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: releng/8.3/sys/net/if.c 255446 2013-09-10 10:13:14Z des $
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_UINT(_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 (*bstp_linkstate_p)(struct ifnet *ifp, int state);
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_LOCK(ifp);
839 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
840 if_delmulti_locked(ifp, ifma, 1);
841 IF_ADDR_UNLOCK(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_LOCK(ifp);
1186 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1187 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1188 IF_ADDR_UNLOCK(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_LOCK(ifp);
1216 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1217 IF_ADDR_UNLOCK(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_LOCK(ifp);
1259 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1260 IF_ADDR_UNLOCK(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_LOCK(ifp);
1303 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1304 ifgr->ifgr_len += sizeof(struct ifg_req);
1305 IF_ADDR_UNLOCK(ifp);
1306 return (0);
1307 }
1308
1309 len = ifgr->ifgr_len;
1310 ifgp = ifgr->ifgr_groups;
1311 /* XXX: wire */
1312 IF_ADDR_LOCK(ifp);
1313 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1314 if (len < sizeof(ifgrq)) {
1315 IF_ADDR_UNLOCK(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_UNLOCK(ifp);
1323 return (error);
1324 }
1325 len -= sizeof(ifgrq);
1326 ifgp++;
1327 }
1328 IF_ADDR_UNLOCK(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), rt->rt_flags|RTF_RNH_LOCKED,
1416 (struct rtentry **) NULL, rt->rt_fibnum);
1417 if (err) {
1418 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1419 }
1420 }
1421
1422 return (0);
1423 }
1424
1425 /*
1426 * Wrapper functions for struct ifnet address list locking macros. These are
1427 * used by kernel modules to avoid encoding programming interface or binary
1428 * interface assumptions that may be violated when kernel-internal locking
1429 * approaches change.
1430 */
1431 void
1432 if_addr_rlock(struct ifnet *ifp)
1433 {
1434
1435 IF_ADDR_LOCK(ifp);
1436 }
1437
1438 void
1439 if_addr_runlock(struct ifnet *ifp)
1440 {
1441
1442 IF_ADDR_UNLOCK(ifp);
1443 }
1444
1445 void
1446 if_maddr_rlock(struct ifnet *ifp)
1447 {
1448
1449 IF_ADDR_LOCK(ifp);
1450 }
1451
1452 void
1453 if_maddr_runlock(struct ifnet *ifp)
1454 {
1455
1456 IF_ADDR_UNLOCK(ifp);
1457 }
1458
1459 /*
1460 * Reference count functions for ifaddrs.
1461 */
1462 void
1463 ifa_init(struct ifaddr *ifa)
1464 {
1465
1466 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1467 refcount_init(&ifa->ifa_refcnt, 1);
1468 }
1469
1470 void
1471 ifa_ref(struct ifaddr *ifa)
1472 {
1473
1474 refcount_acquire(&ifa->ifa_refcnt);
1475 }
1476
1477 void
1478 ifa_free(struct ifaddr *ifa)
1479 {
1480
1481 if (refcount_release(&ifa->ifa_refcnt)) {
1482 mtx_destroy(&ifa->ifa_mtx);
1483 free(ifa, M_IFADDR);
1484 }
1485 }
1486
1487 int
1488 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1489 {
1490 int error = 0;
1491 struct rtentry *rt = NULL;
1492 struct rt_addrinfo info;
1493 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1494
1495 bzero(&info, sizeof(info));
1496 info.rti_ifp = V_loif;
1497 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1498 info.rti_info[RTAX_DST] = ia;
1499 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1500 error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1501
1502 if (error == 0 && rt != NULL) {
1503 RT_LOCK(rt);
1504 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1505 ifa->ifa_ifp->if_type;
1506 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1507 ifa->ifa_ifp->if_index;
1508 RT_REMREF(rt);
1509 RT_UNLOCK(rt);
1510 } else if (error != 0)
1511 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1512
1513 return (error);
1514 }
1515
1516 int
1517 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1518 {
1519 int error = 0;
1520 struct rt_addrinfo info;
1521 struct sockaddr_dl null_sdl;
1522
1523 bzero(&null_sdl, sizeof(null_sdl));
1524 null_sdl.sdl_len = sizeof(null_sdl);
1525 null_sdl.sdl_family = AF_LINK;
1526 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1527 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1528 bzero(&info, sizeof(info));
1529 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1530 info.rti_info[RTAX_DST] = ia;
1531 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1532 error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1533
1534 if (error != 0)
1535 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1536
1537 return (error);
1538 }
1539
1540 /*
1541 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1542 * structs used to represent other address families, it is necessary
1543 * to perform a different comparison.
1544 */
1545
1546 #define sa_equal(a1, a2) \
1547 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1548
1549 #define sa_dl_equal(a1, a2) \
1550 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1551 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1552 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1553 LLADDR((struct sockaddr_dl *)(a2)), \
1554 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1555
1556 /*
1557 * Locate an interface based on a complete address.
1558 */
1559 /*ARGSUSED*/
1560 static struct ifaddr *
1561 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1562 {
1563 struct ifnet *ifp;
1564 struct ifaddr *ifa;
1565
1566 IFNET_RLOCK_NOSLEEP();
1567 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1568 IF_ADDR_LOCK(ifp);
1569 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1570 if (ifa->ifa_addr->sa_family != addr->sa_family)
1571 continue;
1572 if (sa_equal(addr, ifa->ifa_addr)) {
1573 if (getref)
1574 ifa_ref(ifa);
1575 IF_ADDR_UNLOCK(ifp);
1576 goto done;
1577 }
1578 /* IP6 doesn't have broadcast */
1579 if ((ifp->if_flags & IFF_BROADCAST) &&
1580 ifa->ifa_broadaddr &&
1581 ifa->ifa_broadaddr->sa_len != 0 &&
1582 sa_equal(ifa->ifa_broadaddr, addr)) {
1583 if (getref)
1584 ifa_ref(ifa);
1585 IF_ADDR_UNLOCK(ifp);
1586 goto done;
1587 }
1588 }
1589 IF_ADDR_UNLOCK(ifp);
1590 }
1591 ifa = NULL;
1592 done:
1593 IFNET_RUNLOCK_NOSLEEP();
1594 return (ifa);
1595 }
1596
1597 struct ifaddr *
1598 ifa_ifwithaddr(struct sockaddr *addr)
1599 {
1600
1601 return (ifa_ifwithaddr_internal(addr, 1));
1602 }
1603
1604 int
1605 ifa_ifwithaddr_check(struct sockaddr *addr)
1606 {
1607
1608 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1609 }
1610
1611 /*
1612 * Locate an interface based on the broadcast address.
1613 */
1614 /* ARGSUSED */
1615 struct ifaddr *
1616 ifa_ifwithbroadaddr(struct sockaddr *addr)
1617 {
1618 struct ifnet *ifp;
1619 struct ifaddr *ifa;
1620
1621 IFNET_RLOCK_NOSLEEP();
1622 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1623 IF_ADDR_LOCK(ifp);
1624 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1625 if (ifa->ifa_addr->sa_family != addr->sa_family)
1626 continue;
1627 if ((ifp->if_flags & IFF_BROADCAST) &&
1628 ifa->ifa_broadaddr &&
1629 ifa->ifa_broadaddr->sa_len != 0 &&
1630 sa_equal(ifa->ifa_broadaddr, addr)) {
1631 ifa_ref(ifa);
1632 IF_ADDR_UNLOCK(ifp);
1633 goto done;
1634 }
1635 }
1636 IF_ADDR_UNLOCK(ifp);
1637 }
1638 ifa = NULL;
1639 done:
1640 IFNET_RUNLOCK_NOSLEEP();
1641 return (ifa);
1642 }
1643
1644 /*
1645 * Locate the point to point interface with a given destination address.
