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/8.0/sys/net/if.c 197365 2009-09-20 17:46:56Z qingli $
31 */
32
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_carp.h"
37
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/refcount.h>
54 #include <sys/module.h>
55 #include <sys/rwlock.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/taskqueue.h>
60 #include <sys/domain.h>
61 #include <sys/jail.h>
62 #include <machine/stdarg.h>
63 #include <vm/uma.h>
64
65 #include <net/if.h>
66 #include <net/if_arp.h>
67 #include <net/if_clone.h>
68 #include <net/if_dl.h>
69 #include <net/if_types.h>
70 #include <net/if_var.h>
71 #include <net/radix.h>
72 #include <net/route.h>
73 #include <net/vnet.h>
74
75 #if defined(INET) || defined(INET6)
76 /*XXX*/
77 #include <netinet/in.h>
78 #include <netinet/in_var.h>
79 #ifdef INET6
80 #include <netinet6/in6_var.h>
81 #include <netinet6/in6_ifattach.h>
82 #endif
83 #endif
84 #ifdef INET
85 #include <netinet/if_ether.h>
86 #endif
87 #if defined(INET) || defined(INET6)
88 #ifdef DEV_CARP
89 #include <netinet/ip_carp.h>
90 #endif
91 #endif
92
93 #include <security/mac/mac_framework.h>
94
95 struct ifindex_entry {
96 struct ifnet *ife_ifnet;
97 };
98
99 static int slowtimo_started;
100
101 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
102 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
103
104 /* Log link state change events */
105 static int log_link_state_change = 1;
106
107 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
108 &log_link_state_change, 0,
109 "log interface link state change events");
110
111 void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
112 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
113 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
114
115 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
116
117 /*
118 * XXX: Style; these should be sorted alphabetically, and unprototyped
119 * static functions should be prototyped. Currently they are sorted by
120 * declaration order.
121 */
122 static void if_attachdomain(void *);
123 static void if_attachdomain1(struct ifnet *);
124 static int ifconf(u_long, caddr_t);
125 static void if_freemulti(struct ifmultiaddr *);
126 static void if_init(void *);
127 static void if_grow(void);
128 static void if_check(void *);
129 static void if_route(struct ifnet *, int flag, int fam);
130 static int if_setflag(struct ifnet *, int, int, int *, int);
131 static void if_slowtimo(void *);
132 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
133 static void if_unroute(struct ifnet *, int flag, int fam);
134 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
135 static int if_rtdel(struct radix_node *, void *);
136 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
137 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
138 static void do_link_state_change(void *, int);
139 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
140 static int if_getgroupmembers(struct ifgroupreq *);
141 static void if_delgroups(struct ifnet *);
142 static void if_attach_internal(struct ifnet *, int);
143 static void if_detach_internal(struct ifnet *, int);
144
145 #ifdef INET6
146 /*
147 * XXX: declare here to avoid to include many inet6 related files..
148 * should be more generalized?
149 */
150 extern void nd6_setmtu(struct ifnet *);
151 #endif
152
153 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
154 VNET_DEFINE(struct ifgrouphead, ifg_head);
155 VNET_DEFINE(int, if_index);
156 static VNET_DEFINE(int, if_indexlim) = 8;
157
158 /* Table of ifnet by index. */
159 static VNET_DEFINE(struct ifindex_entry *, ifindex_table);
160
161 #define V_if_indexlim VNET(if_indexlim)
162 #define V_ifindex_table VNET(ifindex_table)
163
164 int ifqmaxlen = IFQ_MAXLEN;
165
166 /*
167 * The global network interface list (V_ifnet) and related state (such as
168 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
169 * an rwlock. Either may be acquired shared to stablize the list, but both
170 * must be acquired writable to modify the list. This model allows us to
171 * both stablize the interface list during interrupt thread processing, but
172 * also to stablize it over long-running ioctls, without introducing priority
173 * inversions and deadlocks.
174 */
175 struct rwlock ifnet_rwlock;
176 struct sx ifnet_sxlock;
177
178 /*
179 * The allocation of network interfaces is a rather non-atomic affair; we
180 * need to select an index before we are ready to expose the interface for
181 * use, so will use this pointer value to indicate reservation.
182 */
183 #define IFNET_HOLD (void *)(uintptr_t)(-1)
184
185 static if_com_alloc_t *if_com_alloc[256];
186 static if_com_free_t *if_com_free[256];
187
188 /*
189 * System initialization
190 */
191 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL);
192
193 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
194 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
195 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
196
197 struct ifnet *
198 ifnet_byindex_locked(u_short idx)
199 {
200
201 if (idx > V_if_index)
202 return (NULL);
203 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
204 return (NULL);
205 return (V_ifindex_table[idx].ife_ifnet);
206 }
207
208 struct ifnet *
209 ifnet_byindex(u_short idx)
210 {
211 struct ifnet *ifp;
212
213 IFNET_RLOCK_NOSLEEP();
214 ifp = ifnet_byindex_locked(idx);
215 IFNET_RUNLOCK_NOSLEEP();
216 return (ifp);
217 }
218
219 struct ifnet *
220 ifnet_byindex_ref(u_short idx)
221 {
222 struct ifnet *ifp;
223
224 IFNET_RLOCK_NOSLEEP();
225 ifp = ifnet_byindex_locked(idx);
226 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
227 IFNET_RUNLOCK_NOSLEEP();
228 return (NULL);
229 }
230 if_ref(ifp);
231 IFNET_RUNLOCK_NOSLEEP();
232 return (ifp);
233 }
234
235 /*
236 * Allocate an ifindex array entry; return 0 on success or an error on
237 * failure.
238 */
239 static int
240 ifindex_alloc_locked(u_short *idxp)
241 {
242 u_short idx;
243
244 IFNET_WLOCK_ASSERT();
245
246 /*
247 * Try to find an empty slot below V_if_index. If we fail, take the
248 * next slot.
249 */
250 for (idx = 1; idx <= V_if_index; idx++) {
251 if (V_ifindex_table[idx].ife_ifnet == NULL)
252 break;
253 }
254
255 /* Catch if_index overflow. */
256 if (idx < 1)
257 return (ENOSPC);
258 if (idx > V_if_index)
259 V_if_index = idx;
260 if (V_if_index >= V_if_indexlim)
261 if_grow();
262 *idxp = idx;
263 return (0);
264 }
265
266 static void
267 ifindex_free_locked(u_short idx)
268 {
269
270 IFNET_WLOCK_ASSERT();
271
272 V_ifindex_table[idx].ife_ifnet = NULL;
273 while (V_if_index > 0 &&
274 V_ifindex_table[V_if_index].ife_ifnet == NULL)
275 V_if_index--;
276 }
277
278 static void
279 ifindex_free(u_short idx)
280 {
281
282 IFNET_WLOCK();
283 ifindex_free_locked(idx);
284 IFNET_WUNLOCK();
285 }
286
287 static void
288 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
289 {
290
291 IFNET_WLOCK_ASSERT();
292
293 V_ifindex_table[idx].ife_ifnet = ifp;
294 }
295
296 static void
297 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
298 {
299
300 IFNET_WLOCK();
301 ifnet_setbyindex_locked(idx, ifp);
302 IFNET_WUNLOCK();
303 }
304
305 struct ifaddr *
306 ifaddr_byindex(u_short idx)
307 {
308 struct ifaddr *ifa;
309
310 IFNET_RLOCK_NOSLEEP();
311 ifa = ifnet_byindex_locked(idx)->if_addr;
312 if (ifa != NULL)
313 ifa_ref(ifa);
314 IFNET_RUNLOCK_NOSLEEP();
315 return (ifa);
316 }
317
318 /*
319 * Network interface utility routines.
320 *
321 * Routines with ifa_ifwith* names take sockaddr *'s as
322 * parameters.
323 */
324
325 static void
326 vnet_if_init(const void *unused __unused)
327 {
328
329 TAILQ_INIT(&V_ifnet);
330 TAILQ_INIT(&V_ifg_head);
331 if_grow(); /* create initial table */
332 vnet_if_clone_init();
333 }
334 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
335 NULL);
336
337 /* ARGSUSED*/
338 static void
339 if_init(void *dummy __unused)
340 {
341
342 IFNET_LOCK_INIT();
343 if_clone_init();
344 }
345 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
346
347
348 #ifdef VIMAGE
349 static void
350 vnet_if_uninit(const void *unused __unused)
351 {
352
353 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
354 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
355
356 free((caddr_t)V_ifindex_table, M_IFNET);
357 }
358 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
359 vnet_if_uninit, NULL);
360 #endif
361
362 static void
363 if_grow(void)
364 {
365 u_int n;
366 struct ifindex_entry *e;
367
368 V_if_indexlim <<= 1;
369 n = V_if_indexlim * sizeof(*e);
370 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
371 if (V_ifindex_table != NULL) {
372 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
373 free((caddr_t)V_ifindex_table, M_IFNET);
374 }
375 V_ifindex_table = e;
376 }
377
378 static void
379 if_check(void *dummy __unused)
380 {
381
382 /*
383 * If at least one interface added during boot uses
384 * if_watchdog then start the timer.
385 */
386 if (slowtimo_started)
387 if_slowtimo(0);
388 }
389
390 /*
391 * Allocate a struct ifnet and an index for an interface. A layer 2
392 * common structure will also be allocated if an allocation routine is
393 * registered for the passed type.
394 */
395 struct ifnet *
396 if_alloc(u_char type)
397 {
398 struct ifnet *ifp;
399 u_short idx;
400
401 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
402 IFNET_WLOCK();
403 if (ifindex_alloc_locked(&idx) != 0) {
404 IFNET_WUNLOCK();
405 free(ifp, M_IFNET);
406 return (NULL);
407 }
408 ifnet_setbyindex_locked(idx, IFNET_HOLD);
409 IFNET_WUNLOCK();
410 ifp->if_index = idx;
411 ifp->if_type = type;
412 ifp->if_alloctype = type;
413 if (if_com_alloc[type] != NULL) {
414 ifp->if_l2com = if_com_alloc[type](type, ifp);
415 if (ifp->if_l2com == NULL) {
416 free(ifp, M_IFNET);
417 ifindex_free(idx);
418 return (NULL);
419 }
420 }
421
422 IF_ADDR_LOCK_INIT(ifp);
423 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
424 ifp->if_afdata_initialized = 0;
425 IF_AFDATA_LOCK_INIT(ifp);
426 TAILQ_INIT(&ifp->if_addrhead);
427 TAILQ_INIT(&ifp->if_prefixhead);
428 TAILQ_INIT(&ifp->if_multiaddrs);
429 TAILQ_INIT(&ifp->if_groups);
430 #ifdef MAC
431 mac_ifnet_init(ifp);
432 #endif
433 ifq_init(&ifp->if_snd, ifp);
434
435 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
436 ifnet_setbyindex(ifp->if_index, ifp);
437 return (ifp);
438 }
439
440 /*
441 * Do the actual work of freeing a struct ifnet, associated index, and layer
442 * 2 common structure. This call is made when the last reference to an
443 * interface is released.
444 */
445 static void
446 if_free_internal(struct ifnet *ifp)
447 {
448
449 KASSERT((ifp->if_flags & IFF_DYING),
450 ("if_free_internal: interface not dying"));
451
452 IFNET_WLOCK();
453 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
454 ("%s: freeing unallocated ifnet", ifp->if_xname));
455
456 ifindex_free_locked(ifp->if_index);
457 IFNET_WUNLOCK();
458
459 if (if_com_free[ifp->if_alloctype] != NULL)
460 if_com_free[ifp->if_alloctype](ifp->if_l2com,
461 ifp->if_alloctype);
462
463 #ifdef MAC
464 mac_ifnet_destroy(ifp);
465 #endif /* MAC */
466 IF_AFDATA_DESTROY(ifp);
467 IF_ADDR_LOCK_DESTROY(ifp);
468 ifq_delete(&ifp->if_snd);
469 free(ifp, M_IFNET);
470 }
471
472 /*
473 * This version should only be called by intefaces that switch their type
474 * after calling if_alloc(). if_free_type() will go away again now that we
475 * have if_alloctype to cache the original allocation type. For now, assert
476 * that they match, since we require that in practice.
