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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
40 #include <sys/malloc.h>
44 #include <sys/systm.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
52 #include <sys/refcount.h>
53 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/sockio.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/taskqueue.h>
59 #include <sys/domain.h>
63 #include <machine/stdarg.h>
67 #include <net/if_arp.h>
68 #include <net/if_clone.h>
69 #include <net/if_dl.h>
70 #include <net/if_types.h>
71 #include <net/if_var.h>
72 #include <net/radix.h>
73 #include <net/route.h>
76 #if defined(INET) || defined(INET6)
77 #include <net/ethernet.h>
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip.h>
81 #include <netinet/ip_carp.h>
83 #include <netinet/if_ether.h>
86 #include <netinet6/in6_var.h>
87 #include <netinet6/in6_ifattach.h>
89 #endif /* INET || INET6 */
91 #include <security/mac/mac_framework.h>
93 #ifdef COMPAT_FREEBSD32
94 #include <sys/mount.h>
95 #include <compat/freebsd32/freebsd32.h>
98 struct ifindex_entry {
99 struct ifnet *ife_ifnet;
102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
105 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
106 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
107 &ifqmaxlen, 0, "max send queue size");
109 /* Log link state change events */
110 static int log_link_state_change = 1;
112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
113 &log_link_state_change, 0,
114 "log interface link state change events");
116 /* Log promiscuous mode change events */
117 static int log_promisc_mode_change = 1;
119 TUNABLE_INT("net.link.log_promisc_mode_change", &log_promisc_mode_change);
120 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
121 &log_promisc_mode_change, 1,
122 "log promiscuous mode change events");
124 /* Interface description */
125 static unsigned int ifdescr_maxlen = 1024;
126 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
128 "administrative maximum length for interface description");
130 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
132 /* global sx for non-critical path ifdescr */
133 static struct sx ifdescr_sx;
134 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
136 void (*bridge_linkstate_p)(struct ifnet *ifp);
137 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
138 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
139 /* These are external hooks for CARP. */
140 void (*carp_linkstate_p)(struct ifnet *ifp);
141 void (*carp_demote_adj_p)(int, char *);
142 int (*carp_master_p)(struct ifaddr *);
143 #if defined(INET) || defined(INET6)
144 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
145 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
146 const struct sockaddr *sa);
147 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
148 int (*carp_attach_p)(struct ifaddr *, int);
149 void (*carp_detach_p)(struct ifaddr *);
152 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
155 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
156 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
157 const struct in6_addr *taddr);
160 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
163 * XXX: Style; these should be sorted alphabetically, and unprototyped
164 * static functions should be prototyped. Currently they are sorted by
167 static void if_attachdomain(void *);
168 static void if_attachdomain1(struct ifnet *);
169 static int ifconf(u_long, caddr_t);
170 static void if_freemulti(struct ifmultiaddr *);
171 static void if_init(void *);
172 static void if_grow(void);
173 static void if_input_default(struct ifnet *, struct mbuf *);
174 static void if_route(struct ifnet *, int flag, int fam);
175 static int if_setflag(struct ifnet *, int, int, int *, int);
176 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
177 static void if_unroute(struct ifnet *, int flag, int fam);
178 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
179 static int if_rtdel(struct radix_node *, void *);
180 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
181 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
182 static void do_link_state_change(void *, int);
183 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
184 static int if_getgroupmembers(struct ifgroupreq *);
185 static void if_delgroups(struct ifnet *);
186 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
187 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
191 * XXX: declare here to avoid to include many inet6 related files..
192 * should be more generalized?
194 extern void nd6_setmtu(struct ifnet *);
197 VNET_DEFINE(int, if_index);
198 int ifqmaxlen = IFQ_MAXLEN;
199 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
200 VNET_DEFINE(struct ifgrouphead, ifg_head);
202 static VNET_DEFINE(int, if_indexlim) = 8;
204 /* Table of ifnet by index. */
205 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
207 #define V_if_indexlim VNET(if_indexlim)
208 #define V_ifindex_table VNET(ifindex_table)
211 * The global network interface list (V_ifnet) and related state (such as
212 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
213 * an rwlock. Either may be acquired shared to stablize the list, but both
214 * must be acquired writable to modify the list. This model allows us to
215 * both stablize the interface list during interrupt thread processing, but
216 * also to stablize it over long-running ioctls, without introducing priority
217 * inversions and deadlocks.
219 struct rwlock ifnet_rwlock;
220 struct sx ifnet_sxlock;
223 * The allocation of network interfaces is a rather non-atomic affair; we
224 * need to select an index before we are ready to expose the interface for
225 * use, so will use this pointer value to indicate reservation.
227 #define IFNET_HOLD (void *)(uintptr_t)(-1)
229 static if_com_alloc_t *if_com_alloc[256];
230 static if_com_free_t *if_com_free[256];
232 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
233 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
234 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
237 ifnet_byindex_locked(u_short idx)
240 if (idx > V_if_index)
242 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
244 return (V_ifindex_table[idx].ife_ifnet);
248 ifnet_byindex(u_short idx)
252 IFNET_RLOCK_NOSLEEP();
253 ifp = ifnet_byindex_locked(idx);
254 IFNET_RUNLOCK_NOSLEEP();
259 ifnet_byindex_ref(u_short idx)
263 IFNET_RLOCK_NOSLEEP();
264 ifp = ifnet_byindex_locked(idx);
265 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
266 IFNET_RUNLOCK_NOSLEEP();
270 IFNET_RUNLOCK_NOSLEEP();
275 * Allocate an ifindex array entry; return 0 on success or an error on
279 ifindex_alloc_locked(u_short *idxp)
283 IFNET_WLOCK_ASSERT();
287 * Try to find an empty slot below V_if_index. If we fail, take the
290 for (idx = 1; idx <= V_if_index; idx++) {
291 if (V_ifindex_table[idx].ife_ifnet == NULL)
295 /* Catch if_index overflow. */
298 if (idx >= V_if_indexlim) {
302 if (idx > V_if_index)
309 ifindex_free_locked(u_short idx)
312 IFNET_WLOCK_ASSERT();
314 V_ifindex_table[idx].ife_ifnet = NULL;
315 while (V_if_index > 0 &&
316 V_ifindex_table[V_if_index].ife_ifnet == NULL)
321 ifindex_free(u_short idx)
325 ifindex_free_locked(idx);
330 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
333 IFNET_WLOCK_ASSERT();
335 V_ifindex_table[idx].ife_ifnet = ifp;
339 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
343 ifnet_setbyindex_locked(idx, ifp);
348 ifaddr_byindex(u_short idx)
351 struct ifaddr *ifa = NULL;
353 IFNET_RLOCK_NOSLEEP();
354 ifp = ifnet_byindex_locked(idx);
355 if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
357 IFNET_RUNLOCK_NOSLEEP();
362 * Network interface utility routines.
364 * Routines with ifa_ifwith* names take sockaddr *'s as
369 vnet_if_init(const void *unused __unused)
372 TAILQ_INIT(&V_ifnet);
373 TAILQ_INIT(&V_ifg_head);
375 if_grow(); /* create initial table */
377 vnet_if_clone_init();
379 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
384 if_init(void *dummy __unused)
390 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
395 vnet_if_uninit(const void *unused __unused)
398 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
399 "not empty", __func__, __LINE__, &V_ifnet));
400 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
401 "not empty", __func__, __LINE__, &V_ifg_head));
403 free((caddr_t)V_ifindex_table, M_IFNET);
405 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
406 vnet_if_uninit, NULL);
414 struct ifindex_entry *e;
416 IFNET_WLOCK_ASSERT();
417 oldlim = V_if_indexlim;
419 n = (oldlim << 1) * sizeof(*e);
420 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
422 if (V_if_indexlim != oldlim) {
426 if (V_ifindex_table != NULL) {
427 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
428 free((caddr_t)V_ifindex_table, M_IFNET);
435 * Allocate a struct ifnet and an index for an interface. A layer 2
436 * common structure will also be allocated if an allocation routine is
437 * registered for the passed type.
440 if_alloc(u_char type)
445 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
447 if (ifindex_alloc_locked(&idx) != 0) {
452 ifnet_setbyindex_locked(idx, IFNET_HOLD);
456 ifp->if_alloctype = type;
457 if (if_com_alloc[type] != NULL) {
458 ifp->if_l2com = if_com_alloc[type](type, ifp);
459 if (ifp->if_l2com == NULL) {
466 IF_ADDR_LOCK_INIT(ifp);
467 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
468 ifp->if_afdata_initialized = 0;
469 IF_AFDATA_LOCK_INIT(ifp);
470 TAILQ_INIT(&ifp->if_addrhead);
471 TAILQ_INIT(&ifp->if_multiaddrs);
472 TAILQ_INIT(&ifp->if_groups);
476 ifq_init(&ifp->if_snd, ifp);
478 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
479 ifnet_setbyindex(ifp->if_index, ifp);
484 * Do the actual work of freeing a struct ifnet, and layer 2 common
485 * structure. This call is made when the last reference to an
486 * interface is released.
