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
145 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
146 const struct in6_addr *taddr);
147 #endif
148
149 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
150
151 /*
152 * XXX: Style; these should be sorted alphabetically, and unprototyped
153 * static functions should be prototyped. Currently they are sorted by
154 * declaration order.
155 */
156 static void if_attachdomain(void *);
157 static void if_attachdomain1(struct ifnet *);
158 static int ifconf(u_long, caddr_t);
159 static void if_freemulti(struct ifmultiaddr *);
160 static void if_init(void *);
161 static void if_grow(void);
162 static void if_check(void *);
163 static void if_route(struct ifnet *, int flag, int fam);
164 static int if_setflag(struct ifnet *, int, int, int *, int);
165 static void if_slowtimo(void *);
166 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
167 static void if_unroute(struct ifnet *, int flag, int fam);
168 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
169 static int if_rtdel(struct radix_node *, void *);
170 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
171 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
172 static void do_link_state_change(void *, int);
173 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
174 static int if_getgroupmembers(struct ifgroupreq *);
175 static void if_delgroups(struct ifnet *);
176 static void if_attach_internal(struct ifnet *, int);
177 static void if_detach_internal(struct ifnet *, int);
178
179 #ifdef INET6
180 /*
181 * XXX: declare here to avoid to include many inet6 related files..
182 * should be more generalized?
183 */
184 extern void nd6_setmtu(struct ifnet *);
185 #endif
186
187 VNET_DEFINE(int, if_index);
188 int ifqmaxlen = IFQ_MAXLEN;
189 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
190 VNET_DEFINE(struct ifgrouphead, ifg_head);
191
192 static VNET_DEFINE(int, if_indexlim) = 8;
193
194 /* Table of ifnet by index. */
195 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
196
197 #define V_if_indexlim VNET(if_indexlim)
198 #define V_ifindex_table VNET(ifindex_table)
199
200 /*
201 * The global network interface list (V_ifnet) and related state (such as
202 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
203 * an rwlock. Either may be acquired shared to stablize the list, but both
204 * must be acquired writable to modify the list. This model allows us to
205 * both stablize the interface list during interrupt thread processing, but
206 * also to stablize it over long-running ioctls, without introducing priority
207 * inversions and deadlocks.
208 */
209 struct rwlock ifnet_rwlock;
210 struct sx ifnet_sxlock;
211
212 /*
213 * The allocation of network interfaces is a rather non-atomic affair; we
214 * need to select an index before we are ready to expose the interface for
215 * use, so will use this pointer value to indicate reservation.
216 */
217 #define IFNET_HOLD (void *)(uintptr_t)(-1)
218
219 static if_com_alloc_t *if_com_alloc[256];
220 static if_com_free_t *if_com_free[256];
221
222 /*
223 * System initialization
224 */
225 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL);
226
227 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
228 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
229 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
230
231 struct ifnet *
232 ifnet_byindex_locked(u_short idx)
233 {
234
235 if (idx > V_if_index)
236 return (NULL);
237 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
238 return (NULL);
239 return (V_ifindex_table[idx].ife_ifnet);
240 }
241
242 struct ifnet *
243 ifnet_byindex(u_short idx)
244 {
245 struct ifnet *ifp;
246
247 IFNET_RLOCK_NOSLEEP();
248 ifp = ifnet_byindex_locked(idx);
249 IFNET_RUNLOCK_NOSLEEP();
250 return (ifp);
251 }
252
253 struct ifnet *
254 ifnet_byindex_ref(u_short idx)
255 {
256 struct ifnet *ifp;
257
258 IFNET_RLOCK_NOSLEEP();
259 ifp = ifnet_byindex_locked(idx);
260 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
261 IFNET_RUNLOCK_NOSLEEP();
262 return (NULL);
263 }
264 if_ref(ifp);
265 IFNET_RUNLOCK_NOSLEEP();
266 return (ifp);
267 }
268
269 /*
270 * Allocate an ifindex array entry; return 0 on success or an error on
271 * failure.
272 */
273 static int
274 ifindex_alloc_locked(u_short *idxp)
275 {
276 u_short idx;
277
278 IFNET_WLOCK_ASSERT();
279
280 retry:
281 /*
282 * Try to find an empty slot below V_if_index. If we fail, take the
283 * next slot.
284 */
285 for (idx = 1; idx <= V_if_index; idx++) {
286 if (V_ifindex_table[idx].ife_ifnet == NULL)
287 break;
288 }
289
290 /* Catch if_index overflow. */
291 if (idx < 1)
292 return (ENOSPC);
293 if (idx >= V_if_indexlim) {
294 if_grow();
295 goto retry;
296 }
297 if (idx > V_if_index)
298 V_if_index = idx;
299 *idxp = idx;
300 return (0);
301 }
302
303 static void
304 ifindex_free_locked(u_short idx)
305 {
306
307 IFNET_WLOCK_ASSERT();
308
309 V_ifindex_table[idx].ife_ifnet = NULL;
310 while (V_if_index > 0 &&
311 V_ifindex_table[V_if_index].ife_ifnet == NULL)
312 V_if_index--;
313 }
314
315 static void
316 ifindex_free(u_short idx)
317 {
318
319 IFNET_WLOCK();
320 ifindex_free_locked(idx);
321 IFNET_WUNLOCK();
322 }
323
324 static void
325 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
326 {
327
328 IFNET_WLOCK_ASSERT();
329
330 V_ifindex_table[idx].ife_ifnet = ifp;
331 }
332
333 static void
334 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
335 {
336
337 IFNET_WLOCK();
338 ifnet_setbyindex_locked(idx, ifp);
339 IFNET_WUNLOCK();
340 }
341
342 struct ifaddr *
343 ifaddr_byindex(u_short idx)
344 {
345 struct ifaddr *ifa;
346
347 IFNET_RLOCK_NOSLEEP();
348 ifa = ifnet_byindex_locked(idx)->if_addr;
349 if (ifa != NULL)
350 ifa_ref(ifa);
351 IFNET_RUNLOCK_NOSLEEP();
352 return (ifa);
353 }
354
355 /*
356 * Network interface utility routines.
357 *
358 * Routines with ifa_ifwith* names take sockaddr *'s as
359 * parameters.
360 */
361
362 static void
363 vnet_if_init(const void *unused __unused)
364 {
365
366 TAILQ_INIT(&V_ifnet);
367 TAILQ_INIT(&V_ifg_head);
368 IFNET_WLOCK();
369 if_grow(); /* create initial table */
370 IFNET_WUNLOCK();
371 vnet_if_clone_init();
372 }
373 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
374 NULL);
375
376 /* ARGSUSED*/
377 static void
378 if_init(void *dummy __unused)
379 {
380
381 IFNET_LOCK_INIT();
382 if_clone_init();
383 }
384 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
385
386
387 #ifdef VIMAGE
388 static void
389 vnet_if_uninit(const void *unused __unused)
390 {
391
392 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
393 "not empty", __func__, __LINE__, &V_ifnet));
394 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
395 "not empty", __func__, __LINE__, &V_ifg_head));
396
397 free((caddr_t)V_ifindex_table, M_IFNET);
398 }
399 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
400 vnet_if_uninit, NULL);
401 #endif
402
403 static void
404 if_grow(void)
405 {
406 int oldlim;
407 u_int n;
408 struct ifindex_entry *e;
409
410 IFNET_WLOCK_ASSERT();
411 oldlim = V_if_indexlim;
412 IFNET_WUNLOCK();
413 n = (oldlim << 1) * sizeof(*e);
414 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
415 IFNET_WLOCK();
416 if (V_if_indexlim != oldlim) {
417 free(e, M_IFNET);
418 return;
419 }
420 if (V_ifindex_table != NULL) {
421 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
422 free((caddr_t)V_ifindex_table, M_IFNET);
423 }
424 V_if_indexlim <<= 1;
425 V_ifindex_table = e;
426 }
427
428 static void
429 if_check(void *dummy __unused)
430 {
431
432 /*
433 * If at least one interface added during boot uses
434 * if_watchdog then start the timer.
435 */
436 if (slowtimo_started)
437 if_slowtimo(0);
438 }
439
440 /*
441 * Allocate a struct ifnet and an index for an interface. A layer 2
442 * common structure will also be allocated if an allocation routine is
443 * registered for the passed type.
444 */
445 struct ifnet *
446 if_alloc(u_char type)
447 {
448 struct ifnet *ifp;
449 u_short idx;
450
451 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
452 IFNET_WLOCK();
453 if (ifindex_alloc_locked(&idx) != 0) {
454 IFNET_WUNLOCK();
455 free(ifp, M_IFNET);
456 return (NULL);
457 }
458 ifnet_setbyindex_locked(idx, IFNET_HOLD);
459 IFNET_WUNLOCK();
460 ifp->if_index = idx;
461 ifp->if_type = type;
462 ifp->if_alloctype = type;
463 if (if_com_alloc[type] != NULL) {
464 ifp->if_l2com = if_com_alloc[type](type, ifp);
465 if (ifp->if_l2com == NULL) {
466 free(ifp, M_IFNET);
467 ifindex_free(idx);
468 return (NULL);
469 }
470 }
471
472 IF_ADDR_LOCK_INIT(ifp);
473 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
474 ifp->if_afdata_initialized = 0;
475 IF_AFDATA_LOCK_INIT(ifp);
476 TAILQ_INIT(&ifp->if_addrhead);
477 TAILQ_INIT(&ifp->if_prefixhead);
478 TAILQ_INIT(&ifp->if_multiaddrs);
479 TAILQ_INIT(&ifp->if_groups);
480 #ifdef MAC
481 mac_ifnet_init(ifp);
482 #endif
483 ifq_init(&ifp->if_snd, ifp);
484
485 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
486 ifnet_setbyindex(ifp->if_index, ifp);
487 return (ifp);
488 }
489
490 /*
491 * Do the actual work of freeing a struct ifnet, and layer 2 common
492 * structure. This call is made when the last reference to an
493 * interface is released.
494 */
495 static void
496 if_free_internal(struct ifnet *ifp)
497 {
498
499 KASSERT((ifp->if_flags & IFF_DYING),
500 ("if_free_internal: interface not dying"));
501
502 if (if_com_free[ifp->if_alloctype] != NULL)
503 if_com_free[ifp->if_alloctype](ifp->if_l2com,
504 ifp->if_alloctype);
505
506 #ifdef MAC
507 mac_ifnet_destroy(ifp);
508 #endif /* MAC */
509 if (ifp->if_description != NULL)
510 free(ifp->if_description, M_IFDESCR);
511 IF_AFDATA_DESTROY(ifp);
512 IF_ADDR_LOCK_DESTROY(ifp);
513 ifq_delete(&ifp->if_snd);
514 free(ifp, M_IFNET);
515 }
516
517 /*
518 * This version should only be called by intefaces that switch their type
519 * after calling if_alloc(). if_free_type() will go away again now that we
520 * have if_alloctype to cache the original allocation type. For now, assert
521 * that they match, since we require that in practice.
522 */
523 void
524 if_free_type(struct ifnet *ifp, u_char type)
525 {
526
527 KASSERT(ifp->if_alloctype == type,
528 ("if_free_type: type (%d) != alloctype (%d)", type,
529 ifp->if_alloctype));
530
531 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
532
533 IFNET_WLOCK();
534 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
535 ("%s: freeing unallocated ifnet", ifp->if_xname));
536
537 ifindex_free_locked(ifp->if_index);
538 IFNET_WUNLOCK();
539
540 if (!refcount_release(&ifp->if_refcount))
541 return;
542 if_free_internal(ifp);
543 }
544
545 /*
546 * This is the normal version of if_free(), used by device drivers to free a
547 * detached network interface. The contents of if_free_type() will move into
548 * here when if_free_type() goes away.
549 */
550 void
551 if_free(struct ifnet *ifp)
552 {
553
554 if_free_type(ifp, ifp->if_alloctype);
555 }
556
557 /*
558 * Interfaces to keep an ifnet type-stable despite the possibility of the
559 * driver calling if_free(). If there are additional references, we defer
560 * freeing the underlying data structure.
