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/10.3/sys/net/if.c 294502 2016-01-21 14:08:40Z bz $
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 #include <net/ethernet.h>
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip.h>
81 #include <netinet/ip_carp.h>
82 #ifdef INET
83 #include <netinet/if_ether.h>
84 #endif /* INET */
85 #ifdef INET6
86 #include <netinet6/in6_var.h>
87 #include <netinet6/in6_ifattach.h>
88 #endif /* INET6 */
89 #endif /* INET || INET6 */
90
91 #include <security/mac/mac_framework.h>
92
93 #ifdef COMPAT_FREEBSD32
94 #include <sys/mount.h>
95 #include <compat/freebsd32/freebsd32.h>
96 #endif
97
98 struct ifindex_entry {
99 struct ifnet *ife_ifnet;
100 };
101
102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
104
105 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
106 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
107 &ifqmaxlen, 0, "max send queue size");
108
109 /* Log link state change events */
110 static int log_link_state_change = 1;
111
112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
113 &log_link_state_change, 0,
114 "log interface link state change events");
115
116 /* Interface description */
117 static unsigned int ifdescr_maxlen = 1024;
118 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
119 &ifdescr_maxlen, 0,
120 "administrative maximum length for interface description");
121
122 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
123
124 /* global sx for non-critical path ifdescr */
125 static struct sx ifdescr_sx;
126 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
127
128 void (*bridge_linkstate_p)(struct ifnet *ifp);
129 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
130 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
131 /* These are external hooks for CARP. */
132 void (*carp_linkstate_p)(struct ifnet *ifp);
133 void (*carp_demote_adj_p)(int, char *);
134 int (*carp_master_p)(struct ifaddr *);
135 #if defined(INET) || defined(INET6)
136 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
137 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
138 const struct sockaddr *sa);
139 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
140 int (*carp_attach_p)(struct ifaddr *, int);
141 void (*carp_detach_p)(struct ifaddr *);
142 #endif
143 #ifdef INET
144 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
145 #endif
146 #ifdef INET6
147 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
148 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
149 const struct in6_addr *taddr);
150 #endif
151
152 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
153
154 /*
155 * XXX: Style; these should be sorted alphabetically, and unprototyped
156 * static functions should be prototyped. Currently they are sorted by
157 * declaration order.
158 */
159 static void if_attachdomain(void *);
160 static void if_attachdomain1(struct ifnet *);
161 static int ifconf(u_long, caddr_t);
162 static void if_freemulti(struct ifmultiaddr *);
163 static void if_init(void *);
164 static void if_grow(void);
165 static void if_input_default(struct ifnet *, struct mbuf *);
166 static void if_route(struct ifnet *, int flag, int fam);
167 static int if_setflag(struct ifnet *, int, int, int *, int);
168 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
169 static void if_unroute(struct ifnet *, int flag, int fam);
170 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
171 static int if_rtdel(struct radix_node *, void *);
172 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
173 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
174 static void do_link_state_change(void *, int);
175 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
176 static int if_getgroupmembers(struct ifgroupreq *);
177 static void if_delgroups(struct ifnet *);
178 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
179 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
180
181 #ifdef INET6
182 /*
183 * XXX: declare here to avoid to include many inet6 related files..
184 * should be more generalized?
185 */
186 extern void nd6_setmtu(struct ifnet *);
187 #endif
188
189 VNET_DEFINE(int, if_index);
190 int ifqmaxlen = IFQ_MAXLEN;
191 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
192 VNET_DEFINE(struct ifgrouphead, ifg_head);
193
194 static VNET_DEFINE(int, if_indexlim) = 8;
195
196 /* Table of ifnet by index. */
197 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
198
199 #define V_if_indexlim VNET(if_indexlim)
200 #define V_ifindex_table VNET(ifindex_table)
201
202 /*
203 * The global network interface list (V_ifnet) and related state (such as
204 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
205 * an rwlock. Either may be acquired shared to stablize the list, but both
206 * must be acquired writable to modify the list. This model allows us to
207 * both stablize the interface list during interrupt thread processing, but
208 * also to stablize it over long-running ioctls, without introducing priority
209 * inversions and deadlocks.
210 */
211 struct rwlock ifnet_rwlock;
212 struct sx ifnet_sxlock;
213
214 /*
215 * The allocation of network interfaces is a rather non-atomic affair; we
216 * need to select an index before we are ready to expose the interface for
217 * use, so will use this pointer value to indicate reservation.
218 */
219 #define IFNET_HOLD (void *)(uintptr_t)(-1)
220
221 static if_com_alloc_t *if_com_alloc[256];
222 static if_com_free_t *if_com_free[256];
223
224 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
225 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
226 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
227
228 struct ifnet *
229 ifnet_byindex_locked(u_short idx)
230 {
231
232 if (idx > V_if_index)
233 return (NULL);
234 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
235 return (NULL);
236 return (V_ifindex_table[idx].ife_ifnet);
237 }
238
239 struct ifnet *
240 ifnet_byindex(u_short idx)
241 {
242 struct ifnet *ifp;
243
244 IFNET_RLOCK_NOSLEEP();
245 ifp = ifnet_byindex_locked(idx);
246 IFNET_RUNLOCK_NOSLEEP();
247 return (ifp);
248 }
249
250 struct ifnet *
251 ifnet_byindex_ref(u_short idx)
252 {
253 struct ifnet *ifp;
254
255 IFNET_RLOCK_NOSLEEP();
256 ifp = ifnet_byindex_locked(idx);
257 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
258 IFNET_RUNLOCK_NOSLEEP();
259 return (NULL);
260 }
261 if_ref(ifp);
262 IFNET_RUNLOCK_NOSLEEP();
263 return (ifp);
264 }
265
266 /*
267 * Allocate an ifindex array entry; return 0 on success or an error on
268 * failure.
269 */
270 static int
271 ifindex_alloc_locked(u_short *idxp)
272 {
273 u_short idx;
274
275 IFNET_WLOCK_ASSERT();
276
277 retry:
278 /*
279 * Try to find an empty slot below V_if_index. If we fail, take the
280 * next slot.
281 */
282 for (idx = 1; idx <= V_if_index; idx++) {
283 if (V_ifindex_table[idx].ife_ifnet == NULL)
284 break;
285 }
286
287 /* Catch if_index overflow. */
288 if (idx < 1)
289 return (ENOSPC);
290 if (idx >= V_if_indexlim) {
291 if_grow();
292 goto retry;
293 }
294 if (idx > V_if_index)
295 V_if_index = idx;
296 *idxp = idx;
297 return (0);
298 }
299
300 static void
301 ifindex_free_locked(u_short idx)
302 {
303
304 IFNET_WLOCK_ASSERT();
305
306 V_ifindex_table[idx].ife_ifnet = NULL;
307 while (V_if_index > 0 &&
308 V_ifindex_table[V_if_index].ife_ifnet == NULL)
309 V_if_index--;
310 }
311
312 static void
313 ifindex_free(u_short idx)
314 {
315
316 IFNET_WLOCK();
317 ifindex_free_locked(idx);
318 IFNET_WUNLOCK();
319 }
320
321 static void
322 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
323 {
324
325 IFNET_WLOCK_ASSERT();
326
327 V_ifindex_table[idx].ife_ifnet = ifp;
328 }
329
330 static void
331 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
332 {
333
334 IFNET_WLOCK();
335 ifnet_setbyindex_locked(idx, ifp);
336 IFNET_WUNLOCK();
337 }
338
339 struct ifaddr *
340 ifaddr_byindex(u_short idx)
341 {
342 struct ifaddr *ifa;
343
344 IFNET_RLOCK_NOSLEEP();
345 ifa = ifnet_byindex_locked(idx)->if_addr;
346 if (ifa != NULL)
347 ifa_ref(ifa);
348 IFNET_RUNLOCK_NOSLEEP();
349 return (ifa);
350 }
351
352 /*
353 * Network interface utility routines.
354 *
355 * Routines with ifa_ifwith* names take sockaddr *'s as
356 * parameters.
357 */
358
359 static void
360 vnet_if_init(const void *unused __unused)
361 {
362
363 TAILQ_INIT(&V_ifnet);
364 TAILQ_INIT(&V_ifg_head);
365 IFNET_WLOCK();
366 if_grow(); /* create initial table */
367 IFNET_WUNLOCK();
368 vnet_if_clone_init();
369 }
370 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
371 NULL);
372
373 /* ARGSUSED*/
374 static void
375 if_init(void *dummy __unused)
376 {
377
378 IFNET_LOCK_INIT();
379 if_clone_init();
380 }
381 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
382
383
384 #ifdef VIMAGE
385 static void
386 vnet_if_uninit(const void *unused __unused)
387 {
388
389 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
390 "not empty", __func__, __LINE__, &V_ifnet));
391 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
392 "not empty", __func__, __LINE__, &V_ifg_head));
393
394 free((caddr_t)V_ifindex_table, M_IFNET);
395 }
396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
397 vnet_if_uninit, NULL);
398 #endif
399
400 static void
401 if_grow(void)
402 {
403 int oldlim;
404 u_int n;
405 struct ifindex_entry *e;
406
407 IFNET_WLOCK_ASSERT();
408 oldlim = V_if_indexlim;
409 IFNET_WUNLOCK();
410 n = (oldlim << 1) * sizeof(*e);
411 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
412 IFNET_WLOCK();
413 if (V_if_indexlim != oldlim) {
414 free(e, M_IFNET);
415 return;
416 }
417 if (V_ifindex_table != NULL) {
418 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
419 free((caddr_t)V_ifindex_table, M_IFNET);
420 }
421 V_if_indexlim <<= 1;
422 V_ifindex_table = e;
423 }
424
425 /*
426 * Allocate a struct ifnet and an index for an interface. A layer 2
427 * common structure will also be allocated if an allocation routine is
428 * registered for the passed type.
429 */
430 struct ifnet *
431 if_alloc(u_char type)
432 {
433 struct ifnet *ifp;
434 u_short idx;
435
436 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
437 IFNET_WLOCK();
438 if (ifindex_alloc_locked(&idx) != 0) {
439 IFNET_WUNLOCK();
440 free(ifp, M_IFNET);
441 return (NULL);
442 }
443 ifnet_setbyindex_locked(idx, IFNET_HOLD);
444 IFNET_WUNLOCK();
445 ifp->if_index = idx;
446 ifp->if_type = type;
447 ifp->if_alloctype = type;
448 if (if_com_alloc[type] != NULL) {
449 ifp->if_l2com = if_com_alloc[type](type, ifp);
450 if (ifp->if_l2com == NULL) {
451 free(ifp, M_IFNET);
452 ifindex_free(idx);
453 return (NULL);
454 }
455 }
456
457 IF_ADDR_LOCK_INIT(ifp);
458 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
459 ifp->if_afdata_initialized = 0;
460 IF_AFDATA_LOCK_INIT(ifp);
461 TAILQ_INIT(&ifp->if_addrhead);
462 TAILQ_INIT(&ifp->if_multiaddrs);
463 TAILQ_INIT(&ifp->if_groups);
464 #ifdef MAC
465 mac_ifnet_init(ifp);
466 #endif
467 ifq_init(&ifp->if_snd, ifp);
468
469 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
470 ifnet_setbyindex(ifp->if_index, ifp);
471 return (ifp);
472 }
473
474 /*
475 * Do the actual work of freeing a struct ifnet, and layer 2 common
476 * structure. This call is made when the last reference to an
477 * interface is released.
478 */
479 static void
480 if_free_internal(struct ifnet *ifp)
481 {
482
483 KASSERT((ifp->if_flags & IFF_DYING),
484 ("if_free_internal: interface not dying"));
485
486 if (if_com_free[ifp->if_alloctype] != NULL)
487 if_com_free[ifp->if_alloctype](ifp->if_l2com,
488 ifp->if_alloctype);
489
490 #ifdef MAC
491 mac_ifnet_destroy(ifp);
492 #endif /* MAC */
493 if (ifp->if_description != NULL)
494 free(ifp->if_description, M_IFDESCR);
495 IF_AFDATA_DESTROY(ifp);
496 IF_ADDR_LOCK_DESTROY(ifp);
497 ifq_delete(&ifp->if_snd);
498 free(ifp, M_IFNET);
499 }
500
501 /*
502 * Deregister an interface and free the associated storage.
503 */
504 void
505 if_free(struct ifnet *ifp)
506 {
507
508 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
509
510 CURVNET_SET_QUIET(ifp->if_vnet);
511 IFNET_WLOCK();
512 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
513 ("%s: freeing unallocated ifnet", ifp->if_xname));
514
515 ifindex_free_locked(ifp->if_index);
516 IFNET_WUNLOCK();
517
518 if (refcount_release(&ifp->if_refcount))
519 if_free_internal(ifp);
520 CURVNET_RESTORE();
521 }
522
523 /*
524 * Interfaces to keep an ifnet type-stable despite the possibility of the
525 * driver calling if_free(). If there are additional references, we defer
526 * freeing the underlying data structure.
