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 1.39 2005/06/05 12:35:24 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>
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 ifnet **ifindex2ifnet = NULL;
164 * Attach an interface to the
165 * list of "active" interfaces.
168 if_attach(struct ifnet *ifp)
170 unsigned socksize, ifasize;
171 int namelen, masklen;
172 struct sockaddr_dl *sdl;
176 static int if_indexlim = 8;
177 static boolean_t inited;
184 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
185 ifp->if_index = ++if_index;
188 * The old code would work if the interface passed a pre-existing
189 * chain of ifaddrs to this code. We don't trust our callers to
190 * properly initialize the tailq, however, so we no longer allow
191 * this unlikely case.
193 TAILQ_INIT(&ifp->if_addrhead);
194 TAILQ_INIT(&ifp->if_prefixhead);
195 LIST_INIT(&ifp->if_multiaddrs);
196 getmicrotime(&ifp->if_lastchange);
197 if (ifindex2ifnet == NULL || if_index >= if_indexlim) {
203 /* grow ifindex2ifnet */
204 n = if_indexlim * sizeof(*q);
205 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
207 bcopy(ifindex2ifnet, q, n/2);
208 free(ifindex2ifnet, M_IFADDR);
213 ifindex2ifnet[if_index] = ifp;
216 * create a Link Level name for this device
218 namelen = strlen(ifp->if_xname);
219 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
220 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
221 socksize = masklen + ifp->if_addrlen;
222 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
223 if (socksize < sizeof(*sdl))
224 socksize = sizeof(*sdl);
225 socksize = ROUNDUP(socksize);
226 ifasize = sizeof(struct ifaddr) + 2 * socksize;
227 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
228 sdl = (struct sockaddr_dl *)(ifa + 1);
229 sdl->sdl_len = socksize;
230 sdl->sdl_family = AF_LINK;
231 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
232 sdl->sdl_nlen = namelen;
233 sdl->sdl_index = ifp->if_index;
234 sdl->sdl_type = ifp->if_type;
235 ifp->if_lladdr = ifa;
237 ifa->ifa_rtrequest = link_rtrequest;
238 ifa->ifa_addr = (struct sockaddr *)sdl;
239 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
240 ifa->ifa_netmask = (struct sockaddr *)sdl;
241 sdl->sdl_len = masklen;
243 sdl->sdl_data[--namelen] = 0xff;
244 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
246 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
250 ifq->altq_disc = NULL;
251 ifq->altq_flags &= ALTQF_CANTCHANGE;
252 ifq->altq_tbr = NULL;
254 ifq_set_classic(ifq);
256 if (!SLIST_EMPTY(&domains))
257 if_attachdomain1(ifp);
259 /* Announce the interface. */
260 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
264 if_attachdomain(void *dummy)
270 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
271 if_attachdomain1(ifp);
274 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
275 if_attachdomain, NULL);
278 if_attachdomain1(struct ifnet *ifp)
285 /* address family dependent data region */
286 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
287 SLIST_FOREACH(dp, &domains, dom_next)
288 if (dp->dom_ifattach)
289 ifp->if_afdata[dp->dom_family] =
290 (*dp->dom_ifattach)(ifp);
295 * Detach an interface, removing it from the
296 * list of "active" interfaces.
299 if_detach(struct ifnet *ifp)
302 struct radix_node_head *rnh;
307 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
310 * Remove routes and flush queues.
313 #ifdef DEVICE_POLLING
314 if (ifp->if_flags & IFF_POLLING)
315 ether_poll_deregister(ifp);
319 if (ifq_is_enabled(&ifp->if_snd))
320 altq_disable(&ifp->if_snd);
321 if (ifq_is_attached(&ifp->if_snd))
322 altq_detach(&ifp->if_snd);
325 * Clean up all addresses.
327 ifp->if_lladdr = NULL;
329 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
330 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
332 /* XXX: Ugly!! ad hoc just for INET */
333 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
334 struct ifaliasreq ifr;
336 bzero(&ifr, sizeof ifr);
337 ifr.ifra_addr = *ifa->ifa_addr;
338 if (ifa->ifa_dstaddr)
339 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
340 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
346 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
348 /* ifp_addrhead is already updated */
352 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
358 * Remove all IPv4 kernel structures related to ifp.
