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 1.85.2.23 2003/04/15 18:11:19 fjoe Exp $
35 * $DragonFly: src/sys/net/if.c,v 1.11 2003/12/30 03:56:00 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/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/sockio.h>
52 #include <sys/syslog.h>
53 #include <sys/sysctl.h>
56 #include <net/if_arp.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/radix.h>
61 #include <net/route.h>
62 #include <machine/stdarg.h>
64 #if defined(INET) || defined(INET6)
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/if_ether.h>
70 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
76 #if defined(COMPAT_43)
77 #include <emulation/43bsd/43bsd_socket.h>
78 #endif /* COMPAT_43 */
81 * System initialization
84 static int ifconf (u_long, caddr_t);
85 static void ifinit (void *);
86 static void if_qflush (struct ifqueue *);
87 static void if_slowtimo (void *);
88 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
89 static int if_rtdel (struct radix_node *, void *);
91 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
93 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
94 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
95 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
97 int ifqmaxlen = IFQ_MAXLEN;
98 struct ifnethead ifnet; /* depend on static init XXX */
102 * XXX: declare here to avoid to include many inet6 related files..
103 * should be more generalized?
105 extern void nd6_setmtu (struct ifnet *);
108 struct if_clone *if_clone_lookup (const char *, int *);
109 int if_clone_list (struct if_clonereq *);
111 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
112 int if_cloners_count;
115 * Network interface utility routines.
117 * Routines with ifa_ifwith* names take sockaddr *'s as
129 TAILQ_FOREACH(ifp, &ifnet, if_link) {
130 if (ifp->if_snd.ifq_maxlen == 0) {
131 printf("%s%d XXX: driver didn't set ifq_maxlen\n",
132 ifp->if_name, ifp->if_unit);
133 ifp->if_snd.ifq_maxlen = ifqmaxlen;
141 struct ifaddr **ifnet_addrs;
142 struct ifnet **ifindex2ifnet = NULL;
146 * Attach an interface to the
147 * list of "active" interfaces.
153 unsigned socksize, ifasize;
154 int namelen, masklen;
156 struct sockaddr_dl *sdl;
158 static int if_indexlim = 8;
166 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
167 ifp->if_index = ++if_index;
170 * The old code would work if the interface passed a pre-existing
171 * chain of ifaddrs to this code. We don't trust our callers to
172 * properly initialize the tailq, however, so we no longer allow
173 * this unlikely case.
175 TAILQ_INIT(&ifp->if_addrhead);
176 TAILQ_INIT(&ifp->if_prefixhead);
177 LIST_INIT(&ifp->if_multiaddrs);
178 getmicrotime(&ifp->if_lastchange);
179 if (ifnet_addrs == 0 || if_index >= if_indexlim) {
180 unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
181 caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
184 bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
185 free((caddr_t)ifnet_addrs, M_IFADDR);
187 ifnet_addrs = (struct ifaddr **)q;
189 /* grow ifindex2ifnet */
190 n = if_indexlim * sizeof(struct ifnet *);
191 q = malloc(n, M_IFADDR, M_WAITOK);
194 bcopy((caddr_t)ifindex2ifnet, q, n/2);
195 free((caddr_t)ifindex2ifnet, M_IFADDR);
197 ifindex2ifnet = (struct ifnet **)q;
200 ifindex2ifnet[if_index] = ifp;
203 * create a Link Level name for this device
205 namelen = snprintf(workbuf, sizeof(workbuf),
206 "%s%d", ifp->if_name, ifp->if_unit);
207 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
208 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
209 socksize = masklen + ifp->if_addrlen;
210 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
211 if (socksize < sizeof(*sdl))
212 socksize = sizeof(*sdl);
213 socksize = ROUNDUP(socksize);
214 ifasize = sizeof(*ifa) + 2 * socksize;
215 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
217 bzero((caddr_t)ifa, ifasize);
218 sdl = (struct sockaddr_dl *)(ifa + 1);
219 sdl->sdl_len = socksize;
220 sdl->sdl_family = AF_LINK;
221 bcopy(workbuf, sdl->sdl_data, namelen);
222 sdl->sdl_nlen = namelen;
223 sdl->sdl_index = ifp->if_index;
224 sdl->sdl_type = ifp->if_type;
225 ifnet_addrs[if_index - 1] = ifa;
227 ifa->ifa_rtrequest = link_rtrequest;
228 ifa->ifa_addr = (struct sockaddr *)sdl;
229 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
230 ifa->ifa_netmask = (struct sockaddr *)sdl;
231 sdl->sdl_len = masklen;
233 sdl->sdl_data[--namelen] = 0xff;
234 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
237 /* Announce the interface. */
238 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
242 * Detach an interface, removing it from the
243 * list of "active" interfaces.
