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_input_default(struct ifnet *, struct mbuf *);
164 static int if_requestencap_default(struct ifnet *, struct if_encap_req *);
165 static void if_route(struct ifnet *, int flag, int fam);
166 static int if_setflag(struct ifnet *, int, int, int *, int);
167 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
168 static void if_unroute(struct ifnet *, int flag, int fam);
169 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
170 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
171 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
172 static void do_link_state_change(void *, int);
173 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
174 static int if_getgroupmembers(struct ifgroupreq *);
175 static void if_delgroups(struct ifnet *);
176 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
177 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
181 * XXX: declare here to avoid to include many inet6 related files..
182 * should be more generalized?
184 extern void nd6_setmtu(struct ifnet *);
187 /* ipsec helper hooks */
188 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
189 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
191 VNET_DEFINE(int, if_index);
192 int ifqmaxlen = IFQ_MAXLEN;
193 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
194 VNET_DEFINE(struct ifgrouphead, ifg_head);
196 static VNET_DEFINE(int, if_indexlim) = 8;
198 /* Table of ifnet by index. */
199 VNET_DEFINE(struct ifnet **, ifindex_table);
201 #define V_if_indexlim VNET(if_indexlim)
202 #define V_ifindex_table VNET(ifindex_table)
205 * The global network interface list (V_ifnet) and related state (such as
206 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
207 * an rwlock. Either may be acquired shared to stablize the list, but both
208 * must be acquired writable to modify the list. This model allows us to
209 * both stablize the interface list during interrupt thread processing, but
210 * also to stablize it over long-running ioctls, without introducing priority
211 * inversions and deadlocks.
213 struct rwlock ifnet_rwlock;
214 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
215 struct sx ifnet_sxlock;
216 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
219 * The allocation of network interfaces is a rather non-atomic affair; we
220 * need to select an index before we are ready to expose the interface for
221 * use, so will use this pointer value to indicate reservation.
223 #define IFNET_HOLD (void *)(uintptr_t)(-1)
225 static if_com_alloc_t *if_com_alloc[256];
226 static if_com_free_t *if_com_free[256];
228 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
229 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
230 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
233 ifnet_byindex_locked(u_short idx)
236 if (idx > V_if_index)
238 if (V_ifindex_table[idx] == IFNET_HOLD)
240 return (V_ifindex_table[idx]);
244 ifnet_byindex(u_short idx)
248 IFNET_RLOCK_NOSLEEP();
249 ifp = ifnet_byindex_locked(idx);
250 IFNET_RUNLOCK_NOSLEEP();
255 ifnet_byindex_ref(u_short idx)
259 IFNET_RLOCK_NOSLEEP();
260 ifp = ifnet_byindex_locked(idx);
261 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
262 IFNET_RUNLOCK_NOSLEEP();
266 IFNET_RUNLOCK_NOSLEEP();
271 * Allocate an ifindex array entry; return 0 on success or an error on
279 IFNET_WLOCK_ASSERT();
282 * Try to find an empty slot below V_if_index. If we fail, take the
285 for (idx = 1; idx <= V_if_index; idx++) {
286 if (V_ifindex_table[idx] == NULL)
290 /* Catch if_index overflow. */
291 if (idx >= V_if_indexlim) {
295 if (idx > V_if_index)
301 ifindex_free_locked(u_short idx)
304 IFNET_WLOCK_ASSERT();
306 V_ifindex_table[idx] = NULL;
307 while (V_if_index > 0 &&
308 V_ifindex_table[V_if_index] == NULL)
313 ifindex_free(u_short idx)
317 ifindex_free_locked(idx);
322 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
325 IFNET_WLOCK_ASSERT();
327 V_ifindex_table[idx] = ifp;
331 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
335 ifnet_setbyindex_locked(idx, ifp);
340 ifaddr_byindex(u_short idx)
343 struct ifaddr *ifa = NULL;
345 IFNET_RLOCK_NOSLEEP();
346 ifp = ifnet_byindex_locked(idx);
347 if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
349 IFNET_RUNLOCK_NOSLEEP();
354 * Network interface utility routines.