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]);
394 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
396 ifindex2ifnet[ifp->if_index] = NULL;
397 while (if_index > 0 && ifindex2ifnet[if_index] == NULL)
400 TAILQ_REMOVE(&ifnet, ifp, if_link);
405 * Delete Routes for a Network Interface
407 * Called for each routing entry via the rnh->rnh_walktree() call above
408 * to delete all route entries referencing a detaching network interface.
411 * rn pointer to node in the routing table
412 * arg argument passed to rnh->rnh_walktree() - detaching interface
416 * errno failed - reason indicated
420 if_rtdel(struct radix_node *rn, void *arg)
422 struct rtentry *rt = (struct rtentry *)rn;
423 struct ifnet *ifp = arg;
426 if (rt->rt_ifp == ifp) {
429 * Protect (sorta) against walktree recursion problems
432 if (!(rt->rt_flags & RTF_UP))
435 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
436 rt_mask(rt), rt->rt_flags,
437 (struct rtentry **) NULL);
439 log(LOG_WARNING, "if_rtdel: error %d\n", err);
447 * Create a clone network interface.
450 if_clone_create(char *name, int len)
452 struct if_clone *ifc;
454 int wildcard, bytoff, bitoff;
458 ifc = if_clone_lookup(name, &unit);
462 if (ifunit(name) != NULL)
466 wildcard = (unit < 0);
468 * Find a free unit if none was given.
471 while (bytoff < ifc->ifc_bmlen &&
472 ifc->ifc_units[bytoff] == 0xff)
474 if (bytoff >= ifc->ifc_bmlen)
476 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
478 unit = (bytoff << 3) + bitoff;
481 if (unit > ifc->ifc_maxunit)
484 err = (*ifc->ifc_create)(ifc, unit);
490 bitoff = unit - (bytoff << 3);
494 * Allocate the unit in the bitmap.
496 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
497 ("%s: bit is already set", __func__));
498 ifc->ifc_units[bytoff] |= (1 << bitoff);
500 /* In the wildcard case, we need to update the name. */
502 for (dp = name; *dp != '\0'; dp++);
503 if (snprintf(dp, len - (dp-name), "%d", unit) >
504 len - (dp-name) - 1) {
506 * This can only be a programmer error and
507 * there's no straightforward way to recover if
510 panic("if_clone_create(): interface name too long");
515 EVENTHANDLER_INVOKE(if_clone_event, ifc);
521 * Destroy a clone network interface.
524 if_clone_destroy(const char *name)
526 struct if_clone *ifc;
531 ifc = if_clone_lookup(name, &unit);
535 if (unit < ifc->ifc_minifs)
542 if (ifc->ifc_destroy == NULL)
545 (*ifc->ifc_destroy)(ifp);
548 * Compute offset in the bitmap and deallocate the unit.
551 bitoff = unit - (bytoff << 3);
552 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
553 ("%s: bit is already cleared", __func__));
554 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
559 * Look up a network interface cloner.
562 if_clone_lookup(const char *name, int *unitp)
564 struct if_clone *ifc;
568 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
569 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
570 if (ifc->ifc_name[i] != *cp)
575 ifc = LIST_NEXT(ifc, ifc_list);
579 return ((struct if_clone *)NULL);
585 for (i = 0; *cp != '\0'; cp++) {
586 if (*cp < '0' || *cp > '9') {
587 /* Bogus unit number. */
590 i = (i * 10) + (*cp - '0');
600 * Register a network interface cloner.
603 if_clone_attach(struct if_clone *ifc)
610 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
611 ("%s: %s requested more units then allowed (%d > %d)",
612 __func__, ifc->ifc_name, ifc->ifc_minifs,
613 ifc->ifc_maxunit + 1));
615 * Compute bitmap size and allocate it.