561 */
562 void
563 if_ref(struct ifnet *ifp)
564 {
565
566 /* We don't assert the ifnet list lock here, but arguably should. */
567 refcount_acquire(&ifp->if_refcount);
568 }
569
570 void
571 if_rele(struct ifnet *ifp)
572 {
573
574 if (!refcount_release(&ifp->if_refcount))
575 return;
576 if_free_internal(ifp);
577 }
578
579 void
580 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
581 {
582
583 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
584
585 if (ifq->ifq_maxlen == 0)
586 ifq->ifq_maxlen = ifqmaxlen;
587
588 ifq->altq_type = 0;
589 ifq->altq_disc = NULL;
590 ifq->altq_flags &= ALTQF_CANTCHANGE;
591 ifq->altq_tbr = NULL;
592 ifq->altq_ifp = ifp;
593 }
594
595 void
596 ifq_delete(struct ifaltq *ifq)
597 {
598 mtx_destroy(&ifq->ifq_mtx);
599 }
600
601 /*
602 * Perform generic interface initalization tasks and attach the interface
603 * to the list of "active" interfaces. If vmove flag is set on entry
604 * to if_attach_internal(), perform only a limited subset of initialization
605 * tasks, given that we are moving from one vnet to another an ifnet which
606 * has already been fully initialized.
607 *
608 * XXX:
609 * - The decision to return void and thus require this function to
610 * succeed is questionable.
611 * - We should probably do more sanity checking. For instance we don't
612 * do anything to insure if_xname is unique or non-empty.
613 */
614 void
615 if_attach(struct ifnet *ifp)
616 {
617
618 if_attach_internal(ifp, 0);
619 }
620
621 static void
622 if_attach_internal(struct ifnet *ifp, int vmove)
623 {
624 unsigned socksize, ifasize;
625 int namelen, masklen;
626 struct sockaddr_dl *sdl;
627 struct ifaddr *ifa;
628
629 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
630 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
631 ifp->if_xname);
632
633 #ifdef VIMAGE
634 ifp->if_vnet = curvnet;
635 if (ifp->if_home_vnet == NULL)
636 ifp->if_home_vnet = curvnet;
637 #endif
638
639 if_addgroup(ifp, IFG_ALL);
640
641 getmicrotime(&ifp->if_lastchange);
642 ifp->if_data.ifi_epoch = time_uptime;
643 ifp->if_data.ifi_datalen = sizeof(struct if_data);
644
645 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
646 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
647 ("transmit and qflush must both either be set or both be NULL"));
648 if (ifp->if_transmit == NULL) {
649 ifp->if_transmit = if_transmit;
650 ifp->if_qflush = if_qflush;
651 }
652
653 if (!vmove) {
654 #ifdef MAC
655 mac_ifnet_create(ifp);
656 #endif
657
658 /*
659 * Create a Link Level name for this device.
660 */
661 namelen = strlen(ifp->if_xname);
662 /*
663 * Always save enough space for any possiable name so we
664 * can do a rename in place later.
665 */
666 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
667 socksize = masklen + ifp->if_addrlen;
668 if (socksize < sizeof(*sdl))
669 socksize = sizeof(*sdl);
670 socksize = roundup2(socksize, sizeof(long));
671 ifasize = sizeof(*ifa) + 2 * socksize;
672 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
673 ifa_init(ifa);
674 sdl = (struct sockaddr_dl *)(ifa + 1);
675 sdl->sdl_len = socksize;
676 sdl->sdl_family = AF_LINK;
677 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
678 sdl->sdl_nlen = namelen;
679 sdl->sdl_index = ifp->if_index;
680 sdl->sdl_type = ifp->if_type;
681 ifp->if_addr = ifa;
682 ifa->ifa_ifp = ifp;
683 ifa->ifa_rtrequest = link_rtrequest;
684 ifa->ifa_addr = (struct sockaddr *)sdl;
685 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
686 ifa->ifa_netmask = (struct sockaddr *)sdl;
687 sdl->sdl_len = masklen;
688 while (namelen != 0)
689 sdl->sdl_data[--namelen] = 0xff;
690 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
691 /* Reliably crash if used uninitialized. */
692 ifp->if_broadcastaddr = NULL;
693 }
694 #ifdef VIMAGE
695 else {
696 /*
697 * Update the interface index in the link layer address
698 * of the interface.
699 */
700 for (ifa = ifp->if_addr; ifa != NULL;
701 ifa = TAILQ_NEXT(ifa, ifa_link)) {
702 if (ifa->ifa_addr->sa_family == AF_LINK) {
703 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
704 sdl->sdl_index = ifp->if_index;
705 }
706 }
707 }
708 #endif
709
710 IFNET_WLOCK();
711 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
712 #ifdef VIMAGE
713 curvnet->vnet_ifcnt++;
714 #endif
715 IFNET_WUNLOCK();
716
717 if (domain_init_status >= 2)
718 if_attachdomain1(ifp);
719
720 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
721 if (IS_DEFAULT_VNET(curvnet))
722 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
723
724 /* Announce the interface. */
725 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
726
727 if (!vmove && ifp->if_watchdog != NULL) {
728 if_printf(ifp,
729 "WARNING: using obsoleted if_watchdog interface\n");
730
731 /*
732 * Note that we need if_slowtimo(). If this happens after
733 * boot, then call if_slowtimo() directly.
734 */
735 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold)
736 if_slowtimo(0);
737 }
738 }
739
740 static void
741 if_attachdomain(void *dummy)
742 {
743 struct ifnet *ifp;
744 int s;
745
746 s = splnet();
747 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
748 if_attachdomain1(ifp);
749 splx(s);
750 }
751 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
752 if_attachdomain, NULL);
753
754 static void
755 if_attachdomain1(struct ifnet *ifp)
756 {
757 struct domain *dp;
758 int s;
759
760 s = splnet();
761
762 /*
763 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
764 * cannot lock ifp->if_afdata initialization, entirely.
765 */
766 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
767 splx(s);
768 return;
769 }
770 if (ifp->if_afdata_initialized >= domain_init_status) {
771 IF_AFDATA_UNLOCK(ifp);
772 splx(s);
773 printf("if_attachdomain called more than once on %s\n",
774 ifp->if_xname);
775 return;
776 }
777 ifp->if_afdata_initialized = domain_init_status;
778 IF_AFDATA_UNLOCK(ifp);
779
780 /* address family dependent data region */
781 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
782 for (dp = domains; dp; dp = dp->dom_next) {
783 if (dp->dom_ifattach)
784 ifp->if_afdata[dp->dom_family] =
785 (*dp->dom_ifattach)(ifp);
786 }
787
788 splx(s);
789 }
790
791 /*
792 * Remove any unicast or broadcast network addresses from an interface.
793 */
794 void
795 if_purgeaddrs(struct ifnet *ifp)
796 {
797 struct ifaddr *ifa, *next;
798
799 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
800 if (ifa->ifa_addr->sa_family == AF_LINK)
801 continue;
802 #ifdef INET
803 /* XXX: Ugly!! ad hoc just for INET */
804 if (ifa->ifa_addr->sa_family == AF_INET) {
805 struct ifaliasreq ifr;
806
807 bzero(&ifr, sizeof(ifr));
808 ifr.ifra_addr = *ifa->ifa_addr;
809 if (ifa->ifa_dstaddr)
810 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
811 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
812 NULL) == 0)
813 continue;
814 }
815 #endif /* INET */
816 #ifdef INET6
817 if (ifa->ifa_addr->sa_family == AF_INET6) {
818 in6_purgeaddr(ifa);
819 /* ifp_addrhead is already updated */
820 continue;
821 }
822 #endif /* INET6 */
823 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
824 ifa_free(ifa);
825 }
826 }
827
828 /*
829 * Remove any multicast network addresses from an interface when an ifnet
830 * is going away.
831 */
832 static void
833 if_purgemaddrs(struct ifnet *ifp)
834 {
835 struct ifmultiaddr *ifma;
836 struct ifmultiaddr *next;
837
838 IF_ADDR_WLOCK(ifp);
839 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
840 if_delmulti_locked(ifp, ifma, 1);
841 IF_ADDR_WUNLOCK(ifp);
842 }
843
844 /*
845 * Detach an interface, removing it from the list of "active" interfaces.
846 * If vmove flag is set on entry to if_detach_internal(), perform only a
847 * limited subset of cleanup tasks, given that we are moving an ifnet from
848 * one vnet to another, where it must be fully operational.
849 *
850 * XXXRW: There are some significant questions about event ordering, and
851 * how to prevent things from starting to use the interface during detach.
852 */
853 void
854 if_detach(struct ifnet *ifp)
855 {
856
857 if_detach_internal(ifp, 0);
858 }
859
860 static void
861 if_detach_internal(struct ifnet *ifp, int vmove)
862 {
863 struct ifaddr *ifa;
864 struct radix_node_head *rnh;
865 int i, j;
866 struct domain *dp;
867 struct ifnet *iter;
868 int found = 0;
869
870 IFNET_WLOCK();
871 TAILQ_FOREACH(iter, &V_ifnet, if_link)
872 if (iter == ifp) {
873 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
874 found = 1;
875 break;
876 }
877 #ifdef VIMAGE
878 if (found)
879 curvnet->vnet_ifcnt--;
880 #endif
881 IFNET_WUNLOCK();
882 if (!found) {
883 if (vmove)
884 panic("%s: ifp=%p not on the ifnet tailq %p",
885 __func__, ifp, &V_ifnet);
886 else
887 return; /* XXX this should panic as well? */
888 }
889
890 /*
891 * Remove/wait for pending events.
892 */
893 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
894
895 /*
896 * Remove routes and flush queues.
897 */
898 if_down(ifp);
899 #ifdef ALTQ
900 if (ALTQ_IS_ENABLED(&ifp->if_snd))
901 altq_disable(&ifp->if_snd);
902 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
903 altq_detach(&ifp->if_snd);
904 #endif
905
906 if_purgeaddrs(ifp);
907
908 #ifdef INET
909 in_ifdetach(ifp);
910 #endif
911
912 #ifdef INET6
913 /*
914 * Remove all IPv6 kernel structs related to ifp. This should be done
915 * before removing routing entries below, since IPv6 interface direct
916 * routes are expected to be removed by the IPv6-specific kernel API.
917 * Otherwise, the kernel will detect some inconsistency and bark it.
918 */
919 in6_ifdetach(ifp);
920 #endif
921 if_purgemaddrs(ifp);
922
923 if (!vmove) {
924 /*
925 * Prevent further calls into the device driver via ifnet.
926 */
927 if_dead(ifp);
928
929 /*
930 * Remove link ifaddr pointer and maybe decrement if_index.
931 * Clean up all addresses.
932 */
933 ifp->if_addr = NULL;
934
935 /* We can now free link ifaddr. */
936 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
937 ifa = TAILQ_FIRST(&ifp->if_addrhead);
938 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
939 ifa_free(ifa);
940 }
941 }
942
943 /*
944 * Delete all remaining routes using this interface
945 * Unfortuneatly the only way to do this is to slog through
946 * the entire routing table looking for routes which point
947 * to this interface...oh well...
948 */
949 for (i = 1; i <= AF_MAX; i++) {
950 for (j = 0; j < rt_numfibs; j++) {
951 rnh = rt_tables_get_rnh(j, i);
952 if (rnh == NULL)
953 continue;
954 RADIX_NODE_HEAD_LOCK(rnh);
955 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
956 RADIX_NODE_HEAD_UNLOCK(rnh);
957 }
958 }
959
960 /* Announce that the interface is gone. */
961 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
962 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
963 if (IS_DEFAULT_VNET(curvnet))
964 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
965 if_delgroups(ifp);
966
967 /*
968 * We cannot hold the lock over dom_ifdetach calls as they might
969 * sleep, for example trying to drain a callout, thus open up the
970 * theoretical race with re-attaching.
971 */
972 IF_AFDATA_LOCK(ifp);
973 i = ifp->if_afdata_initialized;
974 ifp->if_afdata_initialized = 0;
975 IF_AFDATA_UNLOCK(ifp);
976 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
977 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
978 (*dp->dom_ifdetach)(ifp,
979 ifp->if_afdata[dp->dom_family]);
980 }
981 }
982
983 #ifdef VIMAGE
984 /*
985 * if_vmove() performs a limited version of if_detach() in current
986 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
987 * An attempt is made to shrink if_index in current vnet, find an
988 * unused if_index in target vnet and calls if_grow() if necessary,
989 * and finally find an unused if_xname for the target vnet.
990 */
991 void
992 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
993 {
994 u_short idx;
995
996 /*
997 * Detach from current vnet, but preserve LLADDR info, do not
998 * mark as dead etc. so that the ifnet can be reattached later.
999 */
1000 if_detach_internal(ifp, 1);
1001
1002 /*
1003 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1004 * the if_index for that vnet if possible.
1005 *
1006 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1007 * or we'd lock on one vnet and unlock on another.
1008 */
1009 IFNET_WLOCK();
1010 ifindex_free_locked(ifp->if_index);
1011 IFNET_WUNLOCK();
1012
1013 /*
1014 * Perform interface-specific reassignment tasks, if provided by
1015 * the driver.