356 * Routines with ifa_ifwith* names take sockaddr *'s as
361 vnet_if_init(const void *unused __unused)
364 TAILQ_INIT(&V_ifnet);
365 TAILQ_INIT(&V_ifg_head);
367 if_grow(); /* create initial table */
369 vnet_if_clone_init();
371 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
376 vnet_if_uninit(const void *unused __unused)
379 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
380 "not empty", __func__, __LINE__, &V_ifnet));
381 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
382 "not empty", __func__, __LINE__, &V_ifg_head));
384 free((caddr_t)V_ifindex_table, M_IFNET);
386 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
387 vnet_if_uninit, NULL);
397 IFNET_WLOCK_ASSERT();
398 oldlim = V_if_indexlim;
400 n = (oldlim << 1) * sizeof(*e);
401 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
403 if (V_if_indexlim != oldlim) {
407 if (V_ifindex_table != NULL) {
408 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
409 free((caddr_t)V_ifindex_table, M_IFNET);
416 * Allocate a struct ifnet and an index for an interface. A layer 2
417 * common structure will also be allocated if an allocation routine is
418 * registered for the passed type.
421 if_alloc(u_char type)
426 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
428 idx = ifindex_alloc();
429 ifnet_setbyindex_locked(idx, IFNET_HOLD);
433 ifp->if_alloctype = type;
434 if (if_com_alloc[type] != NULL) {
435 ifp->if_l2com = if_com_alloc[type](type, ifp);
436 if (ifp->if_l2com == NULL) {
443 IF_ADDR_LOCK_INIT(ifp);
444 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
445 ifp->if_afdata_initialized = 0;
446 IF_AFDATA_LOCK_INIT(ifp);
447 TAILQ_INIT(&ifp->if_addrhead);
448 TAILQ_INIT(&ifp->if_multiaddrs);
449 TAILQ_INIT(&ifp->if_groups);
453 ifq_init(&ifp->if_snd, ifp);
455 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
456 for (int i = 0; i < IFCOUNTERS; i++)
457 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
458 ifp->if_get_counter = if_get_counter_default;
459 ifnet_setbyindex(ifp->if_index, ifp);
464 * Do the actual work of freeing a struct ifnet, and layer 2 common
465 * structure. This call is made when the last reference to an
466 * interface is released.
469 if_free_internal(struct ifnet *ifp)
472 KASSERT((ifp->if_flags & IFF_DYING),
473 ("if_free_internal: interface not dying"));
475 if (if_com_free[ifp->if_alloctype] != NULL)
476 if_com_free[ifp->if_alloctype](ifp->if_l2com,
480 mac_ifnet_destroy(ifp);
482 if (ifp->if_description != NULL)
483 free(ifp->if_description, M_IFDESCR);
484 IF_AFDATA_DESTROY(ifp);
485 IF_ADDR_LOCK_DESTROY(ifp);
486 ifq_delete(&ifp->if_snd);
488 for (int i = 0; i < IFCOUNTERS; i++)
489 counter_u64_free(ifp->if_counters[i]);
495 * Deregister an interface and free the associated storage.
498 if_free(struct ifnet *ifp)
501 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
503 CURVNET_SET_QUIET(ifp->if_vnet);
505 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
506 ("%s: freeing unallocated ifnet", ifp->if_xname));
508 ifindex_free_locked(ifp->if_index);
511 if (refcount_release(&ifp->if_refcount))
512 if_free_internal(ifp);
517 * Interfaces to keep an ifnet type-stable despite the possibility of the
518 * driver calling if_free(). If there are additional references, we defer
519 * freeing the underlying data structure.
522 if_ref(struct ifnet *ifp)
525 /* We don't assert the ifnet list lock here, but arguably should. */
526 refcount_acquire(&ifp->if_refcount);
530 if_rele(struct ifnet *ifp)
533 if (!refcount_release(&ifp->if_refcount))
535 if_free_internal(ifp);
539 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
542 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
544 if (ifq->ifq_maxlen == 0)
545 ifq->ifq_maxlen = ifqmaxlen;
548 ifq->altq_disc = NULL;
549 ifq->altq_flags &= ALTQF_CANTCHANGE;
550 ifq->altq_tbr = NULL;
555 ifq_delete(struct ifaltq *ifq)
557 mtx_destroy(&ifq->ifq_mtx);
561 * Perform generic interface initalization tasks and attach the interface
562 * to the list of "active" interfaces. If vmove flag is set on entry
563 * to if_attach_internal(), perform only a limited subset of initialization
564 * tasks, given that we are moving from one vnet to another an ifnet which
565 * has already been fully initialized.
