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 1.30 2005/03/12 02:36:04 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 * Support for non-ALTQ interfaces.
87 static int ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
88 struct altq_pktattr *);
90 ifq_classic_dequeue(struct ifaltq *, int);
91 static int ifq_classic_request(struct ifaltq *, int, void *);
94 * System initialization
97 static void if_attachdomain(void *);
98 static void if_attachdomain1(struct ifnet *);
99 static int ifconf (u_long, caddr_t, struct thread *);
100 static void ifinit (void *);
101 static void if_slowtimo (void *);
102 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
103 static int if_rtdel (struct radix_node *, void *);
105 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
107 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
108 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
109 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
111 int ifqmaxlen = IFQ_MAXLEN;
112 struct ifnethead ifnet; /* depend on static init XXX */
116 * XXX: declare here to avoid to include many inet6 related files..
117 * should be more generalized?
119 extern void nd6_setmtu (struct ifnet *);
122 struct if_clone *if_clone_lookup (const char *, int *);
123 int if_clone_list (struct if_clonereq *);
125 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
126 int if_cloners_count;
128 struct callout if_slowtimo_timer;
131 * Network interface utility routines.
133 * Routines with ifa_ifwith* names take sockaddr *'s as
143 callout_init(&if_slowtimo_timer);
146 TAILQ_FOREACH(ifp, &ifnet, if_link) {
147 if (ifp->if_snd.ifq_maxlen == 0) {
148 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
149 ifp->if_snd.ifq_maxlen = ifqmaxlen;
158 struct ifaddr **ifnet_addrs;
159 struct ifnet **ifindex2ifnet = NULL;
162 * Attach an interface to the
163 * list of "active" interfaces.
166 if_attach(struct ifnet *ifp)
168 unsigned socksize, ifasize;
169 int namelen, masklen;
170 struct sockaddr_dl *sdl;
174 static int if_indexlim = 8;
175 static boolean_t inited;
182 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
183 ifp->if_index = ++if_index;
186 * The old code would work if the interface passed a pre-existing
187 * chain of ifaddrs to this code. We don't trust our callers to
188 * properly initialize the tailq, however, so we no longer allow
189 * this unlikely case.
191 TAILQ_INIT(&ifp->if_addrhead);
192 TAILQ_INIT(&ifp->if_prefixhead);
193 LIST_INIT(&ifp->if_multiaddrs);
194 getmicrotime(&ifp->if_lastchange);
195 if (ifnet_addrs == NULL || if_index >= if_indexlim) {
200 n = if_indexlim * sizeof(struct ifaddr *);
201 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
202 if (ifnet_addrs != NULL) {
203 bcopy(ifnet_addrs, q, n/2);
204 free(ifnet_addrs, M_IFADDR);
206 ifnet_addrs = (struct ifaddr **)q;
208 /* grow ifindex2ifnet */
209 n = if_indexlim * sizeof(struct ifnet *);
210 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
212 bcopy(ifindex2ifnet, q, n/2);
213 free(ifindex2ifnet, M_IFADDR);
215 ifindex2ifnet = (struct ifnet **)q;
218 ifindex2ifnet[if_index] = ifp;
221 * create a Link Level name for this device
223 namelen = strlen(ifp->if_xname);
224 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
225 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
226 socksize = masklen + ifp->if_addrlen;
227 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
228 if (socksize < sizeof(*sdl))
229 socksize = sizeof(*sdl);
230 socksize = ROUNDUP(socksize);
231 ifasize = sizeof(struct ifaddr) + 2 * socksize;
232 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
233 sdl = (struct sockaddr_dl *)(ifa + 1);
234 sdl->sdl_len = socksize;
235 sdl->sdl_family = AF_LINK;
236 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
237 sdl->sdl_nlen = namelen;
238 sdl->sdl_index = ifp->if_index;
239 sdl->sdl_type = ifp->if_type;
240 ifnet_addrs[if_index - 1] = ifa;
242 ifa->ifa_rtrequest = link_rtrequest;
243 ifa->ifa_addr = (struct sockaddr *)sdl;
244 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
245 ifa->ifa_netmask = (struct sockaddr *)sdl;
246 sdl->sdl_len = masklen;
248 sdl->sdl_data[--namelen] = 0xff;
249 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
251 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
255 ifq->altq_disc = NULL;
256 ifq->altq_flags &= ALTQF_CANTCHANGE;
257 ifq->altq_tbr = NULL;
259 ifq->altq_enqueue = ifq_classic_enqueue;
260 ifq->altq_dequeue = ifq_classic_dequeue;
261 ifq->altq_request = ifq_classic_request;
263 if (!SLIST_EMPTY(&domains))
264 if_attachdomain1(ifp);
266 /* Announce the interface. */
267 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
271 if_attachdomain(void *dummy)
277 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
278 if_attachdomain1(ifp);
281 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
282 if_attachdomain, NULL);
285 if_attachdomain1(struct ifnet *ifp)
292 /* address family dependent data region */
293 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
294 SLIST_FOREACH(dp, &domains, dom_next)
295 if (dp->dom_ifattach)
296 ifp->if_afdata[dp->dom_family] =
297 (*dp->dom_ifattach)(ifp);
302 * Detach an interface, removing it from the
303 * list of "active" interfaces.