1646 */
1647 /*ARGSUSED*/
1648 struct ifaddr *
1649 ifa_ifwithdstaddr(struct sockaddr *addr)
1650 {
1651 struct ifnet *ifp;
1652 struct ifaddr *ifa;
1653
1654 IFNET_RLOCK_NOSLEEP();
1655 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1656 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1657 continue;
1658 IF_ADDR_LOCK(ifp);
1659 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1660 if (ifa->ifa_addr->sa_family != addr->sa_family)
1661 continue;
1662 if (ifa->ifa_dstaddr != NULL &&
1663 sa_equal(addr, ifa->ifa_dstaddr)) {
1664 ifa_ref(ifa);
1665 IF_ADDR_UNLOCK(ifp);
1666 goto done;
1667 }
1668 }
1669 IF_ADDR_UNLOCK(ifp);
1670 }
1671 ifa = NULL;
1672 done:
1673 IFNET_RUNLOCK_NOSLEEP();
1674 return (ifa);
1675 }
1676
1677 /*
1678 * Find an interface on a specific network. If many, choice
1679 * is most specific found.
1680 */
1681 struct ifaddr *
1682 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1683 {
1684 struct ifnet *ifp;
1685 struct ifaddr *ifa;
1686 struct ifaddr *ifa_maybe = NULL;
1687 u_int af = addr->sa_family;
1688 char *addr_data = addr->sa_data, *cplim;
1689
1690 /*
1691 * AF_LINK addresses can be looked up directly by their index number,
1692 * so do that if we can.
1693 */
1694 if (af == AF_LINK) {
1695 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1696 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1697 return (ifaddr_byindex(sdl->sdl_index));
1698 }
1699
1700 /*
1701 * Scan though each interface, looking for ones that have addresses
1702 * in this address family. Maintain a reference on ifa_maybe once
1703 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1704 * when we move onto the next interface.
1705 */
1706 IFNET_RLOCK_NOSLEEP();
1707 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1708 IF_ADDR_LOCK(ifp);
1709 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1710 char *cp, *cp2, *cp3;
1711
1712 if (ifa->ifa_addr->sa_family != af)
1713 next: continue;
1714 if (af == AF_INET &&
1715 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1716 /*
1717 * This is a bit broken as it doesn't
1718 * take into account that the remote end may
1719 * be a single node in the network we are
1720 * looking for.
1721 * The trouble is that we don't know the
1722 * netmask for the remote end.
1723 */
1724 if (ifa->ifa_dstaddr != NULL &&
1725 sa_equal(addr, ifa->ifa_dstaddr)) {
1726 ifa_ref(ifa);
1727 IF_ADDR_UNLOCK(ifp);
1728 goto done;
1729 }
1730 } else {
1731 /*
1732 * if we have a special address handler,
1733 * then use it instead of the generic one.
1734 */
1735 if (ifa->ifa_claim_addr) {
1736 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1737 ifa_ref(ifa);
1738 IF_ADDR_UNLOCK(ifp);
1739 goto done;
1740 }
1741 continue;
1742 }
1743
1744 /*
1745 * Scan all the bits in the ifa's address.
1746 * If a bit dissagrees with what we are
1747 * looking for, mask it with the netmask
1748 * to see if it really matters.
1749 * (A byte at a time)
1750 */
1751 if (ifa->ifa_netmask == 0)
1752 continue;
1753 cp = addr_data;
1754 cp2 = ifa->ifa_addr->sa_data;
1755 cp3 = ifa->ifa_netmask->sa_data;
1756 cplim = ifa->ifa_netmask->sa_len
1757 + (char *)ifa->ifa_netmask;
1758 while (cp3 < cplim)
1759 if ((*cp++ ^ *cp2++) & *cp3++)
1760 goto next; /* next address! */
1761 /*
1762 * If the netmask of what we just found
1763 * is more specific than what we had before
1764 * (if we had one) then remember the new one
1765 * before continuing to search
1766 * for an even better one.
1767 */
1768 if (ifa_maybe == NULL ||
1769 rn_refines((caddr_t)ifa->ifa_netmask,
1770 (caddr_t)ifa_maybe->ifa_netmask)) {
1771 if (ifa_maybe != NULL)
1772 ifa_free(ifa_maybe);
1773 ifa_maybe = ifa;
1774 ifa_ref(ifa_maybe);
1775 }
1776 }
1777 }
1778 IF_ADDR_UNLOCK(ifp);
1779 }
1780 ifa = ifa_maybe;
1781 ifa_maybe = NULL;
1782 done:
1783 IFNET_RUNLOCK_NOSLEEP();
1784 if (ifa_maybe != NULL)
1785 ifa_free(ifa_maybe);
1786 return (ifa);
1787 }
1788
1789 /*
1790 * Find an interface address specific to an interface best matching
1791 * a given address.
1792 */
1793 struct ifaddr *
1794 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1795 {
1796 struct ifaddr *ifa;
1797 char *cp, *cp2, *cp3;
1798 char *cplim;
1799 struct ifaddr *ifa_maybe = NULL;
1800 u_int af = addr->sa_family;
1801
1802 if (af >= AF_MAX)
1803 return (NULL);
1804 IF_ADDR_LOCK(ifp);
1805 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1806 if (ifa->ifa_addr->sa_family != af)
1807 continue;
1808 if (ifa_maybe == NULL)
1809 ifa_maybe = ifa;
1810 if (ifa->ifa_netmask == 0) {
1811 if (sa_equal(addr, ifa->ifa_addr) ||
1812 (ifa->ifa_dstaddr &&
1813 sa_equal(addr, ifa->ifa_dstaddr)))
1814 goto done;
1815 continue;
1816 }
1817 if (ifp->if_flags & IFF_POINTOPOINT) {
1818 if (sa_equal(addr, ifa->ifa_dstaddr))
1819 goto done;
1820 } else {
1821 cp = addr->sa_data;
1822 cp2 = ifa->ifa_addr->sa_data;
1823 cp3 = ifa->ifa_netmask->sa_data;
1824 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1825 for (; cp3 < cplim; cp3++)
1826 if ((*cp++ ^ *cp2++) & *cp3)
1827 break;
1828 if (cp3 == cplim)
1829 goto done;
1830 }
1831 }
1832 ifa = ifa_maybe;
1833 done:
1834 if (ifa != NULL)
1835 ifa_ref(ifa);
1836 IF_ADDR_UNLOCK(ifp);
1837 return (ifa);
1838 }
1839
1840 #include <net/if_llatbl.h>
1841
1842 /*
1843 * Default action when installing a route with a Link Level gateway.
1844 * Lookup an appropriate real ifa to point to.
1845 * This should be moved to /sys/net/link.c eventually.
1846 */
1847 static void
1848 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1849 {
1850 struct ifaddr *ifa, *oifa;
1851 struct sockaddr *dst;
1852 struct ifnet *ifp;
1853
1854 RT_LOCK_ASSERT(rt);
1855
1856 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1857 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1858 return;
1859 ifa = ifaof_ifpforaddr(dst, ifp);
1860 if (ifa) {
1861 oifa = rt->rt_ifa;
1862 rt->rt_ifa = ifa;
1863 ifa_free(oifa);
1864 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1865 ifa->ifa_rtrequest(cmd, rt, info);
1866 }
1867 }
1868
1869 /*
1870 * Mark an interface down and notify protocols of
1871 * the transition.
1872 * NOTE: must be called at splnet or eqivalent.
1873 */
1874 static void
1875 if_unroute(struct ifnet *ifp, int flag, int fam)
1876 {
1877 struct ifaddr *ifa;
1878
1879 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1880
1881 ifp->if_flags &= ~flag;
1882 getmicrotime(&ifp->if_lastchange);
1883 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1884 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1885 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1886 ifp->if_qflush(ifp);
1887
1888 if (ifp->if_carp)
1889 (*carp_linkstate_p)(ifp);
1890 rt_ifmsg(ifp);
1891 }
1892
1893 /*
1894 * Mark an interface up and notify protocols of
1895 * the transition.
1896 * NOTE: must be called at splnet or eqivalent.
1897 */
1898 static void
1899 if_route(struct ifnet *ifp, int flag, int fam)
1900 {
1901 struct ifaddr *ifa;
1902
1903 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1904
1905 ifp->if_flags |= flag;
1906 getmicrotime(&ifp->if_lastchange);
1907 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1908 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1909 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1910 if (ifp->if_carp)
1911 (*carp_linkstate_p)(ifp);
1912 rt_ifmsg(ifp);
1913 #ifdef INET6
1914 in6_if_up(ifp);
1915 #endif
1916 }
1917
1918 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1919 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
1920
1921 /*
1922 * Handle a change in the interface link state. To avoid LORs
1923 * between driver lock and upper layer locks, as well as possible
1924 * recursions, we post event to taskqueue, and all job
1925 * is done in static do_link_state_change().