527 */
528 void
529 if_ref(struct ifnet *ifp)
530 {
531
532 /* We don't assert the ifnet list lock here, but arguably should. */
533 refcount_acquire(&ifp->if_refcount);
534 }
535
536 void
537 if_rele(struct ifnet *ifp)
538 {
539
540 if (!refcount_release(&ifp->if_refcount))
541 return;
542 if_free_internal(ifp);
543 }
544
545 void
546 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
547 {
548
549 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
550
551 if (ifq->ifq_maxlen == 0)
552 ifq->ifq_maxlen = ifqmaxlen;
553
554 ifq->altq_type = 0;
555 ifq->altq_disc = NULL;
556 ifq->altq_flags &= ALTQF_CANTCHANGE;
557 ifq->altq_tbr = NULL;
558 ifq->altq_ifp = ifp;
559 }
560
561 void
562 ifq_delete(struct ifaltq *ifq)
563 {
564 mtx_destroy(&ifq->ifq_mtx);
565 }
566
567 /*
568 * Perform generic interface initalization tasks and attach the interface
569 * to the list of "active" interfaces. If vmove flag is set on entry
570 * to if_attach_internal(), perform only a limited subset of initialization
571 * tasks, given that we are moving from one vnet to another an ifnet which
572 * has already been fully initialized.
573 *
574 * Note that if_detach_internal() removes group membership unconditionally
575 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
576 * Thus, when if_vmove() is applied to a cloned interface, group membership
577 * is lost while a cloned one always joins a group whose name is
578 * ifc->ifc_name. To recover this after if_detach_internal() and
579 * if_attach_internal(), the cloner should be specified to
580 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
581 * attempts to join a group whose name is ifc->ifc_name.
582 *
583 * XXX:
584 * - The decision to return void and thus require this function to
585 * succeed is questionable.
586 * - We should probably do more sanity checking. For instance we don't
587 * do anything to insure if_xname is unique or non-empty.
588 */
589 void
590 if_attach(struct ifnet *ifp)
591 {
592
593 if_attach_internal(ifp, 0, NULL);
594 }
595
596 /*
597 * Compute the least common TSO limit.
598 */
599 void
600 if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
601 {
602 /*
603 * 1) If there is no limit currently, take the limit from
604 * the network adapter.
605 *
606 * 2) If the network adapter has a limit below the current
607 * limit, apply it.
608 */
609 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
610 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
611 pmax->tsomaxbytes = ifp->if_hw_tsomax;
612 }
613 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
614 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
615 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
616 }
617 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
618 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
619 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
620 }
621 }
622
623 /*
624 * Update TSO limit of a network adapter.
625 *
626 * Returns zero if no change. Else non-zero.
627 */
628 int
629 if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
630 {
631 int retval = 0;
632 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
633 ifp->if_hw_tsomax = pmax->tsomaxbytes;
634 retval++;
635 }
636 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
637 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
638 retval++;
639 }
640 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
641 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
642 retval++;
643 }
644 return (retval);
645 }
646
647 static void
648 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
649 {
650 unsigned socksize, ifasize;
651 int namelen, masklen;
652 struct sockaddr_dl *sdl;
653 struct ifaddr *ifa;
654
655 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
656 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
657 ifp->if_xname);
658
659 #ifdef VIMAGE
660 ifp->if_vnet = curvnet;
661 if (ifp->if_home_vnet == NULL)
662 ifp->if_home_vnet = curvnet;
663 #endif
664
665 if_addgroup(ifp, IFG_ALL);
666
667 /* Restore group membership for cloned interfaces. */
668 if (vmove && ifc != NULL)
669 if_clone_addgroup(ifp, ifc);
670
671 getmicrotime(&ifp->if_lastchange);
672 ifp->if_data.ifi_epoch = time_uptime;
673 ifp->if_data.ifi_datalen = sizeof(struct if_data);
674
675 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
676 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
677 ("transmit and qflush must both either be set or both be NULL"));
678 if (ifp->if_transmit == NULL) {
679 ifp->if_transmit = if_transmit;
680 ifp->if_qflush = if_qflush;
681 }
682 if (ifp->if_input == NULL)
683 ifp->if_input = if_input_default;
684
685 if (!vmove) {
686 #ifdef MAC
687 mac_ifnet_create(ifp);
688 #endif
689
690 /*
691 * Create a Link Level name for this device.
692 */
693 namelen = strlen(ifp->if_xname);
694 /*
695 * Always save enough space for any possiable name so we
696 * can do a rename in place later.
697 */
698 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
699 socksize = masklen + ifp->if_addrlen;
700 if (socksize < sizeof(*sdl))
701 socksize = sizeof(*sdl);
702 socksize = roundup2(socksize, sizeof(long));
703 ifasize = sizeof(*ifa) + 2 * socksize;
704 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
705 ifa_init(ifa);
706 sdl = (struct sockaddr_dl *)(ifa + 1);
707 sdl->sdl_len = socksize;
708 sdl->sdl_family = AF_LINK;
709 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
710 sdl->sdl_nlen = namelen;
711 sdl->sdl_index = ifp->if_index;
712 sdl->sdl_type = ifp->if_type;
713 ifp->if_addr = ifa;
714 ifa->ifa_ifp = ifp;
715 ifa->ifa_rtrequest = link_rtrequest;
716 ifa->ifa_addr = (struct sockaddr *)sdl;
717 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
718 ifa->ifa_netmask = (struct sockaddr *)sdl;
719 sdl->sdl_len = masklen;
720 while (namelen != 0)
721 sdl->sdl_data[--namelen] = 0xff;
722 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
723 /* Reliably crash if used uninitialized. */
724 ifp->if_broadcastaddr = NULL;
725
726 #if defined(INET) || defined(INET6)
727 /* Use defaults for TSO, if nothing is set */
728 if (ifp->if_hw_tsomax == 0 &&
729 ifp->if_hw_tsomaxsegcount == 0 &&
730 ifp->if_hw_tsomaxsegsize == 0) {
731 /*
732 * The TSO defaults needs to be such that an
733 * NFS mbuf list of 35 mbufs totalling just
734 * below 64K works and that a chain of mbufs
735 * can be defragged into at most 32 segments:
736 */
737 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
738 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
739 ifp->if_hw_tsomaxsegcount = 35;
740 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
741
742 /* XXX some drivers set IFCAP_TSO after ethernet attach */
743 if (ifp->if_capabilities & IFCAP_TSO) {
744 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
745 ifp->if_hw_tsomax,
746 ifp->if_hw_tsomaxsegcount,
747 ifp->if_hw_tsomaxsegsize);
748 }
749 }
750 #endif
751 }
752 #ifdef VIMAGE
753 else {
754 /*
755 * Update the interface index in the link layer address
756 * of the interface.
757 */
758 for (ifa = ifp->if_addr; ifa != NULL;
759 ifa = TAILQ_NEXT(ifa, ifa_link)) {
760 if (ifa->ifa_addr->sa_family == AF_LINK) {
761 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
762 sdl->sdl_index = ifp->if_index;
763 }
764 }
765 }
766 #endif
767
768 IFNET_WLOCK();
769 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
770 #ifdef VIMAGE
771 curvnet->vnet_ifcnt++;
772 #endif
773 IFNET_WUNLOCK();
774
775 if (domain_init_status >= 2)
776 if_attachdomain1(ifp);
777
778 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
779 if (IS_DEFAULT_VNET(curvnet))
780 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
781
782 /* Announce the interface. */
783 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
784 }
785
786 static void
787 if_attachdomain(void *dummy)
788 {
789 struct ifnet *ifp;
790
791 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
792 if_attachdomain1(ifp);
793 }
794 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
795 if_attachdomain, NULL);
796
797 static void
798 if_attachdomain1(struct ifnet *ifp)
799 {
800 struct domain *dp;
801
802 /*
803 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
804 * cannot lock ifp->if_afdata initialization, entirely.
805 */
806 if (IF_AFDATA_TRYLOCK(ifp) == 0)
807 return;
808 if (ifp->if_afdata_initialized >= domain_init_status) {
809 IF_AFDATA_UNLOCK(ifp);
810 log(LOG_WARNING, "%s called more than once on %s\n",
811 __func__, ifp->if_xname);
812 return;
813 }
814 ifp->if_afdata_initialized = domain_init_status;
815 IF_AFDATA_UNLOCK(ifp);
816
817 /* address family dependent data region */
818 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
819 for (dp = domains; dp; dp = dp->dom_next) {
820 if (dp->dom_ifattach)
821 ifp->if_afdata[dp->dom_family] =
822 (*dp->dom_ifattach)(ifp);
823 }
824 }
825
826 /*
827 * Remove any unicast or broadcast network addresses from an interface.
828 */
829 void
830 if_purgeaddrs(struct ifnet *ifp)
831 {
832 struct ifaddr *ifa, *next;
833
834 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
835 if (ifa->ifa_addr->sa_family == AF_LINK)
836 continue;
837 #ifdef INET
838 /* XXX: Ugly!! ad hoc just for INET */
839 if (ifa->ifa_addr->sa_family == AF_INET) {
840 struct ifaliasreq ifr;
841
842 bzero(&ifr, sizeof(ifr));
843 ifr.ifra_addr = *ifa->ifa_addr;
844 if (ifa->ifa_dstaddr)
845 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
846 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
847 NULL) == 0)
848 continue;
849 }
850 #endif /* INET */
851 #ifdef INET6
852 if (ifa->ifa_addr->sa_family == AF_INET6) {
853 in6_purgeaddr(ifa);
854 /* ifp_addrhead is already updated */
855 continue;
856 }
857 #endif /* INET6 */
858 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
859 ifa_free(ifa);
860 }
861 }
862
863 /*
864 * Remove any multicast network addresses from an interface when an ifnet
865 * is going away.
866 */
867 static void
868 if_purgemaddrs(struct ifnet *ifp)
869 {
870 struct ifmultiaddr *ifma;
871 struct ifmultiaddr *next;
872
873 IF_ADDR_WLOCK(ifp);
874 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
875 if_delmulti_locked(ifp, ifma, 1);
876 IF_ADDR_WUNLOCK(ifp);
877 }
878
879 /*
880 * Detach an interface, removing it from the list of "active" interfaces.
881 * If vmove flag is set on entry to if_detach_internal(), perform only a
882 * limited subset of cleanup tasks, given that we are moving an ifnet from
883 * one vnet to another, where it must be fully operational.
884 *
885 * XXXRW: There are some significant questions about event ordering, and
886 * how to prevent things from starting to use the interface during detach.
887 */
888 void
889 if_detach(struct ifnet *ifp)
890 {
891
892 CURVNET_SET_QUIET(ifp->if_vnet);
893 if_detach_internal(ifp, 0, NULL);
894 CURVNET_RESTORE();
895 }
896
897 static int
898 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
899 {
900 struct ifaddr *ifa;
901 struct radix_node_head *rnh;
902 int i, j;
903 struct domain *dp;
904 struct ifnet *iter;
905 int found = 0;
906
907 IFNET_WLOCK();
908 TAILQ_FOREACH(iter, &V_ifnet, if_link)
909 if (iter == ifp) {
910 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
911 found = 1;
912 break;
913 }
914 #ifdef VIMAGE
915 if (found)
916 curvnet->vnet_ifcnt--;
917 #endif
918 IFNET_WUNLOCK();
919 if (!found) {
920 /*
921 * While we would want to panic here, we cannot
922 * guarantee that the interface is indeed still on
923 * the list given we don't hold locks all the way.
924 */
925 return (ENOENT);
926 #if 0
927 if (vmove)
928 panic("%s: ifp=%p not on the ifnet tailq %p",
929 __func__, ifp, &V_ifnet);
930 else
931 return; /* XXX this should panic as well? */
932 #endif
933 }
934
935 /* Check if this is a cloned interface or not. */
936 if (vmove && ifcp != NULL)
937 *ifcp = if_clone_findifc(ifp);
938
939 /*
940 * Remove/wait for pending events.
941 */
942 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
943
944 /*
945 * Remove routes and flush queues.
946 */
947 if_down(ifp);
948 #ifdef ALTQ
949 if (ALTQ_IS_ENABLED(&ifp->if_snd))
950 altq_disable(&ifp->if_snd);
951 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
952 altq_detach(&ifp->if_snd);
953 #endif
954
955 if_purgeaddrs(ifp);
956
957 #ifdef INET
958 in_ifdetach(ifp);
959 #endif
960
961 #ifdef INET6
962 /*
963 * Remove all IPv6 kernel structs related to ifp. This should be done
964 * before removing routing entries below, since IPv6 interface direct
965 * routes are expected to be removed by the IPv6-specific kernel API.
966 * Otherwise, the kernel will detect some inconsistency and bark it.
967 */
968 in6_ifdetach(ifp);
969 #endif
970 if_purgemaddrs(ifp);
971
972 if (!vmove) {
973 /*
974 * Prevent further calls into the device driver via ifnet.
975 */
976 if_dead(ifp);
977
978 /*
979 * Remove link ifaddr pointer and maybe decrement if_index.
980 * Clean up all addresses.
981 */
982 ifp->if_addr = NULL;
983
984 /* We can now free link ifaddr. */
985 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
986 ifa = TAILQ_FIRST(&ifp->if_addrhead);
987 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
988 ifa_free(ifa);
989 }
990 }
991
992 /*
993 * Delete all remaining routes using this interface
994 * Unfortuneatly the only way to do this is to slog through
995 * the entire routing table looking for routes which point
996 * to this interface...oh well...
997 */
998 for (i = 1; i <= AF_MAX; i++) {
999 for (j = 0; j < rt_numfibs; j++) {
1000 rnh = rt_tables_get_rnh(j, i);
1001 if (rnh == NULL)
1002 continue;
1003 RADIX_NODE_HEAD_LOCK(rnh);
1004 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
1005 RADIX_NODE_HEAD_UNLOCK(rnh);
1006 }
1007 }
1008
1009 /* Announce that the interface is gone. */
1010 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1011 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1012 if (IS_DEFAULT_VNET(curvnet))
1013 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1014 if_delgroups(ifp);
1015
1016 /*
1017 * We cannot hold the lock over dom_ifdetach calls as they might
1018 * sleep, for example trying to drain a callout, thus open up the
1019 * theoretical race with re-attaching.