306 if_detach(struct ifnet *ifp)
309 struct radix_node_head *rnh;
314 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
317 * Remove routes and flush queues.
322 if (ifq_is_enabled(&ifp->if_snd))
323 altq_disable(&ifp->if_snd);
324 if (ifq_is_attached(&ifp->if_snd))
325 altq_detach(&ifp->if_snd);
328 * Remove address from ifnet_addrs[] and maybe decrement if_index.
329 * Clean up all addresses.
331 ifnet_addrs[ifp->if_index - 1] = 0;
332 while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
335 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
336 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
338 /* XXX: Ugly!! ad hoc just for INET */
339 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
340 struct ifaliasreq ifr;
342 bzero(&ifr, sizeof ifr);
343 ifr.ifra_addr = *ifa->ifa_addr;
344 if (ifa->ifa_dstaddr)
345 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
346 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
352 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
354 /* ifp_addrhead is already updated */
358 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
364 * Remove all IPv6 kernel structs related to ifp. This should be done
365 * before removing routing entries below, since IPv6 interface direct
366 * routes are expected to be removed by the IPv6-specific kernel API.
367 * Otherwise, the kernel will detect some inconsistency and bark it.
373 * Delete all remaining routes using this interface
374 * Unfortuneatly the only way to do this is to slog through
375 * the entire routing table looking for routes which point
376 * to this interface...oh well...
378 for (i = 1; i <= AF_MAX; i++) {
379 if ((rnh = rt_tables[i]) == NULL)
381 rnh->rnh_walktree(rnh, if_rtdel, ifp);
384 /* Announce that the interface is gone. */
385 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
387 SLIST_FOREACH(dp, &domains, dom_next)
388 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
389 (*dp->dom_ifdetach)(ifp,
390 ifp->if_afdata[dp->dom_family]);
392 ifindex2ifnet[ifp->if_index] = NULL;
394 TAILQ_REMOVE(&ifnet, ifp, if_link);
399 * Delete Routes for a Network Interface
401 * Called for each routing entry via the rnh->rnh_walktree() call above
402 * to delete all route entries referencing a detaching network interface.
405 * rn pointer to node in the routing table
406 * arg argument passed to rnh->rnh_walktree() - detaching interface
410 * errno failed - reason indicated
414 if_rtdel(struct radix_node *rn, void *arg)
416 struct rtentry *rt = (struct rtentry *)rn;
417 struct ifnet *ifp = arg;
420 if (rt->rt_ifp == ifp) {
423 * Protect (sorta) against walktree recursion problems
426 if (!(rt->rt_flags & RTF_UP))
429 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
430 rt_mask(rt), rt->rt_flags,
431 (struct rtentry **) NULL);
433 log(LOG_WARNING, "if_rtdel: error %d\n", err);
441 * Create a clone network interface.
444 if_clone_create(char *name, int len)
446 struct if_clone *ifc;
448 int wildcard, bytoff, bitoff;
452 ifc = if_clone_lookup(name, &unit);
456 if (ifunit(name) != NULL)
460 wildcard = (unit < 0);
462 * Find a free unit if none was given.
465 while (bytoff < ifc->ifc_bmlen &&
466 ifc->ifc_units[bytoff] == 0xff)
468 if (bytoff >= ifc->ifc_bmlen)
470 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
472 unit = (bytoff << 3) + bitoff;
475 if (unit > ifc->ifc_maxunit)
478 err = (*ifc->ifc_create)(ifc, unit);
484 bitoff = unit - (bytoff << 3);
488 * Allocate the unit in the bitmap.
