2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
39 #include <sys/param.h>
40 #include <sys/types.h>
43 #include <sys/malloc.h>
47 #include <sys/systm.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/sockio.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/taskqueue.h>
57 #include <sys/domain.h>
59 #include <machine/stdarg.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>
69 #if defined(INET) || defined(INET6)
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
79 #include <netinet/if_ether.h>
82 #include <netinet/ip_carp.h>
85 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
86 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
88 /* Log link state change events */
89 static int log_link_state_change = 1;
91 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
92 &log_link_state_change, 0,
93 "log interface link state change events");
95 void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
96 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
98 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
100 static void if_attachdomain(void *);
101 static void if_attachdomain1(struct ifnet *);
102 static int ifconf(u_long, caddr_t);
103 static void if_grow(void);
104 static void if_init(void *);
105 static void if_check(void *);
106 static void if_qflush(struct ifaltq *);
107 static void if_route(struct ifnet *, int flag, int fam);
108 static int if_setflag(struct ifnet *, int, int, int *, int);
109 static void if_slowtimo(void *);
110 static void if_unroute(struct ifnet *, int flag, int fam);
111 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
112 static int if_rtdel(struct radix_node *, void *);
113 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
114 static void if_start_deferred(void *context, int pending);
115 static void do_link_state_change(void *, int);
118 * XXX: declare here to avoid to include many inet6 related files..
119 * should be more generalized?
121 extern void nd6_setmtu(struct ifnet *);
125 struct ifindex_entry *ifindex_table = NULL;
126 int ifqmaxlen = IFQ_MAXLEN;
127 struct ifnethead ifnet; /* depend on static init XXX */
128 struct mtx ifnet_lock;
129 static if_com_alloc_t *if_com_alloc[256];
130 static if_com_free_t *if_com_free[256];
132 static int if_indexlim = 8;
133 static struct knlist ifklist;
135 static void filt_netdetach(struct knote *kn);
136 static int filt_netdev(struct knote *kn, long hint);
138 static struct filterops netdev_filtops =
139 { 1, NULL, filt_netdetach, filt_netdev };
142 * System initialization
144 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
145 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
147 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
148 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
149 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
151 static d_open_t netopen;
152 static d_close_t netclose;
153 static d_ioctl_t netioctl;
154 static d_kqfilter_t netkqfilter;
156 static struct cdevsw net_cdevsw = {
157 .d_version = D_VERSION,
158 .d_flags = D_NEEDGIANT,
163 .d_kqfilter = netkqfilter,
167 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
173 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
179 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
184 /* only support interface specific ioctls */
185 if (IOCGROUP(cmd) != 'i')
190 * special network device, not interface.
192 if (cmd == SIOCGIFCONF)
193 return (ifconf(cmd, data)); /* XXX remove cmd */
197 ifp = ifnet_byindex(idx);
201 error = ifhwioctl(cmd, ifp, data, td);
202 if (error == ENOIOCTL)
208 netkqfilter(struct cdev *dev, struct knote *kn)
210 struct knlist *klist;
214 switch (kn->kn_filter) {
216 kn->kn_fop = &netdev_filtops;
226 ifp = ifnet_byindex(idx);
229 klist = &ifp->if_klist;
232 kn->kn_hook = (caddr_t)klist;
234 knlist_add(klist, kn, 0);
240 filt_netdetach(struct knote *kn)
242 struct knlist *klist = (struct knlist *)kn->kn_hook;
244 knlist_remove(klist, kn, 0);
248 filt_netdev(struct knote *kn, long hint)
250 struct knlist *klist = (struct knlist *)kn->kn_hook;
253 * Currently NOTE_EXIT is abused to indicate device detach.
255 if (hint == NOTE_EXIT) {
256 kn->kn_data = NOTE_LINKINV;
257 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
258 knlist_remove_inevent(klist, kn);
262 kn->kn_data = hint; /* current status */
263 if (kn->kn_sfflags & hint)
264 kn->kn_fflags |= hint;
265 return (kn->kn_fflags != 0);
269 * Network interface utility routines.
271 * Routines with ifa_ifwith* names take sockaddr *'s as
276 if_init(void *dummy __unused)
281 knlist_init(&ifklist, NULL, NULL, NULL, NULL);
282 if_grow(); /* create initial table */
283 ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
284 UID_ROOT, GID_WHEEL, 0600, "network");
292 struct ifindex_entry *e;
295 n = if_indexlim * sizeof(*e);
296 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
297 if (ifindex_table != NULL) {
298 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
299 free((caddr_t)ifindex_table, M_IFNET);
306 if_check(void *dummy __unused)
312 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */
313 TAILQ_FOREACH(ifp, &ifnet, if_link) {
314 if (ifp->if_snd.ifq_maxlen == 0) {
315 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
316 ifp->if_snd.ifq_maxlen = ifqmaxlen;
318 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
320 "XXX: driver didn't initialize queue mtx\n");
321 mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
322 MTX_NETWORK_LOCK, MTX_DEF);
331 * Allocate a struct ifnet and in index for an interface.
334 if_alloc(u_char type)
338 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
341 * Try to find an empty slot below if_index. If we fail, take
344 * XXX: should be locked!