477 */
478 void
479 if_free_type(struct ifnet *ifp, u_char type)
480 {
481
482 KASSERT(ifp->if_alloctype == type,
483 ("if_free_type: type (%d) != alloctype (%d)", type,
484 ifp->if_alloctype));
485
486 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
487 if (!refcount_release(&ifp->if_refcount))
488 return;
489 if_free_internal(ifp);
490 }
491
492 /*
493 * This is the normal version of if_free(), used by device drivers to free a
494 * detached network interface. The contents of if_free_type() will move into
495 * here when if_free_type() goes away.
496 */
497 void
498 if_free(struct ifnet *ifp)
499 {
500
501 if_free_type(ifp, ifp->if_alloctype);
502 }
503
504 /*
505 * Interfaces to keep an ifnet type-stable despite the possibility of the
506 * driver calling if_free(). If there are additional references, we defer
507 * freeing the underlying data structure.
508 */
509 void
510 if_ref(struct ifnet *ifp)
511 {
512
513 /* We don't assert the ifnet list lock here, but arguably should. */
514 refcount_acquire(&ifp->if_refcount);
515 }
516
517 void
518 if_rele(struct ifnet *ifp)
519 {
520
521 if (!refcount_release(&ifp->if_refcount))
522 return;
523 if_free_internal(ifp);
524 }
525
526 void
527 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
528 {
529
530 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
531
532 if (ifq->ifq_maxlen == 0)
533 ifq->ifq_maxlen = ifqmaxlen;
534
535 ifq->altq_type = 0;
536 ifq->altq_disc = NULL;
537 ifq->altq_flags &= ALTQF_CANTCHANGE;
538 ifq->altq_tbr = NULL;
539 ifq->altq_ifp = ifp;
540 }
541
542 void
543 ifq_delete(struct ifaltq *ifq)
544 {
545 mtx_destroy(&ifq->ifq_mtx);
546 }
547
548 /*
549 * Perform generic interface initalization tasks and attach the interface
550 * to the list of "active" interfaces. If vmove flag is set on entry
551 * to if_attach_internal(), perform only a limited subset of initialization
552 * tasks, given that we are moving from one vnet to another an ifnet which
553 * has already been fully initialized.
554 *
555 * XXX:
556 * - The decision to return void and thus require this function to
557 * succeed is questionable.
558 * - We should probably do more sanity checking. For instance we don't
559 * do anything to insure if_xname is unique or non-empty.
560 */
561 void
562 if_attach(struct ifnet *ifp)
563 {
564
565 if_attach_internal(ifp, 0);
566 }
567
568 static void
569 if_attach_internal(struct ifnet *ifp, int vmove)
570 {
571 unsigned socksize, ifasize;
572 int namelen, masklen;
573 struct sockaddr_dl *sdl;
574 struct ifaddr *ifa;
575
576 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
577 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
578 ifp->if_xname);
579
580 #ifdef VIMAGE
581 ifp->if_vnet = curvnet;
582 if (ifp->if_home_vnet == NULL)
583 ifp->if_home_vnet = curvnet;
584 #endif
585
586 if_addgroup(ifp, IFG_ALL);
587
588 getmicrotime(&ifp->if_lastchange);
589 ifp->if_data.ifi_epoch = time_uptime;
590 ifp->if_data.ifi_datalen = sizeof(struct if_data);
591
592 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
593 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
594 ("transmit and qflush must both either be set or both be NULL"));
595 if (ifp->if_transmit == NULL) {
596 ifp->if_transmit = if_transmit;
597 ifp->if_qflush = if_qflush;
598 }
599
600 if (!vmove) {
601 #ifdef MAC
602 mac_ifnet_create(ifp);
603 #endif
604
605 /*
606 * Create a Link Level name for this device.
607 */
608 namelen = strlen(ifp->if_xname);
609 /*
610 * Always save enough space for any possiable name so we
611 * can do a rename in place later.
612 */
613 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
614 socksize = masklen + ifp->if_addrlen;
615 if (socksize < sizeof(*sdl))
616 socksize = sizeof(*sdl);
617 socksize = roundup2(socksize, sizeof(long));
618 ifasize = sizeof(*ifa) + 2 * socksize;
619 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
620 ifa_init(ifa);
621 sdl = (struct sockaddr_dl *)(ifa + 1);
622 sdl->sdl_len = socksize;
623 sdl->sdl_family = AF_LINK;
624 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
625 sdl->sdl_nlen = namelen;
626 sdl->sdl_index = ifp->if_index;
627 sdl->sdl_type = ifp->if_type;
628 ifp->if_addr = ifa;
629 ifa->ifa_ifp = ifp;
630 ifa->ifa_rtrequest = link_rtrequest;
631 ifa->ifa_addr = (struct sockaddr *)sdl;
632 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
633 ifa->ifa_netmask = (struct sockaddr *)sdl;
634 sdl->sdl_len = masklen;
635 while (namelen != 0)
636 sdl->sdl_data[--namelen] = 0xff;
637 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
638 /* Reliably crash if used uninitialized. */
639 ifp->if_broadcastaddr = NULL;
640 }
641 #ifdef VIMAGE
642 else {
643 /*
644 * Update the interface index in the link layer address
645 * of the interface.
646 */
647 for (ifa = ifp->if_addr; ifa != NULL;
648 ifa = TAILQ_NEXT(ifa, ifa_link)) {
649 if (ifa->ifa_addr->sa_family == AF_LINK) {
650 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
651 sdl->sdl_index = ifp->if_index;
652 }
653 }
654 }
655 #endif
656
657 IFNET_WLOCK();
658 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
659 #ifdef VIMAGE
660 curvnet->vnet_ifcnt++;
661 #endif
662 IFNET_WUNLOCK();
663
664 if (domain_init_status >= 2)
665 if_attachdomain1(ifp);
666
667 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
668 if (IS_DEFAULT_VNET(curvnet))
669 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
670
671 /* Announce the interface. */
672 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
673
674 if (!vmove && ifp->if_watchdog != NULL) {
675 if_printf(ifp,
676 "WARNING: using obsoleted if_watchdog interface\n");
677
678 /*
679 * Note that we need if_slowtimo(). If this happens after
680 * boot, then call if_slowtimo() directly.
681 */
682 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold)
683 if_slowtimo(0);
684 }
685 }
686
687 static void
688 if_attachdomain(void *dummy)
689 {
690 struct ifnet *ifp;
691 int s;
692
693 s = splnet();
694 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
695 if_attachdomain1(ifp);
696 splx(s);
697 }
698 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
699 if_attachdomain, NULL);
700
701 static void
702 if_attachdomain1(struct ifnet *ifp)
703 {
704 struct domain *dp;
705 int s;
706
707 s = splnet();
708
709 /*
710 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
711 * cannot lock ifp->if_afdata initialization, entirely.
712 */
713 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
714 splx(s);
715 return;
716 }
717 if (ifp->if_afdata_initialized >= domain_init_status) {
718 IF_AFDATA_UNLOCK(ifp);
719 splx(s);
720 printf("if_attachdomain called more than once on %s\n",
721 ifp->if_xname);
722 return;
723 }
724 ifp->if_afdata_initialized = domain_init_status;
725 IF_AFDATA_UNLOCK(ifp);
726
727 /* address family dependent data region */
728 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
729 for (dp = domains; dp; dp = dp->dom_next) {
730 if (dp->dom_ifattach)
731 ifp->if_afdata[dp->dom_family] =
732 (*dp->dom_ifattach)(ifp);
733 }
734
735 splx(s);
736 }
737
738 /*
739 * Remove any unicast or broadcast network addresses from an interface.
740 */
741 void
742 if_purgeaddrs(struct ifnet *ifp)
743 {
744 struct ifaddr *ifa, *next;
745
746 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
747 if (ifa->ifa_addr->sa_family == AF_LINK)
748 continue;
749 #ifdef INET
750 /* XXX: Ugly!! ad hoc just for INET */
751 if (ifa->ifa_addr->sa_family == AF_INET) {
752 struct ifaliasreq ifr;
753
754 bzero(&ifr, sizeof(ifr));
755 ifr.ifra_addr = *ifa->ifa_addr;
756 if (ifa->ifa_dstaddr)
757 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
758 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
759 NULL) == 0)
760 continue;
761 }
762 #endif /* INET */
763 #ifdef INET6
764 if (ifa->ifa_addr->sa_family == AF_INET6) {
765 in6_purgeaddr(ifa);
766 /* ifp_addrhead is already updated */
767 continue;
768 }
769 #endif /* INET6 */
770 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
771 ifa_free(ifa);
772 }
773 }
774
775 /*
776 * Remove any multicast network addresses from an interface.
777 */
778 void
779 if_purgemaddrs(struct ifnet *ifp)
780 {
781 struct ifmultiaddr *ifma;
782 struct ifmultiaddr *next;
783
784 IF_ADDR_LOCK(ifp);
785 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
786 if_delmulti_locked(ifp, ifma, 1);
787 IF_ADDR_UNLOCK(ifp);
788 }
789
790 /*
791 * Detach an interface, removing it from the list of "active" interfaces.
792 * If vmove flag is set on entry to if_detach_internal(), perform only a
793 * limited subset of cleanup tasks, given that we are moving an ifnet from
794 * one vnet to another, where it must be fully operational.
795 *
796 * XXXRW: There are some significant questions about event ordering, and
797 * how to prevent things from starting to use the interface during detach.
798 */
799 void
800 if_detach(struct ifnet *ifp)
801 {
802
803 if_detach_internal(ifp, 0);
804 }
805
806 static void
807 if_detach_internal(struct ifnet *ifp, int vmove)
808 {
809 struct ifaddr *ifa;
810 struct radix_node_head *rnh;
811 int i, j;
812 struct domain *dp;
813 struct ifnet *iter;
814 int found = 0;
815
816 IFNET_WLOCK();
817 TAILQ_FOREACH(iter, &V_ifnet, if_link)
818 if (iter == ifp) {
819 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
820 found = 1;
821 break;
822 }
823 #ifdef VIMAGE
824 if (found)
825 curvnet->vnet_ifcnt--;
826 #endif
827 IFNET_WUNLOCK();
828 if (!found) {
829 if (vmove)
830 panic("interface not in it's own ifnet list");
831 else
832 return; /* XXX this should panic as well? */
833 }
834
835 /*
836 * Remove/wait for pending events.
837 */
838 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
839
840 /*
841 * Remove routes and flush queues.
842 */
843 if_down(ifp);
844 #ifdef ALTQ
845 if (ALTQ_IS_ENABLED(&ifp->if_snd))
846 altq_disable(&ifp->if_snd);
847 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
848 altq_detach(&ifp->if_snd);
849 #endif
850
851 if_purgeaddrs(ifp);
852
853 #ifdef INET
854 in_ifdetach(ifp);
855 #endif
856
857 #ifdef INET6
858 /*
859 * Remove all IPv6 kernel structs related to ifp. This should be done
860 * before removing routing entries below, since IPv6 interface direct
861 * routes are expected to be removed by the IPv6-specific kernel API.
862 * Otherwise, the kernel will detect some inconsistency and bark it.
863 */
864 in6_ifdetach(ifp);
865 #endif
866 if_purgemaddrs(ifp);
867
868 if (!vmove) {
869 /*
870 * Prevent further calls into the device driver via ifnet.
871 */
872 if_dead(ifp);
873
874 /*
875 * Remove link ifaddr pointer and maybe decrement if_index.
876 * Clean up all addresses.
877 */
878 ifp->if_addr = NULL;
879
880 /* We can now free link ifaddr. */
881 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
882 ifa = TAILQ_FIRST(&ifp->if_addrhead);
883 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
884 ifa_free(ifa);
885 }
886 }
887
888 /*
889 * Delete all remaining routes using this interface
890 * Unfortuneatly the only way to do this is to slog through
891 * the entire routing table looking for routes which point
892 * to this interface...oh well...
