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/ethernet.h>
69 #include <net/if_arp.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
73 #include <net/if_var.h>
74 #include <net/if_media.h>
75 #include <net/if_vlan_var.h>
76 #include <net/radix.h>
77 #include <net/route.h>
80 #if defined(INET) || defined(INET6)
81 #include <net/ethernet.h>
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_carp.h>
87 #include <netinet/if_ether.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet6/in6_ifattach.h>
93 #endif /* INET || INET6 */
95 #include <security/mac/mac_framework.h>
97 #ifdef COMPAT_FREEBSD32
98 #include <sys/mount.h>
99 #include <compat/freebsd32/freebsd32.h>
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 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106 &ifqmaxlen, 0, "max send queue size");
108 /* Log link state change events */
109 static int log_link_state_change = 1;
111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112 &log_link_state_change, 0,
113 "log interface link state change events");
115 /* Interface description */
116 static unsigned int ifdescr_maxlen = 1024;
117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
119 "administrative maximum length for interface description");
121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
123 /* global sx for non-critical path ifdescr */
124 static struct sx ifdescr_sx;
125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
127 void (*bridge_linkstate_p)(struct ifnet *ifp);
128 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
129 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
130 /* These are external hooks for CARP. */
131 void (*carp_linkstate_p)(struct ifnet *ifp);
132 void (*carp_demote_adj_p)(int, char *);
133 int (*carp_master_p)(struct ifaddr *);
134 #if defined(INET) || defined(INET6)
135 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
136 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
137 const struct sockaddr *sa);
138 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
139 int (*carp_attach_p)(struct ifaddr *, int);
140 void (*carp_detach_p)(struct ifaddr *);
143 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
146 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
147 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
148 const struct in6_addr *taddr);
151 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
154 * XXX: Style; these should be sorted alphabetically, and unprototyped
155 * static functions should be prototyped. Currently they are sorted by
158 static void if_attachdomain(void *);
159 static void if_attachdomain1(struct ifnet *);
160 static int ifconf(u_long, caddr_t);
161 static void if_freemulti(struct ifmultiaddr *);
162 static void if_grow(void);
163 static void if_route(struct ifnet *, int flag, int fam);
164 static int if_setflag(struct ifnet *, int, int, int *, int);
165 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
166 static void if_unroute(struct ifnet *, int flag, int fam);
167 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
168 static int if_rtdel(struct radix_node *, void *);
169 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
170 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
171 static void do_link_state_change(void *, int);
172 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
173 static int if_getgroupmembers(struct ifgroupreq *);
174 static void if_delgroups(struct ifnet *);
175 static void if_attach_internal(struct ifnet *, int);
176 static void if_detach_internal(struct ifnet *, int);
180 * XXX: declare here to avoid to include many inet6 related files..
181 * should be more generalized?
183 extern void nd6_setmtu(struct ifnet *);
186 VNET_DEFINE(int, if_index);
187 int ifqmaxlen = IFQ_MAXLEN;
188 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
189 VNET_DEFINE(struct ifgrouphead, ifg_head);
191 static VNET_DEFINE(int, if_indexlim) = 8;
193 /* Table of ifnet by index. */
194 VNET_DEFINE(struct ifnet **, ifindex_table);
196 #define V_if_indexlim VNET(if_indexlim)
197 #define V_ifindex_table VNET(ifindex_table)
200 * The global network interface list (V_ifnet) and related state (such as
201 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
202 * an rwlock. Either may be acquired shared to stablize the list, but both
203 * must be acquired writable to modify the list. This model allows us to
204 * both stablize the interface list during interrupt thread processing, but
205 * also to stablize it over long-running ioctls, without introducing priority
206 * inversions and deadlocks.
208 struct rwlock ifnet_rwlock;
209 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
210 struct sx ifnet_sxlock;
211 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
214 * The allocation of network interfaces is a rather non-atomic affair; we
215 * need to select an index before we are ready to expose the interface for
216 * use, so will use this pointer value to indicate reservation.
218 #define IFNET_HOLD (void *)(uintptr_t)(-1)
220 static if_com_alloc_t *if_com_alloc[256];
221 static if_com_free_t *if_com_free[256];
223 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
224 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
225 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
228 ifnet_byindex_locked(u_short idx)
231 if (idx > V_if_index)
233 if (V_ifindex_table[idx] == IFNET_HOLD)
235 return (V_ifindex_table[idx]);
239 ifnet_byindex(u_short idx)
243 IFNET_RLOCK_NOSLEEP();
244 ifp = ifnet_byindex_locked(idx);
245 IFNET_RUNLOCK_NOSLEEP();
250 ifnet_byindex_ref(u_short idx)
254 IFNET_RLOCK_NOSLEEP();
255 ifp = ifnet_byindex_locked(idx);
256 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
257 IFNET_RUNLOCK_NOSLEEP();
261 IFNET_RUNLOCK_NOSLEEP();
266 * Allocate an ifindex array entry; return 0 on success or an error on
274 IFNET_WLOCK_ASSERT();
277 * Try to find an empty slot below V_if_index. If we fail, take the
280 for (idx = 1; idx <= V_if_index; idx++) {
281 if (V_ifindex_table[idx] == NULL)
285 /* Catch if_index overflow. */
286 if (idx >= V_if_indexlim) {
290 if (idx > V_if_index)
296 ifindex_free_locked(u_short idx)
299 IFNET_WLOCK_ASSERT();
301 V_ifindex_table[idx] = NULL;
302 while (V_if_index > 0 &&
303 V_ifindex_table[V_if_index] == NULL)
308 ifindex_free(u_short idx)
312 ifindex_free_locked(idx);
317 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
320 IFNET_WLOCK_ASSERT();
322 V_ifindex_table[idx] = ifp;
326 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
330 ifnet_setbyindex_locked(idx, ifp);
335 ifaddr_byindex(u_short idx)
339 IFNET_RLOCK_NOSLEEP();
340 ifa = ifnet_byindex_locked(idx)->if_addr;
343 IFNET_RUNLOCK_NOSLEEP();
348 * Network interface utility routines.
350 * Routines with ifa_ifwith* names take sockaddr *'s as
355 vnet_if_init(const void *unused __unused)
358 TAILQ_INIT(&V_ifnet);
359 TAILQ_INIT(&V_ifg_head);
361 if_grow(); /* create initial table */
363 vnet_if_clone_init();
365 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
370 vnet_if_uninit(const void *unused __unused)
373 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
374 "not empty", __func__, __LINE__, &V_ifnet));
375 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
376 "not empty", __func__, __LINE__, &V_ifg_head));
378 free((caddr_t)V_ifindex_table, M_IFNET);
380 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
381 vnet_if_uninit, NULL);
391 IFNET_WLOCK_ASSERT();
392 oldlim = V_if_indexlim;
394 n = (oldlim << 1) * sizeof(*e);
395 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
397 if (V_if_indexlim != oldlim) {
401 if (V_ifindex_table != NULL) {
402 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
403 free((caddr_t)V_ifindex_table, M_IFNET);
410 * Allocate a struct ifnet and an index for an interface. A layer 2
411 * common structure will also be allocated if an allocation routine is
412 * registered for the passed type.
415 if_alloc(u_char type)
420 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
422 idx = ifindex_alloc();
423 ifnet_setbyindex_locked(idx, IFNET_HOLD);
427 ifp->if_alloctype = type;
428 if (if_com_alloc[type] != NULL) {
429 ifp->if_l2com = if_com_alloc[type](type, ifp);
430 if (ifp->if_l2com == NULL) {
437 IF_ADDR_LOCK_INIT(ifp);
438 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
439 ifp->if_afdata_initialized = 0;
440 IF_AFDATA_LOCK_INIT(ifp);
441 TAILQ_INIT(&ifp->if_addrhead);
442 TAILQ_INIT(&ifp->if_multiaddrs);
443 TAILQ_INIT(&ifp->if_groups);
447 ifq_init(&ifp->if_snd, ifp);
449 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
450 for (int i = 0; i < IFCOUNTERS; i++)
451 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
452 ifp->if_get_counter = if_get_counter_default;
453 ifnet_setbyindex(ifp->if_index, ifp);
458 * Do the actual work of freeing a struct ifnet, and layer 2 common
459 * structure. This call is made when the last reference to an
460 * interface is released.
463 if_free_internal(struct ifnet *ifp)
466 KASSERT((ifp->if_flags & IFF_DYING),
467 ("if_free_internal: interface not dying"));
469 if (if_com_free[ifp->if_alloctype] != NULL)
470 if_com_free[ifp->if_alloctype](ifp->if_l2com,
474 mac_ifnet_destroy(ifp);
476 if (ifp->if_description != NULL)
477 free(ifp->if_description, M_IFDESCR);
478 IF_AFDATA_DESTROY(ifp);
479 IF_ADDR_LOCK_DESTROY(ifp);
480 ifq_delete(&ifp->if_snd);
482 for (int i = 0; i < IFCOUNTERS; i++)
483 counter_u64_free(ifp->if_counters[i]);
489 * Deregister an interface and free the associated storage.
492 if_free(struct ifnet *ifp)
495 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
497 CURVNET_SET_QUIET(ifp->if_vnet);
499 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
500 ("%s: freeing unallocated ifnet", ifp->if_xname));
502 ifindex_free_locked(ifp->if_index);
505 if (refcount_release(&ifp->if_refcount))
506 if_free_internal(ifp);
511 * Interfaces to keep an ifnet type-stable despite the possibility of the
512 * driver calling if_free(). If there are additional references, we defer
513 * freeing the underlying data structure.