1020 */
1021 IF_AFDATA_LOCK(ifp);
1022 i = ifp->if_afdata_initialized;
1023 ifp->if_afdata_initialized = 0;
1024 IF_AFDATA_UNLOCK(ifp);
1025 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1026 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1027 (*dp->dom_ifdetach)(ifp,
1028 ifp->if_afdata[dp->dom_family]);
1029 }
1030
1031 return (0);
1032 }
1033
1034 #ifdef VIMAGE
1035 /*
1036 * if_vmove() performs a limited version of if_detach() in current
1037 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1038 * An attempt is made to shrink if_index in current vnet, find an
1039 * unused if_index in target vnet and calls if_grow() if necessary,
1040 * and finally find an unused if_xname for the target vnet.
1041 */
1042 void
1043 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1044 {
1045 u_short idx;
1046 struct if_clone *ifc;
1047 int rc;
1048
1049 /*
1050 * Detach from current vnet, but preserve LLADDR info, do not
1051 * mark as dead etc. so that the ifnet can be reattached later.
1052 * If we cannot find it, we lost the race to someone else.
1053 */
1054 rc = if_detach_internal(ifp, 1, &ifc);
1055 if (rc != 0)
1056 return;
1057
1058 /*
1059 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1060 * the if_index for that vnet if possible.
1061 *
1062 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1063 * or we'd lock on one vnet and unlock on another.
1064 */
1065 IFNET_WLOCK();
1066 ifindex_free_locked(ifp->if_index);
1067 IFNET_WUNLOCK();
1068
1069 /*
1070 * Perform interface-specific reassignment tasks, if provided by
1071 * the driver.
1072 */
1073 if (ifp->if_reassign != NULL)
1074 ifp->if_reassign(ifp, new_vnet, NULL);
1075
1076 /*
1077 * Switch to the context of the target vnet.
1078 */
1079 CURVNET_SET_QUIET(new_vnet);
1080
1081 IFNET_WLOCK();
1082 if (ifindex_alloc_locked(&idx) != 0) {
1083 IFNET_WUNLOCK();
1084 panic("if_index overflow");
1085 }
1086 ifp->if_index = idx;
1087 ifnet_setbyindex_locked(ifp->if_index, ifp);
1088 IFNET_WUNLOCK();
1089
1090 if_attach_internal(ifp, 1, ifc);
1091
1092 CURVNET_RESTORE();
1093 }
1094
1095 /*
1096 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1097 */
1098 static int
1099 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1100 {
1101 struct prison *pr;
1102 struct ifnet *difp;
1103
1104 /* Try to find the prison within our visibility. */
1105 sx_slock(&allprison_lock);
1106 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1107 sx_sunlock(&allprison_lock);
1108 if (pr == NULL)
1109 return (ENXIO);
1110 prison_hold_locked(pr);
1111 mtx_unlock(&pr->pr_mtx);
1112
1113 /* Do not try to move the iface from and to the same prison. */
1114 if (pr->pr_vnet == ifp->if_vnet) {
1115 prison_free(pr);
1116 return (EEXIST);
1117 }
1118
1119 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1120 /* XXX Lock interfaces to avoid races. */
1121 CURVNET_SET_QUIET(pr->pr_vnet);
1122 difp = ifunit(ifname);
1123 CURVNET_RESTORE();
1124 if (difp != NULL) {
1125 prison_free(pr);
1126 return (EEXIST);
1127 }
1128
1129 /* Move the interface into the child jail/vnet. */
1130 if_vmove(ifp, pr->pr_vnet);
1131
1132 /* Report the new if_xname back to the userland. */
1133 sprintf(ifname, "%s", ifp->if_xname);
1134
1135 prison_free(pr);
1136 return (0);
1137 }
1138
1139 static int
1140 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1141 {
1142 struct prison *pr;
1143 struct vnet *vnet_dst;
1144 struct ifnet *ifp;
1145
1146 /* Try to find the prison within our visibility. */
1147 sx_slock(&allprison_lock);
1148 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1149 sx_sunlock(&allprison_lock);
1150 if (pr == NULL)
1151 return (ENXIO);
1152 prison_hold_locked(pr);
1153 mtx_unlock(&pr->pr_mtx);
1154
1155 /* Make sure the named iface exists in the source prison/vnet. */
1156 CURVNET_SET(pr->pr_vnet);
1157 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1158 if (ifp == NULL) {
1159 CURVNET_RESTORE();
1160 prison_free(pr);
1161 return (ENXIO);
1162 }
1163
1164 /* Do not try to move the iface from and to the same prison. */
1165 vnet_dst = TD_TO_VNET(td);
1166 if (vnet_dst == ifp->if_vnet) {
1167 CURVNET_RESTORE();
1168 prison_free(pr);
1169 return (EEXIST);
1170 }
1171
1172 /* Get interface back from child jail/vnet. */
1173 if_vmove(ifp, vnet_dst);
1174 CURVNET_RESTORE();
1175
1176 /* Report the new if_xname back to the userland. */
1177 sprintf(ifname, "%s", ifp->if_xname);
1178
1179 prison_free(pr);
1180 return (0);
1181 }
1182 #endif /* VIMAGE */
1183
1184 /*
1185 * Add a group to an interface
1186 */
1187 int
1188 if_addgroup(struct ifnet *ifp, const char *groupname)
1189 {
1190 struct ifg_list *ifgl;
1191 struct ifg_group *ifg = NULL;
1192 struct ifg_member *ifgm;
1193 int new = 0;
1194
1195 if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
1196 groupname[strlen(groupname) - 1] <= '9')
1197 return (EINVAL);
1198
1199 IFNET_WLOCK();
1200 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1201 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1202 IFNET_WUNLOCK();
1203 return (EEXIST);
1204 }
1205
1206 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1207 M_NOWAIT)) == NULL) {
1208 IFNET_WUNLOCK();
1209 return (ENOMEM);
1210 }
1211
1212 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1213 M_TEMP, M_NOWAIT)) == NULL) {
1214 free(ifgl, M_TEMP);
1215 IFNET_WUNLOCK();
1216 return (ENOMEM);
1217 }
1218
1219 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1220 if (!strcmp(ifg->ifg_group, groupname))
1221 break;
1222
1223 if (ifg == NULL) {
1224 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1225 M_TEMP, M_NOWAIT)) == NULL) {
1226 free(ifgl, M_TEMP);
1227 free(ifgm, M_TEMP);
1228 IFNET_WUNLOCK();
1229 return (ENOMEM);
1230 }
1231 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1232 ifg->ifg_refcnt = 0;
1233 TAILQ_INIT(&ifg->ifg_members);
1234 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1235 new = 1;
1236 }
1237
1238 ifg->ifg_refcnt++;
1239 ifgl->ifgl_group = ifg;
1240 ifgm->ifgm_ifp = ifp;
1241
1242 IF_ADDR_WLOCK(ifp);
1243 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1244 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1245 IF_ADDR_WUNLOCK(ifp);
1246
1247 IFNET_WUNLOCK();
1248
1249 if (new)
1250 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1251 EVENTHANDLER_INVOKE(group_change_event, groupname);
1252
1253 return (0);
1254 }
1255
1256 /*
1257 * Remove a group from an interface
1258 */
1259 int
1260 if_delgroup(struct ifnet *ifp, const char *groupname)
1261 {
1262 struct ifg_list *ifgl;
1263 struct ifg_member *ifgm;
1264
1265 IFNET_WLOCK();
1266 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1267 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1268 break;
1269 if (ifgl == NULL) {
1270 IFNET_WUNLOCK();
1271 return (ENOENT);
1272 }
1273
1274 IF_ADDR_WLOCK(ifp);
1275 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1276 IF_ADDR_WUNLOCK(ifp);
1277
1278 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1279 if (ifgm->ifgm_ifp == ifp)
1280 break;
1281
1282 if (ifgm != NULL) {
1283 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1284 free(ifgm, M_TEMP);
1285 }
1286
1287 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1288 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1289 IFNET_WUNLOCK();
1290 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1291 free(ifgl->ifgl_group, M_TEMP);
1292 } else
1293 IFNET_WUNLOCK();
1294
1295 free(ifgl, M_TEMP);
1296
1297 EVENTHANDLER_INVOKE(group_change_event, groupname);
1298
1299 return (0);
1300 }
1301
1302 /*
1303 * Remove an interface from all groups
1304 */
1305 static void
1306 if_delgroups(struct ifnet *ifp)
1307 {
1308 struct ifg_list *ifgl;
1309 struct ifg_member *ifgm;
1310 char groupname[IFNAMSIZ];
1311
1312 IFNET_WLOCK();
1313 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1314 ifgl = TAILQ_FIRST(&ifp->if_groups);
1315
1316 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1317
1318 IF_ADDR_WLOCK(ifp);
1319 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1320 IF_ADDR_WUNLOCK(ifp);
1321
1322 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1323 if (ifgm->ifgm_ifp == ifp)
1324 break;
1325
1326 if (ifgm != NULL) {
1327 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1328 ifgm_next);
1329 free(ifgm, M_TEMP);
1330 }
1331
1332 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1333 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1334 IFNET_WUNLOCK();
1335 EVENTHANDLER_INVOKE(group_detach_event,
1336 ifgl->ifgl_group);
1337 free(ifgl->ifgl_group, M_TEMP);
1338 } else
1339 IFNET_WUNLOCK();
1340
1341 free(ifgl, M_TEMP);
1342
1343 EVENTHANDLER_INVOKE(group_change_event, groupname);
1344
1345 IFNET_WLOCK();
1346 }
1347 IFNET_WUNLOCK();
1348 }
1349
1350 /*
1351 * Stores all groups from an interface in memory pointed
1352 * to by data
1353 */
1354 static int
1355 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1356 {
1357 int len, error;
1358 struct ifg_list *ifgl;
1359 struct ifg_req ifgrq, *ifgp;
1360 struct ifgroupreq *ifgr = data;
1361
1362 if (ifgr->ifgr_len == 0) {
1363 IF_ADDR_RLOCK(ifp);
1364 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1365 ifgr->ifgr_len += sizeof(struct ifg_req);
1366 IF_ADDR_RUNLOCK(ifp);
1367 return (0);
1368 }
1369
1370 len = ifgr->ifgr_len;
1371 ifgp = ifgr->ifgr_groups;
1372 /* XXX: wire */
1373 IF_ADDR_RLOCK(ifp);
1374 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1375 if (len < sizeof(ifgrq)) {
1376 IF_ADDR_RUNLOCK(ifp);
1377 return (EINVAL);
1378 }
1379 bzero(&ifgrq, sizeof ifgrq);
1380 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1381 sizeof(ifgrq.ifgrq_group));
1382 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1383 IF_ADDR_RUNLOCK(ifp);
1384 return (error);
1385 }
1386 len -= sizeof(ifgrq);
1387 ifgp++;
1388 }
1389 IF_ADDR_RUNLOCK(ifp);
1390
1391 return (0);
1392 }
1393
1394 /*
1395 * Stores all members of a group in memory pointed to by data
1396 */
1397 static int
1398 if_getgroupmembers(struct ifgroupreq *data)
1399 {
1400 struct ifgroupreq *ifgr = data;
1401 struct ifg_group *ifg;
1402 struct ifg_member *ifgm;
1403 struct ifg_req ifgrq, *ifgp;
1404 int len, error;
1405
1406 IFNET_RLOCK();
1407 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1408 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1409 break;
1410 if (ifg == NULL) {
1411 IFNET_RUNLOCK();
1412 return (ENOENT);
1413 }
1414
1415 if (ifgr->ifgr_len == 0) {
1416 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1417 ifgr->ifgr_len += sizeof(ifgrq);
1418 IFNET_RUNLOCK();
1419 return (0);
1420 }
1421
1422 len = ifgr->ifgr_len;
1423 ifgp = ifgr->ifgr_groups;
1424 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1425 if (len < sizeof(ifgrq)) {
1426 IFNET_RUNLOCK();
1427 return (EINVAL);
1428 }
1429 bzero(&ifgrq, sizeof ifgrq);
1430 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1431 sizeof(ifgrq.ifgrq_member));
1432 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1433 IFNET_RUNLOCK();
1434 return (error);
1435 }
1436 len -= sizeof(ifgrq);
1437 ifgp++;
1438 }
1439 IFNET_RUNLOCK();
1440
1441 return (0);
1442 }
1443
1444 /*
1445 * Delete Routes for a Network Interface
1446 *
1447 * Called for each routing entry via the rnh->rnh_walktree() call above
1448 * to delete all route entries referencing a detaching network interface.
1449 *
1450 * Arguments:
1451 * rn pointer to node in the routing table
1452 * arg argument passed to rnh->rnh_walktree() - detaching interface
1453 *
1454 * Returns:
1455 * 0 successful
1456 * errno failed - reason indicated
1457 *
1458 */
1459 static int
1460 if_rtdel(struct radix_node *rn, void *arg)
1461 {
1462 struct rtentry *rt = (struct rtentry *)rn;
1463 struct ifnet *ifp = arg;
1464 int err;
1465
1466 if (rt->rt_ifp == ifp) {
1467
1468 /*
1469 * Protect (sorta) against walktree recursion problems
1470 * with cloned routes
1471 */
1472 if ((rt->rt_flags & RTF_UP) == 0)
1473 return (0);
1474
1475 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1476 rt_mask(rt),
1477 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1478 (struct rtentry **) NULL, rt->rt_fibnum);
1479 if (err) {
1480 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1481 }
1482 }
1483
1484 return (0);
1485 }
1486
1487 /*
1488 * A compatibility function returns ifnet counter values.