250 struct radix_node_head *rnh;
255 * Remove routes and flush queues.
261 * Remove address from ifnet_addrs[] and maybe decrement if_index.
262 * Clean up all addresses.
264 ifnet_addrs[ifp->if_index - 1] = 0;
265 while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
268 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
269 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
271 /* XXX: Ugly!! ad hoc just for INET */
272 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
273 struct ifaliasreq ifr;
275 bzero(&ifr, sizeof(ifr));
276 ifr.ifra_addr = *ifa->ifa_addr;
277 if (ifa->ifa_dstaddr)
278 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
279 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
285 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
287 /* ifp_addrhead is already updated */
291 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
297 * Remove all IPv6 kernel structs related to ifp. This should be done
298 * before removing routing entries below, since IPv6 interface direct
299 * routes are expected to be removed by the IPv6-specific kernel API.
300 * Otherwise, the kernel will detect some inconsistency and bark it.
306 * Delete all remaining routes using this interface
307 * Unfortuneatly the only way to do this is to slog through
308 * the entire routing table looking for routes which point
309 * to this interface...oh well...
311 for (i = 1; i <= AF_MAX; i++) {
312 if ((rnh = rt_tables[i]) == NULL)
314 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
317 /* Announce that the interface is gone. */
318 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
320 TAILQ_REMOVE(&ifnet, ifp, if_link);
325 * Delete Routes for a Network Interface
327 * Called for each routing entry via the rnh->rnh_walktree() call above
328 * to delete all route entries referencing a detaching network interface.
331 * rn pointer to node in the routing table
332 * arg argument passed to rnh->rnh_walktree() - detaching interface
336 * errno failed - reason indicated
341 struct radix_node *rn;
344 struct rtentry *rt = (struct rtentry *)rn;
345 struct ifnet *ifp = arg;
348 if (rt->rt_ifp == ifp) {
351 * Protect (sorta) against walktree recursion problems
354 if ((rt->rt_flags & RTF_UP) == 0)
357 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
358 rt_mask(rt), rt->rt_flags,
359 (struct rtentry **) NULL);
361 log(LOG_WARNING, "if_rtdel: error %d\n", err);
369 * Create a clone network interface.
372 if_clone_create(name, len)
376 struct if_clone *ifc;
378 int wildcard, bytoff, bitoff;
382 ifc = if_clone_lookup(name, &unit);
386 if (ifunit(name) != NULL)
390 wildcard = (unit < 0);
392 * Find a free unit if none was given.
395 while ((bytoff < ifc->ifc_bmlen)
396 && (ifc->ifc_units[bytoff] == 0xff))
398 if (bytoff >= ifc->ifc_bmlen)
400 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
402 unit = (bytoff << 3) + bitoff;
405 if (unit > ifc->ifc_maxunit)
408 err = (*ifc->ifc_create)(ifc, unit);
414 bitoff = unit - (bytoff << 3);
418 * Allocate the unit in the bitmap.
420 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
421 ("%s: bit is already set", __func__));
422 ifc->ifc_units[bytoff] |= (1 << bitoff);
424 /* In the wildcard case, we need to update the name. */
426 for (dp = name; *dp != '\0'; dp++);
427 if (snprintf(dp, len - (dp-name), "%d", unit) >
428 len - (dp-name) - 1) {
430 * This can only be a programmer error and
431 * there's no straightforward way to recover if
434 panic("if_clone_create(): interface name too long");
443 * Destroy a clone network interface.
