1 /*-
2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
30 * $FreeBSD$
31 */
32
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/conf.h>
40 #include <sys/malloc.h>
41 #include <sys/sbuf.h>
42 #include <sys/bus.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
60 #include <sys/jail.h>
61 #include <sys/priv.h>
62
63 #include <machine/stdarg.h>
64 #include <vm/uma.h>
65
66 #include <net/if.h>
67 #include <net/if_arp.h>
68 #include <net/if_clone.h>
69 #include <net/if_dl.h>
70 #include <net/if_types.h>
71 #include <net/if_var.h>
72 #include <net/radix.h>
73 #include <net/route.h>
74 #include <net/vnet.h>
75
76 #if defined(INET) || defined(INET6)
77 #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);
179 static void if_detach_internal(struct ifnet *, int);
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 *
ifnet_byindex_locked(u_short idx)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 *
ifnet_byindex(u_short idx)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 *
ifnet_byindex_ref(u_short idx)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
ifindex_alloc_locked(u_short * idxp)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
ifindex_free_locked(u_short idx)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
ifindex_free(u_short idx)313 ifindex_free(u_short idx)
314 {
315
316 IFNET_WLOCK();
317 ifindex_free_locked(idx);
318 IFNET_WUNLOCK();
319 }
320
321 static void
ifnet_setbyindex_locked(u_short idx,struct ifnet * ifp)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
ifnet_setbyindex(u_short idx,struct ifnet * ifp)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 *
ifaddr_byindex(u_short idx)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
vnet_if_init(const void * unused __unused)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
if_init(void * dummy __unused)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
vnet_if_uninit(const void * unused __unused)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
if_grow(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 *
if_alloc(u_char type)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
if_free_internal(struct ifnet * ifp)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
if_free(struct ifnet * ifp)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
if_ref(struct ifnet * ifp)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
if_rele(struct ifnet * ifp)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
ifq_init(struct ifaltq * ifq,struct ifnet * ifp)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
ifq_delete(struct ifaltq * ifq)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 * XXX:
575 * - The decision to return void and thus require this function to
576 * succeed is questionable.
577 * - We should probably do more sanity checking. For instance we don't
578 * do anything to insure if_xname is unique or non-empty.
579 */
580 void
if_attach(struct ifnet * ifp)581 if_attach(struct ifnet *ifp)
582 {
583
584 if_attach_internal(ifp, 0);
585 }
586
587 /*
588 * Compute the least common TSO limit.
589 */
590 void
if_hw_tsomax_common(struct ifnet * ifp,struct ifnet_hw_tsomax * pmax)591 if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
592 {
593 /*
594 * 1) If there is no limit currently, take the limit from
595 * the network adapter.
596 *
597 * 2) If the network adapter has a limit below the current
598 * limit, apply it.
599 */
600 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
601 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
602 pmax->tsomaxbytes = ifp->if_hw_tsomax;
603 }
604 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
605 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
606 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
607 }
608 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
609 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
610 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
611 }
612 }
613
614 /*
615 * Update TSO limit of a network adapter.
616 *
617 * Returns zero if no change. Else non-zero.
618 */
619 int
if_hw_tsomax_update(struct ifnet * ifp,struct ifnet_hw_tsomax * pmax)620 if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
621 {
622 int retval = 0;
623 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
624 ifp->if_hw_tsomax = pmax->tsomaxbytes;
625 retval++;
626 }
627 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
628 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
629 retval++;
630 }
631 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
632 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
633 retval++;
634 }
635 return (retval);
636 }
637
638 static void
if_attach_internal(struct ifnet * ifp,int vmove)639 if_attach_internal(struct ifnet *ifp, int vmove)
640 {
641 unsigned socksize, ifasize;
642 int namelen, masklen;
643 struct sockaddr_dl *sdl;
644 struct ifaddr *ifa;
645
646 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
647 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
648 ifp->if_xname);
649
650 #ifdef VIMAGE
651 ifp->if_vnet = curvnet;
652 if (ifp->if_home_vnet == NULL)
653 ifp->if_home_vnet = curvnet;
654 #endif
655
656 if_addgroup(ifp, IFG_ALL);
657
658 getmicrotime(&ifp->if_lastchange);
659 ifp->if_data.ifi_epoch = time_uptime;
660 ifp->if_data.ifi_datalen = sizeof(struct if_data);
661
662 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
663 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
664 ("transmit and qflush must both either be set or both be NULL"));
665 if (ifp->if_transmit == NULL) {
666 ifp->if_transmit = if_transmit;
667 ifp->if_qflush = if_qflush;
668 }
669 if (ifp->if_input == NULL)
670 ifp->if_input = if_input_default;
671
672 if (!vmove) {
673 #ifdef MAC
674 mac_ifnet_create(ifp);
675 #endif
676
677 /*
678 * Create a Link Level name for this device.
679 */
680 namelen = strlen(ifp->if_xname);
681 /*
682 * Always save enough space for any possiable name so we
683 * can do a rename in place later.
684 */
685 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
686 socksize = masklen + ifp->if_addrlen;
687 if (socksize < sizeof(*sdl))
688 socksize = sizeof(*sdl);
689 socksize = roundup2(socksize, sizeof(long));
690 ifasize = sizeof(*ifa) + 2 * socksize;
691 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
692 ifa_init(ifa);
693 sdl = (struct sockaddr_dl *)(ifa + 1);
694 sdl->sdl_len = socksize;
695 sdl->sdl_family = AF_LINK;
696 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
697 sdl->sdl_nlen = namelen;
698 sdl->sdl_index = ifp->if_index;
699 sdl->sdl_type = ifp->if_type;
700 ifp->if_addr = ifa;
701 ifa->ifa_ifp = ifp;
702 ifa->ifa_rtrequest = link_rtrequest;
703 ifa->ifa_addr = (struct sockaddr *)sdl;
704 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
705 ifa->ifa_netmask = (struct sockaddr *)sdl;
706 sdl->sdl_len = masklen;
707 while (namelen != 0)
708 sdl->sdl_data[--namelen] = 0xff;
709 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
710 /* Reliably crash if used uninitialized. */
711 ifp->if_broadcastaddr = NULL;
712
713 #if defined(INET) || defined(INET6)
714 /* Use defaults for TSO, if nothing is set */
715 if (ifp->if_hw_tsomax == 0 &&
716 ifp->if_hw_tsomaxsegcount == 0 &&
717 ifp->if_hw_tsomaxsegsize == 0) {
718 /*
719 * The TSO defaults needs to be such that an
720 * NFS mbuf list of 35 mbufs totalling just
721 * below 64K works and that a chain of mbufs
722 * can be defragged into at most 32 segments:
723 */
724 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
725 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
726 ifp->if_hw_tsomaxsegcount = 35;
727 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
728
729 /* XXX some drivers set IFCAP_TSO after ethernet attach */
730 if (ifp->if_capabilities & IFCAP_TSO) {
731 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
732 ifp->if_hw_tsomax,
733 ifp->if_hw_tsomaxsegcount,
734 ifp->if_hw_tsomaxsegsize);
735 }
736 }
737 #endif
738 }
739 #ifdef VIMAGE
740 else {
741 /*
742 * Update the interface index in the link layer address
743 * of the interface.
744 */
745 for (ifa = ifp->if_addr; ifa != NULL;
746 ifa = TAILQ_NEXT(ifa, ifa_link)) {
747 if (ifa->ifa_addr->sa_family == AF_LINK) {
748 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
749 sdl->sdl_index = ifp->if_index;
750 }
751 }
752 }
753 #endif
754
755 IFNET_WLOCK();
756 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
757 #ifdef VIMAGE
758 curvnet->vnet_ifcnt++;
759 #endif
760 IFNET_WUNLOCK();
761
762 if (domain_init_status >= 2)
763 if_attachdomain1(ifp);
764
765 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
766 if (IS_DEFAULT_VNET(curvnet))
767 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
768
769 /* Announce the interface. */
770 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
771 }
772
773 static void
if_attachdomain(void * dummy)774 if_attachdomain(void *dummy)
775 {
776 struct ifnet *ifp;
777
778 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
779 if_attachdomain1(ifp);
780 }
781 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
782 if_attachdomain, NULL);
783
784 static void
if_attachdomain1(struct ifnet * ifp)785 if_attachdomain1(struct ifnet *ifp)
786 {
787 struct domain *dp;
788
789 /*
790 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
791 * cannot lock ifp->if_afdata initialization, entirely.
792 */
793 if (IF_AFDATA_TRYLOCK(ifp) == 0)
794 return;
795 if (ifp->if_afdata_initialized >= domain_init_status) {
796 IF_AFDATA_UNLOCK(ifp);
797 log(LOG_WARNING, "%s called more than once on %s\n",
798 __func__, ifp->if_xname);
799 return;
800 }
801 ifp->if_afdata_initialized = domain_init_status;
802 IF_AFDATA_UNLOCK(ifp);
803
804 /* address family dependent data region */
805 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
806 for (dp = domains; dp; dp = dp->dom_next) {
807 if (dp->dom_ifattach)
808 ifp->if_afdata[dp->dom_family] =
809 (*dp->dom_ifattach)(ifp);
810 }
811 }
812
813 /*
814 * Remove any unicast or broadcast network addresses from an interface.
815 */
816 void
if_purgeaddrs(struct ifnet * ifp)817 if_purgeaddrs(struct ifnet *ifp)
818 {
819 struct ifaddr *ifa, *next;
820
821 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
822 if (ifa->ifa_addr->sa_family == AF_LINK)
823 continue;
824 #ifdef INET
825 /* XXX: Ugly!! ad hoc just for INET */
826 if (ifa->ifa_addr->sa_family == AF_INET) {
827 struct ifaliasreq ifr;
828
829 bzero(&ifr, sizeof(ifr));
830 ifr.ifra_addr = *ifa->ifa_addr;
831 if (ifa->ifa_dstaddr)
832 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
833 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
834 NULL) == 0)
835 continue;
836 }
837 #endif /* INET */
838 #ifdef INET6
839 if (ifa->ifa_addr->sa_family == AF_INET6) {
840 in6_purgeaddr(ifa);
841 /* ifp_addrhead is already updated */
842 continue;
843 }
844 #endif /* INET6 */
845 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
846 ifa_free(ifa);
847 }
848 }
849
850 /*
851 * Remove any multicast network addresses from an interface when an ifnet
852 * is going away.
853 */
854 static void
if_purgemaddrs(struct ifnet * ifp)855 if_purgemaddrs(struct ifnet *ifp)
856 {
857 struct ifmultiaddr *ifma;
858 struct ifmultiaddr *next;
859
860 IF_ADDR_WLOCK(ifp);
861 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
862 if_delmulti_locked(ifp, ifma, 1);
863 IF_ADDR_WUNLOCK(ifp);
864 }
865
866 /*
867 * Detach an interface, removing it from the list of "active" interfaces.
868 * If vmove flag is set on entry to if_detach_internal(), perform only a
869 * limited subset of cleanup tasks, given that we are moving an ifnet from
870 * one vnet to another, where it must be fully operational.
871 *
872 * XXXRW: There are some significant questions about event ordering, and
873 * how to prevent things from starting to use the interface during detach.
874 */
875 void
if_detach(struct ifnet * ifp)876 if_detach(struct ifnet *ifp)
877 {
878
879 CURVNET_SET_QUIET(ifp->if_vnet);
880 if_detach_internal(ifp, 0);
881 CURVNET_RESTORE();
882 }
883
884 static void
if_detach_internal(struct ifnet * ifp,int vmove)885 if_detach_internal(struct ifnet *ifp, int vmove)
886 {
887 struct ifaddr *ifa;
888 struct radix_node_head *rnh;
889 int i, j;
890 struct domain *dp;
891 struct ifnet *iter;
892 int found = 0;
893
894 IFNET_WLOCK();
895 TAILQ_FOREACH(iter, &V_ifnet, if_link)
896 if (iter == ifp) {
897 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
898 found = 1;
899 break;
900 }
901 #ifdef VIMAGE
902 if (found)
903 curvnet->vnet_ifcnt--;
904 #endif
905 IFNET_WUNLOCK();
906 if (!found) {
907 if (vmove)
908 panic("%s: ifp=%p not on the ifnet tailq %p",
909 __func__, ifp, &V_ifnet);
910 else
911 return; /* XXX this should panic as well? */
912 }
913
914 /*
915 * Remove/wait for pending events.