1016 */
1017 if (ifp->if_reassign != NULL)
1018 ifp->if_reassign(ifp, new_vnet, NULL);
1019
1020 /*
1021 * Switch to the context of the target vnet.
1022 */
1023 CURVNET_SET_QUIET(new_vnet);
1024
1025 IFNET_WLOCK();
1026 if (ifindex_alloc_locked(&idx) != 0) {
1027 IFNET_WUNLOCK();
1028 panic("if_index overflow");
1029 }
1030 ifp->if_index = idx;
1031 ifnet_setbyindex_locked(ifp->if_index, ifp);
1032 IFNET_WUNLOCK();
1033
1034 if_attach_internal(ifp, 1);
1035
1036 CURVNET_RESTORE();
1037 }
1038
1039 /*
1040 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1041 */
1042 static int
1043 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1044 {
1045 struct prison *pr;
1046 struct ifnet *difp;
1047
1048 /* Try to find the prison within our visibility. */
1049 sx_slock(&allprison_lock);
1050 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1051 sx_sunlock(&allprison_lock);
1052 if (pr == NULL)
1053 return (ENXIO);
1054 prison_hold_locked(pr);
1055 mtx_unlock(&pr->pr_mtx);
1056
1057 /* Do not try to move the iface from and to the same prison. */
1058 if (pr->pr_vnet == ifp->if_vnet) {
1059 prison_free(pr);
1060 return (EEXIST);
1061 }
1062
1063 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1064 /* XXX Lock interfaces to avoid races. */
1065 CURVNET_SET_QUIET(pr->pr_vnet);
1066 difp = ifunit(ifname);
1067 CURVNET_RESTORE();
1068 if (difp != NULL) {
1069 prison_free(pr);
1070 return (EEXIST);
1071 }
1072
1073 /* Move the interface into the child jail/vnet. */
1074 if_vmove(ifp, pr->pr_vnet);
1075
1076 /* Report the new if_xname back to the userland. */
1077 sprintf(ifname, "%s", ifp->if_xname);
1078
1079 prison_free(pr);
1080 return (0);
1081 }
1082
1083 static int
1084 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1085 {
1086 struct prison *pr;
1087 struct vnet *vnet_dst;
1088 struct ifnet *ifp;
1089
1090 /* Try to find the prison within our visibility. */
1091 sx_slock(&allprison_lock);
1092 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1093 sx_sunlock(&allprison_lock);
1094 if (pr == NULL)
1095 return (ENXIO);
1096 prison_hold_locked(pr);
1097 mtx_unlock(&pr->pr_mtx);
1098
1099 /* Make sure the named iface exists in the source prison/vnet. */
1100 CURVNET_SET(pr->pr_vnet);
1101 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1102 if (ifp == NULL) {
1103 CURVNET_RESTORE();
1104 prison_free(pr);
1105 return (ENXIO);
1106 }
1107
1108 /* Do not try to move the iface from and to the same prison. */
1109 vnet_dst = TD_TO_VNET(td);
1110 if (vnet_dst == ifp->if_vnet) {
1111 CURVNET_RESTORE();
1112 prison_free(pr);
1113 return (EEXIST);
1114 }
1115
1116 /* Get interface back from child jail/vnet. */
1117 if_vmove(ifp, vnet_dst);
1118 CURVNET_RESTORE();
1119
1120 /* Report the new if_xname back to the userland. */
1121 sprintf(ifname, "%s", ifp->if_xname);
1122
1123 prison_free(pr);
1124 return (0);
1125 }
1126 #endif /* VIMAGE */
1127
1128 /*
1129 * Add a group to an interface
1130 */
1131 int
1132 if_addgroup(struct ifnet *ifp, const char *groupname)
1133 {
1134 struct ifg_list *ifgl;
1135 struct ifg_group *ifg = NULL;
1136 struct ifg_member *ifgm;
1137
1138 if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
1139 groupname[strlen(groupname) - 1] <= '9')
1140 return (EINVAL);
1141
1142 IFNET_WLOCK();
1143 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1144 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1145 IFNET_WUNLOCK();
1146 return (EEXIST);
1147 }
1148
1149 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1150 M_NOWAIT)) == NULL) {
1151 IFNET_WUNLOCK();
1152 return (ENOMEM);
1153 }
1154
1155 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1156 M_TEMP, M_NOWAIT)) == NULL) {
1157 free(ifgl, M_TEMP);
1158 IFNET_WUNLOCK();
1159 return (ENOMEM);
1160 }
1161
1162 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1163 if (!strcmp(ifg->ifg_group, groupname))
1164 break;
1165
1166 if (ifg == NULL) {
1167 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1168 M_TEMP, M_NOWAIT)) == NULL) {
1169 free(ifgl, M_TEMP);
1170 free(ifgm, M_TEMP);
1171 IFNET_WUNLOCK();
1172 return (ENOMEM);
1173 }
1174 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1175 ifg->ifg_refcnt = 0;
1176 TAILQ_INIT(&ifg->ifg_members);
1177 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1178 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1179 }
1180
1181 ifg->ifg_refcnt++;
1182 ifgl->ifgl_group = ifg;
1183 ifgm->ifgm_ifp = ifp;
1184
1185 IF_ADDR_WLOCK(ifp);
1186 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1187 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1188 IF_ADDR_WUNLOCK(ifp);
1189
1190 IFNET_WUNLOCK();
1191
1192 EVENTHANDLER_INVOKE(group_change_event, groupname);
1193
1194 return (0);
1195 }
1196
1197 /*
1198 * Remove a group from an interface
1199 */
1200 int
1201 if_delgroup(struct ifnet *ifp, const char *groupname)
1202 {
1203 struct ifg_list *ifgl;
1204 struct ifg_member *ifgm;
1205
1206 IFNET_WLOCK();
1207 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1208 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1209 break;
1210 if (ifgl == NULL) {
1211 IFNET_WUNLOCK();
1212 return (ENOENT);
1213 }
1214
1215 IF_ADDR_WLOCK(ifp);
1216 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1217 IF_ADDR_WUNLOCK(ifp);
1218
1219 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1220 if (ifgm->ifgm_ifp == ifp)
1221 break;
1222
1223 if (ifgm != NULL) {
1224 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1225 free(ifgm, M_TEMP);
1226 }
1227
1228 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1229 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1230 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1231 free(ifgl->ifgl_group, M_TEMP);
1232 }
1233 IFNET_WUNLOCK();
1234
1235 free(ifgl, M_TEMP);
1236
1237 EVENTHANDLER_INVOKE(group_change_event, groupname);
1238
1239 return (0);
1240 }
1241
1242 /*
1243 * Remove an interface from all groups
1244 */
1245 static void
1246 if_delgroups(struct ifnet *ifp)
1247 {
1248 struct ifg_list *ifgl;
1249 struct ifg_member *ifgm;
1250 char groupname[IFNAMSIZ];
1251
1252 IFNET_WLOCK();
1253 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1254 ifgl = TAILQ_FIRST(&ifp->if_groups);
1255
1256 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1257
1258 IF_ADDR_WLOCK(ifp);
1259 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1260 IF_ADDR_WUNLOCK(ifp);
1261
1262 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1263 if (ifgm->ifgm_ifp == ifp)
1264 break;
1265
1266 if (ifgm != NULL) {
1267 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1268 ifgm_next);
1269 free(ifgm, M_TEMP);
1270 }
1271
1272 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1273 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1274 EVENTHANDLER_INVOKE(group_detach_event,
1275 ifgl->ifgl_group);
1276 free(ifgl->ifgl_group, M_TEMP);
1277 }
1278 IFNET_WUNLOCK();
1279
1280 free(ifgl, M_TEMP);
1281
1282 EVENTHANDLER_INVOKE(group_change_event, groupname);
1283
1284 IFNET_WLOCK();
1285 }
1286 IFNET_WUNLOCK();
1287 }
1288
1289 /*
1290 * Stores all groups from an interface in memory pointed
1291 * to by data
1292 */
1293 static int
1294 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1295 {
1296 int len, error;
1297 struct ifg_list *ifgl;
1298 struct ifg_req ifgrq, *ifgp;
1299 struct ifgroupreq *ifgr = data;
1300
1301 if (ifgr->ifgr_len == 0) {
1302 IF_ADDR_RLOCK(ifp);
1303 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1304 ifgr->ifgr_len += sizeof(struct ifg_req);
1305 IF_ADDR_RUNLOCK(ifp);
1306 return (0);
1307 }
1308
1309 len = ifgr->ifgr_len;
1310 ifgp = ifgr->ifgr_groups;
1311 /* XXX: wire */
1312 IF_ADDR_RLOCK(ifp);
1313 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1314 if (len < sizeof(ifgrq)) {
1315 IF_ADDR_RUNLOCK(ifp);
1316 return (EINVAL);
1317 }
1318 bzero(&ifgrq, sizeof ifgrq);
1319 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1320 sizeof(ifgrq.ifgrq_group));
1321 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1322 IF_ADDR_RUNLOCK(ifp);
1323 return (error);
1324 }
1325 len -= sizeof(ifgrq);
1326 ifgp++;
1327 }
1328 IF_ADDR_RUNLOCK(ifp);
1329
1330 return (0);
1331 }
1332
1333 /*
1334 * Stores all members of a group in memory pointed to by data
1335 */
1336 static int
1337 if_getgroupmembers(struct ifgroupreq *data)
1338 {
1339 struct ifgroupreq *ifgr = data;
1340 struct ifg_group *ifg;
1341 struct ifg_member *ifgm;
1342 struct ifg_req ifgrq, *ifgp;
1343 int len, error;
1344
1345 IFNET_RLOCK();
1346 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1347 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1348 break;
1349 if (ifg == NULL) {
1350 IFNET_RUNLOCK();
1351 return (ENOENT);
1352 }
1353
1354 if (ifgr->ifgr_len == 0) {
1355 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1356 ifgr->ifgr_len += sizeof(ifgrq);
1357 IFNET_RUNLOCK();
1358 return (0);
1359 }
1360
1361 len = ifgr->ifgr_len;
1362 ifgp = ifgr->ifgr_groups;
1363 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1364 if (len < sizeof(ifgrq)) {
1365 IFNET_RUNLOCK();
1366 return (EINVAL);
1367 }
1368 bzero(&ifgrq, sizeof ifgrq);
1369 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1370 sizeof(ifgrq.ifgrq_member));
1371 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1372 IFNET_RUNLOCK();
1373 return (error);
1374 }
1375 len -= sizeof(ifgrq);
1376 ifgp++;
1377 }
1378 IFNET_RUNLOCK();
1379
1380 return (0);
1381 }
1382
1383 /*
1384 * Delete Routes for a Network Interface
1385 *
1386 * Called for each routing entry via the rnh->rnh_walktree() call above
1387 * to delete all route entries referencing a detaching network interface.
1388 *
1389 * Arguments:
1390 * rn pointer to node in the routing table
1391 * arg argument passed to rnh->rnh_walktree() - detaching interface
1392 *
1393 * Returns:
1394 * 0 successful
1395 * errno failed - reason indicated
1396 *
1397 */
1398 static int
1399 if_rtdel(struct radix_node *rn, void *arg)
1400 {
1401 struct rtentry *rt = (struct rtentry *)rn;
1402 struct ifnet *ifp = arg;
1403 int err;
1404
1405 if (rt->rt_ifp == ifp) {
1406
1407 /*
1408 * Protect (sorta) against walktree recursion problems
1409 * with cloned routes
1410 */
1411 if ((rt->rt_flags & RTF_UP) == 0)
1412 return (0);
1413
1414 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1415 rt_mask(rt),
1416 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1417 (struct rtentry **) NULL, rt->rt_fibnum);
1418 if (err) {
1419 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1420 }
1421 }
1422
1423 return (0);
1424 }
1425
1426 /*
1427 * Wrapper functions for struct ifnet address list locking macros. These are
1428 * used by kernel modules to avoid encoding programming interface or binary
1429 * interface assumptions that may be violated when kernel-internal locking
1430 * approaches change.
1431 */
1432 void
1433 if_addr_rlock(struct ifnet *ifp)
1434 {
1435
1436 IF_ADDR_RLOCK(ifp);
1437 }
1438
1439 void
1440 if_addr_runlock(struct ifnet *ifp)
1441 {
1442
1443 IF_ADDR_RUNLOCK(ifp);
1444 }
1445
1446 void
1447 if_maddr_rlock(struct ifnet *ifp)
1448 {
1449
1450 IF_ADDR_RLOCK(ifp);
1451 }
1452
1453 void
1454 if_maddr_runlock(struct ifnet *ifp)
1455 {
1456
1457 IF_ADDR_RUNLOCK(ifp);
1458 }
1459
1460 /*
1461 * Reference count functions for ifaddrs.