1489 */
1490 uint64_t
1491 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1492 {
1493
1494 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1495 switch (cnt) {
1496 case IFCOUNTER_IPACKETS:
1497 return (ifp->if_ipackets);
1498 case IFCOUNTER_IERRORS:
1499 return (ifp->if_ierrors);
1500 case IFCOUNTER_OPACKETS:
1501 return (ifp->if_opackets);
1502 case IFCOUNTER_OERRORS:
1503 return (ifp->if_oerrors);
1504 case IFCOUNTER_COLLISIONS:
1505 return (ifp->if_collisions);
1506 case IFCOUNTER_IBYTES:
1507 return (ifp->if_ibytes);
1508 case IFCOUNTER_OBYTES:
1509 return (ifp->if_obytes);
1510 case IFCOUNTER_IMCASTS:
1511 return (ifp->if_imcasts);
1512 case IFCOUNTER_OMCASTS:
1513 return (ifp->if_omcasts);
1514 case IFCOUNTER_IQDROPS:
1515 return (ifp->if_iqdrops);
1516 #ifdef _IFI_OQDROPS
1517 case IFCOUNTER_OQDROPS:
1518 return (ifp->if_oqdrops);
1519 #endif
1520 case IFCOUNTER_NOPROTO:
1521 return (ifp->if_noproto);
1522 default:
1523 break;
1524 };
1525 return (0);
1526 }
1527
1528 /*
1529 * Increase an ifnet counter. Usually used for counters shared
1530 * between the stack and a driver, but function supports them all.
1531 */
1532 void
1533 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1534 {
1535
1536 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1537 switch (cnt) {
1538 case IFCOUNTER_IPACKETS:
1539 ifp->if_ipackets += inc;
1540 break;
1541 case IFCOUNTER_IERRORS:
1542 ifp->if_ierrors += inc;
1543 break;
1544 case IFCOUNTER_OPACKETS:
1545 ifp->if_opackets += inc;
1546 break;
1547 case IFCOUNTER_OERRORS:
1548 ifp->if_oerrors += inc;
1549 break;
1550 case IFCOUNTER_COLLISIONS:
1551 ifp->if_collisions += inc;
1552 break;
1553 case IFCOUNTER_IBYTES:
1554 ifp->if_ibytes += inc;
1555 break;
1556 case IFCOUNTER_OBYTES:
1557 ifp->if_obytes += inc;
1558 break;
1559 case IFCOUNTER_IMCASTS:
1560 ifp->if_imcasts += inc;
1561 break;
1562 case IFCOUNTER_OMCASTS:
1563 ifp->if_omcasts += inc;
1564 break;
1565 case IFCOUNTER_IQDROPS:
1566 ifp->if_iqdrops += inc;
1567 break;
1568 #ifdef _IFI_OQDROPS
1569 case IFCOUNTER_OQDROPS:
1570 ifp->if_oqdrops += inc;
1571 break;
1572 #endif
1573 case IFCOUNTER_NOPROTO:
1574 ifp->if_noproto += inc;
1575 break;
1576 default:
1577 break;
1578 };
1579 }
1580
1581 /*
1582 * Wrapper functions for struct ifnet address list locking macros. These are
1583 * used by kernel modules to avoid encoding programming interface or binary
1584 * interface assumptions that may be violated when kernel-internal locking
1585 * approaches change.
1586 */
1587 void
1588 if_addr_rlock(struct ifnet *ifp)
1589 {
1590
1591 IF_ADDR_RLOCK(ifp);
1592 }
1593
1594 void
1595 if_addr_runlock(struct ifnet *ifp)
1596 {
1597
1598 IF_ADDR_RUNLOCK(ifp);
1599 }
1600
1601 void
1602 if_maddr_rlock(struct ifnet *ifp)
1603 {
1604
1605 IF_ADDR_RLOCK(ifp);
1606 }
1607
1608 void
1609 if_maddr_runlock(struct ifnet *ifp)
1610 {
1611
1612 IF_ADDR_RUNLOCK(ifp);
1613 }
1614
1615 /*
1616 * Initialization, destruction and refcounting functions for ifaddrs.
1617 */
1618 void
1619 ifa_init(struct ifaddr *ifa)
1620 {
1621
1622 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1623 refcount_init(&ifa->ifa_refcnt, 1);
1624 ifa->if_data.ifi_datalen = sizeof(ifa->if_data);
1625 }
1626
1627 void
1628 ifa_ref(struct ifaddr *ifa)
1629 {
1630
1631 refcount_acquire(&ifa->ifa_refcnt);
1632 }
1633
1634 void
1635 ifa_free(struct ifaddr *ifa)
1636 {
1637
1638 if (refcount_release(&ifa->ifa_refcnt)) {
1639 mtx_destroy(&ifa->ifa_mtx);
1640 free(ifa, M_IFADDR);
1641 }
1642 }
1643
1644 int
1645 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1646 {
1647 int error = 0;
1648 struct rtentry *rt = NULL;
1649 struct rt_addrinfo info;
1650 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1651
1652 bzero(&info, sizeof(info));
1653 info.rti_ifp = V_loif;
1654 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1655 info.rti_info[RTAX_DST] = ia;
1656 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1657 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1658
1659 if (error == 0 && rt != NULL) {
1660 RT_LOCK(rt);
1661 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1662 ifa->ifa_ifp->if_type;
1663 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1664 ifa->ifa_ifp->if_index;
1665 RT_REMREF(rt);
1666 RT_UNLOCK(rt);
1667 } else if (error != 0)
1668 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1669
1670 return (error);
1671 }
1672
1673 int
1674 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1675 {
1676 int error = 0;
1677 struct rt_addrinfo info;
1678 struct sockaddr_dl null_sdl;
1679
1680 bzero(&null_sdl, sizeof(null_sdl));
1681 null_sdl.sdl_len = sizeof(null_sdl);
1682 null_sdl.sdl_family = AF_LINK;
1683 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1684 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1685 bzero(&info, sizeof(info));
1686 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1687 info.rti_info[RTAX_DST] = ia;
1688 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1689 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1690
1691 if (error != 0)
1692 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1693
1694 return (error);
1695 }
1696
1697 /*
1698 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1699 * structs used to represent other address families, it is necessary
1700 * to perform a different comparison.
1701 */
1702
1703 #define sa_equal(a1, a2) \
1704 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1705
1706 #define sa_dl_equal(a1, a2) \
1707 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1708 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1709 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1710 LLADDR((struct sockaddr_dl *)(a2)), \
1711 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1712
1713 /*
1714 * Locate an interface based on a complete address.
1715 */
1716 /*ARGSUSED*/
1717 static struct ifaddr *
1718 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1719 {
1720 struct ifnet *ifp;
1721 struct ifaddr *ifa;
1722
1723 IFNET_RLOCK_NOSLEEP();
1724 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1725 IF_ADDR_RLOCK(ifp);
1726 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1727 if (ifa->ifa_addr->sa_family != addr->sa_family)
1728 continue;
1729 if (sa_equal(addr, ifa->ifa_addr)) {
1730 if (getref)
1731 ifa_ref(ifa);
1732 IF_ADDR_RUNLOCK(ifp);
1733 goto done;
1734 }
1735 /* IP6 doesn't have broadcast */
1736 if ((ifp->if_flags & IFF_BROADCAST) &&
1737 ifa->ifa_broadaddr &&
1738 ifa->ifa_broadaddr->sa_len != 0 &&
1739 sa_equal(ifa->ifa_broadaddr, addr)) {
1740 if (getref)
1741 ifa_ref(ifa);
1742 IF_ADDR_RUNLOCK(ifp);
1743 goto done;
1744 }
1745 }
1746 IF_ADDR_RUNLOCK(ifp);
1747 }
1748 ifa = NULL;
1749 done:
1750 IFNET_RUNLOCK_NOSLEEP();
1751 return (ifa);
1752 }
1753
1754 struct ifaddr *
1755 ifa_ifwithaddr(struct sockaddr *addr)
1756 {
1757
1758 return (ifa_ifwithaddr_internal(addr, 1));
1759 }
1760
1761 int
1762 ifa_ifwithaddr_check(struct sockaddr *addr)
1763 {
1764
1765 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1766 }
1767
1768 /*
1769 * Locate an interface based on the broadcast address.
1770 */
1771 /* ARGSUSED */
1772 struct ifaddr *
1773 ifa_ifwithbroadaddr(struct sockaddr *addr)
1774 {
1775 struct ifnet *ifp;
1776 struct ifaddr *ifa;
1777
1778 IFNET_RLOCK_NOSLEEP();
1779 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1780 IF_ADDR_RLOCK(ifp);
1781 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1782 if (ifa->ifa_addr->sa_family != addr->sa_family)
1783 continue;
1784 if ((ifp->if_flags & IFF_BROADCAST) &&
1785 ifa->ifa_broadaddr &&
1786 ifa->ifa_broadaddr->sa_len != 0 &&
1787 sa_equal(ifa->ifa_broadaddr, addr)) {
1788 ifa_ref(ifa);
1789 IF_ADDR_RUNLOCK(ifp);
1790 goto done;
1791 }
1792 }
1793 IF_ADDR_RUNLOCK(ifp);
1794 }
1795 ifa = NULL;
1796 done:
1797 IFNET_RUNLOCK_NOSLEEP();
1798 return (ifa);
1799 }
1800
1801 /*
1802 * Locate the point to point interface with a given destination address.
1803 */
1804 /*ARGSUSED*/
1805 struct ifaddr *
1806 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
1807 {
1808 struct ifnet *ifp;
1809 struct ifaddr *ifa;
1810
1811 IFNET_RLOCK_NOSLEEP();
1812 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1813 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1814 continue;
1815 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1816 continue;
1817 IF_ADDR_RLOCK(ifp);
1818 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1819 if (ifa->ifa_addr->sa_family != addr->sa_family)
1820 continue;
1821 if (ifa->ifa_dstaddr != NULL &&
1822 sa_equal(addr, ifa->ifa_dstaddr)) {
1823 ifa_ref(ifa);
1824 IF_ADDR_RUNLOCK(ifp);
1825 goto done;
1826 }
1827 }
1828 IF_ADDR_RUNLOCK(ifp);
1829 }
1830 ifa = NULL;
1831 done:
1832 IFNET_RUNLOCK_NOSLEEP();
1833 return (ifa);
1834 }
1835
1836 struct ifaddr *
1837 ifa_ifwithdstaddr(struct sockaddr *addr)
1838 {
1839
1840 return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
1841 }
1842
1843 /*
1844 * Find an interface on a specific network. If many, choice
1845 * is most specific found.
1846 */
1847 struct ifaddr *
1848 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
1849 {
1850 struct ifnet *ifp;
1851 struct ifaddr *ifa;
1852 struct ifaddr *ifa_maybe = NULL;
1853 u_int af = addr->sa_family;
1854 char *addr_data = addr->sa_data, *cplim;
1855
1856 /*
1857 * AF_LINK addresses can be looked up directly by their index number,
1858 * so do that if we can.
1859 */
1860 if (af == AF_LINK) {
1861 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1862 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1863 return (ifaddr_byindex(sdl->sdl_index));
1864 }
1865
1866 /*
1867 * Scan though each interface, looking for ones that have addresses
1868 * in this address family and the requested fib. Maintain a reference
1869 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1870 * kept it stable when we move onto the next interface.
1871 */
1872 IFNET_RLOCK_NOSLEEP();
1873 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1874 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1875 continue;
1876 IF_ADDR_RLOCK(ifp);
1877 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1878 char *cp, *cp2, *cp3;
1879
1880 if (ifa->ifa_addr->sa_family != af)
1881 next: continue;
1882 if (af == AF_INET &&
1883 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1884 /*
1885 * This is a bit broken as it doesn't
1886 * take into account that the remote end may
1887 * be a single node in the network we are
1888 * looking for.
1889 * The trouble is that we don't know the
1890 * netmask for the remote end.
1891 */
1892 if (ifa->ifa_dstaddr != NULL &&
1893 sa_equal(addr, ifa->ifa_dstaddr)) {
1894 ifa_ref(ifa);
1895 IF_ADDR_RUNLOCK(ifp);
1896 goto done;
1897 }
1898 } else {
1899 /*
1900 * if we have a special address handler,
1901 * then use it instead of the generic one.
1902 */
1903 if (ifa->ifa_claim_addr) {
1904 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1905 ifa_ref(ifa);
1906 IF_ADDR_RUNLOCK(ifp);
1907 goto done;
1908 }
1909 continue;
1910 }
1911
1912 /*
1913 * Scan all the bits in the ifa's address.
1914 * If a bit dissagrees with what we are
1915 * looking for, mask it with the netmask
1916 * to see if it really matters.