446 if_clone_destroy(name)
449 struct if_clone *ifc;
454 ifc = if_clone_lookup(name, &unit);
458 if (unit < ifc->ifc_minifs)
465 if (ifc->ifc_destroy == NULL)
468 (*ifc->ifc_destroy)(ifp);
471 * Compute offset in the bitmap and deallocate the unit.
474 bitoff = unit - (bytoff << 3);
475 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
476 ("%s: bit is already cleared", __func__));
477 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
482 * Look up a network interface cloner.
485 if_clone_lookup(name, unitp)
489 struct if_clone *ifc;
493 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
494 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
495 if (ifc->ifc_name[i] != *cp)
500 ifc = LIST_NEXT(ifc, ifc_list);
504 return ((struct if_clone *)NULL);
510 for (i = 0; *cp != '\0'; cp++) {
511 if (*cp < '0' || *cp > '9') {
512 /* Bogus unit number. */
515 i = (i * 10) + (*cp - '0');
525 * Register a network interface cloner.
529 struct if_clone *ifc;
536 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
537 ("%s: %s requested more units then allowed (%d > %d)",
538 __func__, ifc->ifc_name, ifc->ifc_minifs,
539 ifc->ifc_maxunit + 1));
541 * Compute bitmap size and allocate it.
543 maxclone = ifc->ifc_maxunit + 1;
545 if ((len << 3) < maxclone)
547 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
548 ifc->ifc_bmlen = len;
550 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
553 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
554 err = (*ifc->ifc_create)(ifc, unit);
556 ("%s: failed to create required interface %s%d",
557 __func__, ifc->ifc_name, unit));
559 /* Allocate the unit in the bitmap. */
561 bitoff = unit - (bytoff << 3);
562 ifc->ifc_units[bytoff] |= (1 << bitoff);
567 * Unregister a network interface cloner.
571 struct if_clone *ifc;
574 LIST_REMOVE(ifc, ifc_list);
575 free(ifc->ifc_units, M_CLONE);
580 * Provide list of interface cloners to userspace.
584 struct if_clonereq *ifcr;
586 char outbuf[IFNAMSIZ], *dst;
587 struct if_clone *ifc;
588 int count, error = 0;
590 ifcr->ifcr_total = if_cloners_count;
591 if ((dst = ifcr->ifcr_buffer) == NULL) {
592 /* Just asking how many there are. */
596 if (ifcr->ifcr_count < 0)
599 count = (if_cloners_count < ifcr->ifcr_count) ?
600 if_cloners_count : ifcr->ifcr_count;
602 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
603 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
604 strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
605 outbuf[IFNAMSIZ - 1] = '\0'; /* sanity */
606 error = copyout(outbuf, dst, IFNAMSIZ);
615 * Locate an interface based on a complete address.
620 struct sockaddr *addr;
625 #define equal(a1, a2) \
626 (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
627 TAILQ_FOREACH(ifp, &ifnet, if_link)
628 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
629 if (ifa->ifa_addr->sa_family != addr->sa_family)
631 if (equal(addr, ifa->ifa_addr))
633 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
634 /* IP6 doesn't have broadcast */
635 ifa->ifa_broadaddr->sa_len != 0 &&
636 equal(ifa->ifa_broadaddr, addr))
639 return ((struct ifaddr *)0);
642 * Locate the point to point interface with a given destination address.
646 ifa_ifwithdstaddr(addr)
647 struct sockaddr *addr;
652 TAILQ_FOREACH(ifp, &ifnet, if_link)
653 if (ifp->if_flags & IFF_POINTOPOINT)
654 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
655 if (ifa->ifa_addr->sa_family != addr->sa_family)
657 if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
660 return ((struct ifaddr *)0);
664 * Find an interface on a specific network. If many, choice
665 * is most specific found.
669 struct sockaddr *addr;
673 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
674 u_int af = addr->sa_family;
675 char *addr_data = addr->sa_data, *cplim;
678 * AF_LINK addresses can be looked up directly by their index number,
679 * so do that if we can.
682 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
683 if (sdl->sdl_index && sdl->sdl_index <= if_index)
684 return (ifnet_addrs[sdl->sdl_index - 1]);
688 * Scan though each interface, looking for ones that have
689 * addresses in this address family.