346 for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) {
347 if (ifnet_byindex(ifp->if_index) == NULL)
350 /* Catch if_index overflow. */
351 if (ifp->if_index < 1) {
355 if (ifp->if_index > if_index)
356 if_index = ifp->if_index;
357 if (if_index >= if_indexlim)
359 ifnet_byindex(ifp->if_index) = ifp;
363 if (if_com_alloc[type] != NULL) {
364 ifp->if_l2com = if_com_alloc[type](type, ifp);
365 if (ifp->if_l2com == NULL) {
370 IF_ADDR_LOCK_INIT(ifp);
376 if_free(struct ifnet *ifp)
379 /* Do not add code to this function! Add it to if_free_type(). */
380 if_free_type(ifp, ifp->if_type);
384 if_free_type(struct ifnet *ifp, u_char type)
387 if (ifp != ifnet_byindex(ifp->if_index)) {
388 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
393 IF_ADDR_LOCK_DESTROY(ifp);
395 ifnet_byindex(ifp->if_index) = NULL;
397 /* XXX: should be locked with if_findindex() */
398 while (if_index > 0 && ifnet_byindex(if_index) == NULL)
401 if (if_com_free[type] != NULL)
402 if_com_free[type](ifp->if_l2com, type);
408 * Attach an interface to the
409 * list of "active" interfaces.
412 if_attach(struct ifnet *ifp)
414 unsigned socksize, ifasize;
415 int namelen, masklen;
416 struct sockaddr_dl *sdl;
419 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
420 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
423 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
424 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
425 IF_AFDATA_LOCK_INIT(ifp);
426 ifp->if_afdata_initialized = 0;
428 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
432 * The old code would work if the interface passed a pre-existing
433 * chain of ifaddrs to this code. We don't trust our callers to
434 * properly initialize the tailq, however, so we no longer allow
435 * this unlikely case.
437 TAILQ_INIT(&ifp->if_addrhead);
438 TAILQ_INIT(&ifp->if_prefixhead);
439 TAILQ_INIT(&ifp->if_multiaddrs);
440 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
441 getmicrotime(&ifp->if_lastchange);
442 ifp->if_data.ifi_epoch = time_uptime;
443 ifp->if_data.ifi_datalen = sizeof(struct if_data);
447 mac_create_ifnet(ifp);
450 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw,
451 unit2minor(ifp->if_index),
452 UID_ROOT, GID_WHEEL, 0600, "%s/%s",
453 net_cdevsw.d_name, ifp->if_xname);
454 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
455 net_cdevsw.d_name, ifp->if_index);
457 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
460 * create a Link Level name for this device
462 namelen = strlen(ifp->if_xname);
464 * Always save enough space for any possiable name so we can do
465 * a rename in place later.
467 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
468 socksize = masklen + ifp->if_addrlen;
469 if (socksize < sizeof(*sdl))
470 socksize = sizeof(*sdl);
471 socksize = roundup2(socksize, sizeof(long));
472 ifasize = sizeof(*ifa) + 2 * socksize;
473 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
475 sdl = (struct sockaddr_dl *)(ifa + 1);
476 sdl->sdl_len = socksize;
477 sdl->sdl_family = AF_LINK;
478 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
479 sdl->sdl_nlen = namelen;
480 sdl->sdl_index = ifp->if_index;
481 sdl->sdl_type = ifp->if_type;
484 ifa->ifa_rtrequest = link_rtrequest;
485 ifa->ifa_addr = (struct sockaddr *)sdl;
486 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
487 ifa->ifa_netmask = (struct sockaddr *)sdl;
488 sdl->sdl_len = masklen;
490 sdl->sdl_data[--namelen] = 0xff;
492 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
493 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
494 ifp->if_snd.altq_type = 0;
495 ifp->if_snd.altq_disc = NULL;
496 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
497 ifp->if_snd.altq_tbr = NULL;
498 ifp->if_snd.altq_ifp = ifp;
500 if (domain_init_status >= 2)
501 if_attachdomain1(ifp);
503 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
505 /* Announce the interface. */
506 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
510 if_attachdomain(void *dummy)
516 TAILQ_FOREACH(ifp, &ifnet, if_link)
517 if_attachdomain1(ifp);
520 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
521 if_attachdomain, NULL);
524 if_attachdomain1(struct ifnet *ifp)
532 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
533 * cannot lock ifp->if_afdata initialization, entirely.
535 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
539 if (ifp->if_afdata_initialized >= domain_init_status) {
540 IF_AFDATA_UNLOCK(ifp);
542 printf("if_attachdomain called more than once on %s\n",
546 ifp->if_afdata_initialized = domain_init_status;
547 IF_AFDATA_UNLOCK(ifp);
549 /* address family dependent data region */
550 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
551 for (dp = domains; dp; dp = dp->dom_next) {
552 if (dp->dom_ifattach)
553 ifp->if_afdata[dp->dom_family] =
554 (*dp->dom_ifattach)(ifp);
561 * Remove any network addresses from an interface.
565 if_purgeaddrs(struct ifnet *ifp)
567 struct ifaddr *ifa, *next;
569 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
571 if (ifa->ifa_addr->sa_family == AF_LINK)
574 /* XXX: Ugly!! ad hoc just for INET */
575 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
576 struct ifaliasreq ifr;
578 bzero(&ifr, sizeof(ifr));
579 ifr.ifra_addr = *ifa->ifa_addr;
580 if (ifa->ifa_dstaddr)
581 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
582 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
588 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
590 /* ifp_addrhead is already updated */
594 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
600 * Detach an interface, removing it from the
601 * list of "active" interfaces.
603 * XXXRW: There are some significant questions about event ordering, and
604 * how to prevent things from starting to use the interface during detach.
607 if_detach(struct ifnet *ifp)
610 struct radix_node_head *rnh;
618 * Remove/wait for pending events.
620 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
623 /* Maybe hook to the generalized departure handler above?!? */
629 * Remove routes and flush queues.
