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_arp.h>
63 #include <net/if_clone.h>
64 #include <net/if_dl.h>
65 #include <net/if_types.h>
66 #include <net/if_var.h>
67 #include <net/radix.h>
68 #include <net/route.h>
70 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
75 #include <netinet6/in6_var.h>
76 #include <netinet6/in6_ifattach.h>
80 #include <netinet/if_ether.h>
83 #include <netinet/ip_carp.h>
86 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
87 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
89 /* Log link state change events */
90 static int log_link_state_change = 1;
92 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
93 &log_link_state_change, 0,
94 "log interface link state change events");
96 void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
97 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
99 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
101 static void if_attachdomain(void *);
102 static void if_attachdomain1(struct ifnet *);
103 static int ifconf(u_long, caddr_t);
104 static void if_grow(void);
105 static void if_init(void *);
106 static void if_check(void *);
107 static int if_findindex(struct ifnet *);
108 static void if_qflush(struct ifaltq *);
109 static void if_route(struct ifnet *, int flag, int fam);
110 static int if_setflag(struct ifnet *, int, int, int *, int);
111 static void if_slowtimo(void *);
112 static void if_unroute(struct ifnet *, int flag, int fam);
113 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
114 static int if_rtdel(struct radix_node *, void *);
115 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
116 static void if_start_deferred(void *context, int pending);
117 static void do_link_state_change(void *, int);
120 * XXX: declare here to avoid to include many inet6 related files..
121 * should be more generalized?
123 extern void nd6_setmtu(struct ifnet *);
127 struct ifindex_entry *ifindex_table = NULL;
128 int ifqmaxlen = IFQ_MAXLEN;
129 struct ifnethead ifnet; /* depend on static init XXX */
130 struct mtx ifnet_lock;
131 static if_com_alloc_t *if_com_alloc[256];
132 static if_com_free_t *if_com_free[256];
134 static int if_indexlim = 8;
135 static struct knlist ifklist;
137 static void filt_netdetach(struct knote *kn);
138 static int filt_netdev(struct knote *kn, long hint);
140 static struct filterops netdev_filtops =
141 { 1, NULL, filt_netdetach, filt_netdev };
144 * System initialization
146 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
147 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
149 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
150 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
151 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
153 static d_open_t netopen;
154 static d_close_t netclose;
155 static d_ioctl_t netioctl;
156 static d_kqfilter_t netkqfilter;
158 static struct cdevsw net_cdevsw = {
159 .d_version = D_VERSION,
160 .d_flags = D_NEEDGIANT,
165 .d_kqfilter = netkqfilter,
169 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
175 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
181 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
186 /* only support interface specific ioctls */
187 if (IOCGROUP(cmd) != 'i')
192 * special network device, not interface.
194 if (cmd == SIOCGIFCONF)
195 return (ifconf(cmd, data)); /* XXX remove cmd */
199 ifp = ifnet_byindex(idx);
203 error = ifhwioctl(cmd, ifp, data, td);
204 if (error == ENOIOCTL)
210 netkqfilter(struct cdev *dev, struct knote *kn)
212 struct knlist *klist;
216 switch (kn->kn_filter) {
218 kn->kn_fop = &netdev_filtops;
228 ifp = ifnet_byindex(idx);
231 klist = &ifp->if_klist;
234 kn->kn_hook = (caddr_t)klist;
236 knlist_add(klist, kn, 0);
242 filt_netdetach(struct knote *kn)
244 struct knlist *klist = (struct knlist *)kn->kn_hook;
246 knlist_remove(klist, kn, 0);
250 filt_netdev(struct knote *kn, long hint)
252 struct knlist *klist = (struct knlist *)kn->kn_hook;
255 * Currently NOTE_EXIT is abused to indicate device detach.
257 if (hint == NOTE_EXIT) {
258 kn->kn_data = NOTE_LINKINV;
259 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
260 knlist_remove_inevent(klist, kn);
264 kn->kn_data = hint; /* current status */
265 if (kn->kn_sfflags & hint)
266 kn->kn_fflags |= hint;
267 return (kn->kn_fflags != 0);
271 * Network interface utility routines.
273 * Routines with ifa_ifwith* names take sockaddr *'s as
278 if_init(void *dummy __unused)
283 knlist_init(&ifklist, NULL, NULL, NULL, NULL);
284 if_grow(); /* create initial table */
285 ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
286 UID_ROOT, GID_WHEEL, 0600, "network");
294 struct ifindex_entry *e;
297 n = if_indexlim * sizeof(*e);
298 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
299 if (ifindex_table != NULL) {
300 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
301 free((caddr_t)ifindex_table, M_IFNET);
308 if_check(void *dummy __unused)
314 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */
315 TAILQ_FOREACH(ifp, &ifnet, if_link) {
316 if (ifp->if_snd.ifq_maxlen == 0) {
317 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
318 ifp->if_snd.ifq_maxlen = ifqmaxlen;
320 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
322 "XXX: driver didn't initialize queue mtx\n");
323 mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
324 MTX_NETWORK_LOCK, MTX_DEF);
332 /* XXX: should be locked. */
334 if_findindex(struct ifnet *ifp)
337 char eaddr[18], devname[32];
338 const char *name, *p;
340 switch (ifp->if_type) {
341 case IFT_ETHER: /* these types use struct arpcom */
347 snprintf(eaddr, 18, "%6D", IFP2ENADDR(ifp), ":");
353 strlcpy(devname, ifp->if_xname, sizeof(devname));
354 name = net_cdevsw.d_name;
356 while ((resource_find_dev(&i, name, &unit, NULL, NULL)) == 0) {
357 if (resource_string_value(name, unit, "ether", &p) == 0)
358 if (strcmp(p, eaddr) == 0)
360 if (resource_string_value(name, unit, "dev", &p) == 0)
361 if (strcmp(p, devname) == 0)
367 if (ifaddr_byindex(unit) == NULL)
369 printf("%s%d in use, cannot hardwire it to %s.\n",
370 name, unit, devname);
372 for (unit = 1; ; unit++) {
373 if (unit <= if_index && ifaddr_byindex(unit) != NULL)
375 if (resource_string_value(name, unit, "ether", &p) == 0 ||
376 resource_string_value(name, unit, "dev", &p) == 0)
384 * Allocate a struct ifnet and in index for an interface.