1926 */
1927 void
1928 if_link_state_change(struct ifnet *ifp, int link_state)
1929 {
1930 /* Return if state hasn't changed. */
1931 if (ifp->if_link_state == link_state)
1932 return;
1933
1934 ifp->if_link_state = link_state;
1935
1936 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1937 }
1938
1939 static void
1940 do_link_state_change(void *arg, int pending)
1941 {
1942 struct ifnet *ifp = (struct ifnet *)arg;
1943 int link_state = ifp->if_link_state;
1944 CURVNET_SET(ifp->if_vnet);
1945
1946 /* Notify that the link state has changed. */
1947 rt_ifmsg(ifp);
1948 if (ifp->if_vlantrunk != NULL)
1949 (*vlan_link_state_p)(ifp, 0);
1950
1951 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1952 IFP2AC(ifp)->ac_netgraph != NULL)
1953 (*ng_ether_link_state_p)(ifp, link_state);
1954 if (ifp->if_carp)
1955 (*carp_linkstate_p)(ifp);
1956 if (ifp->if_bridge) {
1957 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1958 (*bstp_linkstate_p)(ifp, link_state);
1959 }
1960 if (ifp->if_lagg) {
1961 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1962 (*lagg_linkstate_p)(ifp, link_state);
1963 }
1964
1965 if (IS_DEFAULT_VNET(curvnet))
1966 devctl_notify("IFNET", ifp->if_xname,
1967 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1968 NULL);
1969 if (pending > 1)
1970 if_printf(ifp, "%d link states coalesced\n", pending);
1971 if (log_link_state_change)
1972 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1973 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1974 CURVNET_RESTORE();
1975 }
1976
1977 /*
1978 * Mark an interface down and notify protocols of
1979 * the transition.
1980 * NOTE: must be called at splnet or eqivalent.
1981 */
1982 void
1983 if_down(struct ifnet *ifp)
1984 {
1985
1986 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1987 }
1988
1989 /*
1990 * Mark an interface up and notify protocols of
1991 * the transition.
1992 * NOTE: must be called at splnet or eqivalent.
1993 */
1994 void
1995 if_up(struct ifnet *ifp)
1996 {
1997
1998 if_route(ifp, IFF_UP, AF_UNSPEC);
1999 }
2000
2001 /*
2002 * Flush an interface queue.
2003 */
2004 void
2005 if_qflush(struct ifnet *ifp)
2006 {
2007 struct mbuf *m, *n;
2008 struct ifaltq *ifq;
2009
2010 ifq = &ifp->if_snd;
2011 IFQ_LOCK(ifq);
2012 #ifdef ALTQ
2013 if (ALTQ_IS_ENABLED(ifq))
2014 ALTQ_PURGE(ifq);
2015 #endif
2016 n = ifq->ifq_head;
2017 while ((m = n) != 0) {
2018 n = m->m_act;
2019 m_freem(m);
2020 }
2021 ifq->ifq_head = 0;
2022 ifq->ifq_tail = 0;
2023 ifq->ifq_len = 0;
2024 IFQ_UNLOCK(ifq);
2025 }
2026
2027 /*
2028 * Handle interface watchdog timer routines. Called
2029 * from softclock, we decrement timers (if set) and
2030 * call the appropriate interface routine on expiration.
2031 *
2032 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
2033 * holding Giant.
2034 */
2035 static void
2036 if_slowtimo(void *arg)
2037 {
2038 VNET_ITERATOR_DECL(vnet_iter);
2039 struct ifnet *ifp;
2040 int s = splimp();
2041
2042 VNET_LIST_RLOCK_NOSLEEP();
2043 IFNET_RLOCK_NOSLEEP();
2044 VNET_FOREACH(vnet_iter) {
2045 CURVNET_SET(vnet_iter);
2046 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2047 if (ifp->if_timer == 0 || --ifp->if_timer)
2048 continue;
2049 if (ifp->if_watchdog)
2050 (*ifp->if_watchdog)(ifp);
2051 }
2052 CURVNET_RESTORE();
2053 }
2054 IFNET_RUNLOCK_NOSLEEP();
2055 VNET_LIST_RUNLOCK_NOSLEEP();
2056 splx(s);
2057 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
2058 }
2059
2060 /*
2061 * Map interface name to interface structure pointer, with or without
2062 * returning a reference.
2063 */
2064 struct ifnet *
2065 ifunit_ref(const char *name)
2066 {
2067 struct ifnet *ifp;
2068
2069 IFNET_RLOCK_NOSLEEP();
2070 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2071 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2072 !(ifp->if_flags & IFF_DYING))
2073 break;
2074 }
2075 if (ifp != NULL)
2076 if_ref(ifp);
2077 IFNET_RUNLOCK_NOSLEEP();
2078 return (ifp);
2079 }
2080
2081 struct ifnet *
2082 ifunit(const char *name)
2083 {
2084 struct ifnet *ifp;
2085
2086 IFNET_RLOCK_NOSLEEP();
2087 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2088 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2089 break;
2090 }
2091 IFNET_RUNLOCK_NOSLEEP();
2092 return (ifp);
2093 }
2094
2095 /*
2096 * Hardware specific interface ioctls.
2097 */
2098 static int
2099 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2100 {
2101 struct ifreq *ifr;
2102 struct ifstat *ifs;
2103 int error = 0;
2104 int new_flags, temp_flags;
2105 size_t namelen, onamelen;
2106 size_t descrlen;
2107 char *descrbuf, *odescrbuf;
2108 char new_name[IFNAMSIZ];
2109 struct ifaddr *ifa;
2110 struct sockaddr_dl *sdl;
2111
2112 ifr = (struct ifreq *)data;
2113 switch (cmd) {
2114 case SIOCGIFINDEX:
2115 ifr->ifr_index = ifp->if_index;
2116 break;
2117
2118 case SIOCGIFFLAGS:
2119 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2120 ifr->ifr_flags = temp_flags & 0xffff;
2121 ifr->ifr_flagshigh = temp_flags >> 16;
2122 break;
2123
2124 case SIOCGIFCAP:
2125 ifr->ifr_reqcap = ifp->if_capabilities;
2126 ifr->ifr_curcap = ifp->if_capenable;
2127 break;
2128
2129 #ifdef MAC
2130 case SIOCGIFMAC:
2131 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2132 break;
2133 #endif
2134
2135 case SIOCGIFMETRIC:
2136 ifr->ifr_metric = ifp->if_metric;
2137 break;
2138
2139 case SIOCGIFMTU:
2140 ifr->ifr_mtu = ifp->if_mtu;
2141 break;
2142
2143 case SIOCGIFPHYS:
2144 ifr->ifr_phys = ifp->if_physical;
2145 break;
2146
2147 case SIOCGIFDESCR:
2148 error = 0;
2149 sx_slock(&ifdescr_sx);
2150 if (ifp->if_description == NULL)
2151 error = ENOMSG;
2152 else {
2153 /* space for terminating nul */
2154 descrlen = strlen(ifp->if_description) + 1;
2155 if (ifr->ifr_buffer.length < descrlen)
2156 ifr->ifr_buffer.buffer = NULL;
2157 else
2158 error = copyout(ifp->if_description,
2159 ifr->ifr_buffer.buffer, descrlen);
2160 ifr->ifr_buffer.length = descrlen;
2161 }
2162 sx_sunlock(&ifdescr_sx);
2163 break;
2164
2165 case SIOCSIFDESCR:
2166 error = priv_check(td, PRIV_NET_SETIFDESCR);
2167 if (error)
2168 return (error);
2169
2170 /*
2171 * Copy only (length-1) bytes to make sure that
2172 * if_description is always nul terminated. The
2173 * length parameter is supposed to count the
2174 * terminating nul in.