691 TAILQ_FOREACH(ifp, &ifnet, if_link) {
692 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
693 char *cp, *cp2, *cp3;
695 if (ifa->ifa_addr->sa_family != af)
697 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
699 * This is a bit broken as it doesn't
700 * take into account that the remote end may
701 * be a single node in the network we are
703 * The trouble is that we don't know the
704 * netmask for the remote end.
706 if (ifa->ifa_dstaddr != 0
707 && equal(addr, ifa->ifa_dstaddr))
711 * if we have a special address handler,
712 * then use it instead of the generic one.
714 if (ifa->ifa_claim_addr) {
715 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
723 * Scan all the bits in the ifa's address.
724 * If a bit dissagrees with what we are
725 * looking for, mask it with the netmask
726 * to see if it really matters.
729 if (ifa->ifa_netmask == 0)
732 cp2 = ifa->ifa_addr->sa_data;
733 cp3 = ifa->ifa_netmask->sa_data;
734 cplim = ifa->ifa_netmask->sa_len
735 + (char *)ifa->ifa_netmask;
737 if ((*cp++ ^ *cp2++) & *cp3++)
738 goto next; /* next address! */
740 * If the netmask of what we just found
741 * is more specific than what we had before
742 * (if we had one) then remember the new one
743 * before continuing to search
744 * for an even better one.
746 if (ifa_maybe == 0 ||
747 rn_refines((caddr_t)ifa->ifa_netmask,
748 (caddr_t)ifa_maybe->ifa_netmask))
757 * Find an interface address specific to an interface best matching
761 ifaof_ifpforaddr(addr, ifp)
762 struct sockaddr *addr;
766 char *cp, *cp2, *cp3;
768 struct ifaddr *ifa_maybe = 0;
769 u_int af = addr->sa_family;
773 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
774 if (ifa->ifa_addr->sa_family != af)
778 if (ifa->ifa_netmask == 0) {
779 if (equal(addr, ifa->ifa_addr) ||
780 (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
784 if (ifp->if_flags & IFF_POINTOPOINT) {
785 if (equal(addr, ifa->ifa_dstaddr))
789 cp2 = ifa->ifa_addr->sa_data;
790 cp3 = ifa->ifa_netmask->sa_data;
791 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
792 for (; cp3 < cplim; cp3++)
793 if ((*cp++ ^ *cp2++) & *cp3)
802 #include <net/route.h>
805 * Default action when installing a route with a Link Level gateway.
806 * Lookup an appropriate real ifa to point to.
807 * This should be moved to /sys/net/link.c eventually.
810 link_rtrequest(cmd, rt, info)
813 struct rt_addrinfo *info;
816 struct sockaddr *dst;
819 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
820 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
822 ifa = ifaof_ifpforaddr(dst, ifp);
827 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
828 ifa->ifa_rtrequest(cmd, rt, info);
833 * Mark an interface down and notify protocols of
835 * NOTE: must be called at splnet or eqivalent.
838 if_unroute(ifp, flag, fam)
844 ifp->if_flags &= ~flag;
845 getmicrotime(&ifp->if_lastchange);
846 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
847 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
848 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
849 if_qflush(&ifp->if_snd);
854 * Mark an interface up and notify protocols of
856 * NOTE: must be called at splnet or eqivalent.
859 if_route(ifp, flag, fam)
865 ifp->if_flags |= flag;
866 getmicrotime(&ifp->if_lastchange);
867 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
868 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
869 pfctlinput(PRC_IFUP, ifa->ifa_addr);
877 * Mark an interface down and notify protocols of
879 * NOTE: must be called at splnet or eqivalent.
886 if_unroute(ifp, IFF_UP, AF_UNSPEC);
890 * Mark an interface up and notify protocols of
892 * NOTE: must be called at splnet or eqivalent.
899 if_route(ifp, IFF_UP, AF_UNSPEC);
903 * Flush an interface queue.
912 while ((m = n) != 0) {
922 * Handle interface watchdog timer routines. Called
923 * from softclock, we decrement timers (if set) and
924 * call the appropriate interface routine on expiration.