387 if_alloc(u_char type)
391 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
393 /* XXX: This should fail if if_index is too big */
394 ifp->if_index = if_findindex(ifp);
395 if (ifp->if_index > if_index)
396 if_index = ifp->if_index;
397 if (if_index >= if_indexlim)
400 ifnet_byindex(ifp->if_index) = ifp;
404 if (if_com_alloc[type] != NULL) {
405 ifp->if_l2com = if_com_alloc[type](type, ifp);
406 if (ifp->if_l2com == NULL) {
416 if_free(struct ifnet *ifp)
419 if_free_type(ifp, ifp->if_type);
421 IF_ADDR_LOCK_DESTROY(ifp);
425 if_free_type(struct ifnet *ifp, u_char type)
428 if (ifp != ifnet_byindex(ifp->if_index)) {
429 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
434 ifnet_byindex(ifp->if_index) = NULL;
436 /* XXX: should be locked with if_findindex() */
437 while (if_index > 0 && ifaddr_byindex(if_index) == NULL)
440 if (if_com_free[type] != NULL)
441 if_com_free[type](ifp->if_l2com, type);
447 * Attach an interface to the
448 * list of "active" interfaces.
451 if_attach(struct ifnet *ifp)
453 unsigned socksize, ifasize;
454 int namelen, masklen;
455 struct sockaddr_dl *sdl;
458 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
459 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
462 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
463 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
464 IF_AFDATA_LOCK_INIT(ifp);
465 IF_ADDR_LOCK_INIT(ifp);
466 ifp->if_afdata_initialized = 0;
468 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
472 * The old code would work if the interface passed a pre-existing
473 * chain of ifaddrs to this code. We don't trust our callers to
474 * properly initialize the tailq, however, so we no longer allow
475 * this unlikely case.
477 TAILQ_INIT(&ifp->if_addrhead);
478 TAILQ_INIT(&ifp->if_prefixhead);
479 TAILQ_INIT(&ifp->if_multiaddrs);
480 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
481 getmicrotime(&ifp->if_lastchange);
482 ifp->if_data.ifi_epoch = time_uptime;
483 ifp->if_data.ifi_datalen = sizeof(struct if_data);
487 mac_create_ifnet(ifp);
490 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw,
491 unit2minor(ifp->if_index),
492 UID_ROOT, GID_WHEEL, 0600, "%s/%s",
493 net_cdevsw.d_name, ifp->if_xname);
494 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
495 net_cdevsw.d_name, ifp->if_index);
497 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
500 * create a Link Level name for this device
502 namelen = strlen(ifp->if_xname);
504 * Always save enough space for any possiable name so we can do
505 * a rename in place later.
507 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
508 socksize = masklen + ifp->if_addrlen;
509 if (socksize < sizeof(*sdl))
510 socksize = sizeof(*sdl);
511 socksize = roundup2(socksize, sizeof(long));
512 ifasize = sizeof(*ifa) + 2 * socksize;
513 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
515 sdl = (struct sockaddr_dl *)(ifa + 1);
516 sdl->sdl_len = socksize;
517 sdl->sdl_family = AF_LINK;
518 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
519 sdl->sdl_nlen = namelen;
520 sdl->sdl_index = ifp->if_index;
521 sdl->sdl_type = ifp->if_type;
522 ifaddr_byindex(ifp->if_index) = ifa;
524 ifa->ifa_rtrequest = link_rtrequest;
525 ifa->ifa_addr = (struct sockaddr *)sdl;
526 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
527 ifa->ifa_netmask = (struct sockaddr *)sdl;
528 sdl->sdl_len = masklen;
530 sdl->sdl_data[--namelen] = 0xff;
532 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
533 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
534 ifp->if_snd.altq_type = 0;
535 ifp->if_snd.altq_disc = NULL;
536 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
537 ifp->if_snd.altq_tbr = NULL;
538 ifp->if_snd.altq_ifp = ifp;
540 if (domain_init_status >= 2)
541 if_attachdomain1(ifp);
543 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
545 /* Announce the interface. */
546 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
550 if_attachdomain(void *dummy)
556 TAILQ_FOREACH(ifp, &ifnet, if_link)
557 if_attachdomain1(ifp);
560 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
561 if_attachdomain, NULL);
564 if_attachdomain1(struct ifnet *ifp)
572 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
573 * cannot lock ifp->if_afdata initialization, entirely.
575 if (IF_AFDATA_TRYLOCK(ifp) == 0) {
579 if (ifp->if_afdata_initialized >= domain_init_status) {
580 IF_AFDATA_UNLOCK(ifp);
582 printf("if_attachdomain called more than once on %s\n",
586 ifp->if_afdata_initialized = domain_init_status;
587 IF_AFDATA_UNLOCK(ifp);
589 /* address family dependent data region */
590 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
591 for (dp = domains; dp; dp = dp->dom_next) {
592 if (dp->dom_ifattach)
593 ifp->if_afdata[dp->dom_family] =
594 (*dp->dom_ifattach)(ifp);
601 * Remove any network addresses from an interface.
