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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)if.c 8.3 (Berkeley) 1/4/94
34 * $FreeBSD: src/sys/net/if.c,v 1.185 2004/03/13 02:35:03 brooks Exp $
35 * $DragonFly: src/sys/net/if.c,v 1.28 2005/02/11 22:25:57 joerg Exp $
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
42 #include <sys/param.h>
43 #include <sys/malloc.h>
45 #include <sys/systm.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.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/domain.h>
59 #include <net/if_arp.h>
60 #include <net/if_dl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/ifq_var.h>
64 #include <net/radix.h>
65 #include <net/route.h>
66 #include <machine/stdarg.h>
68 #if defined(INET) || defined(INET6)
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.h>
74 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
75 #include <netinet6/in6_var.h>
76 #include <netinet6/in6_ifattach.h>
80 #if defined(COMPAT_43)
81 #include <emulation/43bsd/43bsd_socket.h>
82 #endif /* COMPAT_43 */
85 * System initialization
88 static void if_attachdomain(void *);
89 static void if_attachdomain1(struct ifnet *);
90 static int ifconf (u_long, caddr_t, struct thread *);
91 static void ifinit (void *);
92 static void if_slowtimo (void *);
93 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
94 static int if_rtdel (struct radix_node *, void *);
96 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
98 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
99 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
100 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
102 int ifqmaxlen = IFQ_MAXLEN;
103 struct ifnethead ifnet; /* depend on static init XXX */
107 * XXX: declare here to avoid to include many inet6 related files..
108 * should be more generalized?
110 extern void nd6_setmtu (struct ifnet *);
113 struct if_clone *if_clone_lookup (const char *, int *);
114 int if_clone_list (struct if_clonereq *);
116 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
117 int if_cloners_count;
119 struct callout if_slowtimo_timer;
122 * Network interface utility routines.
124 * Routines with ifa_ifwith* names take sockaddr *'s as
134 callout_init(&if_slowtimo_timer);
137 TAILQ_FOREACH(ifp, &ifnet, if_link) {
138 if (ifp->if_snd.ifq_maxlen == 0) {
139 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
140 ifp->if_snd.ifq_maxlen = ifqmaxlen;
149 struct ifaddr **ifnet_addrs;
150 struct ifnet **ifindex2ifnet = NULL;
153 * Attach an interface to the
154 * list of "active" interfaces.
157 if_attach(struct ifnet *ifp)
159 unsigned socksize, ifasize;
160 int namelen, masklen;
161 struct sockaddr_dl *sdl;
164 static int if_indexlim = 8;
165 static boolean_t inited;
172 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
173 ifp->if_index = ++if_index;
176 * The old code would work if the interface passed a pre-existing
177 * chain of ifaddrs to this code. We don't trust our callers to
178 * properly initialize the tailq, however, so we no longer allow
179 * this unlikely case.
181 TAILQ_INIT(&ifp->if_addrhead);
182 TAILQ_INIT(&ifp->if_prefixhead);
183 LIST_INIT(&ifp->if_multiaddrs);
184 getmicrotime(&ifp->if_lastchange);
185 if (ifnet_addrs == NULL || if_index >= if_indexlim) {
190 n = if_indexlim * sizeof(struct ifaddr *);
191 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
192 if (ifnet_addrs != NULL) {
193 bcopy(ifnet_addrs, q, n/2);
194 free(ifnet_addrs, M_IFADDR);
196 ifnet_addrs = (struct ifaddr **)q;
198 /* grow ifindex2ifnet */
199 n = if_indexlim * sizeof(struct ifnet *);
200 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
202 bcopy(ifindex2ifnet, q, n/2);
203 free(ifindex2ifnet, M_IFADDR);
205 ifindex2ifnet = (struct ifnet **)q;
208 ifindex2ifnet[if_index] = ifp;
211 * create a Link Level name for this device
213 namelen = strlen(ifp->if_xname);
214 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
215 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
216 socksize = masklen + ifp->if_addrlen;
217 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
218 if (socksize < sizeof(*sdl))
219 socksize = sizeof(*sdl);
220 socksize = ROUNDUP(socksize);
221 ifasize = sizeof(struct ifaddr) + 2 * socksize;
222 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
223 sdl = (struct sockaddr_dl *)(ifa + 1);
224 sdl->sdl_len = socksize;
225 sdl->sdl_family = AF_LINK;
226 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
227 sdl->sdl_nlen = namelen;