893 */
894 for (i = 1; i <= AF_MAX; i++) {
895 for (j = 0; j < rt_numfibs; j++) {
896 rnh = rt_tables_get_rnh(j, i);
897 if (rnh == NULL)
898 continue;
899 RADIX_NODE_HEAD_LOCK(rnh);
900 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
901 RADIX_NODE_HEAD_UNLOCK(rnh);
902 }
903 }
904
905 /* Announce that the interface is gone. */
906 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
907 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
908 if (IS_DEFAULT_VNET(curvnet))
909 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
910 if_delgroups(ifp);
911
912 IF_AFDATA_LOCK(ifp);
913 for (dp = domains; dp; dp = dp->dom_next) {
914 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
915 (*dp->dom_ifdetach)(ifp,
916 ifp->if_afdata[dp->dom_family]);
917 }
918 ifp->if_afdata_initialized = 0;
919 IF_AFDATA_UNLOCK(ifp);
920 }
921
922 #ifdef VIMAGE
923 /*
924 * if_vmove() performs a limited version of if_detach() in current
925 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
926 * An attempt is made to shrink if_index in current vnet, find an
927 * unused if_index in target vnet and calls if_grow() if necessary,
928 * and finally find an unused if_xname for the target vnet.
929 */
930 void
931 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
932 {
933 u_short idx;
934
935 /*
936 * Detach from current vnet, but preserve LLADDR info, do not
937 * mark as dead etc. so that the ifnet can be reattached later.
938 */
939 if_detach_internal(ifp, 1);
940
941 /*
942 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
943 * the if_index for that vnet if possible.
944 *
945 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
946 * or we'd lock on one vnet and unlock on another.
947 */
948 IFNET_WLOCK();
949 ifindex_free_locked(ifp->if_index);
950
951 /*
952 * Switch to the context of the target vnet.
953 */
954 CURVNET_SET_QUIET(new_vnet);
955
956 if (ifindex_alloc_locked(&idx) != 0) {
957 IFNET_WUNLOCK();
958 panic("if_index overflow");
959 }
960 ifp->if_index = idx;
961 ifnet_setbyindex_locked(ifp->if_index, ifp);
962 IFNET_WUNLOCK();
963
964 if_attach_internal(ifp, 1);
965
966 CURVNET_RESTORE();
967 }
968
969 /*
970 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
971 */
972 static int
973 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
974 {
975 struct prison *pr;
976 struct ifnet *difp;
977
978 /* Try to find the prison within our visibility. */
979 sx_slock(&allprison_lock);
980 pr = prison_find_child(td->td_ucred->cr_prison, jid);
981 sx_sunlock(&allprison_lock);
982 if (pr == NULL)
983 return (ENXIO);
984 prison_hold_locked(pr);
985 mtx_unlock(&pr->pr_mtx);
986
987 /* Do not try to move the iface from and to the same prison. */
988 if (pr->pr_vnet == ifp->if_vnet) {
989 prison_free(pr);
990 return (EEXIST);
991 }
992
993 /* Make sure the named iface does not exists in the dst. prison/vnet. */
994 /* XXX Lock interfaces to avoid races. */
995 CURVNET_SET_QUIET(pr->pr_vnet);
996 difp = ifunit(ifname);
997 CURVNET_RESTORE();
998 if (difp != NULL) {
999 prison_free(pr);
1000 return (EEXIST);
1001 }
1002
1003 /* Move the interface into the child jail/vnet. */
1004 if_vmove(ifp, pr->pr_vnet);
1005
1006 /* Report the new if_xname back to the userland. */
1007 sprintf(ifname, "%s", ifp->if_xname);
1008
1009 prison_free(pr);
1010 return (0);
1011 }
1012
1013 static int
1014 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1015 {
1016 struct prison *pr;
1017 struct vnet *vnet_dst;
1018 struct ifnet *ifp;
1019
1020 /* Try to find the prison within our visibility. */
1021 sx_slock(&allprison_lock);
1022 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1023 sx_sunlock(&allprison_lock);
1024 if (pr == NULL)
1025 return (ENXIO);
1026 prison_hold_locked(pr);
1027 mtx_unlock(&pr->pr_mtx);
1028
1029 /* Make sure the named iface exists in the source prison/vnet. */
1030 CURVNET_SET(pr->pr_vnet);
1031 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1032 if (ifp == NULL) {
1033 CURVNET_RESTORE();
1034 prison_free(pr);
1035 return (ENXIO);
1036 }
1037
1038 /* Do not try to move the iface from and to the same prison. */
1039 vnet_dst = TD_TO_VNET(td);
1040 if (vnet_dst == ifp->if_vnet) {
1041 CURVNET_RESTORE();
1042 prison_free(pr);
1043 return (EEXIST);
1044 }
1045
1046 /* Get interface back from child jail/vnet. */
1047 if_vmove(ifp, vnet_dst);
1048 CURVNET_RESTORE();
1049
1050 /* Report the new if_xname back to the userland. */
1051 sprintf(ifname, "%s", ifp->if_xname);
1052
1053 prison_free(pr);
1054 return (0);
1055 }
1056 #endif /* VIMAGE */
1057
1058 /*
1059 * Add a group to an interface
1060 */
1061 int
1062 if_addgroup(struct ifnet *ifp, const char *groupname)
1063 {
1064 struct ifg_list *ifgl;
1065 struct ifg_group *ifg = NULL;
1066 struct ifg_member *ifgm;
1067
1068 if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
1069 groupname[strlen(groupname) - 1] <= '9')
1070 return (EINVAL);
1071
1072 IFNET_WLOCK();
1073 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1074 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1075 IFNET_WUNLOCK();
1076 return (EEXIST);
1077 }
1078
1079 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1080 M_NOWAIT)) == NULL) {
1081 IFNET_WUNLOCK();
1082 return (ENOMEM);
1083 }
1084
1085 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1086 M_TEMP, M_NOWAIT)) == NULL) {
1087 free(ifgl, M_TEMP);
1088 IFNET_WUNLOCK();
1089 return (ENOMEM);
1090 }
1091
1092 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1093 if (!strcmp(ifg->ifg_group, groupname))
1094 break;
1095
1096 if (ifg == NULL) {
1097 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1098 M_TEMP, M_NOWAIT)) == NULL) {
1099 free(ifgl, M_TEMP);
1100 free(ifgm, M_TEMP);
1101 IFNET_WUNLOCK();
1102 return (ENOMEM);
1103 }
1104 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1105 ifg->ifg_refcnt = 0;
1106 TAILQ_INIT(&ifg->ifg_members);
1107 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1108 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1109 }
1110
1111 ifg->ifg_refcnt++;
1112 ifgl->ifgl_group = ifg;
1113 ifgm->ifgm_ifp = ifp;
1114
1115 IF_ADDR_LOCK(ifp);
1116 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1117 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1118 IF_ADDR_UNLOCK(ifp);
1119
1120 IFNET_WUNLOCK();
1121
1122 EVENTHANDLER_INVOKE(group_change_event, groupname);
1123
1124 return (0);
1125 }
1126
1127 /*
1128 * Remove a group from an interface
1129 */
1130 int
1131 if_delgroup(struct ifnet *ifp, const char *groupname)
1132 {
1133 struct ifg_list *ifgl;
1134 struct ifg_member *ifgm;
1135
1136 IFNET_WLOCK();
1137 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1138 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1139 break;
1140 if (ifgl == NULL) {
1141 IFNET_WUNLOCK();
1142 return (ENOENT);
1143 }
1144
1145 IF_ADDR_LOCK(ifp);
1146 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1147 IF_ADDR_UNLOCK(ifp);
1148
1149 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1150 if (ifgm->ifgm_ifp == ifp)
1151 break;
1152
1153 if (ifgm != NULL) {
1154 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1155 free(ifgm, M_TEMP);
1156 }
1157
1158 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1159 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1160 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1161 free(ifgl->ifgl_group, M_TEMP);
1162 }
1163 IFNET_WUNLOCK();
1164
1165 free(ifgl, M_TEMP);
1166
1167 EVENTHANDLER_INVOKE(group_change_event, groupname);
1168
1169 return (0);
1170 }
1171
1172 /*
1173 * Remove an interface from all groups
1174 */
1175 static void
1176 if_delgroups(struct ifnet *ifp)
1177 {
1178 struct ifg_list *ifgl;
1179 struct ifg_member *ifgm;
1180 char groupname[IFNAMSIZ];
1181
1182 IFNET_WLOCK();
1183 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1184 ifgl = TAILQ_FIRST(&ifp->if_groups);
1185
1186 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1187
1188 IF_ADDR_LOCK(ifp);
1189 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1190 IF_ADDR_UNLOCK(ifp);
1191
1192 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1193 if (ifgm->ifgm_ifp == ifp)
1194 break;
1195
1196 if (ifgm != NULL) {
1197 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1198 ifgm_next);
1199 free(ifgm, M_TEMP);
1200 }
1201
1202 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1203 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1204 EVENTHANDLER_INVOKE(group_detach_event,
1205 ifgl->ifgl_group);
1206 free(ifgl->ifgl_group, M_TEMP);
1207 }
1208 IFNET_WUNLOCK();
1209
1210 free(ifgl, M_TEMP);
1211
1212 EVENTHANDLER_INVOKE(group_change_event, groupname);
1213
1214 IFNET_WLOCK();
1215 }
1216 IFNET_WUNLOCK();
1217 }
1218
1219 /*
1220 * Stores all groups from an interface in memory pointed
1221 * to by data
1222 */
1223 static int
1224 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1225 {
1226 int len, error;
1227 struct ifg_list *ifgl;
1228 struct ifg_req ifgrq, *ifgp;
1229 struct ifgroupreq *ifgr = data;
1230
1231 if (ifgr->ifgr_len == 0) {
1232 IF_ADDR_LOCK(ifp);
1233 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1234 ifgr->ifgr_len += sizeof(struct ifg_req);
1235 IF_ADDR_UNLOCK(ifp);
1236 return (0);
1237 }
1238
1239 len = ifgr->ifgr_len;
1240 ifgp = ifgr->ifgr_groups;
1241 /* XXX: wire */
1242 IF_ADDR_LOCK(ifp);
1243 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1244 if (len < sizeof(ifgrq)) {
1245 IF_ADDR_UNLOCK(ifp);
1246 return (EINVAL);
1247 }
1248 bzero(&ifgrq, sizeof ifgrq);
1249 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1250 sizeof(ifgrq.ifgrq_group));
1251 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1252 IF_ADDR_UNLOCK(ifp);
1253 return (error);
1254 }
1255 len -= sizeof(ifgrq);
1256 ifgp++;
1257 }
1258 IF_ADDR_UNLOCK(ifp);
1259
1260 return (0);
1261 }
1262
1263 /*
1264 * Stores all members of a group in memory pointed to by data
1265 */
1266 static int
1267 if_getgroupmembers(struct ifgroupreq *data)
1268 {
1269 struct ifgroupreq *ifgr = data;
1270 struct ifg_group *ifg;
1271 struct ifg_member *ifgm;
1272 struct ifg_req ifgrq, *ifgp;
1273 int len, error;
1274
1275 IFNET_RLOCK();
1276 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1277 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1278 break;
1279 if (ifg == NULL) {
1280 IFNET_RUNLOCK();
1281 return (ENOENT);
1282 }
1283
1284 if (ifgr->ifgr_len == 0) {
1285 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1286 ifgr->ifgr_len += sizeof(ifgrq);
1287 IFNET_RUNLOCK();
1288 return (0);
1289 }
1290
1291 len = ifgr->ifgr_len;
1292 ifgp = ifgr->ifgr_groups;
1293 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1294 if (len < sizeof(ifgrq)) {
1295 IFNET_RUNLOCK();
1296 return (EINVAL);
1297 }
1298 bzero(&ifgrq, sizeof ifgrq);
1299 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1300 sizeof(ifgrq.ifgrq_member));
1301 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1302 IFNET_RUNLOCK();
1303 return (error);
1304 }
1305 len -= sizeof(ifgrq);
1306 ifgp++;
1307 }
1308 IFNET_RUNLOCK();
1309
1310 return (0);
1311 }
1312
1313 /*
1314 * Delete Routes for a Network Interface
1315 *
1316 * Called for each routing entry via the rnh->rnh_walktree() call above
1317 * to delete all route entries referencing a detaching network interface.