916 */
917 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
918
919 /*
920 * Remove routes and flush queues.
921 */
922 if_down(ifp);
923 #ifdef ALTQ
924 if (ALTQ_IS_ENABLED(&ifp->if_snd))
925 altq_disable(&ifp->if_snd);
926 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
927 altq_detach(&ifp->if_snd);
928 #endif
929
930 if_purgeaddrs(ifp);
931
932 #ifdef INET
933 in_ifdetach(ifp);
934 #endif
935
936 #ifdef INET6
937 /*
938 * Remove all IPv6 kernel structs related to ifp. This should be done
939 * before removing routing entries below, since IPv6 interface direct
940 * routes are expected to be removed by the IPv6-specific kernel API.
941 * Otherwise, the kernel will detect some inconsistency and bark it.
942 */
943 in6_ifdetach(ifp);
944 #endif
945 if_purgemaddrs(ifp);
946
947 if (!vmove) {
948 /*
949 * Prevent further calls into the device driver via ifnet.
950 */
951 if_dead(ifp);
952
953 /*
954 * Remove link ifaddr pointer and maybe decrement if_index.
955 * Clean up all addresses.
956 */
957 ifp->if_addr = NULL;
958
959 /* We can now free link ifaddr. */
960 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
961 ifa = TAILQ_FIRST(&ifp->if_addrhead);
962 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
963 ifa_free(ifa);
964 }
965 }
966
967 /*
968 * Delete all remaining routes using this interface
969 * Unfortuneatly the only way to do this is to slog through
970 * the entire routing table looking for routes which point
971 * to this interface...oh well...
972 */
973 for (i = 1; i <= AF_MAX; i++) {
974 for (j = 0; j < rt_numfibs; j++) {
975 rnh = rt_tables_get_rnh(j, i);
976 if (rnh == NULL)
977 continue;
978 RADIX_NODE_HEAD_LOCK(rnh);
979 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
980 RADIX_NODE_HEAD_UNLOCK(rnh);
981 }
982 }
983
984 /* Announce that the interface is gone. */
985 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
986 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
987 if (IS_DEFAULT_VNET(curvnet))
988 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
989 if_delgroups(ifp);
990
991 /*
992 * We cannot hold the lock over dom_ifdetach calls as they might
993 * sleep, for example trying to drain a callout, thus open up the
994 * theoretical race with re-attaching.
995 */
996 IF_AFDATA_LOCK(ifp);
997 i = ifp->if_afdata_initialized;
998 ifp->if_afdata_initialized = 0;
999 IF_AFDATA_UNLOCK(ifp);
1000 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1001 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1002 (*dp->dom_ifdetach)(ifp,
1003 ifp->if_afdata[dp->dom_family]);
1004 }
1005 }
1006
1007 #ifdef VIMAGE
1008 /*
1009 * if_vmove() performs a limited version of if_detach() in current
1010 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1011 * An attempt is made to shrink if_index in current vnet, find an
1012 * unused if_index in target vnet and calls if_grow() if necessary,
1013 * and finally find an unused if_xname for the target vnet.
1014 */
1015 void
if_vmove(struct ifnet * ifp,struct vnet * new_vnet)1016 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1017 {
1018 u_short idx;
1019
1020 /*
1021 * Detach from current vnet, but preserve LLADDR info, do not
1022 * mark as dead etc. so that the ifnet can be reattached later.
1023 */
1024 if_detach_internal(ifp, 1);
1025
1026 /*
1027 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1028 * the if_index for that vnet if possible.
1029 *
1030 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1031 * or we'd lock on one vnet and unlock on another.
1032 */
1033 IFNET_WLOCK();
1034 ifindex_free_locked(ifp->if_index);
1035 IFNET_WUNLOCK();
1036
1037 /*
1038 * Perform interface-specific reassignment tasks, if provided by
1039 * the driver.
1040 */
1041 if (ifp->if_reassign != NULL)
1042 ifp->if_reassign(ifp, new_vnet, NULL);
1043
1044 /*
1045 * Switch to the context of the target vnet.
1046 */
1047 CURVNET_SET_QUIET(new_vnet);
1048
1049 IFNET_WLOCK();
1050 if (ifindex_alloc_locked(&idx) != 0) {
1051 IFNET_WUNLOCK();
1052 panic("if_index overflow");
1053 }
1054 ifp->if_index = idx;
1055 ifnet_setbyindex_locked(ifp->if_index, ifp);
1056 IFNET_WUNLOCK();
1057
1058 if_attach_internal(ifp, 1);
1059
1060 CURVNET_RESTORE();
1061 }
1062
1063 /*
1064 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1065 */
1066 static int
if_vmove_loan(struct thread * td,struct ifnet * ifp,char * ifname,int jid)1067 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1068 {
1069 struct prison *pr;
1070 struct ifnet *difp;
1071
1072 /* Try to find the prison within our visibility. */
1073 sx_slock(&allprison_lock);
1074 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1075 sx_sunlock(&allprison_lock);
1076 if (pr == NULL)
1077 return (ENXIO);
1078 prison_hold_locked(pr);
1079 mtx_unlock(&pr->pr_mtx);
1080
1081 /* Do not try to move the iface from and to the same prison. */
1082 if (pr->pr_vnet == ifp->if_vnet) {
1083 prison_free(pr);
1084 return (EEXIST);
1085 }
1086
1087 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1088 /* XXX Lock interfaces to avoid races. */
1089 CURVNET_SET_QUIET(pr->pr_vnet);
1090 difp = ifunit(ifname);
1091 CURVNET_RESTORE();
1092 if (difp != NULL) {
1093 prison_free(pr);
1094 return (EEXIST);
1095 }
1096
1097 /* Move the interface into the child jail/vnet. */
1098 if_vmove(ifp, pr->pr_vnet);
1099
1100 /* Report the new if_xname back to the userland. */
1101 sprintf(ifname, "%s", ifp->if_xname);
1102
1103 prison_free(pr);
1104 return (0);
1105 }
1106
1107 static int
if_vmove_reclaim(struct thread * td,char * ifname,int jid)1108 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1109 {
1110 struct prison *pr;
1111 struct vnet *vnet_dst;
1112 struct ifnet *ifp;
1113
1114 /* Try to find the prison within our visibility. */
1115 sx_slock(&allprison_lock);
1116 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1117 sx_sunlock(&allprison_lock);
1118 if (pr == NULL)
1119 return (ENXIO);
1120 prison_hold_locked(pr);
1121 mtx_unlock(&pr->pr_mtx);
1122
1123 /* Make sure the named iface exists in the source prison/vnet. */
1124 CURVNET_SET(pr->pr_vnet);
1125 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1126 if (ifp == NULL) {
1127 CURVNET_RESTORE();
1128 prison_free(pr);
1129 return (ENXIO);
1130 }
1131
1132 /* Do not try to move the iface from and to the same prison. */
1133 vnet_dst = TD_TO_VNET(td);
1134 if (vnet_dst == ifp->if_vnet) {
1135 CURVNET_RESTORE();
1136 prison_free(pr);
1137 return (EEXIST);
1138 }
1139
1140 /* Get interface back from child jail/vnet. */
1141 if_vmove(ifp, vnet_dst);
1142 CURVNET_RESTORE();
1143
1144 /* Report the new if_xname back to the userland. */
1145 sprintf(ifname, "%s", ifp->if_xname);
1146
1147 prison_free(pr);
1148 return (0);
1149 }
1150 #endif /* VIMAGE */
1151
1152 /*
1153 * Add a group to an interface
1154 */
1155 int
if_addgroup(struct ifnet * ifp,const char * groupname)1156 if_addgroup(struct ifnet *ifp, const char *groupname)
1157 {
1158 struct ifg_list *ifgl;
1159 struct ifg_group *ifg = NULL;
1160 struct ifg_member *ifgm;
1161 int new = 0;
1162
1163 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1164 groupname[strlen(groupname) - 1] <= '9')
1165 return (EINVAL);
1166
1167 IFNET_WLOCK();
1168 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1169 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1170 IFNET_WUNLOCK();
1171 return (EEXIST);
1172 }
1173
1174 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1175 M_NOWAIT)) == NULL) {
1176 IFNET_WUNLOCK();
1177 return (ENOMEM);
1178 }
1179
1180 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1181 M_TEMP, M_NOWAIT)) == NULL) {
1182 free(ifgl, M_TEMP);
1183 IFNET_WUNLOCK();
1184 return (ENOMEM);
1185 }
1186
1187 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1188 if (!strcmp(ifg->ifg_group, groupname))
1189 break;
1190
1191 if (ifg == NULL) {
1192 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1193 M_TEMP, M_NOWAIT)) == NULL) {
1194 free(ifgl, M_TEMP);
1195 free(ifgm, M_TEMP);
1196 IFNET_WUNLOCK();
1197 return (ENOMEM);
1198 }
1199 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1200 ifg->ifg_refcnt = 0;
1201 TAILQ_INIT(&ifg->ifg_members);
1202 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1203 new = 1;
1204 }
1205
1206 ifg->ifg_refcnt++;
1207 ifgl->ifgl_group = ifg;
1208 ifgm->ifgm_ifp = ifp;
1209
1210 IF_ADDR_WLOCK(ifp);
1211 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1212 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1213 IF_ADDR_WUNLOCK(ifp);
1214
1215 IFNET_WUNLOCK();
1216
1217 if (new)
1218 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1219 EVENTHANDLER_INVOKE(group_change_event, groupname);
1220
1221 return (0);
1222 }
1223
1224 /*
1225 * Remove a group from an interface
1226 */
1227 int
if_delgroup(struct ifnet * ifp,const char * groupname)1228 if_delgroup(struct ifnet *ifp, const char *groupname)
1229 {
1230 struct ifg_list *ifgl;
1231 struct ifg_member *ifgm;
1232
1233 IFNET_WLOCK();
1234 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1235 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1236 break;
1237 if (ifgl == NULL) {
1238 IFNET_WUNLOCK();
1239 return (ENOENT);
1240 }
1241
1242 IF_ADDR_WLOCK(ifp);
1243 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1244 IF_ADDR_WUNLOCK(ifp);
1245
1246 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1247 if (ifgm->ifgm_ifp == ifp)
1248 break;
1249
1250 if (ifgm != NULL) {
1251 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1252 free(ifgm, M_TEMP);
1253 }
1254
1255 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1256 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1257 IFNET_WUNLOCK();
1258 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1259 free(ifgl->ifgl_group, M_TEMP);
1260 } else
1261 IFNET_WUNLOCK();
1262
1263 free(ifgl, M_TEMP);
1264
1265 EVENTHANDLER_INVOKE(group_change_event, groupname);
1266
1267 return (0);
1268 }
1269
1270 /*
1271 * Remove an interface from all groups
1272 */
1273 static void
if_delgroups(struct ifnet * ifp)1274 if_delgroups(struct ifnet *ifp)
1275 {
1276 struct ifg_list *ifgl;
1277 struct ifg_member *ifgm;
1278 char groupname[IFNAMSIZ];
1279
1280 IFNET_WLOCK();
1281 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1282 ifgl = TAILQ_FIRST(&ifp->if_groups);
1283
1284 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1285
1286 IF_ADDR_WLOCK(ifp);
1287 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1288 IF_ADDR_WUNLOCK(ifp);
1289
1290 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1291 if (ifgm->ifgm_ifp == ifp)
1292 break;
1293
1294 if (ifgm != NULL) {
1295 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1296 ifgm_next);
1297 free(ifgm, M_TEMP);
1298 }
1299
1300 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1301 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1302 IFNET_WUNLOCK();
1303 EVENTHANDLER_INVOKE(group_detach_event,
1304 ifgl->ifgl_group);
1305 free(ifgl->ifgl_group, M_TEMP);
1306 } else
1307 IFNET_WUNLOCK();
1308
1309 free(ifgl, M_TEMP);
1310
1311 EVENTHANDLER_INVOKE(group_change_event, groupname);
1312
1313 IFNET_WLOCK();
1314 }
1315 IFNET_WUNLOCK();
1316 }
1317
1318 /*
1319 * Stores all groups from an interface in memory pointed
1320 * to by data
1321 */
1322 static int
if_getgroup(struct ifgroupreq * data,struct ifnet * ifp)1323 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1324 {
1325 int len, error;
1326 struct ifg_list *ifgl;
1327 struct ifg_req ifgrq, *ifgp;
1328 struct ifgroupreq *ifgr = data;
1329
1330 if (ifgr->ifgr_len == 0) {
1331 IF_ADDR_RLOCK(ifp);
1332 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1333 ifgr->ifgr_len += sizeof(struct ifg_req);
1334 IF_ADDR_RUNLOCK(ifp);
1335 return (0);
1336 }
1337
1338 len = ifgr->ifgr_len;
1339 ifgp = ifgr->ifgr_groups;
1340 /* XXX: wire */
1341 IF_ADDR_RLOCK(ifp);
1342 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1343 if (len < sizeof(ifgrq)) {
1344 IF_ADDR_RUNLOCK(ifp);
1345 return (EINVAL);
1346 }
1347 bzero(&ifgrq, sizeof ifgrq);
1348 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1349 sizeof(ifgrq.ifgrq_group));
1350 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1351 IF_ADDR_RUNLOCK(ifp);
1352 return (error);
1353 }
1354 len -= sizeof(ifgrq);
1355 ifgp++;
1356 }
1357 IF_ADDR_RUNLOCK(ifp);
1358
1359 return (0);
1360 }
1361
1362 /*
1363 * Stores all members of a group in memory pointed to by data
1364 */
1365 static int
if_getgroupmembers(struct ifgroupreq * data)1366 if_getgroupmembers(struct ifgroupreq *data)
1367 {
1368 struct ifgroupreq *ifgr = data;
1369 struct ifg_group *ifg;
1370 struct ifg_member *ifgm;
1371 struct ifg_req ifgrq, *ifgp;
1372 int len, error;
1373
1374 IFNET_RLOCK();
1375 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1376 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1377 break;
1378 if (ifg == NULL) {
1379 IFNET_RUNLOCK();
1380 return (ENOENT);
1381 }
1382
1383 if (ifgr->ifgr_len == 0) {
1384 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1385 ifgr->ifgr_len += sizeof(ifgrq);
1386 IFNET_RUNLOCK();
1387 return (0);
1388 }
1389
1390 len = ifgr->ifgr_len;
1391 ifgp = ifgr->ifgr_groups;
1392 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1393 if (len < sizeof(ifgrq)) {
1394 IFNET_RUNLOCK();
1395 return (EINVAL);
1396 }
1397 bzero(&ifgrq, sizeof ifgrq);
1398 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1399 sizeof(ifgrq.ifgrq_member));
1400 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1401 IFNET_RUNLOCK();
1402 return (error);
1403 }
1404 len -= sizeof(ifgrq);
1405 ifgp++;
1406 }
1407 IFNET_RUNLOCK();
1408
1409 return (0);
1410 }
1411
1412 /*
1413 * Delete Routes for a Network Interface
1414 *
1415 * Called for each routing entry via the rnh->rnh_walktree() call above
1416 * to delete all route entries referencing a detaching network interface.
