FreeBSD/Linux Kernel Cross Reference
sys/net/if.c
1 /*
2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
30 * $FreeBSD: releng/5.3/sys/net/if.c 145096 2005-04-15 01:53:14Z cperciva $
31 */
32
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_mac.h"
37
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/mac.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/sockio.h>
53 #include <sys/syslog.h>
54 #include <sys/sysctl.h>
55 #include <sys/taskqueue.h>
56 #include <sys/domain.h>
57 #include <sys/jail.h>
58 #include <machine/stdarg.h>
59
60 #include <net/if.h>
61 #include <net/if_arp.h>
62 #include <net/if_clone.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/if_var.h>
66 #include <net/radix.h>
67 #include <net/route.h>
68
69 #if defined(INET) || defined(INET6)
70 /*XXX*/
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #ifdef INET6
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
76 #endif
77 #endif
78 #ifdef INET
79 #include <netinet/if_ether.h>
80 #endif
81
82 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
83
84 static void if_attachdomain(void *);
85 static void if_attachdomain1(struct ifnet *);
86 static int ifconf(u_long, caddr_t);
87 static void if_grow(void);
88 static void if_init(void *);
89 static void if_check(void *);
90 static int if_findindex(struct ifnet *);
91 static void if_qflush(struct ifaltq *);
92 static void if_route(struct ifnet *, int flag, int fam);
93 static void if_slowtimo(void *);
94 static void if_unroute(struct ifnet *, int flag, int fam);
95 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
96 static int if_rtdel(struct radix_node *, void *);
97 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
98 static void if_start_deferred(void *context, int pending);
99 #ifdef INET6
100 /*
101 * XXX: declare here to avoid to include many inet6 related files..
102 * should be more generalized?
103 */
104 extern void nd6_setmtu(struct ifnet *);
105 #endif
106
107 int if_index = 0;
108 struct ifindex_entry *ifindex_table = NULL;
109 int ifqmaxlen = IFQ_MAXLEN;
110 struct ifnethead ifnet; /* depend on static init XXX */
111 struct mtx ifnet_lock;
112
113 static int if_indexlim = 8;
114 static struct knlist ifklist;
115
116 static void filt_netdetach(struct knote *kn);
117 static int filt_netdev(struct knote *kn, long hint);
118
119 static struct filterops netdev_filtops =
120 { 1, NULL, filt_netdetach, filt_netdev };
121
122 /*
123 * System initialization
124 */
125 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
126 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
127
128 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
129 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
130
131 static d_open_t netopen;
132 static d_close_t netclose;
133 static d_ioctl_t netioctl;
134 static d_kqfilter_t netkqfilter;
135
136 static struct cdevsw net_cdevsw = {
137 .d_version = D_VERSION,
138 .d_flags = D_NEEDGIANT,
139 .d_open = netopen,
140 .d_close = netclose,
141 .d_ioctl = netioctl,
142 .d_name = "net",
143 .d_kqfilter = netkqfilter,
144 };
145
146 static int
147 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
148 {
149 return (0);
150 }
151
152 static int
153 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
154 {
155 return (0);
156 }
157
158 static int
159 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
160 {
161 struct ifnet *ifp;
162 int error, idx;
163
164 /* only support interface specific ioctls */
165 if (IOCGROUP(cmd) != 'i')
166 return (EOPNOTSUPP);
167 idx = minor(dev);
168 if (idx == 0) {
169 /*
170 * special network device, not interface.
171 */
172 if (cmd == SIOCGIFCONF)
173 return (ifconf(cmd, data)); /* XXX remove cmd */
174 return (EOPNOTSUPP);
175 }
176
177 ifp = ifnet_byindex(idx);
178 if (ifp == NULL)
179 return (ENXIO);
180
181 error = ifhwioctl(cmd, ifp, data, td);
182 if (error == ENOIOCTL)
183 error = EOPNOTSUPP;
184 return (error);
185 }
186
187 static int
188 netkqfilter(struct cdev *dev, struct knote *kn)
189 {
190 struct knlist *klist;
191 struct ifnet *ifp;
192 int idx;
193
194 switch (kn->kn_filter) {
195 case EVFILT_NETDEV:
196 kn->kn_fop = &netdev_filtops;
197 break;
198 default:
199 return (1);
200 }
201
202 idx = minor(dev);
203 if (idx == 0) {
204 klist = &ifklist;
205 } else {
206 ifp = ifnet_byindex(idx);
207 if (ifp == NULL)
208 return (1);
209 klist = &ifp->if_klist;
210 }
211
212 kn->kn_hook = (caddr_t)klist;
213
214 knlist_add(klist, kn, 0);
215
216 return (0);
217 }
218
219 static void
220 filt_netdetach(struct knote *kn)
221 {
222 struct knlist *klist = (struct knlist *)kn->kn_hook;
223
224 knlist_remove(klist, kn, 0);
225 }
226
227 static int
228 filt_netdev(struct knote *kn, long hint)
229 {
230 struct knlist *klist = (struct knlist *)kn->kn_hook;
231
232 /*
233 * Currently NOTE_EXIT is abused to indicate device detach.
234 */
235 if (hint == NOTE_EXIT) {
236 kn->kn_data = NOTE_LINKINV;
237 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
238 knlist_remove_inevent(klist, kn);
239 return (1);
240 }
241 if (hint != 0)
242 kn->kn_data = hint; /* current status */
243 if (kn->kn_sfflags & hint)
244 kn->kn_fflags |= hint;
245 return (kn->kn_fflags != 0);
246 }
247
248 /*
249 * Network interface utility routines.
250 *
251 * Routines with ifa_ifwith* names take sockaddr *'s as
252 * parameters.
253 */
254 /* ARGSUSED*/
255 static void
256 if_init(void *dummy __unused)
257 {
258
259 IFNET_LOCK_INIT();
260 TAILQ_INIT(&ifnet);
261 knlist_init(&ifklist, NULL);
262 if_grow(); /* create initial table */
263 ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
264 UID_ROOT, GID_WHEEL, 0600, "network");
265 if_clone_init();
266 }
267
268 static void
269 if_grow(void)
270 {
271 u_int n;
272 struct ifindex_entry *e;
273
274 if_indexlim <<= 1;
275 n = if_indexlim * sizeof(*e);
276 e = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
277 if (ifindex_table != NULL) {
278 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
279 free((caddr_t)ifindex_table, M_IFADDR);
280 }
281 ifindex_table = e;
282 }
283
284 /* ARGSUSED*/
285 static void
286 if_check(void *dummy __unused)
287 {
288 struct ifnet *ifp;
289 int s;
290
291 s = splimp();
292 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */
293 TAILQ_FOREACH(ifp, &ifnet, if_link) {
294 if (ifp->if_snd.ifq_maxlen == 0) {
295 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
296 ifp->if_snd.ifq_maxlen = ifqmaxlen;
297 }
298 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
299 if_printf(ifp,
300 "XXX: driver didn't initialize queue mtx\n");
301 mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
302 MTX_NETWORK_LOCK, MTX_DEF);
303 }
304 }
305 IFNET_RUNLOCK();
306 splx(s);
307 if_slowtimo(0);
308 }
309
310 static int
311 if_findindex(struct ifnet *ifp)
312 {
313 int i, unit;
314 char eaddr[18], devname[32];
315 const char *name, *p;
316
317 switch (ifp->if_type) {
318 case IFT_ETHER: /* these types use struct arpcom */
319 case IFT_FDDI:
320 case IFT_XETHER:
321 case IFT_ISO88025:
322 case IFT_L2VLAN:
323 snprintf(eaddr, 18, "%6D", IFP2AC(ifp)->ac_enaddr, ":");
324 break;
325 default:
326 eaddr[0] = '\0';
327 break;
328 }
329 strlcpy(devname, ifp->if_xname, sizeof(devname));
330 name = net_cdevsw.d_name;
331 i = 0;
332 while ((resource_find_dev(&i, name, &unit, NULL, NULL)) == 0) {
333 if (resource_string_value(name, unit, "ether", &p) == 0)
334 if (strcmp(p, eaddr) == 0)
335 goto found;
336 if (resource_string_value(name, unit, "dev", &p) == 0)
337 if (strcmp(p, devname) == 0)
338 goto found;
339 }
340 unit = 0;
341 found:
342 if (unit != 0) {
343 if (ifaddr_byindex(unit) == NULL)
344 return (unit);
345 printf("%s%d in use, cannot hardwire it to %s.\n",
346 name, unit, devname);
347 }
348 for (unit = 1; ; unit++) {
349 if (unit <= if_index && ifaddr_byindex(unit) != NULL)
350 continue;
351 if (resource_string_value(name, unit, "ether", &p) == 0 ||
352 resource_string_value(name, unit, "dev", &p) == 0)
353 continue;
354 break;
355 }
356 return (unit);
357 }
358
359 /*
360 * Attach an interface to the
361 * list of "active" interfaces.