933 TAILQ_FOREACH(ifp, &ifnet, if_link) {
934 if (ifp->if_timer == 0 || --ifp->if_timer)
936 if (ifp->if_watchdog)
937 (*ifp->if_watchdog)(ifp);
940 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
944 * Map interface name to
945 * interface structure pointer.
948 ifunit(const char *name)
950 char namebuf[IFNAMSIZ + 1];
958 if (len < 2 || len > IFNAMSIZ)
962 if (c < '0' || c > '9')
963 return NULL; /* trailing garbage */
968 return NULL; /* no interface name */
969 unit += (c - '0') * m;
971 return NULL; /* number is unreasonable */
974 } while (c >= '0' && c <= '9');
976 bcopy(name, namebuf, len);
979 * Now search all the interfaces for this name/number
981 TAILQ_FOREACH(ifp, &ifnet, if_link) {
982 if (strcmp(ifp->if_name, namebuf))
984 if (unit == ifp->if_unit)
992 * Map interface name in a sockaddr_dl to
993 * interface structure pointer.
999 char ifname[IFNAMSIZ+1];
1000 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1002 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1003 (sdl->sdl_nlen > IFNAMSIZ) )
1007 * ifunit wants a null-terminated name. It may not be null-terminated
1008 * in the sockaddr. We don't want to change the caller's sockaddr,
1009 * and there might not be room to put the trailing null anyway, so we
1010 * make a local copy that we know we can null terminate safely.
1013 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1014 ifname[sdl->sdl_nlen] = '\0';
1015 return ifunit(ifname);
1023 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1036 return (ifconf(cmd, data));
1038 ifr = (struct ifreq *)data;
1043 if ((error = suser(td)) != 0)
1045 return ((cmd == SIOCIFCREATE) ?
1046 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1047 if_clone_destroy(ifr->ifr_name));
1049 case SIOCIFGCLONERS:
1050 return (if_clone_list((struct if_clonereq *)data));
1053 ifp = ifunit(ifr->ifr_name);
1059 ifr->ifr_flags = ifp->if_flags;
1060 ifr->ifr_flagshigh = ifp->if_ipending >> 16;
1064 ifr->ifr_reqcap = ifp->if_capabilities;
1065 ifr->ifr_curcap = ifp->if_capenable;
1069 ifr->ifr_metric = ifp->if_metric;
1073 ifr->ifr_mtu = ifp->if_mtu;
1077 ifr->ifr_phys = ifp->if_physical;
1084 new_flags = (ifr->ifr_flags & 0xffff) |
1085 (ifr->ifr_flagshigh << 16);
1086 if (ifp->if_flags & IFF_SMART) {
1087 /* Smart drivers twiddle their own routes */
1088 } else if (ifp->if_flags & IFF_UP &&
1089 (new_flags & IFF_UP) == 0) {
1093 } else if (new_flags & IFF_UP &&
1094 (ifp->if_flags & IFF_UP) == 0) {
1099 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1100 (new_flags &~ IFF_CANTCHANGE);
1101 ifp->if_ipending = (ifp->if_ipending & IFF_CANTCHANGE) |
1102 (new_flags &~ IFF_CANTCHANGE);
1103 if (new_flags & IFF_PPROMISC) {
1104 /* Permanently promiscuous mode requested */
1105 ifp->if_flags |= IFF_PROMISC;
1106 } else if (ifp->if_pcount == 0) {
1107 ifp->if_flags &= ~IFF_PROMISC;
1110 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1111 getmicrotime(&ifp->if_lastchange);
1118 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1120 (void) (*ifp->if_ioctl)(ifp, cmd, data);
1127 ifp->if_metric = ifr->ifr_metric;
1128 getmicrotime(&ifp->if_lastchange);
1137 error = (*ifp->if_ioctl)(ifp, cmd, data);
1139 getmicrotime(&ifp->if_lastchange);
1144 u_long oldmtu = ifp->if_mtu;
1149 if (ifp->if_ioctl == NULL)
1150 return (EOPNOTSUPP);
1151 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1153 error = (*ifp->if_ioctl)(ifp, cmd, data);
1155 getmicrotime(&ifp->if_lastchange);
1159 * If the link MTU changed, do network layer specific procedure.