1417 *
1418 * Arguments:
1419 * rn pointer to node in the routing table
1420 * arg argument passed to rnh->rnh_walktree() - detaching interface
1421 *
1422 * Returns:
1423 * 0 successful
1424 * errno failed - reason indicated
1425 *
1426 */
1427 static int
if_rtdel(struct radix_node * rn,void * arg)1428 if_rtdel(struct radix_node *rn, void *arg)
1429 {
1430 struct rtentry *rt = (struct rtentry *)rn;
1431 struct ifnet *ifp = arg;
1432 int err;
1433
1434 if (rt->rt_ifp == ifp) {
1435
1436 /*
1437 * Protect (sorta) against walktree recursion problems
1438 * with cloned routes
1439 */
1440 if ((rt->rt_flags & RTF_UP) == 0)
1441 return (0);
1442
1443 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1444 rt_mask(rt),
1445 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1446 (struct rtentry **) NULL, rt->rt_fibnum);
1447 if (err) {
1448 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1449 }
1450 }
1451
1452 return (0);
1453 }
1454
1455 /*
1456 * A compatibility function returns ifnet counter values.
1457 */
1458 uint64_t
if_get_counter_default(struct ifnet * ifp,ift_counter cnt)1459 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1460 {
1461
1462 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1463 switch (cnt) {
1464 case IFCOUNTER_IPACKETS:
1465 return (ifp->if_ipackets);
1466 case IFCOUNTER_IERRORS:
1467 return (ifp->if_ierrors);
1468 case IFCOUNTER_OPACKETS:
1469 return (ifp->if_opackets);
1470 case IFCOUNTER_OERRORS:
1471 return (ifp->if_oerrors);
1472 case IFCOUNTER_COLLISIONS:
1473 return (ifp->if_collisions);
1474 case IFCOUNTER_IBYTES:
1475 return (ifp->if_ibytes);
1476 case IFCOUNTER_OBYTES:
1477 return (ifp->if_obytes);
1478 case IFCOUNTER_IMCASTS:
1479 return (ifp->if_imcasts);
1480 case IFCOUNTER_OMCASTS:
1481 return (ifp->if_omcasts);
1482 case IFCOUNTER_IQDROPS:
1483 return (ifp->if_iqdrops);
1484 #ifdef _IFI_OQDROPS
1485 case IFCOUNTER_OQDROPS:
1486 return (ifp->if_oqdrops);
1487 #endif
1488 case IFCOUNTER_NOPROTO:
1489 return (ifp->if_noproto);
1490 default:
1491 break;
1492 };
1493 return (0);
1494 }
1495
1496 /*
1497 * Increase an ifnet counter. Usually used for counters shared
1498 * between the stack and a driver, but function supports them all.
1499 */
1500 void
if_inc_counter(struct ifnet * ifp,ift_counter cnt,int64_t inc)1501 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1502 {
1503
1504 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1505 switch (cnt) {
1506 case IFCOUNTER_IPACKETS:
1507 ifp->if_ipackets += inc;
1508 break;
1509 case IFCOUNTER_IERRORS:
1510 ifp->if_ierrors += inc;
1511 break;
1512 case IFCOUNTER_OPACKETS:
1513 ifp->if_opackets += inc;
1514 break;
1515 case IFCOUNTER_OERRORS:
1516 ifp->if_oerrors += inc;
1517 break;
1518 case IFCOUNTER_COLLISIONS:
1519 ifp->if_collisions += inc;
1520 break;
1521 case IFCOUNTER_IBYTES:
1522 ifp->if_ibytes += inc;
1523 break;
1524 case IFCOUNTER_OBYTES:
1525 ifp->if_obytes += inc;
1526 break;
1527 case IFCOUNTER_IMCASTS:
1528 ifp->if_imcasts += inc;
1529 break;
1530 case IFCOUNTER_OMCASTS:
1531 ifp->if_omcasts += inc;
1532 break;
1533 case IFCOUNTER_IQDROPS:
1534 ifp->if_iqdrops += inc;
1535 break;
1536 #ifdef _IFI_OQDROPS
1537 case IFCOUNTER_OQDROPS:
1538 ifp->if_oqdrops += inc;
1539 break;
1540 #endif
1541 case IFCOUNTER_NOPROTO:
1542 ifp->if_noproto += inc;
1543 break;
1544 default:
1545 break;
1546 };
1547 }
1548
1549 /*
1550 * Wrapper functions for struct ifnet address list locking macros. These are
1551 * used by kernel modules to avoid encoding programming interface or binary
1552 * interface assumptions that may be violated when kernel-internal locking
1553 * approaches change.
1554 */
1555 void
if_addr_rlock(struct ifnet * ifp)1556 if_addr_rlock(struct ifnet *ifp)
1557 {
1558
1559 IF_ADDR_RLOCK(ifp);
1560 }
1561
1562 void
if_addr_runlock(struct ifnet * ifp)1563 if_addr_runlock(struct ifnet *ifp)
1564 {
1565
1566 IF_ADDR_RUNLOCK(ifp);
1567 }
1568
1569 void
if_maddr_rlock(struct ifnet * ifp)1570 if_maddr_rlock(struct ifnet *ifp)
1571 {
1572
1573 IF_ADDR_RLOCK(ifp);
1574 }
1575
1576 void
if_maddr_runlock(struct ifnet * ifp)1577 if_maddr_runlock(struct ifnet *ifp)
1578 {
1579
1580 IF_ADDR_RUNLOCK(ifp);
1581 }
1582
1583 /*
1584 * Initialization, destruction and refcounting functions for ifaddrs.
1585 */
1586 void
ifa_init(struct ifaddr * ifa)1587 ifa_init(struct ifaddr *ifa)
1588 {
1589
1590 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1591 refcount_init(&ifa->ifa_refcnt, 1);
1592 ifa->if_data.ifi_datalen = sizeof(ifa->if_data);
1593 }
1594
1595 void
ifa_ref(struct ifaddr * ifa)1596 ifa_ref(struct ifaddr *ifa)
1597 {
1598
1599 refcount_acquire(&ifa->ifa_refcnt);
1600 }
1601
1602 void
ifa_free(struct ifaddr * ifa)1603 ifa_free(struct ifaddr *ifa)
1604 {
1605
1606 if (refcount_release(&ifa->ifa_refcnt)) {
1607 mtx_destroy(&ifa->ifa_mtx);
1608 free(ifa, M_IFADDR);
1609 }
1610 }
1611
1612 int
ifa_add_loopback_route(struct ifaddr * ifa,struct sockaddr * ia)1613 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1614 {
1615 int error = 0;
1616 struct rtentry *rt = NULL;
1617 struct rt_addrinfo info;
1618 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1619
1620 bzero(&info, sizeof(info));
1621 info.rti_ifp = V_loif;
1622 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1623 info.rti_info[RTAX_DST] = ia;
1624 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1625 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1626
1627 if (error == 0 && rt != NULL) {
1628 RT_LOCK(rt);
1629 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1630 ifa->ifa_ifp->if_type;
1631 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1632 ifa->ifa_ifp->if_index;
1633 RT_REMREF(rt);
1634 RT_UNLOCK(rt);
1635 } else if (error != 0)
1636 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1637
1638 return (error);
1639 }
1640
1641 int
ifa_del_loopback_route(struct ifaddr * ifa,struct sockaddr * ia)1642 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1643 {
1644 int error = 0;
1645 struct rt_addrinfo info;
1646 struct sockaddr_dl null_sdl;
1647
1648 bzero(&null_sdl, sizeof(null_sdl));
1649 null_sdl.sdl_len = sizeof(null_sdl);
1650 null_sdl.sdl_family = AF_LINK;
1651 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1652 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1653 bzero(&info, sizeof(info));
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_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1658
1659 if (error != 0)
1660 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1661
1662 return (error);
1663 }
1664
1665 /*
1666 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1667 * structs used to represent other address families, it is necessary
1668 * to perform a different comparison.
1669 */
1670
1671 #define sa_equal(a1, a2) \
1672 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1673
1674 #define sa_dl_equal(a1, a2) \
1675 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1676 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1677 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1678 LLADDR((struct sockaddr_dl *)(a2)), \
1679 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1680
1681 /*
1682 * Locate an interface based on a complete address.
1683 */
1684 /*ARGSUSED*/
1685 static struct ifaddr *
ifa_ifwithaddr_internal(struct sockaddr * addr,int getref)1686 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1687 {
1688 struct ifnet *ifp;
1689 struct ifaddr *ifa;
1690
1691 IFNET_RLOCK_NOSLEEP();
1692 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1693 IF_ADDR_RLOCK(ifp);
1694 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1695 if (ifa->ifa_addr->sa_family != addr->sa_family)
1696 continue;
1697 if (sa_equal(addr, ifa->ifa_addr)) {
1698 if (getref)
1699 ifa_ref(ifa);
1700 IF_ADDR_RUNLOCK(ifp);
1701 goto done;
1702 }
1703 /* IP6 doesn't have broadcast */
1704 if ((ifp->if_flags & IFF_BROADCAST) &&
1705 ifa->ifa_broadaddr &&
1706 ifa->ifa_broadaddr->sa_len != 0 &&
1707 sa_equal(ifa->ifa_broadaddr, addr)) {
1708 if (getref)
1709 ifa_ref(ifa);
1710 IF_ADDR_RUNLOCK(ifp);
1711 goto done;
1712 }
1713 }
1714 IF_ADDR_RUNLOCK(ifp);
1715 }
1716 ifa = NULL;
1717 done:
1718 IFNET_RUNLOCK_NOSLEEP();
1719 return (ifa);
1720 }
1721
1722 struct ifaddr *
ifa_ifwithaddr(struct sockaddr * addr)1723 ifa_ifwithaddr(struct sockaddr *addr)
1724 {
1725
1726 return (ifa_ifwithaddr_internal(addr, 1));
1727 }
1728
1729 int
ifa_ifwithaddr_check(struct sockaddr * addr)1730 ifa_ifwithaddr_check(struct sockaddr *addr)
1731 {
1732
1733 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1734 }
1735
1736 /*
1737 * Locate an interface based on the broadcast address.