516 if_ref(struct ifnet *ifp)
519 /* We don't assert the ifnet list lock here, but arguably should. */
520 refcount_acquire(&ifp->if_refcount);
524 if_rele(struct ifnet *ifp)
527 if (!refcount_release(&ifp->if_refcount))
529 if_free_internal(ifp);
533 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
536 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
538 if (ifq->ifq_maxlen == 0)
539 ifq->ifq_maxlen = ifqmaxlen;
542 ifq->altq_disc = NULL;
543 ifq->altq_flags &= ALTQF_CANTCHANGE;
544 ifq->altq_tbr = NULL;
549 ifq_delete(struct ifaltq *ifq)
551 mtx_destroy(&ifq->ifq_mtx);
555 * Perform generic interface initalization tasks and attach the interface
556 * to the list of "active" interfaces. If vmove flag is set on entry
557 * to if_attach_internal(), perform only a limited subset of initialization
558 * tasks, given that we are moving from one vnet to another an ifnet which
559 * has already been fully initialized.
562 * - The decision to return void and thus require this function to
563 * succeed is questionable.
564 * - We should probably do more sanity checking. For instance we don't
565 * do anything to insure if_xname is unique or non-empty.
568 if_attach(struct ifnet *ifp)
571 if_attach_internal(ifp, 0);
575 * Compute the least common TSO limit.
578 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
581 * 1) If there is no limit currently, take the limit from
582 * the network adapter.
584 * 2) If the network adapter has a limit below the current
587 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
588 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
589 pmax->tsomaxbytes = ifp->if_hw_tsomax;
591 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
592 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
593 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
595 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
596 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
597 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
602 * Update TSO limit of a network adapter.
604 * Returns zero if no change. Else non-zero.
607 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
610 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
611 ifp->if_hw_tsomax = pmax->tsomaxbytes;
614 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
615 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
618 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
619 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
626 if_attach_internal(struct ifnet *ifp, int vmove)
628 unsigned socksize, ifasize;
629 int namelen, masklen;
630 struct sockaddr_dl *sdl;
633 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
634 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
638 ifp->if_vnet = curvnet;
639 if (ifp->if_home_vnet == NULL)
640 ifp->if_home_vnet = curvnet;
643 if_addgroup(ifp, IFG_ALL);
645 getmicrotime(&ifp->if_lastchange);
646 ifp->if_epoch = time_uptime;
648 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
649 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
650 ("transmit and qflush must both either be set or both be NULL"));
651 if (ifp->if_transmit == NULL) {
652 ifp->if_transmit = if_transmit;
653 ifp->if_qflush = if_qflush;
658 mac_ifnet_create(ifp);
662 * Create a Link Level name for this device.
664 namelen = strlen(ifp->if_xname);
666 * Always save enough space for any possiable name so we
667 * can do a rename in place later.
669 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
670 socksize = masklen + ifp->if_addrlen;
671 if (socksize < sizeof(*sdl))
672 socksize = sizeof(*sdl);
673 socksize = roundup2(socksize, sizeof(long));
674 ifasize = sizeof(*ifa) + 2 * socksize;
675 ifa = ifa_alloc(ifasize, M_WAITOK);
676 sdl = (struct sockaddr_dl *)(ifa + 1);
677 sdl->sdl_len = socksize;
678 sdl->sdl_family = AF_LINK;
679 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
680 sdl->sdl_nlen = namelen;
681 sdl->sdl_index = ifp->if_index;
682 sdl->sdl_type = ifp->if_type;
685 ifa->ifa_rtrequest = link_rtrequest;
686 ifa->ifa_addr = (struct sockaddr *)sdl;
687 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
688 ifa->ifa_netmask = (struct sockaddr *)sdl;
689 sdl->sdl_len = masklen;
691 sdl->sdl_data[--namelen] = 0xff;
692 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
693 /* Reliably crash if used uninitialized. */
694 ifp->if_broadcastaddr = NULL;
696 #if defined(INET) || defined(INET6)
697 /* Use defaults for TSO, if nothing is set */
698 if (ifp->if_hw_tsomax == 0 &&
699 ifp->if_hw_tsomaxsegcount == 0 &&
700 ifp->if_hw_tsomaxsegsize == 0) {
702 * The TSO defaults needs to be such that an
703 * NFS mbuf list of 35 mbufs totalling just
704 * below 64K works and that a chain of mbufs
705 * can be defragged into at most 32 segments:
707 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
708 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
709 ifp->if_hw_tsomaxsegcount = 35;
710 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
712 /* XXX some drivers set IFCAP_TSO after ethernet attach */
713 if (ifp->if_capabilities & IFCAP_TSO) {
714 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
716 ifp->if_hw_tsomaxsegcount,
717 ifp->if_hw_tsomaxsegsize);
721 * If the "if_hw_tsomax" limit is set, check if it is
724 KASSERT(ifp->if_hw_tsomax == 0 ||
725 ifp->if_hw_tsomax >= (IP_MAXPACKET / 8),
726 ("%s: if_hw_tsomax is outside of range", __func__));
732 * Update the interface index in the link layer address
735 for (ifa = ifp->if_addr; ifa != NULL;
736 ifa = TAILQ_NEXT(ifa, ifa_link)) {
737 if (ifa->ifa_addr->sa_family == AF_LINK) {
738 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
739 sdl->sdl_index = ifp->if_index;
746 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
748 curvnet->vnet_ifcnt++;
752 if (domain_init_status >= 2)
753 if_attachdomain1(ifp);
755 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
756 if (IS_DEFAULT_VNET(curvnet))
757 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
759 /* Announce the interface. */
760 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
764 if_attachdomain(void *dummy)
768 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
769 if_attachdomain1(ifp);
771 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
772 if_attachdomain, NULL);
775 if_attachdomain1(struct ifnet *ifp)
780 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
781 * cannot lock ifp->if_afdata initialization, entirely.
783 if (IF_AFDATA_TRYLOCK(ifp) == 0)
785 if (ifp->if_afdata_initialized >= domain_init_status) {
786 IF_AFDATA_UNLOCK(ifp);
787 log(LOG_WARNING, "%s called more than once on %s\n",
788 __func__, ifp->if_xname);
791 ifp->if_afdata_initialized = domain_init_status;
792 IF_AFDATA_UNLOCK(ifp);
794 /* address family dependent data region */
795 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
796 for (dp = domains; dp; dp = dp->dom_next) {
797 if (dp->dom_ifattach)
798 ifp->if_afdata[dp->dom_family] =
799 (*dp->dom_ifattach)(ifp);
804 * Remove any unicast or broadcast network addresses from an interface.
807 if_purgeaddrs(struct ifnet *ifp)
809 struct ifaddr *ifa, *next;
811 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
812 if (ifa->ifa_addr->sa_family == AF_LINK)
815 /* XXX: Ugly!! ad hoc just for INET */
816 if (ifa->ifa_addr->sa_family == AF_INET) {
817 struct ifaliasreq ifr;
819 bzero(&ifr, sizeof(ifr));
820 ifr.ifra_addr = *ifa->ifa_addr;
821 if (ifa->ifa_dstaddr)
822 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
823 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
829 if (ifa->ifa_addr->sa_family == AF_INET6) {
831 /* ifp_addrhead is already updated */
835 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
841 * Remove any multicast network addresses from an interface when an ifnet
845 if_purgemaddrs(struct ifnet *ifp)
847 struct ifmultiaddr *ifma;
848 struct ifmultiaddr *next;
851 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
852 if_delmulti_locked(ifp, ifma, 1);
853 IF_ADDR_WUNLOCK(ifp);
857 * Detach an interface, removing it from the list of "active" interfaces.
858 * If vmove flag is set on entry to if_detach_internal(), perform only a
859 * limited subset of cleanup tasks, given that we are moving an ifnet from
860 * one vnet to another, where it must be fully operational.
862 * XXXRW: There are some significant questions about event ordering, and
863 * how to prevent things from starting to use the interface during detach.
866 if_detach(struct ifnet *ifp)
869 CURVNET_SET_QUIET(ifp->if_vnet);
870 if_detach_internal(ifp, 0);
875 if_detach_internal(struct ifnet *ifp, int vmove)
878 struct radix_node_head *rnh;
885 TAILQ_FOREACH(iter, &V_ifnet, if_link)
887 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
893 curvnet->vnet_ifcnt--;
898 panic("%s: ifp=%p not on the ifnet tailq %p",
899 __func__, ifp, &V_ifnet);
901 return; /* XXX this should panic as well? */
905 * Remove/wait for pending events.
907 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
910 * Remove routes and flush queues.
914 if (ALTQ_IS_ENABLED(&ifp->if_snd))
915 altq_disable(&ifp->if_snd);
916 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
917 altq_detach(&ifp->if_snd);
928 * Remove all IPv6 kernel structs related to ifp. This should be done
929 * before removing routing entries below, since IPv6 interface direct
930 * routes are expected to be removed by the IPv6-specific kernel API.
931 * Otherwise, the kernel will detect some inconsistency and bark it.
937 /* Announce that the interface is gone. */
938 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
939 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
940 if (IS_DEFAULT_VNET(curvnet))
941 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
945 * Prevent further calls into the device driver via ifnet.
950 * Remove link ifaddr pointer and maybe decrement if_index.
951 * Clean up all addresses.
955 /* We can now free link ifaddr. */
956 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
957 ifa = TAILQ_FIRST(&ifp->if_addrhead);
958 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
964 * Delete all remaining routes using this interface
965 * Unfortuneatly the only way to do this is to slog through
966 * the entire routing table looking for routes which point
967 * to this interface...oh well...