605 if_purgeaddrs(struct ifnet *ifp)
607 struct ifaddr *ifa, *next;
609 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
611 if (ifa->ifa_addr->sa_family == AF_LINK)
614 /* XXX: Ugly!! ad hoc just for INET */
615 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
616 struct ifaliasreq ifr;
618 bzero(&ifr, sizeof(ifr));
619 ifr.ifra_addr = *ifa->ifa_addr;
620 if (ifa->ifa_dstaddr)
621 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
622 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
628 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
630 /* ifp_addrhead is already updated */
634 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
640 * Detach an interface, removing it from the
641 * list of "active" interfaces and freeing the struct ifnet.
644 if_detach(struct ifnet *ifp)
647 struct radix_node_head *rnh;
655 * Remove/wait for pending events.
657 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
660 /* Maybe hook to the generalized departure handler above?!? */
666 * Remove routes and flush queues.
671 if (ALTQ_IS_ENABLED(&ifp->if_snd))
672 altq_disable(&ifp->if_snd);
673 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
674 altq_detach(&ifp->if_snd);
681 * Remove all IPv6 kernel structs related to ifp. This should be done
682 * before removing routing entries below, since IPv6 interface direct
683 * routes are expected to be removed by the IPv6-specific kernel API.
684 * Otherwise, the kernel will detect some inconsistency and bark it.
689 * Remove address from ifindex_table[] and maybe decrement if_index.
690 * Clean up all addresses.
692 ifaddr_byindex(ifp->if_index) = NULL;
693 destroy_dev(ifdev_byindex(ifp->if_index));
694 ifdev_byindex(ifp->if_index) = NULL;
696 /* We can now free link ifaddr. */
697 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
698 ifa = TAILQ_FIRST(&ifp->if_addrhead);
699 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
704 * Delete all remaining routes using this interface
705 * Unfortuneatly the only way to do this is to slog through
706 * the entire routing table looking for routes which point
707 * to this interface...oh well...
709 for (i = 1; i <= AF_MAX; i++) {
710 if ((rnh = rt_tables[i]) == NULL)
712 RADIX_NODE_HEAD_LOCK(rnh);
713 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
714 RADIX_NODE_HEAD_UNLOCK(rnh);
717 /* Announce that the interface is gone. */
718 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
719 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
722 for (dp = domains; dp; dp = dp->dom_next) {
723 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
724 (*dp->dom_ifdetach)(ifp,
725 ifp->if_afdata[dp->dom_family]);
727 IF_AFDATA_UNLOCK(ifp);
730 mac_destroy_ifnet(ifp);
732 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
733 knlist_clear(&ifp->if_klist, 0);
734 knlist_destroy(&ifp->if_klist);
737 TAILQ_FOREACH(iter, &ifnet, if_link)
743 TAILQ_REMOVE(&ifnet, ifp, if_link);
745 mtx_destroy(&ifp->if_snd.ifq_mtx);
746 IF_AFDATA_DESTROY(ifp);
751 * Delete Routes for a Network Interface
753 * Called for each routing entry via the rnh->rnh_walktree() call above
754 * to delete all route entries referencing a detaching network interface.
757 * rn pointer to node in the routing table
758 * arg argument passed to rnh->rnh_walktree() - detaching interface
762 * errno failed - reason indicated
766 if_rtdel(struct radix_node *rn, void *arg)
768 struct rtentry *rt = (struct rtentry *)rn;
769 struct ifnet *ifp = arg;
772 if (rt->rt_ifp == ifp) {
775 * Protect (sorta) against walktree recursion problems
778 if ((rt->rt_flags & RTF_UP) == 0)
781 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
782 rt_mask(rt), rt->rt_flags,
783 (struct rtentry **) NULL);
785 log(LOG_WARNING, "if_rtdel: error %d\n", err);
792 #define sa_equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0)
795 * Locate an interface based on a complete address.
799 ifa_ifwithaddr(struct sockaddr *addr)
805 TAILQ_FOREACH(ifp, &ifnet, if_link)
806 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
807 if (ifa->ifa_addr->sa_family != addr->sa_family)
809 if (sa_equal(addr, ifa->ifa_addr))
811 /* IP6 doesn't have broadcast */
812 if ((ifp->if_flags & IFF_BROADCAST) &&
813 ifa->ifa_broadaddr &&
814 ifa->ifa_broadaddr->sa_len != 0 &&
815 sa_equal(ifa->ifa_broadaddr, addr))
825 * Locate the point to point interface with a given destination address.
829 ifa_ifwithdstaddr(struct sockaddr *addr)
835 TAILQ_FOREACH(ifp, &ifnet, if_link) {
836 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
838 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
839 if (ifa->ifa_addr->sa_family != addr->sa_family)
841 if (ifa->ifa_dstaddr &&
842 sa_equal(addr, ifa->ifa_dstaddr))
853 * Find an interface on a specific network. If many, choice
854 * is most specific found.
857 ifa_ifwithnet(struct sockaddr *addr)
861 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
862 u_int af = addr->sa_family;
863 char *addr_data = addr->sa_data, *cplim;
866 * AF_LINK addresses can be looked up directly by their index number,
867 * so do that if we can.
870 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
871 if (sdl->sdl_index && sdl->sdl_index <= if_index)
872 return (ifaddr_byindex(sdl->sdl_index));
876 * Scan though each interface, looking for ones that have
877 * addresses in this address family.
880 TAILQ_FOREACH(ifp, &ifnet, if_link) {
881 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
882 char *cp, *cp2, *cp3;
884 if (ifa->ifa_addr->sa_family != af)
886 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
888 * This is a bit broken as it doesn't
889 * take into account that the remote end may
890 * be a single node in the network we are
892 * The trouble is that we don't know the
893 * netmask for the remote end.