1462 */
1463 void
1464 ifa_init(struct ifaddr *ifa)
1465 {
1466
1467 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1468 refcount_init(&ifa->ifa_refcnt, 1);
1469 }
1470
1471 void
1472 ifa_ref(struct ifaddr *ifa)
1473 {
1474
1475 refcount_acquire(&ifa->ifa_refcnt);
1476 }
1477
1478 void
1479 ifa_free(struct ifaddr *ifa)
1480 {
1481
1482 if (refcount_release(&ifa->ifa_refcnt)) {
1483 mtx_destroy(&ifa->ifa_mtx);
1484 free(ifa, M_IFADDR);
1485 }
1486 }
1487
1488 int
1489 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1490 {
1491 int error = 0;
1492 struct rtentry *rt = NULL;
1493 struct rt_addrinfo info;
1494 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1495
1496 bzero(&info, sizeof(info));
1497 info.rti_ifp = V_loif;
1498 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1499 info.rti_info[RTAX_DST] = ia;
1500 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1501 error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1502
1503 if (error == 0 && rt != NULL) {
1504 RT_LOCK(rt);
1505 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1506 ifa->ifa_ifp->if_type;
1507 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1508 ifa->ifa_ifp->if_index;
1509 RT_REMREF(rt);
1510 RT_UNLOCK(rt);
1511 } else if (error != 0)
1512 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1513
1514 return (error);
1515 }
1516
1517 int
1518 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1519 {
1520 int error = 0;
1521 struct rt_addrinfo info;
1522 struct sockaddr_dl null_sdl;
1523
1524 bzero(&null_sdl, sizeof(null_sdl));
1525 null_sdl.sdl_len = sizeof(null_sdl);
1526 null_sdl.sdl_family = AF_LINK;
1527 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1528 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1529 bzero(&info, sizeof(info));
1530 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1531 info.rti_info[RTAX_DST] = ia;
1532 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1533 error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1534
1535 if (error != 0)
1536 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1537
1538 return (error);
1539 }
1540
1541 /*
1542 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1543 * structs used to represent other address families, it is necessary
1544 * to perform a different comparison.
1545 */
1546
1547 #define sa_equal(a1, a2) \
1548 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1549
1550 #define sa_dl_equal(a1, a2) \
1551 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1552 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1553 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1554 LLADDR((struct sockaddr_dl *)(a2)), \
1555 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1556
1557 /*
1558 * Locate an interface based on a complete address.
1559 */
1560 /*ARGSUSED*/
1561 static struct ifaddr *
1562 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1563 {
1564 struct ifnet *ifp;
1565 struct ifaddr *ifa;
1566
1567 IFNET_RLOCK_NOSLEEP();
1568 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1569 IF_ADDR_RLOCK(ifp);
1570 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1571 if (ifa->ifa_addr->sa_family != addr->sa_family)
1572 continue;
1573 if (sa_equal(addr, ifa->ifa_addr)) {
1574 if (getref)
1575 ifa_ref(ifa);
1576 IF_ADDR_RUNLOCK(ifp);
1577 goto done;
1578 }
1579 /* IP6 doesn't have broadcast */
1580 if ((ifp->if_flags & IFF_BROADCAST) &&
1581 ifa->ifa_broadaddr &&
1582 ifa->ifa_broadaddr->sa_len != 0 &&
1583 sa_equal(ifa->ifa_broadaddr, addr)) {
1584 if (getref)
1585 ifa_ref(ifa);
1586 IF_ADDR_RUNLOCK(ifp);
1587 goto done;
1588 }
1589 }
1590 IF_ADDR_RUNLOCK(ifp);
1591 }
1592 ifa = NULL;
1593 done:
1594 IFNET_RUNLOCK_NOSLEEP();
1595 return (ifa);
1596 }
1597
1598 struct ifaddr *
1599 ifa_ifwithaddr(struct sockaddr *addr)
1600 {
1601
1602 return (ifa_ifwithaddr_internal(addr, 1));
1603 }
1604
1605 int
1606 ifa_ifwithaddr_check(struct sockaddr *addr)
1607 {
1608
1609 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1610 }
1611
1612 /*
1613 * Locate an interface based on the broadcast address.
1614 */
1615 /* ARGSUSED */
1616 struct ifaddr *
1617 ifa_ifwithbroadaddr(struct sockaddr *addr)
1618 {
1619 struct ifnet *ifp;
1620 struct ifaddr *ifa;
1621
1622 IFNET_RLOCK_NOSLEEP();
1623 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1624 IF_ADDR_RLOCK(ifp);
1625 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1626 if (ifa->ifa_addr->sa_family != addr->sa_family)
1627 continue;
1628 if ((ifp->if_flags & IFF_BROADCAST) &&
1629 ifa->ifa_broadaddr &&
1630 ifa->ifa_broadaddr->sa_len != 0 &&
1631 sa_equal(ifa->ifa_broadaddr, addr)) {
1632 ifa_ref(ifa);
1633 IF_ADDR_RUNLOCK(ifp);
1634 goto done;
1635 }
1636 }
1637 IF_ADDR_RUNLOCK(ifp);
1638 }
1639 ifa = NULL;
1640 done:
1641 IFNET_RUNLOCK_NOSLEEP();
1642 return (ifa);
1643 }
1644
1645 /*
1646 * Locate the point to point interface with a given destination address.
1647 */
1648 /*ARGSUSED*/
1649 struct ifaddr *
1650 ifa_ifwithdstaddr(struct sockaddr *addr)
1651 {
1652 struct ifnet *ifp;
1653 struct ifaddr *ifa;
1654
1655 IFNET_RLOCK_NOSLEEP();
1656 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1657 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1658 continue;
1659 IF_ADDR_RLOCK(ifp);
1660 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1661 if (ifa->ifa_addr->sa_family != addr->sa_family)
1662 continue;
1663 if (ifa->ifa_dstaddr != NULL &&
1664 sa_equal(addr, ifa->ifa_dstaddr)) {
1665 ifa_ref(ifa);
1666 IF_ADDR_RUNLOCK(ifp);
1667 goto done;
1668 }
1669 }
1670 IF_ADDR_RUNLOCK(ifp);
1671 }
1672 ifa = NULL;
1673 done:
1674 IFNET_RUNLOCK_NOSLEEP();
1675 return (ifa);
1676 }
1677
1678 /*
1679 * Find an interface on a specific network. If many, choice
1680 * is most specific found.
1681 */
1682 struct ifaddr *
1683 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1684 {
1685 struct ifnet *ifp;
1686 struct ifaddr *ifa;
1687 struct ifaddr *ifa_maybe = NULL;
1688 u_int af = addr->sa_family;
1689 char *addr_data = addr->sa_data, *cplim;
1690
1691 /*
1692 * AF_LINK addresses can be looked up directly by their index number,
1693 * so do that if we can.
1694 */
1695 if (af == AF_LINK) {
1696 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1697 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1698 return (ifaddr_byindex(sdl->sdl_index));
1699 }
1700
1701 /*
1702 * Scan though each interface, looking for ones that have addresses
1703 * in this address family. Maintain a reference on ifa_maybe once
1704 * we find one, as we release the IF_ADDR_RLOCK() that kept it stable
1705 * when we move onto the next interface.
1706 */
1707 IFNET_RLOCK_NOSLEEP();
1708 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1709 IF_ADDR_RLOCK(ifp);
1710 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1711 char *cp, *cp2, *cp3;
1712
1713 if (ifa->ifa_addr->sa_family != af)
1714 next: continue;
1715 if (af == AF_INET &&
1716 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1717 /*
1718 * This is a bit broken as it doesn't
1719 * take into account that the remote end may
1720 * be a single node in the network we are
1721 * looking for.
1722 * The trouble is that we don't know the
1723 * netmask for the remote end.
1724 */
1725 if (ifa->ifa_dstaddr != NULL &&
1726 sa_equal(addr, ifa->ifa_dstaddr)) {
1727 ifa_ref(ifa);
1728 IF_ADDR_RUNLOCK(ifp);
1729 goto done;
1730 }
1731 } else {
1732 /*
1733 * if we have a special address handler,
1734 * then use it instead of the generic one.
1735 */
1736 if (ifa->ifa_claim_addr) {
1737 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1738 ifa_ref(ifa);
1739 IF_ADDR_RUNLOCK(ifp);
1740 goto done;
1741 }
1742 continue;
1743 }
1744
1745 /*
1746 * Scan all the bits in the ifa's address.
1747 * If a bit dissagrees with what we are
1748 * looking for, mask it with the netmask
1749 * to see if it really matters.
1750 * (A byte at a time)
1751 */
1752 if (ifa->ifa_netmask == 0)
1753 continue;
1754 cp = addr_data;
1755 cp2 = ifa->ifa_addr->sa_data;
1756 cp3 = ifa->ifa_netmask->sa_data;
1757 cplim = ifa->ifa_netmask->sa_len
1758 + (char *)ifa->ifa_netmask;
1759 while (cp3 < cplim)
1760 if ((*cp++ ^ *cp2++) & *cp3++)
1761 goto next; /* next address! */
1762 /*
1763 * If the netmask of what we just found
1764 * is more specific than what we had before
1765 * (if we had one) then remember the new one
1766 * before continuing to search
1767 * for an even better one.
1768 */
1769 if (ifa_maybe == NULL ||
1770 rn_refines((caddr_t)ifa->ifa_netmask,
1771 (caddr_t)ifa_maybe->ifa_netmask)) {
1772 if (ifa_maybe != NULL)
1773 ifa_free(ifa_maybe);
1774 ifa_maybe = ifa;
1775 ifa_ref(ifa_maybe);
1776 }
1777 }
1778 }
1779 IF_ADDR_RUNLOCK(ifp);
1780 }
1781 ifa = ifa_maybe;
1782 ifa_maybe = NULL;
1783 done:
1784 IFNET_RUNLOCK_NOSLEEP();
1785 if (ifa_maybe != NULL)
1786 ifa_free(ifa_maybe);
1787 return (ifa);
1788 }
1789
1790 /*
1791 * Find an interface address specific to an interface best matching
1792 * a given address.
1793 */
1794 struct ifaddr *
1795 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1796 {
1797 struct ifaddr *ifa;
1798 char *cp, *cp2, *cp3;
1799 char *cplim;
1800 struct ifaddr *ifa_maybe = NULL;
1801 u_int af = addr->sa_family;
1802
1803 if (af >= AF_MAX)
1804 return (NULL);
1805 IF_ADDR_RLOCK(ifp);
1806 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1807 if (ifa->ifa_addr->sa_family != af)
1808 continue;
1809 if (ifa_maybe == NULL)
1810 ifa_maybe = ifa;
1811 if (ifa->ifa_netmask == 0) {
1812 if (sa_equal(addr, ifa->ifa_addr) ||
1813 (ifa->ifa_dstaddr &&
1814 sa_equal(addr, ifa->ifa_dstaddr)))
1815 goto done;
1816 continue;
1817 }
1818 if (ifp->if_flags & IFF_POINTOPOINT) {
1819 if (sa_equal(addr, ifa->ifa_dstaddr))
1820 goto done;
1821 } else {
1822 cp = addr->sa_data;
1823 cp2 = ifa->ifa_addr->sa_data;
1824 cp3 = ifa->ifa_netmask->sa_data;
1825 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1826 for (; cp3 < cplim; cp3++)
1827 if ((*cp++ ^ *cp2++) & *cp3)
1828 break;
1829 if (cp3 == cplim)
1830 goto done;
1831 }
1832 }
1833 ifa = ifa_maybe;
1834 done:
1835 if (ifa != NULL)
1836 ifa_ref(ifa);
1837 IF_ADDR_RUNLOCK(ifp);
1838 return (ifa);
1839 }
1840
1841 #include <net/if_llatbl.h>
1842
1843 /*
1844 * Default action when installing a route with a Link Level gateway.
1845 * Lookup an appropriate real ifa to point to.
1846 * This should be moved to /sys/net/link.c eventually.
1847 */
1848 static void
1849 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1850 {
1851 struct ifaddr *ifa, *oifa;
1852 struct sockaddr *dst;
1853 struct ifnet *ifp;
1854
1855 RT_LOCK_ASSERT(rt);
1856
1857 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1858 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1859 return;
1860 ifa = ifaof_ifpforaddr(dst, ifp);
1861 if (ifa) {
1862 oifa = rt->rt_ifa;
1863 rt->rt_ifa = ifa;
1864 ifa_free(oifa);
1865 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1866 ifa->ifa_rtrequest(cmd, rt, info);
1867 }
1868 }
1869
1870 /*
1871 * Mark an interface down and notify protocols of
1872 * the transition.