362 */
363 void
364 if_attach(struct ifnet *ifp)
365 {
366 unsigned socksize, ifasize;
367 int namelen, masklen;
368 struct sockaddr_dl *sdl;
369 struct ifaddr *ifa;
370
371 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
372 IF_AFDATA_LOCK_INIT(ifp);
373 ifp->if_afdata_initialized = 0;
374 IFNET_WLOCK();
375 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
376 IFNET_WUNLOCK();
377 /*
378 * XXX -
379 * The old code would work if the interface passed a pre-existing
380 * chain of ifaddrs to this code. We don't trust our callers to
381 * properly initialize the tailq, however, so we no longer allow
382 * this unlikely case.
383 */
384 TAILQ_INIT(&ifp->if_addrhead);
385 TAILQ_INIT(&ifp->if_prefixhead);
386 TAILQ_INIT(&ifp->if_multiaddrs);
387 knlist_init(&ifp->if_klist, NULL);
388 getmicrotime(&ifp->if_lastchange);
389 ifp->if_data.ifi_epoch = time_second;
390
391 #ifdef MAC
392 mac_init_ifnet(ifp);
393 mac_create_ifnet(ifp);
394 #endif
395
396 ifp->if_index = if_findindex(ifp);
397 if (ifp->if_index > if_index)
398 if_index = ifp->if_index;
399 if (if_index >= if_indexlim)
400 if_grow();
401 ifp->if_data.ifi_datalen = sizeof(struct if_data);
402
403 ifnet_byindex(ifp->if_index) = ifp;
404 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw,
405 unit2minor(ifp->if_index),
406 UID_ROOT, GID_WHEEL, 0600, "%s/%s",
407 net_cdevsw.d_name, ifp->if_xname);
408 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
409 net_cdevsw.d_name, ifp->if_index);
410
411 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
412
413 /*
414 * create a Link Level name for this device
415 */
416 namelen = strlen(ifp->if_xname);
417 /*
418 * Always save enough space for any possiable name so we can do
419 * a rename in place later.
420 */
421 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
422 socksize = masklen + ifp->if_addrlen;
423 if (socksize < sizeof(*sdl))
424 socksize = sizeof(*sdl);
425 socksize = roundup2(socksize, sizeof(long));
426 ifasize = sizeof(*ifa) + 2 * socksize;
427 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
428 IFA_LOCK_INIT(ifa);
429 sdl = (struct sockaddr_dl *)(ifa + 1);
430 sdl->sdl_len = socksize;
431 sdl->sdl_family = AF_LINK;
432 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
433 sdl->sdl_nlen = namelen;
434 sdl->sdl_index = ifp->if_index;
435 sdl->sdl_type = ifp->if_type;
436 ifaddr_byindex(ifp->if_index) = ifa;
437 ifa->ifa_ifp = ifp;
438 ifa->ifa_rtrequest = link_rtrequest;
439 ifa->ifa_addr = (struct sockaddr *)sdl;
440 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
441 ifa->ifa_netmask = (struct sockaddr *)sdl;
442 sdl->sdl_len = masklen;
443 while (namelen != 0)
444 sdl->sdl_data[--namelen] = 0xff;
445 ifa->ifa_refcnt = 1;
446 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
447 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
448 ifp->if_snd.altq_type = 0;
449 ifp->if_snd.altq_disc = NULL;
450 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
451 ifp->if_snd.altq_tbr = NULL;
452 ifp->if_snd.altq_ifp = ifp;
453
454 if (domains)
455 if_attachdomain1(ifp);
456
457 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
458
459 /* Announce the interface. */
460 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
461 }
462
463 static void
464 if_attachdomain(void *dummy)
465 {
466 struct ifnet *ifp;
467 int s;
468
469 s = splnet();
470 TAILQ_FOREACH(ifp, &ifnet, if_link)
471 if_attachdomain1(ifp);
472 splx(s);
473 }
474 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
475 if_attachdomain, NULL);
476
477 static void
478 if_attachdomain1(struct ifnet *ifp)
479 {
480 struct domain *dp;
481 int s;
482
483 s = splnet();
484
485 /*
486 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
487 * cannot lock ifp->if_afdata initialization, entirely.
488 */
489 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
490 splx(s);
491 return;
492 }
493 if (ifp->if_afdata_initialized) {
494 IF_AFDATA_UNLOCK(ifp);
495 splx(s);
496 return;
497 }
498 ifp->if_afdata_initialized = 1;
499 IF_AFDATA_UNLOCK(ifp);
500
501 /* address family dependent data region */
502 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
503 for (dp = domains; dp; dp = dp->dom_next) {
504 if (dp->dom_ifattach)
505 ifp->if_afdata[dp->dom_family] =
506 (*dp->dom_ifattach)(ifp);
507 }
508
509 splx(s);
510 }
511
512 /*
513 * Detach an interface, removing it from the
514 * list of "active" interfaces.
515 */
516 void
517 if_detach(struct ifnet *ifp)
518 {
519 struct ifaddr *ifa, *next;
520 struct radix_node_head *rnh;
521 int s;
522 int i;
523 struct domain *dp;
524 struct ifnet *iter;
525 int found;
526
527 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
528 /*
529 * Remove routes and flush queues.