489 if_free_internal(struct ifnet *ifp)
492 KASSERT((ifp->if_flags & IFF_DYING),
493 ("if_free_internal: interface not dying"));
495 if (if_com_free[ifp->if_alloctype] != NULL)
496 if_com_free[ifp->if_alloctype](ifp->if_l2com,
500 mac_ifnet_destroy(ifp);
502 if (ifp->if_description != NULL)
503 free(ifp->if_description, M_IFDESCR);
504 IF_AFDATA_DESTROY(ifp);
505 IF_ADDR_LOCK_DESTROY(ifp);
506 ifq_delete(&ifp->if_snd);
511 * Deregister an interface and free the associated storage.
514 if_free(struct ifnet *ifp)
517 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
519 CURVNET_SET_QUIET(ifp->if_vnet);
521 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
522 ("%s: freeing unallocated ifnet", ifp->if_xname));
524 ifindex_free_locked(ifp->if_index);
527 if (refcount_release(&ifp->if_refcount))
528 if_free_internal(ifp);
533 * Interfaces to keep an ifnet type-stable despite the possibility of the
534 * driver calling if_free(). If there are additional references, we defer
535 * freeing the underlying data structure.
538 if_ref(struct ifnet *ifp)
541 /* We don't assert the ifnet list lock here, but arguably should. */
542 refcount_acquire(&ifp->if_refcount);
546 if_rele(struct ifnet *ifp)
549 if (!refcount_release(&ifp->if_refcount))
551 if_free_internal(ifp);
555 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
558 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
560 if (ifq->ifq_maxlen == 0)
561 ifq->ifq_maxlen = ifqmaxlen;
564 ifq->altq_disc = NULL;
565 ifq->altq_flags &= ALTQF_CANTCHANGE;
566 ifq->altq_tbr = NULL;
571 ifq_delete(struct ifaltq *ifq)
573 mtx_destroy(&ifq->ifq_mtx);
577 * Perform generic interface initalization tasks and attach the interface
578 * to the list of "active" interfaces. If vmove flag is set on entry
579 * to if_attach_internal(), perform only a limited subset of initialization
580 * tasks, given that we are moving from one vnet to another an ifnet which
581 * has already been fully initialized.
583 * Note that if_detach_internal() removes group membership unconditionally
584 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
585 * Thus, when if_vmove() is applied to a cloned interface, group membership
586 * is lost while a cloned one always joins a group whose name is
587 * ifc->ifc_name. To recover this after if_detach_internal() and
588 * if_attach_internal(), the cloner should be specified to
589 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
590 * attempts to join a group whose name is ifc->ifc_name.
593 * - The decision to return void and thus require this function to
594 * succeed is questionable.
595 * - We should probably do more sanity checking. For instance we don't
596 * do anything to insure if_xname is unique or non-empty.
599 if_attach(struct ifnet *ifp)
602 if_attach_internal(ifp, 0, NULL);
606 * Compute the least common TSO limit.
609 if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
612 * 1) If there is no limit currently, take the limit from
613 * the network adapter.
615 * 2) If the network adapter has a limit below the current
618 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
619 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
620 pmax->tsomaxbytes = ifp->if_hw_tsomax;
622 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
623 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
624 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
626 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
627 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
628 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
633 * Update TSO limit of a network adapter.
635 * Returns zero if no change. Else non-zero.
638 if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
641 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
642 ifp->if_hw_tsomax = pmax->tsomaxbytes;
645 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
646 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
649 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
650 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
657 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
659 unsigned socksize, ifasize;
660 int namelen, masklen;
661 struct sockaddr_dl *sdl;
664 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
665 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
669 ifp->if_vnet = curvnet;
670 if (ifp->if_home_vnet == NULL)
671 ifp->if_home_vnet = curvnet;
674 if_addgroup(ifp, IFG_ALL);
676 /* Restore group membership for cloned interfaces. */
677 if (vmove && ifc != NULL)
678 if_clone_addgroup(ifp, ifc);
680 getmicrotime(&ifp->if_lastchange);
681 ifp->if_data.ifi_epoch = time_uptime;
682 ifp->if_data.ifi_datalen = sizeof(struct if_data);
684 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
685 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
686 ("transmit and qflush must both either be set or both be NULL"));
687 if (ifp->if_transmit == NULL) {
688 ifp->if_transmit = if_transmit;
689 ifp->if_qflush = if_qflush;
691 if (ifp->if_input == NULL)
692 ifp->if_input = if_input_default;
696 mac_ifnet_create(ifp);
700 * Create a Link Level name for this device.
702 namelen = strlen(ifp->if_xname);
704 * Always save enough space for any possiable name so we
705 * can do a rename in place later.
707 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
708 socksize = masklen + ifp->if_addrlen;
709 if (socksize < sizeof(*sdl))
710 socksize = sizeof(*sdl);
711 socksize = roundup2(socksize, sizeof(long));
712 ifasize = sizeof(*ifa) + 2 * socksize;
713 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
715 sdl = (struct sockaddr_dl *)(ifa + 1);
716 sdl->sdl_len = socksize;
717 sdl->sdl_family = AF_LINK;
718 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
719 sdl->sdl_nlen = namelen;
720 sdl->sdl_index = ifp->if_index;
721 sdl->sdl_type = ifp->if_type;
724 ifa->ifa_rtrequest = link_rtrequest;
725 ifa->ifa_addr = (struct sockaddr *)sdl;
726 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
727 ifa->ifa_netmask = (struct sockaddr *)sdl;
728 sdl->sdl_len = masklen;
730 sdl->sdl_data[--namelen] = 0xff;
731 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
732 /* Reliably crash if used uninitialized. */
733 ifp->if_broadcastaddr = NULL;
735 if (ifp->if_type == IFT_ETHER) {
736 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
740 #if defined(INET) || defined(INET6)
741 /* Use defaults for TSO, if nothing is set */
742 if (ifp->if_hw_tsomax == 0 &&
743 ifp->if_hw_tsomaxsegcount == 0 &&
744 ifp->if_hw_tsomaxsegsize == 0) {
746 * The TSO defaults needs to be such that an
747 * NFS mbuf list of 35 mbufs totalling just
748 * below 64K works and that a chain of mbufs
749 * can be defragged into at most 32 segments:
751 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
752 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
753 ifp->if_hw_tsomaxsegcount = 35;
754 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
756 /* XXX some drivers set IFCAP_TSO after ethernet attach */
757 if (ifp->if_capabilities & IFCAP_TSO) {
758 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
760 ifp->if_hw_tsomaxsegcount,
761 ifp->if_hw_tsomaxsegsize);
769 * Update the interface index in the link layer address
772 for (ifa = ifp->if_addr; ifa != NULL;
773 ifa = TAILQ_NEXT(ifa, ifa_link)) {
774 if (ifa->ifa_addr->sa_family == AF_LINK) {
775 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
776 sdl->sdl_index = ifp->if_index;
783 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
785 curvnet->vnet_ifcnt++;
789 if (domain_init_status >= 2)
790 if_attachdomain1(ifp);
792 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
793 if (IS_DEFAULT_VNET(curvnet))
794 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
796 /* Announce the interface. */
797 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
801 if_attachdomain(void *dummy)
805 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
806 if_attachdomain1(ifp);
808 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
809 if_attachdomain, NULL);
812 if_attachdomain1(struct ifnet *ifp)
817 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
818 * cannot lock ifp->if_afdata initialization, entirely.
820 if (IF_AFDATA_TRYLOCK(ifp) == 0)
822 if (ifp->if_afdata_initialized >= domain_init_status) {
823 IF_AFDATA_UNLOCK(ifp);
824 log(LOG_WARNING, "%s called more than once on %s\n",
825 __func__, ifp->if_xname);
828 ifp->if_afdata_initialized = domain_init_status;
829 IF_AFDATA_UNLOCK(ifp);
831 /* address family dependent data region */
832 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
833 for (dp = domains; dp; dp = dp->dom_next) {
834 if (dp->dom_ifattach)
835 ifp->if_afdata[dp->dom_family] =
836 (*dp->dom_ifattach)(ifp);
841 * Remove any unicast or broadcast network addresses from an interface.
844 if_purgeaddrs(struct ifnet *ifp)
846 struct ifaddr *ifa, *next;
848 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
849 if (ifa->ifa_addr->sa_family == AF_LINK)
852 /* XXX: Ugly!! ad hoc just for INET */
853 if (ifa->ifa_addr->sa_family == AF_INET) {
854 struct ifaliasreq ifr;
856 bzero(&ifr, sizeof(ifr));
857 ifr.ifra_addr = *ifa->ifa_addr;
858 if (ifa->ifa_dstaddr)
859 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
860 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
866 if (ifa->ifa_addr->sa_family == AF_INET6) {
868 /* ifp_addrhead is already updated */
872 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
878 * Remove any multicast network addresses from an interface when an ifnet
882 if_purgemaddrs(struct ifnet *ifp)
884 struct ifmultiaddr *ifma;
885 struct ifmultiaddr *next;
888 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
889 if_delmulti_locked(ifp, ifma, 1);
890 IF_ADDR_WUNLOCK(ifp);
894 * Detach an interface, removing it from the list of "active" interfaces.
895 * If vmove flag is set on entry to if_detach_internal(), perform only a
896 * limited subset of cleanup tasks, given that we are moving an ifnet from
897 * one vnet to another, where it must be fully operational.