1873 * NOTE: must be called at splnet or eqivalent.
1874 */
1875 static void
1876 if_unroute(struct ifnet *ifp, int flag, int fam)
1877 {
1878 struct ifaddr *ifa;
1879
1880 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1881
1882 ifp->if_flags &= ~flag;
1883 getmicrotime(&ifp->if_lastchange);
1884 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1885 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1886 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1887 ifp->if_qflush(ifp);
1888
1889 if (ifp->if_carp)
1890 (*carp_linkstate_p)(ifp);
1891 rt_ifmsg(ifp);
1892 }
1893
1894 /*
1895 * Mark an interface up and notify protocols of
1896 * the transition.
1897 * NOTE: must be called at splnet or eqivalent.
1898 */
1899 static void
1900 if_route(struct ifnet *ifp, int flag, int fam)
1901 {
1902 struct ifaddr *ifa;
1903
1904 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1905
1906 ifp->if_flags |= flag;
1907 getmicrotime(&ifp->if_lastchange);
1908 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1909 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1910 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1911 if (ifp->if_carp)
1912 (*carp_linkstate_p)(ifp);
1913 rt_ifmsg(ifp);
1914 #ifdef INET6
1915 in6_if_up(ifp);
1916 #endif
1917 }
1918
1919 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1920 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
1921
1922 /*
1923 * Handle a change in the interface link state. To avoid LORs
1924 * between driver lock and upper layer locks, as well as possible
1925 * recursions, we post event to taskqueue, and all job
1926 * is done in static do_link_state_change().
1927 */
1928 void
1929 if_link_state_change(struct ifnet *ifp, int link_state)
1930 {
1931 /* Return if state hasn't changed. */
1932 if (ifp->if_link_state == link_state)
1933 return;
1934
1935 ifp->if_link_state = link_state;
1936
1937 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1938 }
1939
1940 static void
1941 do_link_state_change(void *arg, int pending)
1942 {
1943 struct ifnet *ifp = (struct ifnet *)arg;
1944 int link_state = ifp->if_link_state;
1945 CURVNET_SET(ifp->if_vnet);
1946
1947 /* Notify that the link state has changed. */
1948 rt_ifmsg(ifp);
1949 if (ifp->if_vlantrunk != NULL)
1950 (*vlan_link_state_p)(ifp, 0);
1951
1952 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1953 IFP2AC(ifp)->ac_netgraph != NULL)
1954 (*ng_ether_link_state_p)(ifp, link_state);
1955 if (ifp->if_carp)
1956 (*carp_linkstate_p)(ifp);
1957 if (ifp->if_bridge)
1958 (*bridge_linkstate_p)(ifp);
1959 if (ifp->if_lagg)
1960 (*lagg_linkstate_p)(ifp, link_state);
1961
1962 if (IS_DEFAULT_VNET(curvnet))
1963 devctl_notify("IFNET", ifp->if_xname,
1964 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1965 NULL);
1966 if (pending > 1)
1967 if_printf(ifp, "%d link states coalesced\n", pending);
1968 if (log_link_state_change)
1969 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1970 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1971 CURVNET_RESTORE();
1972 }
1973
1974 /*
1975 * Mark an interface down and notify protocols of
1976 * the transition.
1977 * NOTE: must be called at splnet or eqivalent.
1978 */
1979 void
1980 if_down(struct ifnet *ifp)
1981 {
1982
1983 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1984 }
1985
1986 /*
1987 * Mark an interface up and notify protocols of
1988 * the transition.
1989 * NOTE: must be called at splnet or eqivalent.
1990 */
1991 void
1992 if_up(struct ifnet *ifp)
1993 {
1994
1995 if_route(ifp, IFF_UP, AF_UNSPEC);
1996 }
1997
1998 /*
1999 * Flush an interface queue.
2000 */
2001 void
2002 if_qflush(struct ifnet *ifp)
2003 {
2004 struct mbuf *m, *n;
2005 struct ifaltq *ifq;
2006
2007 ifq = &ifp->if_snd;
2008 IFQ_LOCK(ifq);
2009 #ifdef ALTQ
2010 if (ALTQ_IS_ENABLED(ifq))
2011 ALTQ_PURGE(ifq);
2012 #endif
2013 n = ifq->ifq_head;
2014 while ((m = n) != 0) {
2015 n = m->m_act;
2016 m_freem(m);
2017 }
2018 ifq->ifq_head = 0;
2019 ifq->ifq_tail = 0;
2020 ifq->ifq_len = 0;
2021 IFQ_UNLOCK(ifq);
2022 }
2023
2024 /*
2025 * Handle interface watchdog timer routines. Called
2026 * from softclock, we decrement timers (if set) and
2027 * call the appropriate interface routine on expiration.
2028 *
2029 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
2030 * holding Giant.
2031 */
2032 static void
2033 if_slowtimo(void *arg)
2034 {
2035 VNET_ITERATOR_DECL(vnet_iter);
2036 struct ifnet *ifp;
2037 int s = splimp();
2038
2039 VNET_LIST_RLOCK_NOSLEEP();
2040 IFNET_RLOCK_NOSLEEP();
2041 VNET_FOREACH(vnet_iter) {
2042 CURVNET_SET(vnet_iter);
2043 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2044 if (ifp->if_timer == 0 || --ifp->if_timer)
2045 continue;
2046 if (ifp->if_watchdog)
2047 (*ifp->if_watchdog)(ifp);
2048 }
2049 CURVNET_RESTORE();
2050 }
2051 IFNET_RUNLOCK_NOSLEEP();
2052 VNET_LIST_RUNLOCK_NOSLEEP();
2053 splx(s);
2054 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
2055 }
2056
2057 /*
2058 * Map interface name to interface structure pointer, with or without
2059 * returning a reference.
2060 */
2061 struct ifnet *
2062 ifunit_ref(const char *name)
2063 {
2064 struct ifnet *ifp;
2065
2066 IFNET_RLOCK_NOSLEEP();
2067 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2068 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2069 !(ifp->if_flags & IFF_DYING))
2070 break;
2071 }
2072 if (ifp != NULL)
2073 if_ref(ifp);
2074 IFNET_RUNLOCK_NOSLEEP();
2075 return (ifp);
2076 }
2077
2078 struct ifnet *
2079 ifunit(const char *name)
2080 {
2081 struct ifnet *ifp;
2082
2083 IFNET_RLOCK_NOSLEEP();
2084 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2085 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2086 break;
2087 }
2088 IFNET_RUNLOCK_NOSLEEP();
2089 return (ifp);
2090 }
2091
2092 /*
2093 * Hardware specific interface ioctls.
2094 */
2095 static int
2096 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2097 {
2098 struct ifreq *ifr;
2099 struct ifstat *ifs;
2100 int error = 0;
2101 int new_flags, temp_flags;
2102 size_t namelen, onamelen;
2103 size_t descrlen;
2104 char *descrbuf, *odescrbuf;
2105 char new_name[IFNAMSIZ];
2106 struct ifaddr *ifa;
2107 struct sockaddr_dl *sdl;
2108
2109 ifr = (struct ifreq *)data;
2110 switch (cmd) {
2111 case SIOCGIFINDEX:
2112 ifr->ifr_index = ifp->if_index;
2113 break;
2114
2115 case SIOCGIFFLAGS:
2116 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2117 ifr->ifr_flags = temp_flags & 0xffff;
2118 ifr->ifr_flagshigh = temp_flags >> 16;
2119 break;
2120
2121 case SIOCGIFCAP:
2122 ifr->ifr_reqcap = ifp->if_capabilities;
2123 ifr->ifr_curcap = ifp->if_capenable;
2124 break;
2125
2126 #ifdef MAC
2127 case SIOCGIFMAC:
2128 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2129 break;
2130 #endif
2131
2132 case SIOCGIFMETRIC:
2133 ifr->ifr_metric = ifp->if_metric;
2134 break;
2135
2136 case SIOCGIFMTU:
2137 ifr->ifr_mtu = ifp->if_mtu;
2138 break;
2139
2140 case SIOCGIFPHYS:
2141 ifr->ifr_phys = ifp->if_physical;
2142 break;
2143
2144 case SIOCGIFDESCR:
2145 error = 0;
2146 sx_slock(&ifdescr_sx);
2147 if (ifp->if_description == NULL)
2148 error = ENOMSG;
2149 else {
2150 /* space for terminating nul */
2151 descrlen = strlen(ifp->if_description) + 1;
2152 if (ifr->ifr_buffer.length < descrlen)
2153 ifr->ifr_buffer.buffer = NULL;
2154 else
2155 error = copyout(ifp->if_description,
2156 ifr->ifr_buffer.buffer, descrlen);
2157 ifr->ifr_buffer.length = descrlen;
2158 }
2159 sx_sunlock(&ifdescr_sx);
2160 break;
2161
2162 case SIOCSIFDESCR:
2163 error = priv_check(td, PRIV_NET_SETIFDESCR);
2164 if (error)
2165 return (error);
2166
2167 /*
2168 * Copy only (length-1) bytes to make sure that
2169 * if_description is always nul terminated. The
2170 * length parameter is supposed to count the
2171 * terminating nul in.
2172 */
2173 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2174 return (ENAMETOOLONG);
2175 else if (ifr->ifr_buffer.length == 0)
2176 descrbuf = NULL;
2177 else {
2178 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2179 M_WAITOK | M_ZERO);
2180 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2181 ifr->ifr_buffer.length - 1);
2182 if (error) {
2183 free(descrbuf, M_IFDESCR);
2184 break;
2185 }
2186 }
2187
2188 sx_xlock(&ifdescr_sx);
2189 odescrbuf = ifp->if_description;
2190 ifp->if_description = descrbuf;
2191 sx_xunlock(&ifdescr_sx);
2192
2193 getmicrotime(&ifp->if_lastchange);
2194 free(odescrbuf, M_IFDESCR);
2195 break;
2196
2197 case SIOCGIFFIB:
2198 ifr->ifr_fib = ifp->if_fib;
2199 break;
2200
2201 case SIOCSIFFIB:
2202 error = priv_check(td, PRIV_NET_SETIFFIB);
2203 if (error)
2204 return (error);
2205 if (ifr->ifr_fib >= rt_numfibs)
2206 return (EINVAL);
2207
2208 ifp->if_fib = ifr->ifr_fib;
2209 break;
2210
2211 case SIOCSIFFLAGS:
2212 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2213 if (error)
2214 return (error);
2215 /*
2216 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2217 * check, so we don't need special handling here yet.
2218 */
2219 new_flags = (ifr->ifr_flags & 0xffff) |
2220 (ifr->ifr_flagshigh << 16);
2221 if (ifp->if_flags & IFF_SMART) {
2222 /* Smart drivers twiddle their own routes */
2223 } else if (ifp->if_flags & IFF_UP &&
2224 (new_flags & IFF_UP) == 0) {
2225 int s = splimp();
2226 if_down(ifp);
2227 splx(s);
2228 } else if (new_flags & IFF_UP &&
2229 (ifp->if_flags & IFF_UP) == 0) {
2230 int s = splimp();
2231 if_up(ifp);
2232 splx(s);
2233 }
2234 /* See if permanently promiscuous mode bit is about to flip */
2235 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2236 if (new_flags & IFF_PPROMISC)
2237 ifp->if_flags |= IFF_PROMISC;
2238 else if (ifp->if_pcount == 0)
2239 ifp->if_flags &= ~IFF_PROMISC;
2240 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2241 ifp->if_xname,
2242 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2243 }
2244 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2245 (new_flags &~ IFF_CANTCHANGE);
2246 if (ifp->if_ioctl) {
2247 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2248 }
2249 getmicrotime(&ifp->if_lastchange);
2250 break;
2251
2252 case SIOCSIFCAP:
2253 error = priv_check(td, PRIV_NET_SETIFCAP);
2254 if (error)
2255 return (error);
2256 if (ifp->if_ioctl == NULL)
2257 return (EOPNOTSUPP);
2258 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2259 return (EINVAL);
2260 error = (*ifp->if_ioctl)(ifp, cmd, data);
2261 if (error == 0)
2262 getmicrotime(&ifp->if_lastchange);
2263 break;
2264
2265 #ifdef MAC
2266 case SIOCSIFMAC:
2267 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2268 break;
2269 #endif
2270
2271 case SIOCSIFNAME:
2272 error = priv_check(td, PRIV_NET_SETIFNAME);
2273 if (error)
2274 return (error);
2275 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2276 if (error != 0)
2277 return (error);
2278 if (new_name[0] == '\0')
2279 return (EINVAL);
2280 if (ifunit(new_name) != NULL)
2281 return (EEXIST);
2282
2283 /*
2284 * XXX: Locking. Nothing else seems to lock if_flags,
2285 * and there are numerous other races with the
2286 * ifunit() checks not being atomic with namespace
2287 * changes (renames, vmoves, if_attach, etc).