617 maxclone = ifc->ifc_maxunit + 1;
619 if ((len << 3) < maxclone)
621 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
622 ifc->ifc_bmlen = len;
624 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
627 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
628 err = (*ifc->ifc_create)(ifc, unit);
630 ("%s: failed to create required interface %s%d",
631 __func__, ifc->ifc_name, unit));
633 /* Allocate the unit in the bitmap. */
635 bitoff = unit - (bytoff << 3);
636 ifc->ifc_units[bytoff] |= (1 << bitoff);
641 * Unregister a network interface cloner.
644 if_clone_detach(struct if_clone *ifc)
647 LIST_REMOVE(ifc, ifc_list);
648 free(ifc->ifc_units, M_CLONE);
653 * Provide list of interface cloners to userspace.
656 if_clone_list(struct if_clonereq *ifcr)
658 char outbuf[IFNAMSIZ], *dst;
659 struct if_clone *ifc;
660 int count, error = 0;
662 ifcr->ifcr_total = if_cloners_count;
663 if ((dst = ifcr->ifcr_buffer) == NULL) {
664 /* Just asking how many there are. */
668 if (ifcr->ifcr_count < 0)
671 count = (if_cloners_count < ifcr->ifcr_count) ?
672 if_cloners_count : ifcr->ifcr_count;
674 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
675 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
676 strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
677 error = copyout(outbuf, dst, IFNAMSIZ);
686 * Locate an interface based on a complete address.
689 ifa_ifwithaddr(struct sockaddr *addr)
694 TAILQ_FOREACH(ifp, &ifnet, if_link)
695 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
696 if (ifa->ifa_addr->sa_family != addr->sa_family)
698 if (sa_equal(addr, ifa->ifa_addr))
700 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
701 /* IPv6 doesn't have broadcast */
702 ifa->ifa_broadaddr->sa_len != 0 &&
703 sa_equal(ifa->ifa_broadaddr, addr))
706 return ((struct ifaddr *)NULL);
709 * Locate the point to point interface with a given destination address.
712 ifa_ifwithdstaddr(struct sockaddr *addr)
717 TAILQ_FOREACH(ifp, &ifnet, if_link)
718 if (ifp->if_flags & IFF_POINTOPOINT)
719 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
720 if (ifa->ifa_addr->sa_family != addr->sa_family)
722 if (ifa->ifa_dstaddr &&
723 sa_equal(addr, ifa->ifa_dstaddr))
726 return ((struct ifaddr *)NULL);
730 * Find an interface on a specific network. If many, choice
731 * is most specific found.
734 ifa_ifwithnet(struct sockaddr *addr)
738 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
739 u_int af = addr->sa_family;
740 char *addr_data = addr->sa_data, *cplim;
743 * AF_LINK addresses can be looked up directly by their index number,
744 * so do that if we can.
747 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
749 if (sdl->sdl_index && sdl->sdl_index <= if_index)
750 return (ifindex2ifnet[sdl->sdl_index]->if_lladdr);
754 * Scan though each interface, looking for ones that have
755 * addresses in this address family.
757 TAILQ_FOREACH(ifp, &ifnet, if_link) {
758 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
759 char *cp, *cp2, *cp3;
761 if (ifa->ifa_addr->sa_family != af)
763 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
765 * This is a bit broken as it doesn't
766 * take into account that the remote end may
767 * be a single node in the network we are
769 * The trouble is that we don't know the
770 * netmask for the remote end.
772 if (ifa->ifa_dstaddr != NULL &&
773 sa_equal(addr, ifa->ifa_dstaddr))
777 * if we have a special address handler,
778 * then use it instead of the generic one.
780 if (ifa->ifa_claim_addr) {
781 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
789 * Scan all the bits in the ifa's address.
790 * If a bit dissagrees with what we are
791 * looking for, mask it with the netmask
792 * to see if it really matters.
795 if (ifa->ifa_netmask == 0)
798 cp2 = ifa->ifa_addr->sa_data;
799 cp3 = ifa->ifa_netmask->sa_data;
800 cplim = ifa->ifa_netmask->sa_len +
801 (char *)ifa->ifa_netmask;
803 if ((*cp++ ^ *cp2++) & *cp3++)
804 goto next; /* next address! */
806 * If the netmask of what we just found
807 * is more specific than what we had before
808 * (if we had one) then remember the new one
809 * before continuing to search
810 * for an even better one.