899 * XXXRW: There are some significant questions about event ordering, and
900 * how to prevent things from starting to use the interface during detach.
903 if_detach(struct ifnet *ifp)
906 CURVNET_SET_QUIET(ifp->if_vnet);
907 if_detach_internal(ifp, 0, NULL);
912 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
915 struct radix_node_head *rnh;
922 TAILQ_FOREACH(iter, &V_ifnet, if_link)
924 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
930 curvnet->vnet_ifcnt--;
935 * While we would want to panic here, we cannot
936 * guarantee that the interface is indeed still on
937 * the list given we don't hold locks all the way.
942 panic("%s: ifp=%p not on the ifnet tailq %p",
943 __func__, ifp, &V_ifnet);
945 return; /* XXX this should panic as well? */
949 /* Check if this is a cloned interface or not. */
950 if (vmove && ifcp != NULL)
951 *ifcp = if_clone_findifc(ifp);
954 * Remove/wait for pending events.
956 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
959 * Remove routes and flush queues.
963 if (ALTQ_IS_ENABLED(&ifp->if_snd))
964 altq_disable(&ifp->if_snd);
965 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
966 altq_detach(&ifp->if_snd);
977 * Remove all IPv6 kernel structs related to ifp. This should be done
978 * before removing routing entries below, since IPv6 interface direct
979 * routes are expected to be removed by the IPv6-specific kernel API.
980 * Otherwise, the kernel will detect some inconsistency and bark it.
988 * Prevent further calls into the device driver via ifnet.
993 * Remove link ifaddr pointer and maybe decrement if_index.
994 * Clean up all addresses.
996 free(ifp->if_hw_addr, M_IFADDR);
997 ifp->if_hw_addr = NULL;
1000 /* We can now free link ifaddr. */
1001 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1002 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1003 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1009 * Delete all remaining routes using this interface
1010 * Unfortuneatly the only way to do this is to slog through
1011 * the entire routing table looking for routes which point
1012 * to this interface...oh well...
1014 for (i = 1; i <= AF_MAX; i++) {
1015 for (j = 0; j < rt_numfibs; j++) {
1016 rnh = rt_tables_get_rnh(j, i);
1019 RADIX_NODE_HEAD_LOCK(rnh);
1020 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
1021 RADIX_NODE_HEAD_UNLOCK(rnh);
1025 /* Announce that the interface is gone. */
1026 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1027 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1028 if (IS_DEFAULT_VNET(curvnet))
1029 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1033 * We cannot hold the lock over dom_ifdetach calls as they might
1034 * sleep, for example trying to drain a callout, thus open up the
1035 * theoretical race with re-attaching.
1037 IF_AFDATA_LOCK(ifp);
1038 i = ifp->if_afdata_initialized;
1039 ifp->if_afdata_initialized = 0;
1040 IF_AFDATA_UNLOCK(ifp);
1041 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1042 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1043 (*dp->dom_ifdetach)(ifp,
1044 ifp->if_afdata[dp->dom_family]);
1052 * if_vmove() performs a limited version of if_detach() in current
1053 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1054 * An attempt is made to shrink if_index in current vnet, find an
1055 * unused if_index in target vnet and calls if_grow() if necessary,
1056 * and finally find an unused if_xname for the target vnet.
1059 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1062 struct if_clone *ifc;
1066 * Detach from current vnet, but preserve LLADDR info, do not
1067 * mark as dead etc. so that the ifnet can be reattached later.
1068 * If we cannot find it, we lost the race to someone else.
1070 rc = if_detach_internal(ifp, 1, &ifc);
1075 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1076 * the if_index for that vnet if possible.
1078 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1079 * or we'd lock on one vnet and unlock on another.
1082 ifindex_free_locked(ifp->if_index);
1086 * Perform interface-specific reassignment tasks, if provided by
1089 if (ifp->if_reassign != NULL)
1090 ifp->if_reassign(ifp, new_vnet, NULL);
1093 * Switch to the context of the target vnet.
1095 CURVNET_SET_QUIET(new_vnet);
1098 if (ifindex_alloc_locked(&idx) != 0) {
1100 panic("if_index overflow");
1102 ifp->if_index = idx;
1103 ifnet_setbyindex_locked(ifp->if_index, ifp);
1106 if_attach_internal(ifp, 1, ifc);
1112 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1115 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1120 /* Try to find the prison within our visibility. */
1121 sx_slock(&allprison_lock);
1122 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1123 sx_sunlock(&allprison_lock);
1126 prison_hold_locked(pr);
1127 mtx_unlock(&pr->pr_mtx);
1129 /* Do not try to move the iface from and to the same prison. */
1130 if (pr->pr_vnet == ifp->if_vnet) {
1135 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1136 /* XXX Lock interfaces to avoid races. */
1137 CURVNET_SET_QUIET(pr->pr_vnet);
1138 difp = ifunit(ifname);
1145 /* Move the interface into the child jail/vnet. */
1146 if_vmove(ifp, pr->pr_vnet);
1148 /* Report the new if_xname back to the userland. */
1149 sprintf(ifname, "%s", ifp->if_xname);
1156 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1159 struct vnet *vnet_dst;
1162 /* Try to find the prison within our visibility. */
1163 sx_slock(&allprison_lock);
1164 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1165 sx_sunlock(&allprison_lock);
1168 prison_hold_locked(pr);
1169 mtx_unlock(&pr->pr_mtx);
1171 /* Make sure the named iface exists in the source prison/vnet. */
1172 CURVNET_SET(pr->pr_vnet);
1173 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1180 /* Do not try to move the iface from and to the same prison. */
1181 vnet_dst = TD_TO_VNET(td);
1182 if (vnet_dst == ifp->if_vnet) {
1188 /* Get interface back from child jail/vnet. */
1189 if_vmove(ifp, vnet_dst);
1192 /* Report the new if_xname back to the userland. */
1193 sprintf(ifname, "%s", ifp->if_xname);
1201 * Add a group to an interface
1204 if_addgroup(struct ifnet *ifp, const char *groupname)
1206 struct ifg_list *ifgl;
1207 struct ifg_group *ifg = NULL;
1208 struct ifg_member *ifgm;
1211 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1212 groupname[strlen(groupname) - 1] <= '9')
1216 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1217 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1222 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1223 M_NOWAIT)) == NULL) {
1228 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1229 M_TEMP, M_NOWAIT)) == NULL) {
1235 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1236 if (!strcmp(ifg->ifg_group, groupname))
1240 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1241 M_TEMP, M_NOWAIT)) == NULL) {
1247 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1248 ifg->ifg_refcnt = 0;
1249 TAILQ_INIT(&ifg->ifg_members);
1250 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1255 ifgl->ifgl_group = ifg;
1256 ifgm->ifgm_ifp = ifp;
1259 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1260 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1261 IF_ADDR_WUNLOCK(ifp);
1266 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1267 EVENTHANDLER_INVOKE(group_change_event, groupname);
1273 * Remove a group from an interface
1276 if_delgroup(struct ifnet *ifp, const char *groupname)
1278 struct ifg_list *ifgl;
1279 struct ifg_member *ifgm;
1282 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1283 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1291 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1292 IF_ADDR_WUNLOCK(ifp);
1294 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1295 if (ifgm->ifgm_ifp == ifp)
1299 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1303 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1304 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1306 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1307 free(ifgl->ifgl_group, M_TEMP);
1313 EVENTHANDLER_INVOKE(group_change_event, groupname);
1319 * Remove an interface from all groups
1322 if_delgroups(struct ifnet *ifp)
1324 struct ifg_list *ifgl;
1325 struct ifg_member *ifgm;
1326 char groupname[IFNAMSIZ];
1329 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1330 ifgl = TAILQ_FIRST(&ifp->if_groups);
1332 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1335 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1336 IF_ADDR_WUNLOCK(ifp);
1338 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1339 if (ifgm->ifgm_ifp == ifp)
1343 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1348 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1349 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1351 EVENTHANDLER_INVOKE(group_detach_event,
1353 free(ifgl->ifgl_group, M_TEMP);
1359 EVENTHANDLER_INVOKE(group_change_event, groupname);
1367 * Stores all groups from an interface in memory pointed
1371 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1374 struct ifg_list *ifgl;
1375 struct ifg_req ifgrq, *ifgp;
1376 struct ifgroupreq *ifgr = data;
1378 if (ifgr->ifgr_len == 0) {
1380 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1381 ifgr->ifgr_len += sizeof(struct ifg_req);
1382 IF_ADDR_RUNLOCK(ifp);
1386 len = ifgr->ifgr_len;
1387 ifgp = ifgr->ifgr_groups;
1390 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1391 if (len < sizeof(ifgrq)) {
1392 IF_ADDR_RUNLOCK(ifp);
1395 bzero(&ifgrq, sizeof ifgrq);
1396 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1397 sizeof(ifgrq.ifgrq_group));
1398 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1399 IF_ADDR_RUNLOCK(ifp);
1402 len -= sizeof(ifgrq);
1405 IF_ADDR_RUNLOCK(ifp);
1411 * Stores all members of a group in memory pointed to by data
1414 if_getgroupmembers(struct ifgroupreq *data)
1416 struct ifgroupreq *ifgr = data;
1417 struct ifg_group *ifg;
1418 struct ifg_member *ifgm;
1419 struct ifg_req ifgrq, *ifgp;
1423 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1424 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1431 if (ifgr->ifgr_len == 0) {
1432 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1433 ifgr->ifgr_len += sizeof(ifgrq);
1438 len = ifgr->ifgr_len;
1439 ifgp = ifgr->ifgr_groups;
1440 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1441 if (len < sizeof(ifgrq)) {
1445 bzero(&ifgrq, sizeof ifgrq);
1446 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1447 sizeof(ifgrq.ifgrq_member));
1448 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1452 len -= sizeof(ifgrq);
1461 * Delete Routes for a Network Interface
1463 * Called for each routing entry via the rnh->rnh_walktree() call above
1464 * to delete all route entries referencing a detaching network interface.