1161 if (ifp->if_mtu != oldmtu) {
1175 /* Don't allow group membership on non-multicast interfaces. */
1176 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1179 /* Don't let users screw up protocols' entries. */
1180 if (ifr->ifr_addr.sa_family != AF_LINK)
1183 if (cmd == SIOCADDMULTI) {
1184 struct ifmultiaddr *ifma;
1185 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1187 error = if_delmulti(ifp, &ifr->ifr_addr);
1190 getmicrotime(&ifp->if_lastchange);
1193 case SIOCSIFPHYADDR:
1194 case SIOCDIFPHYADDR:
1196 case SIOCSIFPHYADDR_IN6:
1198 case SIOCSLIFPHYADDR:
1200 case SIOCSIFGENERIC:
1204 if (ifp->if_ioctl == 0)
1205 return (EOPNOTSUPP);
1206 error = (*ifp->if_ioctl)(ifp, cmd, data);
1208 getmicrotime(&ifp->if_lastchange);
1212 ifs = (struct ifstat *)data;
1213 ifs->ascii[0] = '\0';
1215 case SIOCGIFPSRCADDR:
1216 case SIOCGIFPDSTADDR:
1217 case SIOCGLIFPHYADDR:
1219 case SIOCGIFGENERIC:
1220 if (ifp->if_ioctl == 0)
1221 return (EOPNOTSUPP);
1222 return ((*ifp->if_ioctl)(ifp, cmd, data));
1228 return if_setlladdr(ifp,
1229 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1232 oif_flags = ifp->if_flags;
1233 if (so->so_proto == 0)
1234 return (EOPNOTSUPP);
1236 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1245 case SIOCSIFDSTADDR:
1247 case SIOCSIFBRDADDR:
1248 case SIOCSIFNETMASK:
1249 #if BYTE_ORDER != BIG_ENDIAN
1250 if (ifr->ifr_addr.sa_family == 0 &&
1251 ifr->ifr_addr.sa_len < 16) {
1252 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1253 ifr->ifr_addr.sa_len = 16;
1256 if (ifr->ifr_addr.sa_len == 0)
1257 ifr->ifr_addr.sa_len = 16;
1265 case OSIOCGIFDSTADDR:
1266 cmd = SIOCGIFDSTADDR;
1269 case OSIOCGIFBRDADDR:
1270 cmd = SIOCGIFBRDADDR;
1273 case OSIOCGIFNETMASK:
1274 cmd = SIOCGIFNETMASK;
1276 error = ((*so->so_proto->pr_usrreqs->pru_control)
1277 (so, cmd, data, ifp, td));
1281 case OSIOCGIFDSTADDR:
1282 case OSIOCGIFBRDADDR:
1283 case OSIOCGIFNETMASK:
1284 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1288 #endif /* COMPAT_43 */
1290 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1292 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1293 if (ifp->if_flags & IFF_UP) {
1307 * Set/clear promiscuous mode on interface ifp based on the truth value
1308 * of pswitch. The calls are reference counted so that only the first
1309 * "on" request actually has an effect, as does the final "off" request.
1310 * Results are undefined if the "off" and "on" requests are not matched.
1313 ifpromisc(ifp, pswitch)
1321 oldflags = ifp->if_flags;
1322 if (ifp->if_ipending & IFF_PPROMISC) {
1323 /* Do nothing if device is in permanently promiscuous mode */
1324 ifp->if_pcount += pswitch ? 1 : -1;
1329 * If the device is not configured up, we cannot put it in
1332 if ((ifp->if_flags & IFF_UP) == 0)
1334 if (ifp->if_pcount++ != 0)
1336 ifp->if_flags |= IFF_PROMISC;
1337 log(LOG_INFO, "%s%d: promiscuous mode enabled\n",
1338 ifp->if_name, ifp->if_unit);
1340 if (--ifp->if_pcount > 0)
1342 ifp->if_flags &= ~IFF_PROMISC;
1343 log(LOG_INFO, "%s%d: promiscuous mode disabled\n",
1344 ifp->if_name, ifp->if_unit);
1346 ifr.ifr_flags = ifp->if_flags;
1347 ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1348 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1352 ifp->if_flags = oldflags;
1357 * Return interface configuration
1358 * of system. List may be used
1359 * in later ioctl's (above) to get
1360 * other information.