634 if (ALTQ_IS_ENABLED(&ifp->if_snd))
635 altq_disable(&ifp->if_snd);
636 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
637 altq_detach(&ifp->if_snd);
648 * Remove all IPv6 kernel structs related to ifp. This should be done
649 * before removing routing entries below, since IPv6 interface direct
650 * routes are expected to be removed by the IPv6-specific kernel API.
651 * Otherwise, the kernel will detect some inconsistency and bark it.
656 * Remove link ifaddr pointer and maybe decrement if_index.
657 * Clean up all addresses.
660 destroy_dev(ifdev_byindex(ifp->if_index));
661 ifdev_byindex(ifp->if_index) = NULL;
663 /* We can now free link ifaddr. */
664 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
665 ifa = TAILQ_FIRST(&ifp->if_addrhead);
666 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
671 * Delete all remaining routes using this interface
672 * Unfortuneatly the only way to do this is to slog through
673 * the entire routing table looking for routes which point
674 * to this interface...oh well...
676 for (i = 1; i <= AF_MAX; i++) {
677 if ((rnh = rt_tables[i]) == NULL)
679 RADIX_NODE_HEAD_LOCK(rnh);
680 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
681 RADIX_NODE_HEAD_UNLOCK(rnh);
684 /* Announce that the interface is gone. */
685 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
686 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
689 for (dp = domains; dp; dp = dp->dom_next) {
690 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
691 (*dp->dom_ifdetach)(ifp,
692 ifp->if_afdata[dp->dom_family]);
694 IF_AFDATA_UNLOCK(ifp);
697 mac_destroy_ifnet(ifp);
699 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
700 knlist_clear(&ifp->if_klist, 0);
701 knlist_destroy(&ifp->if_klist);
704 TAILQ_FOREACH(iter, &ifnet, if_link)
710 TAILQ_REMOVE(&ifnet, ifp, if_link);
712 mtx_destroy(&ifp->if_snd.ifq_mtx);
713 IF_AFDATA_DESTROY(ifp);
718 * Delete Routes for a Network Interface
720 * Called for each routing entry via the rnh->rnh_walktree() call above
721 * to delete all route entries referencing a detaching network interface.
724 * rn pointer to node in the routing table
725 * arg argument passed to rnh->rnh_walktree() - detaching interface
729 * errno failed - reason indicated
733 if_rtdel(struct radix_node *rn, void *arg)
735 struct rtentry *rt = (struct rtentry *)rn;
736 struct ifnet *ifp = arg;
739 if (rt->rt_ifp == ifp) {
742 * Protect (sorta) against walktree recursion problems
745 if ((rt->rt_flags & RTF_UP) == 0)
748 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
749 rt_mask(rt), rt->rt_flags,
750 (struct rtentry **) NULL);
752 log(LOG_WARNING, "if_rtdel: error %d\n", err);
759 #define sa_equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0)
762 * Locate an interface based on a complete address.
766 ifa_ifwithaddr(struct sockaddr *addr)
772 TAILQ_FOREACH(ifp, &ifnet, if_link)
773 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
774 if (ifa->ifa_addr->sa_family != addr->sa_family)
776 if (sa_equal(addr, ifa->ifa_addr))
778 /* IP6 doesn't have broadcast */
779 if ((ifp->if_flags & IFF_BROADCAST) &&
780 ifa->ifa_broadaddr &&
781 ifa->ifa_broadaddr->sa_len != 0 &&
782 sa_equal(ifa->ifa_broadaddr, addr))
792 * Locate the point to point interface with a given destination address.
796 ifa_ifwithdstaddr(struct sockaddr *addr)
802 TAILQ_FOREACH(ifp, &ifnet, if_link) {
803 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
805 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
806 if (ifa->ifa_addr->sa_family != addr->sa_family)
808 if (ifa->ifa_dstaddr &&
809 sa_equal(addr, ifa->ifa_dstaddr))
820 * Find an interface on a specific network. If many, choice
821 * is most specific found.
824 ifa_ifwithnet(struct sockaddr *addr)
828 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
829 u_int af = addr->sa_family;
830 char *addr_data = addr->sa_data, *cplim;
833 * AF_LINK addresses can be looked up directly by their index number,
834 * so do that if we can.
837 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
838 if (sdl->sdl_index && sdl->sdl_index <= if_index)
839 return (ifaddr_byindex(sdl->sdl_index));
843 * Scan though each interface, looking for ones that have
844 * addresses in this address family.
847 TAILQ_FOREACH(ifp, &ifnet, if_link) {
848 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
849 char *cp, *cp2, *cp3;
851 if (ifa->ifa_addr->sa_family != af)
853 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
855 * This is a bit broken as it doesn't
856 * take into account that the remote end may
857 * be a single node in the network we are
859 * The trouble is that we don't know the
860 * netmask for the remote end.
862 if (ifa->ifa_dstaddr != 0 &&
863 sa_equal(addr, ifa->ifa_dstaddr))
867 * if we have a special address handler,
868 * then use it instead of the generic one.
870 if (ifa->ifa_claim_addr) {
871 if ((*ifa->ifa_claim_addr)(ifa, addr))
877 * Scan all the bits in the ifa's address.
878 * If a bit dissagrees with what we are
879 * looking for, mask it with the netmask
880 * to see if it really matters.