1467 * rn pointer to node in the routing table
1468 * arg argument passed to rnh->rnh_walktree() - detaching interface
1472 * errno failed - reason indicated
1476 if_rtdel(struct radix_node *rn, void *arg)
1478 struct rtentry *rt = (struct rtentry *)rn;
1479 struct ifnet *ifp = arg;
1482 if (rt->rt_ifp == ifp) {
1485 * Protect (sorta) against walktree recursion problems
1486 * with cloned routes
1488 if ((rt->rt_flags & RTF_UP) == 0)
1491 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1493 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1494 (struct rtentry **) NULL, rt->rt_fibnum);
1496 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1504 * A compatibility function returns ifnet counter values.
1507 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1510 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1512 case IFCOUNTER_IPACKETS:
1513 return (ifp->if_ipackets);
1514 case IFCOUNTER_IERRORS:
1515 return (ifp->if_ierrors);
1516 case IFCOUNTER_OPACKETS:
1517 return (ifp->if_opackets);
1518 case IFCOUNTER_OERRORS:
1519 return (ifp->if_oerrors);
1520 case IFCOUNTER_COLLISIONS:
1521 return (ifp->if_collisions);
1522 case IFCOUNTER_IBYTES:
1523 return (ifp->if_ibytes);
1524 case IFCOUNTER_OBYTES:
1525 return (ifp->if_obytes);
1526 case IFCOUNTER_IMCASTS:
1527 return (ifp->if_imcasts);
1528 case IFCOUNTER_OMCASTS:
1529 return (ifp->if_omcasts);
1530 case IFCOUNTER_IQDROPS:
1531 return (ifp->if_iqdrops);
1533 case IFCOUNTER_OQDROPS:
1534 return (ifp->if_oqdrops);
1536 case IFCOUNTER_NOPROTO:
1537 return (ifp->if_noproto);
1545 * Increase an ifnet counter. Usually used for counters shared
1546 * between the stack and a driver, but function supports them all.
1549 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1552 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1554 case IFCOUNTER_IPACKETS:
1555 ifp->if_ipackets += inc;
1557 case IFCOUNTER_IERRORS:
1558 ifp->if_ierrors += inc;
1560 case IFCOUNTER_OPACKETS:
1561 ifp->if_opackets += inc;
1563 case IFCOUNTER_OERRORS:
1564 ifp->if_oerrors += inc;
1566 case IFCOUNTER_COLLISIONS:
1567 ifp->if_collisions += inc;
1569 case IFCOUNTER_IBYTES:
1570 ifp->if_ibytes += inc;
1572 case IFCOUNTER_OBYTES:
1573 ifp->if_obytes += inc;
1575 case IFCOUNTER_IMCASTS:
1576 ifp->if_imcasts += inc;
1578 case IFCOUNTER_OMCASTS:
1579 ifp->if_omcasts += inc;
1581 case IFCOUNTER_IQDROPS:
1582 ifp->if_iqdrops += inc;
1585 case IFCOUNTER_OQDROPS:
1586 ifp->if_oqdrops += inc;
1589 case IFCOUNTER_NOPROTO:
1590 ifp->if_noproto += inc;
1598 * Wrapper functions for struct ifnet address list locking macros. These are
1599 * used by kernel modules to avoid encoding programming interface or binary
1600 * interface assumptions that may be violated when kernel-internal locking
1601 * approaches change.
1604 if_addr_rlock(struct ifnet *ifp)
1611 if_addr_runlock(struct ifnet *ifp)
1614 IF_ADDR_RUNLOCK(ifp);
1618 if_maddr_rlock(struct ifnet *ifp)
1625 if_maddr_runlock(struct ifnet *ifp)
1628 IF_ADDR_RUNLOCK(ifp);
1632 * Initialization, destruction and refcounting functions for ifaddrs.
1635 ifa_init(struct ifaddr *ifa)
1638 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1639 refcount_init(&ifa->ifa_refcnt, 1);
1640 ifa->if_data.ifi_datalen = sizeof(ifa->if_data);
1644 ifa_ref(struct ifaddr *ifa)
1647 refcount_acquire(&ifa->ifa_refcnt);
1651 ifa_free(struct ifaddr *ifa)
1654 if (refcount_release(&ifa->ifa_refcnt)) {
1655 mtx_destroy(&ifa->ifa_mtx);
1656 free(ifa, M_IFADDR);
1661 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1664 struct rtentry *rt = NULL;
1665 struct rt_addrinfo info;
1666 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1668 bzero(&info, sizeof(info));
1669 info.rti_ifp = V_loif;
1670 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1671 info.rti_info[RTAX_DST] = ia;
1672 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1673 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1675 if (error == 0 && rt != NULL) {
1677 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1678 ifa->ifa_ifp->if_type;
1679 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1680 ifa->ifa_ifp->if_index;
1683 } else if (error != 0)
1684 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1690 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1693 struct rt_addrinfo info;
1694 struct sockaddr_dl null_sdl;
1696 bzero(&null_sdl, sizeof(null_sdl));
1697 null_sdl.sdl_len = sizeof(null_sdl);
1698 null_sdl.sdl_family = AF_LINK;
1699 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1700 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1701 bzero(&info, sizeof(info));
1702 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1703 info.rti_info[RTAX_DST] = ia;
1704 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1705 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1708 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1714 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1715 * structs used to represent other address families, it is necessary
1716 * to perform a different comparison.
1719 #define sa_equal(a1, a2) \
1720 (bcmp((a1), (a2), ((a1))->sa_len) == 0)
1722 #define sa_dl_equal(a1, a2) \
1723 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1724 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1725 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1726 LLADDR((struct sockaddr_dl *)(a2)), \
1727 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1730 * Locate an interface based on a complete address.
1733 static struct ifaddr *
1734 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1739 IFNET_RLOCK_NOSLEEP();
1740 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1742 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1743 if (ifa->ifa_addr->sa_family != addr->sa_family)
1745 if (sa_equal(addr, ifa->ifa_addr)) {
1748 IF_ADDR_RUNLOCK(ifp);
1751 /* IP6 doesn't have broadcast */
1752 if ((ifp->if_flags & IFF_BROADCAST) &&
1753 ifa->ifa_broadaddr &&
1754 ifa->ifa_broadaddr->sa_len != 0 &&
1755 sa_equal(ifa->ifa_broadaddr, addr)) {
1758 IF_ADDR_RUNLOCK(ifp);
1762 IF_ADDR_RUNLOCK(ifp);
1766 IFNET_RUNLOCK_NOSLEEP();
1771 ifa_ifwithaddr(struct sockaddr *addr)
1774 return (ifa_ifwithaddr_internal(addr, 1));
1778 ifa_ifwithaddr_check(struct sockaddr *addr)
1781 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1785 * Locate an interface based on the broadcast address.
1789 ifa_ifwithbroadaddr(struct sockaddr *addr)
1794 IFNET_RLOCK_NOSLEEP();
1795 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1797 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1798 if (ifa->ifa_addr->sa_family != addr->sa_family)
1800 if ((ifp->if_flags & IFF_BROADCAST) &&
1801 ifa->ifa_broadaddr &&
1802 ifa->ifa_broadaddr->sa_len != 0 &&
1803 sa_equal(ifa->ifa_broadaddr, addr)) {
1805 IF_ADDR_RUNLOCK(ifp);
1809 IF_ADDR_RUNLOCK(ifp);
1813 IFNET_RUNLOCK_NOSLEEP();
1818 * Locate the point to point interface with a given destination address.
1822 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
1827 IFNET_RLOCK_NOSLEEP();
1828 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1829 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1831 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1834 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1835 if (ifa->ifa_addr->sa_family != addr->sa_family)
1837 if (ifa->ifa_dstaddr != NULL &&
1838 sa_equal(addr, ifa->ifa_dstaddr)) {
1840 IF_ADDR_RUNLOCK(ifp);
1844 IF_ADDR_RUNLOCK(ifp);
1848 IFNET_RUNLOCK_NOSLEEP();
1853 ifa_ifwithdstaddr(struct sockaddr *addr)
1856 return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
1860 * Find an interface on a specific network. If many, choice
1861 * is most specific found.
1864 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
1868 struct ifaddr *ifa_maybe = NULL;
1869 u_int af = addr->sa_family;
1870 char *addr_data = addr->sa_data, *cplim;
1873 * AF_LINK addresses can be looked up directly by their index number,
1874 * so do that if we can.