490 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
491 ("%s: bit is already set", __func__));
492 ifc->ifc_units[bytoff] |= (1 << bitoff);
494 /* In the wildcard case, we need to update the name. */
496 for (dp = name; *dp != '\0'; dp++);
497 if (snprintf(dp, len - (dp-name), "%d", unit) >
498 len - (dp-name) - 1) {
500 * This can only be a programmer error and
501 * there's no straightforward way to recover if
504 panic("if_clone_create(): interface name too long");
509 EVENTHANDLER_INVOKE(if_clone_event, ifc);
515 * Destroy a clone network interface.
518 if_clone_destroy(const char *name)
520 struct if_clone *ifc;
525 ifc = if_clone_lookup(name, &unit);
529 if (unit < ifc->ifc_minifs)
536 if (ifc->ifc_destroy == NULL)
539 (*ifc->ifc_destroy)(ifp);
542 * Compute offset in the bitmap and deallocate the unit.
545 bitoff = unit - (bytoff << 3);
546 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
547 ("%s: bit is already cleared", __func__));
548 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
553 * Look up a network interface cloner.
556 if_clone_lookup(const char *name, int *unitp)
558 struct if_clone *ifc;
562 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
563 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
564 if (ifc->ifc_name[i] != *cp)
569 ifc = LIST_NEXT(ifc, ifc_list);
573 return ((struct if_clone *)NULL);
579 for (i = 0; *cp != '\0'; cp++) {
580 if (*cp < '0' || *cp > '9') {
581 /* Bogus unit number. */
584 i = (i * 10) + (*cp - '0');
594 * Register a network interface cloner.
597 if_clone_attach(struct if_clone *ifc)
604 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
605 ("%s: %s requested more units then allowed (%d > %d)",
606 __func__, ifc->ifc_name, ifc->ifc_minifs,
607 ifc->ifc_maxunit + 1));
609 * Compute bitmap size and allocate it.
611 maxclone = ifc->ifc_maxunit + 1;
613 if ((len << 3) < maxclone)
615 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
616 ifc->ifc_bmlen = len;
618 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
621 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
622 err = (*ifc->ifc_create)(ifc, unit);
624 ("%s: failed to create required interface %s%d",
625 __func__, ifc->ifc_name, unit));
627 /* Allocate the unit in the bitmap. */
629 bitoff = unit - (bytoff << 3);
630 ifc->ifc_units[bytoff] |= (1 << bitoff);
635 * Unregister a network interface cloner.
638 if_clone_detach(struct if_clone *ifc)
641 LIST_REMOVE(ifc, ifc_list);
642 free(ifc->ifc_units, M_CLONE);
647 * Provide list of interface cloners to userspace.
650 if_clone_list(struct if_clonereq *ifcr)
652 char outbuf[IFNAMSIZ], *dst;
653 struct if_clone *ifc;
654 int count, error = 0;
656 ifcr->ifcr_total = if_cloners_count;
657 if ((dst = ifcr->ifcr_buffer) == NULL) {
658 /* Just asking how many there are. */
662 if (ifcr->ifcr_count < 0)
665 count = (if_cloners_count < ifcr->ifcr_count) ?
666 if_cloners_count : ifcr->ifcr_count;
668 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
669 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
670 strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
671 error = copyout(outbuf, dst, IFNAMSIZ);
680 * Locate an interface based on a complete address.
683 ifa_ifwithaddr(struct sockaddr *addr)
688 TAILQ_FOREACH(ifp, &ifnet, if_link)
689 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
690 if (ifa->ifa_addr->sa_family != addr->sa_family)
692 if (sa_equal(addr, ifa->ifa_addr))
694 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
695 /* IPv6 doesn't have broadcast */
696 ifa->ifa_broadaddr->sa_len != 0 &&
697 sa_equal(ifa->ifa_broadaddr, addr))
700 return ((struct ifaddr *)NULL);
703 * Locate the point to point interface with a given destination address.
706 ifa_ifwithdstaddr(struct sockaddr *addr)
711 TAILQ_FOREACH(ifp, &ifnet, if_link)
712 if (ifp->if_flags & IFF_POINTOPOINT)
713 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
714 if (ifa->ifa_addr->sa_family != addr->sa_family)
716 if (ifa->ifa_dstaddr &&
717 sa_equal(addr, ifa->ifa_dstaddr))
720 return ((struct ifaddr *)NULL);
724 * Find an interface on a specific network. If many, choice
725 * is most specific found.