1364 ifconf(u_long cmd, caddr_t data)
1366 struct ifconf *ifc = (struct ifconf *)data;
1369 struct sockaddr *sa;
1370 struct ifreq ifr, *ifrp;
1371 int space = ifc->ifc_len, error = 0;
1373 ifrp = ifc->ifc_req;
1374 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1378 if (space <= sizeof (ifr))
1380 ifnlen = snprintf(workbuf, sizeof(workbuf),
1381 "%s%d", ifp->if_name, ifp->if_unit);
1382 if(ifnlen + 1 > sizeof ifr.ifr_name) {
1383 error = ENAMETOOLONG;
1386 strcpy(ifr.ifr_name, workbuf);
1390 ifa = ifp->if_addrhead.tqh_first;
1391 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1392 if (space <= sizeof(ifr))
1395 if (curproc->p_ucred->cr_prison && prison_if(curthread, sa))
1399 if (cmd == OSIOCGIFCONF) {
1400 struct osockaddr *osa =
1401 (struct osockaddr *)&ifr.ifr_addr;
1403 osa->sa_family = sa->sa_family;
1404 error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1409 if (sa->sa_len <= sizeof(*sa)) {
1411 error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1415 if (space < sizeof (ifr) + sa->sa_len -
1418 space -= sa->sa_len - sizeof(*sa);
1419 error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1420 sizeof (ifr.ifr_name));
1422 error = copyout((caddr_t)sa,
1423 (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1424 ifrp = (struct ifreq *)
1425 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1429 space -= sizeof (ifr);
1434 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1435 error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1439 space -= sizeof (ifr);
1443 ifc->ifc_len -= space;
1448 * Just like if_promisc(), but for all-multicast-reception mode.
1451 if_allmulti(ifp, onswitch)
1460 if (ifp->if_amcount++ == 0) {
1461 ifp->if_flags |= IFF_ALLMULTI;
1462 ifr.ifr_flags = ifp->if_flags;
1463 ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1464 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1467 if (ifp->if_amcount > 1) {
1470 ifp->if_amcount = 0;
1471 ifp->if_flags &= ~IFF_ALLMULTI;
1472 ifr.ifr_flags = ifp->if_flags;
1473 ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1474 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1485 * Add a multicast listenership to the interface in question.
1486 * The link layer provides a routine which converts
1489 if_addmulti(ifp, sa, retifma)
1490 struct ifnet *ifp; /* interface to manipulate */
1491 struct sockaddr *sa; /* address to add */
1492 struct ifmultiaddr **retifma;
1494 struct sockaddr *llsa, *dupsa;
1496 struct ifmultiaddr *ifma;
1499 * If the matching multicast address already exists
1500 * then don't add a new one, just add a reference
1502 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1503 if (equal(sa, ifma->ifma_addr)) {
1504 ifma->ifma_refcount++;
1512 * Give the link layer a chance to accept/reject it, and also
1513 * find out which AF_LINK address this maps to, if it isn't one
1516 if (ifp->if_resolvemulti) {
1517 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1518 if (error) return error;
1523 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1524 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1525 bcopy(sa, dupsa, sa->sa_len);
1527 ifma->ifma_addr = dupsa;
1528 ifma->ifma_lladdr = llsa;
1529 ifma->ifma_ifp = ifp;
1530 ifma->ifma_refcount = 1;
1531 ifma->ifma_protospec = 0;
1532 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1535 * Some network interfaces can scan the address list at
1536 * interrupt time; lock them out.