1876 if (af == AF_LINK) {
1877 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1878 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1879 return (ifaddr_byindex(sdl->sdl_index));
1883 * Scan though each interface, looking for ones that have addresses
1884 * in this address family and the requested fib. Maintain a reference
1885 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1886 * kept it stable when we move onto the next interface.
1888 IFNET_RLOCK_NOSLEEP();
1889 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1890 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1893 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1894 char *cp, *cp2, *cp3;
1896 if (ifa->ifa_addr->sa_family != af)
1898 if (af == AF_INET &&
1899 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1901 * This is a bit broken as it doesn't
1902 * take into account that the remote end may
1903 * be a single node in the network we are
1905 * The trouble is that we don't know the
1906 * netmask for the remote end.
1908 if (ifa->ifa_dstaddr != NULL &&
1909 sa_equal(addr, ifa->ifa_dstaddr)) {
1911 IF_ADDR_RUNLOCK(ifp);
1916 * if we have a special address handler,
1917 * then use it instead of the generic one.
1919 if (ifa->ifa_claim_addr) {
1920 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1922 IF_ADDR_RUNLOCK(ifp);
1929 * Scan all the bits in the ifa's address.
1930 * If a bit dissagrees with what we are
1931 * looking for, mask it with the netmask
1932 * to see if it really matters.
1933 * (A byte at a time)
1935 if (ifa->ifa_netmask == 0)
1938 cp2 = ifa->ifa_addr->sa_data;
1939 cp3 = ifa->ifa_netmask->sa_data;
1940 cplim = ifa->ifa_netmask->sa_len
1941 + (char *)ifa->ifa_netmask;
1943 if ((*cp++ ^ *cp2++) & *cp3++)
1944 goto next; /* next address! */
1946 * If the netmask of what we just found
1947 * is more specific than what we had before
1948 * (if we had one), or if the virtual status
1949 * of new prefix is better than of the old one,
1950 * then remember the new one before continuing
1951 * to search for an even better one.
1953 if (ifa_maybe == NULL ||
1954 ifa_preferred(ifa_maybe, ifa) ||
1955 rn_refines((caddr_t)ifa->ifa_netmask,
1956 (caddr_t)ifa_maybe->ifa_netmask)) {
1957 if (ifa_maybe != NULL)
1958 ifa_free(ifa_maybe);
1964 IF_ADDR_RUNLOCK(ifp);
1969 IFNET_RUNLOCK_NOSLEEP();
1970 if (ifa_maybe != NULL)
1971 ifa_free(ifa_maybe);
1976 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1979 return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
1983 * Find an interface address specific to an interface best matching
1987 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1990 char *cp, *cp2, *cp3;
1992 struct ifaddr *ifa_maybe = NULL;
1993 u_int af = addr->sa_family;
1998 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1999 if (ifa->ifa_addr->sa_family != af)
2001 if (ifa_maybe == NULL)
2003 if (ifa->ifa_netmask == 0) {
2004 if (sa_equal(addr, ifa->ifa_addr) ||
2005 (ifa->ifa_dstaddr &&
2006 sa_equal(addr, ifa->ifa_dstaddr)))
2010 if (ifp->if_flags & IFF_POINTOPOINT) {
2011 if (sa_equal(addr, ifa->ifa_dstaddr))
2015 cp2 = ifa->ifa_addr->sa_data;
2016 cp3 = ifa->ifa_netmask->sa_data;
2017 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2018 for (; cp3 < cplim; cp3++)
2019 if ((*cp++ ^ *cp2++) & *cp3)
2029 IF_ADDR_RUNLOCK(ifp);
2034 * See whether new ifa is better than current one:
2035 * 1) A non-virtual one is preferred over virtual.
2036 * 2) A virtual in master state preferred over any other state.
2038 * Used in several address selecting functions.
2041 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2044 return (cur->ifa_carp && (!next->ifa_carp ||
2045 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2048 #include <net/if_llatbl.h>
2051 * Default action when installing a route with a Link Level gateway.
2052 * Lookup an appropriate real ifa to point to.
2053 * This should be moved to /sys/net/link.c eventually.
2056 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2058 struct ifaddr *ifa, *oifa;
2059 struct sockaddr *dst;
2064 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
2065 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
2067 ifa = ifaof_ifpforaddr(dst, ifp);
2072 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2073 ifa->ifa_rtrequest(cmd, rt, info);
2078 * Mark an interface down and notify protocols of
2082 if_unroute(struct ifnet *ifp, int flag, int fam)
2086 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2088 ifp->if_flags &= ~flag;
2089 getmicrotime(&ifp->if_lastchange);
2090 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2091 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2092 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2093 ifp->if_qflush(ifp);
2096 (*carp_linkstate_p)(ifp);
2101 * Mark an interface up and notify protocols of
2105 if_route(struct ifnet *ifp, int flag, int fam)
2109 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2111 ifp->if_flags |= flag;
2112 getmicrotime(&ifp->if_lastchange);
2113 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2114 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2115 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2117 (*carp_linkstate_p)(ifp);
2124 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2125 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2126 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2127 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2128 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2129 int (*vlan_setcookie_p)(struct ifnet *, void *);
2130 void *(*vlan_cookie_p)(struct ifnet *);
2133 * Handle a change in the interface link state. To avoid LORs
2134 * between driver lock and upper layer locks, as well as possible
2135 * recursions, we post event to taskqueue, and all job
2136 * is done in static do_link_state_change().
2139 if_link_state_change(struct ifnet *ifp, int link_state)
2141 /* Return if state hasn't changed. */
2142 if (ifp->if_link_state == link_state)
2145 ifp->if_link_state = link_state;
2147 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2151 do_link_state_change(void *arg, int pending)
2153 struct ifnet *ifp = (struct ifnet *)arg;
2154 int link_state = ifp->if_link_state;
2155 CURVNET_SET(ifp->if_vnet);
2157 /* Notify that the link state has changed. */
2159 if (ifp->if_vlantrunk != NULL)
2160 (*vlan_link_state_p)(ifp);
2162 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2163 IFP2AC(ifp)->ac_netgraph != NULL)
2164 (*ng_ether_link_state_p)(ifp, link_state);
2166 (*carp_linkstate_p)(ifp);
2168 (*bridge_linkstate_p)(ifp);
2170 (*lagg_linkstate_p)(ifp, link_state);
2172 if (IS_DEFAULT_VNET(curvnet))
2173 devctl_notify("IFNET", ifp->if_xname,
2174 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2177 if_printf(ifp, "%d link states coalesced\n", pending);
2178 if (log_link_state_change)
2179 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2180 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2181 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2186 * Mark an interface down and notify protocols of
2190 if_down(struct ifnet *ifp)
2193 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2194 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2198 * Mark an interface up and notify protocols of
2202 if_up(struct ifnet *ifp)
2205 if_route(ifp, IFF_UP, AF_UNSPEC);
2206 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2210 * Flush an interface queue.
2213 if_qflush(struct ifnet *ifp)
2221 if (ALTQ_IS_ENABLED(ifq))
2225 while ((m = n) != 0) {
2236 * Map interface name to interface structure pointer, with or without
2237 * returning a reference.
2240 ifunit_ref(const char *name)
2244 IFNET_RLOCK_NOSLEEP();
2245 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2246 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2247 !(ifp->if_flags & IFF_DYING))
2252 IFNET_RUNLOCK_NOSLEEP();
2257 ifunit(const char *name)
2261 IFNET_RLOCK_NOSLEEP();
2262 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2263 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2266 IFNET_RUNLOCK_NOSLEEP();
2271 * Hardware specific interface ioctls.
2274 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2278 int error = 0, do_ifup = 0;
2279 int new_flags, temp_flags;
2280 size_t namelen, onamelen;
2282 char *descrbuf, *odescrbuf;
2283 char new_name[IFNAMSIZ];
2285 struct sockaddr_dl *sdl;
2287 ifr = (struct ifreq *)data;
2290 ifr->ifr_index = ifp->if_index;
2294 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2295 ifr->ifr_flags = temp_flags & 0xffff;
2296 ifr->ifr_flagshigh = temp_flags >> 16;
2300 ifr->ifr_reqcap = ifp->if_capabilities;
2301 ifr->ifr_curcap = ifp->if_capenable;
2306 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2311 ifr->ifr_metric = ifp->if_metric;
2315 ifr->ifr_mtu = ifp->if_mtu;
2319 ifr->ifr_phys = ifp->if_physical;
2324 sx_slock(&ifdescr_sx);
2325 if (ifp->if_description == NULL)
2328 /* space for terminating nul */
2329 descrlen = strlen(ifp->if_description) + 1;
2330 if (ifr->ifr_buffer.length < descrlen)
2331 ifr->ifr_buffer.buffer = NULL;
2333 error = copyout(ifp->if_description,
2334 ifr->ifr_buffer.buffer, descrlen);
2335 ifr->ifr_buffer.length = descrlen;
2337 sx_sunlock(&ifdescr_sx);
2341 error = priv_check(td, PRIV_NET_SETIFDESCR);
2346 * Copy only (length-1) bytes to make sure that
2347 * if_description is always nul terminated. The
2348 * length parameter is supposed to count the
2349 * terminating nul in.