969 for (i = 1; i <= AF_MAX; i++) {
970 for (j = 0; j < rt_numfibs; j++) {
971 rnh = rt_tables_get_rnh(j, i);
974 RADIX_NODE_HEAD_LOCK(rnh);
975 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
976 RADIX_NODE_HEAD_UNLOCK(rnh);
983 * We cannot hold the lock over dom_ifdetach calls as they might
984 * sleep, for example trying to drain a callout, thus open up the
985 * theoretical race with re-attaching.
988 i = ifp->if_afdata_initialized;
989 ifp->if_afdata_initialized = 0;
990 IF_AFDATA_UNLOCK(ifp);
991 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
992 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
993 (*dp->dom_ifdetach)(ifp,
994 ifp->if_afdata[dp->dom_family]);
1000 * if_vmove() performs a limited version of if_detach() in current
1001 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1002 * An attempt is made to shrink if_index in current vnet, find an
1003 * unused if_index in target vnet and calls if_grow() if necessary,
1004 * and finally find an unused if_xname for the target vnet.
1007 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1011 * Detach from current vnet, but preserve LLADDR info, do not
1012 * mark as dead etc. so that the ifnet can be reattached later.
1014 if_detach_internal(ifp, 1);
1017 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1018 * the if_index for that vnet if possible.
1020 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1021 * or we'd lock on one vnet and unlock on another.
1024 ifindex_free_locked(ifp->if_index);
1028 * Perform interface-specific reassignment tasks, if provided by
1031 if (ifp->if_reassign != NULL)
1032 ifp->if_reassign(ifp, new_vnet, NULL);
1035 * Switch to the context of the target vnet.
1037 CURVNET_SET_QUIET(new_vnet);
1040 ifp->if_index = ifindex_alloc();
1041 ifnet_setbyindex_locked(ifp->if_index, ifp);
1044 if_attach_internal(ifp, 1);
1050 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1053 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1058 /* Try to find the prison within our visibility. */
1059 sx_slock(&allprison_lock);
1060 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1061 sx_sunlock(&allprison_lock);
1064 prison_hold_locked(pr);
1065 mtx_unlock(&pr->pr_mtx);
1067 /* Do not try to move the iface from and to the same prison. */
1068 if (pr->pr_vnet == ifp->if_vnet) {
1073 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1074 /* XXX Lock interfaces to avoid races. */
1075 CURVNET_SET_QUIET(pr->pr_vnet);
1076 difp = ifunit(ifname);
1083 /* Move the interface into the child jail/vnet. */
1084 if_vmove(ifp, pr->pr_vnet);
1086 /* Report the new if_xname back to the userland. */
1087 sprintf(ifname, "%s", ifp->if_xname);
1094 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1097 struct vnet *vnet_dst;
1100 /* Try to find the prison within our visibility. */
1101 sx_slock(&allprison_lock);
1102 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1103 sx_sunlock(&allprison_lock);
1106 prison_hold_locked(pr);
1107 mtx_unlock(&pr->pr_mtx);
1109 /* Make sure the named iface exists in the source prison/vnet. */
1110 CURVNET_SET(pr->pr_vnet);
1111 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1118 /* Do not try to move the iface from and to the same prison. */
1119 vnet_dst = TD_TO_VNET(td);
1120 if (vnet_dst == ifp->if_vnet) {
1126 /* Get interface back from child jail/vnet. */
1127 if_vmove(ifp, vnet_dst);
1130 /* Report the new if_xname back to the userland. */
1131 sprintf(ifname, "%s", ifp->if_xname);
1139 * Add a group to an interface
1142 if_addgroup(struct ifnet *ifp, const char *groupname)
1144 struct ifg_list *ifgl;
1145 struct ifg_group *ifg = NULL;
1146 struct ifg_member *ifgm;
1149 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1150 groupname[strlen(groupname) - 1] <= '9')
1154 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1155 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1160 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1161 M_NOWAIT)) == NULL) {
1166 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1167 M_TEMP, M_NOWAIT)) == NULL) {
1173 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1174 if (!strcmp(ifg->ifg_group, groupname))
1178 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1179 M_TEMP, M_NOWAIT)) == NULL) {
1185 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1186 ifg->ifg_refcnt = 0;
1187 TAILQ_INIT(&ifg->ifg_members);
1188 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1193 ifgl->ifgl_group = ifg;
1194 ifgm->ifgm_ifp = ifp;
1197 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1198 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1199 IF_ADDR_WUNLOCK(ifp);
1204 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1205 EVENTHANDLER_INVOKE(group_change_event, groupname);
1211 * Remove a group from an interface
1214 if_delgroup(struct ifnet *ifp, const char *groupname)
1216 struct ifg_list *ifgl;
1217 struct ifg_member *ifgm;
1220 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1221 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1229 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1230 IF_ADDR_WUNLOCK(ifp);
1232 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1233 if (ifgm->ifgm_ifp == ifp)
1237 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1241 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1242 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1244 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1245 free(ifgl->ifgl_group, M_TEMP);
1251 EVENTHANDLER_INVOKE(group_change_event, groupname);
1257 * Remove an interface from all groups
1260 if_delgroups(struct ifnet *ifp)
1262 struct ifg_list *ifgl;
1263 struct ifg_member *ifgm;
1264 char groupname[IFNAMSIZ];
1267 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1268 ifgl = TAILQ_FIRST(&ifp->if_groups);
1270 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1273 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1274 IF_ADDR_WUNLOCK(ifp);
1276 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1277 if (ifgm->ifgm_ifp == ifp)
1281 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1286 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1287 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1289 EVENTHANDLER_INVOKE(group_detach_event,
1291 free(ifgl->ifgl_group, M_TEMP);
1297 EVENTHANDLER_INVOKE(group_change_event, groupname);
1305 * Stores all groups from an interface in memory pointed
1309 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1312 struct ifg_list *ifgl;
1313 struct ifg_req ifgrq, *ifgp;
1314 struct ifgroupreq *ifgr = data;
1316 if (ifgr->ifgr_len == 0) {
1318 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1319 ifgr->ifgr_len += sizeof(struct ifg_req);
1320 IF_ADDR_RUNLOCK(ifp);
1324 len = ifgr->ifgr_len;
1325 ifgp = ifgr->ifgr_groups;
1328 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1329 if (len < sizeof(ifgrq)) {
1330 IF_ADDR_RUNLOCK(ifp);
1333 bzero(&ifgrq, sizeof ifgrq);
1334 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1335 sizeof(ifgrq.ifgrq_group));
1336 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1337 IF_ADDR_RUNLOCK(ifp);
1340 len -= sizeof(ifgrq);
1343 IF_ADDR_RUNLOCK(ifp);
1349 * Stores all members of a group in memory pointed to by data
1352 if_getgroupmembers(struct ifgroupreq *data)
1354 struct ifgroupreq *ifgr = data;
1355 struct ifg_group *ifg;
1356 struct ifg_member *ifgm;
1357 struct ifg_req ifgrq, *ifgp;
1361 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1362 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1369 if (ifgr->ifgr_len == 0) {
1370 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1371 ifgr->ifgr_len += sizeof(ifgrq);
1376 len = ifgr->ifgr_len;
1377 ifgp = ifgr->ifgr_groups;
1378 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1379 if (len < sizeof(ifgrq)) {
1383 bzero(&ifgrq, sizeof ifgrq);
1384 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1385 sizeof(ifgrq.ifgrq_member));
1386 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1390 len -= sizeof(ifgrq);
1399 * Delete Routes for a Network Interface
1401 * Called for each routing entry via the rnh->rnh_walktree() call above
1402 * to delete all route entries referencing a detaching network interface.
1405 * rn pointer to node in the routing table
1406 * arg argument passed to rnh->rnh_walktree() - detaching interface
1410 * errno failed - reason indicated
1414 if_rtdel(struct radix_node *rn, void *arg)
1416 struct rtentry *rt = (struct rtentry *)rn;
1417 struct ifnet *ifp = arg;
1420 if (rt->rt_ifp == ifp) {
1423 * Protect (sorta) against walktree recursion problems
1424 * with cloned routes
1426 if ((rt->rt_flags & RTF_UP) == 0)
1429 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1431 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1432 (struct rtentry **) NULL, rt->rt_fibnum);
1434 log(LOG_WARNING, "if_rtdel: error %d\n", err);
1442 * Return counter values from counter(9)s stored in ifnet.
1445 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1448 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1450 return (counter_u64_fetch(ifp->if_counters[cnt]));
1454 * Increase an ifnet counter. Usually used for counters shared
1455 * between the stack and a driver, but function supports them all.
1458 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1461 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1463 counter_u64_add(ifp->if_counters[cnt], inc);
1467 * Copy data from ifnet to userland API structure if_data.
1470 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1473 ifd->ifi_type = ifp->if_type;
1474 ifd->ifi_physical = 0;
1475 ifd->ifi_addrlen = ifp->if_addrlen;
1476 ifd->ifi_hdrlen = ifp->if_hdrlen;
1477 ifd->ifi_link_state = ifp->if_link_state;
1479 ifd->ifi_datalen = sizeof(struct if_data);
1480 ifd->ifi_mtu = ifp->if_mtu;
1481 ifd->ifi_metric = ifp->if_metric;
1482 ifd->ifi_baudrate = ifp->if_baudrate;
1483 ifd->ifi_hwassist = ifp->if_hwassist;
1484 ifd->ifi_epoch = ifp->if_epoch;
1485 ifd->ifi_lastchange = ifp->if_lastchange;
1487 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1488 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1489 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1490 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1491 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1492 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1493 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1494 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1495 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1496 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1497 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1498 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1502 * Wrapper functions for struct ifnet address list locking macros. These are
1503 * used by kernel modules to avoid encoding programming interface or binary
1504 * interface assumptions that may be violated when kernel-internal locking
1505 * approaches change.