530 */
531 s = splnet();
532 if_down(ifp);
533 #ifdef ALTQ
534 if (ALTQ_IS_ENABLED(&ifp->if_snd))
535 altq_disable(&ifp->if_snd);
536 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
537 altq_detach(&ifp->if_snd);
538 #endif
539
540 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa; ifa = next) {
541 next = TAILQ_NEXT(ifa, ifa_link);
542
543 if (ifa->ifa_addr->sa_family == AF_LINK)
544 continue;
545 #ifdef INET
546 /* XXX: Ugly!! ad hoc just for INET */
547 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
548 struct ifaliasreq ifr;
549
550 bzero(&ifr, sizeof(ifr));
551 ifr.ifra_addr = *ifa->ifa_addr;
552 if (ifa->ifa_dstaddr)
553 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
554 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
555 NULL) == 0)
556 continue;
557 }
558 #endif /* INET */
559 #ifdef INET6
560 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
561 in6_purgeaddr(ifa);
562 /* ifp_addrhead is already updated */
563 continue;
564 }
565 #endif /* INET6 */
566 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
567 IFAFREE(ifa);
568 }
569
570 #ifdef INET6
571 /*
572 * Remove all IPv6 kernel structs related to ifp. This should be done
573 * before removing routing entries below, since IPv6 interface direct
574 * routes are expected to be removed by the IPv6-specific kernel API.
575 * Otherwise, the kernel will detect some inconsistency and bark it.
576 */
577 in6_ifdetach(ifp);
578 #endif
579 /*
580 * Remove address from ifindex_table[] and maybe decrement if_index.
581 * Clean up all addresses.
582 */
583 ifnet_byindex(ifp->if_index) = NULL;
584 ifaddr_byindex(ifp->if_index) = NULL;
585 destroy_dev(ifdev_byindex(ifp->if_index));
586 ifdev_byindex(ifp->if_index) = NULL;
587
588 while (if_index > 0 && ifaddr_byindex(if_index) == NULL)
589 if_index--;
590
591
592 /* We can now free link ifaddr. */
593 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
594 ifa = TAILQ_FIRST(&ifp->if_addrhead);
595 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
596 IFAFREE(ifa);
597 }
598
599 /*
600 * Delete all remaining routes using this interface
601 * Unfortuneatly the only way to do this is to slog through
602 * the entire routing table looking for routes which point
603 * to this interface...oh well...
604 */
605 for (i = 1; i <= AF_MAX; i++) {
606 if ((rnh = rt_tables[i]) == NULL)
607 continue;
608 RADIX_NODE_HEAD_LOCK(rnh);
609 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
610 RADIX_NODE_HEAD_UNLOCK(rnh);
611 }
612
613 /* Announce that the interface is gone. */
614 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
615
616 IF_AFDATA_LOCK(ifp);
617 for (dp = domains; dp; dp = dp->dom_next) {
618 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
619 (*dp->dom_ifdetach)(ifp,
620 ifp->if_afdata[dp->dom_family]);
621 }
622 IF_AFDATA_UNLOCK(ifp);
623
624 #ifdef MAC
625 mac_destroy_ifnet(ifp);
626 #endif /* MAC */
627 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
628 knlist_clear(&ifp->if_klist, 0);
629 knlist_destroy(&ifp->if_klist);
630 IFNET_WLOCK();
631 found = 0;
632 TAILQ_FOREACH(iter, &ifnet, if_link)
633 if (iter == ifp) {
634 found = 1;
635 break;
636 }
637 if (found)
638 TAILQ_REMOVE(&ifnet, ifp, if_link);
639 IFNET_WUNLOCK();
640 mtx_destroy(&ifp->if_snd.ifq_mtx);
641 IF_AFDATA_DESTROY(ifp);
642 splx(s);
643 }
644
645 /*
646 * Delete Routes for a Network Interface
647 *
648 * Called for each routing entry via the rnh->rnh_walktree() call above
649 * to delete all route entries referencing a detaching network interface.
650 *
651 * Arguments:
652 * rn pointer to node in the routing table
653 * arg argument passed to rnh->rnh_walktree() - detaching interface
654 *
655 * Returns:
656 * 0 successful
657 * errno failed - reason indicated
658 *
659 */
660 static int
661 if_rtdel(struct radix_node *rn, void *arg)
662 {
663 struct rtentry *rt = (struct rtentry *)rn;
664 struct ifnet *ifp = arg;
665 int err;
666
667 if (rt->rt_ifp == ifp) {
668
669 /*
670 * Protect (sorta) against walktree recursion problems
671 * with cloned routes
672 */
673 if ((rt->rt_flags & RTF_UP) == 0)
674 return (0);
675
676 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
677 rt_mask(rt), rt->rt_flags,
678 (struct rtentry **) NULL);
679 if (err) {
680 log(LOG_WARNING, "if_rtdel: error %d\n", err);
681 }
682 }
683
684 return (0);
685 }
686
687 #define equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0)
688
689 /*
690 * Locate an interface based on a complete address.
691 */
692 /*ARGSUSED*/
693 struct ifaddr *
694 ifa_ifwithaddr(struct sockaddr *addr)
695 {
696 struct ifnet *ifp;
697 struct ifaddr *ifa;
698
699 IFNET_RLOCK();
700 TAILQ_FOREACH(ifp, &ifnet, if_link)
701 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
702 if (ifa->ifa_addr->sa_family != addr->sa_family)
703 continue;
704 if (equal(addr, ifa->ifa_addr))
705 goto done;
706 /* IP6 doesn't have broadcast */
707 if ((ifp->if_flags & IFF_BROADCAST) &&
708 ifa->ifa_broadaddr &&
709 ifa->ifa_broadaddr->sa_len != 0 &&
710 equal(ifa->ifa_broadaddr, addr))
711 goto done;
712 }
713 ifa = NULL;
714 done:
715 IFNET_RUNLOCK();
716 return (ifa);
717 }
718
719 /*
720 * Locate the point to point interface with a given destination address.
721 */
722 /*ARGSUSED*/
723 struct ifaddr *
724 ifa_ifwithdstaddr(struct sockaddr *addr)
725 {
726 struct ifnet *ifp;
727 struct ifaddr *ifa;
728
729 IFNET_RLOCK();
730 TAILQ_FOREACH(ifp, &ifnet, if_link) {
731 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
732 continue;
733 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
734 if (ifa->ifa_addr->sa_family != addr->sa_family)
735 continue;
736 if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
737 goto done;
738 }
739 }
740 ifa = NULL;
741 done:
742 IFNET_RUNLOCK();
743 return (ifa);
744 }
745
746 /*
747 * Find an interface on a specific network. If many, choice
748 * is most specific found.
749 */
750 struct ifaddr *
751 ifa_ifwithnet(struct sockaddr *addr)
752 {
753 struct ifnet *ifp;
754 struct ifaddr *ifa;
755 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
756 u_int af = addr->sa_family;
757 char *addr_data = addr->sa_data, *cplim;
758
759 /*
760 * AF_LINK addresses can be looked up directly by their index number,
761 * so do that if we can.
762 */
763 if (af == AF_LINK) {
764 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
765 if (sdl->sdl_index && sdl->sdl_index <= if_index)
766 return (ifaddr_byindex(sdl->sdl_index));
767 }
768
769 /*
770 * Scan though each interface, looking for ones that have
771 * addresses in this address family.
772 */
773 IFNET_RLOCK();
774 TAILQ_FOREACH(ifp, &ifnet, if_link) {
775 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
776 char *cp, *cp2, *cp3;
777
778 if (ifa->ifa_addr->sa_family != af)
779 next: continue;
780 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
781 /*
782 * This is a bit broken as it doesn't
783 * take into account that the remote end may
784 * be a single node in the network we are
785 * looking for.
786 * The trouble is that we don't know the
787 * netmask for the remote end.