2351 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2352 return (ENAMETOOLONG);
2353 else if (ifr->ifr_buffer.length == 0)
2356 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2358 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2359 ifr->ifr_buffer.length - 1);
2361 free(descrbuf, M_IFDESCR);
2366 sx_xlock(&ifdescr_sx);
2367 odescrbuf = ifp->if_description;
2368 ifp->if_description = descrbuf;
2369 sx_xunlock(&ifdescr_sx);
2371 getmicrotime(&ifp->if_lastchange);
2372 free(odescrbuf, M_IFDESCR);
2376 ifr->ifr_fib = ifp->if_fib;
2380 error = priv_check(td, PRIV_NET_SETIFFIB);
2383 if (ifr->ifr_fib >= rt_numfibs)
2386 ifp->if_fib = ifr->ifr_fib;
2390 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2394 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2395 * check, so we don't need special handling here yet.
2397 new_flags = (ifr->ifr_flags & 0xffff) |
2398 (ifr->ifr_flagshigh << 16);
2399 if (ifp->if_flags & IFF_SMART) {
2400 /* Smart drivers twiddle their own routes */
2401 } else if (ifp->if_flags & IFF_UP &&
2402 (new_flags & IFF_UP) == 0) {
2404 } else if (new_flags & IFF_UP &&
2405 (ifp->if_flags & IFF_UP) == 0) {
2408 /* See if permanently promiscuous mode bit is about to flip */
2409 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2410 if (new_flags & IFF_PPROMISC)
2411 ifp->if_flags |= IFF_PROMISC;
2412 else if (ifp->if_pcount == 0)
2413 ifp->if_flags &= ~IFF_PROMISC;
2414 if (log_promisc_mode_change)
2415 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2417 ((new_flags & IFF_PPROMISC) ?
2418 "enabled" : "disabled"));
2420 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2421 (new_flags &~ IFF_CANTCHANGE);
2422 if (ifp->if_ioctl) {
2423 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2427 getmicrotime(&ifp->if_lastchange);
2431 error = priv_check(td, PRIV_NET_SETIFCAP);
2434 if (ifp->if_ioctl == NULL)
2435 return (EOPNOTSUPP);
2436 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2438 error = (*ifp->if_ioctl)(ifp, cmd, data);
2440 getmicrotime(&ifp->if_lastchange);
2445 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2450 error = priv_check(td, PRIV_NET_SETIFNAME);
2453 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2456 if (new_name[0] == '\0')
2458 if (new_name[IFNAMSIZ-1] != '\0') {
2459 new_name[IFNAMSIZ-1] = '\0';
2460 if (strlen(new_name) == IFNAMSIZ-1)
2463 if (ifunit(new_name) != NULL)
2467 * XXX: Locking. Nothing else seems to lock if_flags,
2468 * and there are numerous other races with the
2469 * ifunit() checks not being atomic with namespace
2470 * changes (renames, vmoves, if_attach, etc).
2472 ifp->if_flags |= IFF_RENAMING;
2474 /* Announce the departure of the interface. */
2475 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2476 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2478 log(LOG_INFO, "%s: changing name to '%s'\n",
2479 ifp->if_xname, new_name);
2481 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2484 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2485 namelen = strlen(new_name);
2486 onamelen = sdl->sdl_nlen;
2488 * Move the address if needed. This is safe because we
2489 * allocate space for a name of length IFNAMSIZ when we
2490 * create this in if_attach().
2492 if (namelen != onamelen) {
2493 bcopy(sdl->sdl_data + onamelen,
2494 sdl->sdl_data + namelen, sdl->sdl_alen);
2496 bcopy(new_name, sdl->sdl_data, namelen);
2497 sdl->sdl_nlen = namelen;
2498 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2499 bzero(sdl->sdl_data, onamelen);
2500 while (namelen != 0)
2501 sdl->sdl_data[--namelen] = 0xff;
2504 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2505 /* Announce the return of the interface. */
2506 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2508 ifp->if_flags &= ~IFF_RENAMING;
2513 error = priv_check(td, PRIV_NET_SETIFVNET);
2516 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2521 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2524 ifp->if_metric = ifr->ifr_metric;
2525 getmicrotime(&ifp->if_lastchange);
2529 error = priv_check(td, PRIV_NET_SETIFPHYS);
2532 if (ifp->if_ioctl == NULL)
2533 return (EOPNOTSUPP);
2534 error = (*ifp->if_ioctl)(ifp, cmd, data);
2536 getmicrotime(&ifp->if_lastchange);
2541 u_long oldmtu = ifp->if_mtu;
2543 error = priv_check(td, PRIV_NET_SETIFMTU);
2546 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2548 if (ifp->if_ioctl == NULL)
2549 return (EOPNOTSUPP);
2550 error = (*ifp->if_ioctl)(ifp, cmd, data);
2552 getmicrotime(&ifp->if_lastchange);
2556 * If the link MTU changed, do network layer specific procedure.
2558 if (ifp->if_mtu != oldmtu) {
2568 if (cmd == SIOCADDMULTI)
2569 error = priv_check(td, PRIV_NET_ADDMULTI);
2571 error = priv_check(td, PRIV_NET_DELMULTI);
2575 /* Don't allow group membership on non-multicast interfaces. */
2576 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2577 return (EOPNOTSUPP);
2579 /* Don't let users screw up protocols' entries. */
2580 if (ifr->ifr_addr.sa_family != AF_LINK)
2583 if (cmd == SIOCADDMULTI) {
2584 struct ifmultiaddr *ifma;
2587 * Userland is only permitted to join groups once
2588 * via the if_addmulti() KPI, because it cannot hold
2589 * struct ifmultiaddr * between calls. It may also
2590 * lose a race while we check if the membership
2594 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2595 IF_ADDR_RUNLOCK(ifp);
2599 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2601 error = if_delmulti(ifp, &ifr->ifr_addr);
2604 getmicrotime(&ifp->if_lastchange);
2607 case SIOCSIFPHYADDR:
2608 case SIOCDIFPHYADDR:
2610 case SIOCSIFPHYADDR_IN6:
2612 case SIOCSLIFPHYADDR:
2614 case SIOCSIFGENERIC:
2615 error = priv_check(td, PRIV_NET_HWIOCTL);
2618 if (ifp->if_ioctl == NULL)
2619 return (EOPNOTSUPP);
2620 error = (*ifp->if_ioctl)(ifp, cmd, data);
2622 getmicrotime(&ifp->if_lastchange);
2626 ifs = (struct ifstat *)data;
2627 ifs->ascii[0] = '\0';
2629 case SIOCGIFPSRCADDR:
2630 case SIOCGIFPDSTADDR:
2631 case SIOCGLIFPHYADDR:
2634 case SIOCGIFGENERIC:
2636 case SIOCGIFRSSHASH:
2637 if (ifp->if_ioctl == NULL)
2638 return (EOPNOTSUPP);
2639 error = (*ifp->if_ioctl)(ifp, cmd, data);
2643 error = priv_check(td, PRIV_NET_SETLLADDR);
2646 error = if_setlladdr(ifp,
2647 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2648 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2652 error = if_gethwaddr(ifp, ifr);
2657 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2659 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2662 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2668 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2674 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2676 error = priv_check(td, PRIV_NET_DELIFGROUP);
2679 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2691 #ifdef COMPAT_FREEBSD32
2699 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2706 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2713 CURVNET_SET(so->so_vnet);
2717 error = ifconf(cmd, data);
2721 #ifdef COMPAT_FREEBSD32
2724 struct ifconf32 *ifc32;
2727 ifc32 = (struct ifconf32 *)data;
2728 ifc.ifc_len = ifc32->ifc_len;
2729 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2731 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2734 ifc32->ifc_len = ifc.ifc_len;
2739 ifr = (struct ifreq *)data;
2744 error = priv_check(td, PRIV_NET_SETIFVNET);
2746 error = if_vmove_reclaim(td, ifr->ifr_name,
2753 error = priv_check(td, PRIV_NET_IFCREATE);
2755 error = if_clone_create(ifr->ifr_name,
2756 sizeof(ifr->ifr_name),
2757 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2761 error = priv_check(td, PRIV_NET_IFDESTROY);
2763 error = if_clone_destroy(ifr->ifr_name);
2767 case SIOCIFGCLONERS:
2768 error = if_clone_list((struct if_clonereq *)data);
2772 error = if_getgroupmembers((struct ifgroupreq *)data);
2775 #if defined(INET) || defined(INET6)
2778 if (carp_ioctl_p == NULL)
2779 error = EPROTONOSUPPORT;
2781 error = (*carp_ioctl_p)(ifr, cmd, td);
2787 ifp = ifunit_ref(ifr->ifr_name);
2793 error = ifhwioctl(cmd, ifp, data, td);
2794 if (error != ENOIOCTL) {
2800 oif_flags = ifp->if_flags;
2801 if (so->so_proto == NULL) {
2804 return (EOPNOTSUPP);
2808 * Pass the request on to the socket control method, and if the
2809 * latter returns EOPNOTSUPP, directly to the interface.