228 sdl->sdl_index = ifp->if_index;
229 sdl->sdl_type = ifp->if_type;
230 ifnet_addrs[if_index - 1] = ifa;
232 ifa->ifa_rtrequest = link_rtrequest;
233 ifa->ifa_addr = (struct sockaddr *)sdl;
234 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
235 ifa->ifa_netmask = (struct sockaddr *)sdl;
236 sdl->sdl_len = masklen;
238 sdl->sdl_data[--namelen] = 0xff;
239 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
241 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
243 ifp->if_snd.altq_type = 0;
244 ifp->if_snd.altq_disc = NULL;
245 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
246 ifp->if_snd.altq_tbr = NULL;
247 ifp->if_snd.altq_ifp = ifp;
250 if_attachdomain1(ifp);
252 /* Announce the interface. */
253 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
257 if_attachdomain(void *dummy)
263 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
264 if_attachdomain1(ifp);
267 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
268 if_attachdomain, NULL);
271 if_attachdomain1(struct ifnet *ifp)
278 /* address family dependent data region */
279 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
280 for (dp = domains; dp; dp = dp->dom_next) {
281 if (dp->dom_ifattach)
282 ifp->if_afdata[dp->dom_family] =
283 (*dp->dom_ifattach)(ifp);
289 * Detach an interface, removing it from the
290 * list of "active" interfaces.
293 if_detach(struct ifnet *ifp)
296 struct radix_node_head *rnh;
301 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
304 * Remove routes and flush queues.
309 if (ifq_is_enabled(&ifp->if_snd))
310 altq_disable(&ifp->if_snd);
311 if (ifq_is_attached(&ifp->if_snd))
312 altq_detach(&ifp->if_snd);
315 * Remove address from ifnet_addrs[] and maybe decrement if_index.
316 * Clean up all addresses.
318 ifnet_addrs[ifp->if_index - 1] = 0;
319 while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
322 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
323 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
325 /* XXX: Ugly!! ad hoc just for INET */
326 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
327 struct ifaliasreq ifr;
329 bzero(&ifr, sizeof ifr);
330 ifr.ifra_addr = *ifa->ifa_addr;
331 if (ifa->ifa_dstaddr)
332 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
333 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
339 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
341 /* ifp_addrhead is already updated */
345 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
351 * Remove all IPv6 kernel structs related to ifp. This should be done
352 * before removing routing entries below, since IPv6 interface direct
353 * routes are expected to be removed by the IPv6-specific kernel API.
354 * Otherwise, the kernel will detect some inconsistency and bark it.
360 * Delete all remaining routes using this interface
361 * Unfortuneatly the only way to do this is to slog through
362 * the entire routing table looking for routes which point
363 * to this interface...oh well...
365 for (i = 1; i <= AF_MAX; i++) {
366 if ((rnh = rt_tables[i]) == NULL)
368 rnh->rnh_walktree(rnh, if_rtdel, ifp);
371 /* Announce that the interface is gone. */
372 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
374 for (dp = domains; dp; dp = dp->dom_next) {
375 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
376 (*dp->dom_ifdetach)(ifp,
377 ifp->if_afdata[dp->dom_family]);
380 ifindex2ifnet[ifp->if_index] = NULL;
382 TAILQ_REMOVE(&ifnet, ifp, if_link);
387 * Delete Routes for a Network Interface
389 * Called for each routing entry via the rnh->rnh_walktree() call above
390 * to delete all route entries referencing a detaching network interface.
393 * rn pointer to node in the routing table
394 * arg argument passed to rnh->rnh_walktree() - detaching interface
398 * errno failed - reason indicated
402 if_rtdel(struct radix_node *rn, void *arg)
404 struct rtentry *rt = (struct rtentry *)rn;
405 struct ifnet *ifp = arg;
408 if (rt->rt_ifp == ifp) {
411 * Protect (sorta) against walktree recursion problems
414 if (!(rt->rt_flags & RTF_UP))
417 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
418 rt_mask(rt), rt->rt_flags,
419 (struct rtentry **) NULL);
421 log(LOG_WARNING, "if_rtdel: error %d\n", err);
429 * Create a clone network interface.
432 if_clone_create(char *name, int len)
434 struct if_clone *ifc;
436 int wildcard, bytoff, bitoff;
440 ifc = if_clone_lookup(name, &unit);
444 if (ifunit(name) != NULL)
448 wildcard = (unit < 0);
450 * Find a free unit if none was given.