788 */
789 if (ifa->ifa_dstaddr != 0
790 && equal(addr, ifa->ifa_dstaddr))
791 goto done;
792 } else {
793 /*
794 * if we have a special address handler,
795 * then use it instead of the generic one.
796 */
797 if (ifa->ifa_claim_addr) {
798 if ((*ifa->ifa_claim_addr)(ifa, addr))
799 goto done;
800 continue;
801 }
802
803 /*
804 * Scan all the bits in the ifa's address.
805 * If a bit dissagrees with what we are
806 * looking for, mask it with the netmask
807 * to see if it really matters.
808 * (A byte at a time)
809 */
810 if (ifa->ifa_netmask == 0)
811 continue;
812 cp = addr_data;
813 cp2 = ifa->ifa_addr->sa_data;
814 cp3 = ifa->ifa_netmask->sa_data;
815 cplim = ifa->ifa_netmask->sa_len
816 + (char *)ifa->ifa_netmask;
817 while (cp3 < cplim)
818 if ((*cp++ ^ *cp2++) & *cp3++)
819 goto next; /* next address! */
820 /*
821 * If the netmask of what we just found
822 * is more specific than what we had before
823 * (if we had one) then remember the new one
824 * before continuing to search
825 * for an even better one.
826 */
827 if (ifa_maybe == 0 ||
828 rn_refines((caddr_t)ifa->ifa_netmask,
829 (caddr_t)ifa_maybe->ifa_netmask))
830 ifa_maybe = ifa;
831 }
832 }
833 }
834 ifa = ifa_maybe;
835 done:
836 IFNET_RUNLOCK();
837 return (ifa);
838 }
839
840 /*
841 * Find an interface address specific to an interface best matching
842 * a given address.
843 */
844 struct ifaddr *
845 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
846 {
847 struct ifaddr *ifa;
848 char *cp, *cp2, *cp3;
849 char *cplim;
850 struct ifaddr *ifa_maybe = 0;
851 u_int af = addr->sa_family;
852
853 if (af >= AF_MAX)
854 return (0);
855 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
856 if (ifa->ifa_addr->sa_family != af)
857 continue;
858 if (ifa_maybe == 0)
859 ifa_maybe = ifa;
860 if (ifa->ifa_netmask == 0) {
861 if (equal(addr, ifa->ifa_addr) ||
862 (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
863 goto done;
864 continue;
865 }
866 if (ifp->if_flags & IFF_POINTOPOINT) {
867 if (equal(addr, ifa->ifa_dstaddr))
868 goto done;
869 } else {
870 cp = addr->sa_data;
871 cp2 = ifa->ifa_addr->sa_data;
872 cp3 = ifa->ifa_netmask->sa_data;
873 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
874 for (; cp3 < cplim; cp3++)
875 if ((*cp++ ^ *cp2++) & *cp3)
876 break;
877 if (cp3 == cplim)
878 goto done;
879 }
880 }
881 ifa = ifa_maybe;
882 done:
883 return (ifa);
884 }
885
886 #include <net/route.h>
887
888 /*
889 * Default action when installing a route with a Link Level gateway.
890 * Lookup an appropriate real ifa to point to.
891 * This should be moved to /sys/net/link.c eventually.
892 */
893 static void
894 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
895 {
896 struct ifaddr *ifa, *oifa;
897 struct sockaddr *dst;
898 struct ifnet *ifp;
899
900 RT_LOCK_ASSERT(rt);
901
902 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
903 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
904 return;
905 ifa = ifaof_ifpforaddr(dst, ifp);
906 if (ifa) {
907 IFAREF(ifa); /* XXX */
908 oifa = rt->rt_ifa;
909 rt->rt_ifa = ifa;
910 IFAFREE(oifa);
911 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
912 ifa->ifa_rtrequest(cmd, rt, info);
913 }
914 }
915
916 /*
917 * Mark an interface down and notify protocols of
918 * the transition.
919 * NOTE: must be called at splnet or eqivalent.
920 */
921 static void
922 if_unroute(struct ifnet *ifp, int flag, int fam)
923 {
924 struct ifaddr *ifa;
925
926 ifp->if_flags &= ~flag;
927 getmicrotime(&ifp->if_lastchange);
928 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
929 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
930 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
931 if_qflush(&ifp->if_snd);
932 rt_ifmsg(ifp);
933 }
934
935 /*
936 * Mark an interface up and notify protocols of
937 * the transition.
938 * NOTE: must be called at splnet or eqivalent.
939 */
940 static void
941 if_route(struct ifnet *ifp, int flag, int fam)
942 {
943 struct ifaddr *ifa;
944
945 ifp->if_flags |= flag;
946 getmicrotime(&ifp->if_lastchange);
947 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
948 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
949 pfctlinput(PRC_IFUP, ifa->ifa_addr);
950 rt_ifmsg(ifp);
951 #ifdef INET6
952 in6_if_up(ifp);
953 #endif
954 }
955
956 /*
957 * Mark an interface down and notify protocols of
958 * the transition.
959 * NOTE: must be called at splnet or eqivalent.
960 */
961 void
962 if_down(struct ifnet *ifp)
963 {
964
965 if_unroute(ifp, IFF_UP, AF_UNSPEC);
966 }
967
968 /*
969 * Mark an interface up and notify protocols of
970 * the transition.
971 * NOTE: must be called at splnet or eqivalent.
972 */
973 void
974 if_up(struct ifnet *ifp)
975 {
976
977 if_route(ifp, IFF_UP, AF_UNSPEC);
978 }
979
980 /*
981 * Flush an interface queue.
982 */
983 static void
984 if_qflush(struct ifaltq *ifq)
985 {
986 struct mbuf *m, *n;
987
988 IFQ_LOCK(ifq);
989 #ifdef ALTQ
990 if (ALTQ_IS_ENABLED(ifq))
991 ALTQ_PURGE(ifq);
992 #endif
993 n = ifq->ifq_head;
994 while ((m = n) != 0) {
995 n = m->m_act;
996 m_freem(m);
997 }
998 ifq->ifq_head = 0;
999 ifq->ifq_tail = 0;
1000 ifq->ifq_len = 0;
1001 IFQ_UNLOCK(ifq);
1002 }
1003
1004 /*
1005 * Handle interface watchdog timer routines. Called
1006 * from softclock, we decrement timers (if set) and
1007 * call the appropriate interface routine on expiration.
1008 *
1009 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1010 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab
1011 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1012 */
1013 static void
1014 if_slowtimo(void *arg)
1015 {
1016 struct ifnet *ifp;
1017 int s = splimp();
1018
1019 IFNET_RLOCK();
1020 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1021 if (ifp->if_timer == 0 || --ifp->if_timer)
1022 continue;
1023 if (ifp->if_watchdog)
1024 (*ifp->if_watchdog)(ifp);
1025 }
1026 IFNET_RUNLOCK();
1027 splx(s);
1028 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1029 }
1030
1031 /*
1032 * Map interface name to
1033 * interface structure pointer.
1034 */
1035 struct ifnet *
1036 ifunit(const char *name)
1037 {
1038 struct ifnet *ifp;
1039
1040 IFNET_RLOCK();
1041 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1042 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1043 break;
1044 }
1045 IFNET_RUNLOCK();
1046 return (ifp);
1047 }
1048
1049 /*
1050 * Hardware specific interface ioctls.