812 if (ifa_maybe == 0 ||
813 rn_refines((char *)ifa->ifa_netmask,
814 (char *)ifa_maybe->ifa_netmask))
823 * Find an interface address specific to an interface best matching
827 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
830 char *cp, *cp2, *cp3;
832 struct ifaddr *ifa_maybe = 0;
833 u_int af = addr->sa_family;
837 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
838 if (ifa->ifa_addr->sa_family != af)
842 if (ifa->ifa_netmask == NULL) {
843 if (sa_equal(addr, ifa->ifa_addr) ||
844 (ifa->ifa_dstaddr != NULL &&
845 sa_equal(addr, ifa->ifa_dstaddr)))
849 if (ifp->if_flags & IFF_POINTOPOINT) {
850 if (sa_equal(addr, ifa->ifa_dstaddr))
854 cp2 = ifa->ifa_addr->sa_data;
855 cp3 = ifa->ifa_netmask->sa_data;
856 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
857 for (; cp3 < cplim; cp3++)
858 if ((*cp++ ^ *cp2++) & *cp3)
867 #include <net/route.h>
870 * Default action when installing a route with a Link Level gateway.
871 * Lookup an appropriate real ifa to point to.
872 * This should be moved to /sys/net/link.c eventually.
875 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
878 struct sockaddr *dst;
881 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
882 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
884 ifa = ifaof_ifpforaddr(dst, ifp);
889 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
890 ifa->ifa_rtrequest(cmd, rt, info);
895 * Mark an interface down and notify protocols of
897 * NOTE: must be called at splnet or eqivalent.
900 if_unroute(struct ifnet *ifp, int flag, int fam)
904 ifp->if_flags &= ~flag;
905 getmicrotime(&ifp->if_lastchange);
906 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
907 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
908 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
909 ifq_purge(&ifp->if_snd);
914 * Mark an interface up and notify protocols of
916 * NOTE: must be called at splnet or eqivalent.
919 if_route(struct ifnet *ifp, int flag, int fam)
923 ifp->if_flags |= flag;
924 getmicrotime(&ifp->if_lastchange);
925 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
926 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
927 pfctlinput(PRC_IFUP, ifa->ifa_addr);
935 * Mark an interface down and notify protocols of the transition. An
936 * interface going down is also considered to be a synchronizing event.
937 * We must ensure that all packet processing related to the interface
938 * has completed before we return so e.g. the caller can free the ifnet
939 * structure that the mbufs may be referencing.
941 * NOTE: must be called at splnet or eqivalent.
944 if_down(struct ifnet *ifp)
946 if_unroute(ifp, IFF_UP, AF_UNSPEC);
947 netmsg_service_sync();
951 * Mark an interface up and notify protocols of
953 * NOTE: must be called at splnet or eqivalent.
956 if_up(struct ifnet *ifp)
959 if_route(ifp, IFF_UP, AF_UNSPEC);
963 * Handle interface watchdog timer routines. Called
964 * from softclock, we decrement timers (if set) and
965 * call the appropriate interface routine on expiration.
968 if_slowtimo(void *arg)
973 TAILQ_FOREACH(ifp, &ifnet, if_link) {
974 if (ifp->if_timer == 0 || --ifp->if_timer)
976 if (ifp->if_watchdog)
977 (*ifp->if_watchdog)(ifp);
980 callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
984 * Map interface name to
985 * interface structure pointer.
988 ifunit(const char *name)
993 * Search all the interfaces for this name/number
996 TAILQ_FOREACH(ifp, &ifnet, if_link) {
997 if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1005 * Map interface name in a sockaddr_dl to
1006 * interface structure pointer.
1009 if_withname(struct sockaddr *sa)
1011 char ifname[IFNAMSIZ+1];
1012 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1014 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1015 (sdl->sdl_nlen > IFNAMSIZ) )
1019 * ifunit wants a null-terminated name. It may not be null-terminated
1020 * in the sockaddr. We don't want to change the caller's sockaddr,
1021 * and there might not be room to put the trailing null anyway, so we
1022 * make a local copy that we know we can null terminate safely.