883 if (ifa->ifa_netmask == 0)
886 cp2 = ifa->ifa_addr->sa_data;
887 cp3 = ifa->ifa_netmask->sa_data;
888 cplim = ifa->ifa_netmask->sa_len
889 + (char *)ifa->ifa_netmask;
891 if ((*cp++ ^ *cp2++) & *cp3++)
892 goto next; /* next address! */
894 * If the netmask of what we just found
895 * is more specific than what we had before
896 * (if we had one) then remember the new one
897 * before continuing to search
898 * for an even better one.
900 if (ifa_maybe == 0 ||
901 rn_refines((caddr_t)ifa->ifa_netmask,
902 (caddr_t)ifa_maybe->ifa_netmask))
914 * Find an interface address specific to an interface best matching
918 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
921 char *cp, *cp2, *cp3;
923 struct ifaddr *ifa_maybe = 0;
924 u_int af = addr->sa_family;
928 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
929 if (ifa->ifa_addr->sa_family != af)
933 if (ifa->ifa_netmask == 0) {
934 if (sa_equal(addr, ifa->ifa_addr) ||
936 sa_equal(addr, ifa->ifa_dstaddr)))
940 if (ifp->if_flags & IFF_POINTOPOINT) {
941 if (sa_equal(addr, ifa->ifa_dstaddr))
945 cp2 = ifa->ifa_addr->sa_data;
946 cp3 = ifa->ifa_netmask->sa_data;
947 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
948 for (; cp3 < cplim; cp3++)
949 if ((*cp++ ^ *cp2++) & *cp3)
960 #include <net/route.h>
963 * Default action when installing a route with a Link Level gateway.
964 * Lookup an appropriate real ifa to point to.
965 * This should be moved to /sys/net/link.c eventually.
968 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
970 struct ifaddr *ifa, *oifa;
971 struct sockaddr *dst;
976 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
977 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
979 ifa = ifaof_ifpforaddr(dst, ifp);
981 IFAREF(ifa); /* XXX */
985 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
986 ifa->ifa_rtrequest(cmd, rt, info);
991 * Mark an interface down and notify protocols of
993 * NOTE: must be called at splnet or eqivalent.
996 if_unroute(struct ifnet *ifp, int flag, int fam)
1000 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1002 ifp->if_flags &= ~flag;
1003 getmicrotime(&ifp->if_lastchange);
1004 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1005 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1006 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1007 if_qflush(&ifp->if_snd);
1010 carp_carpdev_state(ifp->if_carp);
1016 * Mark an interface up and notify protocols of
1018 * NOTE: must be called at splnet or eqivalent.
1021 if_route(struct ifnet *ifp, int flag, int fam)
1025 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1027 ifp->if_flags |= flag;
1028 getmicrotime(&ifp->if_lastchange);
1029 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1030 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1031 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1034 carp_carpdev_state(ifp->if_carp);
1042 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1045 * Handle a change in the interface link state. To avoid LORs
1046 * between driver lock and upper layer locks, as well as possible
1047 * recursions, we post event to taskqueue, and all job
1048 * is done in static do_link_state_change().
1051 if_link_state_change(struct ifnet *ifp, int link_state)
1053 /* Return if state hasn't changed. */
1054 if (ifp->if_link_state == link_state)
1057 ifp->if_link_state = link_state;
1059 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1063 do_link_state_change(void *arg, int pending)
1065 struct ifnet *ifp = (struct ifnet *)arg;
1066 int link_state = ifp->if_link_state;
1069 /* Notify that the link state has changed. */
1071 if (link_state == LINK_STATE_UP)
1073 else if (link_state == LINK_STATE_DOWN)
1074 link = NOTE_LINKDOWN;
1076 link = NOTE_LINKINV;
1077 KNOTE_UNLOCKED(&ifp->if_klist, link);
1078 if (ifp->if_nvlans != 0)
1079 (*vlan_link_state_p)(ifp, link);
1081 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1082 IFP2AC(ifp)->ac_netgraph != NULL)
1083 (*ng_ether_link_state_p)(ifp, link_state);
1086 carp_carpdev_state(ifp->if_carp);
1088 if (ifp->if_bridge) {
1089 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1090 (*bstp_linkstate_p)(ifp, link_state);
1093 devctl_notify("IFNET", ifp->if_xname,
1094 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
1096 if_printf(ifp, "%d link states coalesced\n", pending);
1097 if (log_link_state_change)
1098 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1099 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1103 * Mark an interface down and notify protocols of
1105 * NOTE: must be called at splnet or eqivalent.
1108 if_down(struct ifnet *ifp)
1111 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1115 * Mark an interface up and notify protocols of
1117 * NOTE: must be called at splnet or eqivalent.
1120 if_up(struct ifnet *ifp)
1123 if_route(ifp, IFF_UP, AF_UNSPEC);
1127 * Flush an interface queue.
1130 if_qflush(struct ifaltq *ifq)
1136 if (ALTQ_IS_ENABLED(ifq))
1140 while ((m = n) != 0) {
1151 * Handle interface watchdog timer routines. Called
1152 * from softclock, we decrement timers (if set) and
1153 * call the appropriate interface routine on expiration.
1155 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1156 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab
1157 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1160 if_slowtimo(void *arg)
1166 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1167 if (ifp->if_timer == 0 || --ifp->if_timer)
1169 if (ifp->if_watchdog)
1170 (*ifp->if_watchdog)(ifp);
1174 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1178 * Map interface name to
1179 * interface structure pointer.
1182 ifunit(const char *name)
1187 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1188 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1196 * Hardware specific interface ioctls.