895 if (ifa->ifa_dstaddr != 0 &&
896 sa_equal(addr, ifa->ifa_dstaddr))
900 * if we have a special address handler,
901 * then use it instead of the generic one.
903 if (ifa->ifa_claim_addr) {
904 if ((*ifa->ifa_claim_addr)(ifa, addr))
910 * Scan all the bits in the ifa's address.
911 * If a bit dissagrees with what we are
912 * looking for, mask it with the netmask
913 * to see if it really matters.
916 if (ifa->ifa_netmask == 0)
919 cp2 = ifa->ifa_addr->sa_data;
920 cp3 = ifa->ifa_netmask->sa_data;
921 cplim = ifa->ifa_netmask->sa_len
922 + (char *)ifa->ifa_netmask;
924 if ((*cp++ ^ *cp2++) & *cp3++)
925 goto next; /* next address! */
927 * If the netmask of what we just found
928 * is more specific than what we had before
929 * (if we had one) then remember the new one
930 * before continuing to search
931 * for an even better one.
933 if (ifa_maybe == 0 ||
934 rn_refines((caddr_t)ifa->ifa_netmask,
935 (caddr_t)ifa_maybe->ifa_netmask))
947 * Find an interface address specific to an interface best matching
951 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
954 char *cp, *cp2, *cp3;
956 struct ifaddr *ifa_maybe = 0;
957 u_int af = addr->sa_family;
961 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
962 if (ifa->ifa_addr->sa_family != af)
966 if (ifa->ifa_netmask == 0) {
967 if (sa_equal(addr, ifa->ifa_addr) ||
969 sa_equal(addr, ifa->ifa_dstaddr)))
973 if (ifp->if_flags & IFF_POINTOPOINT) {
974 if (sa_equal(addr, ifa->ifa_dstaddr))
978 cp2 = ifa->ifa_addr->sa_data;
979 cp3 = ifa->ifa_netmask->sa_data;
980 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
981 for (; cp3 < cplim; cp3++)
982 if ((*cp++ ^ *cp2++) & *cp3)
993 #include <net/route.h>
996 * Default action when installing a route with a Link Level gateway.
997 * Lookup an appropriate real ifa to point to.
998 * This should be moved to /sys/net/link.c eventually.
1001 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1003 struct ifaddr *ifa, *oifa;
1004 struct sockaddr *dst;
1009 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1010 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1012 ifa = ifaof_ifpforaddr(dst, ifp);
1014 IFAREF(ifa); /* XXX */
1018 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1019 ifa->ifa_rtrequest(cmd, rt, info);
1024 * Mark an interface down and notify protocols of
1026 * NOTE: must be called at splnet or eqivalent.
1029 if_unroute(struct ifnet *ifp, int flag, int fam)
1033 ifp->if_flags &= ~flag;
1034 getmicrotime(&ifp->if_lastchange);
1035 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1036 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1037 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1038 if_qflush(&ifp->if_snd);
1041 carp_carpdev_state(ifp->if_carp);
1047 * Mark an interface up and notify protocols of
1049 * NOTE: must be called at splnet or eqivalent.
1052 if_route(struct ifnet *ifp, int flag, int fam)
1056 ifp->if_flags |= flag;
1057 getmicrotime(&ifp->if_lastchange);
1058 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1059 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1060 pfctlinput(PRC_IFUP, ifa->ifa_addr);
1063 carp_carpdev_state(ifp->if_carp);
1071 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */
1074 * Handle a change in the interface link state. To avoid LORs
1075 * between driver lock and upper layer locks, as well as possible
1076 * recursions, we post event to taskqueue, and all job
1077 * is done in static do_link_state_change().
1080 if_link_state_change(struct ifnet *ifp, int link_state)
1082 /* Return if state hasn't changed. */
1083 if (ifp->if_link_state == link_state)
1086 ifp->if_link_state = link_state;
1088 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1092 do_link_state_change(void *arg, int pending)
1094 struct ifnet *ifp = (struct ifnet *)arg;
1095 int link_state = ifp->if_link_state;
1098 /* Notify that the link state has changed. */
1100 if (link_state == LINK_STATE_UP)
1102 else if (link_state == LINK_STATE_DOWN)
1103 link = NOTE_LINKDOWN;
1105 link = NOTE_LINKINV;
1106 KNOTE_UNLOCKED(&ifp->if_klist, link);
1107 if (ifp->if_nvlans != 0)
1108 (*vlan_link_state_p)(ifp, link);
1110 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1111 IFP2AC(ifp)->ac_netgraph != NULL)
1112 (*ng_ether_link_state_p)(ifp, link_state);
1115 carp_carpdev_state(ifp->if_carp);
1117 if (ifp->if_bridge) {
1118 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1119 (*bstp_linkstate_p)(ifp, link_state);
1122 devctl_notify("IFNET", ifp->if_xname,
1123 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
1125 if_printf(ifp, "%d link states coalesced\n", pending);
1126 if (log_link_state_change)
1127 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1128 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1132 * Mark an interface down and notify protocols of
1134 * NOTE: must be called at splnet or eqivalent.
1137 if_down(struct ifnet *ifp)
1140 if_unroute(ifp, IFF_UP, AF_UNSPEC);
1144 * Mark an interface up and notify protocols of
1146 * NOTE: must be called at splnet or eqivalent.
1149 if_up(struct ifnet *ifp)
1152 if_route(ifp, IFF_UP, AF_UNSPEC);
1156 * Flush an interface queue.
1159 if_qflush(struct ifaltq *ifq)
1165 if (ALTQ_IS_ENABLED(ifq))
1169 while ((m = n) != 0) {
1180 * Handle interface watchdog timer routines. Called
1181 * from softclock, we decrement timers (if set) and
1182 * call the appropriate interface routine on expiration.