1025 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1026 ifname[sdl->sdl_nlen] = '\0';
1027 return ifunit(ifname);
1035 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1043 size_t namelen, onamelen;
1044 char new_name[IFNAMSIZ];
1046 struct sockaddr_dl *sdl;
1052 return (ifconf(cmd, data, td));
1054 ifr = (struct ifreq *)data;
1059 if ((error = suser(td)) != 0)
1061 return ((cmd == SIOCIFCREATE) ?
1062 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1063 if_clone_destroy(ifr->ifr_name));
1065 case SIOCIFGCLONERS:
1066 return (if_clone_list((struct if_clonereq *)data));
1069 ifp = ifunit(ifr->ifr_name);
1075 ifr->ifr_flags = ifp->if_flags;
1076 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1080 ifr->ifr_reqcap = ifp->if_capabilities;
1081 ifr->ifr_curcap = ifp->if_capenable;
1085 ifr->ifr_metric = ifp->if_metric;
1089 ifr->ifr_mtu = ifp->if_mtu;
1093 ifr->ifr_phys = ifp->if_physical;
1100 new_flags = (ifr->ifr_flags & 0xffff) |
1101 (ifr->ifr_flagshigh << 16);
1102 if (ifp->if_flags & IFF_SMART) {
1103 /* Smart drivers twiddle their own routes */
1104 } else if (ifp->if_flags & IFF_UP &&
1105 (new_flags & IFF_UP) == 0) {
1109 } else if (new_flags & IFF_UP &&
1110 (ifp->if_flags & IFF_UP) == 0) {
1116 #ifdef DEVICE_POLLING
1117 if ((new_flags ^ ifp->if_flags) & IFF_POLLING) {
1118 if (new_flags & IFF_POLLING) {
1119 ether_poll_register(ifp);
1121 ether_poll_deregister(ifp);
1126 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1127 (new_flags &~ IFF_CANTCHANGE);
1128 if (new_flags & IFF_PPROMISC) {
1129 /* Permanently promiscuous mode requested */
1130 ifp->if_flags |= IFF_PROMISC;
1131 } else if (ifp->if_pcount == 0) {
1132 ifp->if_flags &= ~IFF_PROMISC;
1135 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1136 getmicrotime(&ifp->if_lastchange);
1143 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1145 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1152 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1155 if (new_name[0] == '\0')
1157 if (ifunit(new_name) != NULL)
1160 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1162 /* Announce the departure of the interface. */
1163 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1165 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1166 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1167 /* XXX IFA_LOCK(ifa); */
1168 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1169 namelen = strlen(new_name);
1170 onamelen = sdl->sdl_nlen;
1172 * Move the address if needed. This is safe because we
1173 * allocate space for a name of length IFNAMSIZ when we
1174 * create this in if_attach().
1176 if (namelen != onamelen) {
1177 bcopy(sdl->sdl_data + onamelen,
1178 sdl->sdl_data + namelen, sdl->sdl_alen);
1180 bcopy(new_name, sdl->sdl_data, namelen);
1181 sdl->sdl_nlen = namelen;
1182 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1183 bzero(sdl->sdl_data, onamelen);
1184 while (namelen != 0)
1185 sdl->sdl_data[--namelen] = 0xff;
1186 /* XXX IFA_UNLOCK(ifa) */
1188 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1190 /* Announce the return of the interface. */
1191 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1198 ifp->if_metric = ifr->ifr_metric;
1199 getmicrotime(&ifp->if_lastchange);
1208 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1210 getmicrotime(&ifp->if_lastchange);
1215 u_long oldmtu = ifp->if_mtu;
1220 if (ifp->if_ioctl == NULL)
1221 return (EOPNOTSUPP);
1222 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1224 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1226 getmicrotime(&ifp->if_lastchange);
1230 * If the link MTU changed, do network layer specific procedure.