1051 */
1052 static int
1053 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1054 {
1055 struct ifreq *ifr;
1056 struct ifstat *ifs;
1057 int error = 0;
1058 int new_flags;
1059 size_t namelen, onamelen;
1060 char new_name[IFNAMSIZ];
1061 struct ifaddr *ifa;
1062 struct sockaddr_dl *sdl;
1063
1064 ifr = (struct ifreq *)data;
1065 switch (cmd) {
1066 case SIOCGIFINDEX:
1067 ifr->ifr_index = ifp->if_index;
1068 break;
1069
1070 case SIOCGIFFLAGS:
1071 ifr->ifr_flags = ifp->if_flags & 0xffff;
1072 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1073 break;
1074
1075 case SIOCGIFCAP:
1076 ifr->ifr_reqcap = ifp->if_capabilities;
1077 ifr->ifr_curcap = ifp->if_capenable;
1078 break;
1079
1080 #ifdef MAC
1081 case SIOCGIFMAC:
1082 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
1083 break;
1084 #endif
1085
1086 case SIOCGIFMETRIC:
1087 ifr->ifr_metric = ifp->if_metric;
1088 break;
1089
1090 case SIOCGIFMTU:
1091 ifr->ifr_mtu = ifp->if_mtu;
1092 break;
1093
1094 case SIOCGIFPHYS:
1095 ifr->ifr_phys = ifp->if_physical;
1096 break;
1097
1098 case SIOCSIFFLAGS:
1099 error = suser(td);
1100 if (error)
1101 return (error);
1102 new_flags = (ifr->ifr_flags & 0xffff) |
1103 (ifr->ifr_flagshigh << 16);
1104 if (ifp->if_flags & IFF_SMART) {
1105 /* Smart drivers twiddle their own routes */
1106 } else if (ifp->if_flags & IFF_UP &&
1107 (new_flags & IFF_UP) == 0) {
1108 int s = splimp();
1109 if_down(ifp);
1110 splx(s);
1111 } else if (new_flags & IFF_UP &&
1112 (ifp->if_flags & IFF_UP) == 0) {
1113 int s = splimp();
1114 if_up(ifp);
1115 splx(s);
1116 }
1117 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1118 (new_flags &~ IFF_CANTCHANGE);
1119 if (new_flags & IFF_PPROMISC) {
1120 /* Permanently promiscuous mode requested */
1121 ifp->if_flags |= IFF_PROMISC;
1122 } else if (ifp->if_pcount == 0) {
1123 ifp->if_flags &= ~IFF_PROMISC;
1124 }
1125 if (ifp->if_ioctl) {
1126 IFF_LOCKGIANT(ifp);
1127 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1128 IFF_UNLOCKGIANT(ifp);
1129 }
1130 getmicrotime(&ifp->if_lastchange);
1131 break;
1132
1133 case SIOCSIFCAP:
1134 error = suser(td);
1135 if (error)
1136 return (error);
1137 if (ifp->if_ioctl == NULL)
1138 return (EOPNOTSUPP);
1139 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1140 return (EINVAL);
1141 IFF_LOCKGIANT(ifp);
1142 error = (*ifp->if_ioctl)(ifp, cmd, data);
1143 IFF_UNLOCKGIANT(ifp);
1144 if (error == 0)
1145 getmicrotime(&ifp->if_lastchange);
1146 break;
1147
1148 #ifdef MAC
1149 case SIOCSIFMAC:
1150 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
1151 break;
1152 #endif
1153
1154 case SIOCSIFNAME:
1155 error = suser(td);
1156 if (error != 0)
1157 return (error);
1158 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1159 if (error != 0)
1160 return (error);
1161 if (new_name[0] == '\0')
1162 return (EINVAL);
1163 if (ifunit(new_name) != NULL)
1164 return (EEXIST);
1165
1166 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1167 /* Announce the departure of the interface. */
1168 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1169
1170 log(LOG_INFO, "%s: changing name to '%s'\n",
1171 ifp->if_xname, new_name);
1172
1173 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1174 ifa = ifaddr_byindex(ifp->if_index);
1175 IFA_LOCK(ifa);
1176 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1177 namelen = strlen(new_name);
1178 onamelen = sdl->sdl_nlen;
1179 /*
1180 * Move the address if needed. This is safe because we
1181 * allocate space for a name of length IFNAMSIZ when we
1182 * create this in if_attach().
1183 */
1184 if (namelen != onamelen) {
1185 bcopy(sdl->sdl_data + onamelen,
1186 sdl->sdl_data + namelen, sdl->sdl_alen);
1187 }
1188 bcopy(new_name, sdl->sdl_data, namelen);
1189 sdl->sdl_nlen = namelen;
1190 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1191 bzero(sdl->sdl_data, onamelen);
1192 while (namelen != 0)
1193 sdl->sdl_data[--namelen] = 0xff;
1194 IFA_UNLOCK(ifa);
1195
1196 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1197 /* Announce the return of the interface. */
1198 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1199 break;
1200
1201 case SIOCSIFMETRIC:
1202 error = suser(td);
1203 if (error)
1204 return (error);
1205 ifp->if_metric = ifr->ifr_metric;
1206 getmicrotime(&ifp->if_lastchange);
1207 break;
1208
1209 case SIOCSIFPHYS:
1210 error = suser(td);
1211 if (error)
1212 return (error);
1213 if (ifp->if_ioctl == NULL)
1214 return (EOPNOTSUPP);
1215 IFF_LOCKGIANT(ifp);
1216 error = (*ifp->if_ioctl)(ifp, cmd, data);
1217 IFF_UNLOCKGIANT(ifp);
1218 if (error == 0)
1219 getmicrotime(&ifp->if_lastchange);
1220 break;
1221
1222 case SIOCSIFMTU:
1223 {
1224 u_long oldmtu = ifp->if_mtu;
1225
1226 error = suser(td);
1227 if (error)
1228 return (error);
1229 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1230 return (EINVAL);
1231 if (ifp->if_ioctl == NULL)
1232 return (EOPNOTSUPP);
1233 IFF_LOCKGIANT(ifp);
1234 error = (*ifp->if_ioctl)(ifp, cmd, data);
1235 IFF_UNLOCKGIANT(ifp);
1236 if (error == 0) {
1237 getmicrotime(&ifp->if_lastchange);
1238 rt_ifmsg(ifp);
1239 }
1240 /*
1241 * If the link MTU changed, do network layer specific procedure.