728 ifa_ifwithnet(struct sockaddr *addr)
732 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
733 u_int af = addr->sa_family;
734 char *addr_data = addr->sa_data, *cplim;
737 * AF_LINK addresses can be looked up directly by their index number,
738 * so do that if we can.
741 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
743 if (sdl->sdl_index && sdl->sdl_index <= if_index)
744 return (ifnet_addrs[sdl->sdl_index - 1]);
748 * Scan though each interface, looking for ones that have
749 * addresses in this address family.
751 TAILQ_FOREACH(ifp, &ifnet, if_link) {
752 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
753 char *cp, *cp2, *cp3;
755 if (ifa->ifa_addr->sa_family != af)
757 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
759 * This is a bit broken as it doesn't
760 * take into account that the remote end may
761 * be a single node in the network we are
763 * The trouble is that we don't know the
764 * netmask for the remote end.
766 if (ifa->ifa_dstaddr != NULL &&
767 sa_equal(addr, ifa->ifa_dstaddr))
771 * if we have a special address handler,
772 * then use it instead of the generic one.
774 if (ifa->ifa_claim_addr) {
775 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
783 * Scan all the bits in the ifa's address.
784 * If a bit dissagrees with what we are
785 * looking for, mask it with the netmask
786 * to see if it really matters.
789 if (ifa->ifa_netmask == 0)
792 cp2 = ifa->ifa_addr->sa_data;
793 cp3 = ifa->ifa_netmask->sa_data;
794 cplim = ifa->ifa_netmask->sa_len +
795 (char *)ifa->ifa_netmask;
797 if ((*cp++ ^ *cp2++) & *cp3++)
798 goto next; /* next address! */
800 * If the netmask of what we just found
801 * is more specific than what we had before
802 * (if we had one) then remember the new one
803 * before continuing to search
804 * for an even better one.
806 if (ifa_maybe == 0 ||
807 rn_refines((char *)ifa->ifa_netmask,
808 (char *)ifa_maybe->ifa_netmask))
817 * Find an interface address specific to an interface best matching
821 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
824 char *cp, *cp2, *cp3;
826 struct ifaddr *ifa_maybe = 0;
827 u_int af = addr->sa_family;
831 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
832 if (ifa->ifa_addr->sa_family != af)
836 if (ifa->ifa_netmask == NULL) {
837 if (sa_equal(addr, ifa->ifa_addr) ||
838 (ifa->ifa_dstaddr != NULL &&
839 sa_equal(addr, ifa->ifa_dstaddr)))
843 if (ifp->if_flags & IFF_POINTOPOINT) {
844 if (sa_equal(addr, ifa->ifa_dstaddr))
848 cp2 = ifa->ifa_addr->sa_data;
849 cp3 = ifa->ifa_netmask->sa_data;
850 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
851 for (; cp3 < cplim; cp3++)
852 if ((*cp++ ^ *cp2++) & *cp3)
861 #include <net/route.h>
864 * Default action when installing a route with a Link Level gateway.
865 * Lookup an appropriate real ifa to point to.
866 * This should be moved to /sys/net/link.c eventually.
869 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
872 struct sockaddr *dst;
875 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
876 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
878 ifa = ifaof_ifpforaddr(dst, ifp);
883 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
884 ifa->ifa_rtrequest(cmd, rt, info);
889 * Mark an interface down and notify protocols of
891 * NOTE: must be called at splnet or eqivalent.
894 if_unroute(struct ifnet *ifp, int flag, int fam)
898 ifp->if_flags &= ~flag;
899 getmicrotime(&ifp->if_lastchange);
900 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
901 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
902 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
903 ifq_purge(&ifp->if_snd);
908 * Mark an interface up and notify protocols of
910 * NOTE: must be called at splnet or eqivalent.
913 if_route(struct ifnet *ifp, int flag, int fam)
917 ifp->if_flags |= flag;
918 getmicrotime(&ifp->if_lastchange);
919 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
920 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
921 pfctlinput(PRC_IFUP, ifa->ifa_addr);
929 * Mark an interface down and notify protocols of the transition. An
930 * interface going down is also considered to be a synchronizing event.