1738 */
1739 /* ARGSUSED */
1740 struct ifaddr *
ifa_ifwithbroadaddr(struct sockaddr * addr)1741 ifa_ifwithbroadaddr(struct sockaddr *addr)
1742 {
1743 struct ifnet *ifp;
1744 struct ifaddr *ifa;
1745
1746 IFNET_RLOCK_NOSLEEP();
1747 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1748 IF_ADDR_RLOCK(ifp);
1749 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1750 if (ifa->ifa_addr->sa_family != addr->sa_family)
1751 continue;
1752 if ((ifp->if_flags & IFF_BROADCAST) &&
1753 ifa->ifa_broadaddr &&
1754 ifa->ifa_broadaddr->sa_len != 0 &&
1755 sa_equal(ifa->ifa_broadaddr, addr)) {
1756 ifa_ref(ifa);
1757 IF_ADDR_RUNLOCK(ifp);
1758 goto done;
1759 }
1760 }
1761 IF_ADDR_RUNLOCK(ifp);
1762 }
1763 ifa = NULL;
1764 done:
1765 IFNET_RUNLOCK_NOSLEEP();
1766 return (ifa);
1767 }
1768
1769 /*
1770 * Locate the point to point interface with a given destination address.
1771 */
1772 /*ARGSUSED*/
1773 struct ifaddr *
ifa_ifwithdstaddr_fib(struct sockaddr * addr,int fibnum)1774 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
1775 {
1776 struct ifnet *ifp;
1777 struct ifaddr *ifa;
1778
1779 IFNET_RLOCK_NOSLEEP();
1780 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1781 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1782 continue;
1783 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1784 continue;
1785 IF_ADDR_RLOCK(ifp);
1786 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1787 if (ifa->ifa_addr->sa_family != addr->sa_family)
1788 continue;
1789 if (ifa->ifa_dstaddr != NULL &&
1790 sa_equal(addr, ifa->ifa_dstaddr)) {
1791 ifa_ref(ifa);
1792 IF_ADDR_RUNLOCK(ifp);
1793 goto done;
1794 }
1795 }
1796 IF_ADDR_RUNLOCK(ifp);
1797 }
1798 ifa = NULL;
1799 done:
1800 IFNET_RUNLOCK_NOSLEEP();
1801 return (ifa);
1802 }
1803
1804 struct ifaddr *
ifa_ifwithdstaddr(struct sockaddr * addr)1805 ifa_ifwithdstaddr(struct sockaddr *addr)
1806 {
1807
1808 return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
1809 }
1810
1811 /*
1812 * Find an interface on a specific network. If many, choice
1813 * is most specific found.
1814 */
1815 struct ifaddr *
ifa_ifwithnet_fib(struct sockaddr * addr,int ignore_ptp,int fibnum)1816 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
1817 {
1818 struct ifnet *ifp;
1819 struct ifaddr *ifa;
1820 struct ifaddr *ifa_maybe = NULL;
1821 u_int af = addr->sa_family;
1822 char *addr_data = addr->sa_data, *cplim;
1823
1824 /*
1825 * AF_LINK addresses can be looked up directly by their index number,
1826 * so do that if we can.
1827 */
1828 if (af == AF_LINK) {
1829 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1830 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1831 return (ifaddr_byindex(sdl->sdl_index));
1832 }
1833
1834 /*
1835 * Scan though each interface, looking for ones that have addresses
1836 * in this address family and the requested fib. Maintain a reference
1837 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1838 * kept it stable when we move onto the next interface.
1839 */
1840 IFNET_RLOCK_NOSLEEP();
1841 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1842 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1843 continue;
1844 IF_ADDR_RLOCK(ifp);
1845 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1846 char *cp, *cp2, *cp3;
1847
1848 if (ifa->ifa_addr->sa_family != af)
1849 next: continue;
1850 if (af == AF_INET &&
1851 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1852 /*
1853 * This is a bit broken as it doesn't
1854 * take into account that the remote end may
1855 * be a single node in the network we are
1856 * looking for.
1857 * The trouble is that we don't know the
1858 * netmask for the remote end.
1859 */
1860 if (ifa->ifa_dstaddr != NULL &&
1861 sa_equal(addr, ifa->ifa_dstaddr)) {
1862 ifa_ref(ifa);
1863 IF_ADDR_RUNLOCK(ifp);
1864 goto done;
1865 }
1866 } else {
1867 /*
1868 * if we have a special address handler,
1869 * then use it instead of the generic one.
1870 */
1871 if (ifa->ifa_claim_addr) {
1872 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1873 ifa_ref(ifa);
1874 IF_ADDR_RUNLOCK(ifp);
1875 goto done;
1876 }
1877 continue;
1878 }
1879
1880 /*
1881 * Scan all the bits in the ifa's address.
1882 * If a bit dissagrees with what we are
1883 * looking for, mask it with the netmask
1884 * to see if it really matters.
1885 * (A byte at a time)
1886 */
1887 if (ifa->ifa_netmask == 0)
1888 continue;
1889 cp = addr_data;
1890 cp2 = ifa->ifa_addr->sa_data;
1891 cp3 = ifa->ifa_netmask->sa_data;
1892 cplim = ifa->ifa_netmask->sa_len
1893 + (char *)ifa->ifa_netmask;
1894 while (cp3 < cplim)
1895 if ((*cp++ ^ *cp2++) & *cp3++)
1896 goto next; /* next address! */
1897 /*
1898 * If the netmask of what we just found
1899 * is more specific than what we had before
1900 * (if we had one), or if the virtual status
1901 * of new prefix is better than of the old one,
1902 * then remember the new one before continuing
1903 * to search for an even better one.
1904 */
1905 if (ifa_maybe == NULL ||
1906 ifa_preferred(ifa_maybe, ifa) ||
1907 rn_refines((caddr_t)ifa->ifa_netmask,
1908 (caddr_t)ifa_maybe->ifa_netmask)) {
1909 if (ifa_maybe != NULL)
1910 ifa_free(ifa_maybe);
1911 ifa_maybe = ifa;
1912 ifa_ref(ifa_maybe);
1913 }
1914 }
1915 }
1916 IF_ADDR_RUNLOCK(ifp);
1917 }
1918 ifa = ifa_maybe;
1919 ifa_maybe = NULL;
1920 done:
1921 IFNET_RUNLOCK_NOSLEEP();
1922 if (ifa_maybe != NULL)
1923 ifa_free(ifa_maybe);
1924 return (ifa);
1925 }
1926
1927 struct ifaddr *
ifa_ifwithnet(struct sockaddr * addr,int ignore_ptp)1928 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1929 {
1930
1931 return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
1932 }
1933
1934 /*
1935 * Find an interface address specific to an interface best matching
1936 * a given address.
1937 */
1938 struct ifaddr *
ifaof_ifpforaddr(struct sockaddr * addr,struct ifnet * ifp)1939 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1940 {
1941 struct ifaddr *ifa;
1942 char *cp, *cp2, *cp3;
1943 char *cplim;
1944 struct ifaddr *ifa_maybe = NULL;
1945 u_int af = addr->sa_family;
1946
1947 if (af >= AF_MAX)
1948 return (NULL);
1949 IF_ADDR_RLOCK(ifp);
1950 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1951 if (ifa->ifa_addr->sa_family != af)
1952 continue;
1953 if (ifa_maybe == NULL)
1954 ifa_maybe = ifa;
1955 if (ifa->ifa_netmask == 0) {
1956 if (sa_equal(addr, ifa->ifa_addr) ||
1957 (ifa->ifa_dstaddr &&
1958 sa_equal(addr, ifa->ifa_dstaddr)))
1959 goto done;
1960 continue;
1961 }
1962 if (ifp->if_flags & IFF_POINTOPOINT) {
1963 if (sa_equal(addr, ifa->ifa_dstaddr))
1964 goto done;
1965 } else {
1966 cp = addr->sa_data;
1967 cp2 = ifa->ifa_addr->sa_data;
1968 cp3 = ifa->ifa_netmask->sa_data;
1969 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1970 for (; cp3 < cplim; cp3++)
1971 if ((*cp++ ^ *cp2++) & *cp3)
1972 break;
1973 if (cp3 == cplim)
1974 goto done;
1975 }
1976 }
1977 ifa = ifa_maybe;
1978 done:
1979 if (ifa != NULL)
1980 ifa_ref(ifa);
1981 IF_ADDR_RUNLOCK(ifp);
1982 return (ifa);
1983 }
1984
1985 /*
1986 * See whether new ifa is better than current one:
1987 * 1) A non-virtual one is preferred over virtual.
1988 * 2) A virtual in master state preferred over any other state.
1989 *
1990 * Used in several address selecting functions.
1991 */
1992 int
ifa_preferred(struct ifaddr * cur,struct ifaddr * next)1993 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1994 {
1995
1996 return (cur->ifa_carp && (!next->ifa_carp ||
1997 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1998 }
1999
2000 #include <net/if_llatbl.h>
2001
2002 /*
2003 * Default action when installing a route with a Link Level gateway.
2004 * Lookup an appropriate real ifa to point to.
2005 * This should be moved to /sys/net/link.c eventually.
2006 */
2007 static void
link_rtrequest(int cmd,struct rtentry * rt,struct rt_addrinfo * info)2008 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2009 {
2010 struct ifaddr *ifa, *oifa;
2011 struct sockaddr *dst;
2012 struct ifnet *ifp;
2013
2014 RT_LOCK_ASSERT(rt);
2015
2016 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
2017 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
2018 return;
2019 ifa = ifaof_ifpforaddr(dst, ifp);
2020 if (ifa) {
2021 oifa = rt->rt_ifa;
2022 rt->rt_ifa = ifa;
2023 ifa_free(oifa);
2024 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2025 ifa->ifa_rtrequest(cmd, rt, info);
2026 }
2027 }
2028
2029 /*
2030 * Mark an interface down and notify protocols of
2031 * the transition.
2032 */
2033 static void
if_unroute(struct ifnet * ifp,int flag,int fam)2034 if_unroute(struct ifnet *ifp, int flag, int fam)
2035 {
2036 struct ifaddr *ifa;
2037
2038 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2039
2040 ifp->if_flags &= ~flag;
2041 getmicrotime(&ifp->if_lastchange);
2042 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2043 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2044 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2045 ifp->if_qflush(ifp);
2046
2047 if (ifp->if_carp)
2048 (*carp_linkstate_p)(ifp);
2049 rt_ifmsg(ifp);
2050 }
2051
2052 /*
2053 * Mark an interface up and notify protocols of
2054 * the transition.
2055 */
2056 static void
if_route(struct ifnet * ifp,int flag,int fam)2057 if_route(struct ifnet *ifp, int flag, int fam)
2058 {
2059 struct ifaddr *ifa;
2060
2061 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2062
2063 ifp->if_flags |= flag;
2064 getmicrotime(&ifp->if_lastchange);
2065 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2066 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2067 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2068 if (ifp->if_carp)
2069 (*carp_linkstate_p)(ifp);
2070 rt_ifmsg(ifp);
2071 #ifdef INET6
2072 in6_if_up(ifp);
2073 #endif
2074 }
2075
2076 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2077 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2078 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2079 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2080 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2081 int (*vlan_setcookie_p)(struct ifnet *, void *);
2082 void *(*vlan_cookie_p)(struct ifnet *);
2083
2084 /*
2085 * Handle a change in the interface link state. To avoid LORs
2086 * between driver lock and upper layer locks, as well as possible
2087 * recursions, we post event to taskqueue, and all job
2088 * is done in static do_link_state_change().
