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 1.34 2005/05/24 20:58:43 dillon 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>
57 #include <sys/thread.h>
60 #include <net/if_arp.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/ifq_var.h>
65 #include <net/radix.h>
66 #include <net/route.h>
67 #include <machine/stdarg.h>
69 #include <sys/thread2.h>
71 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
77 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
78 #include <netinet6/in6_var.h>
79 #include <netinet6/in6_ifattach.h>
83 #if defined(COMPAT_43)
84 #include <emulation/43bsd/43bsd_socket.h>
85 #endif /* COMPAT_43 */
88 * Support for non-ALTQ interfaces.
90 static int ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
91 struct altq_pktattr *);
93 ifq_classic_dequeue(struct ifaltq *, int);
94 static int ifq_classic_request(struct ifaltq *, int, void *);
97 * System initialization
100 static void if_attachdomain(void *);
101 static void if_attachdomain1(struct ifnet *);
102 static int ifconf (u_long, caddr_t, struct thread *);
103 static void ifinit (void *);
104 static void if_slowtimo (void *);
105 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
106 static int if_rtdel (struct radix_node *, void *);
108 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
110 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
111 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
112 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
114 int ifqmaxlen = IFQ_MAXLEN;
115 struct ifnethead ifnet; /* depend on static init XXX */
119 * XXX: declare here to avoid to include many inet6 related files..
120 * should be more generalized?
122 extern void nd6_setmtu (struct ifnet *);
125 struct if_clone *if_clone_lookup (const char *, int *);
126 int if_clone_list (struct if_clonereq *);
128 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
129 int if_cloners_count;
131 struct callout if_slowtimo_timer;
134 * Network interface utility routines.
136 * Routines with ifa_ifwith* names take sockaddr *'s as
146 callout_init(&if_slowtimo_timer);
149 TAILQ_FOREACH(ifp, &ifnet, if_link) {
150 if (ifp->if_snd.ifq_maxlen == 0) {
151 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
152 ifp->if_snd.ifq_maxlen = ifqmaxlen;
161 struct ifaddr **ifnet_addrs;
162 struct ifnet **ifindex2ifnet = NULL;
165 * Attach an interface to the
166 * list of "active" interfaces.
169 if_attach(struct ifnet *ifp)
171 unsigned socksize, ifasize;
172 int namelen, masklen;
173 struct sockaddr_dl *sdl;
177 static int if_indexlim = 8;
178 static boolean_t inited;
185 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
186 ifp->if_index = ++if_index;
189 * The old code would work if the interface passed a pre-existing
190 * chain of ifaddrs to this code. We don't trust our callers to
191 * properly initialize the tailq, however, so we no longer allow
192 * this unlikely case.
194 TAILQ_INIT(&ifp->if_addrhead);
195 TAILQ_INIT(&ifp->if_prefixhead);
196 LIST_INIT(&ifp->if_multiaddrs);
197 getmicrotime(&ifp->if_lastchange);
198 if (ifnet_addrs == NULL || if_index >= if_indexlim) {
203 n = if_indexlim * sizeof(struct ifaddr *);
204 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
205 if (ifnet_addrs != NULL) {
206 bcopy(ifnet_addrs, q, n/2);
207 free(ifnet_addrs, M_IFADDR);
209 ifnet_addrs = (struct ifaddr **)q;
211 /* grow ifindex2ifnet */
212 n = if_indexlim * sizeof(struct ifnet *);
213 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
215 bcopy(ifindex2ifnet, q, n/2);
216 free(ifindex2ifnet, M_IFADDR);
218 ifindex2ifnet = (struct ifnet **)q;
221 ifindex2ifnet[if_index] = ifp;
224 * create a Link Level name for this device
226 namelen = strlen(ifp->if_xname);
227 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
228 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
229 socksize = masklen + ifp->if_addrlen;
230 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
231 if (socksize < sizeof(*sdl))
232 socksize = sizeof(*sdl);
233 socksize = ROUNDUP(socksize);
234 ifasize = sizeof(struct ifaddr) + 2 * socksize;
235 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
236 sdl = (struct sockaddr_dl *)(ifa + 1);
237 sdl->sdl_len = socksize;
238 sdl->sdl_family = AF_LINK;
239 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
240 sdl->sdl_nlen = namelen;
241 sdl->sdl_index = ifp->if_index;
242 sdl->sdl_type = ifp->if_type;
243 ifnet_addrs[if_index - 1] = ifa;
245 ifa->ifa_rtrequest = link_rtrequest;
246 ifa->ifa_addr = (struct sockaddr *)sdl;
247 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
248 ifa->ifa_netmask = (struct sockaddr *)sdl;
249 sdl->sdl_len = masklen;
251 sdl->sdl_data[--namelen] = 0xff;
252 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
254 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
258 ifq->altq_disc = NULL;
259 ifq->altq_flags &= ALTQF_CANTCHANGE;
260 ifq->altq_tbr = NULL;
262 ifq_set_classic(ifq);
264 if (!SLIST_EMPTY(&domains))
265 if_attachdomain1(ifp);
267 /* Announce the interface. */
268 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
272 if_attachdomain(void *dummy)
278 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
279 if_attachdomain1(ifp);
282 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
283 if_attachdomain, NULL);
286 if_attachdomain1(struct ifnet *ifp)
293 /* address family dependent data region */
294 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
295 SLIST_FOREACH(dp, &domains, dom_next)
296 if (dp->dom_ifattach)
297 ifp->if_afdata[dp->dom_family] =
298 (*dp->dom_ifattach)(ifp);
303 * Detach an interface, removing it from the
304 * list of "active" interfaces.