1184 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1185 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab
1186 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1189 if_slowtimo(void *arg)
1195 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1196 if (ifp->if_timer == 0 || --ifp->if_timer)
1198 if (ifp->if_watchdog)
1199 (*ifp->if_watchdog)(ifp);
1203 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1207 * Map interface name to
1208 * interface structure pointer.
1211 ifunit(const char *name)
1216 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1217 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1225 * Hardware specific interface ioctls.
1228 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1234 size_t namelen, onamelen;
1235 char new_name[IFNAMSIZ];
1237 struct sockaddr_dl *sdl;
1239 ifr = (struct ifreq *)data;
1242 ifr->ifr_index = ifp->if_index;
1246 ifr->ifr_flags = ifp->if_flags & 0xffff;
1247 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1251 ifr->ifr_reqcap = ifp->if_capabilities;
1252 ifr->ifr_curcap = ifp->if_capenable;
1257 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
1262 ifr->ifr_metric = ifp->if_metric;
1266 ifr->ifr_mtu = ifp->if_mtu;
1270 ifr->ifr_phys = ifp->if_physical;
1277 new_flags = (ifr->ifr_flags & 0xffff) |
1278 (ifr->ifr_flagshigh << 16);
1279 if (ifp->if_flags & IFF_SMART) {
1280 /* Smart drivers twiddle their own routes */
1281 } else if (ifp->if_flags & IFF_UP &&
1282 (new_flags & IFF_UP) == 0) {
1286 } else if (new_flags & IFF_UP &&
1287 (ifp->if_flags & IFF_UP) == 0) {
1292 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1293 (new_flags &~ IFF_CANTCHANGE);
1294 if (new_flags & IFF_PPROMISC) {
1295 /* Permanently promiscuous mode requested */
1296 ifp->if_flags |= IFF_PROMISC;
1297 } else if (ifp->if_pcount == 0) {
1298 ifp->if_flags &= ~IFF_PROMISC;
1300 if (ifp->if_ioctl) {
1302 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1303 IFF_UNLOCKGIANT(ifp);
1305 getmicrotime(&ifp->if_lastchange);
1312 if (ifp->if_ioctl == NULL)
1313 return (EOPNOTSUPP);
1314 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1317 error = (*ifp->if_ioctl)(ifp, cmd, data);
1318 IFF_UNLOCKGIANT(ifp);
1320 getmicrotime(&ifp->if_lastchange);
1325 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
1333 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1336 if (new_name[0] == '\0')
1338 if (ifunit(new_name) != NULL)
1341 /* Announce the departure of the interface. */
1342 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1343 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1345 log(LOG_INFO, "%s: changing name to '%s'\n",
1346 ifp->if_xname, new_name);
1348 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1349 ifa = ifaddr_byindex(ifp->if_index);
1351 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1352 namelen = strlen(new_name);
1353 onamelen = sdl->sdl_nlen;
1355 * Move the address if needed. This is safe because we
1356 * allocate space for a name of length IFNAMSIZ when we
1357 * create this in if_attach().
1359 if (namelen != onamelen) {
1360 bcopy(sdl->sdl_data + onamelen,
1361 sdl->sdl_data + namelen, sdl->sdl_alen);
1363 bcopy(new_name, sdl->sdl_data, namelen);
1364 sdl->sdl_nlen = namelen;
1365 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1366 bzero(sdl->sdl_data, onamelen);
1367 while (namelen != 0)
1368 sdl->sdl_data[--namelen] = 0xff;
1371 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1372 /* Announce the return of the interface. */
1373 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1380 ifp->if_metric = ifr->ifr_metric;
1381 getmicrotime(&ifp->if_lastchange);
1388 if (ifp->if_ioctl == NULL)
1389 return (EOPNOTSUPP);
1391 error = (*ifp->if_ioctl)(ifp, cmd, data);
1392 IFF_UNLOCKGIANT(ifp);
1394 getmicrotime(&ifp->if_lastchange);
1399 u_long oldmtu = ifp->if_mtu;
1404 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1406 if (ifp->if_ioctl == NULL)
1407 return (EOPNOTSUPP);
1409 error = (*ifp->if_ioctl)(ifp, cmd, data);
1410 IFF_UNLOCKGIANT(ifp);
1412 getmicrotime(&ifp->if_lastchange);
1416 * If the link MTU changed, do network layer specific procedure.
1418 if (ifp->if_mtu != oldmtu) {
1432 /* Don't allow group membership on non-multicast interfaces. */
1433 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1434 return (EOPNOTSUPP);
1436 /* Don't let users screw up protocols' entries. */
1437 if (ifr->ifr_addr.sa_family != AF_LINK)
1440 if (cmd == SIOCADDMULTI) {
1441 struct ifmultiaddr *ifma;
1442 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1444 error = if_delmulti(ifp, &ifr->ifr_addr);
1447 getmicrotime(&ifp->if_lastchange);
1450 case SIOCSIFPHYADDR:
1451 case SIOCDIFPHYADDR:
1453 case SIOCSIFPHYADDR_IN6:
1455 case SIOCSLIFPHYADDR:
1457 case SIOCSIFGENERIC:
1461 if (ifp->if_ioctl == NULL)
1462 return (EOPNOTSUPP);
1464 error = (*ifp->if_ioctl)(ifp, cmd, data);
1465 IFF_UNLOCKGIANT(ifp);
1467 getmicrotime(&ifp->if_lastchange);
1471 ifs = (struct ifstat *)data;
1472 ifs->ascii[0] = '\0';
1474 case SIOCGIFPSRCADDR:
1475 case SIOCGIFPDSTADDR:
1476 case SIOCGLIFPHYADDR:
1478 case SIOCGIFGENERIC:
1479 if (ifp->if_ioctl == NULL)
1480 return (EOPNOTSUPP);
1482 error = (*ifp->if_ioctl)(ifp, cmd, data);
1483 IFF_UNLOCKGIANT(ifp);
1490 error = if_setlladdr(ifp,
1491 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1505 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1515 return (ifconf(cmd, data));
1517 ifr = (struct ifreq *)data;
1522 if ((error = suser(td)) != 0)
1524 return ((cmd == SIOCIFCREATE) ?