1917 * (A byte at a time)
1918 */
1919 if (ifa->ifa_netmask == 0)
1920 continue;
1921 cp = addr_data;
1922 cp2 = ifa->ifa_addr->sa_data;
1923 cp3 = ifa->ifa_netmask->sa_data;
1924 cplim = ifa->ifa_netmask->sa_len
1925 + (char *)ifa->ifa_netmask;
1926 while (cp3 < cplim)
1927 if ((*cp++ ^ *cp2++) & *cp3++)
1928 goto next; /* next address! */
1929 /*
1930 * If the netmask of what we just found
1931 * is more specific than what we had before
1932 * (if we had one), or if the virtual status
1933 * of new prefix is better than of the old one,
1934 * then remember the new one before continuing
1935 * to search for an even better one.
1936 */
1937 if (ifa_maybe == NULL ||
1938 ifa_preferred(ifa_maybe, ifa) ||
1939 rn_refines((caddr_t)ifa->ifa_netmask,
1940 (caddr_t)ifa_maybe->ifa_netmask)) {
1941 if (ifa_maybe != NULL)
1942 ifa_free(ifa_maybe);
1943 ifa_maybe = ifa;
1944 ifa_ref(ifa_maybe);
1945 }
1946 }
1947 }
1948 IF_ADDR_RUNLOCK(ifp);
1949 }
1950 ifa = ifa_maybe;
1951 ifa_maybe = NULL;
1952 done:
1953 IFNET_RUNLOCK_NOSLEEP();
1954 if (ifa_maybe != NULL)
1955 ifa_free(ifa_maybe);
1956 return (ifa);
1957 }
1958
1959 struct ifaddr *
1960 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1961 {
1962
1963 return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
1964 }
1965
1966 /*
1967 * Find an interface address specific to an interface best matching
1968 * a given address.
1969 */
1970 struct ifaddr *
1971 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1972 {
1973 struct ifaddr *ifa;
1974 char *cp, *cp2, *cp3;
1975 char *cplim;
1976 struct ifaddr *ifa_maybe = NULL;
1977 u_int af = addr->sa_family;
1978
1979 if (af >= AF_MAX)
1980 return (NULL);
1981 IF_ADDR_RLOCK(ifp);
1982 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1983 if (ifa->ifa_addr->sa_family != af)
1984 continue;
1985 if (ifa_maybe == NULL)
1986 ifa_maybe = ifa;
1987 if (ifa->ifa_netmask == 0) {
1988 if (sa_equal(addr, ifa->ifa_addr) ||
1989 (ifa->ifa_dstaddr &&
1990 sa_equal(addr, ifa->ifa_dstaddr)))
1991 goto done;
1992 continue;
1993 }
1994 if (ifp->if_flags & IFF_POINTOPOINT) {
1995 if (sa_equal(addr, ifa->ifa_dstaddr))
1996 goto done;
1997 } else {
1998 cp = addr->sa_data;
1999 cp2 = ifa->ifa_addr->sa_data;
2000 cp3 = ifa->ifa_netmask->sa_data;
2001 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2002 for (; cp3 < cplim; cp3++)
2003 if ((*cp++ ^ *cp2++) & *cp3)
2004 break;
2005 if (cp3 == cplim)
2006 goto done;
2007 }
2008 }
2009 ifa = ifa_maybe;
2010 done:
2011 if (ifa != NULL)
2012 ifa_ref(ifa);
2013 IF_ADDR_RUNLOCK(ifp);
2014 return (ifa);
2015 }
2016
2017 /*
2018 * See whether new ifa is better than current one:
2019 * 1) A non-virtual one is preferred over virtual.
2020 * 2) A virtual in master state preferred over any other state.
2021 *
2022 * Used in several address selecting functions.
2023 */
2024 int
2025 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2026 {
2027
2028 return (cur->ifa_carp && (!next->ifa_carp ||
2029 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2030 }
2031
2032 #include <net/if_llatbl.h>
2033
2034 /*
2035 * Default action when installing a route with a Link Level gateway.
2036 * Lookup an appropriate real ifa to point to.
2037 * This should be moved to /sys/net/link.c eventually.
2038 */
2039 static void
2040 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2041 {
2042 struct ifaddr *ifa, *oifa;
2043 struct sockaddr *dst;
2044 struct ifnet *ifp;
2045
2046 RT_LOCK_ASSERT(rt);
2047
2048 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
2049 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
2050 return;
2051 ifa = ifaof_ifpforaddr(dst, ifp);
2052 if (ifa) {
2053 oifa = rt->rt_ifa;
2054 rt->rt_ifa = ifa;
2055 ifa_free(oifa);
2056 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2057 ifa->ifa_rtrequest(cmd, rt, info);
2058 }
2059 }
2060
2061 /*
2062 * Mark an interface down and notify protocols of
2063 * the transition.
2064 */
2065 static void
2066 if_unroute(struct ifnet *ifp, int flag, int fam)
2067 {
2068 struct ifaddr *ifa;
2069
2070 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2071
2072 ifp->if_flags &= ~flag;
2073 getmicrotime(&ifp->if_lastchange);
2074 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2075 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2076 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2077 ifp->if_qflush(ifp);
2078
2079 if (ifp->if_carp)
2080 (*carp_linkstate_p)(ifp);
2081 rt_ifmsg(ifp);
2082 }
2083
2084 /*
2085 * Mark an interface up and notify protocols of
2086 * the transition.
2087 */
2088 static void
2089 if_route(struct ifnet *ifp, int flag, int fam)
2090 {
2091 struct ifaddr *ifa;
2092
2093 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2094
2095 ifp->if_flags |= flag;
2096 getmicrotime(&ifp->if_lastchange);
2097 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2098 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2099 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2100 if (ifp->if_carp)
2101 (*carp_linkstate_p)(ifp);
2102 rt_ifmsg(ifp);
2103 #ifdef INET6
2104 in6_if_up(ifp);
2105 #endif
2106 }
2107
2108 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2109 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2110 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2111 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2112 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2113 int (*vlan_setcookie_p)(struct ifnet *, void *);
2114 void *(*vlan_cookie_p)(struct ifnet *);
2115
2116 /*
2117 * Handle a change in the interface link state. To avoid LORs
2118 * between driver lock and upper layer locks, as well as possible
2119 * recursions, we post event to taskqueue, and all job
2120 * is done in static do_link_state_change().
2121 */
2122 void
2123 if_link_state_change(struct ifnet *ifp, int link_state)
2124 {
2125 /* Return if state hasn't changed. */
2126 if (ifp->if_link_state == link_state)
2127 return;
2128
2129 ifp->if_link_state = link_state;
2130
2131 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2132 }
2133
2134 static void
2135 do_link_state_change(void *arg, int pending)
2136 {
2137 struct ifnet *ifp = (struct ifnet *)arg;
2138 int link_state = ifp->if_link_state;
2139 CURVNET_SET(ifp->if_vnet);
2140
2141 /* Notify that the link state has changed. */
2142 rt_ifmsg(ifp);
2143 if (ifp->if_vlantrunk != NULL)
2144 (*vlan_link_state_p)(ifp);
2145
2146 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2147 IFP2AC(ifp)->ac_netgraph != NULL)
2148 (*ng_ether_link_state_p)(ifp, link_state);
2149 if (ifp->if_carp)
2150 (*carp_linkstate_p)(ifp);
2151 if (ifp->if_bridge)
2152 (*bridge_linkstate_p)(ifp);
2153 if (ifp->if_lagg)
2154 (*lagg_linkstate_p)(ifp, link_state);
2155
2156 if (IS_DEFAULT_VNET(curvnet))
2157 devctl_notify("IFNET", ifp->if_xname,
2158 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2159 NULL);
2160 if (pending > 1)
2161 if_printf(ifp, "%d link states coalesced\n", pending);
2162 if (log_link_state_change)
2163 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2164 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2165 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2166 CURVNET_RESTORE();
2167 }
2168
2169 /*
2170 * Mark an interface down and notify protocols of
2171 * the transition.
2172 */
2173 void
2174 if_down(struct ifnet *ifp)
2175 {
2176
2177 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2178 }
2179
2180 /*
2181 * Mark an interface up and notify protocols of
2182 * the transition.
2183 */
2184 void
2185 if_up(struct ifnet *ifp)
2186 {
2187
2188 if_route(ifp, IFF_UP, AF_UNSPEC);
2189 }
2190
2191 /*
2192 * Flush an interface queue.
2193 */
2194 void
2195 if_qflush(struct ifnet *ifp)
2196 {
2197 struct mbuf *m, *n;
2198 struct ifaltq *ifq;
2199
2200 ifq = &ifp->if_snd;
2201 IFQ_LOCK(ifq);
2202 #ifdef ALTQ
2203 if (ALTQ_IS_ENABLED(ifq))
2204 ALTQ_PURGE(ifq);
2205 #endif
2206 n = ifq->ifq_head;
2207 while ((m = n) != 0) {
2208 n = m->m_nextpkt;
2209 m_freem(m);
2210 }
2211 ifq->ifq_head = 0;
2212 ifq->ifq_tail = 0;
2213 ifq->ifq_len = 0;
2214 IFQ_UNLOCK(ifq);
2215 }
2216
2217 /*
2218 * Map interface name to interface structure pointer, with or without
2219 * returning a reference.
2220 */
2221 struct ifnet *
2222 ifunit_ref(const char *name)
2223 {
2224 struct ifnet *ifp;
2225
2226 IFNET_RLOCK_NOSLEEP();
2227 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2228 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2229 !(ifp->if_flags & IFF_DYING))
2230 break;
2231 }
2232 if (ifp != NULL)
2233 if_ref(ifp);
2234 IFNET_RUNLOCK_NOSLEEP();
2235 return (ifp);
2236 }
2237
2238 struct ifnet *
2239 ifunit(const char *name)
2240 {
2241 struct ifnet *ifp;
2242
2243 IFNET_RLOCK_NOSLEEP();
2244 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2245 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2246 break;
2247 }
2248 IFNET_RUNLOCK_NOSLEEP();
2249 return (ifp);
2250 }
2251
2252 /*
2253 * Hardware specific interface ioctls.
2254 */
2255 static int
2256 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2257 {
2258 struct ifreq *ifr;
2259 struct ifstat *ifs;
2260 int error = 0;
2261 int new_flags, temp_flags;
2262 size_t namelen, onamelen;
2263 size_t descrlen;
2264 char *descrbuf, *odescrbuf;
2265 char new_name[IFNAMSIZ];
2266 struct ifaddr *ifa;
2267 struct sockaddr_dl *sdl;
2268
2269 ifr = (struct ifreq *)data;
2270 switch (cmd) {
2271 case SIOCGIFINDEX:
2272 ifr->ifr_index = ifp->if_index;
2273 break;
2274
2275 case SIOCGIFFLAGS:
2276 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2277 ifr->ifr_flags = temp_flags & 0xffff;
2278 ifr->ifr_flagshigh = temp_flags >> 16;
2279 break;
2280
2281 case SIOCGIFCAP:
2282 ifr->ifr_reqcap = ifp->if_capabilities;
2283 ifr->ifr_curcap = ifp->if_capenable;
2284 break;
2285
2286 #ifdef MAC
2287 case SIOCGIFMAC:
2288 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2289 break;
2290 #endif
2291
2292 case SIOCGIFMETRIC:
2293 ifr->ifr_metric = ifp->if_metric;
2294 break;
2295
2296 case SIOCGIFMTU:
2297 ifr->ifr_mtu = ifp->if_mtu;
2298 break;
2299
2300 case SIOCGIFPHYS:
2301 ifr->ifr_phys = ifp->if_physical;
2302 break;
2303
2304 case SIOCGIFDESCR:
2305 error = 0;
2306 sx_slock(&ifdescr_sx);
2307 if (ifp->if_description == NULL)
2308 error = ENOMSG;
2309 else {
2310 /* space for terminating nul */
2311 descrlen = strlen(ifp->if_description) + 1;
2312 if (ifr->ifr_buffer.length < descrlen)
2313 ifr->ifr_buffer.buffer = NULL;
2314 else
2315 error = copyout(ifp->if_description,
2316 ifr->ifr_buffer.buffer, descrlen);
2317 ifr->ifr_buffer.length = descrlen;
2318 }
2319 sx_sunlock(&ifdescr_sx);
2320 break;
2321
2322 case SIOCSIFDESCR:
2323 error = priv_check(td, PRIV_NET_SETIFDESCR);
2324 if (error)
2325 return (error);
2326
2327 /*
2328 * Copy only (length-1) bytes to make sure that
2329 * if_description is always nul terminated. The
2330 * length parameter is supposed to count the
2331 * terminating nul in.
2332 */
2333 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2334 return (ENAMETOOLONG);
2335 else if (ifr->ifr_buffer.length == 0)
2336 descrbuf = NULL;
2337 else {
2338 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2339 M_WAITOK | M_ZERO);
2340 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2341 ifr->ifr_buffer.length - 1);
2342 if (error) {
2343 free(descrbuf, M_IFDESCR);
2344 break;
2345 }
2346 }
2347
2348 sx_xlock(&ifdescr_sx);
2349 odescrbuf = ifp->if_description;
2350 ifp->if_description = descrbuf;
2351 sx_xunlock(&ifdescr_sx);
2352
2353 getmicrotime(&ifp->if_lastchange);
2354 free(odescrbuf, M_IFDESCR);
2355 break;
2356
2357 case SIOCGIFFIB:
2358 ifr->ifr_fib = ifp->if_fib;
2359 break;
2360
2361 case SIOCSIFFIB:
2362 error = priv_check(td, PRIV_NET_SETIFFIB);
2363 if (error)
2364 return (error);
2365 if (ifr->ifr_fib >= rt_numfibs)
2366 return (EINVAL);
2367
2368 ifp->if_fib = ifr->ifr_fib;
2369 break;
2370
2371 case SIOCSIFFLAGS:
2372 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2373 if (error)
2374 return (error);
2375 /*
2376 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2377 * check, so we don't need special handling here yet.