307 if_detach(struct ifnet *ifp)
310 struct radix_node_head *rnh;
315 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
318 * Remove routes and flush queues.
323 if (ifq_is_enabled(&ifp->if_snd))
324 altq_disable(&ifp->if_snd);
325 if (ifq_is_attached(&ifp->if_snd))
326 altq_detach(&ifp->if_snd);
329 * Remove address from ifnet_addrs[] and maybe decrement if_index.
330 * Clean up all addresses.
332 ifnet_addrs[ifp->if_index - 1] = 0;
333 while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
336 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
337 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
339 /* XXX: Ugly!! ad hoc just for INET */
340 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
341 struct ifaliasreq ifr;
343 bzero(&ifr, sizeof ifr);
344 ifr.ifra_addr = *ifa->ifa_addr;
345 if (ifa->ifa_dstaddr)
346 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
347 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
353 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
355 /* ifp_addrhead is already updated */
359 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
365 * Remove all IPv6 kernel structs related to ifp. This should be done
366 * before removing routing entries below, since IPv6 interface direct
367 * routes are expected to be removed by the IPv6-specific kernel API.
368 * Otherwise, the kernel will detect some inconsistency and bark it.
374 * Delete all remaining routes using this interface
375 * Unfortuneatly the only way to do this is to slog through
376 * the entire routing table looking for routes which point
377 * to this interface...oh well...
379 for (i = 1; i <= AF_MAX; i++) {
380 if ((rnh = rt_tables[i]) == NULL)
382 rnh->rnh_walktree(rnh, if_rtdel, ifp);
385 /* Announce that the interface is gone. */
386 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
388 SLIST_FOREACH(dp, &domains, dom_next)
389 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
390 (*dp->dom_ifdetach)(ifp,
391 ifp->if_afdata[dp->dom_family]);
393 ifindex2ifnet[ifp->if_index] = NULL;
395 TAILQ_REMOVE(&ifnet, ifp, if_link);
400 * Delete Routes for a Network Interface
402 * Called for each routing entry via the rnh->rnh_walktree() call above
403 * to delete all route entries referencing a detaching network interface.
406 * rn pointer to node in the routing table
407 * arg argument passed to rnh->rnh_walktree() - detaching interface
411 * errno failed - reason indicated
415 if_rtdel(struct radix_node *rn, void *arg)
417 struct rtentry *rt = (struct rtentry *)rn;
418 struct ifnet *ifp = arg;
421 if (rt->rt_ifp == ifp) {
424 * Protect (sorta) against walktree recursion problems
427 if (!(rt->rt_flags & RTF_UP))
430 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
431 rt_mask(rt), rt->rt_flags,
432 (struct rtentry **) NULL);
434 log(LOG_WARNING, "if_rtdel: error %d\n", err);
442 * Create a clone network interface.
445 if_clone_create(char *name, int len)
447 struct if_clone *ifc;
449 int wildcard, bytoff, bitoff;
453 ifc = if_clone_lookup(name, &unit);
457 if (ifunit(name) != NULL)
461 wildcard = (unit < 0);
463 * Find a free unit if none was given.