1199 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1204 int new_flags, temp_flags;
1205 size_t namelen, onamelen;
1206 char new_name[IFNAMSIZ];
1208 struct sockaddr_dl *sdl;
1210 ifr = (struct ifreq *)data;
1213 ifr->ifr_index = ifp->if_index;
1217 temp_flags = ifp->if_flags | ifp->if_drv_flags;
1218 ifr->ifr_flags = temp_flags & 0xffff;
1219 ifr->ifr_flagshigh = temp_flags >> 16;
1223 ifr->ifr_reqcap = ifp->if_capabilities;
1224 ifr->ifr_curcap = ifp->if_capenable;
1229 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
1234 ifr->ifr_metric = ifp->if_metric;
1238 ifr->ifr_mtu = ifp->if_mtu;
1242 ifr->ifr_phys = ifp->if_physical;
1250 * Currently, no driver owned flags pass the IFF_CANTCHANGE
1251 * check, so we don't need special handling here yet.
1253 new_flags = (ifr->ifr_flags & 0xffff) |
1254 (ifr->ifr_flagshigh << 16);
1255 if (ifp->if_flags & IFF_SMART) {
1256 /* Smart drivers twiddle their own routes */
1257 } else if (ifp->if_flags & IFF_UP &&
1258 (new_flags & IFF_UP) == 0) {
1262 } else if (new_flags & IFF_UP &&
1263 (ifp->if_flags & IFF_UP) == 0) {
1268 /* See if permanently promiscuous mode bit is about to flip */
1269 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
1270 if (new_flags & IFF_PPROMISC)
1271 ifp->if_flags |= IFF_PROMISC;
1272 else if (ifp->if_pcount == 0)
1273 ifp->if_flags &= ~IFF_PROMISC;
1274 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
1276 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
1278 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1279 (new_flags &~ IFF_CANTCHANGE);
1280 if (ifp->if_ioctl) {
1282 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1283 IFF_UNLOCKGIANT(ifp);
1285 getmicrotime(&ifp->if_lastchange);
1292 if (ifp->if_ioctl == NULL)
1293 return (EOPNOTSUPP);
1294 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1297 error = (*ifp->if_ioctl)(ifp, cmd, data);
1298 IFF_UNLOCKGIANT(ifp);
1300 getmicrotime(&ifp->if_lastchange);
1305 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
1313 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1316 if (new_name[0] == '\0')
1318 if (ifunit(new_name) != NULL)
1321 /* Announce the departure of the interface. */
1322 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1323 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1325 log(LOG_INFO, "%s: changing name to '%s'\n",
1326 ifp->if_xname, new_name);
1328 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1331 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1332 namelen = strlen(new_name);
1333 onamelen = sdl->sdl_nlen;
1335 * Move the address if needed. This is safe because we
1336 * allocate space for a name of length IFNAMSIZ when we
1337 * create this in if_attach().
1339 if (namelen != onamelen) {
1340 bcopy(sdl->sdl_data + onamelen,
1341 sdl->sdl_data + namelen, sdl->sdl_alen);
1343 bcopy(new_name, sdl->sdl_data, namelen);
1344 sdl->sdl_nlen = namelen;
1345 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1346 bzero(sdl->sdl_data, onamelen);
1347 while (namelen != 0)
1348 sdl->sdl_data[--namelen] = 0xff;
1351 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1352 /* Announce the return of the interface. */
1353 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1360 ifp->if_metric = ifr->ifr_metric;
1361 getmicrotime(&ifp->if_lastchange);
1368 if (ifp->if_ioctl == NULL)
1369 return (EOPNOTSUPP);
1371 error = (*ifp->if_ioctl)(ifp, cmd, data);
1372 IFF_UNLOCKGIANT(ifp);
1374 getmicrotime(&ifp->if_lastchange);
1379 u_long oldmtu = ifp->if_mtu;
1384 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1386 if (ifp->if_ioctl == NULL)
1387 return (EOPNOTSUPP);
1389 error = (*ifp->if_ioctl)(ifp, cmd, data);
1390 IFF_UNLOCKGIANT(ifp);
1392 getmicrotime(&ifp->if_lastchange);
1396 * If the link MTU changed, do network layer specific procedure.
1398 if (ifp->if_mtu != oldmtu) {
1412 /* Don't allow group membership on non-multicast interfaces. */
1413 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1414 return (EOPNOTSUPP);
1416 /* Don't let users screw up protocols' entries. */
1417 if (ifr->ifr_addr.sa_family != AF_LINK)
1420 if (cmd == SIOCADDMULTI) {
1421 struct ifmultiaddr *ifma;
1422 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1424 error = if_delmulti(ifp, &ifr->ifr_addr);
1427 getmicrotime(&ifp->if_lastchange);
1430 case SIOCSIFPHYADDR:
1431 case SIOCDIFPHYADDR:
1433 case SIOCSIFPHYADDR_IN6:
1435 case SIOCSLIFPHYADDR:
1437 case SIOCSIFGENERIC:
1441 if (ifp->if_ioctl == NULL)
1442 return (EOPNOTSUPP);
1444 error = (*ifp->if_ioctl)(ifp, cmd, data);
1445 IFF_UNLOCKGIANT(ifp);
1447 getmicrotime(&ifp->if_lastchange);
1451 ifs = (struct ifstat *)data;
1452 ifs->ascii[0] = '\0';
1454 case SIOCGIFPSRCADDR:
1455 case SIOCGIFPDSTADDR:
1456 case SIOCGLIFPHYADDR:
1458 case SIOCGIFGENERIC:
1459 if (ifp->if_ioctl == NULL)
1460 return (EOPNOTSUPP);
1462 error = (*ifp->if_ioctl)(ifp, cmd, data);
1463 IFF_UNLOCKGIANT(ifp);
1470 error = if_setlladdr(ifp,
1471 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1485 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1495 return (ifconf(cmd, data));
1497 ifr = (struct ifreq *)data;
1502 if ((error = suser(td)) != 0)
1504 return ((cmd == SIOCIFCREATE) ?