2378 */
2379 new_flags = (ifr->ifr_flags & 0xffff) |
2380 (ifr->ifr_flagshigh << 16);
2381 if (ifp->if_flags & IFF_SMART) {
2382 /* Smart drivers twiddle their own routes */
2383 } else if (ifp->if_flags & IFF_UP &&
2384 (new_flags & IFF_UP) == 0) {
2385 if_down(ifp);
2386 } else if (new_flags & IFF_UP &&
2387 (ifp->if_flags & IFF_UP) == 0) {
2388 if_up(ifp);
2389 }
2390 /* See if permanently promiscuous mode bit is about to flip */
2391 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2392 if (new_flags & IFF_PPROMISC)
2393 ifp->if_flags |= IFF_PROMISC;
2394 else if (ifp->if_pcount == 0)
2395 ifp->if_flags &= ~IFF_PROMISC;
2396 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2397 ifp->if_xname,
2398 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2399 }
2400 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2401 (new_flags &~ IFF_CANTCHANGE);
2402 if (ifp->if_ioctl) {
2403 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2404 }
2405 getmicrotime(&ifp->if_lastchange);
2406 break;
2407
2408 case SIOCSIFCAP:
2409 error = priv_check(td, PRIV_NET_SETIFCAP);
2410 if (error)
2411 return (error);
2412 if (ifp->if_ioctl == NULL)
2413 return (EOPNOTSUPP);
2414 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2415 return (EINVAL);
2416 error = (*ifp->if_ioctl)(ifp, cmd, data);
2417 if (error == 0)
2418 getmicrotime(&ifp->if_lastchange);
2419 break;
2420
2421 #ifdef MAC
2422 case SIOCSIFMAC:
2423 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2424 break;
2425 #endif
2426
2427 case SIOCSIFNAME:
2428 error = priv_check(td, PRIV_NET_SETIFNAME);
2429 if (error)
2430 return (error);
2431 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2432 if (error != 0)
2433 return (error);
2434 if (new_name[0] == '\0')
2435 return (EINVAL);
2436 if (ifunit(new_name) != NULL)
2437 return (EEXIST);
2438
2439 /*
2440 * XXX: Locking. Nothing else seems to lock if_flags,
2441 * and there are numerous other races with the
2442 * ifunit() checks not being atomic with namespace
2443 * changes (renames, vmoves, if_attach, etc).
2444 */
2445 ifp->if_flags |= IFF_RENAMING;
2446
2447 /* Announce the departure of the interface. */
2448 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2449 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2450
2451 log(LOG_INFO, "%s: changing name to '%s'\n",
2452 ifp->if_xname, new_name);
2453
2454 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2455 ifa = ifp->if_addr;
2456 IFA_LOCK(ifa);
2457 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2458 namelen = strlen(new_name);
2459 onamelen = sdl->sdl_nlen;
2460 /*
2461 * Move the address if needed. This is safe because we
2462 * allocate space for a name of length IFNAMSIZ when we
2463 * create this in if_attach().
2464 */
2465 if (namelen != onamelen) {
2466 bcopy(sdl->sdl_data + onamelen,
2467 sdl->sdl_data + namelen, sdl->sdl_alen);
2468 }
2469 bcopy(new_name, sdl->sdl_data, namelen);
2470 sdl->sdl_nlen = namelen;
2471 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2472 bzero(sdl->sdl_data, onamelen);
2473 while (namelen != 0)
2474 sdl->sdl_data[--namelen] = 0xff;
2475 IFA_UNLOCK(ifa);
2476
2477 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2478 /* Announce the return of the interface. */
2479 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2480
2481 ifp->if_flags &= ~IFF_RENAMING;
2482 break;
2483
2484 #ifdef VIMAGE
2485 case SIOCSIFVNET:
2486 error = priv_check(td, PRIV_NET_SETIFVNET);
2487 if (error)
2488 return (error);
2489 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2490 break;
2491 #endif
2492
2493 case SIOCSIFMETRIC:
2494 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2495 if (error)
2496 return (error);
2497 ifp->if_metric = ifr->ifr_metric;
2498 getmicrotime(&ifp->if_lastchange);
2499 break;
2500
2501 case SIOCSIFPHYS:
2502 error = priv_check(td, PRIV_NET_SETIFPHYS);
2503 if (error)
2504 return (error);
2505 if (ifp->if_ioctl == NULL)
2506 return (EOPNOTSUPP);
2507 error = (*ifp->if_ioctl)(ifp, cmd, data);
2508 if (error == 0)
2509 getmicrotime(&ifp->if_lastchange);
2510 break;
2511
2512 case SIOCSIFMTU:
2513 {
2514 u_long oldmtu = ifp->if_mtu;
2515
2516 error = priv_check(td, PRIV_NET_SETIFMTU);
2517 if (error)
2518 return (error);
2519 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2520 return (EINVAL);
2521 if (ifp->if_ioctl == NULL)
2522 return (EOPNOTSUPP);
2523 error = (*ifp->if_ioctl)(ifp, cmd, data);
2524 if (error == 0) {
2525 getmicrotime(&ifp->if_lastchange);
2526 rt_ifmsg(ifp);
2527 }
2528 /*
2529 * If the link MTU changed, do network layer specific procedure.
2530 */
2531 if (ifp->if_mtu != oldmtu) {
2532 #ifdef INET6
2533 nd6_setmtu(ifp);
2534 #endif
2535 }
2536 break;
2537 }
2538
2539 case SIOCADDMULTI:
2540 case SIOCDELMULTI:
2541 if (cmd == SIOCADDMULTI)
2542 error = priv_check(td, PRIV_NET_ADDMULTI);
2543 else
2544 error = priv_check(td, PRIV_NET_DELMULTI);
2545 if (error)
2546 return (error);
2547
2548 /* Don't allow group membership on non-multicast interfaces. */
2549 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2550 return (EOPNOTSUPP);
2551
2552 /* Don't let users screw up protocols' entries. */
2553 if (ifr->ifr_addr.sa_family != AF_LINK)
2554 return (EINVAL);
2555
2556 if (cmd == SIOCADDMULTI) {
2557 struct ifmultiaddr *ifma;
2558
2559 /*
2560 * Userland is only permitted to join groups once
2561 * via the if_addmulti() KPI, because it cannot hold
2562 * struct ifmultiaddr * between calls. It may also
2563 * lose a race while we check if the membership
2564 * already exists.
2565 */
2566 IF_ADDR_RLOCK(ifp);
2567 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2568 IF_ADDR_RUNLOCK(ifp);
2569 if (ifma != NULL)
2570 error = EADDRINUSE;
2571 else
2572 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2573 } else {
2574 error = if_delmulti(ifp, &ifr->ifr_addr);
2575 }
2576 if (error == 0)
2577 getmicrotime(&ifp->if_lastchange);
2578 break;
2579
2580 case SIOCSIFPHYADDR:
2581 case SIOCDIFPHYADDR:
2582 #ifdef INET6
2583 case SIOCSIFPHYADDR_IN6:
2584 #endif
2585 case SIOCSLIFPHYADDR:
2586 case SIOCSIFMEDIA:
2587 case SIOCSIFGENERIC:
2588 error = priv_check(td, PRIV_NET_HWIOCTL);
2589 if (error)
2590 return (error);
2591 if (ifp->if_ioctl == NULL)
2592 return (EOPNOTSUPP);
2593 error = (*ifp->if_ioctl)(ifp, cmd, data);
2594 if (error == 0)
2595 getmicrotime(&ifp->if_lastchange);
2596 break;
2597
2598 case SIOCGIFSTATUS:
2599 ifs = (struct ifstat *)data;
2600 ifs->ascii[0] = '\0';
2601
2602 case SIOCGIFPSRCADDR:
2603 case SIOCGIFPDSTADDR:
2604 case SIOCGLIFPHYADDR:
2605 case SIOCGIFMEDIA:
2606 case SIOCGIFXMEDIA:
2607 case SIOCGIFGENERIC:
2608 if (ifp->if_ioctl == NULL)
2609 return (EOPNOTSUPP);
2610 error = (*ifp->if_ioctl)(ifp, cmd, data);
2611 break;
2612
2613 case SIOCSIFLLADDR:
2614 error = priv_check(td, PRIV_NET_SETLLADDR);
2615 if (error)
2616 return (error);
2617 error = if_setlladdr(ifp,
2618 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2619 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2620 break;
2621
2622 case SIOCAIFGROUP:
2623 {
2624 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2625
2626 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2627 if (error)
2628 return (error);
2629 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2630 return (error);
2631 break;
2632 }
2633
2634 case SIOCGIFGROUP:
2635 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2636 return (error);
2637 break;
2638
2639 case SIOCDIFGROUP:
2640 {
2641 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2642
2643 error = priv_check(td, PRIV_NET_DELIFGROUP);
2644 if (error)
2645 return (error);
2646 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2647 return (error);
2648 break;
2649 }
2650
2651 default:
2652 error = ENOIOCTL;
2653 break;
2654 }
2655 return (error);
2656 }
2657
2658 #ifdef COMPAT_FREEBSD32
2659 struct ifconf32 {
2660 int32_t ifc_len;
2661 union {
2662 uint32_t ifcu_buf;
2663 uint32_t ifcu_req;
2664 } ifc_ifcu;
2665 };
2666 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2667 #endif
2668
2669 /*
2670 * Interface ioctls.
2671 */
2672 int
2673 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2674 {
2675 struct ifnet *ifp;
2676 struct ifreq *ifr;
2677 int error;
2678 int oif_flags;
2679
2680 CURVNET_SET(so->so_vnet);
2681 switch (cmd) {
2682 case SIOCGIFCONF:
2683 case OSIOCGIFCONF:
2684 error = ifconf(cmd, data);
2685 CURVNET_RESTORE();
2686 return (error);
2687
2688 #ifdef COMPAT_FREEBSD32
2689 case SIOCGIFCONF32:
2690 {
2691 struct ifconf32 *ifc32;
2692 struct ifconf ifc;
2693
2694 ifc32 = (struct ifconf32 *)data;
2695 ifc.ifc_len = ifc32->ifc_len;
2696 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2697
2698 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2699 CURVNET_RESTORE();
2700 if (error == 0)
2701 ifc32->ifc_len = ifc.ifc_len;
2702 return (error);
2703 }
2704 #endif
2705 }
2706 ifr = (struct ifreq *)data;
2707
2708 switch (cmd) {
2709 #ifdef VIMAGE
2710 case SIOCSIFRVNET:
2711 error = priv_check(td, PRIV_NET_SETIFVNET);
2712 if (error == 0)
2713 error = if_vmove_reclaim(td, ifr->ifr_name,
2714 ifr->ifr_jid);
2715 CURVNET_RESTORE();
2716 return (error);
2717 #endif
2718 case SIOCIFCREATE:
2719 case SIOCIFCREATE2:
2720 error = priv_check(td, PRIV_NET_IFCREATE);
2721 if (error == 0)
2722 error = if_clone_create(ifr->ifr_name,
2723 sizeof(ifr->ifr_name),
2724 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2725 CURVNET_RESTORE();
2726 return (error);
2727 case SIOCIFDESTROY:
2728 error = priv_check(td, PRIV_NET_IFDESTROY);
2729 if (error == 0)
2730 error = if_clone_destroy(ifr->ifr_name);
2731 CURVNET_RESTORE();
2732 return (error);
2733
2734 case SIOCIFGCLONERS:
2735 error = if_clone_list((struct if_clonereq *)data);
2736 CURVNET_RESTORE();
2737 return (error);
2738 case SIOCGIFGMEMB:
2739 error = if_getgroupmembers((struct ifgroupreq *)data);
2740 CURVNET_RESTORE();
2741 return (error);
2742 #if defined(INET) || defined(INET6)
2743 case SIOCSVH:
2744 case SIOCGVH:
2745 if (carp_ioctl_p == NULL)
2746 error = EPROTONOSUPPORT;
2747 else
2748 error = (*carp_ioctl_p)(ifr, cmd, td);
2749 CURVNET_RESTORE();
2750 return (error);
2751 #endif
2752 }
2753
2754 ifp = ifunit_ref(ifr->ifr_name);
2755 if (ifp == NULL) {
2756 CURVNET_RESTORE();
2757 return (ENXIO);
2758 }
2759
2760 error = ifhwioctl(cmd, ifp, data, td);
2761 if (error != ENOIOCTL) {
2762 if_rele(ifp);
2763 CURVNET_RESTORE();
2764 return (error);
2765 }
2766
2767 oif_flags = ifp->if_flags;
2768 if (so->so_proto == NULL) {
2769 if_rele(ifp);
2770 CURVNET_RESTORE();
2771 return (EOPNOTSUPP);
2772 }
2773
2774 /*
2775 * Pass the request on to the socket control method, and if the
2776 * latter returns EOPNOTSUPP, directly to the interface.
2777 *
2778 * Make an exception for the legacy SIOCSIF* requests. Drivers
2779 * trust SIOCSIFADDR et al to come from an already privileged
2780 * layer, and do not perform any credentials checks or input
2781 * validation.