1318 *
1319 * Arguments:
1320 * rn pointer to node in the routing table
1321 * arg argument passed to rnh->rnh_walktree() - detaching interface
1322 *
1323 * Returns:
1324 * 0 successful
1325 * errno failed - reason indicated
1326 *
1327 */
1328 static int
1329 if_rtdel(struct radix_node *rn, void *arg)
1330 {
1331 struct rtentry *rt = (struct rtentry *)rn;
1332 struct ifnet *ifp = arg;
1333 int err;
1334
1335 if (rt->rt_ifp == ifp) {
1336
1337 /*
1338 * Protect (sorta) against walktree recursion problems
1339 * with cloned routes
1340 */
1341 if ((rt->rt_flags & RTF_UP) == 0)
1342 return (0);
1343
1344 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1345 rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
1346 (struct rtentry **) NULL, rt->rt_fibnum);
1347 if (err) {
1348 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1349 }
1350 }
1351
1352 return (0);
1353 }
1354
1355 /*
1356 * Wrapper functions for struct ifnet address list locking macros. These are
1357 * used by kernel modules to avoid encoding programming interface or binary
1358 * interface assumptions that may be violated when kernel-internal locking
1359 * approaches change.
1360 */
1361 void
1362 if_addr_rlock(struct ifnet *ifp)
1363 {
1364
1365 IF_ADDR_LOCK(ifp);
1366 }
1367
1368 void
1369 if_addr_runlock(struct ifnet *ifp)
1370 {
1371
1372 IF_ADDR_UNLOCK(ifp);
1373 }
1374
1375 void
1376 if_maddr_rlock(struct ifnet *ifp)
1377 {
1378
1379 IF_ADDR_LOCK(ifp);
1380 }
1381
1382 void
1383 if_maddr_runlock(struct ifnet *ifp)
1384 {
1385
1386 IF_ADDR_UNLOCK(ifp);
1387 }
1388
1389 /*
1390 * Reference count functions for ifaddrs.
1391 */
1392 void
1393 ifa_init(struct ifaddr *ifa)
1394 {
1395
1396 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1397 refcount_init(&ifa->ifa_refcnt, 1);
1398 }
1399
1400 void
1401 ifa_ref(struct ifaddr *ifa)
1402 {
1403
1404 refcount_acquire(&ifa->ifa_refcnt);
1405 }
1406
1407 void
1408 ifa_free(struct ifaddr *ifa)
1409 {
1410
1411 if (refcount_release(&ifa->ifa_refcnt)) {
1412 mtx_destroy(&ifa->ifa_mtx);
1413 free(ifa, M_IFADDR);
1414 }
1415 }
1416
1417 int
1418 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1419 {
1420 int error = 0;
1421 struct rtentry *rt = NULL;
1422 struct rt_addrinfo info;
1423 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1424
1425 bzero(&info, sizeof(info));
1426 info.rti_ifp = V_loif;
1427 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1428 info.rti_info[RTAX_DST] = ia;
1429 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1430 error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1431
1432 if (error == 0 && rt != NULL) {
1433 RT_LOCK(rt);
1434 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1435 ifa->ifa_ifp->if_type;
1436 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1437 ifa->ifa_ifp->if_index;
1438 RT_REMREF(rt);
1439 RT_UNLOCK(rt);
1440 } else if (error != 0)
1441 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1442
1443 return (error);
1444 }
1445
1446 int
1447 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1448 {
1449 int error = 0;
1450 struct rt_addrinfo info;
1451 struct sockaddr_dl null_sdl;
1452
1453 bzero(&null_sdl, sizeof(null_sdl));
1454 null_sdl.sdl_len = sizeof(null_sdl);
1455 null_sdl.sdl_family = AF_LINK;
1456 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1457 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1458 bzero(&info, sizeof(info));
1459 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1460 info.rti_info[RTAX_DST] = ia;
1461 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1462 error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1463
1464 if (error != 0)
1465 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1466
1467 return (error);
1468 }
1469
1470 /*
1471 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1472 * structs used to represent other address families, it is necessary
1473 * to perform a different comparison.
1474 */
1475
1476 #define sa_equal(a1, a2) \
1477 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1478
1479 #define sa_dl_equal(a1, a2) \
1480 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1481 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1482 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1483 LLADDR((struct sockaddr_dl *)(a2)), \
1484 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1485
1486 /*
1487 * Locate an interface based on a complete address.
1488 */
1489 /*ARGSUSED*/
1490 static struct ifaddr *
1491 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1492 {
1493 struct ifnet *ifp;
1494 struct ifaddr *ifa;
1495
1496 IFNET_RLOCK_NOSLEEP();
1497 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1498 IF_ADDR_LOCK(ifp);
1499 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1500 if (ifa->ifa_addr->sa_family != addr->sa_family)
1501 continue;
1502 if (sa_equal(addr, ifa->ifa_addr)) {
1503 if (getref)
1504 ifa_ref(ifa);
1505 IF_ADDR_UNLOCK(ifp);
1506 goto done;
1507 }
1508 /* IP6 doesn't have broadcast */
1509 if ((ifp->if_flags & IFF_BROADCAST) &&
1510 ifa->ifa_broadaddr &&
1511 ifa->ifa_broadaddr->sa_len != 0 &&
1512 sa_equal(ifa->ifa_broadaddr, addr)) {
1513 if (getref)
1514 ifa_ref(ifa);
1515 IF_ADDR_UNLOCK(ifp);
1516 goto done;
1517 }
1518 }
1519 IF_ADDR_UNLOCK(ifp);
1520 }
1521 ifa = NULL;
1522 done:
1523 IFNET_RUNLOCK_NOSLEEP();
1524 return (ifa);
1525 }
1526
1527 struct ifaddr *
1528 ifa_ifwithaddr(struct sockaddr *addr)
1529 {
1530
1531 return (ifa_ifwithaddr_internal(addr, 1));
1532 }
1533
1534 int
1535 ifa_ifwithaddr_check(struct sockaddr *addr)
1536 {
1537
1538 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1539 }
1540
1541 /*
1542 * Locate an interface based on the broadcast address.
1543 */
1544 /* ARGSUSED */
1545 struct ifaddr *
1546 ifa_ifwithbroadaddr(struct sockaddr *addr)
1547 {
1548 struct ifnet *ifp;
1549 struct ifaddr *ifa;
1550
1551 IFNET_RLOCK_NOSLEEP();
1552 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1553 IF_ADDR_LOCK(ifp);
1554 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1555 if (ifa->ifa_addr->sa_family != addr->sa_family)
1556 continue;
1557 if ((ifp->if_flags & IFF_BROADCAST) &&
1558 ifa->ifa_broadaddr &&
1559 ifa->ifa_broadaddr->sa_len != 0 &&
1560 sa_equal(ifa->ifa_broadaddr, addr)) {
1561 ifa_ref(ifa);
1562 IF_ADDR_UNLOCK(ifp);
1563 goto done;
1564 }
1565 }
1566 IF_ADDR_UNLOCK(ifp);
1567 }
1568 ifa = NULL;
1569 done:
1570 IFNET_RUNLOCK_NOSLEEP();
1571 return (ifa);
1572 }
1573
1574 /*
1575 * Locate the point to point interface with a given destination address.
1576 */
1577 /*ARGSUSED*/
1578 struct ifaddr *
1579 ifa_ifwithdstaddr(struct sockaddr *addr)
1580 {
1581 struct ifnet *ifp;
1582 struct ifaddr *ifa;
1583
1584 IFNET_RLOCK_NOSLEEP();
1585 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1586 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1587 continue;
1588 IF_ADDR_LOCK(ifp);
1589 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1590 if (ifa->ifa_addr->sa_family != addr->sa_family)
1591 continue;
1592 if (ifa->ifa_dstaddr != NULL &&
1593 sa_equal(addr, ifa->ifa_dstaddr)) {
1594 ifa_ref(ifa);
1595 IF_ADDR_UNLOCK(ifp);
1596 goto done;
1597 }
1598 }
1599 IF_ADDR_UNLOCK(ifp);
1600 }
1601 ifa = NULL;
1602 done:
1603 IFNET_RUNLOCK_NOSLEEP();
1604 return (ifa);
1605 }
1606
1607 /*
1608 * Find an interface on a specific network. If many, choice
1609 * is most specific found.
1610 */
1611 struct ifaddr *
1612 ifa_ifwithnet(struct sockaddr *addr)
1613 {
1614 struct ifnet *ifp;
1615 struct ifaddr *ifa;
1616 struct ifaddr *ifa_maybe = NULL;
1617 u_int af = addr->sa_family;
1618 char *addr_data = addr->sa_data, *cplim;
1619
1620 /*
1621 * AF_LINK addresses can be looked up directly by their index number,
1622 * so do that if we can.
1623 */
1624 if (af == AF_LINK) {
1625 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1626 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1627 return (ifaddr_byindex(sdl->sdl_index));
1628 }
1629
1630 /*
1631 * Scan though each interface, looking for ones that have addresses
1632 * in this address family. Maintain a reference on ifa_maybe once
1633 * we find one, as we release the IF_ADDR_LOCK() that kept it stable
1634 * when we move onto the next interface.
1635 */
1636 IFNET_RLOCK_NOSLEEP();
1637 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1638 IF_ADDR_LOCK(ifp);
1639 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1640 char *cp, *cp2, *cp3;
1641
1642 if (ifa->ifa_addr->sa_family != af)
1643 next: continue;
1644 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
1645 /*
1646 * This is a bit broken as it doesn't
1647 * take into account that the remote end may
1648 * be a single node in the network we are
1649 * looking for.
1650 * The trouble is that we don't know the
1651 * netmask for the remote end.
1652 */
1653 if (ifa->ifa_dstaddr != NULL &&
1654 sa_equal(addr, ifa->ifa_dstaddr)) {
1655 ifa_ref(ifa);
1656 IF_ADDR_UNLOCK(ifp);
1657 goto done;
1658 }
1659 } else {
1660 /*
1661 * if we have a special address handler,
1662 * then use it instead of the generic one.
1663 */
1664 if (ifa->ifa_claim_addr) {
1665 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1666 ifa_ref(ifa);
1667 IF_ADDR_UNLOCK(ifp);
1668 goto done;
1669 }
1670 continue;
1671 }
1672
1673 /*
1674 * Scan all the bits in the ifa's address.
1675 * If a bit dissagrees with what we are
1676 * looking for, mask it with the netmask
1677 * to see if it really matters.
1678 * (A byte at a time)
1679 */
1680 if (ifa->ifa_netmask == 0)
1681 continue;
1682 cp = addr_data;
1683 cp2 = ifa->ifa_addr->sa_data;
1684 cp3 = ifa->ifa_netmask->sa_data;
1685 cplim = ifa->ifa_netmask->sa_len
1686 + (char *)ifa->ifa_netmask;
1687 while (cp3 < cplim)
1688 if ((*cp++ ^ *cp2++) & *cp3++)
1689 goto next; /* next address! */
1690 /*
1691 * If the netmask of what we just found
1692 * is more specific than what we had before
1693 * (if we had one) then remember the new one
1694 * before continuing to search
1695 * for an even better one.
1696 */
1697 if (ifa_maybe == NULL ||
1698 rn_refines((caddr_t)ifa->ifa_netmask,
1699 (caddr_t)ifa_maybe->ifa_netmask)) {
1700 if (ifa_maybe != NULL)
1701 ifa_free(ifa_maybe);
1702 ifa_maybe = ifa;
1703 ifa_ref(ifa_maybe);
1704 }
1705 }
1706 }
1707 IF_ADDR_UNLOCK(ifp);
1708 }
1709 ifa = ifa_maybe;
1710 ifa_maybe = NULL;
1711 done:
1712 IFNET_RUNLOCK_NOSLEEP();
1713 if (ifa_maybe != NULL)
1714 ifa_free(ifa_maybe);
1715 return (ifa);
1716 }
1717
1718 /*
1719 * Find an interface address specific to an interface best matching
1720 * a given address.