931 * We must ensure that all packet processing related to the interface
932 * has completed before we return so e.g. the caller can free the ifnet
933 * structure that the mbufs may be referencing.
935 * NOTE: must be called at splnet or eqivalent.
938 if_down(struct ifnet *ifp)
941 if_unroute(ifp, IFF_UP, AF_UNSPEC);
942 netmsg_service_sync();
946 * Mark an interface up and notify protocols of
948 * NOTE: must be called at splnet or eqivalent.
951 if_up(struct ifnet *ifp)
954 if_route(ifp, IFF_UP, AF_UNSPEC);
958 * Handle interface watchdog timer routines. Called
959 * from softclock, we decrement timers (if set) and
960 * call the appropriate interface routine on expiration.
963 if_slowtimo(void *arg)
968 TAILQ_FOREACH(ifp, &ifnet, if_link) {
969 if (ifp->if_timer == 0 || --ifp->if_timer)
971 if (ifp->if_watchdog)
972 (*ifp->if_watchdog)(ifp);
975 callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
979 * Map interface name to
980 * interface structure pointer.
983 ifunit(const char *name)
988 * Search all the interfaces for this name/number
991 TAILQ_FOREACH(ifp, &ifnet, if_link) {
992 if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1000 * Map interface name in a sockaddr_dl to
1001 * interface structure pointer.
1004 if_withname(struct sockaddr *sa)
1006 char ifname[IFNAMSIZ+1];
1007 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1009 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1010 (sdl->sdl_nlen > IFNAMSIZ) )
1014 * ifunit wants a null-terminated name. It may not be null-terminated
1015 * in the sockaddr. We don't want to change the caller's sockaddr,
1016 * and there might not be room to put the trailing null anyway, so we
1017 * make a local copy that we know we can null terminate safely.
1020 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1021 ifname[sdl->sdl_nlen] = '\0';
1022 return ifunit(ifname);
1030 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1038 size_t namelen, onamelen;
1039 char new_name[IFNAMSIZ];
1041 struct sockaddr_dl *sdl;
1047 return (ifconf(cmd, data, td));
1049 ifr = (struct ifreq *)data;
1054 if ((error = suser(td)) != 0)
1056 return ((cmd == SIOCIFCREATE) ?
1057 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1058 if_clone_destroy(ifr->ifr_name));
1060 case SIOCIFGCLONERS:
1061 return (if_clone_list((struct if_clonereq *)data));
1064 ifp = ifunit(ifr->ifr_name);
1070 ifr->ifr_flags = ifp->if_flags;
1071 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1075 ifr->ifr_reqcap = ifp->if_capabilities;
1076 ifr->ifr_curcap = ifp->if_capenable;
1080 ifr->ifr_metric = ifp->if_metric;
1084 ifr->ifr_mtu = ifp->if_mtu;
1088 ifr->ifr_phys = ifp->if_physical;
1095 new_flags = (ifr->ifr_flags & 0xffff) |
1096 (ifr->ifr_flagshigh << 16);
1097 if (ifp->if_flags & IFF_SMART) {
1098 /* Smart drivers twiddle their own routes */
1099 } else if (ifp->if_flags & IFF_UP &&
1100 (new_flags & IFF_UP) == 0) {
1104 } else if (new_flags & IFF_UP &&
1105 (ifp->if_flags & IFF_UP) == 0) {
1110 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1111 (new_flags &~ IFF_CANTCHANGE);
1112 if (new_flags & IFF_PPROMISC) {
1113 /* Permanently promiscuous mode requested */
1114 ifp->if_flags |= IFF_PROMISC;
1115 } else if (ifp->if_pcount == 0) {
1116 ifp->if_flags &= ~IFF_PROMISC;
1119 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1120 getmicrotime(&ifp->if_lastchange);
1127 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1129 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1136 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1139 if (new_name[0] == '\0')
1141 if (ifunit(new_name) != NULL)
1144 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1146 /* Announce the departure of the interface. */
1147 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1149 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1150 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1151 /* XXX IFA_LOCK(ifa); */
1152 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1153 namelen = strlen(new_name);
1154 onamelen = sdl->sdl_nlen;
1156 * Move the address if needed. This is safe because we
1157 * allocate space for a name of length IFNAMSIZ when we
1158 * create this in if_attach().