2175 */
2176 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2177 return (ENAMETOOLONG);
2178 else if (ifr->ifr_buffer.length == 0)
2179 descrbuf = NULL;
2180 else {
2181 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2182 M_WAITOK | M_ZERO);
2183 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2184 ifr->ifr_buffer.length - 1);
2185 if (error) {
2186 free(descrbuf, M_IFDESCR);
2187 break;
2188 }
2189 }
2190
2191 sx_xlock(&ifdescr_sx);
2192 odescrbuf = ifp->if_description;
2193 ifp->if_description = descrbuf;
2194 sx_xunlock(&ifdescr_sx);
2195
2196 getmicrotime(&ifp->if_lastchange);
2197 free(odescrbuf, M_IFDESCR);
2198 break;
2199
2200 case SIOCGIFFIB:
2201 ifr->ifr_fib = ifp->if_fib;
2202 break;
2203
2204 case SIOCSIFFIB:
2205 error = priv_check(td, PRIV_NET_SETIFFIB);
2206 if (error)
2207 return (error);
2208 if (ifr->ifr_fib >= rt_numfibs)
2209 return (EINVAL);
2210
2211 ifp->if_fib = ifr->ifr_fib;
2212 break;
2213
2214 case SIOCSIFFLAGS:
2215 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2216 if (error)
2217 return (error);
2218 /*
2219 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2220 * check, so we don't need special handling here yet.
2221 */
2222 new_flags = (ifr->ifr_flags & 0xffff) |
2223 (ifr->ifr_flagshigh << 16);
2224 if (ifp->if_flags & IFF_SMART) {
2225 /* Smart drivers twiddle their own routes */
2226 } else if (ifp->if_flags & IFF_UP &&
2227 (new_flags & IFF_UP) == 0) {
2228 int s = splimp();
2229 if_down(ifp);
2230 splx(s);
2231 } else if (new_flags & IFF_UP &&
2232 (ifp->if_flags & IFF_UP) == 0) {
2233 int s = splimp();
2234 if_up(ifp);
2235 splx(s);
2236 }
2237 /* See if permanently promiscuous mode bit is about to flip */
2238 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2239 if (new_flags & IFF_PPROMISC)
2240 ifp->if_flags |= IFF_PROMISC;
2241 else if (ifp->if_pcount == 0)
2242 ifp->if_flags &= ~IFF_PROMISC;
2243 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2244 ifp->if_xname,
2245 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2246 }
2247 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2248 (new_flags &~ IFF_CANTCHANGE);
2249 if (ifp->if_ioctl) {
2250 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2251 }
2252 getmicrotime(&ifp->if_lastchange);
2253 break;
2254
2255 case SIOCSIFCAP:
2256 error = priv_check(td, PRIV_NET_SETIFCAP);
2257 if (error)
2258 return (error);
2259 if (ifp->if_ioctl == NULL)
2260 return (EOPNOTSUPP);
2261 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2262 return (EINVAL);
2263 error = (*ifp->if_ioctl)(ifp, cmd, data);
2264 if (error == 0)
2265 getmicrotime(&ifp->if_lastchange);
2266 break;
2267
2268 #ifdef MAC
2269 case SIOCSIFMAC:
2270 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2271 break;
2272 #endif
2273
2274 case SIOCSIFNAME:
2275 error = priv_check(td, PRIV_NET_SETIFNAME);
2276 if (error)
2277 return (error);
2278 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2279 if (error != 0)
2280 return (error);
2281 if (new_name[0] == '\0')
2282 return (EINVAL);
2283 if (ifunit(new_name) != NULL)
2284 return (EEXIST);
2285
2286 /*
2287 * XXX: Locking. Nothing else seems to lock if_flags,
2288 * and there are numerous other races with the
2289 * ifunit() checks not being atomic with namespace
2290 * changes (renames, vmoves, if_attach, etc).
2291 */
2292 ifp->if_flags |= IFF_RENAMING;
2293
2294 /* Announce the departure of the interface. */
2295 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2296 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2297
2298 log(LOG_INFO, "%s: changing name to '%s'\n",
2299 ifp->if_xname, new_name);
2300
2301 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2302 ifa = ifp->if_addr;
2303 IFA_LOCK(ifa);
2304 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2305 namelen = strlen(new_name);
2306 onamelen = sdl->sdl_nlen;
2307 /*
2308 * Move the address if needed. This is safe because we
2309 * allocate space for a name of length IFNAMSIZ when we
2310 * create this in if_attach().
2311 */
2312 if (namelen != onamelen) {
2313 bcopy(sdl->sdl_data + onamelen,
2314 sdl->sdl_data + namelen, sdl->sdl_alen);
2315 }
2316 bcopy(new_name, sdl->sdl_data, namelen);
2317 sdl->sdl_nlen = namelen;
2318 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2319 bzero(sdl->sdl_data, onamelen);
2320 while (namelen != 0)
2321 sdl->sdl_data[--namelen] = 0xff;
2322 IFA_UNLOCK(ifa);
2323
2324 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2325 /* Announce the return of the interface. */
2326 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2327
2328 ifp->if_flags &= ~IFF_RENAMING;
2329 break;
2330
2331 #ifdef VIMAGE
2332 case SIOCSIFVNET:
2333 error = priv_check(td, PRIV_NET_SETIFVNET);
2334 if (error)
2335 return (error);
2336 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2337 break;
2338 #endif
2339
2340 case SIOCSIFMETRIC:
2341 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2342 if (error)
2343 return (error);
2344 ifp->if_metric = ifr->ifr_metric;
2345 getmicrotime(&ifp->if_lastchange);
2346 break;
2347
2348 case SIOCSIFPHYS:
2349 error = priv_check(td, PRIV_NET_SETIFPHYS);
2350 if (error)
2351 return (error);
2352 if (ifp->if_ioctl == NULL)
2353 return (EOPNOTSUPP);
2354 error = (*ifp->if_ioctl)(ifp, cmd, data);
2355 if (error == 0)
2356 getmicrotime(&ifp->if_lastchange);
2357 break;
2358
2359 case SIOCSIFMTU:
2360 {
2361 u_long oldmtu = ifp->if_mtu;
2362
2363 error = priv_check(td, PRIV_NET_SETIFMTU);
2364 if (error)
2365 return (error);
2366 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2367 return (EINVAL);
2368 if (ifp->if_ioctl == NULL)
2369 return (EOPNOTSUPP);
2370 error = (*ifp->if_ioctl)(ifp, cmd, data);
2371 if (error == 0) {
2372 getmicrotime(&ifp->if_lastchange);
2373 rt_ifmsg(ifp);
2374 }
2375 /*
2376 * If the link MTU changed, do network layer specific procedure.
2377 */
2378 if (ifp->if_mtu != oldmtu) {
2379 #ifdef INET6
2380 nd6_setmtu(ifp);
2381 #endif
2382 }
2383 break;
2384 }
2385
2386 case SIOCADDMULTI:
2387 case SIOCDELMULTI:
2388 if (cmd == SIOCADDMULTI)
2389 error = priv_check(td, PRIV_NET_ADDMULTI);
2390 else
2391 error = priv_check(td, PRIV_NET_DELMULTI);
2392 if (error)
2393 return (error);
2394
2395 /* Don't allow group membership on non-multicast interfaces. */
2396 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2397 return (EOPNOTSUPP);
2398
2399 /* Don't let users screw up protocols' entries. */
2400 if (ifr->ifr_addr.sa_family != AF_LINK)
2401 return (EINVAL);
2402
2403 if (cmd == SIOCADDMULTI) {
2404 struct ifmultiaddr *ifma;
2405
2406 /*
2407 * Userland is only permitted to join groups once
2408 * via the if_addmulti() KPI, because it cannot hold
2409 * struct ifmultiaddr * between calls. It may also
2410 * lose a race while we check if the membership
2411 * already exists.