1721 */
1722 struct ifaddr *
1723 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1724 {
1725 struct ifaddr *ifa;
1726 char *cp, *cp2, *cp3;
1727 char *cplim;
1728 struct ifaddr *ifa_maybe = NULL;
1729 u_int af = addr->sa_family;
1730
1731 if (af >= AF_MAX)
1732 return (0);
1733 IF_ADDR_LOCK(ifp);
1734 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1735 if (ifa->ifa_addr->sa_family != af)
1736 continue;
1737 if (ifa_maybe == NULL)
1738 ifa_maybe = ifa;
1739 if (ifa->ifa_netmask == 0) {
1740 if (sa_equal(addr, ifa->ifa_addr) ||
1741 (ifa->ifa_dstaddr &&
1742 sa_equal(addr, ifa->ifa_dstaddr)))
1743 goto done;
1744 continue;
1745 }
1746 if (ifp->if_flags & IFF_POINTOPOINT) {
1747 if (sa_equal(addr, ifa->ifa_dstaddr))
1748 goto done;
1749 } else {
1750 cp = addr->sa_data;
1751 cp2 = ifa->ifa_addr->sa_data;
1752 cp3 = ifa->ifa_netmask->sa_data;
1753 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1754 for (; cp3 < cplim; cp3++)
1755 if ((*cp++ ^ *cp2++) & *cp3)
1756 break;
1757 if (cp3 == cplim)
1758 goto done;
1759 }
1760 }
1761 ifa = ifa_maybe;
1762 done:
1763 if (ifa != NULL)
1764 ifa_ref(ifa);
1765 IF_ADDR_UNLOCK(ifp);
1766 return (ifa);
1767 }
1768
1769 #include <net/if_llatbl.h>
1770
1771 /*
1772 * Default action when installing a route with a Link Level gateway.
1773 * Lookup an appropriate real ifa to point to.
1774 * This should be moved to /sys/net/link.c eventually.
1775 */
1776 static void
1777 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1778 {
1779 struct ifaddr *ifa, *oifa;
1780 struct sockaddr *dst;
1781 struct ifnet *ifp;
1782
1783 RT_LOCK_ASSERT(rt);
1784
1785 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1786 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1787 return;
1788 ifa = ifaof_ifpforaddr(dst, ifp);
1789 if (ifa) {
1790 oifa = rt->rt_ifa;
1791 rt->rt_ifa = ifa;
1792 ifa_free(oifa);
1793 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1794 ifa->ifa_rtrequest(cmd, rt, info);
1795 }
1796 }
1797
1798 /*
1799 * Mark an interface down and notify protocols of
1800 * the transition.
1801 * NOTE: must be called at splnet or eqivalent.
1802 */
1803 static void
1804 if_unroute(struct ifnet *ifp, int flag, int fam)
1805 {
1806 struct ifaddr *ifa;
1807
1808 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1809
1810 ifp->if_flags &= ~flag;
1811 getmicrotime(&ifp->if_lastchange);
1812 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1813 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1814 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1815 ifp->if_qflush(ifp);
1816
1817 #if defined(INET) || defined(INET6)
1818 #ifdef DEV_CARP
1819 if (ifp->if_carp)
1820 carp_carpdev_state(ifp->if_carp);
1821 #endif
1822 #endif
1823 rt_ifmsg(ifp);
1824 }
1825
1826 /*
1827 * Mark an interface up and notify protocols of
1828 * the transition.
1829 * NOTE: must be called at splnet or eqivalent.
1830 */
1831 static void
1832 if_route(struct ifnet *ifp, int flag, int fam)
1833 {
1834 struct ifaddr *ifa;
1835
1836 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1837
1838 ifp->if_flags |= flag;
1839 getmicrotime(&ifp->if_lastchange);
1840 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1841 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1842 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1843 #if defined(INET) || defined(INET6)
1844 #ifdef DEV_CARP
1845 if (ifp->if_carp)
1846 carp_carpdev_state(ifp->if_carp);
1847 #endif
1848 #endif
1849 rt_ifmsg(ifp);
1850 #ifdef INET6
1851 in6_if_up(ifp);
1852 #endif
1853 }
1854
1855 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1856 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
1857
1858 /*
1859 * Handle a change in the interface link state. To avoid LORs
1860 * between driver lock and upper layer locks, as well as possible
1861 * recursions, we post event to taskqueue, and all job
1862 * is done in static do_link_state_change().
1863 */
1864 void
1865 if_link_state_change(struct ifnet *ifp, int link_state)
1866 {
1867 /* Return if state hasn't changed. */
1868 if (ifp->if_link_state == link_state)
1869 return;
1870
1871 ifp->if_link_state = link_state;
1872
1873 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1874 }
1875
1876 static void
1877 do_link_state_change(void *arg, int pending)
1878 {
1879 struct ifnet *ifp = (struct ifnet *)arg;
1880 int link_state = ifp->if_link_state;
1881 int link;
1882 CURVNET_SET(ifp->if_vnet);
1883
1884 /* Notify that the link state has changed. */
1885 rt_ifmsg(ifp);
1886 if (link_state == LINK_STATE_UP)
1887 link = NOTE_LINKUP;
1888 else if (link_state == LINK_STATE_DOWN)
1889 link = NOTE_LINKDOWN;
1890 else
1891 link = NOTE_LINKINV;
1892 if (ifp->if_vlantrunk != NULL)
1893 (*vlan_link_state_p)(ifp, link);
1894
1895 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1896 IFP2AC(ifp)->ac_netgraph != NULL)
1897 (*ng_ether_link_state_p)(ifp, link_state);
1898 #if defined(INET) || defined(INET6)
1899 #ifdef DEV_CARP
1900 if (ifp->if_carp)
1901 carp_carpdev_state(ifp->if_carp);
1902 #endif
1903 #endif
1904 if (ifp->if_bridge) {
1905 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1906 (*bstp_linkstate_p)(ifp, link_state);
1907 }
1908 if (ifp->if_lagg) {
1909 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1910 (*lagg_linkstate_p)(ifp, link_state);
1911 }
1912
1913 if (IS_DEFAULT_VNET(curvnet))
1914 devctl_notify("IFNET", ifp->if_xname,
1915 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1916 NULL);
1917 if (pending > 1)
1918 if_printf(ifp, "%d link states coalesced\n", pending);
1919 if (log_link_state_change)
1920 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1921 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1922 CURVNET_RESTORE();
1923 }
1924
1925 /*
1926 * Mark an interface down and notify protocols of
1927 * the transition.
1928 * NOTE: must be called at splnet or eqivalent.
1929 */
1930 void
1931 if_down(struct ifnet *ifp)
1932 {
1933
1934 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1935 }
1936
1937 /*
1938 * Mark an interface up and notify protocols of
1939 * the transition.
1940 * NOTE: must be called at splnet or eqivalent.
1941 */
1942 void
1943 if_up(struct ifnet *ifp)
1944 {
1945
1946 if_route(ifp, IFF_UP, AF_UNSPEC);
1947 }
1948
1949 /*
1950 * Flush an interface queue.
1951 */
1952 void
1953 if_qflush(struct ifnet *ifp)
1954 {
1955 struct mbuf *m, *n;
1956 struct ifaltq *ifq;
1957
1958 ifq = &ifp->if_snd;
1959 IFQ_LOCK(ifq);
1960 #ifdef ALTQ
1961 if (ALTQ_IS_ENABLED(ifq))
1962 ALTQ_PURGE(ifq);
1963 #endif
1964 n = ifq->ifq_head;
1965 while ((m = n) != 0) {
1966 n = m->m_act;
1967 m_freem(m);
1968 }
1969 ifq->ifq_head = 0;
1970 ifq->ifq_tail = 0;
1971 ifq->ifq_len = 0;
1972 IFQ_UNLOCK(ifq);
1973 }
1974
1975 /*
1976 * Handle interface watchdog timer routines. Called
1977 * from softclock, we decrement timers (if set) and
1978 * call the appropriate interface routine on expiration.
1979 *
1980 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1981 * holding Giant.
1982 */
1983 static void
1984 if_slowtimo(void *arg)
1985 {
1986 VNET_ITERATOR_DECL(vnet_iter);
1987 struct ifnet *ifp;
1988 int s = splimp();
1989
1990 VNET_LIST_RLOCK_NOSLEEP();
1991 IFNET_RLOCK_NOSLEEP();
1992 VNET_FOREACH(vnet_iter) {
1993 CURVNET_SET(vnet_iter);
1994 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1995 if (ifp->if_timer == 0 || --ifp->if_timer)
1996 continue;
1997 if (ifp->if_watchdog)
1998 (*ifp->if_watchdog)(ifp);
1999 }
2000 CURVNET_RESTORE();
2001 }
2002 IFNET_RUNLOCK_NOSLEEP();
2003 VNET_LIST_RUNLOCK_NOSLEEP();
2004 splx(s);
2005 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
2006 }
2007
2008 /*
2009 * Map interface name to interface structure pointer, with or without
2010 * returning a reference.
2011 */
2012 struct ifnet *
2013 ifunit_ref(const char *name)
2014 {
2015 struct ifnet *ifp;
2016
2017 IFNET_RLOCK_NOSLEEP();
2018 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2019 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2020 !(ifp->if_flags & IFF_DYING))
2021 break;
2022 }
2023 if (ifp != NULL)
2024 if_ref(ifp);
2025 IFNET_RUNLOCK_NOSLEEP();
2026 return (ifp);
2027 }
2028
2029 struct ifnet *
2030 ifunit(const char *name)
2031 {
2032 struct ifnet *ifp;
2033
2034 IFNET_RLOCK_NOSLEEP();
2035 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2036 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2037 break;
2038 }
2039 IFNET_RUNLOCK_NOSLEEP();
2040 return (ifp);
2041 }
2042
2043 /*
2044 * Hardware specific interface ioctls.
2045 */
2046 static int
2047 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2048 {
2049 struct ifreq *ifr;
2050 struct ifstat *ifs;
2051 int error = 0;
2052 int new_flags, temp_flags;
2053 size_t namelen, onamelen;
2054 char new_name[IFNAMSIZ];
2055 struct ifaddr *ifa;
2056 struct sockaddr_dl *sdl;
2057
2058 ifr = (struct ifreq *)data;
2059 switch (cmd) {
2060 case SIOCGIFINDEX:
2061 ifr->ifr_index = ifp->if_index;
2062 break;
2063
2064 case SIOCGIFFLAGS:
2065 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2066 ifr->ifr_flags = temp_flags & 0xffff;
2067 ifr->ifr_flagshigh = temp_flags >> 16;
2068 break;
2069
2070 case SIOCGIFCAP:
2071 ifr->ifr_reqcap = ifp->if_capabilities;
2072 ifr->ifr_curcap = ifp->if_capenable;
2073 break;
2074
2075 #ifdef MAC
2076 case SIOCGIFMAC:
2077 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2078 break;
2079 #endif
2080
2081 case SIOCGIFMETRIC:
2082 ifr->ifr_metric = ifp->if_metric;
2083 break;
2084
2085 case SIOCGIFMTU:
2086 ifr->ifr_mtu = ifp->if_mtu;
2087 break;
2088
2089 case SIOCGIFPHYS:
2090 ifr->ifr_phys = ifp->if_physical;
2091 break;
2092
2093 case SIOCSIFFLAGS:
2094 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2095 if (error)
2096 return (error);
2097 /*
2098 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2099 * check, so we don't need special handling here yet.