1508 if_addr_rlock(struct ifnet *ifp)
1515 if_addr_runlock(struct ifnet *ifp)
1518 IF_ADDR_RUNLOCK(ifp);
1522 if_maddr_rlock(if_t ifp)
1525 IF_ADDR_RLOCK((struct ifnet *)ifp);
1529 if_maddr_runlock(if_t ifp)
1532 IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1536 * Initialization, destruction and refcounting functions for ifaddrs.
1539 ifa_alloc(size_t size, int flags)
1543 KASSERT(size >= sizeof(struct ifaddr),
1544 ("%s: invalid size %zu", __func__, size));
1546 ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1550 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1552 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1554 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1556 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1559 refcount_init(&ifa->ifa_refcnt, 1);
1564 /* free(NULL) is okay */
1565 counter_u64_free(ifa->ifa_opackets);
1566 counter_u64_free(ifa->ifa_ipackets);
1567 counter_u64_free(ifa->ifa_obytes);
1568 counter_u64_free(ifa->ifa_ibytes);
1569 free(ifa, M_IFADDR);
1575 ifa_ref(struct ifaddr *ifa)
1578 refcount_acquire(&ifa->ifa_refcnt);
1582 ifa_free(struct ifaddr *ifa)
1585 if (refcount_release(&ifa->ifa_refcnt)) {
1586 counter_u64_free(ifa->ifa_opackets);
1587 counter_u64_free(ifa->ifa_ipackets);
1588 counter_u64_free(ifa->ifa_obytes);
1589 counter_u64_free(ifa->ifa_ibytes);
1590 free(ifa, M_IFADDR);
1595 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1598 struct rtentry *rt = NULL;
1599 struct rt_addrinfo info;
1600 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1602 bzero(&info, sizeof(info));
1603 info.rti_ifp = V_loif;
1604 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1605 info.rti_info[RTAX_DST] = ia;
1606 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1607 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
1609 if (error == 0 && rt != NULL) {
1611 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1612 ifa->ifa_ifp->if_type;
1613 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1614 ifa->ifa_ifp->if_index;
1617 } else if (error != 0)
1618 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
1624 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1627 struct rt_addrinfo info;
1628 struct sockaddr_dl null_sdl;
1630 bzero(&null_sdl, sizeof(null_sdl));
1631 null_sdl.sdl_len = sizeof(null_sdl);
1632 null_sdl.sdl_family = AF_LINK;
1633 null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1634 null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1635 bzero(&info, sizeof(info));
1636 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1637 info.rti_info[RTAX_DST] = ia;
1638 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1639 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
1642 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
1648 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib)
1652 rt = rtalloc1_fib(sa, 0, 0, fib);
1654 log(LOG_DEBUG, "%s: fail", __func__);
1655 return (EHOSTUNREACH);
1657 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1658 ifa->ifa_ifp->if_type;
1659 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1660 ifa->ifa_ifp->if_index;
1667 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1668 * structs used to represent other address families, it is necessary
1669 * to perform a different comparison.
1672 #define sa_dl_equal(a1, a2) \
1673 ((((struct sockaddr_dl *)(a1))->sdl_len == \
1674 ((struct sockaddr_dl *)(a2))->sdl_len) && \
1675 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
1676 LLADDR((struct sockaddr_dl *)(a2)), \
1677 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1680 * Locate an interface based on a complete address.
1683 static struct ifaddr *
1684 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1689 IFNET_RLOCK_NOSLEEP();
1690 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1692 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1693 if (ifa->ifa_addr->sa_family != addr->sa_family)
1695 if (sa_equal(addr, ifa->ifa_addr)) {
1698 IF_ADDR_RUNLOCK(ifp);
1701 /* IP6 doesn't have broadcast */
1702 if ((ifp->if_flags & IFF_BROADCAST) &&
1703 ifa->ifa_broadaddr &&
1704 ifa->ifa_broadaddr->sa_len != 0 &&
1705 sa_equal(ifa->ifa_broadaddr, addr)) {
1708 IF_ADDR_RUNLOCK(ifp);
1712 IF_ADDR_RUNLOCK(ifp);
1716 IFNET_RUNLOCK_NOSLEEP();
1721 ifa_ifwithaddr(struct sockaddr *addr)
1724 return (ifa_ifwithaddr_internal(addr, 1));
1728 ifa_ifwithaddr_check(struct sockaddr *addr)
1731 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1735 * Locate an interface based on the broadcast address.
1739 ifa_ifwithbroadaddr(struct sockaddr *addr, int fibnum)
1744 IFNET_RLOCK_NOSLEEP();
1745 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1746 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1749 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1750 if (ifa->ifa_addr->sa_family != addr->sa_family)
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)) {
1757 IF_ADDR_RUNLOCK(ifp);
1761 IF_ADDR_RUNLOCK(ifp);
1765 IFNET_RUNLOCK_NOSLEEP();
1770 * Locate the point to point interface with a given destination address.
1774 ifa_ifwithdstaddr(struct sockaddr *addr, int fibnum)
1779 IFNET_RLOCK_NOSLEEP();
1780 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1781 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1783 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1786 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1787 if (ifa->ifa_addr->sa_family != addr->sa_family)
1789 if (ifa->ifa_dstaddr != NULL &&
1790 sa_equal(addr, ifa->ifa_dstaddr)) {
1792 IF_ADDR_RUNLOCK(ifp);
1796 IF_ADDR_RUNLOCK(ifp);
1800 IFNET_RUNLOCK_NOSLEEP();
1805 * Find an interface on a specific network. If many, choice
1806 * is most specific found.
1809 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp, int fibnum)
1813 struct ifaddr *ifa_maybe = NULL;
1814 u_int af = addr->sa_family;
1815 char *addr_data = addr->sa_data, *cplim;
1818 * AF_LINK addresses can be looked up directly by their index number,
1819 * so do that if we can.
1821 if (af == AF_LINK) {
1822 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1823 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1824 return (ifaddr_byindex(sdl->sdl_index));
1828 * Scan though each interface, looking for ones that have addresses
1829 * in this address family and the requested fib. Maintain a reference
1830 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1831 * kept it stable when we move onto the next interface.
1833 IFNET_RLOCK_NOSLEEP();
1834 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1835 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1838 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1839 char *cp, *cp2, *cp3;
1841 if (ifa->ifa_addr->sa_family != af)
1843 if (af == AF_INET &&
1844 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1846 * This is a bit broken as it doesn't
1847 * take into account that the remote end may
1848 * be a single node in the network we are
1850 * The trouble is that we don't know the
1851 * netmask for the remote end.
1853 if (ifa->ifa_dstaddr != NULL &&
1854 sa_equal(addr, ifa->ifa_dstaddr)) {
1856 IF_ADDR_RUNLOCK(ifp);
1861 * Scan all the bits in the ifa's address.
1862 * If a bit dissagrees with what we are
1863 * looking for, mask it with the netmask
1864 * to see if it really matters.
1865 * (A byte at a time)
1867 if (ifa->ifa_netmask == 0)
1870 cp2 = ifa->ifa_addr->sa_data;
1871 cp3 = ifa->ifa_netmask->sa_data;
1872 cplim = ifa->ifa_netmask->sa_len
1873 + (char *)ifa->ifa_netmask;
1875 if ((*cp++ ^ *cp2++) & *cp3++)
1876 goto next; /* next address! */
1878 * If the netmask of what we just found
1879 * is more specific than what we had before
1880 * (if we had one), or if the virtual status
1881 * of new prefix is better than of the old one,
1882 * then remember the new one before continuing
1883 * to search for an even better one.
1885 if (ifa_maybe == NULL ||
1886 ifa_preferred(ifa_maybe, ifa) ||
1887 rn_refines((caddr_t)ifa->ifa_netmask,
1888 (caddr_t)ifa_maybe->ifa_netmask)) {
1889 if (ifa_maybe != NULL)
1890 ifa_free(ifa_maybe);
1896 IF_ADDR_RUNLOCK(ifp);
1901 IFNET_RUNLOCK_NOSLEEP();
1902 if (ifa_maybe != NULL)
1903 ifa_free(ifa_maybe);
1908 * Find an interface address specific to an interface best matching
1912 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1915 char *cp, *cp2, *cp3;
1917 struct ifaddr *ifa_maybe = NULL;
1918 u_int af = addr->sa_family;
1923 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1924 if (ifa->ifa_addr->sa_family != af)
1926 if (ifa_maybe == NULL)
1928 if (ifa->ifa_netmask == 0) {
1929 if (sa_equal(addr, ifa->ifa_addr) ||
1930 (ifa->ifa_dstaddr &&
1931 sa_equal(addr, ifa->ifa_dstaddr)))
1935 if (ifp->if_flags & IFF_POINTOPOINT) {
1936 if (sa_equal(addr, ifa->ifa_dstaddr))
1940 cp2 = ifa->ifa_addr->sa_data;
1941 cp3 = ifa->ifa_netmask->sa_data;
1942 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1943 for (; cp3 < cplim; cp3++)
1944 if ((*cp++ ^ *cp2++) & *cp3)
1954 IF_ADDR_RUNLOCK(ifp);
1959 * See whether new ifa is better than current one:
1960 * 1) A non-virtual one is preferred over virtual.