453 while (bytoff < ifc->ifc_bmlen &&
454 ifc->ifc_units[bytoff] == 0xff)
456 if (bytoff >= ifc->ifc_bmlen)
458 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
460 unit = (bytoff << 3) + bitoff;
463 if (unit > ifc->ifc_maxunit)
466 err = (*ifc->ifc_create)(ifc, unit);
472 bitoff = unit - (bytoff << 3);
476 * Allocate the unit in the bitmap.
478 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
479 ("%s: bit is already set", __func__));
480 ifc->ifc_units[bytoff] |= (1 << bitoff);
482 /* In the wildcard case, we need to update the name. */
484 for (dp = name; *dp != '\0'; dp++);
485 if (snprintf(dp, len - (dp-name), "%d", unit) >
486 len - (dp-name) - 1) {
488 * This can only be a programmer error and
489 * there's no straightforward way to recover if
492 panic("if_clone_create(): interface name too long");
497 EVENTHANDLER_INVOKE(if_clone_event, ifc);
503 * Destroy a clone network interface.
506 if_clone_destroy(const char *name)
508 struct if_clone *ifc;
513 ifc = if_clone_lookup(name, &unit);
517 if (unit < ifc->ifc_minifs)
524 if (ifc->ifc_destroy == NULL)
527 (*ifc->ifc_destroy)(ifp);
530 * Compute offset in the bitmap and deallocate the unit.
533 bitoff = unit - (bytoff << 3);
534 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
535 ("%s: bit is already cleared", __func__));
536 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
541 * Look up a network interface cloner.
544 if_clone_lookup(const char *name, int *unitp)
546 struct if_clone *ifc;
550 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
551 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
552 if (ifc->ifc_name[i] != *cp)
557 ifc = LIST_NEXT(ifc, ifc_list);
561 return ((struct if_clone *)NULL);
567 for (i = 0; *cp != '\0'; cp++) {
568 if (*cp < '0' || *cp > '9') {
569 /* Bogus unit number. */
572 i = (i * 10) + (*cp - '0');
582 * Register a network interface cloner.
585 if_clone_attach(struct if_clone *ifc)
592 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
593 ("%s: %s requested more units then allowed (%d > %d)",
594 __func__, ifc->ifc_name, ifc->ifc_minifs,
595 ifc->ifc_maxunit + 1));
597 * Compute bitmap size and allocate it.
599 maxclone = ifc->ifc_maxunit + 1;
601 if ((len << 3) < maxclone)
603 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
604 ifc->ifc_bmlen = len;
606 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
609 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
610 err = (*ifc->ifc_create)(ifc, unit);
612 ("%s: failed to create required interface %s%d",
613 __func__, ifc->ifc_name, unit));
615 /* Allocate the unit in the bitmap. */
617 bitoff = unit - (bytoff << 3);
618 ifc->ifc_units[bytoff] |= (1 << bitoff);
623 * Unregister a network interface cloner.
626 if_clone_detach(struct if_clone *ifc)
629 LIST_REMOVE(ifc, ifc_list);
630 free(ifc->ifc_units, M_CLONE);
635 * Provide list of interface cloners to userspace.
638 if_clone_list(struct if_clonereq *ifcr)
640 char outbuf[IFNAMSIZ], *dst;
641 struct if_clone *ifc;
642 int count, error = 0;
644 ifcr->ifcr_total = if_cloners_count;
645 if ((dst = ifcr->ifcr_buffer) == NULL) {
646 /* Just asking how many there are. */
650 if (ifcr->ifcr_count < 0)
653 count = (if_cloners_count < ifcr->ifcr_count) ?
654 if_cloners_count : ifcr->ifcr_count;
656 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
657 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
658 strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
659 error = copyout(outbuf, dst, IFNAMSIZ);
668 * Locate an interface based on a complete address.
671 ifa_ifwithaddr(struct sockaddr *addr)
676 TAILQ_FOREACH(ifp, &ifnet, if_link)
677 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
678 if (ifa->ifa_addr->sa_family != addr->sa_family)
680 if (sa_equal(addr, ifa->ifa_addr))
682 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
683 /* IPv6 doesn't have broadcast */
684 ifa->ifa_broadaddr->sa_len != 0 &&
685 sa_equal(ifa->ifa_broadaddr, addr))
688 return ((struct ifaddr *)NULL);
691 * Locate the point to point interface with a given destination address.