466 while (bytoff < ifc->ifc_bmlen &&
467 ifc->ifc_units[bytoff] == 0xff)
469 if (bytoff >= ifc->ifc_bmlen)
471 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
473 unit = (bytoff << 3) + bitoff;
476 if (unit > ifc->ifc_maxunit)
479 err = (*ifc->ifc_create)(ifc, unit);
485 bitoff = unit - (bytoff << 3);
489 * Allocate the unit in the bitmap.
491 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
492 ("%s: bit is already set", __func__));
493 ifc->ifc_units[bytoff] |= (1 << bitoff);
495 /* In the wildcard case, we need to update the name. */
497 for (dp = name; *dp != '\0'; dp++);
498 if (snprintf(dp, len - (dp-name), "%d", unit) >
499 len - (dp-name) - 1) {
501 * This can only be a programmer error and
502 * there's no straightforward way to recover if
505 panic("if_clone_create(): interface name too long");
510 EVENTHANDLER_INVOKE(if_clone_event, ifc);
516 * Destroy a clone network interface.
519 if_clone_destroy(const char *name)
521 struct if_clone *ifc;
526 ifc = if_clone_lookup(name, &unit);
530 if (unit < ifc->ifc_minifs)
537 if (ifc->ifc_destroy == NULL)
540 (*ifc->ifc_destroy)(ifp);
543 * Compute offset in the bitmap and deallocate the unit.
546 bitoff = unit - (bytoff << 3);
547 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
548 ("%s: bit is already cleared", __func__));
549 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
554 * Look up a network interface cloner.
557 if_clone_lookup(const char *name, int *unitp)
559 struct if_clone *ifc;
563 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
564 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
565 if (ifc->ifc_name[i] != *cp)
570 ifc = LIST_NEXT(ifc, ifc_list);
574 return ((struct if_clone *)NULL);
580 for (i = 0; *cp != '\0'; cp++) {
581 if (*cp < '0' || *cp > '9') {
582 /* Bogus unit number. */
585 i = (i * 10) + (*cp - '0');
595 * Register a network interface cloner.
598 if_clone_attach(struct if_clone *ifc)
605 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
606 ("%s: %s requested more units then allowed (%d > %d)",
607 __func__, ifc->ifc_name, ifc->ifc_minifs,
608 ifc->ifc_maxunit + 1));
610 * Compute bitmap size and allocate it.
612 maxclone = ifc->ifc_maxunit + 1;
614 if ((len << 3) < maxclone)
616 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
617 ifc->ifc_bmlen = len;
619 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
622 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
623 err = (*ifc->ifc_create)(ifc, unit);
625 ("%s: failed to create required interface %s%d",
626 __func__, ifc->ifc_name, unit));
628 /* Allocate the unit in the bitmap. */
630 bitoff = unit - (bytoff << 3);
631 ifc->ifc_units[bytoff] |= (1 << bitoff);
636 * Unregister a network interface cloner.
639 if_clone_detach(struct if_clone *ifc)
642 LIST_REMOVE(ifc, ifc_list);
643 free(ifc->ifc_units, M_CLONE);
648 * Provide list of interface cloners to userspace.
651 if_clone_list(struct if_clonereq *ifcr)
653 char outbuf[IFNAMSIZ], *dst;
654 struct if_clone *ifc;
655 int count, error = 0;
657 ifcr->ifcr_total = if_cloners_count;
658 if ((dst = ifcr->ifcr_buffer) == NULL) {
659 /* Just asking how many there are. */
663 if (ifcr->ifcr_count < 0)
666 count = (if_cloners_count < ifcr->ifcr_count) ?
667 if_cloners_count : ifcr->ifcr_count;
669 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
670 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
671 strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
672 error = copyout(outbuf, dst, IFNAMSIZ);
681 * Locate an interface based on a complete address.
684 ifa_ifwithaddr(struct sockaddr *addr)
689 TAILQ_FOREACH(ifp, &ifnet, if_link)
690 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
691 if (ifa->ifa_addr->sa_family != addr->sa_family)
693 if (sa_equal(addr, ifa->ifa_addr))
695 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
696 /* IPv6 doesn't have broadcast */
697 ifa->ifa_broadaddr->sa_len != 0 &&
698 sa_equal(ifa->ifa_broadaddr, addr))
701 return ((struct ifaddr *)NULL);
704 * Locate the point to point interface with a given destination address.