2412 */
2413 IF_ADDR_LOCK(ifp);
2414 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2415 IF_ADDR_UNLOCK(ifp);
2416 if (ifma != NULL)
2417 error = EADDRINUSE;
2418 else
2419 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2420 } else {
2421 error = if_delmulti(ifp, &ifr->ifr_addr);
2422 }
2423 if (error == 0)
2424 getmicrotime(&ifp->if_lastchange);
2425 break;
2426
2427 case SIOCSIFPHYADDR:
2428 case SIOCDIFPHYADDR:
2429 #ifdef INET6
2430 case SIOCSIFPHYADDR_IN6:
2431 #endif
2432 case SIOCSLIFPHYADDR:
2433 case SIOCSIFMEDIA:
2434 case SIOCSIFGENERIC:
2435 error = priv_check(td, PRIV_NET_HWIOCTL);
2436 if (error)
2437 return (error);
2438 if (ifp->if_ioctl == NULL)
2439 return (EOPNOTSUPP);
2440 error = (*ifp->if_ioctl)(ifp, cmd, data);
2441 if (error == 0)
2442 getmicrotime(&ifp->if_lastchange);
2443 break;
2444
2445 case SIOCGIFSTATUS:
2446 ifs = (struct ifstat *)data;
2447 ifs->ascii[0] = '\0';
2448
2449 case SIOCGIFPSRCADDR:
2450 case SIOCGIFPDSTADDR:
2451 case SIOCGLIFPHYADDR:
2452 case SIOCGIFMEDIA:
2453 case SIOCGIFGENERIC:
2454 if (ifp->if_ioctl == NULL)
2455 return (EOPNOTSUPP);
2456 error = (*ifp->if_ioctl)(ifp, cmd, data);
2457 break;
2458
2459 case SIOCSIFLLADDR:
2460 error = priv_check(td, PRIV_NET_SETLLADDR);
2461 if (error)
2462 return (error);
2463 error = if_setlladdr(ifp,
2464 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2465 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2466 break;
2467
2468 case SIOCAIFGROUP:
2469 {
2470 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2471
2472 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2473 if (error)
2474 return (error);
2475 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2476 return (error);
2477 break;
2478 }
2479
2480 case SIOCGIFGROUP:
2481 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2482 return (error);
2483 break;
2484
2485 case SIOCDIFGROUP:
2486 {
2487 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2488
2489 error = priv_check(td, PRIV_NET_DELIFGROUP);
2490 if (error)
2491 return (error);
2492 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2493 return (error);
2494 break;
2495 }
2496
2497 default:
2498 error = ENOIOCTL;
2499 break;
2500 }
2501 return (error);
2502 }
2503
2504 #ifdef COMPAT_FREEBSD32
2505 struct ifconf32 {
2506 int32_t ifc_len;
2507 union {
2508 uint32_t ifcu_buf;
2509 uint32_t ifcu_req;
2510 } ifc_ifcu;
2511 };
2512 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2513 #endif
2514
2515 /*
2516 * Interface ioctls.
2517 */
2518 int
2519 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2520 {
2521 struct ifnet *ifp;
2522 struct ifreq *ifr;
2523 int error;
2524 int oif_flags;
2525
2526 CURVNET_SET(so->so_vnet);
2527 switch (cmd) {
2528 case SIOCGIFCONF:
2529 case OSIOCGIFCONF:
2530 error = ifconf(cmd, data);
2531 CURVNET_RESTORE();
2532 return (error);
2533
2534 #ifdef COMPAT_FREEBSD32
2535 case SIOCGIFCONF32:
2536 {
2537 struct ifconf32 *ifc32;
2538 struct ifconf ifc;
2539
2540 ifc32 = (struct ifconf32 *)data;
2541 ifc.ifc_len = ifc32->ifc_len;
2542 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2543
2544 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2545 CURVNET_RESTORE();
2546 if (error == 0)
2547 ifc32->ifc_len = ifc.ifc_len;
2548 return (error);
2549 }
2550 #endif
2551 }
2552 ifr = (struct ifreq *)data;
2553
2554 switch (cmd) {
2555 #ifdef VIMAGE
2556 case SIOCSIFRVNET:
2557 error = priv_check(td, PRIV_NET_SETIFVNET);
2558 if (error == 0)
2559 error = if_vmove_reclaim(td, ifr->ifr_name,
2560 ifr->ifr_jid);
2561 CURVNET_RESTORE();
2562 return (error);
2563 #endif
2564 case SIOCIFCREATE:
2565 case SIOCIFCREATE2:
2566 error = priv_check(td, PRIV_NET_IFCREATE);
2567 if (error == 0)
2568 error = if_clone_create(ifr->ifr_name,
2569 sizeof(ifr->ifr_name),
2570 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2571 CURVNET_RESTORE();
2572 return (error);
2573 case SIOCIFDESTROY:
2574 error = priv_check(td, PRIV_NET_IFDESTROY);
2575 if (error == 0)
2576 error = if_clone_destroy(ifr->ifr_name);
2577 CURVNET_RESTORE();
2578 return (error);
2579
2580 case SIOCIFGCLONERS:
2581 error = if_clone_list((struct if_clonereq *)data);
2582 CURVNET_RESTORE();
2583 return (error);
2584 case SIOCGIFGMEMB:
2585 error = if_getgroupmembers((struct ifgroupreq *)data);
2586 CURVNET_RESTORE();
2587 return (error);
2588 }
2589
2590 ifp = ifunit_ref(ifr->ifr_name);
2591 if (ifp == NULL) {
2592 CURVNET_RESTORE();
2593 return (ENXIO);
2594 }
2595
2596 error = ifhwioctl(cmd, ifp, data, td);
2597 if (error != ENOIOCTL) {
2598 if_rele(ifp);
2599 CURVNET_RESTORE();
2600 return (error);
2601 }
2602
2603 oif_flags = ifp->if_flags;
2604 if (so->so_proto == NULL) {
2605 if_rele(ifp);
2606 CURVNET_RESTORE();
2607 return (EOPNOTSUPP);
2608 }
2609
2610 /*
2611 * Pass the request on to the socket control method, and if the
2612 * latter returns EOPNOTSUPP, directly to the interface.
2613 *
2614 * Make an exception for the legacy SIOCSIF* requests. Drivers
2615 * trust SIOCSIFADDR et al to come from an already privileged
2616 * layer, and do not perform any credentials checks or input
2617 * validation.
2618 */
2619 #ifndef COMPAT_43
2620 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2621 data,
2622 ifp, td));
2623 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2624 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2625 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2626 error = (*ifp->if_ioctl)(ifp, cmd, data);
2627 #else
2628 {
2629 u_long ocmd = cmd;
2630
2631 switch (cmd) {
2632
2633 case SIOCSIFDSTADDR:
2634 case SIOCSIFADDR:
2635 case SIOCSIFBRDADDR:
2636 case SIOCSIFNETMASK:
2637 #if BYTE_ORDER != BIG_ENDIAN
2638 if (ifr->ifr_addr.sa_family == 0 &&
2639 ifr->ifr_addr.sa_len < 16) {
2640 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2641 ifr->ifr_addr.sa_len = 16;
2642 }
2643 #else
2644 if (ifr->ifr_addr.sa_len == 0)
2645 ifr->ifr_addr.sa_len = 16;
2646 #endif
2647 break;
2648
2649 case OSIOCGIFADDR:
2650 cmd = SIOCGIFADDR;
2651 break;
2652
2653 case OSIOCGIFDSTADDR:
2654 cmd = SIOCGIFDSTADDR;
2655 break;
2656
2657 case OSIOCGIFBRDADDR:
2658 cmd = SIOCGIFBRDADDR;
2659 break;
2660
2661 case OSIOCGIFNETMASK:
2662 cmd = SIOCGIFNETMASK;
2663 }
2664 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2665 cmd,
2666 data,
2667 ifp, td));
2668 if (error == EOPNOTSUPP && ifp != NULL &&
2669 ifp->if_ioctl != NULL &&
2670 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2671 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2672 error = (*ifp->if_ioctl)(ifp, cmd, data);
2673 switch (ocmd) {
2674
2675 case OSIOCGIFADDR:
2676 case OSIOCGIFDSTADDR:
2677 case OSIOCGIFBRDADDR:
2678 case OSIOCGIFNETMASK:
2679 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2680
2681 }
2682 }
2683 #endif /* COMPAT_43 */
2684
2685 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2686 #ifdef INET6
2687 if (ifp->if_flags & IFF_UP) {
2688 int s = splimp();
2689 in6_if_up(ifp);
2690 splx(s);
2691 }
2692 #endif
2693 }
2694 if_rele(ifp);
2695 CURVNET_RESTORE();
2696 return (error);
2697 }
2698
2699 /*
2700 * The code common to handling reference counted flags,
2701 * e.g., in ifpromisc() and if_allmulti().
2702 * The "pflag" argument can specify a permanent mode flag to check,
2703 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2704 *
2705 * Only to be used on stack-owned flags, not driver-owned flags.