2288 */
2289 ifp->if_flags |= IFF_RENAMING;
2290
2291 /* Announce the departure of the interface. */
2292 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2293 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2294
2295 log(LOG_INFO, "%s: changing name to '%s'\n",
2296 ifp->if_xname, new_name);
2297
2298 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2299 ifa = ifp->if_addr;
2300 IFA_LOCK(ifa);
2301 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2302 namelen = strlen(new_name);
2303 onamelen = sdl->sdl_nlen;
2304 /*
2305 * Move the address if needed. This is safe because we
2306 * allocate space for a name of length IFNAMSIZ when we
2307 * create this in if_attach().
2308 */
2309 if (namelen != onamelen) {
2310 bcopy(sdl->sdl_data + onamelen,
2311 sdl->sdl_data + namelen, sdl->sdl_alen);
2312 }
2313 bcopy(new_name, sdl->sdl_data, namelen);
2314 sdl->sdl_nlen = namelen;
2315 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2316 bzero(sdl->sdl_data, onamelen);
2317 while (namelen != 0)
2318 sdl->sdl_data[--namelen] = 0xff;
2319 IFA_UNLOCK(ifa);
2320
2321 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2322 /* Announce the return of the interface. */
2323 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2324
2325 ifp->if_flags &= ~IFF_RENAMING;
2326 break;
2327
2328 #ifdef VIMAGE
2329 case SIOCSIFVNET:
2330 error = priv_check(td, PRIV_NET_SETIFVNET);
2331 if (error)
2332 return (error);
2333 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2334 break;
2335 #endif
2336
2337 case SIOCSIFMETRIC:
2338 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2339 if (error)
2340 return (error);
2341 ifp->if_metric = ifr->ifr_metric;
2342 getmicrotime(&ifp->if_lastchange);
2343 break;
2344
2345 case SIOCSIFPHYS:
2346 error = priv_check(td, PRIV_NET_SETIFPHYS);
2347 if (error)
2348 return (error);
2349 if (ifp->if_ioctl == NULL)
2350 return (EOPNOTSUPP);
2351 error = (*ifp->if_ioctl)(ifp, cmd, data);
2352 if (error == 0)
2353 getmicrotime(&ifp->if_lastchange);
2354 break;
2355
2356 case SIOCSIFMTU:
2357 {
2358 u_long oldmtu = ifp->if_mtu;
2359
2360 error = priv_check(td, PRIV_NET_SETIFMTU);
2361 if (error)
2362 return (error);
2363 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2364 return (EINVAL);
2365 if (ifp->if_ioctl == NULL)
2366 return (EOPNOTSUPP);
2367 error = (*ifp->if_ioctl)(ifp, cmd, data);
2368 if (error == 0) {
2369 getmicrotime(&ifp->if_lastchange);
2370 rt_ifmsg(ifp);
2371 }
2372 /*
2373 * If the link MTU changed, do network layer specific procedure.
2374 */
2375 if (ifp->if_mtu != oldmtu) {
2376 #ifdef INET6
2377 nd6_setmtu(ifp);
2378 #endif
2379 }
2380 break;
2381 }
2382
2383 case SIOCADDMULTI:
2384 case SIOCDELMULTI:
2385 if (cmd == SIOCADDMULTI)
2386 error = priv_check(td, PRIV_NET_ADDMULTI);
2387 else
2388 error = priv_check(td, PRIV_NET_DELMULTI);
2389 if (error)
2390 return (error);
2391
2392 /* Don't allow group membership on non-multicast interfaces. */
2393 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2394 return (EOPNOTSUPP);
2395
2396 /* Don't let users screw up protocols' entries. */
2397 if (ifr->ifr_addr.sa_family != AF_LINK)
2398 return (EINVAL);
2399
2400 if (cmd == SIOCADDMULTI) {
2401 struct ifmultiaddr *ifma;
2402
2403 /*
2404 * Userland is only permitted to join groups once
2405 * via the if_addmulti() KPI, because it cannot hold
2406 * struct ifmultiaddr * between calls. It may also
2407 * lose a race while we check if the membership
2408 * already exists.
2409 */
2410 IF_ADDR_RLOCK(ifp);
2411 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2412 IF_ADDR_RUNLOCK(ifp);
2413 if (ifma != NULL)
2414 error = EADDRINUSE;
2415 else
2416 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2417 } else {
2418 error = if_delmulti(ifp, &ifr->ifr_addr);
2419 }
2420 if (error == 0)
2421 getmicrotime(&ifp->if_lastchange);
2422 break;
2423
2424 case SIOCSIFPHYADDR:
2425 case SIOCDIFPHYADDR:
2426 #ifdef INET6
2427 case SIOCSIFPHYADDR_IN6:
2428 #endif
2429 case SIOCSLIFPHYADDR:
2430 case SIOCSIFMEDIA:
2431 case SIOCSIFGENERIC:
2432 error = priv_check(td, PRIV_NET_HWIOCTL);
2433 if (error)
2434 return (error);
2435 if (ifp->if_ioctl == NULL)
2436 return (EOPNOTSUPP);
2437 error = (*ifp->if_ioctl)(ifp, cmd, data);
2438 if (error == 0)
2439 getmicrotime(&ifp->if_lastchange);
2440 break;
2441
2442 case SIOCGIFSTATUS:
2443 ifs = (struct ifstat *)data;
2444 ifs->ascii[0] = '\0';
2445
2446 case SIOCGIFPSRCADDR:
2447 case SIOCGIFPDSTADDR:
2448 case SIOCGLIFPHYADDR:
2449 case SIOCGIFMEDIA:
2450 case SIOCGIFGENERIC:
2451 if (ifp->if_ioctl == NULL)
2452 return (EOPNOTSUPP);
2453 error = (*ifp->if_ioctl)(ifp, cmd, data);
2454 break;
2455
2456 case SIOCSIFLLADDR:
2457 error = priv_check(td, PRIV_NET_SETLLADDR);
2458 if (error)
2459 return (error);
2460 error = if_setlladdr(ifp,
2461 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2462 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2463 break;
2464
2465 case SIOCAIFGROUP:
2466 {
2467 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2468
2469 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2470 if (error)
2471 return (error);
2472 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2473 return (error);
2474 break;
2475 }
2476
2477 case SIOCGIFGROUP:
2478 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2479 return (error);
2480 break;
2481
2482 case SIOCDIFGROUP:
2483 {
2484 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2485
2486 error = priv_check(td, PRIV_NET_DELIFGROUP);
2487 if (error)
2488 return (error);
2489 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2490 return (error);
2491 break;
2492 }
2493
2494 default:
2495 error = ENOIOCTL;
2496 break;
2497 }
2498 return (error);
2499 }
2500
2501 #ifdef COMPAT_FREEBSD32
2502 struct ifconf32 {
2503 int32_t ifc_len;
2504 union {
2505 uint32_t ifcu_buf;
2506 uint32_t ifcu_req;
2507 } ifc_ifcu;
2508 };
2509 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2510 #endif
2511
2512 /*
2513 * Interface ioctls.
2514 */
2515 int
2516 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2517 {
2518 struct ifnet *ifp;
2519 struct ifreq *ifr;
2520 int error;
2521 int oif_flags;
2522
2523 CURVNET_SET(so->so_vnet);
2524 switch (cmd) {
2525 case SIOCGIFCONF:
2526 case OSIOCGIFCONF:
2527 error = ifconf(cmd, data);
2528 CURVNET_RESTORE();
2529 return (error);
2530
2531 #ifdef COMPAT_FREEBSD32
2532 case SIOCGIFCONF32:
2533 {
2534 struct ifconf32 *ifc32;
2535 struct ifconf ifc;
2536
2537 ifc32 = (struct ifconf32 *)data;
2538 ifc.ifc_len = ifc32->ifc_len;
2539 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2540
2541 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2542 CURVNET_RESTORE();
2543 if (error == 0)
2544 ifc32->ifc_len = ifc.ifc_len;
2545 return (error);
2546 }
2547 #endif
2548 }
2549 ifr = (struct ifreq *)data;
2550
2551 switch (cmd) {
2552 #ifdef VIMAGE
2553 case SIOCSIFRVNET:
2554 error = priv_check(td, PRIV_NET_SETIFVNET);
2555 if (error == 0)
2556 error = if_vmove_reclaim(td, ifr->ifr_name,
2557 ifr->ifr_jid);
2558 CURVNET_RESTORE();
2559 return (error);
2560 #endif
2561 case SIOCIFCREATE:
2562 case SIOCIFCREATE2:
2563 error = priv_check(td, PRIV_NET_IFCREATE);
2564 if (error == 0)
2565 error = if_clone_create(ifr->ifr_name,
2566 sizeof(ifr->ifr_name),
2567 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2568 CURVNET_RESTORE();
2569 return (error);
2570 case SIOCIFDESTROY:
2571 error = priv_check(td, PRIV_NET_IFDESTROY);
2572 if (error == 0)
2573 error = if_clone_destroy(ifr->ifr_name);
2574 CURVNET_RESTORE();
2575 return (error);
2576
2577 case SIOCIFGCLONERS:
2578 error = if_clone_list((struct if_clonereq *)data);
2579 CURVNET_RESTORE();
2580 return (error);
2581 case SIOCGIFGMEMB:
2582 error = if_getgroupmembers((struct ifgroupreq *)data);
2583 CURVNET_RESTORE();
2584 return (error);
2585 }
2586
2587 ifp = ifunit_ref(ifr->ifr_name);
2588 if (ifp == NULL) {
2589 CURVNET_RESTORE();
2590 return (ENXIO);
2591 }
2592
2593 error = ifhwioctl(cmd, ifp, data, td);
2594 if (error != ENOIOCTL) {
2595 if_rele(ifp);
2596 CURVNET_RESTORE();
2597 return (error);
2598 }
2599
2600 oif_flags = ifp->if_flags;
2601 if (so->so_proto == NULL) {
2602 if_rele(ifp);
2603 CURVNET_RESTORE();
2604 return (EOPNOTSUPP);
2605 }
2606
2607 /*
2608 * Pass the request on to the socket control method, and if the
2609 * latter returns EOPNOTSUPP, directly to the interface.
2610 *
2611 * Make an exception for the legacy SIOCSIF* requests. Drivers
2612 * trust SIOCSIFADDR et al to come from an already privileged
2613 * layer, and do not perform any credentials checks or input
2614 * validation.
2615 */
2616 #ifndef COMPAT_43
2617 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2618 data,
2619 ifp, td));
2620 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2621 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2622 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2623 error = (*ifp->if_ioctl)(ifp, cmd, data);
2624 #else
2625 {
2626 u_long ocmd = cmd;
2627
2628 switch (cmd) {
2629
2630 case SIOCSIFDSTADDR:
2631 case SIOCSIFADDR:
2632 case SIOCSIFBRDADDR:
2633 case SIOCSIFNETMASK:
2634 #if BYTE_ORDER != BIG_ENDIAN
2635 if (ifr->ifr_addr.sa_family == 0 &&
2636 ifr->ifr_addr.sa_len < 16) {
2637 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2638 ifr->ifr_addr.sa_len = 16;
2639 }
2640 #else
2641 if (ifr->ifr_addr.sa_len == 0)
2642 ifr->ifr_addr.sa_len = 16;
2643 #endif
2644 break;
2645
2646 case OSIOCGIFADDR:
2647 cmd = SIOCGIFADDR;
2648 break;
2649
2650 case OSIOCGIFDSTADDR:
2651 cmd = SIOCGIFDSTADDR;
2652 break;
2653
2654 case OSIOCGIFBRDADDR:
2655 cmd = SIOCGIFBRDADDR;
2656 break;
2657
2658 case OSIOCGIFNETMASK:
2659 cmd = SIOCGIFNETMASK;
2660 }
2661 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2662 cmd,
2663 data,
2664 ifp, td));
2665 if (error == EOPNOTSUPP && ifp != NULL &&
2666 ifp->if_ioctl != NULL &&
2667 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2668 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2669 error = (*ifp->if_ioctl)(ifp, cmd, data);
2670 switch (ocmd) {
2671
2672 case OSIOCGIFADDR:
2673 case OSIOCGIFDSTADDR:
2674 case OSIOCGIFBRDADDR:
2675 case OSIOCGIFNETMASK:
2676 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2677
2678 }
2679 }
2680 #endif /* COMPAT_43 */
2681
2682 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2683 #ifdef INET6
2684 if (ifp->if_flags & IFF_UP) {
2685 int s = splimp();
2686 in6_if_up(ifp);
2687 splx(s);
2688 }
2689 #endif
2690 }
2691 if_rele(ifp);
2692 CURVNET_RESTORE();
2693 return (error);
2694 }
2695
2696 /*
2697 * The code common to handling reference counted flags,
2698 * e.g., in ifpromisc() and if_allmulti().