694 ifa_ifwithdstaddr(struct sockaddr *addr)
699 TAILQ_FOREACH(ifp, &ifnet, if_link)
700 if (ifp->if_flags & IFF_POINTOPOINT)
701 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
702 if (ifa->ifa_addr->sa_family != addr->sa_family)
704 if (ifa->ifa_dstaddr &&
705 sa_equal(addr, ifa->ifa_dstaddr))
708 return ((struct ifaddr *)NULL);
712 * Find an interface on a specific network. If many, choice
713 * is most specific found.
716 ifa_ifwithnet(struct sockaddr *addr)
720 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
721 u_int af = addr->sa_family;
722 char *addr_data = addr->sa_data, *cplim;
725 * AF_LINK addresses can be looked up directly by their index number,
726 * so do that if we can.
729 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
731 if (sdl->sdl_index && sdl->sdl_index <= if_index)
732 return (ifnet_addrs[sdl->sdl_index - 1]);
736 * Scan though each interface, looking for ones that have
737 * addresses in this address family.
739 TAILQ_FOREACH(ifp, &ifnet, if_link) {
740 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
741 char *cp, *cp2, *cp3;
743 if (ifa->ifa_addr->sa_family != af)
745 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
747 * This is a bit broken as it doesn't
748 * take into account that the remote end may
749 * be a single node in the network we are
751 * The trouble is that we don't know the
752 * netmask for the remote end.
754 if (ifa->ifa_dstaddr != NULL &&
755 sa_equal(addr, ifa->ifa_dstaddr))
759 * if we have a special address handler,
760 * then use it instead of the generic one.
762 if (ifa->ifa_claim_addr) {
763 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
771 * Scan all the bits in the ifa's address.
772 * If a bit dissagrees with what we are
773 * looking for, mask it with the netmask
774 * to see if it really matters.
777 if (ifa->ifa_netmask == 0)
780 cp2 = ifa->ifa_addr->sa_data;
781 cp3 = ifa->ifa_netmask->sa_data;
782 cplim = ifa->ifa_netmask->sa_len +
783 (char *)ifa->ifa_netmask;
785 if ((*cp++ ^ *cp2++) & *cp3++)
786 goto next; /* next address! */
788 * If the netmask of what we just found
789 * is more specific than what we had before
790 * (if we had one) then remember the new one
791 * before continuing to search
792 * for an even better one.
794 if (ifa_maybe == 0 ||
795 rn_refines((char *)ifa->ifa_netmask,
796 (char *)ifa_maybe->ifa_netmask))
805 * Find an interface address specific to an interface best matching
809 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
812 char *cp, *cp2, *cp3;
814 struct ifaddr *ifa_maybe = 0;
815 u_int af = addr->sa_family;
819 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
820 if (ifa->ifa_addr->sa_family != af)
824 if (ifa->ifa_netmask == NULL) {
825 if (sa_equal(addr, ifa->ifa_addr) ||
826 (ifa->ifa_dstaddr != NULL &&
827 sa_equal(addr, ifa->ifa_dstaddr)))
831 if (ifp->if_flags & IFF_POINTOPOINT) {
832 if (sa_equal(addr, ifa->ifa_dstaddr))
836 cp2 = ifa->ifa_addr->sa_data;
837 cp3 = ifa->ifa_netmask->sa_data;
838 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
839 for (; cp3 < cplim; cp3++)
840 if ((*cp++ ^ *cp2++) & *cp3)
849 #include <net/route.h>
852 * Default action when installing a route with a Link Level gateway.
853 * Lookup an appropriate real ifa to point to.
854 * This should be moved to /sys/net/link.c eventually.
857 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
860 struct sockaddr *dst;
863 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
864 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
866 ifa = ifaof_ifpforaddr(dst, ifp);
871 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
872 ifa->ifa_rtrequest(cmd, rt, info);
877 * Mark an interface down and notify protocols of
879 * NOTE: must be called at splnet or eqivalent.
882 if_unroute(struct ifnet *ifp, int flag, int fam)
886 ifp->if_flags &= ~flag;
887 getmicrotime(&ifp->if_lastchange);
888 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
889 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
890 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
891 ifq_purge(&ifp->if_snd);
896 * Mark an interface up and notify protocols of
898 * NOTE: must be called at splnet or eqivalent.