2089 */
2090 void
if_link_state_change(struct ifnet * ifp,int link_state)2091 if_link_state_change(struct ifnet *ifp, int link_state)
2092 {
2093 /* Return if state hasn't changed. */
2094 if (ifp->if_link_state == link_state)
2095 return;
2096
2097 ifp->if_link_state = link_state;
2098
2099 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2100 }
2101
2102 static void
do_link_state_change(void * arg,int pending)2103 do_link_state_change(void *arg, int pending)
2104 {
2105 struct ifnet *ifp = (struct ifnet *)arg;
2106 int link_state = ifp->if_link_state;
2107 CURVNET_SET(ifp->if_vnet);
2108
2109 /* Notify that the link state has changed. */
2110 rt_ifmsg(ifp);
2111 if (ifp->if_vlantrunk != NULL)
2112 (*vlan_link_state_p)(ifp);
2113
2114 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2115 IFP2AC(ifp)->ac_netgraph != NULL)
2116 (*ng_ether_link_state_p)(ifp, link_state);
2117 if (ifp->if_carp)
2118 (*carp_linkstate_p)(ifp);
2119 if (ifp->if_bridge)
2120 (*bridge_linkstate_p)(ifp);
2121 if (ifp->if_lagg)
2122 (*lagg_linkstate_p)(ifp, link_state);
2123
2124 if (IS_DEFAULT_VNET(curvnet))
2125 devctl_notify("IFNET", ifp->if_xname,
2126 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2127 NULL);
2128 if (pending > 1)
2129 if_printf(ifp, "%d link states coalesced\n", pending);
2130 if (log_link_state_change)
2131 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2132 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2133 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2134 CURVNET_RESTORE();
2135 }
2136
2137 /*
2138 * Mark an interface down and notify protocols of
2139 * the transition.
2140 */
2141 void
if_down(struct ifnet * ifp)2142 if_down(struct ifnet *ifp)
2143 {
2144
2145 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2146 }
2147
2148 /*
2149 * Mark an interface up and notify protocols of
2150 * the transition.
2151 */
2152 void
if_up(struct ifnet * ifp)2153 if_up(struct ifnet *ifp)
2154 {
2155
2156 if_route(ifp, IFF_UP, AF_UNSPEC);
2157 }
2158
2159 /*
2160 * Flush an interface queue.
2161 */
2162 void
if_qflush(struct ifnet * ifp)2163 if_qflush(struct ifnet *ifp)
2164 {
2165 struct mbuf *m, *n;
2166 struct ifaltq *ifq;
2167
2168 ifq = &ifp->if_snd;
2169 IFQ_LOCK(ifq);
2170 #ifdef ALTQ
2171 if (ALTQ_IS_ENABLED(ifq))
2172 ALTQ_PURGE(ifq);
2173 #endif
2174 n = ifq->ifq_head;
2175 while ((m = n) != 0) {
2176 n = m->m_nextpkt;
2177 m_freem(m);
2178 }
2179 ifq->ifq_head = 0;
2180 ifq->ifq_tail = 0;
2181 ifq->ifq_len = 0;
2182 IFQ_UNLOCK(ifq);
2183 }
2184
2185 /*
2186 * Map interface name to interface structure pointer, with or without
2187 * returning a reference.
2188 */
2189 struct ifnet *
ifunit_ref(const char * name)2190 ifunit_ref(const char *name)
2191 {
2192 struct ifnet *ifp;
2193
2194 IFNET_RLOCK_NOSLEEP();
2195 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2196 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2197 !(ifp->if_flags & IFF_DYING))
2198 break;
2199 }
2200 if (ifp != NULL)
2201 if_ref(ifp);
2202 IFNET_RUNLOCK_NOSLEEP();
2203 return (ifp);
2204 }
2205
2206 struct ifnet *
ifunit(const char * name)2207 ifunit(const char *name)
2208 {
2209 struct ifnet *ifp;
2210
2211 IFNET_RLOCK_NOSLEEP();
2212 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2213 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2214 break;
2215 }
2216 IFNET_RUNLOCK_NOSLEEP();
2217 return (ifp);
2218 }
2219
2220 /*
2221 * Hardware specific interface ioctls.
2222 */
2223 static int
ifhwioctl(u_long cmd,struct ifnet * ifp,caddr_t data,struct thread * td)2224 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2225 {
2226 struct ifreq *ifr;
2227 struct ifstat *ifs;
2228 int error = 0;
2229 int new_flags, temp_flags;
2230 size_t namelen, onamelen;
2231 size_t descrlen;
2232 char *descrbuf, *odescrbuf;
2233 char new_name[IFNAMSIZ];
2234 struct ifaddr *ifa;
2235 struct sockaddr_dl *sdl;
2236
2237 ifr = (struct ifreq *)data;
2238 switch (cmd) {
2239 case SIOCGIFINDEX:
2240 ifr->ifr_index = ifp->if_index;
2241 break;
2242
2243 case SIOCGIFFLAGS:
2244 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2245 ifr->ifr_flags = temp_flags & 0xffff;
2246 ifr->ifr_flagshigh = temp_flags >> 16;
2247 break;
2248
2249 case SIOCGIFCAP:
2250 ifr->ifr_reqcap = ifp->if_capabilities;
2251 ifr->ifr_curcap = ifp->if_capenable;
2252 break;
2253
2254 #ifdef MAC
2255 case SIOCGIFMAC:
2256 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2257 break;
2258 #endif
2259
2260 case SIOCGIFMETRIC:
2261 ifr->ifr_metric = ifp->if_metric;
2262 break;
2263
2264 case SIOCGIFMTU:
2265 ifr->ifr_mtu = ifp->if_mtu;
2266 break;
2267
2268 case SIOCGIFPHYS:
2269 ifr->ifr_phys = ifp->if_physical;
2270 break;
2271
2272 case SIOCGIFDESCR:
2273 error = 0;
2274 sx_slock(&ifdescr_sx);
2275 if (ifp->if_description == NULL)
2276 error = ENOMSG;
2277 else {
2278 /* space for terminating nul */
2279 descrlen = strlen(ifp->if_description) + 1;
2280 if (ifr->ifr_buffer.length < descrlen)
2281 ifr->ifr_buffer.buffer = NULL;
2282 else
2283 error = copyout(ifp->if_description,
2284 ifr->ifr_buffer.buffer, descrlen);
2285 ifr->ifr_buffer.length = descrlen;
2286 }
2287 sx_sunlock(&ifdescr_sx);
2288 break;
2289
2290 case SIOCSIFDESCR:
2291 error = priv_check(td, PRIV_NET_SETIFDESCR);
2292 if (error)
2293 return (error);
2294
2295 /*
2296 * Copy only (length-1) bytes to make sure that
2297 * if_description is always nul terminated. The
2298 * length parameter is supposed to count the
2299 * terminating nul in.
2300 */
2301 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2302 return (ENAMETOOLONG);
2303 else if (ifr->ifr_buffer.length == 0)
2304 descrbuf = NULL;
2305 else {
2306 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2307 M_WAITOK | M_ZERO);
2308 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2309 ifr->ifr_buffer.length - 1);
2310 if (error) {
2311 free(descrbuf, M_IFDESCR);
2312 break;
2313 }
2314 }
2315
2316 sx_xlock(&ifdescr_sx);
2317 odescrbuf = ifp->if_description;
2318 ifp->if_description = descrbuf;
2319 sx_xunlock(&ifdescr_sx);
2320
2321 getmicrotime(&ifp->if_lastchange);
2322 free(odescrbuf, M_IFDESCR);
2323 break;
2324
2325 case SIOCGIFFIB:
2326 ifr->ifr_fib = ifp->if_fib;
2327 break;
2328
2329 case SIOCSIFFIB:
2330 error = priv_check(td, PRIV_NET_SETIFFIB);
2331 if (error)
2332 return (error);
2333 if (ifr->ifr_fib >= rt_numfibs)
2334 return (EINVAL);
2335
2336 ifp->if_fib = ifr->ifr_fib;
2337 break;
2338
2339 case SIOCSIFFLAGS:
2340 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2341 if (error)
2342 return (error);
2343 /*
2344 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2345 * check, so we don't need special handling here yet.
2346 */
2347 new_flags = (ifr->ifr_flags & 0xffff) |
2348 (ifr->ifr_flagshigh << 16);
2349 if (ifp->if_flags & IFF_SMART) {
2350 /* Smart drivers twiddle their own routes */
2351 } else if (ifp->if_flags & IFF_UP &&
2352 (new_flags & IFF_UP) == 0) {
2353 if_down(ifp);
2354 } else if (new_flags & IFF_UP &&
2355 (ifp->if_flags & IFF_UP) == 0) {
2356 if_up(ifp);
2357 }
2358 /* See if permanently promiscuous mode bit is about to flip */
2359 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2360 if (new_flags & IFF_PPROMISC)
2361 ifp->if_flags |= IFF_PROMISC;
2362 else if (ifp->if_pcount == 0)
2363 ifp->if_flags &= ~IFF_PROMISC;
2364 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2365 ifp->if_xname,
2366 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2367 }
2368 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2369 (new_flags &~ IFF_CANTCHANGE);
2370 if (ifp->if_ioctl) {
2371 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2372 }
2373 getmicrotime(&ifp->if_lastchange);
2374 break;
2375
2376 case SIOCSIFCAP:
2377 error = priv_check(td, PRIV_NET_SETIFCAP);
2378 if (error)
2379 return (error);
2380 if (ifp->if_ioctl == NULL)
2381 return (EOPNOTSUPP);
2382 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2383 return (EINVAL);
2384 error = (*ifp->if_ioctl)(ifp, cmd, data);
2385 if (error == 0)
2386 getmicrotime(&ifp->if_lastchange);
2387 break;
2388
2389 #ifdef MAC
2390 case SIOCSIFMAC:
2391 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2392 break;
2393 #endif
2394
2395 case SIOCSIFNAME:
2396 error = priv_check(td, PRIV_NET_SETIFNAME);
2397 if (error)
2398 return (error);
2399 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2400 if (error != 0)
2401 return (error);
2402 if (new_name[0] == '\0')
2403 return (EINVAL);
2404 if (ifunit(new_name) != NULL)
2405 return (EEXIST);
2406
2407 /*
2408 * XXX: Locking. Nothing else seems to lock if_flags,
2409 * and there are numerous other races with the
2410 * ifunit() checks not being atomic with namespace
2411 * changes (renames, vmoves, if_attach, etc).
2412 */
2413 ifp->if_flags |= IFF_RENAMING;
2414
2415 /* Announce the departure of the interface. */
2416 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2417 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2418
2419 log(LOG_INFO, "%s: changing name to '%s'\n",
2420 ifp->if_xname, new_name);
2421
2422 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2423 ifa = ifp->if_addr;
2424 IFA_LOCK(ifa);
2425 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2426 namelen = strlen(new_name);
2427 onamelen = sdl->sdl_nlen;
2428 /*
2429 * Move the address if needed. This is safe because we
2430 * allocate space for a name of length IFNAMSIZ when we
2431 * create this in if_attach().
2432 */
2433 if (namelen != onamelen) {
2434 bcopy(sdl->sdl_data + onamelen,
2435 sdl->sdl_data + namelen, sdl->sdl_alen);
2436 }
2437 bcopy(new_name, sdl->sdl_data, namelen);
2438 sdl->sdl_nlen = namelen;
2439 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2440 bzero(sdl->sdl_data, onamelen);
2441 while (namelen != 0)
2442 sdl->sdl_data[--namelen] = 0xff;
2443 IFA_UNLOCK(ifa);
2444
2445 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2446 /* Announce the return of the interface. */
2447 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2448
2449 ifp->if_flags &= ~IFF_RENAMING;
2450 break;
2451
2452 #ifdef VIMAGE
2453 case SIOCSIFVNET:
2454 error = priv_check(td, PRIV_NET_SETIFVNET);
2455 if (error)
2456 return (error);
2457 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2458 break;
2459 #endif
2460
2461 case SIOCSIFMETRIC:
2462 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2463 if (error)
2464 return (error);
2465 ifp->if_metric = ifr->ifr_metric;
2466 getmicrotime(&ifp->if_lastchange);
2467 break;
2468
2469 case SIOCSIFPHYS:
2470 error = priv_check(td, PRIV_NET_SETIFPHYS);
2471 if (error)
2472 return (error);
2473 if (ifp->if_ioctl == NULL)
2474 return (EOPNOTSUPP);
2475 error = (*ifp->if_ioctl)(ifp, cmd, data);
2476 if (error == 0)
2477 getmicrotime(&ifp->if_lastchange);
2478 break;
2479
2480 case SIOCSIFMTU:
2481 {
2482 u_long oldmtu = ifp->if_mtu;
2483
2484 error = priv_check(td, PRIV_NET_SETIFMTU);
2485 if (error)
2486 return (error);
2487 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2488 return (EINVAL);
2489 if (ifp->if_ioctl == NULL)
2490 return (EOPNOTSUPP);
2491 error = (*ifp->if_ioctl)(ifp, cmd, data);
2492 if (error == 0) {
2493 getmicrotime(&ifp->if_lastchange);
2494 rt_ifmsg(ifp);
2495 }
2496 /*
2497 * If the link MTU changed, do network layer specific procedure.