1242 */
1243 if (ifp->if_mtu != oldmtu) {
1244 #ifdef INET6
1245 nd6_setmtu(ifp);
1246 #endif
1247 }
1248 break;
1249 }
1250
1251 case SIOCADDMULTI:
1252 case SIOCDELMULTI:
1253 error = suser(td);
1254 if (error)
1255 return (error);
1256
1257 /* Don't allow group membership on non-multicast interfaces. */
1258 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1259 return (EOPNOTSUPP);
1260
1261 /* Don't let users screw up protocols' entries. */
1262 if (ifr->ifr_addr.sa_family != AF_LINK)
1263 return (EINVAL);
1264
1265 if (cmd == SIOCADDMULTI) {
1266 struct ifmultiaddr *ifma;
1267 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1268 } else {
1269 error = if_delmulti(ifp, &ifr->ifr_addr);
1270 }
1271 if (error == 0)
1272 getmicrotime(&ifp->if_lastchange);
1273 break;
1274
1275 case SIOCSIFPHYADDR:
1276 case SIOCDIFPHYADDR:
1277 #ifdef INET6
1278 case SIOCSIFPHYADDR_IN6:
1279 #endif
1280 case SIOCSLIFPHYADDR:
1281 case SIOCSIFMEDIA:
1282 case SIOCSIFGENERIC:
1283 error = suser(td);
1284 if (error)
1285 return (error);
1286 if (ifp->if_ioctl == NULL)
1287 return (EOPNOTSUPP);
1288 IFF_LOCKGIANT(ifp);
1289 error = (*ifp->if_ioctl)(ifp, cmd, data);
1290 IFF_UNLOCKGIANT(ifp);
1291 if (error == 0)
1292 getmicrotime(&ifp->if_lastchange);
1293 break;
1294
1295 case SIOCGIFSTATUS:
1296 ifs = (struct ifstat *)data;
1297 ifs->ascii[0] = '\0';
1298
1299 case SIOCGIFPSRCADDR:
1300 case SIOCGIFPDSTADDR:
1301 case SIOCGLIFPHYADDR:
1302 case SIOCGIFMEDIA:
1303 case SIOCGIFGENERIC:
1304 if (ifp->if_ioctl == NULL)
1305 return (EOPNOTSUPP);
1306 IFF_LOCKGIANT(ifp);
1307 error = (*ifp->if_ioctl)(ifp, cmd, data);
1308 IFF_UNLOCKGIANT(ifp);
1309 break;
1310
1311 case SIOCSIFLLADDR:
1312 error = suser(td);
1313 if (error)
1314 return (error);
1315 error = if_setlladdr(ifp,
1316 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1317 break;
1318
1319 default:
1320 error = ENOIOCTL;
1321 break;
1322 }
1323 return (error);
1324 }
1325
1326 /*
1327 * Interface ioctls.
1328 */
1329 int
1330 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1331 {
1332 struct ifnet *ifp;
1333 struct ifreq *ifr;
1334 int error;
1335 int oif_flags;
1336
1337 switch (cmd) {
1338 case SIOCGIFCONF:
1339 case OSIOCGIFCONF:
1340 return (ifconf(cmd, data));
1341 }
1342 ifr = (struct ifreq *)data;
1343
1344 switch (cmd) {
1345 case SIOCIFCREATE:
1346 case SIOCIFDESTROY:
1347 if ((error = suser(td)) != 0)
1348 return (error);
1349 return ((cmd == SIOCIFCREATE) ?
1350 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1351 if_clone_destroy(ifr->ifr_name));
1352
1353 case SIOCIFGCLONERS:
1354 return (if_clone_list((struct if_clonereq *)data));
1355 }
1356
1357 ifp = ifunit(ifr->ifr_name);
1358 if (ifp == 0)
1359 return (ENXIO);
1360
1361 error = ifhwioctl(cmd, ifp, data, td);
1362 if (error != ENOIOCTL)
1363 return (error);
1364
1365 oif_flags = ifp->if_flags;
1366 if (so->so_proto == 0)
1367 return (EOPNOTSUPP);
1368 #ifndef COMPAT_43
1369 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1370 data,
1371 ifp, td));
1372 #else
1373 {
1374 int ocmd = cmd;
1375
1376 switch (cmd) {
1377
1378 case SIOCSIFDSTADDR:
1379 case SIOCSIFADDR:
1380 case SIOCSIFBRDADDR:
1381 case SIOCSIFNETMASK:
1382 #if BYTE_ORDER != BIG_ENDIAN
1383 if (ifr->ifr_addr.sa_family == 0 &&
1384 ifr->ifr_addr.sa_len < 16) {
1385 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1386 ifr->ifr_addr.sa_len = 16;
1387 }
1388 #else
1389 if (ifr->ifr_addr.sa_len == 0)
1390 ifr->ifr_addr.sa_len = 16;
1391 #endif
1392 break;
1393
1394 case OSIOCGIFADDR:
1395 cmd = SIOCGIFADDR;
1396 break;
1397
1398 case OSIOCGIFDSTADDR:
1399 cmd = SIOCGIFDSTADDR;
1400 break;
1401
1402 case OSIOCGIFBRDADDR:
1403 cmd = SIOCGIFBRDADDR;
1404 break;
1405
1406 case OSIOCGIFNETMASK:
1407 cmd = SIOCGIFNETMASK;
1408 }
1409 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
1410 cmd,
1411 data,
1412 ifp, td));
1413 switch (ocmd) {
1414
1415 case OSIOCGIFADDR:
1416 case OSIOCGIFDSTADDR:
1417 case OSIOCGIFBRDADDR:
1418 case OSIOCGIFNETMASK:
1419 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1420
1421 }
1422 }
1423 #endif /* COMPAT_43 */
1424
1425 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1426 #ifdef INET6
1427 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1428 if (ifp->if_flags & IFF_UP) {
1429 int s = splimp();
1430 in6_if_up(ifp);
1431 splx(s);
1432 }
1433 #endif
1434 }
1435 return (error);
1436 }
1437
1438 /*
1439 * Set/clear promiscuous mode on interface ifp based on the truth value
1440 * of pswitch. The calls are reference counted so that only the first
1441 * "on" request actually has an effect, as does the final "off" request.
1442 * Results are undefined if the "off" and "on" requests are not matched.
1443 */
1444 int
1445 ifpromisc(struct ifnet *ifp, int pswitch)
1446 {
1447 struct ifreq ifr;
1448 int error;
1449 int oldflags, oldpcount;
1450
1451 oldpcount = ifp->if_pcount;
1452 oldflags = ifp->if_flags;
1453 if (ifp->if_flags & IFF_PPROMISC) {
1454 /* Do nothing if device is in permanently promiscuous mode */
1455 ifp->if_pcount += pswitch ? 1 : -1;
1456 return (0);
1457 }
1458 if (pswitch) {
1459 /*
1460 * If the device is not configured up, we cannot put it in
1461 * promiscuous mode.
1462 */
1463 if ((ifp->if_flags & IFF_UP) == 0)
1464 return (ENETDOWN);
1465 if (ifp->if_pcount++ != 0)
1466 return (0);
1467 ifp->if_flags |= IFF_PROMISC;
1468 } else {
1469 if (--ifp->if_pcount > 0)
1470 return (0);
1471 ifp->if_flags &= ~IFF_PROMISC;
1472 }
1473 ifr.ifr_flags = ifp->if_flags & 0xffff;
1474 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1475 IFF_LOCKGIANT(ifp);
1476 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1477 IFF_UNLOCKGIANT(ifp);
1478 if (error == 0) {
1479 log(LOG_INFO, "%s: promiscuous mode %s\n",
1480 ifp->if_xname,
1481 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
1482 rt_ifmsg(ifp);
1483 } else {
1484 ifp->if_pcount = oldpcount;
1485 ifp->if_flags = oldflags;
1486 }
1487 return error;
1488 }
1489
1490 /*
1491 * Return interface configuration
1492 * of system. List may be used
1493 * in later ioctl's (above) to get
1494 * other information.