901 if_route(struct ifnet *ifp, int flag, int fam)
905 ifp->if_flags |= flag;
906 getmicrotime(&ifp->if_lastchange);
907 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
908 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
909 pfctlinput(PRC_IFUP, ifa->ifa_addr);
917 * Mark an interface down and notify protocols of the transition. An
918 * interface going down is also considered to be a synchronizing event.
919 * We must ensure that all packet processing related to the interface
920 * has completed before we return so e.g. the caller can free the ifnet
921 * structure that the mbufs may be referencing.
923 * NOTE: must be called at splnet or eqivalent.
926 if_down(struct ifnet *ifp)
929 if_unroute(ifp, IFF_UP, AF_UNSPEC);
930 netmsg_service_sync();
934 * Mark an interface up and notify protocols of
936 * NOTE: must be called at splnet or eqivalent.
939 if_up(struct ifnet *ifp)
942 if_route(ifp, IFF_UP, AF_UNSPEC);
946 * Handle interface watchdog timer routines. Called
947 * from softclock, we decrement timers (if set) and
948 * call the appropriate interface routine on expiration.
951 if_slowtimo(void *arg)
956 TAILQ_FOREACH(ifp, &ifnet, if_link) {
957 if (ifp->if_timer == 0 || --ifp->if_timer)
959 if (ifp->if_watchdog)
960 (*ifp->if_watchdog)(ifp);
963 callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
967 * Map interface name to
968 * interface structure pointer.
971 ifunit(const char *name)
976 * Search all the interfaces for this name/number
979 TAILQ_FOREACH(ifp, &ifnet, if_link) {
980 if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
988 * Map interface name in a sockaddr_dl to
989 * interface structure pointer.
992 if_withname(struct sockaddr *sa)
994 char ifname[IFNAMSIZ+1];
995 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
997 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
998 (sdl->sdl_nlen > IFNAMSIZ) )
1002 * ifunit wants a null-terminated name. It may not be null-terminated
1003 * in the sockaddr. We don't want to change the caller's sockaddr,
1004 * and there might not be room to put the trailing null anyway, so we
1005 * make a local copy that we know we can null terminate safely.
1008 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1009 ifname[sdl->sdl_nlen] = '\0';
1010 return ifunit(ifname);
1018 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1026 size_t namelen, onamelen;
1027 char new_name[IFNAMSIZ];
1029 struct sockaddr_dl *sdl;
1035 return (ifconf(cmd, data, td));
1037 ifr = (struct ifreq *)data;
1042 if ((error = suser(td)) != 0)
1044 return ((cmd == SIOCIFCREATE) ?
1045 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1046 if_clone_destroy(ifr->ifr_name));
1048 case SIOCIFGCLONERS:
1049 return (if_clone_list((struct if_clonereq *)data));
1052 ifp = ifunit(ifr->ifr_name);
1058 ifr->ifr_flags = ifp->if_flags;
1059 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1063 ifr->ifr_reqcap = ifp->if_capabilities;
1064 ifr->ifr_curcap = ifp->if_capenable;
1068 ifr->ifr_metric = ifp->if_metric;
1072 ifr->ifr_mtu = ifp->if_mtu;
1076 ifr->ifr_phys = ifp->if_physical;
1083 new_flags = (ifr->ifr_flags & 0xffff) |
1084 (ifr->ifr_flagshigh << 16);
1085 if (ifp->if_flags & IFF_SMART) {
1086 /* Smart drivers twiddle their own routes */
1087 } else if (ifp->if_flags & IFF_UP &&
1088 (new_flags & IFF_UP) == 0) {
1092 } else if (new_flags & IFF_UP &&
1093 (ifp->if_flags & IFF_UP) == 0) {
1098 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1099 (new_flags &~ IFF_CANTCHANGE);
1100 if (new_flags & IFF_PPROMISC) {
1101 /* Permanently promiscuous mode requested */
1102 ifp->if_flags |= IFF_PROMISC;
1103 } else if (ifp->if_pcount == 0) {
1104 ifp->if_flags &= ~IFF_PROMISC;
1107 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1108 getmicrotime(&ifp->if_lastchange);
1115 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1117 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1124 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1127 if (new_name[0] == '\0')
1129 if (ifunit(new_name) != NULL)
1132 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1134 /* Announce the departure of the interface. */
1135 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1137 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1138 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1139 /* XXX IFA_LOCK(ifa); */
1140 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1141 namelen = strlen(new_name);
1142 onamelen = sdl->sdl_nlen;
1144 * Move the address if needed. This is safe because we
1145 * allocate space for a name of length IFNAMSIZ when we
1146 * create this in if_attach().