1525 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1526 if_clone_destroy(ifr->ifr_name));
1528 case SIOCIFGCLONERS:
1529 return (if_clone_list((struct if_clonereq *)data));
1532 ifp = ifunit(ifr->ifr_name);
1536 error = ifhwioctl(cmd, ifp, data, td);
1537 if (error != ENOIOCTL)
1540 oif_flags = ifp->if_flags;
1541 if (so->so_proto == 0)
1542 return (EOPNOTSUPP);
1544 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1553 case SIOCSIFDSTADDR:
1555 case SIOCSIFBRDADDR:
1556 case SIOCSIFNETMASK:
1557 #if BYTE_ORDER != BIG_ENDIAN
1558 if (ifr->ifr_addr.sa_family == 0 &&
1559 ifr->ifr_addr.sa_len < 16) {
1560 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1561 ifr->ifr_addr.sa_len = 16;
1564 if (ifr->ifr_addr.sa_len == 0)
1565 ifr->ifr_addr.sa_len = 16;
1573 case OSIOCGIFDSTADDR:
1574 cmd = SIOCGIFDSTADDR;
1577 case OSIOCGIFBRDADDR:
1578 cmd = SIOCGIFBRDADDR;
1581 case OSIOCGIFNETMASK:
1582 cmd = SIOCGIFNETMASK;
1584 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
1591 case OSIOCGIFDSTADDR:
1592 case OSIOCGIFBRDADDR:
1593 case OSIOCGIFNETMASK:
1594 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1598 #endif /* COMPAT_43 */
1600 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1602 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1603 if (ifp->if_flags & IFF_UP) {
1614 * The code common to hadling reference counted flags,
1615 * e.g., in ifpromisc() and if_allmulti().
1616 * The "pflag" argument can specify a permanent mode flag,
1617 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
1620 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
1624 int oldflags, oldcount;
1626 /* Sanity checks to catch programming errors */
1628 if (*refcount < 0) {
1630 "refusing to increment negative refcount %d "
1631 "for interface flag %d\n", *refcount, flag);
1635 if (*refcount <= 0) {
1637 "refusing to decrement non-positive refcount %d"
1638 "for interface flag %d\n", *refcount, flag);
1643 /* In case this mode is permanent, just touch refcount */
1644 if (ifp->if_flags & pflag) {
1645 *refcount += onswitch ? 1 : -1;
1649 /* Save ifnet parameters for if_ioctl() may fail */
1650 oldcount = *refcount;
1651 oldflags = ifp->if_flags;
1654 * See if we aren't the only and touching refcount is enough.
1655 * Actually toggle interface flag if we are the first or last.
1660 ifp->if_flags |= flag;
1664 ifp->if_flags &= ~flag;
1667 /* Call down the driver since we've changed interface flags */
1668 if (ifp->if_ioctl == NULL) {
1672 ifr.ifr_flags = ifp->if_flags & 0xffff;
1673 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1675 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1676 IFF_UNLOCKGIANT(ifp);
1679 /* Notify userland that interface flags have changed */
1684 /* Recover after driver error */
1685 *refcount = oldcount;
1686 ifp->if_flags = oldflags;
1691 * Set/clear promiscuous mode on interface ifp based on the truth value
1692 * of pswitch. The calls are reference counted so that only the first
1693 * "on" request actually has an effect, as does the final "off" request.
1694 * Results are undefined if the "off" and "on" requests are not matched.
1697 ifpromisc(struct ifnet *ifp, int pswitch)
1700 int oldflags = ifp->if_flags;
1702 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
1703 &ifp->if_pcount, pswitch);
1704 /* If promiscuous mode status has changed, log a message */
1705 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
1706 log(LOG_INFO, "%s: promiscuous mode %s\n",
1708 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
1713 * Return interface configuration
1714 * of system. List may be used
1715 * in later ioctl's (above) to get
1716 * other information.
1720 ifconf(u_long cmd, caddr_t data)
1722 struct ifconf *ifc = (struct ifconf *)data;
1727 int error, full = 0, valid_len, max_len;
1729 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
1730 max_len = MAXPHYS - 1;
1732 /* Prevent hostile input from being able to crash the system */
1733 if (ifc->ifc_len <= 0)
1737 if (ifc->ifc_len <= max_len) {
1738 max_len = ifc->ifc_len;
1741 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
1745 IFNET_RLOCK(); /* could sleep XXX */
1746 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1750 * Zero the ifr_name buffer to make sure we don't
1751 * disclose the contents of the stack.