2782 */
2783 #ifndef COMPAT_43
2784 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2785 data,
2786 ifp, td));
2787 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2788 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2789 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2790 error = (*ifp->if_ioctl)(ifp, cmd, data);
2791 #else
2792 {
2793 u_long ocmd = cmd;
2794
2795 switch (cmd) {
2796
2797 case SIOCSIFDSTADDR:
2798 case SIOCSIFADDR:
2799 case SIOCSIFBRDADDR:
2800 case SIOCSIFNETMASK:
2801 #if BYTE_ORDER != BIG_ENDIAN
2802 if (ifr->ifr_addr.sa_family == 0 &&
2803 ifr->ifr_addr.sa_len < 16) {
2804 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2805 ifr->ifr_addr.sa_len = 16;
2806 }
2807 #else
2808 if (ifr->ifr_addr.sa_len == 0)
2809 ifr->ifr_addr.sa_len = 16;
2810 #endif
2811 break;
2812
2813 case OSIOCGIFADDR:
2814 cmd = SIOCGIFADDR;
2815 break;
2816
2817 case OSIOCGIFDSTADDR:
2818 cmd = SIOCGIFDSTADDR;
2819 break;
2820
2821 case OSIOCGIFBRDADDR:
2822 cmd = SIOCGIFBRDADDR;
2823 break;
2824
2825 case OSIOCGIFNETMASK:
2826 cmd = SIOCGIFNETMASK;
2827 }
2828 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2829 cmd,
2830 data,
2831 ifp, td));
2832 if (error == EOPNOTSUPP && ifp != NULL &&
2833 ifp->if_ioctl != NULL &&
2834 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2835 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2836 error = (*ifp->if_ioctl)(ifp, cmd, data);
2837 switch (ocmd) {
2838
2839 case OSIOCGIFADDR:
2840 case OSIOCGIFDSTADDR:
2841 case OSIOCGIFBRDADDR:
2842 case OSIOCGIFNETMASK:
2843 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2844
2845 }
2846 }
2847 #endif /* COMPAT_43 */
2848
2849 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2850 #ifdef INET6
2851 if (ifp->if_flags & IFF_UP)
2852 in6_if_up(ifp);
2853 #endif
2854 }
2855 if_rele(ifp);
2856 CURVNET_RESTORE();
2857 return (error);
2858 }
2859
2860 /*
2861 * The code common to handling reference counted flags,
2862 * e.g., in ifpromisc() and if_allmulti().
2863 * The "pflag" argument can specify a permanent mode flag to check,
2864 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2865 *
2866 * Only to be used on stack-owned flags, not driver-owned flags.
2867 */
2868 static int
2869 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2870 {
2871 struct ifreq ifr;
2872 int error;
2873 int oldflags, oldcount;
2874
2875 /* Sanity checks to catch programming errors */
2876 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2877 ("%s: setting driver-owned flag %d", __func__, flag));
2878
2879 if (onswitch)
2880 KASSERT(*refcount >= 0,
2881 ("%s: increment negative refcount %d for flag %d",
2882 __func__, *refcount, flag));
2883 else
2884 KASSERT(*refcount > 0,
2885 ("%s: decrement non-positive refcount %d for flag %d",
2886 __func__, *refcount, flag));
2887
2888 /* In case this mode is permanent, just touch refcount */
2889 if (ifp->if_flags & pflag) {
2890 *refcount += onswitch ? 1 : -1;
2891 return (0);
2892 }
2893
2894 /* Save ifnet parameters for if_ioctl() may fail */
2895 oldcount = *refcount;
2896 oldflags = ifp->if_flags;
2897
2898 /*
2899 * See if we aren't the only and touching refcount is enough.
2900 * Actually toggle interface flag if we are the first or last.
2901 */
2902 if (onswitch) {
2903 if ((*refcount)++)
2904 return (0);
2905 ifp->if_flags |= flag;
2906 } else {
2907 if (--(*refcount))
2908 return (0);
2909 ifp->if_flags &= ~flag;
2910 }
2911
2912 /* Call down the driver since we've changed interface flags */
2913 if (ifp->if_ioctl == NULL) {
2914 error = EOPNOTSUPP;
2915 goto recover;
2916 }
2917 ifr.ifr_flags = ifp->if_flags & 0xffff;
2918 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2919 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2920 if (error)
2921 goto recover;
2922 /* Notify userland that interface flags have changed */
2923 rt_ifmsg(ifp);
2924 return (0);
2925
2926 recover:
2927 /* Recover after driver error */
2928 *refcount = oldcount;
2929 ifp->if_flags = oldflags;
2930 return (error);
2931 }
2932
2933 /*
2934 * Set/clear promiscuous mode on interface ifp based on the truth value
2935 * of pswitch. The calls are reference counted so that only the first
2936 * "on" request actually has an effect, as does the final "off" request.
2937 * Results are undefined if the "off" and "on" requests are not matched.
2938 */
2939 int
2940 ifpromisc(struct ifnet *ifp, int pswitch)
2941 {
2942 int error;
2943 int oldflags = ifp->if_flags;
2944
2945 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2946 &ifp->if_pcount, pswitch);
2947 /* If promiscuous mode status has changed, log a message */
2948 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2949 log(LOG_INFO, "%s: promiscuous mode %s\n",
2950 ifp->if_xname,
2951 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2952 return (error);
2953 }
2954
2955 /*
2956 * Return interface configuration
2957 * of system. List may be used
2958 * in later ioctl's (above) to get
2959 * other information.
2960 */
2961 /*ARGSUSED*/
2962 static int
2963 ifconf(u_long cmd, caddr_t data)
2964 {
2965 struct ifconf *ifc = (struct ifconf *)data;
2966 struct ifnet *ifp;
2967 struct ifaddr *ifa;
2968 struct ifreq ifr;
2969 struct sbuf *sb;
2970 int error, full = 0, valid_len, max_len;
2971
2972 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2973 max_len = MAXPHYS - 1;
2974
2975 /* Prevent hostile input from being able to crash the system */
2976 if (ifc->ifc_len <= 0)
2977 return (EINVAL);
2978
2979 again:
2980 if (ifc->ifc_len <= max_len) {
2981 max_len = ifc->ifc_len;
2982 full = 1;
2983 }
2984 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2985 max_len = 0;
2986 valid_len = 0;
2987
2988 IFNET_RLOCK();
2989 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2990 int addrs;
2991
2992 /*
2993 * Zero the ifr_name buffer to make sure we don't
2994 * disclose the contents of the stack.
2995 */
2996 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2997
2998 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2999 >= sizeof(ifr.ifr_name)) {
3000 sbuf_delete(sb);
3001 IFNET_RUNLOCK();
3002 return (ENAMETOOLONG);
3003 }
3004
3005 addrs = 0;
3006 IF_ADDR_RLOCK(ifp);
3007 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3008 struct sockaddr *sa = ifa->ifa_addr;
3009
3010 if (prison_if(curthread->td_ucred, sa) != 0)
3011 continue;
3012 addrs++;
3013 #ifdef COMPAT_43
3014 if (cmd == OSIOCGIFCONF) {
3015 struct osockaddr *osa =
3016 (struct osockaddr *)&ifr.ifr_addr;
3017 ifr.ifr_addr = *sa;
3018 osa->sa_family = sa->sa_family;
3019 sbuf_bcat(sb, &ifr, sizeof(ifr));
3020 max_len += sizeof(ifr);
3021 } else
3022 #endif
3023 if (sa->sa_len <= sizeof(*sa)) {
3024 ifr.ifr_addr = *sa;
3025 sbuf_bcat(sb, &ifr, sizeof(ifr));
3026 max_len += sizeof(ifr);
3027 } else {
3028 sbuf_bcat(sb, &ifr,
3029 offsetof(struct ifreq, ifr_addr));
3030 max_len += offsetof(struct ifreq, ifr_addr);
3031 sbuf_bcat(sb, sa, sa->sa_len);
3032 max_len += sa->sa_len;
3033 }
3034
3035 if (sbuf_error(sb) == 0)
3036 valid_len = sbuf_len(sb);
3037 }
3038 IF_ADDR_RUNLOCK(ifp);
3039 if (addrs == 0) {
3040 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3041 sbuf_bcat(sb, &ifr, sizeof(ifr));
3042 max_len += sizeof(ifr);
3043
3044 if (sbuf_error(sb) == 0)
3045 valid_len = sbuf_len(sb);
3046 }
3047 }
3048 IFNET_RUNLOCK();
3049
3050 /*
3051 * If we didn't allocate enough space (uncommon), try again. If
3052 * we have already allocated as much space as we are allowed,
3053 * return what we've got.
3054 */
3055 if (valid_len != max_len && !full) {
3056 sbuf_delete(sb);
3057 goto again;
3058 }
3059
3060 ifc->ifc_len = valid_len;
3061 sbuf_finish(sb);
3062 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3063 sbuf_delete(sb);
3064 return (error);
3065 }
3066
3067 /*
3068 * Just like ifpromisc(), but for all-multicast-reception mode.
3069 */
3070 int
3071 if_allmulti(struct ifnet *ifp, int onswitch)
3072 {
3073
3074 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3075 }
3076
3077 struct ifmultiaddr *
3078 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
3079 {
3080 struct ifmultiaddr *ifma;
3081
3082 IF_ADDR_LOCK_ASSERT(ifp);
3083
3084 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3085 if (sa->sa_family == AF_LINK) {
3086 if (sa_dl_equal(ifma->ifma_addr, sa))
3087 break;
3088 } else {
3089 if (sa_equal(ifma->ifma_addr, sa))
3090 break;
3091 }
3092 }
3093
3094 return ifma;
3095 }
3096
3097 /*
3098 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3099 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3100 * the ifnet multicast address list here, so the caller must do that and
3101 * other setup work (such as notifying the device driver). The reference
3102 * count is initialized to 1.
3103 */
3104 static struct ifmultiaddr *
3105 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3106 int mflags)
3107 {
3108 struct ifmultiaddr *ifma;
3109 struct sockaddr *dupsa;
3110
3111 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3112 M_ZERO);
3113 if (ifma == NULL)
3114 return (NULL);
3115
3116 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3117 if (dupsa == NULL) {
3118 free(ifma, M_IFMADDR);
3119 return (NULL);
3120 }
3121 bcopy(sa, dupsa, sa->sa_len);
3122 ifma->ifma_addr = dupsa;
3123
3124 ifma->ifma_ifp = ifp;
3125 ifma->ifma_refcount = 1;
3126 ifma->ifma_protospec = NULL;
3127
3128 if (llsa == NULL) {
3129 ifma->ifma_lladdr = NULL;
3130 return (ifma);
3131 }
3132
3133 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3134 if (dupsa == NULL) {
3135 free(ifma->ifma_addr, M_IFMADDR);
3136 free(ifma, M_IFMADDR);
3137 return (NULL);
3138 }
3139 bcopy(llsa, dupsa, llsa->sa_len);
3140 ifma->ifma_lladdr = dupsa;
3141
3142 return (ifma);
3143 }
3144
3145 /*
3146 * if_freemulti: free ifmultiaddr structure and possibly attached related
3147 * addresses. The caller is responsible for implementing reference
3148 * counting, notifying the driver, handling routing messages, and releasing
3149 * any dependent link layer state.
3150 */
3151 static void
3152 if_freemulti(struct ifmultiaddr *ifma)
3153 {
3154
3155 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3156 ifma->ifma_refcount));
3157 KASSERT(ifma->ifma_protospec == NULL,
3158 ("if_freemulti: protospec not NULL"));
3159
3160 if (ifma->ifma_lladdr != NULL)
3161 free(ifma->ifma_lladdr, M_IFMADDR);
3162 free(ifma->ifma_addr, M_IFMADDR);
3163 free(ifma, M_IFMADDR);
3164 }
3165
3166 /*
3167 * Register an additional multicast address with a network interface.
3168 *
3169 * - If the address is already present, bump the reference count on the
3170 * address and return.
3171 * - If the address is not link-layer, look up a link layer address.
3172 * - Allocate address structures for one or both addresses, and attach to the
3173 * multicast address list on the interface. If automatically adding a link
3174 * layer address, the protocol address will own a reference to the link
3175 * layer address, to be freed when it is freed.
3176 * - Notify the network device driver of an addition to the multicast address
3177 * list.
3178 *
3179 * 'sa' points to caller-owned memory with the desired multicast address.
3180 *
3181 * 'retifma' will be used to return a pointer to the resulting multicast
3182 * address reference, if desired.
3183 */
3184 int
3185 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3186 struct ifmultiaddr **retifma)
3187 {
3188 struct ifmultiaddr *ifma, *ll_ifma;
3189 struct sockaddr *llsa;
3190 int error;
3191
3192 /*
3193 * If the address is already present, return a new reference to it;
3194 * otherwise, allocate storage and set up a new address.
3195 */
3196 IF_ADDR_WLOCK(ifp);
3197 ifma = if_findmulti(ifp, sa);
3198 if (ifma != NULL) {
3199 ifma->ifma_refcount++;
3200 if (retifma != NULL)
3201 *retifma = ifma;
3202 IF_ADDR_WUNLOCK(ifp);
3203 return (0);
3204 }
3205
3206 /*
3207 * The address isn't already present; resolve the protocol address
3208 * into a link layer address, and then look that up, bump its
3209 * refcount or allocate an ifma for that also. If 'llsa' was
3210 * returned, we will need to free it later.
3211 */
3212 llsa = NULL;
3213 ll_ifma = NULL;
3214 if (ifp->if_resolvemulti != NULL) {
3215 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3216 if (error)
3217 goto unlock_out;
3218 }
3219
3220 /*
3221 * Allocate the new address. Don't hook it up yet, as we may also
3222 * need to allocate a link layer multicast address.