1148 if (namelen != onamelen) {
1149 bcopy(sdl->sdl_data + onamelen,
1150 sdl->sdl_data + namelen, sdl->sdl_alen);
1152 bcopy(new_name, sdl->sdl_data, namelen);
1153 sdl->sdl_nlen = namelen;
1154 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1155 bzero(sdl->sdl_data, onamelen);
1156 while (namelen != 0)
1157 sdl->sdl_data[--namelen] = 0xff;
1158 /* XXX IFA_UNLOCK(ifa) */
1160 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1162 /* Announce the return of the interface. */
1163 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1170 ifp->if_metric = ifr->ifr_metric;
1171 getmicrotime(&ifp->if_lastchange);
1180 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1182 getmicrotime(&ifp->if_lastchange);
1187 u_long oldmtu = ifp->if_mtu;
1192 if (ifp->if_ioctl == NULL)
1193 return (EOPNOTSUPP);
1194 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1196 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1198 getmicrotime(&ifp->if_lastchange);
1202 * If the link MTU changed, do network layer specific procedure.
1204 if (ifp->if_mtu != oldmtu) {
1218 /* Don't allow group membership on non-multicast interfaces. */
1219 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1222 /* Don't let users screw up protocols' entries. */
1223 if (ifr->ifr_addr.sa_family != AF_LINK)
1226 if (cmd == SIOCADDMULTI) {
1227 struct ifmultiaddr *ifma;
1228 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1230 error = if_delmulti(ifp, &ifr->ifr_addr);
1233 getmicrotime(&ifp->if_lastchange);
1236 case SIOCSIFPHYADDR:
1237 case SIOCDIFPHYADDR:
1239 case SIOCSIFPHYADDR_IN6:
1241 case SIOCSLIFPHYADDR:
1243 case SIOCSIFGENERIC:
1247 if (ifp->if_ioctl == 0)
1248 return (EOPNOTSUPP);
1249 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1251 getmicrotime(&ifp->if_lastchange);
1255 ifs = (struct ifstat *)data;
1256 ifs->ascii[0] = '\0';
1258 case SIOCGIFPSRCADDR:
1259 case SIOCGIFPDSTADDR:
1260 case SIOCGLIFPHYADDR:
1262 case SIOCGIFGENERIC:
1263 if (ifp->if_ioctl == 0)
1264 return (EOPNOTSUPP);
1265 return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1271 return if_setlladdr(ifp,
1272 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1275 oif_flags = ifp->if_flags;
1276 if (so->so_proto == 0)
1277 return (EOPNOTSUPP);
1279 error = so_pru_control(so, cmd, data, ifp, td);
1286 case SIOCSIFDSTADDR:
1288 case SIOCSIFBRDADDR:
1289 case SIOCSIFNETMASK:
1290 #if BYTE_ORDER != BIG_ENDIAN
1291 if (ifr->ifr_addr.sa_family == 0 &&
1292 ifr->ifr_addr.sa_len < 16) {
1293 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1294 ifr->ifr_addr.sa_len = 16;
1297 if (ifr->ifr_addr.sa_len == 0)
1298 ifr->ifr_addr.sa_len = 16;
1306 case OSIOCGIFDSTADDR:
1307 cmd = SIOCGIFDSTADDR;
1310 case OSIOCGIFBRDADDR:
1311 cmd = SIOCGIFBRDADDR;
1314 case OSIOCGIFNETMASK:
1315 cmd = SIOCGIFNETMASK;
1317 error = so_pru_control(so, cmd, data, ifp, td);
1321 case OSIOCGIFDSTADDR:
1322 case OSIOCGIFBRDADDR:
1323 case OSIOCGIFNETMASK:
1324 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1328 #endif /* COMPAT_43 */
1330 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1332 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1333 if (ifp->if_flags & IFF_UP) {
1347 * Set/clear promiscuous mode on interface ifp based on the truth value
1348 * of pswitch. The calls are reference counted so that only the first
1349 * "on" request actually has an effect, as does the final "off" request.
1350 * Results are undefined if the "off" and "on" requests are not matched.