2100 */
2101 new_flags = (ifr->ifr_flags & 0xffff) |
2102 (ifr->ifr_flagshigh << 16);
2103 if (ifp->if_flags & IFF_SMART) {
2104 /* Smart drivers twiddle their own routes */
2105 } else if (ifp->if_flags & IFF_UP &&
2106 (new_flags & IFF_UP) == 0) {
2107 int s = splimp();
2108 if_down(ifp);
2109 splx(s);
2110 } else if (new_flags & IFF_UP &&
2111 (ifp->if_flags & IFF_UP) == 0) {
2112 int s = splimp();
2113 if_up(ifp);
2114 splx(s);
2115 }
2116 /* See if permanently promiscuous mode bit is about to flip */
2117 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2118 if (new_flags & IFF_PPROMISC)
2119 ifp->if_flags |= IFF_PROMISC;
2120 else if (ifp->if_pcount == 0)
2121 ifp->if_flags &= ~IFF_PROMISC;
2122 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2123 ifp->if_xname,
2124 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2125 }
2126 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2127 (new_flags &~ IFF_CANTCHANGE);
2128 if (ifp->if_ioctl) {
2129 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2130 }
2131 getmicrotime(&ifp->if_lastchange);
2132 break;
2133
2134 case SIOCSIFCAP:
2135 error = priv_check(td, PRIV_NET_SETIFCAP);
2136 if (error)
2137 return (error);
2138 if (ifp->if_ioctl == NULL)
2139 return (EOPNOTSUPP);
2140 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2141 return (EINVAL);
2142 error = (*ifp->if_ioctl)(ifp, cmd, data);
2143 if (error == 0)
2144 getmicrotime(&ifp->if_lastchange);
2145 break;
2146
2147 #ifdef MAC
2148 case SIOCSIFMAC:
2149 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2150 break;
2151 #endif
2152
2153 case SIOCSIFNAME:
2154 error = priv_check(td, PRIV_NET_SETIFNAME);
2155 if (error)
2156 return (error);
2157 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2158 if (error != 0)
2159 return (error);
2160 if (new_name[0] == '\0')
2161 return (EINVAL);
2162 if (ifunit(new_name) != NULL)
2163 return (EEXIST);
2164
2165 /* Announce the departure of the interface. */
2166 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2167 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2168
2169 log(LOG_INFO, "%s: changing name to '%s'\n",
2170 ifp->if_xname, new_name);
2171
2172 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2173 ifa = ifp->if_addr;
2174 IFA_LOCK(ifa);
2175 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2176 namelen = strlen(new_name);
2177 onamelen = sdl->sdl_nlen;
2178 /*
2179 * Move the address if needed. This is safe because we
2180 * allocate space for a name of length IFNAMSIZ when we
2181 * create this in if_attach().
2182 */
2183 if (namelen != onamelen) {
2184 bcopy(sdl->sdl_data + onamelen,
2185 sdl->sdl_data + namelen, sdl->sdl_alen);
2186 }
2187 bcopy(new_name, sdl->sdl_data, namelen);
2188 sdl->sdl_nlen = namelen;
2189 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2190 bzero(sdl->sdl_data, onamelen);
2191 while (namelen != 0)
2192 sdl->sdl_data[--namelen] = 0xff;
2193 IFA_UNLOCK(ifa);
2194
2195 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2196 /* Announce the return of the interface. */
2197 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2198 break;
2199
2200 #ifdef VIMAGE
2201 case SIOCSIFVNET:
2202 error = priv_check(td, PRIV_NET_SETIFVNET);
2203 if (error)
2204 return (error);
2205 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2206 break;
2207 #endif
2208
2209 case SIOCSIFMETRIC:
2210 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2211 if (error)
2212 return (error);
2213 ifp->if_metric = ifr->ifr_metric;
2214 getmicrotime(&ifp->if_lastchange);
2215 break;
2216
2217 case SIOCSIFPHYS:
2218 error = priv_check(td, PRIV_NET_SETIFPHYS);
2219 if (error)
2220 return (error);
2221 if (ifp->if_ioctl == NULL)
2222 return (EOPNOTSUPP);
2223 error = (*ifp->if_ioctl)(ifp, cmd, data);
2224 if (error == 0)
2225 getmicrotime(&ifp->if_lastchange);
2226 break;
2227
2228 case SIOCSIFMTU:
2229 {
2230 u_long oldmtu = ifp->if_mtu;
2231
2232 error = priv_check(td, PRIV_NET_SETIFMTU);
2233 if (error)
2234 return (error);
2235 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2236 return (EINVAL);
2237 if (ifp->if_ioctl == NULL)
2238 return (EOPNOTSUPP);
2239 error = (*ifp->if_ioctl)(ifp, cmd, data);
2240 if (error == 0) {
2241 getmicrotime(&ifp->if_lastchange);
2242 rt_ifmsg(ifp);
2243 }
2244 /*
2245 * If the link MTU changed, do network layer specific procedure.
2246 */
2247 if (ifp->if_mtu != oldmtu) {
2248 #ifdef INET6
2249 nd6_setmtu(ifp);
2250 #endif
2251 }
2252 break;
2253 }
2254
2255 case SIOCADDMULTI:
2256 case SIOCDELMULTI:
2257 if (cmd == SIOCADDMULTI)
2258 error = priv_check(td, PRIV_NET_ADDMULTI);
2259 else
2260 error = priv_check(td, PRIV_NET_DELMULTI);
2261 if (error)
2262 return (error);
2263
2264 /* Don't allow group membership on non-multicast interfaces. */
2265 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2266 return (EOPNOTSUPP);
2267
2268 /* Don't let users screw up protocols' entries. */
2269 if (ifr->ifr_addr.sa_family != AF_LINK)
2270 return (EINVAL);
2271
2272 if (cmd == SIOCADDMULTI) {
2273 struct ifmultiaddr *ifma;
2274
2275 /*
2276 * Userland is only permitted to join groups once
2277 * via the if_addmulti() KPI, because it cannot hold
2278 * struct ifmultiaddr * between calls. It may also
2279 * lose a race while we check if the membership
2280 * already exists.
2281 */
2282 IF_ADDR_LOCK(ifp);
2283 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2284 IF_ADDR_UNLOCK(ifp);
2285 if (ifma != NULL)
2286 error = EADDRINUSE;
2287 else
2288 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2289 } else {
2290 error = if_delmulti(ifp, &ifr->ifr_addr);
2291 }
2292 if (error == 0)
2293 getmicrotime(&ifp->if_lastchange);
2294 break;
2295
2296 case SIOCSIFPHYADDR:
2297 case SIOCDIFPHYADDR:
2298 #ifdef INET6
2299 case SIOCSIFPHYADDR_IN6:
2300 #endif
2301 case SIOCSLIFPHYADDR:
2302 case SIOCSIFMEDIA:
2303 case SIOCSIFGENERIC:
2304 error = priv_check(td, PRIV_NET_HWIOCTL);
2305 if (error)
2306 return (error);
2307 if (ifp->if_ioctl == NULL)
2308 return (EOPNOTSUPP);
2309 error = (*ifp->if_ioctl)(ifp, cmd, data);
2310 if (error == 0)
2311 getmicrotime(&ifp->if_lastchange);
2312 break;
2313
2314 case SIOCGIFSTATUS:
2315 ifs = (struct ifstat *)data;
2316 ifs->ascii[0] = '\0';
2317
2318 case SIOCGIFPSRCADDR:
2319 case SIOCGIFPDSTADDR:
2320 case SIOCGLIFPHYADDR:
2321 case SIOCGIFMEDIA:
2322 case SIOCGIFGENERIC:
2323 if (ifp->if_ioctl == NULL)
2324 return (EOPNOTSUPP);
2325 error = (*ifp->if_ioctl)(ifp, cmd, data);
2326 break;
2327
2328 case SIOCSIFLLADDR:
2329 error = priv_check(td, PRIV_NET_SETLLADDR);
2330 if (error)
2331 return (error);
2332 error = if_setlladdr(ifp,
2333 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2334 break;
2335
2336 case SIOCAIFGROUP:
2337 {
2338 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2339
2340 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2341 if (error)
2342 return (error);
2343 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2344 return (error);
2345 break;
2346 }
2347
2348 case SIOCGIFGROUP:
2349 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2350 return (error);
2351 break;
2352
2353 case SIOCDIFGROUP:
2354 {
2355 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2356
2357 error = priv_check(td, PRIV_NET_DELIFGROUP);
2358 if (error)
2359 return (error);
2360 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2361 return (error);
2362 break;
2363 }
2364
2365 default:
2366 error = ENOIOCTL;
2367 break;
2368 }
2369 return (error);
2370 }
2371
2372 /*
2373 * Interface ioctls.
2374 */
2375 int
2376 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2377 {
2378 struct ifnet *ifp;
2379 struct ifreq *ifr;
2380 int error;
2381 int oif_flags;
2382
2383 switch (cmd) {
2384 case SIOCGIFCONF:
2385 case OSIOCGIFCONF:
2386 #ifdef __amd64__
2387 case SIOCGIFCONF32:
2388 #endif
2389 return (ifconf(cmd, data));
2390 }
2391 ifr = (struct ifreq *)data;
2392
2393 switch (cmd) {
2394 #ifdef VIMAGE
2395 case SIOCSIFRVNET:
2396 error = priv_check(td, PRIV_NET_SETIFVNET);
2397 if (error)
2398 return (error);
2399 return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
2400 #endif
2401 case SIOCIFCREATE:
2402 case SIOCIFCREATE2:
2403 error = priv_check(td, PRIV_NET_IFCREATE);
2404 if (error)
2405 return (error);
2406 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
2407 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
2408 case SIOCIFDESTROY:
2409 error = priv_check(td, PRIV_NET_IFDESTROY);
2410 if (error)
2411 return (error);
2412 return if_clone_destroy(ifr->ifr_name);
2413
2414 case SIOCIFGCLONERS:
2415 return (if_clone_list((struct if_clonereq *)data));
2416 case SIOCGIFGMEMB:
2417 return (if_getgroupmembers((struct ifgroupreq *)data));
2418 }
2419
2420 ifp = ifunit_ref(ifr->ifr_name);
2421 if (ifp == NULL)
2422 return (ENXIO);
2423
2424 error = ifhwioctl(cmd, ifp, data, td);
2425 if (error != ENOIOCTL) {
2426 if_rele(ifp);
2427 return (error);
2428 }
2429
2430 oif_flags = ifp->if_flags;
2431 if (so->so_proto == NULL) {
2432 if_rele(ifp);
2433 return (EOPNOTSUPP);
2434 }
2435 #ifndef COMPAT_43
2436 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2437 data,
2438 ifp, td));
2439 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2440 error = (*ifp->if_ioctl)(ifp, cmd, data);
2441 #else
2442 {
2443 u_long ocmd = cmd;
2444
2445 switch (cmd) {
2446
2447 case SIOCSIFDSTADDR:
2448 case SIOCSIFADDR:
2449 case SIOCSIFBRDADDR:
2450 case SIOCSIFNETMASK:
2451 #if BYTE_ORDER != BIG_ENDIAN
2452 if (ifr->ifr_addr.sa_family == 0 &&
2453 ifr->ifr_addr.sa_len < 16) {
2454 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2455 ifr->ifr_addr.sa_len = 16;
2456 }
2457 #else
2458 if (ifr->ifr_addr.sa_len == 0)
2459 ifr->ifr_addr.sa_len = 16;
2460 #endif
2461 break;
2462
2463 case OSIOCGIFADDR:
2464 cmd = SIOCGIFADDR;
2465 break;
2466
2467 case OSIOCGIFDSTADDR:
2468 cmd = SIOCGIFDSTADDR;
2469 break;
2470
2471 case OSIOCGIFBRDADDR:
2472 cmd = SIOCGIFBRDADDR;
2473 break;
2474
2475 case OSIOCGIFNETMASK:
2476 cmd = SIOCGIFNETMASK;
2477 }
2478 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2479 cmd,
2480 data,
2481 ifp, td));
2482 if (error == EOPNOTSUPP && ifp != NULL &&
2483 ifp->if_ioctl != NULL)
2484 error = (*ifp->if_ioctl)(ifp, cmd, data);
2485 switch (ocmd) {
2486
2487 case OSIOCGIFADDR:
2488 case OSIOCGIFDSTADDR:
2489 case OSIOCGIFBRDADDR:
2490 case OSIOCGIFNETMASK:
2491 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2492
2493 }
2494 }
2495 #endif /* COMPAT_43 */
2496
2497 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2498 #ifdef INET6
2499 if (ifp->if_flags & IFF_UP) {
2500 int s = splimp();
2501 in6_if_up(ifp);
2502 splx(s);
2503 }
2504 #endif
2505 }
2506 if_rele(ifp);
2507 return (error);
2508 }
2509
2510 /*
2511 * The code common to handling reference counted flags,
2512 * e.g., in ifpromisc() and if_allmulti().
2513 * The "pflag" argument can specify a permanent mode flag to check,
2514 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2515 *
2516 * Only to be used on stack-owned flags, not driver-owned flags.