2498 */
2499 if (ifp->if_mtu != oldmtu) {
2500 #ifdef INET6
2501 nd6_setmtu(ifp);
2502 #endif
2503 }
2504 break;
2505 }
2506
2507 case SIOCADDMULTI:
2508 case SIOCDELMULTI:
2509 if (cmd == SIOCADDMULTI)
2510 error = priv_check(td, PRIV_NET_ADDMULTI);
2511 else
2512 error = priv_check(td, PRIV_NET_DELMULTI);
2513 if (error)
2514 return (error);
2515
2516 /* Don't allow group membership on non-multicast interfaces. */
2517 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2518 return (EOPNOTSUPP);
2519
2520 /* Don't let users screw up protocols' entries. */
2521 if (ifr->ifr_addr.sa_family != AF_LINK)
2522 return (EINVAL);
2523
2524 if (cmd == SIOCADDMULTI) {
2525 struct ifmultiaddr *ifma;
2526
2527 /*
2528 * Userland is only permitted to join groups once
2529 * via the if_addmulti() KPI, because it cannot hold
2530 * struct ifmultiaddr * between calls. It may also
2531 * lose a race while we check if the membership
2532 * already exists.
2533 */
2534 IF_ADDR_RLOCK(ifp);
2535 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2536 IF_ADDR_RUNLOCK(ifp);
2537 if (ifma != NULL)
2538 error = EADDRINUSE;
2539 else
2540 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2541 } else {
2542 error = if_delmulti(ifp, &ifr->ifr_addr);
2543 }
2544 if (error == 0)
2545 getmicrotime(&ifp->if_lastchange);
2546 break;
2547
2548 case SIOCSIFPHYADDR:
2549 case SIOCDIFPHYADDR:
2550 #ifdef INET6
2551 case SIOCSIFPHYADDR_IN6:
2552 #endif
2553 case SIOCSLIFPHYADDR:
2554 case SIOCSIFMEDIA:
2555 case SIOCSIFGENERIC:
2556 error = priv_check(td, PRIV_NET_HWIOCTL);
2557 if (error)
2558 return (error);
2559 if (ifp->if_ioctl == NULL)
2560 return (EOPNOTSUPP);
2561 error = (*ifp->if_ioctl)(ifp, cmd, data);
2562 if (error == 0)
2563 getmicrotime(&ifp->if_lastchange);
2564 break;
2565
2566 case SIOCGIFSTATUS:
2567 ifs = (struct ifstat *)data;
2568 ifs->ascii[0] = '\0';
2569
2570 case SIOCGIFPSRCADDR:
2571 case SIOCGIFPDSTADDR:
2572 case SIOCGLIFPHYADDR:
2573 case SIOCGIFMEDIA:
2574 case SIOCGIFXMEDIA:
2575 case SIOCGIFGENERIC:
2576 if (ifp->if_ioctl == NULL)
2577 return (EOPNOTSUPP);
2578 error = (*ifp->if_ioctl)(ifp, cmd, data);
2579 break;
2580
2581 case SIOCSIFLLADDR:
2582 error = priv_check(td, PRIV_NET_SETLLADDR);
2583 if (error)
2584 return (error);
2585 error = if_setlladdr(ifp,
2586 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2587 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2588 break;
2589
2590 case SIOCAIFGROUP:
2591 {
2592 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2593
2594 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2595 if (error)
2596 return (error);
2597 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2598 return (error);
2599 break;
2600 }
2601
2602 case SIOCGIFGROUP:
2603 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2604 return (error);
2605 break;
2606
2607 case SIOCDIFGROUP:
2608 {
2609 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2610
2611 error = priv_check(td, PRIV_NET_DELIFGROUP);
2612 if (error)
2613 return (error);
2614 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2615 return (error);
2616 break;
2617 }
2618
2619 default:
2620 error = ENOIOCTL;
2621 break;
2622 }
2623 return (error);
2624 }
2625
2626 #ifdef COMPAT_FREEBSD32
2627 struct ifconf32 {
2628 int32_t ifc_len;
2629 union {
2630 uint32_t ifcu_buf;
2631 uint32_t ifcu_req;
2632 } ifc_ifcu;
2633 };
2634 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2635 #endif
2636
2637 /*
2638 * Interface ioctls.
2639 */
2640 int
ifioctl(struct socket * so,u_long cmd,caddr_t data,struct thread * td)2641 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2642 {
2643 struct ifnet *ifp;
2644 struct ifreq *ifr;
2645 int error;
2646 int oif_flags;
2647
2648 CURVNET_SET(so->so_vnet);
2649 switch (cmd) {
2650 case SIOCGIFCONF:
2651 case OSIOCGIFCONF:
2652 error = ifconf(cmd, data);
2653 CURVNET_RESTORE();
2654 return (error);
2655
2656 #ifdef COMPAT_FREEBSD32
2657 case SIOCGIFCONF32:
2658 {
2659 struct ifconf32 *ifc32;
2660 struct ifconf ifc;
2661
2662 ifc32 = (struct ifconf32 *)data;
2663 ifc.ifc_len = ifc32->ifc_len;
2664 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2665
2666 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2667 CURVNET_RESTORE();
2668 if (error == 0)
2669 ifc32->ifc_len = ifc.ifc_len;
2670 return (error);
2671 }
2672 #endif
2673 }
2674 ifr = (struct ifreq *)data;
2675
2676 switch (cmd) {
2677 #ifdef VIMAGE
2678 case SIOCSIFRVNET:
2679 error = priv_check(td, PRIV_NET_SETIFVNET);
2680 if (error == 0)
2681 error = if_vmove_reclaim(td, ifr->ifr_name,
2682 ifr->ifr_jid);
2683 CURVNET_RESTORE();
2684 return (error);
2685 #endif
2686 case SIOCIFCREATE:
2687 case SIOCIFCREATE2:
2688 error = priv_check(td, PRIV_NET_IFCREATE);
2689 if (error == 0)
2690 error = if_clone_create(ifr->ifr_name,
2691 sizeof(ifr->ifr_name),
2692 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2693 CURVNET_RESTORE();
2694 return (error);
2695 case SIOCIFDESTROY:
2696 error = priv_check(td, PRIV_NET_IFDESTROY);
2697 if (error == 0)
2698 error = if_clone_destroy(ifr->ifr_name);
2699 CURVNET_RESTORE();
2700 return (error);
2701
2702 case SIOCIFGCLONERS:
2703 error = if_clone_list((struct if_clonereq *)data);
2704 CURVNET_RESTORE();
2705 return (error);
2706 case SIOCGIFGMEMB:
2707 error = if_getgroupmembers((struct ifgroupreq *)data);
2708 CURVNET_RESTORE();
2709 return (error);
2710 #if defined(INET) || defined(INET6)
2711 case SIOCSVH:
2712 case SIOCGVH:
2713 if (carp_ioctl_p == NULL)
2714 error = EPROTONOSUPPORT;
2715 else
2716 error = (*carp_ioctl_p)(ifr, cmd, td);
2717 CURVNET_RESTORE();
2718 return (error);
2719 #endif
2720 }
2721
2722 ifp = ifunit_ref(ifr->ifr_name);
2723 if (ifp == NULL) {
2724 CURVNET_RESTORE();
2725 return (ENXIO);
2726 }
2727
2728 error = ifhwioctl(cmd, ifp, data, td);
2729 if (error != ENOIOCTL) {
2730 if_rele(ifp);
2731 CURVNET_RESTORE();
2732 return (error);
2733 }
2734
2735 oif_flags = ifp->if_flags;
2736 if (so->so_proto == NULL) {
2737 if_rele(ifp);
2738 CURVNET_RESTORE();
2739 return (EOPNOTSUPP);
2740 }
2741
2742 /*
2743 * Pass the request on to the socket control method, and if the
2744 * latter returns EOPNOTSUPP, directly to the interface.
2745 *
2746 * Make an exception for the legacy SIOCSIF* requests. Drivers
2747 * trust SIOCSIFADDR et al to come from an already privileged
2748 * layer, and do not perform any credentials checks or input
2749 * validation.
2750 */
2751 #ifndef COMPAT_43
2752 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2753 data,
2754 ifp, td));
2755 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2756 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2757 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2758 error = (*ifp->if_ioctl)(ifp, cmd, data);
2759 #else
2760 {
2761 u_long ocmd = cmd;
2762
2763 switch (cmd) {
2764
2765 case SIOCSIFDSTADDR:
2766 case SIOCSIFADDR:
2767 case SIOCSIFBRDADDR:
2768 case SIOCSIFNETMASK:
2769 #if BYTE_ORDER != BIG_ENDIAN
2770 if (ifr->ifr_addr.sa_family == 0 &&
2771 ifr->ifr_addr.sa_len < 16) {
2772 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2773 ifr->ifr_addr.sa_len = 16;
2774 }
2775 #else
2776 if (ifr->ifr_addr.sa_len == 0)
2777 ifr->ifr_addr.sa_len = 16;
2778 #endif
2779 break;
2780
2781 case OSIOCGIFADDR:
2782 cmd = SIOCGIFADDR;
2783 break;
2784
2785 case OSIOCGIFDSTADDR:
2786 cmd = SIOCGIFDSTADDR;
2787 break;
2788
2789 case OSIOCGIFBRDADDR:
2790 cmd = SIOCGIFBRDADDR;
2791 break;
2792
2793 case OSIOCGIFNETMASK:
2794 cmd = SIOCGIFNETMASK;
2795 }
2796 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2797 cmd,
2798 data,
2799 ifp, td));
2800 if (error == EOPNOTSUPP && ifp != NULL &&
2801 ifp->if_ioctl != NULL &&
2802 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2803 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2804 error = (*ifp->if_ioctl)(ifp, cmd, data);
2805 switch (ocmd) {
2806
2807 case OSIOCGIFADDR:
2808 case OSIOCGIFDSTADDR:
2809 case OSIOCGIFBRDADDR:
2810 case OSIOCGIFNETMASK:
2811 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2812
2813 }
2814 }
2815 #endif /* COMPAT_43 */
2816
2817 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2818 #ifdef INET6
2819 if (ifp->if_flags & IFF_UP)
2820 in6_if_up(ifp);
2821 #endif
2822 }
2823 if_rele(ifp);
2824 CURVNET_RESTORE();
2825 return (error);
2826 }
2827
2828 /*
2829 * The code common to handling reference counted flags,
2830 * e.g., in ifpromisc() and if_allmulti().
2831 * The "pflag" argument can specify a permanent mode flag to check,
2832 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2833 *
2834 * Only to be used on stack-owned flags, not driver-owned flags.
2835 */
2836 static int
if_setflag(struct ifnet * ifp,int flag,int pflag,int * refcount,int onswitch)2837 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2838 {
2839 struct ifreq ifr;
2840 int error;
2841 int oldflags, oldcount;
2842
2843 /* Sanity checks to catch programming errors */
2844 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2845 ("%s: setting driver-owned flag %d", __func__, flag));
2846
2847 if (onswitch)
2848 KASSERT(*refcount >= 0,
2849 ("%s: increment negative refcount %d for flag %d",
2850 __func__, *refcount, flag));
2851 else
2852 KASSERT(*refcount > 0,
2853 ("%s: decrement non-positive refcount %d for flag %d",
2854 __func__, *refcount, flag));
2855
2856 /* In case this mode is permanent, just touch refcount */
2857 if (ifp->if_flags & pflag) {
2858 *refcount += onswitch ? 1 : -1;
2859 return (0);
2860 }
2861
2862 /* Save ifnet parameters for if_ioctl() may fail */
2863 oldcount = *refcount;
2864 oldflags = ifp->if_flags;
2865
2866 /*
2867 * See if we aren't the only and touching refcount is enough.
2868 * Actually toggle interface flag if we are the first or last.