2699 * The "pflag" argument can specify a permanent mode flag to check,
2700 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2701 *
2702 * Only to be used on stack-owned flags, not driver-owned flags.
2703 */
2704 static int
2705 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2706 {
2707 struct ifreq ifr;
2708 int error;
2709 int oldflags, oldcount;
2710
2711 /* Sanity checks to catch programming errors */
2712 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2713 ("%s: setting driver-owned flag %d", __func__, flag));
2714
2715 if (onswitch)
2716 KASSERT(*refcount >= 0,
2717 ("%s: increment negative refcount %d for flag %d",
2718 __func__, *refcount, flag));
2719 else
2720 KASSERT(*refcount > 0,
2721 ("%s: decrement non-positive refcount %d for flag %d",
2722 __func__, *refcount, flag));
2723
2724 /* In case this mode is permanent, just touch refcount */
2725 if (ifp->if_flags & pflag) {
2726 *refcount += onswitch ? 1 : -1;
2727 return (0);
2728 }
2729
2730 /* Save ifnet parameters for if_ioctl() may fail */
2731 oldcount = *refcount;
2732 oldflags = ifp->if_flags;
2733
2734 /*
2735 * See if we aren't the only and touching refcount is enough.
2736 * Actually toggle interface flag if we are the first or last.
2737 */
2738 if (onswitch) {
2739 if ((*refcount)++)
2740 return (0);
2741 ifp->if_flags |= flag;
2742 } else {
2743 if (--(*refcount))
2744 return (0);
2745 ifp->if_flags &= ~flag;
2746 }
2747
2748 /* Call down the driver since we've changed interface flags */
2749 if (ifp->if_ioctl == NULL) {
2750 error = EOPNOTSUPP;
2751 goto recover;
2752 }
2753 ifr.ifr_flags = ifp->if_flags & 0xffff;
2754 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2755 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2756 if (error)
2757 goto recover;
2758 /* Notify userland that interface flags have changed */
2759 rt_ifmsg(ifp);
2760 return (0);
2761
2762 recover:
2763 /* Recover after driver error */
2764 *refcount = oldcount;
2765 ifp->if_flags = oldflags;
2766 return (error);
2767 }
2768
2769 /*
2770 * Set/clear promiscuous mode on interface ifp based on the truth value
2771 * of pswitch. The calls are reference counted so that only the first
2772 * "on" request actually has an effect, as does the final "off" request.
2773 * Results are undefined if the "off" and "on" requests are not matched.
2774 */
2775 int
2776 ifpromisc(struct ifnet *ifp, int pswitch)
2777 {
2778 int error;
2779 int oldflags = ifp->if_flags;
2780
2781 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2782 &ifp->if_pcount, pswitch);
2783 /* If promiscuous mode status has changed, log a message */
2784 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2785 log(LOG_INFO, "%s: promiscuous mode %s\n",
2786 ifp->if_xname,
2787 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2788 return (error);
2789 }
2790
2791 /*
2792 * Return interface configuration
2793 * of system. List may be used
2794 * in later ioctl's (above) to get
2795 * other information.
2796 */
2797 /*ARGSUSED*/
2798 static int
2799 ifconf(u_long cmd, caddr_t data)
2800 {
2801 struct ifconf *ifc = (struct ifconf *)data;
2802 struct ifnet *ifp;
2803 struct ifaddr *ifa;
2804 struct ifreq ifr;
2805 struct sbuf *sb;
2806 int error, full = 0, valid_len, max_len;
2807
2808 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2809 max_len = MAXPHYS - 1;
2810
2811 /* Prevent hostile input from being able to crash the system */
2812 if (ifc->ifc_len <= 0)
2813 return (EINVAL);
2814
2815 again:
2816 if (ifc->ifc_len <= max_len) {
2817 max_len = ifc->ifc_len;
2818 full = 1;
2819 }
2820 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2821 max_len = 0;
2822 valid_len = 0;
2823
2824 IFNET_RLOCK();
2825 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2826 int addrs;
2827
2828 /*
2829 * Zero the ifr_name buffer to make sure we don't
2830 * disclose the contents of the stack.
2831 */
2832 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2833
2834 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2835 >= sizeof(ifr.ifr_name)) {
2836 sbuf_delete(sb);
2837 IFNET_RUNLOCK();
2838 return (ENAMETOOLONG);
2839 }
2840
2841 addrs = 0;
2842 IF_ADDR_RLOCK(ifp);
2843 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2844 struct sockaddr *sa = ifa->ifa_addr;
2845
2846 if (prison_if(curthread->td_ucred, sa) != 0)
2847 continue;
2848 addrs++;
2849 #ifdef COMPAT_43
2850 if (cmd == OSIOCGIFCONF) {
2851 struct osockaddr *osa =
2852 (struct osockaddr *)&ifr.ifr_addr;
2853 ifr.ifr_addr = *sa;
2854 osa->sa_family = sa->sa_family;
2855 sbuf_bcat(sb, &ifr, sizeof(ifr));
2856 max_len += sizeof(ifr);
2857 } else
2858 #endif
2859 if (sa->sa_len <= sizeof(*sa)) {
2860 ifr.ifr_addr = *sa;
2861 sbuf_bcat(sb, &ifr, sizeof(ifr));
2862 max_len += sizeof(ifr);
2863 } else {
2864 sbuf_bcat(sb, &ifr,
2865 offsetof(struct ifreq, ifr_addr));
2866 max_len += offsetof(struct ifreq, ifr_addr);
2867 sbuf_bcat(sb, sa, sa->sa_len);
2868 max_len += sa->sa_len;
2869 }
2870
2871 if (!sbuf_overflowed(sb))
2872 valid_len = sbuf_len(sb);
2873 }
2874 IF_ADDR_RUNLOCK(ifp);
2875 if (addrs == 0) {
2876 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2877 sbuf_bcat(sb, &ifr, sizeof(ifr));
2878 max_len += sizeof(ifr);
2879
2880 if (!sbuf_overflowed(sb))
2881 valid_len = sbuf_len(sb);
2882 }
2883 }
2884 IFNET_RUNLOCK();
2885
2886 /*
2887 * If we didn't allocate enough space (uncommon), try again. If
2888 * we have already allocated as much space as we are allowed,
2889 * return what we've got.
2890 */
2891 if (valid_len != max_len && !full) {
2892 sbuf_delete(sb);
2893 goto again;
2894 }
2895
2896 ifc->ifc_len = valid_len;
2897 sbuf_finish(sb);
2898 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2899 sbuf_delete(sb);
2900 return (error);
2901 }
2902
2903 /*
2904 * Just like ifpromisc(), but for all-multicast-reception mode.
2905 */
2906 int
2907 if_allmulti(struct ifnet *ifp, int onswitch)
2908 {
2909
2910 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2911 }
2912
2913 struct ifmultiaddr *
2914 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2915 {
2916 struct ifmultiaddr *ifma;
2917
2918 IF_ADDR_LOCK_ASSERT(ifp);
2919
2920 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2921 if (sa->sa_family == AF_LINK) {
2922 if (sa_dl_equal(ifma->ifma_addr, sa))
2923 break;
2924 } else {
2925 if (sa_equal(ifma->ifma_addr, sa))
2926 break;
2927 }
2928 }
2929
2930 return ifma;
2931 }
2932
2933 /*
2934 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2935 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2936 * the ifnet multicast address list here, so the caller must do that and
2937 * other setup work (such as notifying the device driver). The reference
2938 * count is initialized to 1.
2939 */
2940 static struct ifmultiaddr *
2941 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2942 int mflags)
2943 {
2944 struct ifmultiaddr *ifma;
2945 struct sockaddr *dupsa;
2946
2947 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2948 M_ZERO);
2949 if (ifma == NULL)
2950 return (NULL);
2951
2952 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2953 if (dupsa == NULL) {
2954 free(ifma, M_IFMADDR);
2955 return (NULL);
2956 }
2957 bcopy(sa, dupsa, sa->sa_len);
2958 ifma->ifma_addr = dupsa;
2959
2960 ifma->ifma_ifp = ifp;
2961 ifma->ifma_refcount = 1;
2962 ifma->ifma_protospec = NULL;
2963
2964 if (llsa == NULL) {
2965 ifma->ifma_lladdr = NULL;
2966 return (ifma);
2967 }
2968
2969 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2970 if (dupsa == NULL) {
2971 free(ifma->ifma_addr, M_IFMADDR);
2972 free(ifma, M_IFMADDR);
2973 return (NULL);
2974 }
2975 bcopy(llsa, dupsa, llsa->sa_len);
2976 ifma->ifma_lladdr = dupsa;
2977
2978 return (ifma);
2979 }
2980
2981 /*
2982 * if_freemulti: free ifmultiaddr structure and possibly attached related
2983 * addresses. The caller is responsible for implementing reference
2984 * counting, notifying the driver, handling routing messages, and releasing
2985 * any dependent link layer state.
2986 */
2987 static void
2988 if_freemulti(struct ifmultiaddr *ifma)
2989 {
2990
2991 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2992 ifma->ifma_refcount));
2993 KASSERT(ifma->ifma_protospec == NULL,
2994 ("if_freemulti: protospec not NULL"));
2995
2996 if (ifma->ifma_lladdr != NULL)
2997 free(ifma->ifma_lladdr, M_IFMADDR);
2998 free(ifma->ifma_addr, M_IFMADDR);
2999 free(ifma, M_IFMADDR);
3000 }
3001
3002 /*
3003 * Register an additional multicast address with a network interface.
3004 *
3005 * - If the address is already present, bump the reference count on the
3006 * address and return.
3007 * - If the address is not link-layer, look up a link layer address.
3008 * - Allocate address structures for one or both addresses, and attach to the
3009 * multicast address list on the interface. If automatically adding a link
3010 * layer address, the protocol address will own a reference to the link
3011 * layer address, to be freed when it is freed.
3012 * - Notify the network device driver of an addition to the multicast address
3013 * list.
3014 *
3015 * 'sa' points to caller-owned memory with the desired multicast address.
3016 *
3017 * 'retifma' will be used to return a pointer to the resulting multicast
3018 * address reference, if desired.
3019 */
3020 int
3021 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3022 struct ifmultiaddr **retifma)
3023 {
3024 struct ifmultiaddr *ifma, *ll_ifma;
3025 struct sockaddr *llsa;
3026 int error;
3027
3028 /*
3029 * If the address is already present, return a new reference to it;
3030 * otherwise, allocate storage and set up a new address.
3031 */
3032 IF_ADDR_WLOCK(ifp);
3033 ifma = if_findmulti(ifp, sa);
3034 if (ifma != NULL) {
3035 ifma->ifma_refcount++;
3036 if (retifma != NULL)
3037 *retifma = ifma;
3038 IF_ADDR_WUNLOCK(ifp);
3039 return (0);
3040 }
3041
3042 /*
3043 * The address isn't already present; resolve the protocol address
3044 * into a link layer address, and then look that up, bump its
3045 * refcount or allocate an ifma for that also. If 'llsa' was
3046 * returned, we will need to free it later.
3047 */
3048 llsa = NULL;
3049 ll_ifma = NULL;
3050 if (ifp->if_resolvemulti != NULL) {
3051 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3052 if (error)
3053 goto unlock_out;
3054 }
3055
3056 /*
3057 * Allocate the new address. Don't hook it up yet, as we may also
3058 * need to allocate a link layer multicast address.
3059 */
3060 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3061 if (ifma == NULL) {
3062 error = ENOMEM;
3063 goto free_llsa_out;
3064 }
3065
3066 /*
3067 * If a link layer address is found, we'll need to see if it's
3068 * already present in the address list, or allocate is as well.
3069 * When this block finishes, the link layer address will be on the
3070 * list.