1539 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1544 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1545 if (equal(ifma->ifma_addr, llsa))
1549 ifma->ifma_refcount++;
1551 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1552 M_IFMADDR, M_WAITOK);
1553 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1554 M_IFMADDR, M_WAITOK);
1555 bcopy(llsa, dupsa, llsa->sa_len);
1556 ifma->ifma_addr = dupsa;
1557 ifma->ifma_ifp = ifp;
1558 ifma->ifma_refcount = 1;
1560 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1565 * We are certain we have added something, so call down to the
1566 * interface to let them know about it.
1569 ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1576 * Remove a reference to a multicast address on this interface. Yell
1577 * if the request does not match an existing membership.
1580 if_delmulti(ifp, sa)
1582 struct sockaddr *sa;
1584 struct ifmultiaddr *ifma;
1587 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1588 if (equal(sa, ifma->ifma_addr))
1593 if (ifma->ifma_refcount > 1) {
1594 ifma->ifma_refcount--;
1598 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1599 sa = ifma->ifma_lladdr;
1601 LIST_REMOVE(ifma, ifma_link);
1603 * Make sure the interface driver is notified
1604 * in the case of a link layer mcast group being left.
1606 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1607 ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1609 free(ifma->ifma_addr, M_IFMADDR);
1610 free(ifma, M_IFMADDR);
1615 * Now look for the link-layer address which corresponds to
1616 * this network address. It had been squirreled away in
1617 * ifma->ifma_lladdr for this purpose (so we don't have
1618 * to call ifp->if_resolvemulti() again), and we saved that
1619 * value in sa above. If some nasty deleted the
1620 * link-layer address out from underneath us, we can deal because
1621 * the address we stored was is not the same as the one which was
1622 * in the record for the link-layer address. (So we don't complain
1625 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1626 if (equal(sa, ifma->ifma_addr))
1631 if (ifma->ifma_refcount > 1) {
1632 ifma->ifma_refcount--;
1637 LIST_REMOVE(ifma, ifma_link);
1638 ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1640 free(ifma->ifma_addr, M_IFMADDR);
1641 free(sa, M_IFMADDR);
1642 free(ifma, M_IFMADDR);
1648 * Set the link layer address on an interface.
1650 * At this time we only support certain types of interfaces,
1651 * and we don't allow the length of the address to change.
1654 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1656 struct sockaddr_dl *sdl;
1660 ifa = ifnet_addrs[ifp->if_index - 1];
1663 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1666 if (len != sdl->sdl_alen) /* don't allow length to change */
1668 switch (ifp->if_type) {
1669 case IFT_ETHER: /* these types use struct arpcom */
1674 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1677 bcopy(lladdr, LLADDR(sdl), len);
1683 * If the interface is already up, we need
1684 * to re-init it in order to reprogram its
1687 if ((ifp->if_flags & IFF_UP) != 0) {
1688 ifp->if_flags &= ~IFF_UP;
1689 ifr.ifr_flags = ifp->if_flags;
1690 ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1691 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1692 ifp->if_flags |= IFF_UP;
1693 ifr.ifr_flags = ifp->if_flags;
1694 ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1695 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1698 * Also send gratuitous ARPs to notify other nodes about
1699 * the address change.
1701 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1702 if (ifa->ifa_addr != NULL &&
1703 ifa->ifa_addr->sa_family == AF_INET)
1704 arp_ifinit(ifp, ifa);
1711 struct ifmultiaddr *
1712 ifmaof_ifpforaddr(sa, ifp)
1713 struct sockaddr *sa;
1716 struct ifmultiaddr *ifma;
1718 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1719 if (equal(ifma->ifma_addr, sa))
1726 * The name argument must be a pointer to storage which will last as
1727 * long as the interface does. For physical devices, the result of
1728 * device_get_name(dev) is a good choice and for pseudo-devices a
1729 * static string works well.
1732 if_initname(struct ifnet *ifp, const char *name, int unit)
1734 ifp->if_name = name; /* XXX change to dname */
1735 ifp->if_unit = unit; /* XXX change to dunit */
1736 if (unit != IF_DUNIT_NONE)
1737 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1739 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1743 if_printf(struct ifnet *ifp, const char *fmt, ...)
1748 retval = printf("%s%d: ", ifp->if_name, ifp->if_unit);
1749 __va_start(ap, fmt);
1750 retval += vprintf(fmt, ap);
1755 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1756 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");