1505 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1506 if_clone_destroy(ifr->ifr_name));
1508 case SIOCIFGCLONERS:
1509 return (if_clone_list((struct if_clonereq *)data));
1512 ifp = ifunit(ifr->ifr_name);
1516 error = ifhwioctl(cmd, ifp, data, td);
1517 if (error != ENOIOCTL)
1520 oif_flags = ifp->if_flags;
1521 if (so->so_proto == 0)
1522 return (EOPNOTSUPP);
1524 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1533 case SIOCSIFDSTADDR:
1535 case SIOCSIFBRDADDR:
1536 case SIOCSIFNETMASK:
1537 #if BYTE_ORDER != BIG_ENDIAN
1538 if (ifr->ifr_addr.sa_family == 0 &&
1539 ifr->ifr_addr.sa_len < 16) {
1540 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1541 ifr->ifr_addr.sa_len = 16;
1544 if (ifr->ifr_addr.sa_len == 0)
1545 ifr->ifr_addr.sa_len = 16;
1553 case OSIOCGIFDSTADDR:
1554 cmd = SIOCGIFDSTADDR;
1557 case OSIOCGIFBRDADDR:
1558 cmd = SIOCGIFBRDADDR;
1561 case OSIOCGIFNETMASK:
1562 cmd = SIOCGIFNETMASK;
1564 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
1571 case OSIOCGIFDSTADDR:
1572 case OSIOCGIFBRDADDR:
1573 case OSIOCGIFNETMASK:
1574 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1578 #endif /* COMPAT_43 */
1580 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1582 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1583 if (ifp->if_flags & IFF_UP) {
1594 * The code common to handling reference counted flags,
1595 * e.g., in ifpromisc() and if_allmulti().
1596 * The "pflag" argument can specify a permanent mode flag to check,
1597 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
1599 * Only to be used on stack-owned flags, not driver-owned flags.
1602 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
1606 int oldflags, oldcount;
1608 /* Sanity checks to catch programming errors */
1609 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
1610 ("%s: setting driver-owned flag %d", __func__, flag));
1613 KASSERT(*refcount >= 0,
1614 ("%s: increment negative refcount %d for flag %d",
1615 __func__, *refcount, flag));
1617 KASSERT(*refcount > 0,
1618 ("%s: decrement non-positive refcount %d for flag %d",
1619 __func__, *refcount, flag));
1621 /* In case this mode is permanent, just touch refcount */
1622 if (ifp->if_flags & pflag) {
1623 *refcount += onswitch ? 1 : -1;
1627 /* Save ifnet parameters for if_ioctl() may fail */
1628 oldcount = *refcount;
1629 oldflags = ifp->if_flags;
1632 * See if we aren't the only and touching refcount is enough.
1633 * Actually toggle interface flag if we are the first or last.
1638 ifp->if_flags |= flag;
1642 ifp->if_flags &= ~flag;
1645 /* Call down the driver since we've changed interface flags */
1646 if (ifp->if_ioctl == NULL) {
1650 ifr.ifr_flags = ifp->if_flags & 0xffff;
1651 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1653 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1654 IFF_UNLOCKGIANT(ifp);
1657 /* Notify userland that interface flags have changed */
1662 /* Recover after driver error */
1663 *refcount = oldcount;
1664 ifp->if_flags = oldflags;
1669 * Set/clear promiscuous mode on interface ifp based on the truth value
1670 * of pswitch. The calls are reference counted so that only the first
1671 * "on" request actually has an effect, as does the final "off" request.
1672 * Results are undefined if the "off" and "on" requests are not matched.
1675 ifpromisc(struct ifnet *ifp, int pswitch)
1678 int oldflags = ifp->if_flags;
1680 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
1681 &ifp->if_pcount, pswitch);
1682 /* If promiscuous mode status has changed, log a message */
1683 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
1684 log(LOG_INFO, "%s: promiscuous mode %s\n",
1686 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
1691 * Return interface configuration
1692 * of system. List may be used
1693 * in later ioctl's (above) to get
1694 * other information.
1698 ifconf(u_long cmd, caddr_t data)
1700 struct ifconf *ifc = (struct ifconf *)data;
1705 int error, full = 0, valid_len, max_len;
1707 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
1708 max_len = MAXPHYS - 1;
1710 /* Prevent hostile input from being able to crash the system */
1711 if (ifc->ifc_len <= 0)
1715 if (ifc->ifc_len <= max_len) {
1716 max_len = ifc->ifc_len;
1719 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
1723 IFNET_RLOCK(); /* could sleep XXX */
1724 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1728 * Zero the ifr_name buffer to make sure we don't
1729 * disclose the contents of the stack.