1232 if (ifp->if_mtu != oldmtu) {
1246 /* Don't allow group membership on non-multicast interfaces. */
1247 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1250 /* Don't let users screw up protocols' entries. */
1251 if (ifr->ifr_addr.sa_family != AF_LINK)
1254 if (cmd == SIOCADDMULTI) {
1255 struct ifmultiaddr *ifma;
1256 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1258 error = if_delmulti(ifp, &ifr->ifr_addr);
1261 getmicrotime(&ifp->if_lastchange);
1264 case SIOCSIFPHYADDR:
1265 case SIOCDIFPHYADDR:
1267 case SIOCSIFPHYADDR_IN6:
1269 case SIOCSLIFPHYADDR:
1271 case SIOCSIFGENERIC:
1275 if (ifp->if_ioctl == 0)
1276 return (EOPNOTSUPP);
1277 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1279 getmicrotime(&ifp->if_lastchange);
1283 ifs = (struct ifstat *)data;
1284 ifs->ascii[0] = '\0';
1286 case SIOCGIFPSRCADDR:
1287 case SIOCGIFPDSTADDR:
1288 case SIOCGLIFPHYADDR:
1290 case SIOCGIFGENERIC:
1291 if (ifp->if_ioctl == 0)
1292 return (EOPNOTSUPP);
1293 return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1299 return if_setlladdr(ifp,
1300 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1303 oif_flags = ifp->if_flags;
1304 if (so->so_proto == 0)
1305 return (EOPNOTSUPP);
1307 error = so_pru_control(so, cmd, data, ifp, td);
1314 case SIOCSIFDSTADDR:
1316 case SIOCSIFBRDADDR:
1317 case SIOCSIFNETMASK:
1318 #if BYTE_ORDER != BIG_ENDIAN
1319 if (ifr->ifr_addr.sa_family == 0 &&
1320 ifr->ifr_addr.sa_len < 16) {
1321 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1322 ifr->ifr_addr.sa_len = 16;
1325 if (ifr->ifr_addr.sa_len == 0)
1326 ifr->ifr_addr.sa_len = 16;
1334 case OSIOCGIFDSTADDR:
1335 cmd = SIOCGIFDSTADDR;
1338 case OSIOCGIFBRDADDR:
1339 cmd = SIOCGIFBRDADDR;
1342 case OSIOCGIFNETMASK:
1343 cmd = SIOCGIFNETMASK;
1345 error = so_pru_control(so, cmd, data, ifp, td);
1349 case OSIOCGIFDSTADDR:
1350 case OSIOCGIFBRDADDR:
1351 case OSIOCGIFNETMASK:
1352 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1356 #endif /* COMPAT_43 */
1358 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1360 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1361 if (ifp->if_flags & IFF_UP) {
1375 * Set/clear promiscuous mode on interface ifp based on the truth value
1376 * of pswitch. The calls are reference counted so that only the first
1377 * "on" request actually has an effect, as does the final "off" request.
1378 * Results are undefined if the "off" and "on" requests are not matched.
1381 ifpromisc(struct ifnet *ifp, int pswitch)
1387 oldflags = ifp->if_flags;
1388 if (ifp->if_flags & IFF_PPROMISC) {
1389 /* Do nothing if device is in permanently promiscuous mode */
1390 ifp->if_pcount += pswitch ? 1 : -1;
1395 * If the device is not configured up, we cannot put it in
1398 if ((ifp->if_flags & IFF_UP) == 0)
1400 if (ifp->if_pcount++ != 0)
1402 ifp->if_flags |= IFF_PROMISC;
1403 log(LOG_INFO, "%s: promiscuous mode enabled\n",
1406 if (--ifp->if_pcount > 0)
1408 ifp->if_flags &= ~IFF_PROMISC;
1409 log(LOG_INFO, "%s: promiscuous mode disabled\n",
1412 ifr.ifr_flags = ifp->if_flags;
1413 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1414 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1415 (struct ucred *)NULL);
1419 ifp->if_flags = oldflags;
1424 * Return interface configuration
1425 * of system. List may be used
1426 * in later ioctl's (above) to get
1427 * other information.