2811 * Make an exception for the legacy SIOCSIF* requests. Drivers
2812 * trust SIOCSIFADDR et al to come from an already privileged
2813 * layer, and do not perform any credentials checks or input
2817 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2820 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2821 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2822 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2823 error = (*ifp->if_ioctl)(ifp, cmd, data);
2830 case SIOCSIFDSTADDR:
2832 case SIOCSIFBRDADDR:
2833 case SIOCSIFNETMASK:
2834 #if BYTE_ORDER != BIG_ENDIAN
2835 if (ifr->ifr_addr.sa_family == 0 &&
2836 ifr->ifr_addr.sa_len < 16) {
2837 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2838 ifr->ifr_addr.sa_len = 16;
2841 if (ifr->ifr_addr.sa_len == 0)
2842 ifr->ifr_addr.sa_len = 16;
2850 case OSIOCGIFDSTADDR:
2851 cmd = SIOCGIFDSTADDR;
2854 case OSIOCGIFBRDADDR:
2855 cmd = SIOCGIFBRDADDR;
2858 case OSIOCGIFNETMASK:
2859 cmd = SIOCGIFNETMASK;
2861 error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
2865 if (error == EOPNOTSUPP && ifp != NULL &&
2866 ifp->if_ioctl != NULL &&
2867 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2868 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2869 error = (*ifp->if_ioctl)(ifp, cmd, data);
2873 case OSIOCGIFDSTADDR:
2874 case OSIOCGIFBRDADDR:
2875 case OSIOCGIFNETMASK:
2876 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2880 #endif /* COMPAT_43 */
2882 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2884 if (ifp->if_flags & IFF_UP)
2894 * The code common to handling reference counted flags,
2895 * e.g., in ifpromisc() and if_allmulti().
2896 * The "pflag" argument can specify a permanent mode flag to check,
2897 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2899 * Only to be used on stack-owned flags, not driver-owned flags.
2902 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2906 int oldflags, oldcount;
2908 /* Sanity checks to catch programming errors */
2909 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2910 ("%s: setting driver-owned flag %d", __func__, flag));
2913 KASSERT(*refcount >= 0,
2914 ("%s: increment negative refcount %d for flag %d",
2915 __func__, *refcount, flag));
2917 KASSERT(*refcount > 0,
2918 ("%s: decrement non-positive refcount %d for flag %d",
2919 __func__, *refcount, flag));
2921 /* In case this mode is permanent, just touch refcount */
2922 if (ifp->if_flags & pflag) {
2923 *refcount += onswitch ? 1 : -1;
2927 /* Save ifnet parameters for if_ioctl() may fail */
2928 oldcount = *refcount;
2929 oldflags = ifp->if_flags;
2932 * See if we aren't the only and touching refcount is enough.
2933 * Actually toggle interface flag if we are the first or last.
2938 ifp->if_flags |= flag;
2942 ifp->if_flags &= ~flag;
2945 /* Call down the driver since we've changed interface flags */
2946 if (ifp->if_ioctl == NULL) {
2950 ifr.ifr_flags = ifp->if_flags & 0xffff;
2951 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2952 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2955 /* Notify userland that interface flags have changed */
2960 /* Recover after driver error */
2961 *refcount = oldcount;
2962 ifp->if_flags = oldflags;
2967 * Set/clear promiscuous mode on interface ifp based on the truth value
2968 * of pswitch. The calls are reference counted so that only the first
2969 * "on" request actually has an effect, as does the final "off" request.
2970 * Results are undefined if the "off" and "on" requests are not matched.
2973 ifpromisc(struct ifnet *ifp, int pswitch)
2976 int oldflags = ifp->if_flags;
2978 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2979 &ifp->if_pcount, pswitch);
2980 /* If promiscuous mode status has changed, log a message */
2981 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
2982 log_promisc_mode_change)
2983 log(LOG_INFO, "%s: promiscuous mode %s\n",
2985 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2990 * Return interface configuration
2991 * of system. List may be used
2992 * in later ioctl's (above) to get
2993 * other information.
2997 ifconf(u_long cmd, caddr_t data)
2999 struct ifconf *ifc = (struct ifconf *)data;
3004 int error, full = 0, valid_len, max_len;
3006 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3007 max_len = MAXPHYS - 1;
3009 /* Prevent hostile input from being able to crash the system */
3010 if (ifc->ifc_len <= 0)
3014 if (ifc->ifc_len <= max_len) {
3015 max_len = ifc->ifc_len;
3018 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3023 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3027 * Zero the ifr_name buffer to make sure we don't
3028 * disclose the contents of the stack.
3030 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
3032 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3033 >= sizeof(ifr.ifr_name)) {
3036 return (ENAMETOOLONG);
3041 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3042 struct sockaddr *sa = ifa->ifa_addr;
3044 if (prison_if(curthread->td_ucred, sa) != 0)
3048 if (cmd == OSIOCGIFCONF) {
3049 struct osockaddr *osa =
3050 (struct osockaddr *)&ifr.ifr_addr;
3052 osa->sa_family = sa->sa_family;
3053 sbuf_bcat(sb, &ifr, sizeof(ifr));
3054 max_len += sizeof(ifr);
3057 if (sa->sa_len <= sizeof(*sa)) {
3059 sbuf_bcat(sb, &ifr, sizeof(ifr));
3060 max_len += sizeof(ifr);
3063 offsetof(struct ifreq, ifr_addr));
3064 max_len += offsetof(struct ifreq, ifr_addr);
3065 sbuf_bcat(sb, sa, sa->sa_len);
3066 max_len += sa->sa_len;
3069 if (sbuf_error(sb) == 0)
3070 valid_len = sbuf_len(sb);
3072 IF_ADDR_RUNLOCK(ifp);
3074 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
3075 sbuf_bcat(sb, &ifr, sizeof(ifr));
3076 max_len += sizeof(ifr);
3078 if (sbuf_error(sb) == 0)
3079 valid_len = sbuf_len(sb);
3085 * If we didn't allocate enough space (uncommon), try again. If
3086 * we have already allocated as much space as we are allowed,
3087 * return what we've got.
3089 if (valid_len != max_len && !full) {
3094 ifc->ifc_len = valid_len;
3096 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3102 * Just like ifpromisc(), but for all-multicast-reception mode.
3105 if_allmulti(struct ifnet *ifp, int onswitch)
3108 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3111 struct ifmultiaddr *
3112 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
3114 struct ifmultiaddr *ifma;
3116 IF_ADDR_LOCK_ASSERT(ifp);
3118 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3119 if (sa->sa_family == AF_LINK) {
3120 if (sa_dl_equal(ifma->ifma_addr, sa))
3123 if (sa_equal(ifma->ifma_addr, sa))
3132 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3133 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3134 * the ifnet multicast address list here, so the caller must do that and
3135 * other setup work (such as notifying the device driver). The reference
3136 * count is initialized to 1.
3138 static struct ifmultiaddr *
3139 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3142 struct ifmultiaddr *ifma;
3143 struct sockaddr *dupsa;
3145 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3150 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3151 if (dupsa == NULL) {
3152 free(ifma, M_IFMADDR);
3155 bcopy(sa, dupsa, sa->sa_len);
3156 ifma->ifma_addr = dupsa;
3158 ifma->ifma_ifp = ifp;
3159 ifma->ifma_refcount = 1;
3160 ifma->ifma_protospec = NULL;
3163 ifma->ifma_lladdr = NULL;
3167 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3168 if (dupsa == NULL) {
3169 free(ifma->ifma_addr, M_IFMADDR);
3170 free(ifma, M_IFMADDR);
3173 bcopy(llsa, dupsa, llsa->sa_len);
3174 ifma->ifma_lladdr = dupsa;
3180 * if_freemulti: free ifmultiaddr structure and possibly attached related
3181 * addresses. The caller is responsible for implementing reference
3182 * counting, notifying the driver, handling routing messages, and releasing
3183 * any dependent link layer state.
3186 if_freemulti(struct ifmultiaddr *ifma)
3189 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3190 ifma->ifma_refcount));
3191 KASSERT(ifma->ifma_protospec == NULL,
3192 ("if_freemulti: protospec not NULL"));
3194 if (ifma->ifma_lladdr != NULL)
3195 free(ifma->ifma_lladdr, M_IFMADDR);
3196 free(ifma->ifma_addr, M_IFMADDR);
3197 free(ifma, M_IFMADDR);
3201 * Register an additional multicast address with a network interface.
3203 * - If the address is already present, bump the reference count on the
3204 * address and return.
3205 * - If the address is not link-layer, look up a link layer address.
3206 * - Allocate address structures for one or both addresses, and attach to the
3207 * multicast address list on the interface. If automatically adding a link
3208 * layer address, the protocol address will own a reference to the link
3209 * layer address, to be freed when it is freed.
3210 * - Notify the network device driver of an addition to the multicast address
3213 * 'sa' points to caller-owned memory with the desired multicast address.
3215 * 'retifma' will be used to return a pointer to the resulting multicast
3216 * address reference, if desired.
3219 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3220 struct ifmultiaddr **retifma)
3222 struct ifmultiaddr *ifma, *ll_ifma;
3223 struct sockaddr *llsa;
3227 * If the address is already present, return a new reference to it;
3228 * otherwise, allocate storage and set up a new address.