2869 */
2870 if (onswitch) {
2871 if ((*refcount)++)
2872 return (0);
2873 ifp->if_flags |= flag;
2874 } else {
2875 if (--(*refcount))
2876 return (0);
2877 ifp->if_flags &= ~flag;
2878 }
2879
2880 /* Call down the driver since we've changed interface flags */
2881 if (ifp->if_ioctl == NULL) {
2882 error = EOPNOTSUPP;
2883 goto recover;
2884 }
2885 ifr.ifr_flags = ifp->if_flags & 0xffff;
2886 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2887 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2888 if (error)
2889 goto recover;
2890 /* Notify userland that interface flags have changed */
2891 rt_ifmsg(ifp);
2892 return (0);
2893
2894 recover:
2895 /* Recover after driver error */
2896 *refcount = oldcount;
2897 ifp->if_flags = oldflags;
2898 return (error);
2899 }
2900
2901 /*
2902 * Set/clear promiscuous mode on interface ifp based on the truth value
2903 * of pswitch. The calls are reference counted so that only the first
2904 * "on" request actually has an effect, as does the final "off" request.
2905 * Results are undefined if the "off" and "on" requests are not matched.
2906 */
2907 int
ifpromisc(struct ifnet * ifp,int pswitch)2908 ifpromisc(struct ifnet *ifp, int pswitch)
2909 {
2910 int error;
2911 int oldflags = ifp->if_flags;
2912
2913 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2914 &ifp->if_pcount, pswitch);
2915 /* If promiscuous mode status has changed, log a message */
2916 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2917 log(LOG_INFO, "%s: promiscuous mode %s\n",
2918 ifp->if_xname,
2919 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2920 return (error);
2921 }
2922
2923 /*
2924 * Return interface configuration
2925 * of system. List may be used
2926 * in later ioctl's (above) to get
2927 * other information.
2928 */
2929 /*ARGSUSED*/
2930 static int
ifconf(u_long cmd,caddr_t data)2931 ifconf(u_long cmd, caddr_t data)
2932 {
2933 struct ifconf *ifc = (struct ifconf *)data;
2934 struct ifnet *ifp;
2935 struct ifaddr *ifa;
2936 struct ifreq ifr;
2937 struct sbuf *sb;
2938 int error, full = 0, valid_len, max_len;
2939
2940 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2941 max_len = MAXPHYS - 1;
2942
2943 /* Prevent hostile input from being able to crash the system */
2944 if (ifc->ifc_len <= 0)
2945 return (EINVAL);
2946
2947 again:
2948 if (ifc->ifc_len <= max_len) {
2949 max_len = ifc->ifc_len;
2950 full = 1;
2951 }
2952 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2953 max_len = 0;
2954 valid_len = 0;
2955
2956 IFNET_RLOCK();
2957 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2958 int addrs;
2959
2960 /*
2961 * Zero the ifr_name buffer to make sure we don't
2962 * disclose the contents of the stack.
2963 */
2964 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2965
2966 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2967 >= sizeof(ifr.ifr_name)) {
2968 sbuf_delete(sb);
2969 IFNET_RUNLOCK();
2970 return (ENAMETOOLONG);
2971 }
2972
2973 addrs = 0;
2974 IF_ADDR_RLOCK(ifp);
2975 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2976 struct sockaddr *sa = ifa->ifa_addr;
2977
2978 if (prison_if(curthread->td_ucred, sa) != 0)
2979 continue;
2980 addrs++;
2981 #ifdef COMPAT_43
2982 if (cmd == OSIOCGIFCONF) {
2983 struct osockaddr *osa =
2984 (struct osockaddr *)&ifr.ifr_addr;
2985 ifr.ifr_addr = *sa;
2986 osa->sa_family = sa->sa_family;
2987 sbuf_bcat(sb, &ifr, sizeof(ifr));
2988 max_len += sizeof(ifr);
2989 } else
2990 #endif
2991 if (sa->sa_len <= sizeof(*sa)) {
2992 ifr.ifr_addr = *sa;
2993 sbuf_bcat(sb, &ifr, sizeof(ifr));
2994 max_len += sizeof(ifr);
2995 } else {
2996 sbuf_bcat(sb, &ifr,
2997 offsetof(struct ifreq, ifr_addr));
2998 max_len += offsetof(struct ifreq, ifr_addr);
2999 sbuf_bcat(sb, sa, sa->sa_len);
3000 max_len += sa->sa_len;
3001 }
3002
3003 if (sbuf_error(sb) == 0)
3004 valid_len = sbuf_len(sb);
3005 }
3006 IF_ADDR_RUNLOCK(ifp);
3007 if (addrs == 0) {
3008 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3009 sbuf_bcat(sb, &ifr, sizeof(ifr));
3010 max_len += sizeof(ifr);
3011
3012 if (sbuf_error(sb) == 0)
3013 valid_len = sbuf_len(sb);
3014 }
3015 }
3016 IFNET_RUNLOCK();
3017
3018 /*
3019 * If we didn't allocate enough space (uncommon), try again. If
3020 * we have already allocated as much space as we are allowed,
3021 * return what we've got.
3022 */
3023 if (valid_len != max_len && !full) {
3024 sbuf_delete(sb);
3025 goto again;
3026 }
3027
3028 ifc->ifc_len = valid_len;
3029 sbuf_finish(sb);
3030 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3031 sbuf_delete(sb);
3032 return (error);
3033 }
3034
3035 /*
3036 * Just like ifpromisc(), but for all-multicast-reception mode.
3037 */
3038 int
if_allmulti(struct ifnet * ifp,int onswitch)3039 if_allmulti(struct ifnet *ifp, int onswitch)
3040 {
3041
3042 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3043 }
3044
3045 struct ifmultiaddr *
if_findmulti(struct ifnet * ifp,struct sockaddr * sa)3046 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
3047 {
3048 struct ifmultiaddr *ifma;
3049
3050 IF_ADDR_LOCK_ASSERT(ifp);
3051
3052 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3053 if (sa->sa_family == AF_LINK) {
3054 if (sa_dl_equal(ifma->ifma_addr, sa))
3055 break;
3056 } else {
3057 if (sa_equal(ifma->ifma_addr, sa))
3058 break;
3059 }
3060 }
3061
3062 return ifma;
3063 }
3064
3065 /*
3066 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3067 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3068 * the ifnet multicast address list here, so the caller must do that and
3069 * other setup work (such as notifying the device driver). The reference
3070 * count is initialized to 1.
3071 */
3072 static struct ifmultiaddr *
if_allocmulti(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr * llsa,int mflags)3073 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3074 int mflags)
3075 {
3076 struct ifmultiaddr *ifma;
3077 struct sockaddr *dupsa;
3078
3079 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3080 M_ZERO);
3081 if (ifma == NULL)
3082 return (NULL);
3083
3084 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3085 if (dupsa == NULL) {
3086 free(ifma, M_IFMADDR);
3087 return (NULL);
3088 }
3089 bcopy(sa, dupsa, sa->sa_len);
3090 ifma->ifma_addr = dupsa;
3091
3092 ifma->ifma_ifp = ifp;
3093 ifma->ifma_refcount = 1;
3094 ifma->ifma_protospec = NULL;
3095
3096 if (llsa == NULL) {
3097 ifma->ifma_lladdr = NULL;
3098 return (ifma);
3099 }
3100
3101 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3102 if (dupsa == NULL) {
3103 free(ifma->ifma_addr, M_IFMADDR);
3104 free(ifma, M_IFMADDR);
3105 return (NULL);
3106 }
3107 bcopy(llsa, dupsa, llsa->sa_len);
3108 ifma->ifma_lladdr = dupsa;
3109
3110 return (ifma);
3111 }
3112
3113 /*
3114 * if_freemulti: free ifmultiaddr structure and possibly attached related
3115 * addresses. The caller is responsible for implementing reference
3116 * counting, notifying the driver, handling routing messages, and releasing
3117 * any dependent link layer state.
3118 */
3119 static void
if_freemulti(struct ifmultiaddr * ifma)3120 if_freemulti(struct ifmultiaddr *ifma)
3121 {
3122
3123 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3124 ifma->ifma_refcount));
3125 KASSERT(ifma->ifma_protospec == NULL,
3126 ("if_freemulti: protospec not NULL"));
3127
3128 if (ifma->ifma_lladdr != NULL)
3129 free(ifma->ifma_lladdr, M_IFMADDR);
3130 free(ifma->ifma_addr, M_IFMADDR);
3131 free(ifma, M_IFMADDR);
3132 }
3133
3134 /*
3135 * Register an additional multicast address with a network interface.
3136 *
3137 * - If the address is already present, bump the reference count on the
3138 * address and return.
3139 * - If the address is not link-layer, look up a link layer address.
3140 * - Allocate address structures for one or both addresses, and attach to the
3141 * multicast address list on the interface. If automatically adding a link
3142 * layer address, the protocol address will own a reference to the link
3143 * layer address, to be freed when it is freed.
3144 * - Notify the network device driver of an addition to the multicast address
3145 * list.
3146 *
3147 * 'sa' points to caller-owned memory with the desired multicast address.
3148 *
3149 * 'retifma' will be used to return a pointer to the resulting multicast
3150 * address reference, if desired.
3151 */
3152 int
if_addmulti(struct ifnet * ifp,struct sockaddr * sa,struct ifmultiaddr ** retifma)3153 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3154 struct ifmultiaddr **retifma)
3155 {
3156 struct ifmultiaddr *ifma, *ll_ifma;
3157 struct sockaddr *llsa;
3158 int error;
3159
3160 /*
3161 * If the address is already present, return a new reference to it;
3162 * otherwise, allocate storage and set up a new address.
3163 */
3164 IF_ADDR_WLOCK(ifp);
3165 ifma = if_findmulti(ifp, sa);
3166 if (ifma != NULL) {
3167 ifma->ifma_refcount++;
3168 if (retifma != NULL)
3169 *retifma = ifma;
3170 IF_ADDR_WUNLOCK(ifp);
3171 return (0);
3172 }
3173
3174 /*
3175 * The address isn't already present; resolve the protocol address
3176 * into a link layer address, and then look that up, bump its
3177 * refcount or allocate an ifma for that also. If 'llsa' was
3178 * returned, we will need to free it later.
3179 */
3180 llsa = NULL;
3181 ll_ifma = NULL;
3182 if (ifp->if_resolvemulti != NULL) {
3183 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3184 if (error)
3185 goto unlock_out;
3186 }
3187
3188 /*
3189 * Allocate the new address. Don't hook it up yet, as we may also
3190 * need to allocate a link layer multicast address.
3191 */
3192 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3193 if (ifma == NULL) {
3194 error = ENOMEM;
3195 goto free_llsa_out;
3196 }
3197
3198 /*
3199 * If a link layer address is found, we'll need to see if it's
3200 * already present in the address list, or allocate is as well.
3201 * When this block finishes, the link layer address will be on the
3202 * list.
3203 */
3204 if (llsa != NULL) {
3205 ll_ifma = if_findmulti(ifp, llsa);
3206 if (ll_ifma == NULL) {
3207 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3208 if (ll_ifma == NULL) {
3209 --ifma->ifma_refcount;
3210 if_freemulti(ifma);
3211 error = ENOMEM;
3212 goto free_llsa_out;
3213 }
3214 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3215 ifma_link);
3216 } else
3217 ll_ifma->ifma_refcount++;
3218 ifma->ifma_llifma = ll_ifma;
3219 }
3220
3221 /*
3222 * We now have a new multicast address, ifma, and possibly a new or
3223 * referenced link layer address. Add the primary address to the
3224 * ifnet address list.
3225 */
3226 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3227
3228 if (retifma != NULL)
3229 *retifma = ifma;
3230
3231 /*
3232 * Must generate the message while holding the lock so that 'ifma'
3233 * pointer is still valid.
3234 */
3235 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3236 IF_ADDR_WUNLOCK(ifp);
3237
3238 /*
3239 * We are certain we have added something, so call down to the
3240 * interface to let them know about it.
3241 */
3242 if (ifp->if_ioctl != NULL) {
3243 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3244 }
3245
3246 if (llsa != NULL)
3247 free(llsa, M_IFMADDR);
3248
3249 return (0);
3250
3251 free_llsa_out:
3252 if (llsa != NULL)
3253 free(llsa, M_IFMADDR);
3254
3255 unlock_out:
3256 IF_ADDR_WUNLOCK(ifp);
3257 return (error);
3258 }
3259
3260 /*
3261 * Delete a multicast group membership by network-layer group address.