1961 * 2) A virtual in master state preferred over any other state.
1963 * Used in several address selecting functions.
1966 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1969 return (cur->ifa_carp && (!next->ifa_carp ||
1970 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1973 #include <net/if_llatbl.h>
1976 * Default action when installing a route with a Link Level gateway.
1977 * Lookup an appropriate real ifa to point to.
1978 * This should be moved to /sys/net/link.c eventually.
1981 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1983 struct ifaddr *ifa, *oifa;
1984 struct sockaddr *dst;
1987 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1988 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1990 ifa = ifaof_ifpforaddr(dst, ifp);
1995 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1996 ifa->ifa_rtrequest(cmd, rt, info);
2000 struct sockaddr_dl *
2001 link_alloc_sdl(size_t size, int flags)
2004 return (malloc(size, M_TEMP, flags));
2008 link_free_sdl(struct sockaddr *sa)
2014 * Fills in given sdl with interface basic info.
2015 * Returns pointer to filled sdl.
2017 struct sockaddr_dl *
2018 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2020 struct sockaddr_dl *sdl;
2022 sdl = (struct sockaddr_dl *)paddr;
2023 memset(sdl, 0, sizeof(struct sockaddr_dl));
2024 sdl->sdl_len = sizeof(struct sockaddr_dl);
2025 sdl->sdl_family = AF_LINK;
2026 sdl->sdl_index = ifp->if_index;
2027 sdl->sdl_type = iftype;
2033 * Mark an interface down and notify protocols of
2037 if_unroute(struct ifnet *ifp, int flag, int fam)
2041 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2043 ifp->if_flags &= ~flag;
2044 getmicrotime(&ifp->if_lastchange);
2045 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2046 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2047 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2048 ifp->if_qflush(ifp);
2051 (*carp_linkstate_p)(ifp);
2056 * Mark an interface up and notify protocols of
2060 if_route(struct ifnet *ifp, int flag, int fam)
2064 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2066 ifp->if_flags |= flag;
2067 getmicrotime(&ifp->if_lastchange);
2068 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2069 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2070 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2072 (*carp_linkstate_p)(ifp);
2079 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2080 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2081 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2082 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2083 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2084 int (*vlan_setcookie_p)(struct ifnet *, void *);
2085 void *(*vlan_cookie_p)(struct ifnet *);
2088 * Handle a change in the interface link state. To avoid LORs
2089 * between driver lock and upper layer locks, as well as possible
2090 * recursions, we post event to taskqueue, and all job
2091 * is done in static do_link_state_change().
2094 if_link_state_change(struct ifnet *ifp, int link_state)
2096 /* Return if state hasn't changed. */
2097 if (ifp->if_link_state == link_state)
2100 ifp->if_link_state = link_state;
2102 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2106 do_link_state_change(void *arg, int pending)
2108 struct ifnet *ifp = (struct ifnet *)arg;
2109 int link_state = ifp->if_link_state;
2110 CURVNET_SET(ifp->if_vnet);
2112 /* Notify that the link state has changed. */
2114 if (ifp->if_vlantrunk != NULL)
2115 (*vlan_link_state_p)(ifp);
2117 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2118 ifp->if_l2com != NULL)
2119 (*ng_ether_link_state_p)(ifp, link_state);
2121 (*carp_linkstate_p)(ifp);
2123 (*bridge_linkstate_p)(ifp);
2125 (*lagg_linkstate_p)(ifp, link_state);
2127 if (IS_DEFAULT_VNET(curvnet))
2128 devctl_notify("IFNET", ifp->if_xname,
2129 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2132 if_printf(ifp, "%d link states coalesced\n", pending);
2133 if (log_link_state_change)
2134 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2135 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2136 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2141 * Mark an interface down and notify protocols of
2145 if_down(struct ifnet *ifp)
2148 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2152 * Mark an interface up and notify protocols of
2156 if_up(struct ifnet *ifp)
2159 if_route(ifp, IFF_UP, AF_UNSPEC);
2163 * Flush an interface queue.
2166 if_qflush(struct ifnet *ifp)
2174 if (ALTQ_IS_ENABLED(ifq))
2178 while ((m = n) != 0) {
2189 * Map interface name to interface structure pointer, with or without
2190 * returning a reference.
2193 ifunit_ref(const char *name)
2197 IFNET_RLOCK_NOSLEEP();
2198 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2199 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2200 !(ifp->if_flags & IFF_DYING))
2205 IFNET_RUNLOCK_NOSLEEP();
2210 ifunit(const char *name)
2214 IFNET_RLOCK_NOSLEEP();
2215 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2216 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2219 IFNET_RUNLOCK_NOSLEEP();
2224 * Hardware specific interface ioctls.
2227 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2231 int new_flags, temp_flags;
2232 size_t namelen, onamelen;
2234 char *descrbuf, *odescrbuf;
2235 char new_name[IFNAMSIZ];
2237 struct sockaddr_dl *sdl;
2239 ifr = (struct ifreq *)data;
2242 ifr->ifr_index = ifp->if_index;
2246 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2247 ifr->ifr_flags = temp_flags & 0xffff;
2248 ifr->ifr_flagshigh = temp_flags >> 16;
2252 ifr->ifr_reqcap = ifp->if_capabilities;
2253 ifr->ifr_curcap = ifp->if_capenable;
2258 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2263 ifr->ifr_metric = ifp->if_metric;
2267 ifr->ifr_mtu = ifp->if_mtu;
2271 /* XXXGL: did this ever worked? */
2277 sx_slock(&ifdescr_sx);
2278 if (ifp->if_description == NULL)
2281 /* space for terminating nul */
2282 descrlen = strlen(ifp->if_description) + 1;
2283 if (ifr->ifr_buffer.length < descrlen)
2284 ifr->ifr_buffer.buffer = NULL;
2286 error = copyout(ifp->if_description,
2287 ifr->ifr_buffer.buffer, descrlen);
2288 ifr->ifr_buffer.length = descrlen;
2290 sx_sunlock(&ifdescr_sx);
2294 error = priv_check(td, PRIV_NET_SETIFDESCR);
2299 * Copy only (length-1) bytes to make sure that
2300 * if_description is always nul terminated. The
2301 * length parameter is supposed to count the
2302 * terminating nul in.
2304 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2305 return (ENAMETOOLONG);
2306 else if (ifr->ifr_buffer.length == 0)
2309 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2311 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2312 ifr->ifr_buffer.length - 1);
2314 free(descrbuf, M_IFDESCR);
2319 sx_xlock(&ifdescr_sx);
2320 odescrbuf = ifp->if_description;
2321 ifp->if_description = descrbuf;
2322 sx_xunlock(&ifdescr_sx);
2324 getmicrotime(&ifp->if_lastchange);
2325 free(odescrbuf, M_IFDESCR);
2329 ifr->ifr_fib = ifp->if_fib;
2333 error = priv_check(td, PRIV_NET_SETIFFIB);
2336 if (ifr->ifr_fib >= rt_numfibs)
2339 ifp->if_fib = ifr->ifr_fib;
2343 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2347 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2348 * check, so we don't need special handling here yet.
2350 new_flags = (ifr->ifr_flags & 0xffff) |
2351 (ifr->ifr_flagshigh << 16);
2352 if (ifp->if_flags & IFF_UP &&
2353 (new_flags & IFF_UP) == 0) {
2355 } else if (new_flags & IFF_UP &&
2356 (ifp->if_flags & IFF_UP) == 0) {
2359 /* See if permanently promiscuous mode bit is about to flip */
2360 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2361 if (new_flags & IFF_PPROMISC)
2362 ifp->if_flags |= IFF_PROMISC;
2363 else if (ifp->if_pcount == 0)
2364 ifp->if_flags &= ~IFF_PROMISC;
2365 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2367 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2369 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2370 (new_flags &~ IFF_CANTCHANGE);
2371 if (ifp->if_ioctl) {
2372 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2374 getmicrotime(&ifp->if_lastchange);
2378 error = priv_check(td, PRIV_NET_SETIFCAP);
2381 if (ifp->if_ioctl == NULL)
2382 return (EOPNOTSUPP);
2383 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2385 error = (*ifp->if_ioctl)(ifp, cmd, data);
2387 getmicrotime(&ifp->if_lastchange);
2392 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2397 error = priv_check(td, PRIV_NET_SETIFNAME);
2400 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2403 if (new_name[0] == '\0')
2405 if (ifunit(new_name) != NULL)
2409 * XXX: Locking. Nothing else seems to lock if_flags,
2410 * and there are numerous other races with the
2411 * ifunit() checks not being atomic with namespace
2412 * changes (renames, vmoves, if_attach, etc).
2414 ifp->if_flags |= IFF_RENAMING;
2416 /* Announce the departure of the interface. */
2417 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2418 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2420 log(LOG_INFO, "%s: changing name to '%s'\n",
2421 ifp->if_xname, new_name);
2424 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2426 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2427 namelen = strlen(new_name);
2428 onamelen = sdl->sdl_nlen;
2430 * Move the address if needed. This is safe because we
2431 * allocate space for a name of length IFNAMSIZ when we
2432 * create this in if_attach().