1753 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
1755 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1756 >= sizeof(ifr.ifr_name))
1757 return (ENAMETOOLONG);
1760 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1761 struct sockaddr *sa = ifa->ifa_addr;
1763 if (jailed(curthread->td_ucred) &&
1764 prison_if(curthread->td_ucred, sa))
1768 if (cmd == OSIOCGIFCONF) {
1769 struct osockaddr *osa =
1770 (struct osockaddr *)&ifr.ifr_addr;
1772 osa->sa_family = sa->sa_family;
1773 sbuf_bcat(sb, &ifr, sizeof(ifr));
1774 max_len += sizeof(ifr);
1777 if (sa->sa_len <= sizeof(*sa)) {
1779 sbuf_bcat(sb, &ifr, sizeof(ifr));
1780 max_len += sizeof(ifr);
1783 offsetof(struct ifreq, ifr_addr));
1784 max_len += offsetof(struct ifreq, ifr_addr);
1785 sbuf_bcat(sb, sa, sa->sa_len);
1786 max_len += sa->sa_len;
1789 if (!sbuf_overflowed(sb))
1790 valid_len = sbuf_len(sb);
1793 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1794 sbuf_bcat(sb, &ifr, sizeof(ifr));
1795 max_len += sizeof(ifr);
1797 if (!sbuf_overflowed(sb))
1798 valid_len = sbuf_len(sb);
1804 * If we didn't allocate enough space (uncommon), try again. If
1805 * we have already allocated as much space as we are allowed,
1806 * return what we've got.
1808 if (valid_len != max_len && !full) {
1813 ifc->ifc_len = valid_len;
1815 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
1821 * Just like ifpromisc(), but for all-multicast-reception mode.
1824 if_allmulti(struct ifnet *ifp, int onswitch)
1827 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
1830 static struct ifmultiaddr *
1831 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
1833 struct ifmultiaddr *ifma;
1835 IF_ADDR_LOCK_ASSERT(ifp);
1837 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1838 if (sa_equal(ifma->ifma_addr, sa))
1846 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
1847 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
1848 * the ifnet multicast address list here, so the caller must do that and
1849 * other setup work (such as notifying the device driver). The reference
1850 * count is initialized to 1.
1852 static struct ifmultiaddr *
1853 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
1856 struct ifmultiaddr *ifma;
1857 struct sockaddr *dupsa;
1859 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
1864 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
1865 if (dupsa == NULL) {
1866 FREE(ifma, M_IFMADDR);
1869 bcopy(sa, dupsa, sa->sa_len);
1870 ifma->ifma_addr = dupsa;
1872 ifma->ifma_ifp = ifp;
1873 ifma->ifma_refcount = 1;
1874 ifma->ifma_protospec = NULL;
1877 ifma->ifma_lladdr = NULL;
1881 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
1882 if (dupsa == NULL) {
1883 FREE(ifma->ifma_addr, M_IFMADDR);
1884 FREE(ifma, M_IFMADDR);
1887 bcopy(llsa, dupsa, llsa->sa_len);
1888 ifma->ifma_lladdr = dupsa;
1894 * if_freemulti: free ifmultiaddr structure and possibly attached related
1895 * addresses. The caller is responsible for implementing reference
1896 * counting, notifying the driver, handling routing messages, and releasing
1897 * any dependent link layer state.
1900 if_freemulti(struct ifmultiaddr *ifma)
1903 KASSERT(ifma->ifma_refcount == 1, ("if_freemulti: refcount %d",
1904 ifma->ifma_refcount));
1905 KASSERT(ifma->ifma_protospec == NULL,
1906 ("if_freemulti: protospec not NULL"));
1908 if (ifma->ifma_lladdr != NULL)
1909 FREE(ifma->ifma_lladdr, M_IFMADDR);
1910 FREE(ifma->ifma_addr, M_IFMADDR);
1911 FREE(ifma, M_IFMADDR);
1915 * Register an additional multicast address with a network interface.
1917 * - If the address is already present, bump the reference count on the
1918 * address and return.
1919 * - If the address is not link-layer, look up a link layer address.
1920 * - Allocate address structures for one or both addresses, and attach to the
1921 * multicast address list on the interface. If automatically adding a link
1922 * layer address, the protocol address will own a reference to the link
1923 * layer address, to be freed when it is freed.
1924 * - Notify the network device driver of an addition to the multicast address
1927 * 'sa' points to caller-owned memory with the desired multicast address.
1929 * 'retifma' will be used to return a pointer to the resulting multicast
1930 * address reference, if desired.
1933 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
1934 struct ifmultiaddr **retifma)
1936 struct ifmultiaddr *ifma, *ll_ifma;
1937 struct sockaddr *llsa;
1941 * If the address is already present, return a new reference to it;
1942 * otherwise, allocate storage and set up a new address.
1945 ifma = if_findmulti(ifp, sa);
1947 ifma->ifma_refcount++;
1948 if (retifma != NULL)
1950 IF_ADDR_UNLOCK(ifp);
1955 * The address isn't already present; resolve the protocol address
1956 * into a link layer address, and then look that up, bump its
1957 * refcount or allocate an ifma for that also. If 'llsa' was
1958 * returned, we will need to free it later.
1962 if (ifp->if_resolvemulti != NULL) {
1963 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1969 * Allocate the new address. Don't hook it up yet, as we may also
1970 * need to allocate a link layer multicast address.
1972 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
1979 * If a link layer address is found, we'll need to see if it's
1980 * already present in the address list, or allocate is as well.
1981 * When this block finishes, the link layer address will be on the
1985 ll_ifma = if_findmulti(ifp, llsa);
1986 if (ll_ifma == NULL) {
1987 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
1988 if (ll_ifma == NULL) {
1993 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
1996 ll_ifma->ifma_refcount++;
2000 * We now have a new multicast address, ifma, and possibly a new or
2001 * referenced link layer address. Add the primary address to the
2002 * ifnet address list.
2004 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2006 if (retifma != NULL)
2010 * Must generate the message while holding the lock so that 'ifma'
2011 * pointer is still valid.
2013 * XXXRW: How come we don't announce ll_ifma?
2015 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2016 IF_ADDR_UNLOCK(ifp);
2019 * We are certain we have added something, so call down to the
2020 * interface to let them know about it.