3262 *
3263 * Returns ENOENT if the entry could not be found. If ifp no longer
3264 * exists, results are undefined. This entry point should only be used
3265 * from subsystems which do appropriate locking to hold ifp for the
3266 * duration of the call.
3267 * Network-layer protocol domains must use if_delmulti_ifma().
3268 */
3269 int
if_delmulti(struct ifnet * ifp,struct sockaddr * sa)3270 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3271 {
3272 struct ifmultiaddr *ifma;
3273 int lastref;
3274 #ifdef INVARIANTS
3275 struct ifnet *oifp;
3276
3277 IFNET_RLOCK_NOSLEEP();
3278 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3279 if (ifp == oifp)
3280 break;
3281 if (ifp != oifp)
3282 ifp = NULL;
3283 IFNET_RUNLOCK_NOSLEEP();
3284
3285 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3286 #endif
3287 if (ifp == NULL)
3288 return (ENOENT);
3289
3290 IF_ADDR_WLOCK(ifp);
3291 lastref = 0;
3292 ifma = if_findmulti(ifp, sa);
3293 if (ifma != NULL)
3294 lastref = if_delmulti_locked(ifp, ifma, 0);
3295 IF_ADDR_WUNLOCK(ifp);
3296
3297 if (ifma == NULL)
3298 return (ENOENT);
3299
3300 if (lastref && ifp->if_ioctl != NULL) {
3301 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3302 }
3303
3304 return (0);
3305 }
3306
3307 /*
3308 * Delete all multicast group membership for an interface.
3309 * Should be used to quickly flush all multicast filters.
3310 */
3311 void
if_delallmulti(struct ifnet * ifp)3312 if_delallmulti(struct ifnet *ifp)
3313 {
3314 struct ifmultiaddr *ifma;
3315 struct ifmultiaddr *next;
3316
3317 IF_ADDR_WLOCK(ifp);
3318 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3319 if_delmulti_locked(ifp, ifma, 0);
3320 IF_ADDR_WUNLOCK(ifp);
3321 }
3322
3323 /*
3324 * Delete a multicast group membership by group membership pointer.
3325 * Network-layer protocol domains must use this routine.
3326 *
3327 * It is safe to call this routine if the ifp disappeared.
3328 */
3329 void
if_delmulti_ifma(struct ifmultiaddr * ifma)3330 if_delmulti_ifma(struct ifmultiaddr *ifma)
3331 {
3332 struct ifnet *ifp;
3333 int lastref;
3334
3335 ifp = ifma->ifma_ifp;
3336 #ifdef DIAGNOSTIC
3337 if (ifp == NULL) {
3338 printf("%s: ifma_ifp seems to be detached\n", __func__);
3339 } else {
3340 struct ifnet *oifp;
3341
3342 IFNET_RLOCK_NOSLEEP();
3343 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3344 if (ifp == oifp)
3345 break;
3346 if (ifp != oifp) {
3347 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3348 ifp = NULL;
3349 }
3350 IFNET_RUNLOCK_NOSLEEP();
3351 }
3352 #endif
3353 /*
3354 * If and only if the ifnet instance exists: Acquire the address lock.
3355 */
3356 if (ifp != NULL)
3357 IF_ADDR_WLOCK(ifp);
3358
3359 lastref = if_delmulti_locked(ifp, ifma, 0);
3360
3361 if (ifp != NULL) {
3362 /*
3363 * If and only if the ifnet instance exists:
3364 * Release the address lock.
3365 * If the group was left: update the hardware hash filter.
3366 */
3367 IF_ADDR_WUNLOCK(ifp);
3368 if (lastref && ifp->if_ioctl != NULL) {
3369 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3370 }
3371 }
3372 }
3373
3374 /*
3375 * Perform deletion of network-layer and/or link-layer multicast address.
3376 *
3377 * Return 0 if the reference count was decremented.
3378 * Return 1 if the final reference was released, indicating that the
3379 * hardware hash filter should be reprogrammed.
3380 */
3381 static int
if_delmulti_locked(struct ifnet * ifp,struct ifmultiaddr * ifma,int detaching)3382 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3383 {
3384 struct ifmultiaddr *ll_ifma;
3385
3386 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3387 KASSERT(ifma->ifma_ifp == ifp,
3388 ("%s: inconsistent ifp %p", __func__, ifp));
3389 IF_ADDR_WLOCK_ASSERT(ifp);
3390 }
3391
3392 ifp = ifma->ifma_ifp;
3393
3394 /*
3395 * If the ifnet is detaching, null out references to ifnet,
3396 * so that upper protocol layers will notice, and not attempt
3397 * to obtain locks for an ifnet which no longer exists. The
3398 * routing socket announcement must happen before the ifnet
3399 * instance is detached from the system.
3400 */
3401 if (detaching) {
3402 #ifdef DIAGNOSTIC
3403 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3404 #endif
3405 /*
3406 * ifp may already be nulled out if we are being reentered
3407 * to delete the ll_ifma.
3408 */
3409 if (ifp != NULL) {
3410 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3411 ifma->ifma_ifp = NULL;
3412 }
3413 }
3414
3415 if (--ifma->ifma_refcount > 0)
3416 return 0;
3417
3418 /*
3419 * If this ifma is a network-layer ifma, a link-layer ifma may
3420 * have been associated with it. Release it first if so.
3421 */
3422 ll_ifma = ifma->ifma_llifma;
3423 if (ll_ifma != NULL) {
3424 KASSERT(ifma->ifma_lladdr != NULL,
3425 ("%s: llifma w/o lladdr", __func__));
3426 if (detaching)
3427 ll_ifma->ifma_ifp = NULL; /* XXX */
3428 if (--ll_ifma->ifma_refcount == 0) {
3429 if (ifp != NULL) {
3430 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3431 ifma_link);
3432 }
3433 if_freemulti(ll_ifma);
3434 }
3435 }
3436
3437 if (ifp != NULL)
3438 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3439
3440 if_freemulti(ifma);
3441
3442 /*
3443 * The last reference to this instance of struct ifmultiaddr
3444 * was released; the hardware should be notified of this change.
3445 */
3446 return 1;
3447 }
3448
3449 /*
3450 * Set the link layer address on an interface.
3451 *
3452 * At this time we only support certain types of interfaces,
3453 * and we don't allow the length of the address to change.
3454 */
3455 int
if_setlladdr(struct ifnet * ifp,const u_char * lladdr,int len)3456 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3457 {
3458 struct sockaddr_dl *sdl;
3459 struct ifaddr *ifa;
3460 struct ifreq ifr;
3461
3462 IF_ADDR_RLOCK(ifp);
3463 ifa = ifp->if_addr;
3464 if (ifa == NULL) {
3465 IF_ADDR_RUNLOCK(ifp);
3466 return (EINVAL);
3467 }
3468 ifa_ref(ifa);
3469 IF_ADDR_RUNLOCK(ifp);
3470 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3471 if (sdl == NULL) {
3472 ifa_free(ifa);
3473 return (EINVAL);
3474 }
3475 if (len != sdl->sdl_alen) { /* don't allow length to change */
3476 ifa_free(ifa);
3477 return (EINVAL);
3478 }
3479 switch (ifp->if_type) {
3480 case IFT_ETHER:
3481 case IFT_FDDI:
3482 case IFT_XETHER:
3483 case IFT_ISO88025:
3484 case IFT_L2VLAN:
3485 case IFT_BRIDGE:
3486 case IFT_ARCNET:
3487 case IFT_IEEE8023ADLAG:
3488 case IFT_IEEE80211:
3489 bcopy(lladdr, LLADDR(sdl), len);
3490 ifa_free(ifa);
3491 break;
3492 default:
3493 ifa_free(ifa);
3494 return (ENODEV);
3495 }
3496
3497 /*
3498 * If the interface is already up, we need
3499 * to re-init it in order to reprogram its
3500 * address filter.
3501 */
3502 if ((ifp->if_flags & IFF_UP) != 0) {
3503 if (ifp->if_ioctl) {
3504 ifp->if_flags &= ~IFF_UP;
3505 ifr.ifr_flags = ifp->if_flags & 0xffff;
3506 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3507 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3508 ifp->if_flags |= IFF_UP;
3509 ifr.ifr_flags = ifp->if_flags & 0xffff;
3510 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3511 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3512 }
3513 #ifdef INET
3514 /*
3515 * Also send gratuitous ARPs to notify other nodes about
3516 * the address change.
3517 */
3518 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3519 if (ifa->ifa_addr->sa_family == AF_INET)
3520 arp_ifinit(ifp, ifa);
3521 }
3522 #endif
3523 }
3524 return (0);
3525 }
3526
3527 /*
3528 * The name argument must be a pointer to storage which will last as
3529 * long as the interface does. For physical devices, the result of
3530 * device_get_name(dev) is a good choice and for pseudo-devices a
3531 * static string works well.
3532 */
3533 void
if_initname(struct ifnet * ifp,const char * name,int unit)3534 if_initname(struct ifnet *ifp, const char *name, int unit)
3535 {
3536 ifp->if_dname = name;
3537 ifp->if_dunit = unit;
3538 if (unit != IF_DUNIT_NONE)
3539 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3540 else
3541 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3542 }
3543
3544 int
if_printf(struct ifnet * ifp,const char * fmt,...)3545 if_printf(struct ifnet *ifp, const char * fmt, ...)
3546 {
3547 va_list ap;
3548 int retval;
3549
3550 retval = printf("%s: ", ifp->if_xname);
3551 va_start(ap, fmt);
3552 retval += vprintf(fmt, ap);
3553 va_end(ap);
3554 return (retval);
3555 }
3556
3557 void
if_start(struct ifnet * ifp)3558 if_start(struct ifnet *ifp)
3559 {
3560
3561 (*(ifp)->if_start)(ifp);
3562 }
3563
3564 /*
3565 * Backwards compatibility interface for drivers
3566 * that have not implemented it
3567 */
3568 static int
if_transmit(struct ifnet * ifp,struct mbuf * m)3569 if_transmit(struct ifnet *ifp, struct mbuf *m)
3570 {
3571 int error;
3572
3573 IFQ_HANDOFF(ifp, m, error);
3574 return (error);
3575 }
3576
3577 static void
if_input_default(struct ifnet * ifp __unused,struct mbuf * m)3578 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3579 {
3580
3581 m_freem(m);
3582 }
3583
3584 int
if_handoff(struct ifqueue * ifq,struct mbuf * m,struct ifnet * ifp,int adjust)3585 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3586 {
3587 int active = 0;
3588
3589 IF_LOCK(ifq);
3590 if (_IF_QFULL(ifq)) {
3591 _IF_DROP(ifq);
3592 IF_UNLOCK(ifq);
3593 m_freem(m);
3594 return (0);
3595 }
3596 if (ifp != NULL) {
3597 ifp->if_obytes += m->m_pkthdr.len + adjust;
3598 if (m->m_flags & (M_BCAST|M_MCAST))
3599 ifp->if_omcasts++;
3600 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3601 }
3602 _IF_ENQUEUE(ifq, m);
3603 IF_UNLOCK(ifq);
3604 if (ifp != NULL && !active)
3605 (*(ifp)->if_start)(ifp);
3606 return (1);
3607 }
3608
3609 void
if_register_com_alloc(u_char type,if_com_alloc_t * a,if_com_free_t * f)3610 if_register_com_alloc(u_char type,
3611 if_com_alloc_t *a, if_com_free_t *f)
3612 {
3613
3614 KASSERT(if_com_alloc[type] == NULL,
3615 ("if_register_com_alloc: %d already registered", type));
3616 KASSERT(if_com_free[type] == NULL,
3617 ("if_register_com_alloc: %d free already registered", type));
3618
3619 if_com_alloc[type] = a;
3620 if_com_free[type] = f;
3621 }
3622
3623 void
if_deregister_com_alloc(u_char type)3624 if_deregister_com_alloc(u_char type)
3625 {
3626
3627 KASSERT(if_com_alloc[type] != NULL,
3628 ("if_deregister_com_alloc: %d not registered", type));
3629 KASSERT(if_com_free[type] != NULL,
3630 ("if_deregister_com_alloc: %d free not registered", type));
3631 if_com_alloc[type] = NULL;
3632 if_com_free[type] = NULL;
3633 }
3634