1160 if (namelen != onamelen) {
1161 bcopy(sdl->sdl_data + onamelen,
1162 sdl->sdl_data + namelen, sdl->sdl_alen);
1164 bcopy(new_name, sdl->sdl_data, namelen);
1165 sdl->sdl_nlen = namelen;
1166 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1167 bzero(sdl->sdl_data, onamelen);
1168 while (namelen != 0)
1169 sdl->sdl_data[--namelen] = 0xff;
1170 /* XXX IFA_UNLOCK(ifa) */
1172 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1174 /* Announce the return of the interface. */
1175 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1182 ifp->if_metric = ifr->ifr_metric;
1183 getmicrotime(&ifp->if_lastchange);
1192 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1194 getmicrotime(&ifp->if_lastchange);
1199 u_long oldmtu = ifp->if_mtu;
1204 if (ifp->if_ioctl == NULL)
1205 return (EOPNOTSUPP);
1206 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1208 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1210 getmicrotime(&ifp->if_lastchange);
1214 * If the link MTU changed, do network layer specific procedure.
1216 if (ifp->if_mtu != oldmtu) {
1230 /* Don't allow group membership on non-multicast interfaces. */
1231 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1234 /* Don't let users screw up protocols' entries. */
1235 if (ifr->ifr_addr.sa_family != AF_LINK)
1238 if (cmd == SIOCADDMULTI) {
1239 struct ifmultiaddr *ifma;
1240 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1242 error = if_delmulti(ifp, &ifr->ifr_addr);
1245 getmicrotime(&ifp->if_lastchange);
1248 case SIOCSIFPHYADDR:
1249 case SIOCDIFPHYADDR:
1251 case SIOCSIFPHYADDR_IN6:
1253 case SIOCSLIFPHYADDR:
1255 case SIOCSIFGENERIC:
1259 if (ifp->if_ioctl == 0)
1260 return (EOPNOTSUPP);
1261 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1263 getmicrotime(&ifp->if_lastchange);
1267 ifs = (struct ifstat *)data;
1268 ifs->ascii[0] = '\0';
1270 case SIOCGIFPSRCADDR:
1271 case SIOCGIFPDSTADDR:
1272 case SIOCGLIFPHYADDR:
1274 case SIOCGIFGENERIC:
1275 if (ifp->if_ioctl == 0)
1276 return (EOPNOTSUPP);
1277 return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1283 return if_setlladdr(ifp,
1284 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1287 oif_flags = ifp->if_flags;
1288 if (so->so_proto == 0)
1289 return (EOPNOTSUPP);
1291 error = so_pru_control(so, cmd, data, ifp, td);
1298 case SIOCSIFDSTADDR:
1300 case SIOCSIFBRDADDR:
1301 case SIOCSIFNETMASK:
1302 #if BYTE_ORDER != BIG_ENDIAN
1303 if (ifr->ifr_addr.sa_family == 0 &&
1304 ifr->ifr_addr.sa_len < 16) {
1305 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1306 ifr->ifr_addr.sa_len = 16;
1309 if (ifr->ifr_addr.sa_len == 0)
1310 ifr->ifr_addr.sa_len = 16;
1318 case OSIOCGIFDSTADDR:
1319 cmd = SIOCGIFDSTADDR;
1322 case OSIOCGIFBRDADDR:
1323 cmd = SIOCGIFBRDADDR;
1326 case OSIOCGIFNETMASK:
1327 cmd = SIOCGIFNETMASK;
1329 error = so_pru_control(so, cmd, data, ifp, td);
1333 case OSIOCGIFDSTADDR:
1334 case OSIOCGIFBRDADDR:
1335 case OSIOCGIFNETMASK:
1336 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1340 #endif /* COMPAT_43 */
1342 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1344 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1345 if (ifp->if_flags & IFF_UP) {
1359 * Set/clear promiscuous mode on interface ifp based on the truth value
1360 * of pswitch. The calls are reference counted so that only the first
1361 * "on" request actually has an effect, as does the final "off" request.
1362 * Results are undefined if the "off" and "on" requests are not matched.