2434 if (namelen != onamelen) {
2435 bcopy(sdl->sdl_data + onamelen,
2436 sdl->sdl_data + namelen, sdl->sdl_alen);
2438 bcopy(new_name, sdl->sdl_data, namelen);
2439 sdl->sdl_nlen = namelen;
2440 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2441 bzero(sdl->sdl_data, onamelen);
2442 while (namelen != 0)
2443 sdl->sdl_data[--namelen] = 0xff;
2444 IF_ADDR_WUNLOCK(ifp);
2446 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2447 /* Announce the return of the interface. */
2448 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2450 ifp->if_flags &= ~IFF_RENAMING;
2455 error = priv_check(td, PRIV_NET_SETIFVNET);
2458 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2463 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2466 ifp->if_metric = ifr->ifr_metric;
2467 getmicrotime(&ifp->if_lastchange);
2471 error = priv_check(td, PRIV_NET_SETIFPHYS);
2474 if (ifp->if_ioctl == NULL)
2475 return (EOPNOTSUPP);
2476 error = (*ifp->if_ioctl)(ifp, cmd, data);
2478 getmicrotime(&ifp->if_lastchange);
2483 u_long oldmtu = ifp->if_mtu;
2485 error = priv_check(td, PRIV_NET_SETIFMTU);
2488 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2490 if (ifp->if_ioctl == NULL)
2491 return (EOPNOTSUPP);
2492 error = (*ifp->if_ioctl)(ifp, cmd, data);
2494 getmicrotime(&ifp->if_lastchange);
2498 * If the link MTU changed, do network layer specific procedure.
2500 if (ifp->if_mtu != oldmtu) {
2510 if (cmd == SIOCADDMULTI)
2511 error = priv_check(td, PRIV_NET_ADDMULTI);
2513 error = priv_check(td, PRIV_NET_DELMULTI);
2517 /* Don't allow group membership on non-multicast interfaces. */
2518 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2519 return (EOPNOTSUPP);
2521 /* Don't let users screw up protocols' entries. */
2522 if (ifr->ifr_addr.sa_family != AF_LINK)
2525 if (cmd == SIOCADDMULTI) {
2526 struct ifmultiaddr *ifma;
2529 * Userland is only permitted to join groups once
2530 * via the if_addmulti() KPI, because it cannot hold
2531 * struct ifmultiaddr * between calls. It may also
2532 * lose a race while we check if the membership
2536 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2537 IF_ADDR_RUNLOCK(ifp);
2541 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2543 error = if_delmulti(ifp, &ifr->ifr_addr);
2546 getmicrotime(&ifp->if_lastchange);
2549 case SIOCSIFPHYADDR:
2550 case SIOCDIFPHYADDR:
2552 case SIOCSIFPHYADDR_IN6:
2555 case SIOCSIFGENERIC:
2556 error = priv_check(td, PRIV_NET_HWIOCTL);
2559 if (ifp->if_ioctl == NULL)
2560 return (EOPNOTSUPP);
2561 error = (*ifp->if_ioctl)(ifp, cmd, data);
2563 getmicrotime(&ifp->if_lastchange);
2567 case SIOCGIFPSRCADDR:
2568 case SIOCGIFPDSTADDR:
2570 case SIOCGIFGENERIC:
2571 if (ifp->if_ioctl == NULL)
2572 return (EOPNOTSUPP);
2573 error = (*ifp->if_ioctl)(ifp, cmd, data);
2577 error = priv_check(td, PRIV_NET_SETLLADDR);
2580 error = if_setlladdr(ifp,
2581 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2582 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2587 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2589 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2592 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2598 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2604 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2606 error = priv_check(td, PRIV_NET_DELIFGROUP);
2609 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2621 #ifdef COMPAT_FREEBSD32
2629 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2636 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2643 CURVNET_SET(so->so_vnet);
2646 error = ifconf(cmd, data);
2650 #ifdef COMPAT_FREEBSD32
2653 struct ifconf32 *ifc32;
2656 ifc32 = (struct ifconf32 *)data;
2657 ifc.ifc_len = ifc32->ifc_len;
2658 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2660 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2663 ifc32->ifc_len = ifc.ifc_len;
2668 ifr = (struct ifreq *)data;
2673 error = priv_check(td, PRIV_NET_SETIFVNET);
2675 error = if_vmove_reclaim(td, ifr->ifr_name,
2682 error = priv_check(td, PRIV_NET_IFCREATE);
2684 error = if_clone_create(ifr->ifr_name,
2685 sizeof(ifr->ifr_name),
2686 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2690 error = priv_check(td, PRIV_NET_IFDESTROY);
2692 error = if_clone_destroy(ifr->ifr_name);
2696 case SIOCIFGCLONERS:
2697 error = if_clone_list((struct if_clonereq *)data);
2701 error = if_getgroupmembers((struct ifgroupreq *)data);
2704 #if defined(INET) || defined(INET6)
2707 if (carp_ioctl_p == NULL)
2708 error = EPROTONOSUPPORT;
2710 error = (*carp_ioctl_p)(ifr, cmd, td);
2716 ifp = ifunit_ref(ifr->ifr_name);
2722 error = ifhwioctl(cmd, ifp, data, td);
2723 if (error != ENOIOCTL) {
2729 oif_flags = ifp->if_flags;
2730 if (so->so_proto == NULL) {
2733 return (EOPNOTSUPP);
2737 * Pass the request on to the socket control method, and if the
2738 * latter returns EOPNOTSUPP, directly to the interface.
2740 * Make an exception for the legacy SIOCSIF* requests. Drivers
2741 * trust SIOCSIFADDR et al to come from an already privileged
2742 * layer, and do not perform any credentials checks or input
2745 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2747 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2748 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2749 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2750 error = (*ifp->if_ioctl)(ifp, cmd, data);
2752 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2754 if (ifp->if_flags & IFF_UP)
2764 * The code common to handling reference counted flags,
2765 * e.g., in ifpromisc() and if_allmulti().
2766 * The "pflag" argument can specify a permanent mode flag to check,
2767 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2769 * Only to be used on stack-owned flags, not driver-owned flags.
2772 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2776 int oldflags, oldcount;
2778 /* Sanity checks to catch programming errors */
2779 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2780 ("%s: setting driver-owned flag %d", __func__, flag));
2783 KASSERT(*refcount >= 0,
2784 ("%s: increment negative refcount %d for flag %d",
2785 __func__, *refcount, flag));
2787 KASSERT(*refcount > 0,
2788 ("%s: decrement non-positive refcount %d for flag %d",
2789 __func__, *refcount, flag));
2791 /* In case this mode is permanent, just touch refcount */
2792 if (ifp->if_flags & pflag) {
2793 *refcount += onswitch ? 1 : -1;
2797 /* Save ifnet parameters for if_ioctl() may fail */
2798 oldcount = *refcount;
2799 oldflags = ifp->if_flags;
2802 * See if we aren't the only and touching refcount is enough.
2803 * Actually toggle interface flag if we are the first or last.
2808 ifp->if_flags |= flag;
2812 ifp->if_flags &= ~flag;
2815 /* Call down the driver since we've changed interface flags */
2816 if (ifp->if_ioctl == NULL) {
2820 ifr.ifr_flags = ifp->if_flags & 0xffff;
2821 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2822 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2825 /* Notify userland that interface flags have changed */
2830 /* Recover after driver error */
2831 *refcount = oldcount;
2832 ifp->if_flags = oldflags;
2837 * Set/clear promiscuous mode on interface ifp based on the truth value
2838 * of pswitch. The calls are reference counted so that only the first
2839 * "on" request actually has an effect, as does the final "off" request.
2840 * Results are undefined if the "off" and "on" requests are not matched.
2843 ifpromisc(struct ifnet *ifp, int pswitch)
2846 int oldflags = ifp->if_flags;
2848 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2849 &ifp->if_pcount, pswitch);
2850 /* If promiscuous mode status has changed, log a message */
2851 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2852 log(LOG_INFO, "%s: promiscuous mode %s\n",
2854 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2859 * Return interface configuration
2860 * of system. List may be used
2861 * in later ioctl's (above) to get
2862 * other information.
2866 ifconf(u_long cmd, caddr_t data)
2868 struct ifconf *ifc = (struct ifconf *)data;
2873 int error, full = 0, valid_len, max_len;
2875 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2876 max_len = MAXPHYS - 1;
2878 /* Prevent hostile input from being able to crash the system */
2879 if (ifc->ifc_len <= 0)
2883 if (ifc->ifc_len <= max_len) {
2884 max_len = ifc->ifc_len;
2887 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2892 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2896 * Zero the ifr_name buffer to make sure we don't
2897 * disclose the contents of the stack.
2899 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2901 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2902 >= sizeof(ifr.ifr_name)) {
2905 return (ENAMETOOLONG);
2910 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2911 struct sockaddr *sa = ifa->ifa_addr;
2913 if (prison_if(curthread->td_ucred, sa) != 0)
2916 if (sa->sa_len <= sizeof(*sa)) {
2918 sbuf_bcat(sb, &ifr, sizeof(ifr));
2919 max_len += sizeof(ifr);
2922 offsetof(struct ifreq, ifr_addr));
2923 max_len += offsetof(struct ifreq, ifr_addr);
2924 sbuf_bcat(sb, sa, sa->sa_len);
2925 max_len += sa->sa_len;
2928 if (sbuf_error(sb) == 0)
2929 valid_len = sbuf_len(sb);
2931 IF_ADDR_RUNLOCK(ifp);
2933 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2934 sbuf_bcat(sb, &ifr, sizeof(ifr));
2935 max_len += sizeof(ifr);
2937 if (sbuf_error(sb) == 0)
2938 valid_len = sbuf_len(sb);
2944 * If we didn't allocate enough space (uncommon), try again. If
2945 * we have already allocated as much space as we are allowed,
2946 * return what we've got.