2706 */
2707 static int
2708 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2709 {
2710 struct ifreq ifr;
2711 int error;
2712 int oldflags, oldcount;
2713
2714 /* Sanity checks to catch programming errors */
2715 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2716 ("%s: setting driver-owned flag %d", __func__, flag));
2717
2718 if (onswitch)
2719 KASSERT(*refcount >= 0,
2720 ("%s: increment negative refcount %d for flag %d",
2721 __func__, *refcount, flag));
2722 else
2723 KASSERT(*refcount > 0,
2724 ("%s: decrement non-positive refcount %d for flag %d",
2725 __func__, *refcount, flag));
2726
2727 /* In case this mode is permanent, just touch refcount */
2728 if (ifp->if_flags & pflag) {
2729 *refcount += onswitch ? 1 : -1;
2730 return (0);
2731 }
2732
2733 /* Save ifnet parameters for if_ioctl() may fail */
2734 oldcount = *refcount;
2735 oldflags = ifp->if_flags;
2736
2737 /*
2738 * See if we aren't the only and touching refcount is enough.
2739 * Actually toggle interface flag if we are the first or last.
2740 */
2741 if (onswitch) {
2742 if ((*refcount)++)
2743 return (0);
2744 ifp->if_flags |= flag;
2745 } else {
2746 if (--(*refcount))
2747 return (0);
2748 ifp->if_flags &= ~flag;
2749 }
2750
2751 /* Call down the driver since we've changed interface flags */
2752 if (ifp->if_ioctl == NULL) {
2753 error = EOPNOTSUPP;
2754 goto recover;
2755 }
2756 ifr.ifr_flags = ifp->if_flags & 0xffff;
2757 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2758 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2759 if (error)
2760 goto recover;
2761 /* Notify userland that interface flags have changed */
2762 rt_ifmsg(ifp);
2763 return (0);
2764
2765 recover:
2766 /* Recover after driver error */
2767 *refcount = oldcount;
2768 ifp->if_flags = oldflags;
2769 return (error);
2770 }
2771
2772 /*
2773 * Set/clear promiscuous mode on interface ifp based on the truth value
2774 * of pswitch. The calls are reference counted so that only the first
2775 * "on" request actually has an effect, as does the final "off" request.
2776 * Results are undefined if the "off" and "on" requests are not matched.
2777 */
2778 int
2779 ifpromisc(struct ifnet *ifp, int pswitch)
2780 {
2781 int error;
2782 int oldflags = ifp->if_flags;
2783
2784 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2785 &ifp->if_pcount, pswitch);
2786 /* If promiscuous mode status has changed, log a message */
2787 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2788 log(LOG_INFO, "%s: promiscuous mode %s\n",
2789 ifp->if_xname,
2790 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2791 return (error);
2792 }
2793
2794 /*
2795 * Return interface configuration
2796 * of system. List may be used
2797 * in later ioctl's (above) to get
2798 * other information.
2799 */
2800 /*ARGSUSED*/
2801 static int
2802 ifconf(u_long cmd, caddr_t data)
2803 {
2804 struct ifconf *ifc = (struct ifconf *)data;
2805 struct ifnet *ifp;
2806 struct ifaddr *ifa;
2807 struct ifreq ifr;
2808 struct sbuf *sb;
2809 int error, full = 0, valid_len, max_len;
2810
2811 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2812 max_len = MAXPHYS - 1;
2813
2814 /* Prevent hostile input from being able to crash the system */
2815 if (ifc->ifc_len <= 0)
2816 return (EINVAL);
2817
2818 again:
2819 if (ifc->ifc_len <= max_len) {
2820 max_len = ifc->ifc_len;
2821 full = 1;
2822 }
2823 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2824 max_len = 0;
2825 valid_len = 0;
2826
2827 IFNET_RLOCK();
2828 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2829 int addrs;
2830
2831 /*
2832 * Zero the ifr_name buffer to make sure we don't
2833 * disclose the contents of the stack.
2834 */
2835 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2836
2837 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2838 >= sizeof(ifr.ifr_name)) {
2839 sbuf_delete(sb);
2840 IFNET_RUNLOCK();
2841 return (ENAMETOOLONG);
2842 }
2843
2844 addrs = 0;
2845 IF_ADDR_LOCK(ifp);
2846 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2847 struct sockaddr *sa = ifa->ifa_addr;
2848
2849 if (prison_if(curthread->td_ucred, sa) != 0)
2850 continue;
2851 addrs++;
2852 #ifdef COMPAT_43
2853 if (cmd == OSIOCGIFCONF) {
2854 struct osockaddr *osa =
2855 (struct osockaddr *)&ifr.ifr_addr;
2856 ifr.ifr_addr = *sa;
2857 osa->sa_family = sa->sa_family;
2858 sbuf_bcat(sb, &ifr, sizeof(ifr));
2859 max_len += sizeof(ifr);
2860 } else
2861 #endif
2862 if (sa->sa_len <= sizeof(*sa)) {
2863 ifr.ifr_addr = *sa;
2864 sbuf_bcat(sb, &ifr, sizeof(ifr));
2865 max_len += sizeof(ifr);
2866 } else {
2867 sbuf_bcat(sb, &ifr,
2868 offsetof(struct ifreq, ifr_addr));
2869 max_len += offsetof(struct ifreq, ifr_addr);
2870 sbuf_bcat(sb, sa, sa->sa_len);
2871 max_len += sa->sa_len;
2872 }
2873
2874 if (!sbuf_overflowed(sb))
2875 valid_len = sbuf_len(sb);
2876 }
2877 IF_ADDR_UNLOCK(ifp);
2878 if (addrs == 0) {
2879 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2880 sbuf_bcat(sb, &ifr, sizeof(ifr));
2881 max_len += sizeof(ifr);
2882
2883 if (!sbuf_overflowed(sb))
2884 valid_len = sbuf_len(sb);
2885 }
2886 }
2887 IFNET_RUNLOCK();
2888
2889 /*
2890 * If we didn't allocate enough space (uncommon), try again. If
2891 * we have already allocated as much space as we are allowed,
2892 * return what we've got.
2893 */
2894 if (valid_len != max_len && !full) {
2895 sbuf_delete(sb);
2896 goto again;
2897 }
2898
2899 ifc->ifc_len = valid_len;
2900 sbuf_finish(sb);
2901 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2902 sbuf_delete(sb);
2903 return (error);
2904 }
2905
2906 /*
2907 * Just like ifpromisc(), but for all-multicast-reception mode.
2908 */
2909 int
2910 if_allmulti(struct ifnet *ifp, int onswitch)
2911 {
2912
2913 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2914 }
2915
2916 struct ifmultiaddr *
2917 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2918 {
2919 struct ifmultiaddr *ifma;
2920
2921 IF_ADDR_LOCK_ASSERT(ifp);
2922
2923 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2924 if (sa->sa_family == AF_LINK) {
2925 if (sa_dl_equal(ifma->ifma_addr, sa))
2926 break;
2927 } else {
2928 if (sa_equal(ifma->ifma_addr, sa))
2929 break;
2930 }
2931 }
2932
2933 return ifma;
2934 }
2935
2936 /*
2937 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2938 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2939 * the ifnet multicast address list here, so the caller must do that and
2940 * other setup work (such as notifying the device driver). The reference
2941 * count is initialized to 1.
2942 */
2943 static struct ifmultiaddr *
2944 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2945 int mflags)
2946 {
2947 struct ifmultiaddr *ifma;
2948 struct sockaddr *dupsa;
2949
2950 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2951 M_ZERO);
2952 if (ifma == NULL)
2953 return (NULL);
2954
2955 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2956 if (dupsa == NULL) {
2957 free(ifma, M_IFMADDR);
2958 return (NULL);
2959 }
2960 bcopy(sa, dupsa, sa->sa_len);
2961 ifma->ifma_addr = dupsa;
2962
2963 ifma->ifma_ifp = ifp;
2964 ifma->ifma_refcount = 1;
2965 ifma->ifma_protospec = NULL;
2966
2967 if (llsa == NULL) {
2968 ifma->ifma_lladdr = NULL;
2969 return (ifma);
2970 }
2971
2972 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2973 if (dupsa == NULL) {
2974 free(ifma->ifma_addr, M_IFMADDR);
2975 free(ifma, M_IFMADDR);
2976 return (NULL);
2977 }
2978 bcopy(llsa, dupsa, llsa->sa_len);
2979 ifma->ifma_lladdr = dupsa;
2980
2981 return (ifma);
2982 }
2983
2984 /*
2985 * if_freemulti: free ifmultiaddr structure and possibly attached related
2986 * addresses. The caller is responsible for implementing reference
2987 * counting, notifying the driver, handling routing messages, and releasing
2988 * any dependent link layer state.