1365 ifpromisc(struct ifnet *ifp, int pswitch)
1371 oldflags = ifp->if_flags;
1372 if (ifp->if_flags & IFF_PPROMISC) {
1373 /* Do nothing if device is in permanently promiscuous mode */
1374 ifp->if_pcount += pswitch ? 1 : -1;
1379 * If the device is not configured up, we cannot put it in
1382 if ((ifp->if_flags & IFF_UP) == 0)
1384 if (ifp->if_pcount++ != 0)
1386 ifp->if_flags |= IFF_PROMISC;
1387 log(LOG_INFO, "%s: promiscuous mode enabled\n",
1390 if (--ifp->if_pcount > 0)
1392 ifp->if_flags &= ~IFF_PROMISC;
1393 log(LOG_INFO, "%s: promiscuous mode disabled\n",
1396 ifr.ifr_flags = ifp->if_flags;
1397 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1398 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1399 (struct ucred *)NULL);
1403 ifp->if_flags = oldflags;
1408 * Return interface configuration
1409 * of system. List may be used
1410 * in later ioctl's (above) to get
1411 * other information.
1414 ifconf(u_long cmd, caddr_t data, struct thread *td)
1416 struct ifconf *ifc = (struct ifconf *)data;
1419 struct sockaddr *sa;
1420 struct ifreq ifr, *ifrp;
1421 int space = ifc->ifc_len, error = 0;
1423 ifrp = ifc->ifc_req;
1424 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1427 if (space <= sizeof ifr)
1429 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1430 >= sizeof(ifr.ifr_name)) {
1431 error = ENAMETOOLONG;
1436 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1437 if (space <= sizeof ifr)
1440 if (td->td_proc->p_ucred->cr_prison &&
1445 if (cmd == OSIOCGIFCONF) {
1446 struct osockaddr *osa =
1447 (struct osockaddr *)&ifr.ifr_addr;
1449 osa->sa_family = sa->sa_family;
1450 error = copyout(&ifr, ifrp, sizeof ifr);
1454 if (sa->sa_len <= sizeof(*sa)) {
1456 error = copyout(&ifr, ifrp, sizeof ifr);
1459 if (space < (sizeof ifr) + sa->sa_len -
1462 space -= sa->sa_len - sizeof(*sa);
1463 error = copyout(&ifr, ifrp,
1464 sizeof ifr.ifr_name);
1466 error = copyout(sa, &ifrp->ifr_addr,
1468 ifrp = (struct ifreq *)
1469 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1473 space -= sizeof ifr;
1478 bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1479 error = copyout(&ifr, ifrp, sizeof ifr);
1482 space -= sizeof ifr;
1486 ifc->ifc_len -= space;
1491 * Just like if_promisc(), but for all-multicast-reception mode.
1494 if_allmulti(struct ifnet *ifp, int onswitch)
1501 if (ifp->if_amcount++ == 0) {
1502 ifp->if_flags |= IFF_ALLMULTI;
1503 ifr.ifr_flags = ifp->if_flags;
1504 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1505 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1506 (struct ucred *)NULL);
1509 if (ifp->if_amcount > 1) {
1512 ifp->if_amcount = 0;
1513 ifp->if_flags &= ~IFF_ALLMULTI;
1514 ifr.ifr_flags = ifp->if_flags;
1515 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1516 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1517 (struct ucred *)NULL);
1528 * Add a multicast listenership to the interface in question.
1529 * The link layer provides a routine which converts
1533 struct ifnet *ifp, /* interface to manipulate */
1534 struct sockaddr *sa, /* address to add */
1535 struct ifmultiaddr **retifma)
1537 struct sockaddr *llsa, *dupsa;
1539 struct ifmultiaddr *ifma;
1542 * If the matching multicast address already exists
1543 * then don't add a new one, just add a reference
1545 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1546 if (sa_equal(sa, ifma->ifma_addr)) {
1547 ifma->ifma_refcount++;
1555 * Give the link layer a chance to accept/reject it, and also
1556 * find out which AF_LINK address this maps to, if it isn't one
1559 if (ifp->if_resolvemulti) {
1560 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1561 if (error) return error;
1566 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1567 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1568 bcopy(sa, dupsa, sa->sa_len);
1570 ifma->ifma_addr = dupsa;
1571 ifma->ifma_lladdr = llsa;
1572 ifma->ifma_ifp = ifp;
1573 ifma->ifma_refcount = 1;
1574 ifma->ifma_protospec = 0;
1575 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1578 * Some network interfaces can scan the address list at
1579 * interrupt time; lock them out.