1495 */
1496 /*ARGSUSED*/
1497 static int
1498 ifconf(u_long cmd, caddr_t data)
1499 {
1500 struct ifconf *ifc = (struct ifconf *)data;
1501 struct ifnet *ifp;
1502 struct ifaddr *ifa;
1503 struct ifreq ifr;
1504 struct sbuf *sb;
1505 int error, full = 0, valid_len, max_len;
1506
1507 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
1508 max_len = MAXPHYS - 1;
1509
1510 again:
1511 if (ifc->ifc_len <= max_len) {
1512 max_len = ifc->ifc_len;
1513 full = 1;
1514 }
1515 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
1516 max_len = 0;
1517 valid_len = 0;
1518
1519 IFNET_RLOCK(); /* could sleep XXX */
1520 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1521 int addrs;
1522
1523 /*
1524 * Zero the ifr_name buffer to make sure we don't
1525 * disclose the contents of the stack.
1526 */
1527 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
1528
1529 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1530 >= sizeof(ifr.ifr_name))
1531 return (ENAMETOOLONG);
1532
1533 addrs = 0;
1534 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1535 struct sockaddr *sa = ifa->ifa_addr;
1536
1537 if (jailed(curthread->td_ucred) &&
1538 prison_if(curthread->td_ucred, sa))
1539 continue;
1540 addrs++;
1541 #ifdef COMPAT_43
1542 if (cmd == OSIOCGIFCONF) {
1543 struct osockaddr *osa =
1544 (struct osockaddr *)&ifr.ifr_addr;
1545 ifr.ifr_addr = *sa;
1546 osa->sa_family = sa->sa_family;
1547 sbuf_bcat(sb, &ifr, sizeof(ifr));
1548 max_len += sizeof(ifr);
1549 } else
1550 #endif
1551 if (sa->sa_len <= sizeof(*sa)) {
1552 ifr.ifr_addr = *sa;
1553 sbuf_bcat(sb, &ifr, sizeof(ifr));
1554 max_len += sizeof(ifr);
1555 } else {
1556 sbuf_bcat(sb, &ifr,
1557 offsetof(struct ifreq, ifr_addr));
1558 max_len += offsetof(struct ifreq, ifr_addr);
1559 sbuf_bcat(sb, sa, sa->sa_len);
1560 max_len += sa->sa_len;
1561 }
1562
1563 if (!sbuf_overflowed(sb))
1564 valid_len = sbuf_len(sb);
1565 }
1566 if (addrs == 0) {
1567 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1568 sbuf_bcat(sb, &ifr, sizeof(ifr));
1569 max_len += sizeof(ifr);
1570
1571 if (!sbuf_overflowed(sb))
1572 valid_len = sbuf_len(sb);
1573 }
1574 }
1575 IFNET_RUNLOCK();
1576
1577 /*
1578 * If we didn't allocate enough space (uncommon), try again. If
1579 * we have already allocated as much space as we are allowed,
1580 * return what we've got.
1581 */
1582 if (valid_len != max_len && !full) {
1583 sbuf_delete(sb);
1584 goto again;
1585 }
1586
1587 ifc->ifc_len = valid_len;
1588 sbuf_finish(sb);
1589 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
1590 sbuf_delete(sb);
1591 return (error);
1592 }
1593
1594 /*
1595 * Just like if_promisc(), but for all-multicast-reception mode.
1596 */
1597 int
1598 if_allmulti(struct ifnet *ifp, int onswitch)
1599 {
1600 int error = 0;
1601 int s = splimp();
1602 struct ifreq ifr;
1603
1604 if (onswitch) {
1605 if (ifp->if_amcount++ == 0) {
1606 ifp->if_flags |= IFF_ALLMULTI;
1607 ifr.ifr_flags = ifp->if_flags & 0xffff;
1608 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1609 IFF_LOCKGIANT(ifp);
1610 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1611 IFF_UNLOCKGIANT(ifp);
1612 }
1613 } else {
1614 if (ifp->if_amcount > 1) {
1615 ifp->if_amcount--;
1616 } else {
1617 ifp->if_amcount = 0;
1618 ifp->if_flags &= ~IFF_ALLMULTI;
1619 ifr.ifr_flags = ifp->if_flags & 0xffff;;
1620 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1621 IFF_LOCKGIANT(ifp);
1622 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1623 IFF_UNLOCKGIANT(ifp);
1624 }
1625 }
1626 splx(s);
1627
1628 if (error == 0)
1629 rt_ifmsg(ifp);
1630 return error;
1631 }
1632
1633 /*
1634 * Add a multicast listenership to the interface in question.
1635 * The link layer provides a routine which converts
1636 */
1637 int
1638 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, struct ifmultiaddr **retifma)
1639 {
1640 struct sockaddr *llsa, *dupsa;
1641 int error, s;
1642 struct ifmultiaddr *ifma;
1643
1644 /*
1645 * If the matching multicast address already exists
1646 * then don't add a new one, just add a reference
1647 */
1648 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1649 if (equal(sa, ifma->ifma_addr)) {
1650 ifma->ifma_refcount++;
1651 if (retifma)
1652 *retifma = ifma;
1653 return 0;
1654 }
1655 }
1656
1657 /*
1658 * Give the link layer a chance to accept/reject it, and also
1659 * find out which AF_LINK address this maps to, if it isn't one
1660 * already.
1661 */
1662 if (ifp->if_resolvemulti) {
1663 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1664 if (error) return error;
1665 } else {
1666 llsa = 0;
1667 }
1668
1669 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1670 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1671 bcopy(sa, dupsa, sa->sa_len);
1672
1673 ifma->ifma_addr = dupsa;
1674 ifma->ifma_lladdr = llsa;
1675 ifma->ifma_ifp = ifp;
1676 ifma->ifma_refcount = 1;
1677 ifma->ifma_protospec = 0;
1678 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1679
1680 /*
1681 * Some network interfaces can scan the address list at
1682 * interrupt time; lock them out.
1683 */
1684 s = splimp();
1685 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1686 splx(s);
1687 if (retifma != NULL)
1688 *retifma = ifma;
1689
1690 if (llsa != 0) {
1691 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1692 if (equal(ifma->ifma_addr, llsa))
1693 break;
1694 }
1695 if (ifma) {
1696 ifma->ifma_refcount++;
1697 } else {
1698 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1699 M_IFMADDR, M_WAITOK);
1700 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1701 M_IFMADDR, M_WAITOK);
1702 bcopy(llsa, dupsa, llsa->sa_len);
1703 ifma->ifma_addr = dupsa;
1704 ifma->ifma_lladdr = NULL;
1705 ifma->ifma_ifp = ifp;
1706 ifma->ifma_refcount = 1;
1707 ifma->ifma_protospec = 0;
1708 s = splimp();
1709 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1710 splx(s);
1711 }
1712 }
1713 /*
1714 * We are certain we have added something, so call down to the
1715 * interface to let them know about it.
1716 */
1717 s = splimp();
1718 IFF_LOCKGIANT(ifp);
1719 ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1720 IFF_UNLOCKGIANT(ifp);
1721 splx(s);
1722
1723 return 0;
1724 }
1725
1726 /*
1727 * Remove a reference to a multicast address on this interface. Yell
1728 * if the request does not match an existing membership.
1729 */
1730 int
1731 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1732 {
1733 struct ifmultiaddr *ifma;
1734 int s;
1735
1736 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1737 if (equal(sa, ifma->ifma_addr))
1738 break;
1739 if (ifma == 0)
1740 return ENOENT;
1741
1742 if (ifma->ifma_refcount > 1) {
1743 ifma->ifma_refcount--;
1744 return 0;
1745 }
1746
1747 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1748 sa = ifma->ifma_lladdr;
1749 s = splimp();
1750 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1751 /*
1752 * Make sure the interface driver is notified
1753 * in the case of a link layer mcast group being left.