2517 */
2518 static int
2519 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2520 {
2521 struct ifreq ifr;
2522 int error;
2523 int oldflags, oldcount;
2524
2525 /* Sanity checks to catch programming errors */
2526 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2527 ("%s: setting driver-owned flag %d", __func__, flag));
2528
2529 if (onswitch)
2530 KASSERT(*refcount >= 0,
2531 ("%s: increment negative refcount %d for flag %d",
2532 __func__, *refcount, flag));
2533 else
2534 KASSERT(*refcount > 0,
2535 ("%s: decrement non-positive refcount %d for flag %d",
2536 __func__, *refcount, flag));
2537
2538 /* In case this mode is permanent, just touch refcount */
2539 if (ifp->if_flags & pflag) {
2540 *refcount += onswitch ? 1 : -1;
2541 return (0);
2542 }
2543
2544 /* Save ifnet parameters for if_ioctl() may fail */
2545 oldcount = *refcount;
2546 oldflags = ifp->if_flags;
2547
2548 /*
2549 * See if we aren't the only and touching refcount is enough.
2550 * Actually toggle interface flag if we are the first or last.
2551 */
2552 if (onswitch) {
2553 if ((*refcount)++)
2554 return (0);
2555 ifp->if_flags |= flag;
2556 } else {
2557 if (--(*refcount))
2558 return (0);
2559 ifp->if_flags &= ~flag;
2560 }
2561
2562 /* Call down the driver since we've changed interface flags */
2563 if (ifp->if_ioctl == NULL) {
2564 error = EOPNOTSUPP;
2565 goto recover;
2566 }
2567 ifr.ifr_flags = ifp->if_flags & 0xffff;
2568 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2569 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2570 if (error)
2571 goto recover;
2572 /* Notify userland that interface flags have changed */
2573 rt_ifmsg(ifp);
2574 return (0);
2575
2576 recover:
2577 /* Recover after driver error */
2578 *refcount = oldcount;
2579 ifp->if_flags = oldflags;
2580 return (error);
2581 }
2582
2583 /*
2584 * Set/clear promiscuous mode on interface ifp based on the truth value
2585 * of pswitch. The calls are reference counted so that only the first
2586 * "on" request actually has an effect, as does the final "off" request.
2587 * Results are undefined if the "off" and "on" requests are not matched.
2588 */
2589 int
2590 ifpromisc(struct ifnet *ifp, int pswitch)
2591 {
2592 int error;
2593 int oldflags = ifp->if_flags;
2594
2595 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2596 &ifp->if_pcount, pswitch);
2597 /* If promiscuous mode status has changed, log a message */
2598 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2599 log(LOG_INFO, "%s: promiscuous mode %s\n",
2600 ifp->if_xname,
2601 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2602 return (error);
2603 }
2604
2605 /*
2606 * Return interface configuration
2607 * of system. List may be used
2608 * in later ioctl's (above) to get
2609 * other information.
2610 */
2611 /*ARGSUSED*/
2612 static int
2613 ifconf(u_long cmd, caddr_t data)
2614 {
2615 struct ifconf *ifc = (struct ifconf *)data;
2616 #ifdef __amd64__
2617 struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2618 struct ifconf ifc_swab;
2619 #endif
2620 struct ifnet *ifp;
2621 struct ifaddr *ifa;
2622 struct ifreq ifr;
2623 struct sbuf *sb;
2624 int error, full = 0, valid_len, max_len;
2625
2626 #ifdef __amd64__
2627 if (cmd == SIOCGIFCONF32) {
2628 ifc_swab.ifc_len = ifc32->ifc_len;
2629 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2630 ifc = &ifc_swab;
2631 }
2632 #endif
2633 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2634 max_len = MAXPHYS - 1;
2635
2636 /* Prevent hostile input from being able to crash the system */
2637 if (ifc->ifc_len <= 0)
2638 return (EINVAL);
2639
2640 again:
2641 if (ifc->ifc_len <= max_len) {
2642 max_len = ifc->ifc_len;
2643 full = 1;
2644 }
2645 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2646 max_len = 0;
2647 valid_len = 0;
2648
2649 IFNET_RLOCK();
2650 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2651 int addrs;
2652
2653 /*
2654 * Zero the ifr_name buffer to make sure we don't
2655 * disclose the contents of the stack.
2656 */
2657 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2658
2659 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2660 >= sizeof(ifr.ifr_name)) {
2661 sbuf_delete(sb);
2662 IFNET_RUNLOCK();
2663 return (ENAMETOOLONG);
2664 }
2665
2666 addrs = 0;
2667 IF_ADDR_LOCK(ifp);
2668 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2669 struct sockaddr *sa = ifa->ifa_addr;
2670
2671 if (prison_if(curthread->td_ucred, sa) != 0)
2672 continue;
2673 addrs++;
2674 #ifdef COMPAT_43
2675 if (cmd == OSIOCGIFCONF) {
2676 struct osockaddr *osa =
2677 (struct osockaddr *)&ifr.ifr_addr;
2678 ifr.ifr_addr = *sa;
2679 osa->sa_family = sa->sa_family;
2680 sbuf_bcat(sb, &ifr, sizeof(ifr));
2681 max_len += sizeof(ifr);
2682 } else
2683 #endif
2684 if (sa->sa_len <= sizeof(*sa)) {
2685 ifr.ifr_addr = *sa;
2686 sbuf_bcat(sb, &ifr, sizeof(ifr));
2687 max_len += sizeof(ifr);
2688 } else {
2689 sbuf_bcat(sb, &ifr,
2690 offsetof(struct ifreq, ifr_addr));
2691 max_len += offsetof(struct ifreq, ifr_addr);
2692 sbuf_bcat(sb, sa, sa->sa_len);
2693 max_len += sa->sa_len;
2694 }
2695
2696 if (!sbuf_overflowed(sb))
2697 valid_len = sbuf_len(sb);
2698 }
2699 IF_ADDR_UNLOCK(ifp);
2700 if (addrs == 0) {
2701 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2702 sbuf_bcat(sb, &ifr, sizeof(ifr));
2703 max_len += sizeof(ifr);
2704
2705 if (!sbuf_overflowed(sb))
2706 valid_len = sbuf_len(sb);
2707 }
2708 }
2709 IFNET_RUNLOCK();
2710
2711 /*
2712 * If we didn't allocate enough space (uncommon), try again. If
2713 * we have already allocated as much space as we are allowed,
2714 * return what we've got.
2715 */
2716 if (valid_len != max_len && !full) {
2717 sbuf_delete(sb);
2718 goto again;
2719 }
2720
2721 ifc->ifc_len = valid_len;
2722 #ifdef __amd64__
2723 if (cmd == SIOCGIFCONF32)
2724 ifc32->ifc_len = valid_len;
2725 #endif
2726 sbuf_finish(sb);
2727 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2728 sbuf_delete(sb);
2729 return (error);
2730 }
2731
2732 /*
2733 * Just like ifpromisc(), but for all-multicast-reception mode.
2734 */
2735 int
2736 if_allmulti(struct ifnet *ifp, int onswitch)
2737 {
2738
2739 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2740 }
2741
2742 struct ifmultiaddr *
2743 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2744 {
2745 struct ifmultiaddr *ifma;
2746
2747 IF_ADDR_LOCK_ASSERT(ifp);
2748
2749 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2750 if (sa->sa_family == AF_LINK) {
2751 if (sa_dl_equal(ifma->ifma_addr, sa))
2752 break;
2753 } else {
2754 if (sa_equal(ifma->ifma_addr, sa))
2755 break;
2756 }
2757 }
2758
2759 return ifma;
2760 }
2761
2762 /*
2763 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2764 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2765 * the ifnet multicast address list here, so the caller must do that and
2766 * other setup work (such as notifying the device driver). The reference
2767 * count is initialized to 1.
2768 */
2769 static struct ifmultiaddr *
2770 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2771 int mflags)
2772 {
2773 struct ifmultiaddr *ifma;
2774 struct sockaddr *dupsa;
2775
2776 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2777 M_ZERO);
2778 if (ifma == NULL)
2779 return (NULL);
2780
2781 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2782 if (dupsa == NULL) {
2783 free(ifma, M_IFMADDR);
2784 return (NULL);
2785 }
2786 bcopy(sa, dupsa, sa->sa_len);
2787 ifma->ifma_addr = dupsa;
2788
2789 ifma->ifma_ifp = ifp;
2790 ifma->ifma_refcount = 1;
2791 ifma->ifma_protospec = NULL;
2792
2793 if (llsa == NULL) {
2794 ifma->ifma_lladdr = NULL;
2795 return (ifma);
2796 }
2797
2798 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2799 if (dupsa == NULL) {
2800 free(ifma->ifma_addr, M_IFMADDR);
2801 free(ifma, M_IFMADDR);
2802 return (NULL);
2803 }
2804 bcopy(llsa, dupsa, llsa->sa_len);
2805 ifma->ifma_lladdr = dupsa;
2806
2807 return (ifma);
2808 }
2809
2810 /*
2811 * if_freemulti: free ifmultiaddr structure and possibly attached related
2812 * addresses. The caller is responsible for implementing reference
2813 * counting, notifying the driver, handling routing messages, and releasing
2814 * any dependent link layer state.
2815 */
2816 static void
2817 if_freemulti(struct ifmultiaddr *ifma)
2818 {
2819
2820 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2821 ifma->ifma_refcount));
2822 KASSERT(ifma->ifma_protospec == NULL,
2823 ("if_freemulti: protospec not NULL"));
2824
2825 if (ifma->ifma_lladdr != NULL)
2826 free(ifma->ifma_lladdr, M_IFMADDR);
2827 free(ifma->ifma_addr, M_IFMADDR);
2828 free(ifma, M_IFMADDR);
2829 }
2830
2831 /*
2832 * Register an additional multicast address with a network interface.
2833 *
2834 * - If the address is already present, bump the reference count on the
2835 * address and return.
2836 * - If the address is not link-layer, look up a link layer address.
2837 * - Allocate address structures for one or both addresses, and attach to the
2838 * multicast address list on the interface. If automatically adding a link
2839 * layer address, the protocol address will own a reference to the link
2840 * layer address, to be freed when it is freed.
2841 * - Notify the network device driver of an addition to the multicast address
2842 * list.
2843 *
2844 * 'sa' points to caller-owned memory with the desired multicast address.
2845 *
2846 * 'retifma' will be used to return a pointer to the resulting multicast
2847 * address reference, if desired.
2848 */
2849 int
2850 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2851 struct ifmultiaddr **retifma)
2852 {
2853 struct ifmultiaddr *ifma, *ll_ifma;
2854 struct sockaddr *llsa;
2855 int error;
2856
2857 /*
2858 * If the address is already present, return a new reference to it;
2859 * otherwise, allocate storage and set up a new address.
2860 */
2861 IF_ADDR_LOCK(ifp);
2862 ifma = if_findmulti(ifp, sa);
2863 if (ifma != NULL) {
2864 ifma->ifma_refcount++;
2865 if (retifma != NULL)
2866 *retifma = ifma;
2867 IF_ADDR_UNLOCK(ifp);
2868 return (0);
2869 }
2870
2871 /*
2872 * The address isn't already present; resolve the protocol address
2873 * into a link layer address, and then look that up, bump its
2874 * refcount or allocate an ifma for that also. If 'llsa' was
2875 * returned, we will need to free it later.
2876 */
2877 llsa = NULL;
2878 ll_ifma = NULL;
2879 if (ifp->if_resolvemulti != NULL) {
2880 error = ifp->if_resolvemulti(ifp, &llsa, sa);
2881 if (error)
2882 goto unlock_out;
2883 }
2884
2885 /*
2886 * Allocate the new address. Don't hook it up yet, as we may also
2887 * need to allocate a link layer multicast address.
2888 */
2889 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2890 if (ifma == NULL) {
2891 error = ENOMEM;
2892 goto free_llsa_out;
2893 }
2894
2895 /*
2896 * If a link layer address is found, we'll need to see if it's
2897 * already present in the address list, or allocate is as well.
2898 * When this block finishes, the link layer address will be on the
2899 * list.