1353 ifpromisc(struct ifnet *ifp, int pswitch)
1359 oldflags = ifp->if_flags;
1360 if (ifp->if_flags & IFF_PPROMISC) {
1361 /* Do nothing if device is in permanently promiscuous mode */
1362 ifp->if_pcount += pswitch ? 1 : -1;
1367 * If the device is not configured up, we cannot put it in
1370 if ((ifp->if_flags & IFF_UP) == 0)
1372 if (ifp->if_pcount++ != 0)
1374 ifp->if_flags |= IFF_PROMISC;
1375 log(LOG_INFO, "%s: promiscuous mode enabled\n",
1378 if (--ifp->if_pcount > 0)
1380 ifp->if_flags &= ~IFF_PROMISC;
1381 log(LOG_INFO, "%s: promiscuous mode disabled\n",
1384 ifr.ifr_flags = ifp->if_flags;
1385 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1386 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1387 (struct ucred *)NULL);
1391 ifp->if_flags = oldflags;
1396 * Return interface configuration
1397 * of system. List may be used
1398 * in later ioctl's (above) to get
1399 * other information.
1402 ifconf(u_long cmd, caddr_t data, struct thread *td)
1404 struct ifconf *ifc = (struct ifconf *)data;
1407 struct sockaddr *sa;
1408 struct ifreq ifr, *ifrp;
1409 int space = ifc->ifc_len, error = 0;
1411 ifrp = ifc->ifc_req;
1412 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1415 if (space <= sizeof ifr)
1417 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1418 >= sizeof(ifr.ifr_name)) {
1419 error = ENAMETOOLONG;
1424 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1425 if (space <= sizeof ifr)
1428 if (td->td_proc->p_ucred->cr_prison &&
1433 if (cmd == OSIOCGIFCONF) {
1434 struct osockaddr *osa =
1435 (struct osockaddr *)&ifr.ifr_addr;
1437 osa->sa_family = sa->sa_family;
1438 error = copyout(&ifr, ifrp, sizeof ifr);
1442 if (sa->sa_len <= sizeof(*sa)) {
1444 error = copyout(&ifr, ifrp, sizeof ifr);
1447 if (space < (sizeof ifr) + sa->sa_len -
1450 space -= sa->sa_len - sizeof(*sa);
1451 error = copyout(&ifr, ifrp,
1452 sizeof ifr.ifr_name);
1454 error = copyout(sa, &ifrp->ifr_addr,
1456 ifrp = (struct ifreq *)
1457 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1461 space -= sizeof ifr;
1466 bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1467 error = copyout(&ifr, ifrp, sizeof ifr);
1470 space -= sizeof ifr;
1474 ifc->ifc_len -= space;
1479 * Just like if_promisc(), but for all-multicast-reception mode.
1482 if_allmulti(struct ifnet *ifp, int onswitch)
1489 if (ifp->if_amcount++ == 0) {
1490 ifp->if_flags |= IFF_ALLMULTI;
1491 ifr.ifr_flags = ifp->if_flags;
1492 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1493 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1494 (struct ucred *)NULL);
1497 if (ifp->if_amcount > 1) {
1500 ifp->if_amcount = 0;
1501 ifp->if_flags &= ~IFF_ALLMULTI;
1502 ifr.ifr_flags = ifp->if_flags;
1503 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1504 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1505 (struct ucred *)NULL);
1516 * Add a multicast listenership to the interface in question.
1517 * The link layer provides a routine which converts
1521 struct ifnet *ifp, /* interface to manipulate */
1522 struct sockaddr *sa, /* address to add */
1523 struct ifmultiaddr **retifma)
1525 struct sockaddr *llsa, *dupsa;
1527 struct ifmultiaddr *ifma;
1530 * If the matching multicast address already exists
1531 * then don't add a new one, just add a reference
1533 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1534 if (sa_equal(sa, ifma->ifma_addr)) {
1535 ifma->ifma_refcount++;
1543 * Give the link layer a chance to accept/reject it, and also
1544 * find out which AF_LINK address this maps to, if it isn't one
1547 if (ifp->if_resolvemulti) {
1548 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1549 if (error) return error;
1554 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1555 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1556 bcopy(sa, dupsa, sa->sa_len);
1558 ifma->ifma_addr = dupsa;
1559 ifma->ifma_lladdr = llsa;
1560 ifma->ifma_ifp = ifp;
1561 ifma->ifma_refcount = 1;
1562 ifma->ifma_protospec = 0;
1563 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1566 * Some network interfaces can scan the address list at
1567 * interrupt time; lock them out.