2948 if (valid_len != max_len && !full) {
2953 ifc->ifc_len = valid_len;
2955 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2961 * Just like ifpromisc(), but for all-multicast-reception mode.
2964 if_allmulti(struct ifnet *ifp, int onswitch)
2967 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2970 struct ifmultiaddr *
2971 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2973 struct ifmultiaddr *ifma;
2975 IF_ADDR_LOCK_ASSERT(ifp);
2977 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2978 if (sa->sa_family == AF_LINK) {
2979 if (sa_dl_equal(ifma->ifma_addr, sa))
2982 if (sa_equal(ifma->ifma_addr, sa))
2991 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2992 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2993 * the ifnet multicast address list here, so the caller must do that and
2994 * other setup work (such as notifying the device driver). The reference
2995 * count is initialized to 1.
2997 static struct ifmultiaddr *
2998 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3001 struct ifmultiaddr *ifma;
3002 struct sockaddr *dupsa;
3004 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3009 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3010 if (dupsa == NULL) {
3011 free(ifma, M_IFMADDR);
3014 bcopy(sa, dupsa, sa->sa_len);
3015 ifma->ifma_addr = dupsa;
3017 ifma->ifma_ifp = ifp;
3018 ifma->ifma_refcount = 1;
3019 ifma->ifma_protospec = NULL;
3022 ifma->ifma_lladdr = NULL;
3026 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3027 if (dupsa == NULL) {
3028 free(ifma->ifma_addr, M_IFMADDR);
3029 free(ifma, M_IFMADDR);
3032 bcopy(llsa, dupsa, llsa->sa_len);
3033 ifma->ifma_lladdr = dupsa;
3039 * if_freemulti: free ifmultiaddr structure and possibly attached related
3040 * addresses. The caller is responsible for implementing reference
3041 * counting, notifying the driver, handling routing messages, and releasing
3042 * any dependent link layer state.
3045 if_freemulti(struct ifmultiaddr *ifma)
3048 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3049 ifma->ifma_refcount));
3051 if (ifma->ifma_lladdr != NULL)
3052 free(ifma->ifma_lladdr, M_IFMADDR);
3053 free(ifma->ifma_addr, M_IFMADDR);
3054 free(ifma, M_IFMADDR);
3058 * Register an additional multicast address with a network interface.
3060 * - If the address is already present, bump the reference count on the
3061 * address and return.
3062 * - If the address is not link-layer, look up a link layer address.
3063 * - Allocate address structures for one or both addresses, and attach to the
3064 * multicast address list on the interface. If automatically adding a link
3065 * layer address, the protocol address will own a reference to the link
3066 * layer address, to be freed when it is freed.
3067 * - Notify the network device driver of an addition to the multicast address
3070 * 'sa' points to caller-owned memory with the desired multicast address.
3072 * 'retifma' will be used to return a pointer to the resulting multicast
3073 * address reference, if desired.
3076 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3077 struct ifmultiaddr **retifma)
3079 struct ifmultiaddr *ifma, *ll_ifma;
3080 struct sockaddr *llsa;
3081 struct sockaddr_dl sdl;
3085 * If the address is already present, return a new reference to it;
3086 * otherwise, allocate storage and set up a new address.
3089 ifma = if_findmulti(ifp, sa);
3091 ifma->ifma_refcount++;
3092 if (retifma != NULL)
3094 IF_ADDR_WUNLOCK(ifp);
3099 * The address isn't already present; resolve the protocol address
3100 * into a link layer address, and then look that up, bump its
3101 * refcount or allocate an ifma for that also.
3102 * Most link layer resolving functions returns address data which
3103 * fits inside default sockaddr_dl structure. However callback
3104 * can allocate another sockaddr structure, in that case we need to
3109 if (ifp->if_resolvemulti != NULL) {
3110 /* Provide called function with buffer size information */
3111 sdl.sdl_len = sizeof(sdl);
3112 llsa = (struct sockaddr *)&sdl;
3113 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3119 * Allocate the new address. Don't hook it up yet, as we may also
3120 * need to allocate a link layer multicast address.
3122 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3129 * If a link layer address is found, we'll need to see if it's
3130 * already present in the address list, or allocate is as well.
3131 * When this block finishes, the link layer address will be on the
3135 ll_ifma = if_findmulti(ifp, llsa);
3136 if (ll_ifma == NULL) {
3137 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3138 if (ll_ifma == NULL) {
3139 --ifma->ifma_refcount;
3144 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3147 ll_ifma->ifma_refcount++;
3148 ifma->ifma_llifma = ll_ifma;
3152 * We now have a new multicast address, ifma, and possibly a new or
3153 * referenced link layer address. Add the primary address to the
3154 * ifnet address list.
3156 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3158 if (retifma != NULL)
3162 * Must generate the message while holding the lock so that 'ifma'
3163 * pointer is still valid.
3165 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3166 IF_ADDR_WUNLOCK(ifp);
3169 * We are certain we have added something, so call down to the
3170 * interface to let them know about it.
3172 if (ifp->if_ioctl != NULL) {
3173 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3176 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3177 link_free_sdl(llsa);
3182 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3183 link_free_sdl(llsa);
3186 IF_ADDR_WUNLOCK(ifp);
3191 * Delete a multicast group membership by network-layer group address.
3193 * Returns ENOENT if the entry could not be found. If ifp no longer
3194 * exists, results are undefined. This entry point should only be used
3195 * from subsystems which do appropriate locking to hold ifp for the
3196 * duration of the call.
3197 * Network-layer protocol domains must use if_delmulti_ifma().
3200 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3202 struct ifmultiaddr *ifma;
3207 IFNET_RLOCK_NOSLEEP();
3208 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3213 IFNET_RUNLOCK_NOSLEEP();
3215 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3222 ifma = if_findmulti(ifp, sa);
3224 lastref = if_delmulti_locked(ifp, ifma, 0);
3225 IF_ADDR_WUNLOCK(ifp);
3230 if (lastref && ifp->if_ioctl != NULL) {
3231 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3238 * Delete all multicast group membership for an interface.
3239 * Should be used to quickly flush all multicast filters.
3242 if_delallmulti(struct ifnet *ifp)
3244 struct ifmultiaddr *ifma;
3245 struct ifmultiaddr *next;
3248 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3249 if_delmulti_locked(ifp, ifma, 0);
3250 IF_ADDR_WUNLOCK(ifp);
3254 * Delete a multicast group membership by group membership pointer.
3255 * Network-layer protocol domains must use this routine.
3257 * It is safe to call this routine if the ifp disappeared.
3260 if_delmulti_ifma(struct ifmultiaddr *ifma)
3265 ifp = ifma->ifma_ifp;
3268 printf("%s: ifma_ifp seems to be detached\n", __func__);
3272 IFNET_RLOCK_NOSLEEP();
3273 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3277 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3280 IFNET_RUNLOCK_NOSLEEP();
3284 * If and only if the ifnet instance exists: Acquire the address lock.
3289 lastref = if_delmulti_locked(ifp, ifma, 0);
3293 * If and only if the ifnet instance exists:
3294 * Release the address lock.
3295 * If the group was left: update the hardware hash filter.
3297 IF_ADDR_WUNLOCK(ifp);
3298 if (lastref && ifp->if_ioctl != NULL) {
3299 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3305 * Perform deletion of network-layer and/or link-layer multicast address.
3307 * Return 0 if the reference count was decremented.
3308 * Return 1 if the final reference was released, indicating that the
3309 * hardware hash filter should be reprogrammed.
3312 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3314 struct ifmultiaddr *ll_ifma;
3316 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3317 KASSERT(ifma->ifma_ifp == ifp,
3318 ("%s: inconsistent ifp %p", __func__, ifp));
3319 IF_ADDR_WLOCK_ASSERT(ifp);
3322 ifp = ifma->ifma_ifp;
3325 * If the ifnet is detaching, null out references to ifnet,
3326 * so that upper protocol layers will notice, and not attempt
3327 * to obtain locks for an ifnet which no longer exists. The
3328 * routing socket announcement must happen before the ifnet
3329 * instance is detached from the system.
3333 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3336 * ifp may already be nulled out if we are being reentered
3337 * to delete the ll_ifma.
3340 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3341 ifma->ifma_ifp = NULL;
3345 if (--ifma->ifma_refcount > 0)
3349 * If this ifma is a network-layer ifma, a link-layer ifma may
3350 * have been associated with it. Release it first if so.
3352 ll_ifma = ifma->ifma_llifma;
3353 if (ll_ifma != NULL) {
3354 KASSERT(ifma->ifma_lladdr != NULL,
3355 ("%s: llifma w/o lladdr", __func__));
3357 ll_ifma->ifma_ifp = NULL; /* XXX */
3358 if (--ll_ifma->ifma_refcount == 0) {
3360 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3363 if_freemulti(ll_ifma);
3368 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3373 * The last reference to this instance of struct ifmultiaddr
3374 * was released; the hardware should be notified of this change.
3380 * Set the link layer address on an interface.
3382 * At this time we only support certain types of interfaces,
3383 * and we don't allow the length of the address to change.