1430 ifconf(u_long cmd, caddr_t data, struct thread *td)
1432 struct ifconf *ifc = (struct ifconf *)data;
1435 struct sockaddr *sa;
1436 struct ifreq ifr, *ifrp;
1437 int space = ifc->ifc_len, error = 0;
1439 ifrp = ifc->ifc_req;
1440 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1443 if (space <= sizeof ifr)
1447 * Zero the stack declared structure first to prevent
1448 * memory disclosure.
1450 bzero(&ifr, sizeof(ifr));
1451 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1452 >= sizeof(ifr.ifr_name)) {
1453 error = ENAMETOOLONG;
1458 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1459 if (space <= sizeof ifr)
1462 if (td->td_proc->p_ucred->cr_prison &&
1467 if (cmd == OSIOCGIFCONF) {
1468 struct osockaddr *osa =
1469 (struct osockaddr *)&ifr.ifr_addr;
1471 osa->sa_family = sa->sa_family;
1472 error = copyout(&ifr, ifrp, sizeof ifr);
1476 if (sa->sa_len <= sizeof(*sa)) {
1478 error = copyout(&ifr, ifrp, sizeof ifr);
1481 if (space < (sizeof ifr) + sa->sa_len -
1484 space -= sa->sa_len - sizeof(*sa);
1485 error = copyout(&ifr, ifrp,
1486 sizeof ifr.ifr_name);
1488 error = copyout(sa, &ifrp->ifr_addr,
1490 ifrp = (struct ifreq *)
1491 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1495 space -= sizeof ifr;
1500 bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1501 error = copyout(&ifr, ifrp, sizeof ifr);
1504 space -= sizeof ifr;
1508 ifc->ifc_len -= space;
1513 * Just like if_promisc(), but for all-multicast-reception mode.
1516 if_allmulti(struct ifnet *ifp, int onswitch)
1523 if (ifp->if_amcount++ == 0) {
1524 ifp->if_flags |= IFF_ALLMULTI;
1525 ifr.ifr_flags = ifp->if_flags;
1526 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1527 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1528 (struct ucred *)NULL);
1531 if (ifp->if_amcount > 1) {
1534 ifp->if_amcount = 0;
1535 ifp->if_flags &= ~IFF_ALLMULTI;
1536 ifr.ifr_flags = ifp->if_flags;
1537 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1538 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1539 (struct ucred *)NULL);
1550 * Add a multicast listenership to the interface in question.
1551 * The link layer provides a routine which converts
1555 struct ifnet *ifp, /* interface to manipulate */
1556 struct sockaddr *sa, /* address to add */
1557 struct ifmultiaddr **retifma)
1559 struct sockaddr *llsa, *dupsa;
1561 struct ifmultiaddr *ifma;
1564 * If the matching multicast address already exists
1565 * then don't add a new one, just add a reference
1567 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1568 if (sa_equal(sa, ifma->ifma_addr)) {
1569 ifma->ifma_refcount++;
1577 * Give the link layer a chance to accept/reject it, and also
1578 * find out which AF_LINK address this maps to, if it isn't one
1581 if (ifp->if_resolvemulti) {
1582 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1583 if (error) return error;
1588 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1589 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1590 bcopy(sa, dupsa, sa->sa_len);
1592 ifma->ifma_addr = dupsa;
1593 ifma->ifma_lladdr = llsa;
1594 ifma->ifma_ifp = ifp;
1595 ifma->ifma_refcount = 1;
1596 ifma->ifma_protospec = 0;
1597 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1600 * Some network interfaces can scan the address list at
1601 * interrupt time; lock them out.
1604 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1609 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1610 if (sa_equal(ifma->ifma_addr, llsa))
1614 ifma->ifma_refcount++;
1616 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1617 M_IFMADDR, M_WAITOK);
1618 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1619 M_IFMADDR, M_WAITOK);
1620 bcopy(llsa, dupsa, llsa->sa_len);
1621 ifma->ifma_addr = dupsa;
1622 ifma->ifma_ifp = ifp;
1623 ifma->ifma_refcount = 1;
1625 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1630 * We are certain we have added something, so call down to the
1631 * interface to let them know about it.
1634 ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1641 * Remove a reference to a multicast address on this interface. Yell
1642 * if the request does not match an existing membership.