707 ifa_ifwithdstaddr(struct sockaddr *addr)
712 TAILQ_FOREACH(ifp, &ifnet, if_link)
713 if (ifp->if_flags & IFF_POINTOPOINT)
714 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
715 if (ifa->ifa_addr->sa_family != addr->sa_family)
717 if (ifa->ifa_dstaddr &&
718 sa_equal(addr, ifa->ifa_dstaddr))
721 return ((struct ifaddr *)NULL);
725 * Find an interface on a specific network. If many, choice
726 * is most specific found.
729 ifa_ifwithnet(struct sockaddr *addr)
733 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
734 u_int af = addr->sa_family;
735 char *addr_data = addr->sa_data, *cplim;
738 * AF_LINK addresses can be looked up directly by their index number,
739 * so do that if we can.
742 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
744 if (sdl->sdl_index && sdl->sdl_index <= if_index)
745 return (ifnet_addrs[sdl->sdl_index - 1]);
749 * Scan though each interface, looking for ones that have
750 * addresses in this address family.
752 TAILQ_FOREACH(ifp, &ifnet, if_link) {
753 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
754 char *cp, *cp2, *cp3;
756 if (ifa->ifa_addr->sa_family != af)
758 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
760 * This is a bit broken as it doesn't
761 * take into account that the remote end may
762 * be a single node in the network we are
764 * The trouble is that we don't know the
765 * netmask for the remote end.
767 if (ifa->ifa_dstaddr != NULL &&
768 sa_equal(addr, ifa->ifa_dstaddr))
772 * if we have a special address handler,
773 * then use it instead of the generic one.
775 if (ifa->ifa_claim_addr) {
776 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
784 * Scan all the bits in the ifa's address.
785 * If a bit dissagrees with what we are
786 * looking for, mask it with the netmask
787 * to see if it really matters.
790 if (ifa->ifa_netmask == 0)
793 cp2 = ifa->ifa_addr->sa_data;
794 cp3 = ifa->ifa_netmask->sa_data;
795 cplim = ifa->ifa_netmask->sa_len +
796 (char *)ifa->ifa_netmask;
798 if ((*cp++ ^ *cp2++) & *cp3++)
799 goto next; /* next address! */
801 * If the netmask of what we just found
802 * is more specific than what we had before
803 * (if we had one) then remember the new one
804 * before continuing to search
805 * for an even better one.
807 if (ifa_maybe == 0 ||
808 rn_refines((char *)ifa->ifa_netmask,
809 (char *)ifa_maybe->ifa_netmask))
818 * Find an interface address specific to an interface best matching
822 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
825 char *cp, *cp2, *cp3;
827 struct ifaddr *ifa_maybe = 0;
828 u_int af = addr->sa_family;
832 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
833 if (ifa->ifa_addr->sa_family != af)
837 if (ifa->ifa_netmask == NULL) {
838 if (sa_equal(addr, ifa->ifa_addr) ||
839 (ifa->ifa_dstaddr != NULL &&
840 sa_equal(addr, ifa->ifa_dstaddr)))
844 if (ifp->if_flags & IFF_POINTOPOINT) {
845 if (sa_equal(addr, ifa->ifa_dstaddr))
849 cp2 = ifa->ifa_addr->sa_data;
850 cp3 = ifa->ifa_netmask->sa_data;
851 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
852 for (; cp3 < cplim; cp3++)
853 if ((*cp++ ^ *cp2++) & *cp3)
862 #include <net/route.h>
865 * Default action when installing a route with a Link Level gateway.
866 * Lookup an appropriate real ifa to point to.
867 * This should be moved to /sys/net/link.c eventually.
870 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
873 struct sockaddr *dst;
876 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
877 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
879 ifa = ifaof_ifpforaddr(dst, ifp);
884 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
885 ifa->ifa_rtrequest(cmd, rt, info);
890 * Mark an interface down and notify protocols of
892 * NOTE: must be called at splnet or eqivalent.
895 if_unroute(struct ifnet *ifp, int flag, int fam)
899 ifp->if_flags &= ~flag;
900 getmicrotime(&ifp->if_lastchange);
901 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
902 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
903 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
904 ifq_purge(&ifp->if_snd);
909 * Mark an interface up and notify protocols of
911 * NOTE: must be called at splnet or eqivalent.
914 if_route(struct ifnet *ifp, int flag, int fam)
918 ifp->if_flags |= flag;
919 getmicrotime(&ifp->if_lastchange);
920 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
921 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
922 pfctlinput(PRC_IFUP, ifa->ifa_addr);
930 * Mark an interface down and notify protocols of the transition. An
931 * interface going down is also considered to be a synchronizing event.