2989 */
2990 static void
2991 if_freemulti(struct ifmultiaddr *ifma)
2992 {
2993
2994 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2995 ifma->ifma_refcount));
2996 KASSERT(ifma->ifma_protospec == NULL,
2997 ("if_freemulti: protospec not NULL"));
2998
2999 if (ifma->ifma_lladdr != NULL)
3000 free(ifma->ifma_lladdr, M_IFMADDR);
3001 free(ifma->ifma_addr, M_IFMADDR);
3002 free(ifma, M_IFMADDR);
3003 }
3004
3005 /*
3006 * Register an additional multicast address with a network interface.
3007 *
3008 * - If the address is already present, bump the reference count on the
3009 * address and return.
3010 * - If the address is not link-layer, look up a link layer address.
3011 * - Allocate address structures for one or both addresses, and attach to the
3012 * multicast address list on the interface. If automatically adding a link
3013 * layer address, the protocol address will own a reference to the link
3014 * layer address, to be freed when it is freed.
3015 * - Notify the network device driver of an addition to the multicast address
3016 * list.
3017 *
3018 * 'sa' points to caller-owned memory with the desired multicast address.
3019 *
3020 * 'retifma' will be used to return a pointer to the resulting multicast
3021 * address reference, if desired.
3022 */
3023 int
3024 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3025 struct ifmultiaddr **retifma)
3026 {
3027 struct ifmultiaddr *ifma, *ll_ifma;
3028 struct sockaddr *llsa;
3029 int error;
3030
3031 /*
3032 * If the address is already present, return a new reference to it;
3033 * otherwise, allocate storage and set up a new address.
3034 */
3035 IF_ADDR_LOCK(ifp);
3036 ifma = if_findmulti(ifp, sa);
3037 if (ifma != NULL) {
3038 ifma->ifma_refcount++;
3039 if (retifma != NULL)
3040 *retifma = ifma;
3041 IF_ADDR_UNLOCK(ifp);
3042 return (0);
3043 }
3044
3045 /*
3046 * The address isn't already present; resolve the protocol address
3047 * into a link layer address, and then look that up, bump its
3048 * refcount or allocate an ifma for that also. If 'llsa' was
3049 * returned, we will need to free it later.
3050 */
3051 llsa = NULL;
3052 ll_ifma = NULL;
3053 if (ifp->if_resolvemulti != NULL) {
3054 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3055 if (error)
3056 goto unlock_out;
3057 }
3058
3059 /*
3060 * Allocate the new address. Don't hook it up yet, as we may also
3061 * need to allocate a link layer multicast address.
3062 */
3063 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3064 if (ifma == NULL) {
3065 error = ENOMEM;
3066 goto free_llsa_out;
3067 }
3068
3069 /*
3070 * If a link layer address is found, we'll need to see if it's
3071 * already present in the address list, or allocate is as well.
3072 * When this block finishes, the link layer address will be on the
3073 * list.
3074 */
3075 if (llsa != NULL) {
3076 ll_ifma = if_findmulti(ifp, llsa);
3077 if (ll_ifma == NULL) {
3078 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3079 if (ll_ifma == NULL) {
3080 --ifma->ifma_refcount;
3081 if_freemulti(ifma);
3082 error = ENOMEM;
3083 goto free_llsa_out;
3084 }
3085 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3086 ifma_link);
3087 } else
3088 ll_ifma->ifma_refcount++;
3089 ifma->ifma_llifma = ll_ifma;
3090 }
3091
3092 /*
3093 * We now have a new multicast address, ifma, and possibly a new or
3094 * referenced link layer address. Add the primary address to the
3095 * ifnet address list.
3096 */
3097 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3098
3099 if (retifma != NULL)
3100 *retifma = ifma;
3101
3102 /*
3103 * Must generate the message while holding the lock so that 'ifma'
3104 * pointer is still valid.
3105 */
3106 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3107 IF_ADDR_UNLOCK(ifp);
3108
3109 /*
3110 * We are certain we have added something, so call down to the
3111 * interface to let them know about it.
3112 */
3113 if (ifp->if_ioctl != NULL) {
3114 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3115 }
3116
3117 if (llsa != NULL)
3118 free(llsa, M_IFMADDR);
3119
3120 return (0);
3121
3122 free_llsa_out:
3123 if (llsa != NULL)
3124 free(llsa, M_IFMADDR);
3125
3126 unlock_out:
3127 IF_ADDR_UNLOCK(ifp);
3128 return (error);
3129 }
3130
3131 /*
3132 * Delete a multicast group membership by network-layer group address.
3133 *
3134 * Returns ENOENT if the entry could not be found. If ifp no longer
3135 * exists, results are undefined. This entry point should only be used
3136 * from subsystems which do appropriate locking to hold ifp for the
3137 * duration of the call.
3138 * Network-layer protocol domains must use if_delmulti_ifma().
3139 */
3140 int
3141 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3142 {
3143 struct ifmultiaddr *ifma;
3144 int lastref;
3145 #ifdef INVARIANTS
3146 struct ifnet *oifp;
3147
3148 IFNET_RLOCK_NOSLEEP();
3149 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3150 if (ifp == oifp)
3151 break;
3152 if (ifp != oifp)
3153 ifp = NULL;
3154 IFNET_RUNLOCK_NOSLEEP();
3155
3156 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3157 #endif
3158 if (ifp == NULL)
3159 return (ENOENT);
3160
3161 IF_ADDR_LOCK(ifp);
3162 lastref = 0;
3163 ifma = if_findmulti(ifp, sa);
3164 if (ifma != NULL)
3165 lastref = if_delmulti_locked(ifp, ifma, 0);
3166 IF_ADDR_UNLOCK(ifp);
3167
3168 if (ifma == NULL)
3169 return (ENOENT);
3170
3171 if (lastref && ifp->if_ioctl != NULL) {
3172 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3173 }
3174
3175 return (0);
3176 }
3177
3178 /*
3179 * Delete all multicast group membership for an interface.
3180 * Should be used to quickly flush all multicast filters.
3181 */
3182 void
3183 if_delallmulti(struct ifnet *ifp)
3184 {
3185 struct ifmultiaddr *ifma;
3186 struct ifmultiaddr *next;
3187
3188 IF_ADDR_LOCK(ifp);
3189 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3190 if_delmulti_locked(ifp, ifma, 0);
3191 IF_ADDR_UNLOCK(ifp);
3192 }
3193
3194 /*
3195 * Delete a multicast group membership by group membership pointer.
3196 * Network-layer protocol domains must use this routine.
3197 *
3198 * It is safe to call this routine if the ifp disappeared.
3199 */
3200 void
3201 if_delmulti_ifma(struct ifmultiaddr *ifma)
3202 {
3203 struct ifnet *ifp;
3204 int lastref;
3205
3206 ifp = ifma->ifma_ifp;
3207 #ifdef DIAGNOSTIC
3208 if (ifp == NULL) {
3209 printf("%s: ifma_ifp seems to be detached\n", __func__);
3210 } else {
3211 struct ifnet *oifp;
3212
3213 IFNET_RLOCK_NOSLEEP();
3214 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3215 if (ifp == oifp)
3216 break;
3217 if (ifp != oifp) {
3218 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3219 ifp = NULL;
3220 }
3221 IFNET_RUNLOCK_NOSLEEP();
3222 }
3223 #endif
3224 /*
3225 * If and only if the ifnet instance exists: Acquire the address lock.
3226 */
3227 if (ifp != NULL)
3228 IF_ADDR_LOCK(ifp);
3229
3230 lastref = if_delmulti_locked(ifp, ifma, 0);
3231
3232 if (ifp != NULL) {
3233 /*
3234 * If and only if the ifnet instance exists:
3235 * Release the address lock.