1645 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1647 struct ifmultiaddr *ifma;
1650 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1651 if (sa_equal(sa, ifma->ifma_addr))
1656 if (ifma->ifma_refcount > 1) {
1657 ifma->ifma_refcount--;
1661 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1662 sa = ifma->ifma_lladdr;
1664 LIST_REMOVE(ifma, ifma_link);
1666 * Make sure the interface driver is notified
1667 * in the case of a link layer mcast group being left.
1669 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1670 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1672 free(ifma->ifma_addr, M_IFMADDR);
1673 free(ifma, M_IFMADDR);
1678 * Now look for the link-layer address which corresponds to
1679 * this network address. It had been squirreled away in
1680 * ifma->ifma_lladdr for this purpose (so we don't have
1681 * to call ifp->if_resolvemulti() again), and we saved that
1682 * value in sa above. If some nasty deleted the
1683 * link-layer address out from underneath us, we can deal because
1684 * the address we stored was is not the same as the one which was
1685 * in the record for the link-layer address. (So we don't complain
1688 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1689 if (sa_equal(sa, ifma->ifma_addr))
1694 if (ifma->ifma_refcount > 1) {
1695 ifma->ifma_refcount--;
1700 LIST_REMOVE(ifma, ifma_link);
1701 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1703 free(ifma->ifma_addr, M_IFMADDR);
1704 free(sa, M_IFMADDR);
1705 free(ifma, M_IFMADDR);
1711 * Set the link layer address on an interface.
1713 * At this time we only support certain types of interfaces,
1714 * and we don't allow the length of the address to change.
1717 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1719 struct sockaddr_dl *sdl;
1723 sdl = IF_LLSOCKADDR(ifp);
1726 if (len != sdl->sdl_alen) /* don't allow length to change */
1728 switch (ifp->if_type) {
1729 case IFT_ETHER: /* these types use struct arpcom */
1734 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1737 bcopy(lladdr, LLADDR(sdl), len);
1743 * If the interface is already up, we need
1744 * to re-init it in order to reprogram its
1747 if ((ifp->if_flags & IFF_UP) != 0) {
1748 ifp->if_flags &= ~IFF_UP;
1749 ifr.ifr_flags = ifp->if_flags;
1750 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1751 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1752 (struct ucred *)NULL);
1753 ifp->if_flags |= IFF_UP;
1754 ifr.ifr_flags = ifp->if_flags;
1755 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1756 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1757 (struct ucred *)NULL);
1760 * Also send gratuitous ARPs to notify other nodes about
1761 * the address change.
1763 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1764 if (ifa->ifa_addr != NULL &&
1765 ifa->ifa_addr->sa_family == AF_INET)
1766 arp_ifinit(ifp, ifa);
1773 struct ifmultiaddr *
1774 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1776 struct ifmultiaddr *ifma;
1778 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1779 if (sa_equal(ifma->ifma_addr, sa))
1786 * The name argument must be a pointer to storage which will last as
1787 * long as the interface does. For physical devices, the result of
1788 * device_get_name(dev) is a good choice and for pseudo-devices a
1789 * static string works well.
1792 if_initname(struct ifnet *ifp, const char *name, int unit)
1794 ifp->if_dname = name;
1795 ifp->if_dunit = unit;
1796 if (unit != IF_DUNIT_NONE)
1797 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1799 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1803 if_printf(struct ifnet *ifp, const char *fmt, ...)
1808 retval = printf("%s: ", ifp->if_xname);
1809 __va_start(ap, fmt);
1810 retval += vprintf(fmt, ap);
1815 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1816 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1819 ifq_set_classic(struct ifaltq *ifq)
1821 ifq->altq_enqueue = ifq_classic_enqueue;
1822 ifq->altq_dequeue = ifq_classic_dequeue;
1823 ifq->altq_request = ifq_classic_request;
1827 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1828 struct altq_pktattr *pa __unused)
1831 if (IF_QFULL(ifq)) {
1842 static struct mbuf *
1843 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1856 panic("unsupported ALTQ dequeue op: %d", op);
1863 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1871 panic("unspported ALTQ request: %d", req);