3071 */
3072 if (llsa != NULL) {
3073 ll_ifma = if_findmulti(ifp, llsa);
3074 if (ll_ifma == NULL) {
3075 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3076 if (ll_ifma == NULL) {
3077 --ifma->ifma_refcount;
3078 if_freemulti(ifma);
3079 error = ENOMEM;
3080 goto free_llsa_out;
3081 }
3082 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3083 ifma_link);
3084 } else
3085 ll_ifma->ifma_refcount++;
3086 ifma->ifma_llifma = ll_ifma;
3087 }
3088
3089 /*
3090 * We now have a new multicast address, ifma, and possibly a new or
3091 * referenced link layer address. Add the primary address to the
3092 * ifnet address list.
3093 */
3094 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3095
3096 if (retifma != NULL)
3097 *retifma = ifma;
3098
3099 /*
3100 * Must generate the message while holding the lock so that 'ifma'
3101 * pointer is still valid.
3102 */
3103 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3104 IF_ADDR_WUNLOCK(ifp);
3105
3106 /*
3107 * We are certain we have added something, so call down to the
3108 * interface to let them know about it.
3109 */
3110 if (ifp->if_ioctl != NULL) {
3111 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3112 }
3113
3114 if (llsa != NULL)
3115 free(llsa, M_IFMADDR);
3116
3117 return (0);
3118
3119 free_llsa_out:
3120 if (llsa != NULL)
3121 free(llsa, M_IFMADDR);
3122
3123 unlock_out:
3124 IF_ADDR_WUNLOCK(ifp);
3125 return (error);
3126 }
3127
3128 /*
3129 * Delete a multicast group membership by network-layer group address.
3130 *
3131 * Returns ENOENT if the entry could not be found. If ifp no longer
3132 * exists, results are undefined. This entry point should only be used
3133 * from subsystems which do appropriate locking to hold ifp for the
3134 * duration of the call.
3135 * Network-layer protocol domains must use if_delmulti_ifma().
3136 */
3137 int
3138 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3139 {
3140 struct ifmultiaddr *ifma;
3141 int lastref;
3142 #ifdef INVARIANTS
3143 struct ifnet *oifp;
3144
3145 IFNET_RLOCK_NOSLEEP();
3146 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3147 if (ifp == oifp)
3148 break;
3149 if (ifp != oifp)
3150 ifp = NULL;
3151 IFNET_RUNLOCK_NOSLEEP();
3152
3153 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3154 #endif
3155 if (ifp == NULL)
3156 return (ENOENT);
3157
3158 IF_ADDR_WLOCK(ifp);
3159 lastref = 0;
3160 ifma = if_findmulti(ifp, sa);
3161 if (ifma != NULL)
3162 lastref = if_delmulti_locked(ifp, ifma, 0);
3163 IF_ADDR_WUNLOCK(ifp);
3164
3165 if (ifma == NULL)
3166 return (ENOENT);
3167
3168 if (lastref && ifp->if_ioctl != NULL) {
3169 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3170 }
3171
3172 return (0);
3173 }
3174
3175 /*
3176 * Delete all multicast group membership for an interface.
3177 * Should be used to quickly flush all multicast filters.
3178 */
3179 void
3180 if_delallmulti(struct ifnet *ifp)
3181 {
3182 struct ifmultiaddr *ifma;
3183 struct ifmultiaddr *next;
3184
3185 IF_ADDR_WLOCK(ifp);
3186 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3187 if_delmulti_locked(ifp, ifma, 0);
3188 IF_ADDR_WUNLOCK(ifp);
3189 }
3190
3191 /*
3192 * Delete a multicast group membership by group membership pointer.
3193 * Network-layer protocol domains must use this routine.
3194 *
3195 * It is safe to call this routine if the ifp disappeared.
3196 */
3197 void
3198 if_delmulti_ifma(struct ifmultiaddr *ifma)
3199 {
3200 struct ifnet *ifp;
3201 int lastref;
3202
3203 ifp = ifma->ifma_ifp;
3204 #ifdef DIAGNOSTIC
3205 if (ifp == NULL) {
3206 printf("%s: ifma_ifp seems to be detached\n", __func__);
3207 } else {
3208 struct ifnet *oifp;
3209
3210 IFNET_RLOCK_NOSLEEP();
3211 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3212 if (ifp == oifp)
3213 break;
3214 if (ifp != oifp) {
3215 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3216 ifp = NULL;
3217 }
3218 IFNET_RUNLOCK_NOSLEEP();
3219 }
3220 #endif
3221 /*
3222 * If and only if the ifnet instance exists: Acquire the address lock.
3223 */
3224 if (ifp != NULL)
3225 IF_ADDR_WLOCK(ifp);
3226
3227 lastref = if_delmulti_locked(ifp, ifma, 0);
3228
3229 if (ifp != NULL) {
3230 /*
3231 * If and only if the ifnet instance exists:
3232 * Release the address lock.
3233 * If the group was left: update the hardware hash filter.
3234 */
3235 IF_ADDR_WUNLOCK(ifp);
3236 if (lastref && ifp->if_ioctl != NULL) {
3237 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3238 }
3239 }
3240 }
3241
3242 /*
3243 * Perform deletion of network-layer and/or link-layer multicast address.
3244 *
3245 * Return 0 if the reference count was decremented.
3246 * Return 1 if the final reference was released, indicating that the
3247 * hardware hash filter should be reprogrammed.
3248 */
3249 static int
3250 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3251 {
3252 struct ifmultiaddr *ll_ifma;
3253
3254 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3255 KASSERT(ifma->ifma_ifp == ifp,
3256 ("%s: inconsistent ifp %p", __func__, ifp));
3257 IF_ADDR_WLOCK_ASSERT(ifp);
3258 }
3259
3260 ifp = ifma->ifma_ifp;
3261
3262 /*
3263 * If the ifnet is detaching, null out references to ifnet,
3264 * so that upper protocol layers will notice, and not attempt
3265 * to obtain locks for an ifnet which no longer exists. The
3266 * routing socket announcement must happen before the ifnet
3267 * instance is detached from the system.
3268 */
3269 if (detaching) {
3270 #ifdef DIAGNOSTIC
3271 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3272 #endif
3273 /*
3274 * ifp may already be nulled out if we are being reentered
3275 * to delete the ll_ifma.
3276 */
3277 if (ifp != NULL) {
3278 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3279 ifma->ifma_ifp = NULL;
3280 }
3281 }
3282
3283 if (--ifma->ifma_refcount > 0)
3284 return 0;
3285
3286 /*
3287 * If this ifma is a network-layer ifma, a link-layer ifma may
3288 * have been associated with it. Release it first if so.
3289 */
3290 ll_ifma = ifma->ifma_llifma;
3291 if (ll_ifma != NULL) {
3292 KASSERT(ifma->ifma_lladdr != NULL,
3293 ("%s: llifma w/o lladdr", __func__));
3294 if (detaching)
3295 ll_ifma->ifma_ifp = NULL; /* XXX */
3296 if (--ll_ifma->ifma_refcount == 0) {
3297 if (ifp != NULL) {
3298 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3299 ifma_link);
3300 }
3301 if_freemulti(ll_ifma);
3302 }
3303 }
3304
3305 if (ifp != NULL)
3306 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3307
3308 if_freemulti(ifma);
3309
3310 /*
3311 * The last reference to this instance of struct ifmultiaddr
3312 * was released; the hardware should be notified of this change.
3313 */
3314 return 1;
3315 }
3316
3317 /*
3318 * Set the link layer address on an interface.
3319 *
3320 * At this time we only support certain types of interfaces,
3321 * and we don't allow the length of the address to change.
3322 */
3323 int
3324 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3325 {
3326 struct sockaddr_dl *sdl;
3327 struct ifaddr *ifa;
3328 struct ifreq ifr;
3329
3330 IF_ADDR_RLOCK(ifp);
3331 ifa = ifp->if_addr;
3332 if (ifa == NULL) {
3333 IF_ADDR_RUNLOCK(ifp);
3334 return (EINVAL);
3335 }
3336 ifa_ref(ifa);
3337 IF_ADDR_RUNLOCK(ifp);
3338 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3339 if (sdl == NULL) {
3340 ifa_free(ifa);
3341 return (EINVAL);
3342 }
3343 if (len != sdl->sdl_alen) { /* don't allow length to change */
3344 ifa_free(ifa);
3345 return (EINVAL);
3346 }
3347 switch (ifp->if_type) {
3348 case IFT_ETHER:
3349 case IFT_FDDI:
3350 case IFT_XETHER:
3351 case IFT_ISO88025:
3352 case IFT_L2VLAN:
3353 case IFT_BRIDGE:
3354 case IFT_ARCNET:
3355 case IFT_IEEE8023ADLAG:
3356 case IFT_IEEE80211:
3357 bcopy(lladdr, LLADDR(sdl), len);
3358 ifa_free(ifa);
3359 break;
3360 default:
3361 ifa_free(ifa);
3362 return (ENODEV);
3363 }
3364
3365 /*
3366 * If the interface is already up, we need
3367 * to re-init it in order to reprogram its
3368 * address filter.
3369 */
3370 if ((ifp->if_flags & IFF_UP) != 0) {
3371 if (ifp->if_ioctl) {
3372 ifp->if_flags &= ~IFF_UP;
3373 ifr.ifr_flags = ifp->if_flags & 0xffff;
3374 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3375 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3376 ifp->if_flags |= IFF_UP;
3377 ifr.ifr_flags = ifp->if_flags & 0xffff;
3378 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3379 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3380 }
3381 #ifdef INET
3382 /*
3383 * Also send gratuitous ARPs to notify other nodes about
3384 * the address change.
3385 */
3386 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3387 if (ifa->ifa_addr->sa_family == AF_INET)
3388 arp_ifinit(ifp, ifa);
3389 }
3390 #endif
3391 }
3392 return (0);
3393 }
3394
3395 /*
3396 * The name argument must be a pointer to storage which will last as
3397 * long as the interface does. For physical devices, the result of
3398 * device_get_name(dev) is a good choice and for pseudo-devices a
3399 * static string works well.
3400 */
3401 void
3402 if_initname(struct ifnet *ifp, const char *name, int unit)
3403 {
3404 ifp->if_dname = name;
3405 ifp->if_dunit = unit;
3406 if (unit != IF_DUNIT_NONE)
3407 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3408 else
3409 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3410 }
3411
3412 int
3413 if_printf(struct ifnet *ifp, const char * fmt, ...)
3414 {
3415 va_list ap;
3416 int retval;
3417
3418 retval = printf("%s: ", ifp->if_xname);
3419 va_start(ap, fmt);
3420 retval += vprintf(fmt, ap);
3421 va_end(ap);
3422 return (retval);
3423 }
3424
3425 void
3426 if_start(struct ifnet *ifp)
3427 {
3428
3429 (*(ifp)->if_start)(ifp);
3430 }
3431
3432 /*
3433 * Backwards compatibility interface for drivers
3434 * that have not implemented it
3435 */
3436 static int
3437 if_transmit(struct ifnet *ifp, struct mbuf *m)
3438 {
3439 int error;
3440
3441 IFQ_HANDOFF(ifp, m, error);
3442 return (error);
3443 }
3444
3445 int
3446 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3447 {
3448 int active = 0;
3449
3450 IF_LOCK(ifq);
3451 if (_IF_QFULL(ifq)) {
3452 _IF_DROP(ifq);
3453 IF_UNLOCK(ifq);
3454 m_freem(m);
3455 return (0);
3456 }
3457 if (ifp != NULL) {
3458 ifp->if_obytes += m->m_pkthdr.len + adjust;
3459 if (m->m_flags & (M_BCAST|M_MCAST))
3460 ifp->if_omcasts++;
3461 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3462 }
3463 _IF_ENQUEUE(ifq, m);
3464 IF_UNLOCK(ifq);
3465 if (ifp != NULL && !active)
3466 (*(ifp)->if_start)(ifp);
3467 return (1);
3468 }
3469
3470 void
3471 if_register_com_alloc(u_char type,
3472 if_com_alloc_t *a, if_com_free_t *f)
3473 {
3474
3475 KASSERT(if_com_alloc[type] == NULL,
3476 ("if_register_com_alloc: %d already registered", type));
3477 KASSERT(if_com_free[type] == NULL,
3478 ("if_register_com_alloc: %d free already registered", type));
3479
3480 if_com_alloc[type] = a;
3481 if_com_free[type] = f;
3482 }
3483
3484 void
3485 if_deregister_com_alloc(u_char type)
3486 {
3487
3488 KASSERT(if_com_alloc[type] != NULL,
3489 ("if_deregister_com_alloc: %d not registered", type));
3490 KASSERT(if_com_free[type] != NULL,
3491 ("if_deregister_com_alloc: %d free not registered", type));
3492 if_com_alloc[type] = NULL;
3493 if_com_free[type] = NULL;
3494 }
Cache object: cc6b5936f9fac83f2973b04017a91adf
|