3223 */
3224 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3225 if (ifma == NULL) {
3226 error = ENOMEM;
3227 goto free_llsa_out;
3228 }
3229
3230 /*
3231 * If a link layer address is found, we'll need to see if it's
3232 * already present in the address list, or allocate is as well.
3233 * When this block finishes, the link layer address will be on the
3234 * list.
3235 */
3236 if (llsa != NULL) {
3237 ll_ifma = if_findmulti(ifp, llsa);
3238 if (ll_ifma == NULL) {
3239 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3240 if (ll_ifma == NULL) {
3241 --ifma->ifma_refcount;
3242 if_freemulti(ifma);
3243 error = ENOMEM;
3244 goto free_llsa_out;
3245 }
3246 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3247 ifma_link);
3248 } else
3249 ll_ifma->ifma_refcount++;
3250 ifma->ifma_llifma = ll_ifma;
3251 }
3252
3253 /*
3254 * We now have a new multicast address, ifma, and possibly a new or
3255 * referenced link layer address. Add the primary address to the
3256 * ifnet address list.
3257 */
3258 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3259
3260 if (retifma != NULL)
3261 *retifma = ifma;
3262
3263 /*
3264 * Must generate the message while holding the lock so that 'ifma'
3265 * pointer is still valid.
3266 */
3267 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3268 IF_ADDR_WUNLOCK(ifp);
3269
3270 /*
3271 * We are certain we have added something, so call down to the
3272 * interface to let them know about it.
3273 */
3274 if (ifp->if_ioctl != NULL) {
3275 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3276 }
3277
3278 if (llsa != NULL)
3279 free(llsa, M_IFMADDR);
3280
3281 return (0);
3282
3283 free_llsa_out:
3284 if (llsa != NULL)
3285 free(llsa, M_IFMADDR);
3286
3287 unlock_out:
3288 IF_ADDR_WUNLOCK(ifp);
3289 return (error);
3290 }
3291
3292 /*
3293 * Delete a multicast group membership by network-layer group address.
3294 *
3295 * Returns ENOENT if the entry could not be found. If ifp no longer
3296 * exists, results are undefined. This entry point should only be used
3297 * from subsystems which do appropriate locking to hold ifp for the
3298 * duration of the call.
3299 * Network-layer protocol domains must use if_delmulti_ifma().
3300 */
3301 int
3302 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3303 {
3304 struct ifmultiaddr *ifma;
3305 int lastref;
3306 #ifdef INVARIANTS
3307 struct ifnet *oifp;
3308
3309 IFNET_RLOCK_NOSLEEP();
3310 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3311 if (ifp == oifp)
3312 break;
3313 if (ifp != oifp)
3314 ifp = NULL;
3315 IFNET_RUNLOCK_NOSLEEP();
3316
3317 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3318 #endif
3319 if (ifp == NULL)
3320 return (ENOENT);
3321
3322 IF_ADDR_WLOCK(ifp);
3323 lastref = 0;
3324 ifma = if_findmulti(ifp, sa);
3325 if (ifma != NULL)
3326 lastref = if_delmulti_locked(ifp, ifma, 0);
3327 IF_ADDR_WUNLOCK(ifp);
3328
3329 if (ifma == NULL)
3330 return (ENOENT);
3331
3332 if (lastref && ifp->if_ioctl != NULL) {
3333 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3334 }
3335
3336 return (0);
3337 }
3338
3339 /*
3340 * Delete all multicast group membership for an interface.
3341 * Should be used to quickly flush all multicast filters.
3342 */
3343 void
3344 if_delallmulti(struct ifnet *ifp)
3345 {
3346 struct ifmultiaddr *ifma;
3347 struct ifmultiaddr *next;
3348
3349 IF_ADDR_WLOCK(ifp);
3350 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3351 if_delmulti_locked(ifp, ifma, 0);
3352 IF_ADDR_WUNLOCK(ifp);
3353 }
3354
3355 /*
3356 * Delete a multicast group membership by group membership pointer.
3357 * Network-layer protocol domains must use this routine.
3358 *
3359 * It is safe to call this routine if the ifp disappeared.
3360 */
3361 void
3362 if_delmulti_ifma(struct ifmultiaddr *ifma)
3363 {
3364 struct ifnet *ifp;
3365 int lastref;
3366
3367 ifp = ifma->ifma_ifp;
3368 #ifdef DIAGNOSTIC
3369 if (ifp == NULL) {
3370 printf("%s: ifma_ifp seems to be detached\n", __func__);
3371 } else {
3372 struct ifnet *oifp;
3373
3374 IFNET_RLOCK_NOSLEEP();
3375 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3376 if (ifp == oifp)
3377 break;
3378 if (ifp != oifp) {
3379 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3380 ifp = NULL;
3381 }
3382 IFNET_RUNLOCK_NOSLEEP();
3383 }
3384 #endif
3385 /*
3386 * If and only if the ifnet instance exists: Acquire the address lock.
3387 */
3388 if (ifp != NULL)
3389 IF_ADDR_WLOCK(ifp);
3390
3391 lastref = if_delmulti_locked(ifp, ifma, 0);
3392
3393 if (ifp != NULL) {
3394 /*
3395 * If and only if the ifnet instance exists:
3396 * Release the address lock.
3397 * If the group was left: update the hardware hash filter.
3398 */
3399 IF_ADDR_WUNLOCK(ifp);
3400 if (lastref && ifp->if_ioctl != NULL) {
3401 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3402 }
3403 }
3404 }
3405
3406 /*
3407 * Perform deletion of network-layer and/or link-layer multicast address.
3408 *
3409 * Return 0 if the reference count was decremented.
3410 * Return 1 if the final reference was released, indicating that the
3411 * hardware hash filter should be reprogrammed.
3412 */
3413 static int
3414 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3415 {
3416 struct ifmultiaddr *ll_ifma;
3417
3418 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3419 KASSERT(ifma->ifma_ifp == ifp,
3420 ("%s: inconsistent ifp %p", __func__, ifp));
3421 IF_ADDR_WLOCK_ASSERT(ifp);
3422 }
3423
3424 ifp = ifma->ifma_ifp;
3425
3426 /*
3427 * If the ifnet is detaching, null out references to ifnet,
3428 * so that upper protocol layers will notice, and not attempt
3429 * to obtain locks for an ifnet which no longer exists. The
3430 * routing socket announcement must happen before the ifnet
3431 * instance is detached from the system.
3432 */
3433 if (detaching) {
3434 #ifdef DIAGNOSTIC
3435 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3436 #endif
3437 /*
3438 * ifp may already be nulled out if we are being reentered
3439 * to delete the ll_ifma.
3440 */
3441 if (ifp != NULL) {
3442 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3443 ifma->ifma_ifp = NULL;
3444 }
3445 }
3446
3447 if (--ifma->ifma_refcount > 0)
3448 return 0;
3449
3450 /*
3451 * If this ifma is a network-layer ifma, a link-layer ifma may
3452 * have been associated with it. Release it first if so.
3453 */
3454 ll_ifma = ifma->ifma_llifma;
3455 if (ll_ifma != NULL) {
3456 KASSERT(ifma->ifma_lladdr != NULL,
3457 ("%s: llifma w/o lladdr", __func__));
3458 if (detaching)
3459 ll_ifma->ifma_ifp = NULL; /* XXX */
3460 if (--ll_ifma->ifma_refcount == 0) {
3461 if (ifp != NULL) {
3462 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3463 ifma_link);
3464 }
3465 if_freemulti(ll_ifma);
3466 }
3467 }
3468
3469 if (ifp != NULL)
3470 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3471
3472 if_freemulti(ifma);
3473
3474 /*
3475 * The last reference to this instance of struct ifmultiaddr
3476 * was released; the hardware should be notified of this change.
3477 */
3478 return 1;
3479 }
3480
3481 /*
3482 * Set the link layer address on an interface.
3483 *
3484 * At this time we only support certain types of interfaces,
3485 * and we don't allow the length of the address to change.
3486 */
3487 int
3488 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3489 {
3490 struct sockaddr_dl *sdl;
3491 struct ifaddr *ifa;
3492 struct ifreq ifr;
3493
3494 IF_ADDR_RLOCK(ifp);
3495 ifa = ifp->if_addr;
3496 if (ifa == NULL) {
3497 IF_ADDR_RUNLOCK(ifp);
3498 return (EINVAL);
3499 }
3500 ifa_ref(ifa);
3501 IF_ADDR_RUNLOCK(ifp);
3502 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3503 if (sdl == NULL) {
3504 ifa_free(ifa);
3505 return (EINVAL);
3506 }
3507 if (len != sdl->sdl_alen) { /* don't allow length to change */
3508 ifa_free(ifa);
3509 return (EINVAL);
3510 }
3511 switch (ifp->if_type) {
3512 case IFT_ETHER:
3513 case IFT_FDDI:
3514 case IFT_XETHER:
3515 case IFT_ISO88025:
3516 case IFT_L2VLAN:
3517 case IFT_BRIDGE:
3518 case IFT_ARCNET:
3519 case IFT_IEEE8023ADLAG:
3520 case IFT_IEEE80211:
3521 bcopy(lladdr, LLADDR(sdl), len);
3522 ifa_free(ifa);
3523 break;
3524 default:
3525 ifa_free(ifa);
3526 return (ENODEV);
3527 }
3528
3529 /*
3530 * If the interface is already up, we need
3531 * to re-init it in order to reprogram its
3532 * address filter.
3533 */
3534 if ((ifp->if_flags & IFF_UP) != 0) {
3535 if (ifp->if_ioctl) {
3536 ifp->if_flags &= ~IFF_UP;
3537 ifr.ifr_flags = ifp->if_flags & 0xffff;
3538 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3539 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3540 ifp->if_flags |= IFF_UP;
3541 ifr.ifr_flags = ifp->if_flags & 0xffff;
3542 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3543 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3544 }
3545 #ifdef INET
3546 /*
3547 * Also send gratuitous ARPs to notify other nodes about
3548 * the address change.
3549 */
3550 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3551 if (ifa->ifa_addr->sa_family == AF_INET)
3552 arp_ifinit(ifp, ifa);
3553 }
3554 #endif
3555 }
3556 return (0);
3557 }
3558
3559 /*
3560 * The name argument must be a pointer to storage which will last as
3561 * long as the interface does. For physical devices, the result of
3562 * device_get_name(dev) is a good choice and for pseudo-devices a
3563 * static string works well.
3564 */
3565 void
3566 if_initname(struct ifnet *ifp, const char *name, int unit)
3567 {
3568 ifp->if_dname = name;
3569 ifp->if_dunit = unit;
3570 if (unit != IF_DUNIT_NONE)
3571 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3572 else
3573 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3574 }
3575
3576 int
3577 if_printf(struct ifnet *ifp, const char * fmt, ...)
3578 {
3579 va_list ap;
3580 int retval;
3581
3582 retval = printf("%s: ", ifp->if_xname);
3583 va_start(ap, fmt);
3584 retval += vprintf(fmt, ap);
3585 va_end(ap);
3586 return (retval);
3587 }
3588
3589 void
3590 if_start(struct ifnet *ifp)
3591 {
3592
3593 (*(ifp)->if_start)(ifp);
3594 }
3595
3596 /*
3597 * Backwards compatibility interface for drivers
3598 * that have not implemented it
3599 */
3600 static int
3601 if_transmit(struct ifnet *ifp, struct mbuf *m)
3602 {
3603 int error;
3604
3605 IFQ_HANDOFF(ifp, m, error);
3606 return (error);
3607 }
3608
3609 static void
3610 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3611 {
3612
3613 m_freem(m);
3614 }
3615
3616 int
3617 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3618 {
3619 int active = 0;
3620
3621 IF_LOCK(ifq);
3622 if (_IF_QFULL(ifq)) {
3623 _IF_DROP(ifq);
3624 IF_UNLOCK(ifq);
3625 m_freem(m);
3626 return (0);
3627 }
3628 if (ifp != NULL) {
3629 ifp->if_obytes += m->m_pkthdr.len + adjust;
3630 if (m->m_flags & (M_BCAST|M_MCAST))
3631 ifp->if_omcasts++;
3632 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3633 }
3634 _IF_ENQUEUE(ifq, m);
3635 IF_UNLOCK(ifq);
3636 if (ifp != NULL && !active)
3637 (*(ifp)->if_start)(ifp);
3638 return (1);
3639 }
3640
3641 void
3642 if_register_com_alloc(u_char type,
3643 if_com_alloc_t *a, if_com_free_t *f)
3644 {
3645
3646 KASSERT(if_com_alloc[type] == NULL,
3647 ("if_register_com_alloc: %d already registered", type));
3648 KASSERT(if_com_free[type] == NULL,
3649 ("if_register_com_alloc: %d free already registered", type));
3650
3651 if_com_alloc[type] = a;
3652 if_com_free[type] = f;
3653 }
3654
3655 void
3656 if_deregister_com_alloc(u_char type)
3657 {
3658
3659 KASSERT(if_com_alloc[type] != NULL,
3660 ("if_deregister_com_alloc: %d not registered", type));
3661 KASSERT(if_com_free[type] != NULL,
3662 ("if_deregister_com_alloc: %d free not registered", type));
3663 if_com_alloc[type] = NULL;
3664 if_com_free[type] = NULL;
3665 }
Cache object: 464220972aef68a130622a1309d7f6f8
|