1582 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1587 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1588 if (sa_equal(ifma->ifma_addr, llsa))
1592 ifma->ifma_refcount++;
1594 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1595 M_IFMADDR, M_WAITOK);
1596 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1597 M_IFMADDR, M_WAITOK);
1598 bcopy(llsa, dupsa, llsa->sa_len);
1599 ifma->ifma_addr = dupsa;
1600 ifma->ifma_ifp = ifp;
1601 ifma->ifma_refcount = 1;
1603 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1608 * We are certain we have added something, so call down to the
1609 * interface to let them know about it.
1612 ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1619 * Remove a reference to a multicast address on this interface. Yell
1620 * if the request does not match an existing membership.
1623 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1625 struct ifmultiaddr *ifma;
1628 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1629 if (sa_equal(sa, ifma->ifma_addr))
1634 if (ifma->ifma_refcount > 1) {
1635 ifma->ifma_refcount--;
1639 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1640 sa = ifma->ifma_lladdr;
1642 LIST_REMOVE(ifma, ifma_link);
1644 * Make sure the interface driver is notified
1645 * in the case of a link layer mcast group being left.
1647 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1648 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1650 free(ifma->ifma_addr, M_IFMADDR);
1651 free(ifma, M_IFMADDR);
1656 * Now look for the link-layer address which corresponds to
1657 * this network address. It had been squirreled away in
1658 * ifma->ifma_lladdr for this purpose (so we don't have
1659 * to call ifp->if_resolvemulti() again), and we saved that
1660 * value in sa above. If some nasty deleted the
1661 * link-layer address out from underneath us, we can deal because
1662 * the address we stored was is not the same as the one which was
1663 * in the record for the link-layer address. (So we don't complain
1666 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1667 if (sa_equal(sa, ifma->ifma_addr))
1672 if (ifma->ifma_refcount > 1) {
1673 ifma->ifma_refcount--;
1678 LIST_REMOVE(ifma, ifma_link);
1679 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1681 free(ifma->ifma_addr, M_IFMADDR);
1682 free(sa, M_IFMADDR);
1683 free(ifma, M_IFMADDR);
1689 * Set the link layer address on an interface.
1691 * At this time we only support certain types of interfaces,
1692 * and we don't allow the length of the address to change.
1695 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1697 struct sockaddr_dl *sdl;
1701 ifa = ifnet_addrs[ifp->if_index - 1];
1704 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1707 if (len != sdl->sdl_alen) /* don't allow length to change */
1709 switch (ifp->if_type) {
1710 case IFT_ETHER: /* these types use struct arpcom */
1715 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1718 bcopy(lladdr, LLADDR(sdl), len);
1724 * If the interface is already up, we need
1725 * to re-init it in order to reprogram its
1728 if ((ifp->if_flags & IFF_UP) != 0) {
1729 ifp->if_flags &= ~IFF_UP;
1730 ifr.ifr_flags = ifp->if_flags;
1731 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1732 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1733 (struct ucred *)NULL);
1734 ifp->if_flags |= IFF_UP;
1735 ifr.ifr_flags = ifp->if_flags;
1736 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1737 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1738 (struct ucred *)NULL);
1741 * Also send gratuitous ARPs to notify other nodes about
1742 * the address change.
1744 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1745 if (ifa->ifa_addr != NULL &&
1746 ifa->ifa_addr->sa_family == AF_INET)
1747 arp_ifinit(ifp, ifa);
1754 struct ifmultiaddr *
1755 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1757 struct ifmultiaddr *ifma;
1759 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1760 if (sa_equal(ifma->ifma_addr, sa))
1767 * The name argument must be a pointer to storage which will last as
1768 * long as the interface does. For physical devices, the result of
1769 * device_get_name(dev) is a good choice and for pseudo-devices a
1770 * static string works well.
1773 if_initname(struct ifnet *ifp, const char *name, int unit)
1775 ifp->if_dname = name;
1776 ifp->if_dunit = unit;
1777 if (unit != IF_DUNIT_NONE)
1778 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1780 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1784 if_printf(struct ifnet *ifp, const char *fmt, ...)
1789 retval = printf("%s: ", ifp->if_xname);
1790 __va_start(ap, fmt);
1791 retval += vprintf(fmt, ap);
1796 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1797 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1800 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1801 struct altq_pktattr *pa __unused)
1803 if (IF_QFULL(ifq)) {
1812 static struct mbuf *
1813 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1825 panic("unsupported ALTQ dequeue op: %d", op);
1832 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1839 panic("unspported ALTQ request: %d", req);