932 * We must ensure that all packet processing related to the interface
933 * has completed before we return so e.g. the caller can free the ifnet
934 * structure that the mbufs may be referencing.
936 * NOTE: must be called at splnet or eqivalent.
939 if_down(struct ifnet *ifp)
942 if_unroute(ifp, IFF_UP, AF_UNSPEC);
943 netmsg_service_sync();
947 * Mark an interface up and notify protocols of
949 * NOTE: must be called at splnet or eqivalent.
952 if_up(struct ifnet *ifp)
955 if_route(ifp, IFF_UP, AF_UNSPEC);
959 * Handle interface watchdog timer routines. Called
960 * from softclock, we decrement timers (if set) and
961 * call the appropriate interface routine on expiration.
964 if_slowtimo(void *arg)
969 TAILQ_FOREACH(ifp, &ifnet, if_link) {
970 if (ifp->if_timer == 0 || --ifp->if_timer)
972 if (ifp->if_watchdog)
973 (*ifp->if_watchdog)(ifp);
976 callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
980 * Map interface name to
981 * interface structure pointer.
984 ifunit(const char *name)
989 * Search all the interfaces for this name/number
992 TAILQ_FOREACH(ifp, &ifnet, if_link) {
993 if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1001 * Map interface name in a sockaddr_dl to
1002 * interface structure pointer.
1005 if_withname(struct sockaddr *sa)
1007 char ifname[IFNAMSIZ+1];
1008 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1010 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1011 (sdl->sdl_nlen > IFNAMSIZ) )
1015 * ifunit wants a null-terminated name. It may not be null-terminated
1016 * in the sockaddr. We don't want to change the caller's sockaddr,
1017 * and there might not be room to put the trailing null anyway, so we
1018 * make a local copy that we know we can null terminate safely.
1021 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1022 ifname[sdl->sdl_nlen] = '\0';
1023 return ifunit(ifname);
1031 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1039 size_t namelen, onamelen;
1040 char new_name[IFNAMSIZ];
1042 struct sockaddr_dl *sdl;
1048 return (ifconf(cmd, data, td));
1050 ifr = (struct ifreq *)data;
1055 if ((error = suser(td)) != 0)
1057 return ((cmd == SIOCIFCREATE) ?
1058 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1059 if_clone_destroy(ifr->ifr_name));
1061 case SIOCIFGCLONERS:
1062 return (if_clone_list((struct if_clonereq *)data));
1065 ifp = ifunit(ifr->ifr_name);
1071 ifr->ifr_flags = ifp->if_flags;
1072 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1076 ifr->ifr_reqcap = ifp->if_capabilities;
1077 ifr->ifr_curcap = ifp->if_capenable;
1081 ifr->ifr_metric = ifp->if_metric;
1085 ifr->ifr_mtu = ifp->if_mtu;
1089 ifr->ifr_phys = ifp->if_physical;
1096 new_flags = (ifr->ifr_flags & 0xffff) |
1097 (ifr->ifr_flagshigh << 16);
1098 if (ifp->if_flags & IFF_SMART) {
1099 /* Smart drivers twiddle their own routes */
1100 } else if (ifp->if_flags & IFF_UP &&
1101 (new_flags & IFF_UP) == 0) {
1105 } else if (new_flags & IFF_UP &&
1106 (ifp->if_flags & IFF_UP) == 0) {
1111 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1112 (new_flags &~ IFF_CANTCHANGE);
1113 if (new_flags & IFF_PPROMISC) {
1114 /* Permanently promiscuous mode requested */
1115 ifp->if_flags |= IFF_PROMISC;
1116 } else if (ifp->if_pcount == 0) {
1117 ifp->if_flags &= ~IFF_PROMISC;
1120 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1121 getmicrotime(&ifp->if_lastchange);
1128 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1130 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1137 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1140 if (new_name[0] == '\0')
1142 if (ifunit(new_name) != NULL)
1145 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1147 /* Announce the departure of the interface. */
1148 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1150 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1151 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1152 /* XXX IFA_LOCK(ifa); */
1153 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1154 namelen = strlen(new_name);
1155 onamelen = sdl->sdl_nlen;
1157 * Move the address if needed. This is safe because we
1158 * allocate space for a name of length IFNAMSIZ when we
1159 * create this in if_attach().