3231 ifma = if_findmulti(ifp, sa);
3233 ifma->ifma_refcount++;
3234 if (retifma != NULL)
3236 IF_ADDR_WUNLOCK(ifp);
3241 * The address isn't already present; resolve the protocol address
3242 * into a link layer address, and then look that up, bump its
3243 * refcount or allocate an ifma for that also. If 'llsa' was
3244 * returned, we will need to free it later.
3248 if (ifp->if_resolvemulti != NULL) {
3249 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3255 * Allocate the new address. Don't hook it up yet, as we may also
3256 * need to allocate a link layer multicast address.
3258 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3265 * If a link layer address is found, we'll need to see if it's
3266 * already present in the address list, or allocate is as well.
3267 * When this block finishes, the link layer address will be on the
3271 ll_ifma = if_findmulti(ifp, llsa);
3272 if (ll_ifma == NULL) {
3273 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3274 if (ll_ifma == NULL) {
3275 --ifma->ifma_refcount;
3280 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3283 ll_ifma->ifma_refcount++;
3284 ifma->ifma_llifma = ll_ifma;
3288 * We now have a new multicast address, ifma, and possibly a new or
3289 * referenced link layer address. Add the primary address to the
3290 * ifnet address list.
3292 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3294 if (retifma != NULL)
3298 * Must generate the message while holding the lock so that 'ifma'
3299 * pointer is still valid.
3301 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3302 IF_ADDR_WUNLOCK(ifp);
3305 * We are certain we have added something, so call down to the
3306 * interface to let them know about it.
3308 if (ifp->if_ioctl != NULL) {
3309 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3313 free(llsa, M_IFMADDR);
3319 free(llsa, M_IFMADDR);
3322 IF_ADDR_WUNLOCK(ifp);
3327 * Delete a multicast group membership by network-layer group address.
3329 * Returns ENOENT if the entry could not be found. If ifp no longer
3330 * exists, results are undefined. This entry point should only be used
3331 * from subsystems which do appropriate locking to hold ifp for the
3332 * duration of the call.
3333 * Network-layer protocol domains must use if_delmulti_ifma().
3336 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3338 struct ifmultiaddr *ifma;
3343 IFNET_RLOCK_NOSLEEP();
3344 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3349 IFNET_RUNLOCK_NOSLEEP();
3351 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3358 ifma = if_findmulti(ifp, sa);
3360 lastref = if_delmulti_locked(ifp, ifma, 0);
3361 IF_ADDR_WUNLOCK(ifp);
3366 if (lastref && ifp->if_ioctl != NULL) {
3367 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3374 * Delete all multicast group membership for an interface.
3375 * Should be used to quickly flush all multicast filters.
3378 if_delallmulti(struct ifnet *ifp)
3380 struct ifmultiaddr *ifma;
3381 struct ifmultiaddr *next;
3384 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3385 if_delmulti_locked(ifp, ifma, 0);
3386 IF_ADDR_WUNLOCK(ifp);
3390 * Delete a multicast group membership by group membership pointer.
3391 * Network-layer protocol domains must use this routine.
3393 * It is safe to call this routine if the ifp disappeared.
3396 if_delmulti_ifma(struct ifmultiaddr *ifma)
3401 ifp = ifma->ifma_ifp;
3404 printf("%s: ifma_ifp seems to be detached\n", __func__);
3408 IFNET_RLOCK_NOSLEEP();
3409 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3413 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3416 IFNET_RUNLOCK_NOSLEEP();
3420 * If and only if the ifnet instance exists: Acquire the address lock.
3425 lastref = if_delmulti_locked(ifp, ifma, 0);
3429 * If and only if the ifnet instance exists:
3430 * Release the address lock.
3431 * If the group was left: update the hardware hash filter.
3433 IF_ADDR_WUNLOCK(ifp);
3434 if (lastref && ifp->if_ioctl != NULL) {
3435 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3441 * Perform deletion of network-layer and/or link-layer multicast address.
3443 * Return 0 if the reference count was decremented.
3444 * Return 1 if the final reference was released, indicating that the
3445 * hardware hash filter should be reprogrammed.
3448 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3450 struct ifmultiaddr *ll_ifma;
3452 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3453 KASSERT(ifma->ifma_ifp == ifp,
3454 ("%s: inconsistent ifp %p", __func__, ifp));
3455 IF_ADDR_WLOCK_ASSERT(ifp);
3458 ifp = ifma->ifma_ifp;
3461 * If the ifnet is detaching, null out references to ifnet,
3462 * so that upper protocol layers will notice, and not attempt
3463 * to obtain locks for an ifnet which no longer exists. The
3464 * routing socket announcement must happen before the ifnet
3465 * instance is detached from the system.
3469 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3472 * ifp may already be nulled out if we are being reentered
3473 * to delete the ll_ifma.
3476 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3477 ifma->ifma_ifp = NULL;
3481 if (--ifma->ifma_refcount > 0)
3485 * If this ifma is a network-layer ifma, a link-layer ifma may
3486 * have been associated with it. Release it first if so.
3488 ll_ifma = ifma->ifma_llifma;
3489 if (ll_ifma != NULL) {
3490 KASSERT(ifma->ifma_lladdr != NULL,
3491 ("%s: llifma w/o lladdr", __func__));
3493 ll_ifma->ifma_ifp = NULL; /* XXX */
3494 if (--ll_ifma->ifma_refcount == 0) {
3496 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3499 if_freemulti(ll_ifma);
3504 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3509 * The last reference to this instance of struct ifmultiaddr
3510 * was released; the hardware should be notified of this change.
3516 * Set the link layer address on an interface.
3518 * At this time we only support certain types of interfaces,
3519 * and we don't allow the length of the address to change.
3522 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3524 struct sockaddr_dl *sdl;
3531 IF_ADDR_RUNLOCK(ifp);
3535 IF_ADDR_RUNLOCK(ifp);
3536 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3541 if (len != sdl->sdl_alen) { /* don't allow length to change */
3545 switch (ifp->if_type) {
3553 case IFT_IEEE8023ADLAG:
3555 bcopy(lladdr, LLADDR(sdl), len);
3564 * If the interface is already up, we need
3565 * to re-init it in order to reprogram its
3568 if ((ifp->if_flags & IFF_UP) != 0) {
3569 if (ifp->if_ioctl) {
3570 ifp->if_flags &= ~IFF_UP;
3571 ifr.ifr_flags = ifp->if_flags & 0xffff;
3572 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3573 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3574 ifp->if_flags |= IFF_UP;
3575 ifr.ifr_flags = ifp->if_flags & 0xffff;
3576 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3577 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3581 * Also send gratuitous ARPs to notify other nodes about
3582 * the address change.
3584 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3585 if (ifa->ifa_addr->sa_family == AF_INET)
3586 arp_ifinit(ifp, ifa);
3594 * Get the link layer address that was read from the hardware at attach.
3596 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3597 * their component interfaces as IFT_IEEE8023ADLAG.
3600 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3603 if (ifp->if_hw_addr == NULL)
3606 switch (ifp->if_type) {
3608 case IFT_IEEE8023ADLAG:
3609 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3617 * The name argument must be a pointer to storage which will last as
3618 * long as the interface does. For physical devices, the result of
3619 * device_get_name(dev) is a good choice and for pseudo-devices a
3620 * static string works well.
3623 if_initname(struct ifnet *ifp, const char *name, int unit)
3625 ifp->if_dname = name;
3626 ifp->if_dunit = unit;
3627 if (unit != IF_DUNIT_NONE)
3628 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3630 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3634 if_printf(struct ifnet *ifp, const char * fmt, ...)
3639 retval = printf("%s: ", ifp->if_xname);
3641 retval += vprintf(fmt, ap);
3647 if_start(struct ifnet *ifp)
3650 (*(ifp)->if_start)(ifp);
3654 * Backwards compatibility interface for drivers
3655 * that have not implemented it
3658 if_transmit(struct ifnet *ifp, struct mbuf *m)
3662 IFQ_HANDOFF(ifp, m, error);
3667 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3674 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3679 if (_IF_QFULL(ifq)) {
3686 ifp->if_obytes += m->m_pkthdr.len + adjust;
3687 if (m->m_flags & (M_BCAST|M_MCAST))
3689 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3691 _IF_ENQUEUE(ifq, m);
3693 if (ifp != NULL && !active)
3694 (*(ifp)->if_start)(ifp);
3699 if_register_com_alloc(u_char type,
3700 if_com_alloc_t *a, if_com_free_t *f)
3703 KASSERT(if_com_alloc[type] == NULL,
3704 ("if_register_com_alloc: %d already registered", type));
3705 KASSERT(if_com_free[type] == NULL,
3706 ("if_register_com_alloc: %d free already registered", type));
3708 if_com_alloc[type] = a;
3709 if_com_free[type] = f;
3713 if_deregister_com_alloc(u_char type)
3716 KASSERT(if_com_alloc[type] != NULL,
3717 ("if_deregister_com_alloc: %d not registered", type));
3718 KASSERT(if_com_free[type] != NULL,
3719 ("if_deregister_com_alloc: %d free not registered", type));
3720 if_com_alloc[type] = NULL;
3721 if_com_free[type] = NULL;