1731 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
1733 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1734 >= sizeof(ifr.ifr_name)) {
1737 return (ENAMETOOLONG);
1741 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1742 struct sockaddr *sa = ifa->ifa_addr;
1744 if (jailed(curthread->td_ucred) &&
1745 prison_if(curthread->td_ucred, sa))
1749 if (cmd == OSIOCGIFCONF) {
1750 struct osockaddr *osa =
1751 (struct osockaddr *)&ifr.ifr_addr;
1753 osa->sa_family = sa->sa_family;
1754 sbuf_bcat(sb, &ifr, sizeof(ifr));
1755 max_len += sizeof(ifr);
1758 if (sa->sa_len <= sizeof(*sa)) {
1760 sbuf_bcat(sb, &ifr, sizeof(ifr));
1761 max_len += sizeof(ifr);
1764 offsetof(struct ifreq, ifr_addr));
1765 max_len += offsetof(struct ifreq, ifr_addr);
1766 sbuf_bcat(sb, sa, sa->sa_len);
1767 max_len += sa->sa_len;
1770 if (!sbuf_overflowed(sb))
1771 valid_len = sbuf_len(sb);
1774 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1775 sbuf_bcat(sb, &ifr, sizeof(ifr));
1776 max_len += sizeof(ifr);
1778 if (!sbuf_overflowed(sb))
1779 valid_len = sbuf_len(sb);
1785 * If we didn't allocate enough space (uncommon), try again. If
1786 * we have already allocated as much space as we are allowed,
1787 * return what we've got.
1789 if (valid_len != max_len && !full) {
1794 ifc->ifc_len = valid_len;
1796 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
1802 * Just like ifpromisc(), but for all-multicast-reception mode.
1805 if_allmulti(struct ifnet *ifp, int onswitch)
1808 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
1811 static struct ifmultiaddr *
1812 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
1814 struct ifmultiaddr *ifma;
1816 IF_ADDR_LOCK_ASSERT(ifp);
1818 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1819 if (sa_equal(ifma->ifma_addr, sa))
1827 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
1828 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
1829 * the ifnet multicast address list here, so the caller must do that and
1830 * other setup work (such as notifying the device driver). The reference
1831 * count is initialized to 1.
1833 static struct ifmultiaddr *
1834 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
1837 struct ifmultiaddr *ifma;
1838 struct sockaddr *dupsa;
1840 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
1845 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
1846 if (dupsa == NULL) {
1847 FREE(ifma, M_IFMADDR);
1850 bcopy(sa, dupsa, sa->sa_len);
1851 ifma->ifma_addr = dupsa;
1853 ifma->ifma_ifp = ifp;
1854 ifma->ifma_refcount = 1;
1855 ifma->ifma_protospec = NULL;
1858 ifma->ifma_lladdr = NULL;
1862 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
1863 if (dupsa == NULL) {
1864 FREE(ifma->ifma_addr, M_IFMADDR);
1865 FREE(ifma, M_IFMADDR);
1868 bcopy(llsa, dupsa, llsa->sa_len);
1869 ifma->ifma_lladdr = dupsa;
1875 * if_freemulti: free ifmultiaddr structure and possibly attached related
1876 * addresses. The caller is responsible for implementing reference
1877 * counting, notifying the driver, handling routing messages, and releasing
1878 * any dependent link layer state.
1881 if_freemulti(struct ifmultiaddr *ifma)
1884 KASSERT(ifma->ifma_refcount == 1, ("if_freemulti: refcount %d",
1885 ifma->ifma_refcount));
1886 KASSERT(ifma->ifma_protospec == NULL,
1887 ("if_freemulti: protospec not NULL"));
1889 if (ifma->ifma_lladdr != NULL)
1890 FREE(ifma->ifma_lladdr, M_IFMADDR);
1891 FREE(ifma->ifma_addr, M_IFMADDR);
1892 FREE(ifma, M_IFMADDR);
1896 * Register an additional multicast address with a network interface.
1898 * - If the address is already present, bump the reference count on the
1899 * address and return.
1900 * - If the address is not link-layer, look up a link layer address.
1901 * - Allocate address structures for one or both addresses, and attach to the
1902 * multicast address list on the interface. If automatically adding a link
1903 * layer address, the protocol address will own a reference to the link
1904 * layer address, to be freed when it is freed.
1905 * - Notify the network device driver of an addition to the multicast address
1908 * 'sa' points to caller-owned memory with the desired multicast address.
1910 * 'retifma' will be used to return a pointer to the resulting multicast
1911 * address reference, if desired.
1914 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
1915 struct ifmultiaddr **retifma)
1917 struct ifmultiaddr *ifma, *ll_ifma;
1918 struct sockaddr *llsa;
1922 * If the address is already present, return a new reference to it;
1923 * otherwise, allocate storage and set up a new address.
1926 ifma = if_findmulti(ifp, sa);
1928 ifma->ifma_refcount++;
1929 if (retifma != NULL)
1931 IF_ADDR_UNLOCK(ifp);
1936 * The address isn't already present; resolve the protocol address
1937 * into a link layer address, and then look that up, bump its
1938 * refcount or allocate an ifma for that also. If 'llsa' was
1939 * returned, we will need to free it later.
1943 if (ifp->if_resolvemulti != NULL) {
1944 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1950 * Allocate the new address. Don't hook it up yet, as we may also
1951 * need to allocate a link layer multicast address.
1953 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
1960 * If a link layer address is found, we'll need to see if it's
1961 * already present in the address list, or allocate is as well.
1962 * When this block finishes, the link layer address will be on the
1966 ll_ifma = if_findmulti(ifp, llsa);
1967 if (ll_ifma == NULL) {
1968 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
1969 if (ll_ifma == NULL) {
1974 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
1977 ll_ifma->ifma_refcount++;
1981 * We now have a new multicast address, ifma, and possibly a new or
1982 * referenced link layer address. Add the primary address to the
1983 * ifnet address list.
1985 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1987 if (retifma != NULL)
1991 * Must generate the message while holding the lock so that 'ifma'
1992 * pointer is still valid.
1994 * XXXRW: How come we don't announce ll_ifma?