1161 if (namelen != onamelen) {
1162 bcopy(sdl->sdl_data + onamelen,
1163 sdl->sdl_data + namelen, sdl->sdl_alen);
1165 bcopy(new_name, sdl->sdl_data, namelen);
1166 sdl->sdl_nlen = namelen;
1167 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1168 bzero(sdl->sdl_data, onamelen);
1169 while (namelen != 0)
1170 sdl->sdl_data[--namelen] = 0xff;
1171 /* XXX IFA_UNLOCK(ifa) */
1173 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1175 /* Announce the return of the interface. */
1176 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1183 ifp->if_metric = ifr->ifr_metric;
1184 getmicrotime(&ifp->if_lastchange);
1193 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1195 getmicrotime(&ifp->if_lastchange);
1200 u_long oldmtu = ifp->if_mtu;
1205 if (ifp->if_ioctl == NULL)
1206 return (EOPNOTSUPP);
1207 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1209 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1211 getmicrotime(&ifp->if_lastchange);
1215 * If the link MTU changed, do network layer specific procedure.
1217 if (ifp->if_mtu != oldmtu) {
1231 /* Don't allow group membership on non-multicast interfaces. */
1232 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1235 /* Don't let users screw up protocols' entries. */
1236 if (ifr->ifr_addr.sa_family != AF_LINK)
1239 if (cmd == SIOCADDMULTI) {
1240 struct ifmultiaddr *ifma;
1241 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1243 error = if_delmulti(ifp, &ifr->ifr_addr);
1246 getmicrotime(&ifp->if_lastchange);
1249 case SIOCSIFPHYADDR:
1250 case SIOCDIFPHYADDR:
1252 case SIOCSIFPHYADDR_IN6:
1254 case SIOCSLIFPHYADDR:
1256 case SIOCSIFGENERIC:
1260 if (ifp->if_ioctl == 0)
1261 return (EOPNOTSUPP);
1262 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1264 getmicrotime(&ifp->if_lastchange);
1268 ifs = (struct ifstat *)data;
1269 ifs->ascii[0] = '\0';
1271 case SIOCGIFPSRCADDR:
1272 case SIOCGIFPDSTADDR:
1273 case SIOCGLIFPHYADDR:
1275 case SIOCGIFGENERIC:
1276 if (ifp->if_ioctl == 0)
1277 return (EOPNOTSUPP);
1278 return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1284 return if_setlladdr(ifp,
1285 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1288 oif_flags = ifp->if_flags;
1289 if (so->so_proto == 0)
1290 return (EOPNOTSUPP);
1292 error = so_pru_control(so, cmd, data, ifp, td);
1299 case SIOCSIFDSTADDR:
1301 case SIOCSIFBRDADDR:
1302 case SIOCSIFNETMASK:
1303 #if BYTE_ORDER != BIG_ENDIAN
1304 if (ifr->ifr_addr.sa_family == 0 &&
1305 ifr->ifr_addr.sa_len < 16) {
1306 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1307 ifr->ifr_addr.sa_len = 16;
1310 if (ifr->ifr_addr.sa_len == 0)
1311 ifr->ifr_addr.sa_len = 16;
1319 case OSIOCGIFDSTADDR:
1320 cmd = SIOCGIFDSTADDR;
1323 case OSIOCGIFBRDADDR:
1324 cmd = SIOCGIFBRDADDR;
1327 case OSIOCGIFNETMASK:
1328 cmd = SIOCGIFNETMASK;
1330 error = so_pru_control(so, cmd, data, ifp, td);
1334 case OSIOCGIFDSTADDR:
1335 case OSIOCGIFBRDADDR:
1336 case OSIOCGIFNETMASK:
1337 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1341 #endif /* COMPAT_43 */
1343 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1345 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1346 if (ifp->if_flags & IFF_UP) {
1360 * Set/clear promiscuous mode on interface ifp based on the truth value
1361 * of pswitch. The calls are reference counted so that only the first
1362 * "on" request actually has an effect, as does the final "off" request.
1363 * Results are undefined if the "off" and "on" requests are not matched.
1366 ifpromisc(struct ifnet *ifp, int pswitch)
1372 oldflags = ifp->if_flags;
1373 if (ifp->if_flags & IFF_PPROMISC) {
1374 /* Do nothing if device is in permanently promiscuous mode */
1375 ifp->if_pcount += pswitch ? 1 : -1;
1380 * If the device is not configured up, we cannot put it in
1383 if ((ifp->if_flags & IFF_UP) == 0)
1385 if (ifp->if_pcount++ != 0)
1387 ifp->if_flags |= IFF_PROMISC;
1388 log(LOG_INFO, "%s: promiscuous mode enabled\n",
1391 if (--ifp->if_pcount > 0)
1393 ifp->if_flags &= ~IFF_PROMISC;
1394 log(LOG_INFO, "%s: promiscuous mode disabled\n",
1397 ifr.ifr_flags = ifp->if_flags;
1398 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1399 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1400 (struct ucred *)NULL);
1404 ifp->if_flags = oldflags;
1409 * Return interface configuration
1410 * of system. List may be used
1411 * in later ioctl's (above) to get
1412 * other information.