3386 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3388 struct sockaddr_dl *sdl;
3395 IF_ADDR_RUNLOCK(ifp);
3399 IF_ADDR_RUNLOCK(ifp);
3400 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3405 if (len != sdl->sdl_alen) { /* don't allow length to change */
3409 switch (ifp->if_type) {
3417 case IFT_IEEE8023ADLAG:
3419 bcopy(lladdr, LLADDR(sdl), len);
3428 * If the interface is already up, we need
3429 * to re-init it in order to reprogram its
3432 if ((ifp->if_flags & IFF_UP) != 0) {
3433 if (ifp->if_ioctl) {
3434 ifp->if_flags &= ~IFF_UP;
3435 ifr.ifr_flags = ifp->if_flags & 0xffff;
3436 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3437 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3438 ifp->if_flags |= IFF_UP;
3439 ifr.ifr_flags = ifp->if_flags & 0xffff;
3440 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3441 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3445 * Also send gratuitous ARPs to notify other nodes about
3446 * the address change.
3448 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3449 if (ifa->ifa_addr->sa_family == AF_INET)
3450 arp_ifinit(ifp, ifa);
3458 * The name argument must be a pointer to storage which will last as
3459 * long as the interface does. For physical devices, the result of
3460 * device_get_name(dev) is a good choice and for pseudo-devices a
3461 * static string works well.
3464 if_initname(struct ifnet *ifp, const char *name, int unit)
3466 ifp->if_dname = name;
3467 ifp->if_dunit = unit;
3468 if (unit != IF_DUNIT_NONE)
3469 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3471 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3475 if_printf(struct ifnet *ifp, const char * fmt, ...)
3480 retval = printf("%s: ", ifp->if_xname);
3482 retval += vprintf(fmt, ap);
3488 if_start(struct ifnet *ifp)
3491 (*(ifp)->if_start)(ifp);
3495 * Backwards compatibility interface for drivers
3496 * that have not implemented it
3499 if_transmit(struct ifnet *ifp, struct mbuf *m)
3503 IFQ_HANDOFF(ifp, m, error);
3508 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3513 if (_IF_QFULL(ifq)) {
3515 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3520 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3521 if (m->m_flags & (M_BCAST|M_MCAST))
3522 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3523 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3525 _IF_ENQUEUE(ifq, m);
3527 if (ifp != NULL && !active)
3528 (*(ifp)->if_start)(ifp);
3533 if_register_com_alloc(u_char type,
3534 if_com_alloc_t *a, if_com_free_t *f)
3537 KASSERT(if_com_alloc[type] == NULL,
3538 ("if_register_com_alloc: %d already registered", type));
3539 KASSERT(if_com_free[type] == NULL,
3540 ("if_register_com_alloc: %d free already registered", type));
3542 if_com_alloc[type] = a;
3543 if_com_free[type] = f;
3547 if_deregister_com_alloc(u_char type)
3550 KASSERT(if_com_alloc[type] != NULL,
3551 ("if_deregister_com_alloc: %d not registered", type));
3552 KASSERT(if_com_free[type] != NULL,
3553 ("if_deregister_com_alloc: %d free not registered", type));
3554 if_com_alloc[type] = NULL;
3555 if_com_free[type] = NULL;
3558 /* API for driver access to network stack owned ifnet.*/
3560 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3564 oldbrate = ifp->if_baudrate;
3565 ifp->if_baudrate = baudrate;
3570 if_getbaudrate(if_t ifp)
3573 return (((struct ifnet *)ifp)->if_baudrate);
3577 if_setcapabilities(if_t ifp, int capabilities)
3579 ((struct ifnet *)ifp)->if_capabilities = capabilities;
3584 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3586 ((struct ifnet *)ifp)->if_capabilities |= setbit;
3587 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3593 if_getcapabilities(if_t ifp)
3595 return ((struct ifnet *)ifp)->if_capabilities;
3599 if_setcapenable(if_t ifp, int capabilities)
3601 ((struct ifnet *)ifp)->if_capenable = capabilities;
3606 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3609 ((struct ifnet *)ifp)->if_capenable |= setcap;
3611 ((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3617 if_getdname(if_t ifp)
3619 return ((struct ifnet *)ifp)->if_dname;
3623 if_togglecapenable(if_t ifp, int togglecap)
3625 ((struct ifnet *)ifp)->if_capenable ^= togglecap;
3630 if_getcapenable(if_t ifp)
3632 return ((struct ifnet *)ifp)->if_capenable;
3636 * This is largely undesirable because it ties ifnet to a device, but does
3637 * provide flexiblity for an embedded product vendor. Should be used with
3638 * the understanding that it violates the interface boundaries, and should be
3639 * a last resort only.
3642 if_setdev(if_t ifp, void *dev)
3648 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3650 ((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3651 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3657 if_getdrvflags(if_t ifp)
3659 return ((struct ifnet *)ifp)->if_drv_flags;
3663 if_setdrvflags(if_t ifp, int flags)
3665 ((struct ifnet *)ifp)->if_drv_flags = flags;
3671 if_setflags(if_t ifp, int flags)
3673 ((struct ifnet *)ifp)->if_flags = flags;
3678 if_setflagbits(if_t ifp, int set, int clear)
3680 ((struct ifnet *)ifp)->if_flags |= set;
3681 ((struct ifnet *)ifp)->if_flags &= ~clear;
3687 if_getflags(if_t ifp)
3689 return ((struct ifnet *)ifp)->if_flags;
3693 if_clearhwassist(if_t ifp)
3695 ((struct ifnet *)ifp)->if_hwassist = 0;
3700 if_sethwassistbits(if_t ifp, int toset, int toclear)
3702 ((struct ifnet *)ifp)->if_hwassist |= toset;
3703 ((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3709 if_sethwassist(if_t ifp, int hwassist_bit)
3711 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3716 if_gethwassist(if_t ifp)
3718 return ((struct ifnet *)ifp)->if_hwassist;
3722 if_setmtu(if_t ifp, int mtu)
3724 ((struct ifnet *)ifp)->if_mtu = mtu;
3731 return ((struct ifnet *)ifp)->if_mtu;
3735 if_getmtu_family(if_t ifp, int family)
3739 for (dp = domains; dp; dp = dp->dom_next) {
3740 if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3741 return (dp->dom_ifmtu((struct ifnet *)ifp));
3744 return (((struct ifnet *)ifp)->if_mtu);
3748 if_setsoftc(if_t ifp, void *softc)
3750 ((struct ifnet *)ifp)->if_softc = softc;
3755 if_getsoftc(if_t ifp)
3757 return ((struct ifnet *)ifp)->if_softc;
3761 if_setrcvif(struct mbuf *m, if_t ifp)
3763 m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3767 if_setvtag(struct mbuf *m, uint16_t tag)
3769 m->m_pkthdr.ether_vtag = tag;
3773 if_getvtag(struct mbuf *m)
3776 return (m->m_pkthdr.ether_vtag);
3780 if_sendq_empty(if_t ifp)
3782 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3786 if_getifaddr(if_t ifp)
3788 return ((struct ifnet *)ifp)->if_addr;
3792 if_getamcount(if_t ifp)
3794 return ((struct ifnet *)ifp)->if_amcount;
3799 if_setsendqready(if_t ifp)
3801 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3806 if_setsendqlen(if_t ifp, int tx_desc_count)
3808 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3809 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3815 if_vlantrunkinuse(if_t ifp)
3817 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3821 if_input(if_t ifp, struct mbuf* sendmp)
3823 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3829 #ifndef ETH_ADDR_LEN
3830 #define ETH_ADDR_LEN 6
3834 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3836 struct ifmultiaddr *ifma;
3837 uint8_t *lmta = (uint8_t *)mta;
3840 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3841 if (ifma->ifma_addr->sa_family != AF_LINK)
3847 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3848 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3857 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3861 if_maddr_rlock(ifp);
3862 error = if_setupmultiaddr(ifp, mta, cnt, max);
3863 if_maddr_runlock(ifp);
3868 if_multiaddr_count(if_t ifp, int max)
3870 struct ifmultiaddr *ifma;
3874 if_maddr_rlock(ifp);
3875 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3876 if (ifma->ifma_addr->sa_family != AF_LINK)
3882 if_maddr_runlock(ifp);
3887 if_dequeue(if_t ifp)
3890 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3896 if_sendq_prepend(if_t ifp, struct mbuf *m)
3898 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3903 if_setifheaderlen(if_t ifp, int len)
3905 ((struct ifnet *)ifp)->if_hdrlen = len;
3910 if_getlladdr(if_t ifp)
3912 return (IF_LLADDR((struct ifnet *)ifp));
3916 if_gethandle(u_char type)
3918 return (if_alloc(type));
3922 if_bpfmtap(if_t ifh, struct mbuf *m)
3924 struct ifnet *ifp = (struct ifnet *)ifh;
3930 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3932 struct ifnet *ifp = (struct ifnet *)ifh;
3934 ETHER_BPF_MTAP(ifp, m);
3938 if_vlancap(if_t ifh)
3940 struct ifnet *ifp = (struct ifnet *)ifh;
3941 VLAN_CAPABILITIES(ifp);
3945 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3947 ((struct ifnet *)ifp)->if_init = init_fn;
3951 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3953 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3957 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3959 ((struct ifnet *)ifp)->if_start = (void *)start_fn;
3963 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3965 ((struct ifnet *)ifp)->if_transmit = start_fn;
3968 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3970 ((struct ifnet *)ifp)->if_qflush = flush_fn;
3975 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
3978 ifp->if_get_counter = fn;
3981 /* Revisit these - These are inline functions originally. */
3983 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
3985 return drbr_inuse_drv(ifh, br);
3989 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
3991 return drbr_dequeue(ifh, br);
3995 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
3997 return drbr_needs_enqueue(ifh, br);
4001 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4003 return drbr_enqueue(ifh, br, m);