1996 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1997 IF_ADDR_UNLOCK(ifp);
2000 * We are certain we have added something, so call down to the
2001 * interface to let them know about it.
2003 if (ifp->if_ioctl != NULL) {
2005 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2006 IFF_UNLOCKGIANT(ifp);
2010 FREE(llsa, M_IFMADDR);
2016 FREE(llsa, M_IFMADDR);
2019 IF_ADDR_UNLOCK(ifp);
2024 * Remove a reference to a multicast address on this interface. Yell
2025 * if the request does not match an existing membership.
2028 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2030 struct ifmultiaddr *ifma, *ll_ifma;
2033 ifma = if_findmulti(ifp, sa);
2035 IF_ADDR_UNLOCK(ifp);
2039 if (ifma->ifma_refcount > 1) {
2040 ifma->ifma_refcount--;
2041 IF_ADDR_UNLOCK(ifp);
2045 sa = ifma->ifma_lladdr;
2047 ll_ifma = if_findmulti(ifp, sa);
2052 * XXXRW: How come we don't announce ll_ifma?
2054 rt_newmaddrmsg(RTM_DELMADDR, ifma);
2056 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
2059 if (ll_ifma != NULL) {
2060 if (ll_ifma->ifma_refcount == 1) {
2061 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link);
2062 if_freemulti(ll_ifma);
2064 ll_ifma->ifma_refcount--;
2066 IF_ADDR_UNLOCK(ifp);
2069 * Make sure the interface driver is notified
2070 * in the case of a link layer mcast group being left.
2072 if (ifp->if_ioctl) {
2074 (void) (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2075 IFF_UNLOCKGIANT(ifp);
2082 * Set the link layer address on an interface.
2084 * At this time we only support certain types of interfaces,
2085 * and we don't allow the length of the address to change.
2088 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
2090 struct sockaddr_dl *sdl;
2097 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2100 if (len != sdl->sdl_alen) /* don't allow length to change */
2102 switch (ifp->if_type) {
2110 bcopy(lladdr, LLADDR(sdl), len);
2116 * If the interface is already up, we need
2117 * to re-init it in order to reprogram its
2120 if ((ifp->if_flags & IFF_UP) != 0) {
2121 if (ifp->if_ioctl) {
2123 ifp->if_flags &= ~IFF_UP;
2124 ifr.ifr_flags = ifp->if_flags & 0xffff;
2125 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2126 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2127 ifp->if_flags |= IFF_UP;
2128 ifr.ifr_flags = ifp->if_flags & 0xffff;
2129 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2130 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2131 IFF_UNLOCKGIANT(ifp);
2135 * Also send gratuitous ARPs to notify other nodes about
2136 * the address change.
2138 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2139 if (ifa->ifa_addr != NULL &&
2140 ifa->ifa_addr->sa_family == AF_INET)
2141 arp_ifinit(ifp, ifa);
2149 * The name argument must be a pointer to storage which will last as
2150 * long as the interface does. For physical devices, the result of
2151 * device_get_name(dev) is a good choice and for pseudo-devices a
2152 * static string works well.
2155 if_initname(struct ifnet *ifp, const char *name, int unit)
2157 ifp->if_dname = name;
2158 ifp->if_dunit = unit;
2159 if (unit != IF_DUNIT_NONE)
2160 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
2162 strlcpy(ifp->if_xname, name, IFNAMSIZ);
2166 if_printf(struct ifnet *ifp, const char * fmt, ...)
2171 retval = printf("%s: ", ifp->if_xname);
2173 retval += vprintf(fmt, ap);
2179 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
2180 * be called without Giant. However, we often can't acquire the Giant lock
2181 * at those points; instead, we run it via a task queue that holds Giant via
2182 * if_start_deferred.
2184 * XXXRW: We need to make sure that the ifnet isn't fully detached until any
2185 * outstanding if_start_deferred() tasks that will run after the free. This
2186 * probably means waiting in if_detach().
2189 if_start(struct ifnet *ifp)
2194 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) {
2195 if (mtx_owned(&Giant))
2196 (*(ifp)->if_start)(ifp);
2198 taskqueue_enqueue(taskqueue_swi_giant,
2199 &ifp->if_starttask);
2201 (*(ifp)->if_start)(ifp);
2205 if_start_deferred(void *context, int pending)
2210 * This code must be entered with Giant, and should never run if
2211 * we're not running with debug.mpsafenet.
2213 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet"));
2217 (ifp->if_start)(ifp);
2221 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
2226 if (_IF_QFULL(ifq)) {
2233 ifp->if_obytes += m->m_pkthdr.len + adjust;
2234 if (m->m_flags & (M_BCAST|M_MCAST))
2236 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
2238 _IF_ENQUEUE(ifq, m);
2240 if (ifp != NULL && !active)
2246 if_register_com_alloc(u_char type,
2247 if_com_alloc_t *a, if_com_free_t *f)
2250 KASSERT(if_com_alloc[type] == NULL,
2251 ("if_register_com_alloc: %d already registered", type));
2252 KASSERT(if_com_free[type] == NULL,
2253 ("if_register_com_alloc: %d free already registered", type));
2255 if_com_alloc[type] = a;
2256 if_com_free[type] = f;
2260 if_deregister_com_alloc(u_char type)
2263 KASSERT(if_com_alloc[type] == NULL,
2264 ("if_deregister_com_alloc: %d not registered", type));
2265 KASSERT(if_com_free[type] == NULL,
2266 ("if_deregister_com_alloc: %d free not registered", type));
2267 if_com_alloc[type] = NULL;
2268 if_com_free[type] = NULL;