1415 ifconf(u_long cmd, caddr_t data, struct thread *td)
1417 struct ifconf *ifc = (struct ifconf *)data;
1420 struct sockaddr *sa;
1421 struct ifreq ifr, *ifrp;
1422 int space = ifc->ifc_len, error = 0;
1424 ifrp = ifc->ifc_req;
1425 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1428 if (space <= sizeof ifr)
1432 * Zero the stack declared structure first to prevent
1433 * memory disclosure.
1435 bzero(&ifr, sizeof(ifr));
1436 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1437 >= sizeof(ifr.ifr_name)) {
1438 error = ENAMETOOLONG;
1443 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1444 if (space <= sizeof ifr)
1447 if (td->td_proc->p_ucred->cr_prison &&
1452 if (cmd == OSIOCGIFCONF) {
1453 struct osockaddr *osa =
1454 (struct osockaddr *)&ifr.ifr_addr;
1456 osa->sa_family = sa->sa_family;
1457 error = copyout(&ifr, ifrp, sizeof ifr);
1461 if (sa->sa_len <= sizeof(*sa)) {
1463 error = copyout(&ifr, ifrp, sizeof ifr);
1466 if (space < (sizeof ifr) + sa->sa_len -
1469 space -= sa->sa_len - sizeof(*sa);
1470 error = copyout(&ifr, ifrp,
1471 sizeof ifr.ifr_name);
1473 error = copyout(sa, &ifrp->ifr_addr,
1475 ifrp = (struct ifreq *)
1476 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1480 space -= sizeof ifr;
1485 bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1486 error = copyout(&ifr, ifrp, sizeof ifr);
1489 space -= sizeof ifr;
1493 ifc->ifc_len -= space;
1498 * Just like if_promisc(), but for all-multicast-reception mode.
1501 if_allmulti(struct ifnet *ifp, int onswitch)
1508 if (ifp->if_amcount++ == 0) {
1509 ifp->if_flags |= IFF_ALLMULTI;
1510 ifr.ifr_flags = ifp->if_flags;
1511 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1512 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1513 (struct ucred *)NULL);
1516 if (ifp->if_amcount > 1) {
1519 ifp->if_amcount = 0;
1520 ifp->if_flags &= ~IFF_ALLMULTI;
1521 ifr.ifr_flags = ifp->if_flags;
1522 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1523 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1524 (struct ucred *)NULL);
1535 * Add a multicast listenership to the interface in question.
1536 * The link layer provides a routine which converts
1540 struct ifnet *ifp, /* interface to manipulate */
1541 struct sockaddr *sa, /* address to add */
1542 struct ifmultiaddr **retifma)
1544 struct sockaddr *llsa, *dupsa;
1546 struct ifmultiaddr *ifma;
1549 * If the matching multicast address already exists
1550 * then don't add a new one, just add a reference
1552 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1553 if (sa_equal(sa, ifma->ifma_addr)) {
1554 ifma->ifma_refcount++;
1562 * Give the link layer a chance to accept/reject it, and also
1563 * find out which AF_LINK address this maps to, if it isn't one
1566 if (ifp->if_resolvemulti) {
1567 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1568 if (error) return error;
1573 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1574 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1575 bcopy(sa, dupsa, sa->sa_len);
1577 ifma->ifma_addr = dupsa;
1578 ifma->ifma_lladdr = llsa;
1579 ifma->ifma_ifp = ifp;
1580 ifma->ifma_refcount = 1;
1581 ifma->ifma_protospec = 0;
1582 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1585 * Some network interfaces can scan the address list at
1586 * interrupt time; lock them out.
1589 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1594 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1595 if (sa_equal(ifma->ifma_addr, llsa))
1599 ifma->ifma_refcount++;
1601 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1602 M_IFMADDR, M_WAITOK);
1603 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1604 M_IFMADDR, M_WAITOK);
1605 bcopy(llsa, dupsa, llsa->sa_len);
1606 ifma->ifma_addr = dupsa;
1607 ifma->ifma_ifp = ifp;
1608 ifma->ifma_refcount = 1;
1610 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1615 * We are certain we have added something, so call down to the
1616 * interface to let them know about it.