1570 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1575 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1576 if (sa_equal(ifma->ifma_addr, llsa))
1580 ifma->ifma_refcount++;
1582 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1583 M_IFMADDR, M_WAITOK);
1584 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1585 M_IFMADDR, M_WAITOK);
1586 bcopy(llsa, dupsa, llsa->sa_len);
1587 ifma->ifma_addr = dupsa;
1588 ifma->ifma_ifp = ifp;
1589 ifma->ifma_refcount = 1;
1591 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1596 * We are certain we have added something, so call down to the
1597 * interface to let them know about it.
1600 ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1607 * Remove a reference to a multicast address on this interface. Yell
1608 * if the request does not match an existing membership.
1611 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1613 struct ifmultiaddr *ifma;
1616 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1617 if (sa_equal(sa, ifma->ifma_addr))
1622 if (ifma->ifma_refcount > 1) {
1623 ifma->ifma_refcount--;
1627 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1628 sa = ifma->ifma_lladdr;
1630 LIST_REMOVE(ifma, ifma_link);
1632 * Make sure the interface driver is notified
1633 * in the case of a link layer mcast group being left.
1635 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1636 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1638 free(ifma->ifma_addr, M_IFMADDR);
1639 free(ifma, M_IFMADDR);
1644 * Now look for the link-layer address which corresponds to
1645 * this network address. It had been squirreled away in
1646 * ifma->ifma_lladdr for this purpose (so we don't have
1647 * to call ifp->if_resolvemulti() again), and we saved that
1648 * value in sa above. If some nasty deleted the
1649 * link-layer address out from underneath us, we can deal because
1650 * the address we stored was is not the same as the one which was
1651 * in the record for the link-layer address. (So we don't complain
1654 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1655 if (sa_equal(sa, ifma->ifma_addr))
1660 if (ifma->ifma_refcount > 1) {
1661 ifma->ifma_refcount--;
1666 LIST_REMOVE(ifma, ifma_link);
1667 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1669 free(ifma->ifma_addr, M_IFMADDR);
1670 free(sa, M_IFMADDR);
1671 free(ifma, M_IFMADDR);
1677 * Set the link layer address on an interface.
1679 * At this time we only support certain types of interfaces,
1680 * and we don't allow the length of the address to change.
1683 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1685 struct sockaddr_dl *sdl;
1689 ifa = ifnet_addrs[ifp->if_index - 1];
1692 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1695 if (len != sdl->sdl_alen) /* don't allow length to change */
1697 switch (ifp->if_type) {
1698 case IFT_ETHER: /* these types use struct arpcom */
1703 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1706 bcopy(lladdr, LLADDR(sdl), len);
1712 * If the interface is already up, we need
1713 * to re-init it in order to reprogram its
1716 if ((ifp->if_flags & IFF_UP) != 0) {
1717 ifp->if_flags &= ~IFF_UP;
1718 ifr.ifr_flags = ifp->if_flags;
1719 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1720 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1721 (struct ucred *)NULL);
1722 ifp->if_flags |= IFF_UP;
1723 ifr.ifr_flags = ifp->if_flags;
1724 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1725 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1726 (struct ucred *)NULL);
1729 * Also send gratuitous ARPs to notify other nodes about
1730 * the address change.
1732 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1733 if (ifa->ifa_addr != NULL &&
1734 ifa->ifa_addr->sa_family == AF_INET)
1735 arp_ifinit(ifp, ifa);
1742 struct ifmultiaddr *
1743 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1745 struct ifmultiaddr *ifma;
1747 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1748 if (sa_equal(ifma->ifma_addr, sa))
1755 * The name argument must be a pointer to storage which will last as
1756 * long as the interface does. For physical devices, the result of
1757 * device_get_name(dev) is a good choice and for pseudo-devices a
1758 * static string works well.
1761 if_initname(struct ifnet *ifp, const char *name, int unit)
1763 ifp->if_dname = name;
1764 ifp->if_dunit = unit;
1765 if (unit != IF_DUNIT_NONE)
1766 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1768 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1772 if_printf(struct ifnet *ifp, const char *fmt, ...)
1777 retval = printf("%s: ", ifp->if_xname);
1778 __va_start(ap, fmt);
1779 retval += vprintf(fmt, ap);
1784 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1785 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");