3236 * If the group was left: update the hardware hash filter.
3237 */
3238 IF_ADDR_UNLOCK(ifp);
3239 if (lastref && ifp->if_ioctl != NULL) {
3240 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3241 }
3242 }
3243 }
3244
3245 /*
3246 * Perform deletion of network-layer and/or link-layer multicast address.
3247 *
3248 * Return 0 if the reference count was decremented.
3249 * Return 1 if the final reference was released, indicating that the
3250 * hardware hash filter should be reprogrammed.
3251 */
3252 static int
3253 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3254 {
3255 struct ifmultiaddr *ll_ifma;
3256
3257 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3258 KASSERT(ifma->ifma_ifp == ifp,
3259 ("%s: inconsistent ifp %p", __func__, ifp));
3260 IF_ADDR_LOCK_ASSERT(ifp);
3261 }
3262
3263 ifp = ifma->ifma_ifp;
3264
3265 /*
3266 * If the ifnet is detaching, null out references to ifnet,
3267 * so that upper protocol layers will notice, and not attempt
3268 * to obtain locks for an ifnet which no longer exists. The
3269 * routing socket announcement must happen before the ifnet
3270 * instance is detached from the system.
3271 */
3272 if (detaching) {
3273 #ifdef DIAGNOSTIC
3274 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3275 #endif
3276 /*
3277 * ifp may already be nulled out if we are being reentered
3278 * to delete the ll_ifma.
3279 */
3280 if (ifp != NULL) {
3281 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3282 ifma->ifma_ifp = NULL;
3283 }
3284 }
3285
3286 if (--ifma->ifma_refcount > 0)
3287 return 0;
3288
3289 /*
3290 * If this ifma is a network-layer ifma, a link-layer ifma may
3291 * have been associated with it. Release it first if so.
3292 */
3293 ll_ifma = ifma->ifma_llifma;
3294 if (ll_ifma != NULL) {
3295 KASSERT(ifma->ifma_lladdr != NULL,
3296 ("%s: llifma w/o lladdr", __func__));
3297 if (detaching)
3298 ll_ifma->ifma_ifp = NULL; /* XXX */
3299 if (--ll_ifma->ifma_refcount == 0) {
3300 if (ifp != NULL) {
3301 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3302 ifma_link);
3303 }
3304 if_freemulti(ll_ifma);
3305 }
3306 }
3307
3308 if (ifp != NULL)
3309 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3310
3311 if_freemulti(ifma);
3312
3313 /*
3314 * The last reference to this instance of struct ifmultiaddr
3315 * was released; the hardware should be notified of this change.
3316 */
3317 return 1;
3318 }
3319
3320 /*
3321 * Set the link layer address on an interface.
3322 *
3323 * At this time we only support certain types of interfaces,
3324 * and we don't allow the length of the address to change.
3325 */
3326 int
3327 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3328 {
3329 struct sockaddr_dl *sdl;
3330 struct ifaddr *ifa;
3331 struct ifreq ifr;
3332
3333 IF_ADDR_LOCK(ifp);
3334 ifa = ifp->if_addr;
3335 if (ifa == NULL) {
3336 IF_ADDR_UNLOCK(ifp);
3337 return (EINVAL);
3338 }
3339 ifa_ref(ifa);
3340 IF_ADDR_UNLOCK(ifp);
3341 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3342 if (sdl == NULL) {
3343 ifa_free(ifa);
3344 return (EINVAL);
3345 }
3346 if (len != sdl->sdl_alen) { /* don't allow length to change */
3347 ifa_free(ifa);
3348 return (EINVAL);
3349 }
3350 switch (ifp->if_type) {
3351 case IFT_ETHER:
3352 case IFT_FDDI:
3353 case IFT_XETHER:
3354 case IFT_ISO88025:
3355 case IFT_L2VLAN:
3356 case IFT_BRIDGE:
3357 case IFT_ARCNET:
3358 case IFT_IEEE8023ADLAG:
3359 case IFT_IEEE80211:
3360 bcopy(lladdr, LLADDR(sdl), len);
3361 ifa_free(ifa);
3362 break;
3363 default:
3364 ifa_free(ifa);
3365 return (ENODEV);
3366 }
3367
3368 /*
3369 * If the interface is already up, we need
3370 * to re-init it in order to reprogram its
3371 * address filter.
3372 */
3373 if ((ifp->if_flags & IFF_UP) != 0) {
3374 if (ifp->if_ioctl) {
3375 ifp->if_flags &= ~IFF_UP;
3376 ifr.ifr_flags = ifp->if_flags & 0xffff;
3377 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3378 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3379 ifp->if_flags |= IFF_UP;
3380 ifr.ifr_flags = ifp->if_flags & 0xffff;
3381 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3382 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3383 }
3384 #ifdef INET
3385 /*
3386 * Also send gratuitous ARPs to notify other nodes about
3387 * the address change.
3388 */
3389 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3390 if (ifa->ifa_addr->sa_family == AF_INET)
3391 arp_ifinit(ifp, ifa);
3392 }
3393 #endif
3394 }
3395 return (0);
3396 }
3397
3398 /*
3399 * The name argument must be a pointer to storage which will last as
3400 * long as the interface does. For physical devices, the result of
3401 * device_get_name(dev) is a good choice and for pseudo-devices a
3402 * static string works well.
3403 */
3404 void
3405 if_initname(struct ifnet *ifp, const char *name, int unit)
3406 {
3407 ifp->if_dname = name;
3408 ifp->if_dunit = unit;
3409 if (unit != IF_DUNIT_NONE)
3410 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3411 else
3412 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3413 }
3414
3415 int
3416 if_printf(struct ifnet *ifp, const char * fmt, ...)
3417 {
3418 va_list ap;
3419 int retval;
3420
3421 retval = printf("%s: ", ifp->if_xname);
3422 va_start(ap, fmt);
3423 retval += vprintf(fmt, ap);
3424 va_end(ap);
3425 return (retval);
3426 }
3427
3428 void
3429 if_start(struct ifnet *ifp)
3430 {
3431
3432 (*(ifp)->if_start)(ifp);
3433 }
3434
3435 /*
3436 * Backwards compatibility interface for drivers
3437 * that have not implemented it
3438 */
3439 static int
3440 if_transmit(struct ifnet *ifp, struct mbuf *m)
3441 {
3442 int error;
3443
3444 IFQ_HANDOFF(ifp, m, error);
3445 return (error);
3446 }
3447
3448 int
3449 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3450 {
3451 int active = 0;
3452
3453 IF_LOCK(ifq);
3454 if (_IF_QFULL(ifq)) {
3455 _IF_DROP(ifq);
3456 IF_UNLOCK(ifq);
3457 m_freem(m);
3458 return (0);
3459 }
3460 if (ifp != NULL) {
3461 ifp->if_obytes += m->m_pkthdr.len + adjust;
3462 if (m->m_flags & (M_BCAST|M_MCAST))
3463 ifp->if_omcasts++;
3464 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3465 }
3466 _IF_ENQUEUE(ifq, m);
3467 IF_UNLOCK(ifq);
3468 if (ifp != NULL && !active)
3469 (*(ifp)->if_start)(ifp);
3470 return (1);
3471 }
3472
3473 void
3474 if_register_com_alloc(u_char type,
3475 if_com_alloc_t *a, if_com_free_t *f)
3476 {
3477
3478 KASSERT(if_com_alloc[type] == NULL,
3479 ("if_register_com_alloc: %d already registered", type));
3480 KASSERT(if_com_free[type] == NULL,
3481 ("if_register_com_alloc: %d free already registered", type));
3482
3483 if_com_alloc[type] = a;
3484 if_com_free[type] = f;
3485 }
3486
3487 void
3488 if_deregister_com_alloc(u_char type)
3489 {
3490
3491 KASSERT(if_com_alloc[type] != NULL,
3492 ("if_deregister_com_alloc: %d not registered", type));
3493 KASSERT(if_com_free[type] != NULL,
3494 ("if_deregister_com_alloc: %d free not registered", type));
3495 if_com_alloc[type] = NULL;
3496 if_com_free[type] = NULL;
3497 }
Cache object: 7efb92e601373ee9d757d8b0480cea87
|