2022 if (ifp->if_ioctl != NULL) {
2024 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2025 IFF_UNLOCKGIANT(ifp);
2029 FREE(llsa, M_IFMADDR);
2035 FREE(llsa, M_IFMADDR);
2038 IF_ADDR_UNLOCK(ifp);
2043 * Remove a reference to a multicast address on this interface. Yell
2044 * if the request does not match an existing membership.
2047 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2049 struct ifmultiaddr *ifma, *ll_ifma;
2052 ifma = if_findmulti(ifp, sa);
2054 IF_ADDR_UNLOCK(ifp);
2058 if (ifma->ifma_refcount > 1) {
2059 ifma->ifma_refcount--;
2060 IF_ADDR_UNLOCK(ifp);
2064 sa = ifma->ifma_lladdr;
2066 ll_ifma = if_findmulti(ifp, sa);
2071 * XXXRW: How come we don't announce ll_ifma?
2073 rt_newmaddrmsg(RTM_DELMADDR, ifma);
2075 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
2078 if (ll_ifma != NULL) {
2079 if (ll_ifma->ifma_refcount == 1) {
2080 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link);
2081 if_freemulti(ll_ifma);
2083 ll_ifma->ifma_refcount--;
2085 IF_ADDR_UNLOCK(ifp);
2088 * Make sure the interface driver is notified
2089 * in the case of a link layer mcast group being left.
2091 if (ifp->if_ioctl) {
2093 (void) (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2094 IFF_UNLOCKGIANT(ifp);
2101 * Set the link layer address on an interface.
2103 * At this time we only support certain types of interfaces,
2104 * and we don't allow the length of the address to change.
2107 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
2109 struct sockaddr_dl *sdl;
2113 ifa = ifaddr_byindex(ifp->if_index);
2116 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2119 if (len != sdl->sdl_alen) /* don't allow length to change */
2121 switch (ifp->if_type) {
2122 case IFT_ETHER: /* these types use struct arpcom */
2128 bcopy(lladdr, IFP2ENADDR(ifp), len);
2130 * XXX We also need to store the lladdr in LLADDR(sdl),
2131 * which is done below. This is a pain because we must
2132 * remember to keep the info in sync.
2136 bcopy(lladdr, LLADDR(sdl), len);
2142 * If the interface is already up, we need
2143 * to re-init it in order to reprogram its
2146 if ((ifp->if_flags & IFF_UP) != 0) {
2147 if (ifp->if_ioctl) {
2149 ifp->if_flags &= ~IFF_UP;
2150 ifr.ifr_flags = ifp->if_flags & 0xffff;
2151 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2152 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2153 ifp->if_flags |= IFF_UP;
2154 ifr.ifr_flags = ifp->if_flags & 0xffff;
2155 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2156 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2157 IFF_UNLOCKGIANT(ifp);
2161 * Also send gratuitous ARPs to notify other nodes about
2162 * the address change.
2164 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2165 if (ifa->ifa_addr != NULL &&
2166 ifa->ifa_addr->sa_family == AF_INET)
2167 arp_ifinit(ifp, ifa);
2175 * The name argument must be a pointer to storage which will last as
2176 * long as the interface does. For physical devices, the result of
2177 * device_get_name(dev) is a good choice and for pseudo-devices a
2178 * static string works well.
2181 if_initname(struct ifnet *ifp, const char *name, int unit)
2183 ifp->if_dname = name;
2184 ifp->if_dunit = unit;
2185 if (unit != IF_DUNIT_NONE)
2186 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
2188 strlcpy(ifp->if_xname, name, IFNAMSIZ);
2192 if_printf(struct ifnet *ifp, const char * fmt, ...)
2197 retval = printf("%s: ", ifp->if_xname);
2199 retval += vprintf(fmt, ap);
2205 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
2206 * be called without Giant. However, we often can't acquire the Giant lock
2207 * at those points; instead, we run it via a task queue that holds Giant via
2208 * if_start_deferred.
2210 * XXXRW: We need to make sure that the ifnet isn't fully detached until any
2211 * outstanding if_start_deferred() tasks that will run after the free. This
2212 * probably means waiting in if_detach().
2215 if_start(struct ifnet *ifp)
2220 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) {
2221 if (mtx_owned(&Giant))
2222 (*(ifp)->if_start)(ifp);
2224 taskqueue_enqueue(taskqueue_swi_giant,
2225 &ifp->if_starttask);
2227 (*(ifp)->if_start)(ifp);
2231 if_start_deferred(void *context, int pending)
2236 * This code must be entered with Giant, and should never run if
2237 * we're not running with debug.mpsafenet.
2239 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet"));
2243 (ifp->if_start)(ifp);
2247 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
2252 if (_IF_QFULL(ifq)) {
2259 ifp->if_obytes += m->m_pkthdr.len + adjust;
2260 if (m->m_flags & (M_BCAST|M_MCAST))
2262 active = ifp->if_flags & IFF_OACTIVE;
2264 _IF_ENQUEUE(ifq, m);
2266 if (ifp != NULL && !active)
2272 if_register_com_alloc(u_char type,
2273 if_com_alloc_t *a, if_com_free_t *f)
2276 KASSERT(if_com_alloc[type] == NULL,
2277 ("if_register_com_alloc: %d already registered", type));
2278 KASSERT(if_com_free[type] == NULL,
2279 ("if_register_com_alloc: %d free already registered", type));
2281 if_com_alloc[type] = a;
2282 if_com_free[type] = f;
2286 if_deregister_com_alloc(u_char type)
2289 KASSERT(if_com_alloc[type] == NULL,
2290 ("if_deregister_com_alloc: %d not registered", type));
2291 KASSERT(if_com_free[type] == NULL,
2292 ("if_deregister_com_alloc: %d free not registered", type));
2293 if_com_alloc[type] = NULL;
2294 if_com_free[type] = NULL;