1619 ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1626 * Remove a reference to a multicast address on this interface. Yell
1627 * if the request does not match an existing membership.
1630 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1632 struct ifmultiaddr *ifma;
1635 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1636 if (sa_equal(sa, ifma->ifma_addr))
1641 if (ifma->ifma_refcount > 1) {
1642 ifma->ifma_refcount--;
1646 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1647 sa = ifma->ifma_lladdr;
1649 LIST_REMOVE(ifma, ifma_link);
1651 * Make sure the interface driver is notified
1652 * in the case of a link layer mcast group being left.
1654 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1655 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1657 free(ifma->ifma_addr, M_IFMADDR);
1658 free(ifma, M_IFMADDR);
1663 * Now look for the link-layer address which corresponds to
1664 * this network address. It had been squirreled away in
1665 * ifma->ifma_lladdr for this purpose (so we don't have
1666 * to call ifp->if_resolvemulti() again), and we saved that
1667 * value in sa above. If some nasty deleted the
1668 * link-layer address out from underneath us, we can deal because
1669 * the address we stored was is not the same as the one which was
1670 * in the record for the link-layer address. (So we don't complain
1673 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1674 if (sa_equal(sa, ifma->ifma_addr))
1679 if (ifma->ifma_refcount > 1) {
1680 ifma->ifma_refcount--;
1685 LIST_REMOVE(ifma, ifma_link);
1686 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1688 free(ifma->ifma_addr, M_IFMADDR);
1689 free(sa, M_IFMADDR);
1690 free(ifma, M_IFMADDR);
1696 * Set the link layer address on an interface.
1698 * At this time we only support certain types of interfaces,
1699 * and we don't allow the length of the address to change.
1702 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1704 struct sockaddr_dl *sdl;
1708 ifa = ifnet_addrs[ifp->if_index - 1];
1711 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1714 if (len != sdl->sdl_alen) /* don't allow length to change */
1716 switch (ifp->if_type) {
1717 case IFT_ETHER: /* these types use struct arpcom */
1722 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1725 bcopy(lladdr, LLADDR(sdl), len);
1731 * If the interface is already up, we need
1732 * to re-init it in order to reprogram its
1735 if ((ifp->if_flags & IFF_UP) != 0) {
1736 ifp->if_flags &= ~IFF_UP;
1737 ifr.ifr_flags = ifp->if_flags;
1738 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1739 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1740 (struct ucred *)NULL);
1741 ifp->if_flags |= IFF_UP;
1742 ifr.ifr_flags = ifp->if_flags;
1743 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1744 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1745 (struct ucred *)NULL);
1748 * Also send gratuitous ARPs to notify other nodes about
1749 * the address change.
1751 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1752 if (ifa->ifa_addr != NULL &&
1753 ifa->ifa_addr->sa_family == AF_INET)
1754 arp_ifinit(ifp, ifa);
1761 struct ifmultiaddr *
1762 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1764 struct ifmultiaddr *ifma;
1766 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1767 if (sa_equal(ifma->ifma_addr, sa))
1774 * The name argument must be a pointer to storage which will last as
1775 * long as the interface does. For physical devices, the result of
1776 * device_get_name(dev) is a good choice and for pseudo-devices a
1777 * static string works well.
1780 if_initname(struct ifnet *ifp, const char *name, int unit)
1782 ifp->if_dname = name;
1783 ifp->if_dunit = unit;
1784 if (unit != IF_DUNIT_NONE)
1785 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1787 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1791 if_printf(struct ifnet *ifp, const char *fmt, ...)
1796 retval = printf("%s: ", ifp->if_xname);
1797 __va_start(ap, fmt);
1798 retval += vprintf(fmt, ap);
1803 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1804 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1807 ifq_set_classic(struct ifaltq *ifq)
1809 ifq->altq_enqueue = ifq_classic_enqueue;
1810 ifq->altq_dequeue = ifq_classic_dequeue;
1811 ifq->altq_request = ifq_classic_request;
1815 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1816 struct altq_pktattr *pa __unused)
1819 if (IF_QFULL(ifq)) {
1830 static struct mbuf *
1831 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1844 panic("unsupported ALTQ dequeue op: %d", op);
1851 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1859 panic("unspported ALTQ request: %d", req);