1754 */
1755 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0) {
1756 IFF_LOCKGIANT(ifp);
1757 ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1758 IFF_UNLOCKGIANT(ifp);
1759 }
1760 splx(s);
1761 free(ifma->ifma_addr, M_IFMADDR);
1762 free(ifma, M_IFMADDR);
1763 if (sa == 0)
1764 return 0;
1765
1766 /*
1767 * Now look for the link-layer address which corresponds to
1768 * this network address. It had been squirreled away in
1769 * ifma->ifma_lladdr for this purpose (so we don't have
1770 * to call ifp->if_resolvemulti() again), and we saved that
1771 * value in sa above. If some nasty deleted the
1772 * link-layer address out from underneath us, we can deal because
1773 * the address we stored was is not the same as the one which was
1774 * in the record for the link-layer address. (So we don't complain
1775 * in that case.)
1776 */
1777 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1778 if (equal(sa, ifma->ifma_addr))
1779 break;
1780 if (ifma == 0)
1781 return 0;
1782
1783 if (ifma->ifma_refcount > 1) {
1784 ifma->ifma_refcount--;
1785 return 0;
1786 }
1787
1788 s = splimp();
1789 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1790 IFF_LOCKGIANT(ifp);
1791 ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1792 IFF_UNLOCKGIANT(ifp);
1793 splx(s);
1794 free(ifma->ifma_addr, M_IFMADDR);
1795 free(sa, M_IFMADDR);
1796 free(ifma, M_IFMADDR);
1797
1798 return 0;
1799 }
1800
1801 /*
1802 * Set the link layer address on an interface.
1803 *
1804 * At this time we only support certain types of interfaces,
1805 * and we don't allow the length of the address to change.
1806 */
1807 int
1808 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1809 {
1810 struct sockaddr_dl *sdl;
1811 struct ifaddr *ifa;
1812 struct ifreq ifr;
1813
1814 ifa = ifaddr_byindex(ifp->if_index);
1815 if (ifa == NULL)
1816 return (EINVAL);
1817 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1818 if (sdl == NULL)
1819 return (EINVAL);
1820 if (len != sdl->sdl_alen) /* don't allow length to change */
1821 return (EINVAL);
1822 switch (ifp->if_type) {
1823 case IFT_ETHER: /* these types use struct arpcom */
1824 case IFT_FDDI:
1825 case IFT_XETHER:
1826 case IFT_ISO88025:
1827 case IFT_L2VLAN:
1828 bcopy(lladdr, IFP2AC(ifp)->ac_enaddr, len);
1829 /*
1830 * XXX We also need to store the lladdr in LLADDR(sdl),
1831 * which is done below. This is a pain because we must
1832 * remember to keep the info in sync.
1833 */
1834 /* FALLTHROUGH */
1835 case IFT_ARCNET:
1836 bcopy(lladdr, LLADDR(sdl), len);
1837 break;
1838 default:
1839 return (ENODEV);
1840 }
1841 /*
1842 * If the interface is already up, we need
1843 * to re-init it in order to reprogram its
1844 * address filter.
1845 */
1846 if ((ifp->if_flags & IFF_UP) != 0) {
1847 IFF_LOCKGIANT(ifp);
1848 ifp->if_flags &= ~IFF_UP;
1849 ifr.ifr_flags = ifp->if_flags & 0xffff;
1850 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1851 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1852 ifp->if_flags |= IFF_UP;
1853 ifr.ifr_flags = ifp->if_flags & 0xffff;
1854 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1855 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1856 IFF_UNLOCKGIANT(ifp);
1857 #ifdef INET
1858 /*
1859 * Also send gratuitous ARPs to notify other nodes about
1860 * the address change.
1861 */
1862 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1863 if (ifa->ifa_addr != NULL &&
1864 ifa->ifa_addr->sa_family == AF_INET)
1865 arp_ifinit(ifp, ifa);
1866 }
1867 #endif
1868 }
1869 return (0);
1870 }
1871
1872 struct ifmultiaddr *
1873 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1874 {
1875 struct ifmultiaddr *ifma;
1876
1877 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1878 if (equal(ifma->ifma_addr, sa))
1879 break;
1880
1881 return ifma;
1882 }
1883
1884 /*
1885 * The name argument must be a pointer to storage which will last as
1886 * long as the interface does. For physical devices, the result of
1887 * device_get_name(dev) is a good choice and for pseudo-devices a
1888 * static string works well.
1889 */
1890 void
1891 if_initname(struct ifnet *ifp, const char *name, int unit)
1892 {
1893 ifp->if_dname = name;
1894 ifp->if_dunit = unit;
1895 if (unit != IF_DUNIT_NONE)
1896 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1897 else
1898 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1899 }
1900
1901 int
1902 if_printf(struct ifnet *ifp, const char * fmt, ...)
1903 {
1904 va_list ap;
1905 int retval;
1906
1907 retval = printf("%s: ", ifp->if_xname);
1908 va_start(ap, fmt);
1909 retval += vprintf(fmt, ap);
1910 va_end(ap);
1911 return (retval);
1912 }
1913
1914 /*
1915 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
1916 * be called without Giant. However, we often can't acquire the Giant lock
1917 * at those points; instead, we run it via a task queue that holds Giant via
1918 * if_start_deferred.
1919 *
1920 * XXXRW: We need to make sure that the ifnet isn't fully detached until any
1921 * outstanding if_start_deferred() tasks that will run after the free. This
1922 * probably means waiting in if_detach().
1923 */
1924 void
1925 if_start(struct ifnet *ifp)
1926 {
1927
1928 NET_ASSERT_GIANT();
1929
1930 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) {
1931 if (mtx_owned(&Giant))
1932 (*(ifp)->if_start)(ifp);
1933 else
1934 taskqueue_enqueue(taskqueue_swi_giant,
1935 &ifp->if_starttask);
1936 } else
1937 (*(ifp)->if_start)(ifp);
1938 }
1939
1940 static void
1941 if_start_deferred(void *context, int pending)
1942 {
1943 struct ifnet *ifp;
1944
1945 /*
1946 * This code must be entered with Giant, and should never run if
1947 * we're not running with debug.mpsafenet.
1948 */
1949 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet"));
1950 GIANT_REQUIRED;
1951
1952 ifp = (struct ifnet *)context;
1953 (ifp->if_start)(ifp);
1954 }
1955
1956 int
1957 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
1958 {
1959 int active = 0;
1960
1961 IF_LOCK(ifq);
1962 if (_IF_QFULL(ifq)) {
1963 _IF_DROP(ifq);
1964 IF_UNLOCK(ifq);
1965 m_freem(m);
1966 return (0);
1967 }
1968 if (ifp != NULL) {
1969 ifp->if_obytes += m->m_pkthdr.len + adjust;
1970 if (m->m_flags & (M_BCAST|M_MCAST))
1971 ifp->if_omcasts++;
1972 active = ifp->if_flags & IFF_OACTIVE;
1973 }
1974 _IF_ENQUEUE(ifq, m);
1975 IF_UNLOCK(ifq);
1976 if (ifp != NULL && !active)
1977 if_start(ifp);
1978 return (1);
1979 }
1980
1981 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1982 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
Cache object: acafa0a0068432f0fa503b7cb219e4d1
|