2900 */
2901 if (llsa != NULL) {
2902 ll_ifma = if_findmulti(ifp, llsa);
2903 if (ll_ifma == NULL) {
2904 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2905 if (ll_ifma == NULL) {
2906 --ifma->ifma_refcount;
2907 if_freemulti(ifma);
2908 error = ENOMEM;
2909 goto free_llsa_out;
2910 }
2911 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2912 ifma_link);
2913 } else
2914 ll_ifma->ifma_refcount++;
2915 ifma->ifma_llifma = ll_ifma;
2916 }
2917
2918 /*
2919 * We now have a new multicast address, ifma, and possibly a new or
2920 * referenced link layer address. Add the primary address to the
2921 * ifnet address list.
2922 */
2923 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2924
2925 if (retifma != NULL)
2926 *retifma = ifma;
2927
2928 /*
2929 * Must generate the message while holding the lock so that 'ifma'
2930 * pointer is still valid.
2931 */
2932 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2933 IF_ADDR_UNLOCK(ifp);
2934
2935 /*
2936 * We are certain we have added something, so call down to the
2937 * interface to let them know about it.
2938 */
2939 if (ifp->if_ioctl != NULL) {
2940 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2941 }
2942
2943 if (llsa != NULL)
2944 free(llsa, M_IFMADDR);
2945
2946 return (0);
2947
2948 free_llsa_out:
2949 if (llsa != NULL)
2950 free(llsa, M_IFMADDR);
2951
2952 unlock_out:
2953 IF_ADDR_UNLOCK(ifp);
2954 return (error);
2955 }
2956
2957 /*
2958 * Delete a multicast group membership by network-layer group address.
2959 *
2960 * Returns ENOENT if the entry could not be found. If ifp no longer
2961 * exists, results are undefined. This entry point should only be used
2962 * from subsystems which do appropriate locking to hold ifp for the
2963 * duration of the call.
2964 * Network-layer protocol domains must use if_delmulti_ifma().
2965 */
2966 int
2967 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2968 {
2969 struct ifmultiaddr *ifma;
2970 int lastref;
2971 #ifdef INVARIANTS
2972 struct ifnet *oifp;
2973
2974 IFNET_RLOCK_NOSLEEP();
2975 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
2976 if (ifp == oifp)
2977 break;
2978 if (ifp != oifp)
2979 ifp = NULL;
2980 IFNET_RUNLOCK_NOSLEEP();
2981
2982 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
2983 #endif
2984 if (ifp == NULL)
2985 return (ENOENT);
2986
2987 IF_ADDR_LOCK(ifp);
2988 lastref = 0;
2989 ifma = if_findmulti(ifp, sa);
2990 if (ifma != NULL)
2991 lastref = if_delmulti_locked(ifp, ifma, 0);
2992 IF_ADDR_UNLOCK(ifp);
2993
2994 if (ifma == NULL)
2995 return (ENOENT);
2996
2997 if (lastref && ifp->if_ioctl != NULL) {
2998 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2999 }
3000
3001 return (0);
3002 }
3003
3004 /*
3005 * Delete a multicast group membership by group membership pointer.
3006 * Network-layer protocol domains must use this routine.
3007 *
3008 * It is safe to call this routine if the ifp disappeared.
3009 */
3010 void
3011 if_delmulti_ifma(struct ifmultiaddr *ifma)
3012 {
3013 struct ifnet *ifp;
3014 int lastref;
3015
3016 ifp = ifma->ifma_ifp;
3017 #ifdef DIAGNOSTIC
3018 if (ifp == NULL) {
3019 printf("%s: ifma_ifp seems to be detached\n", __func__);
3020 } else {
3021 struct ifnet *oifp;
3022
3023 IFNET_RLOCK_NOSLEEP();
3024 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3025 if (ifp == oifp)
3026 break;
3027 if (ifp != oifp) {
3028 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3029 ifp = NULL;
3030 }
3031 IFNET_RUNLOCK_NOSLEEP();
3032 }
3033 #endif
3034 /*
3035 * If and only if the ifnet instance exists: Acquire the address lock.
3036 */
3037 if (ifp != NULL)
3038 IF_ADDR_LOCK(ifp);
3039
3040 lastref = if_delmulti_locked(ifp, ifma, 0);
3041
3042 if (ifp != NULL) {
3043 /*
3044 * If and only if the ifnet instance exists:
3045 * Release the address lock.
3046 * If the group was left: update the hardware hash filter.
3047 */
3048 IF_ADDR_UNLOCK(ifp);
3049 if (lastref && ifp->if_ioctl != NULL) {
3050 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3051 }
3052 }
3053 }
3054
3055 /*
3056 * Perform deletion of network-layer and/or link-layer multicast address.
3057 *
3058 * Return 0 if the reference count was decremented.
3059 * Return 1 if the final reference was released, indicating that the
3060 * hardware hash filter should be reprogrammed.
3061 */
3062 static int
3063 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3064 {
3065 struct ifmultiaddr *ll_ifma;
3066
3067 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3068 KASSERT(ifma->ifma_ifp == ifp,
3069 ("%s: inconsistent ifp %p", __func__, ifp));
3070 IF_ADDR_LOCK_ASSERT(ifp);
3071 }
3072
3073 ifp = ifma->ifma_ifp;
3074
3075 /*
3076 * If the ifnet is detaching, null out references to ifnet,
3077 * so that upper protocol layers will notice, and not attempt
3078 * to obtain locks for an ifnet which no longer exists. The
3079 * routing socket announcement must happen before the ifnet
3080 * instance is detached from the system.
3081 */
3082 if (detaching) {
3083 #ifdef DIAGNOSTIC
3084 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3085 #endif
3086 /*
3087 * ifp may already be nulled out if we are being reentered
3088 * to delete the ll_ifma.
3089 */
3090 if (ifp != NULL) {
3091 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3092 ifma->ifma_ifp = NULL;
3093 }
3094 }
3095
3096 if (--ifma->ifma_refcount > 0)
3097 return 0;
3098
3099 /*
3100 * If this ifma is a network-layer ifma, a link-layer ifma may
3101 * have been associated with it. Release it first if so.
3102 */
3103 ll_ifma = ifma->ifma_llifma;
3104 if (ll_ifma != NULL) {
3105 KASSERT(ifma->ifma_lladdr != NULL,
3106 ("%s: llifma w/o lladdr", __func__));
3107 if (detaching)
3108 ll_ifma->ifma_ifp = NULL; /* XXX */
3109 if (--ll_ifma->ifma_refcount == 0) {
3110 if (ifp != NULL) {
3111 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3112 ifma_link);
3113 }
3114 if_freemulti(ll_ifma);
3115 }
3116 }
3117
3118 if (ifp != NULL)
3119 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3120
3121 if_freemulti(ifma);
3122
3123 /*
3124 * The last reference to this instance of struct ifmultiaddr
3125 * was released; the hardware should be notified of this change.
3126 */
3127 return 1;
3128 }
3129
3130 /*
3131 * Set the link layer address on an interface.
3132 *
3133 * At this time we only support certain types of interfaces,
3134 * and we don't allow the length of the address to change.
3135 */
3136 int
3137 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3138 {
3139 struct sockaddr_dl *sdl;
3140 struct ifaddr *ifa;
3141 struct ifreq ifr;
3142
3143 IF_ADDR_LOCK(ifp);
3144 ifa = ifp->if_addr;
3145 if (ifa == NULL) {
3146 IF_ADDR_UNLOCK(ifp);
3147 return (EINVAL);
3148 }
3149 ifa_ref(ifa);
3150 IF_ADDR_UNLOCK(ifp);
3151 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3152 if (sdl == NULL) {
3153 ifa_free(ifa);
3154 return (EINVAL);
3155 }
3156 if (len != sdl->sdl_alen) { /* don't allow length to change */
3157 ifa_free(ifa);
3158 return (EINVAL);
3159 }
3160 switch (ifp->if_type) {
3161 case IFT_ETHER:
3162 case IFT_FDDI:
3163 case IFT_XETHER:
3164 case IFT_ISO88025:
3165 case IFT_L2VLAN:
3166 case IFT_BRIDGE:
3167 case IFT_ARCNET:
3168 case IFT_IEEE8023ADLAG:
3169 case IFT_IEEE80211:
3170 bcopy(lladdr, LLADDR(sdl), len);
3171 ifa_free(ifa);
3172 break;
3173 default:
3174 ifa_free(ifa);
3175 return (ENODEV);
3176 }
3177
3178 /*
3179 * If the interface is already up, we need
3180 * to re-init it in order to reprogram its
3181 * address filter.
3182 */
3183 if ((ifp->if_flags & IFF_UP) != 0) {
3184 if (ifp->if_ioctl) {
3185 ifp->if_flags &= ~IFF_UP;
3186 ifr.ifr_flags = ifp->if_flags & 0xffff;
3187 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3188 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3189 ifp->if_flags |= IFF_UP;
3190 ifr.ifr_flags = ifp->if_flags & 0xffff;
3191 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3192 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3193 }
3194 #ifdef INET
3195 /*
3196 * Also send gratuitous ARPs to notify other nodes about
3197 * the address change.
3198 */
3199 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3200 if (ifa->ifa_addr->sa_family == AF_INET)
3201 arp_ifinit(ifp, ifa);
3202 }
3203 #endif
3204 }
3205 return (0);
3206 }
3207
3208 /*
3209 * The name argument must be a pointer to storage which will last as
3210 * long as the interface does. For physical devices, the result of
3211 * device_get_name(dev) is a good choice and for pseudo-devices a
3212 * static string works well.
3213 */
3214 void
3215 if_initname(struct ifnet *ifp, const char *name, int unit)
3216 {
3217 ifp->if_dname = name;
3218 ifp->if_dunit = unit;
3219 if (unit != IF_DUNIT_NONE)
3220 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3221 else
3222 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3223 }
3224
3225 int
3226 if_printf(struct ifnet *ifp, const char * fmt, ...)
3227 {
3228 va_list ap;
3229 int retval;
3230
3231 retval = printf("%s: ", ifp->if_xname);
3232 va_start(ap, fmt);
3233 retval += vprintf(fmt, ap);
3234 va_end(ap);
3235 return (retval);
3236 }
3237
3238 void
3239 if_start(struct ifnet *ifp)
3240 {
3241
3242 (*(ifp)->if_start)(ifp);
3243 }
3244
3245 /*
3246 * Backwards compatibility interface for drivers
3247 * that have not implemented it
3248 */
3249 static int
3250 if_transmit(struct ifnet *ifp, struct mbuf *m)
3251 {
3252 int error;
3253
3254 IFQ_HANDOFF(ifp, m, error);
3255 return (error);
3256 }
3257
3258 int
3259 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3260 {
3261 int active = 0;
3262
3263 IF_LOCK(ifq);
3264 if (_IF_QFULL(ifq)) {
3265 _IF_DROP(ifq);
3266 IF_UNLOCK(ifq);
3267 m_freem(m);
3268 return (0);
3269 }
3270 if (ifp != NULL) {
3271 ifp->if_obytes += m->m_pkthdr.len + adjust;
3272 if (m->m_flags & (M_BCAST|M_MCAST))
3273 ifp->if_omcasts++;
3274 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3275 }
3276 _IF_ENQUEUE(ifq, m);
3277 IF_UNLOCK(ifq);
3278 if (ifp != NULL && !active)
3279 (*(ifp)->if_start)(ifp);
3280 return (1);
3281 }
3282
3283 void
3284 if_register_com_alloc(u_char type,
3285 if_com_alloc_t *a, if_com_free_t *f)
3286 {
3287
3288 KASSERT(if_com_alloc[type] == NULL,
3289 ("if_register_com_alloc: %d already registered", type));
3290 KASSERT(if_com_free[type] == NULL,
3291 ("if_register_com_alloc: %d free already registered", type));
3292
3293 if_com_alloc[type] = a;
3294 if_com_free[type] = f;
3295 }
3296
3297 void
3298 if_deregister_com_alloc(u_char type)
3299 {
3300
3301 KASSERT(if_com_alloc[type] != NULL,
3302 ("if_deregister_com_alloc: %d not registered", type));
3303 KASSERT(if_com_free[type] != NULL,
3304 ("if_deregister_com_alloc: %d free not registered", type));
3305 if_com_alloc[type] = NULL;
3306 if_com_free[type] = NULL;
3307 }
Cache object: 8f0a28e9418f5579be53cbe48dd52aa6
|