567 * Note that if_detach_internal() removes group membership unconditionally
568 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
569 * Thus, when if_vmove() is applied to a cloned interface, group membership
570 * is lost while a cloned one always joins a group whose name is
571 * ifc->ifc_name. To recover this after if_detach_internal() and
572 * if_attach_internal(), the cloner should be specified to
573 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
574 * attempts to join a group whose name is ifc->ifc_name.
577 * - The decision to return void and thus require this function to
578 * succeed is questionable.
579 * - We should probably do more sanity checking. For instance we don't
580 * do anything to insure if_xname is unique or non-empty.
583 if_attach(struct ifnet *ifp)
586 if_attach_internal(ifp, 0, NULL);
590 * Compute the least common TSO limit.
593 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
596 * 1) If there is no limit currently, take the limit from
597 * the network adapter.
599 * 2) If the network adapter has a limit below the current
602 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
603 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
604 pmax->tsomaxbytes = ifp->if_hw_tsomax;
606 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
607 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
608 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
610 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
611 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
612 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
617 * Update TSO limit of a network adapter.
619 * Returns zero if no change. Else non-zero.
622 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
625 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
626 ifp->if_hw_tsomax = pmax->tsomaxbytes;
629 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
630 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
633 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
634 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
641 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
643 unsigned socksize, ifasize;
644 int namelen, masklen;
645 struct sockaddr_dl *sdl;
648 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
649 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
653 ifp->if_vnet = curvnet;
654 if (ifp->if_home_vnet == NULL)
655 ifp->if_home_vnet = curvnet;
658 if_addgroup(ifp, IFG_ALL);
660 /* Restore group membership for cloned interfaces. */
661 if (vmove && ifc != NULL)
662 if_clone_addgroup(ifp, ifc);
664 getmicrotime(&ifp->if_lastchange);
665 ifp->if_epoch = time_uptime;
667 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
668 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
669 ("transmit and qflush must both either be set or both be NULL"));
670 if (ifp->if_transmit == NULL) {
671 ifp->if_transmit = if_transmit;
672 ifp->if_qflush = if_qflush;
674 if (ifp->if_input == NULL)
675 ifp->if_input = if_input_default;
677 if (ifp->if_requestencap == NULL)
678 ifp->if_requestencap = if_requestencap_default;
682 mac_ifnet_create(ifp);
686 * Create a Link Level name for this device.
688 namelen = strlen(ifp->if_xname);
690 * Always save enough space for any possiable name so we
691 * can do a rename in place later.
693 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
694 socksize = masklen + ifp->if_addrlen;
695 if (socksize < sizeof(*sdl))
696 socksize = sizeof(*sdl);
697 socksize = roundup2(socksize, sizeof(long));
698 ifasize = sizeof(*ifa) + 2 * socksize;
699 ifa = ifa_alloc(ifasize, M_WAITOK);
700 sdl = (struct sockaddr_dl *)(ifa + 1);
701 sdl->sdl_len = socksize;
702 sdl->sdl_family = AF_LINK;
703 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
704 sdl->sdl_nlen = namelen;
705 sdl->sdl_index = ifp->if_index;
706 sdl->sdl_type = ifp->if_type;
709 ifa->ifa_rtrequest = link_rtrequest;
710 ifa->ifa_addr = (struct sockaddr *)sdl;
711 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
712 ifa->ifa_netmask = (struct sockaddr *)sdl;
713 sdl->sdl_len = masklen;
715 sdl->sdl_data[--namelen] = 0xff;
716 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
717 /* Reliably crash if used uninitialized. */
718 ifp->if_broadcastaddr = NULL;
720 #if defined(INET) || defined(INET6)
721 /* Use defaults for TSO, if nothing is set */
722 if (ifp->if_hw_tsomax == 0 &&
723 ifp->if_hw_tsomaxsegcount == 0 &&
724 ifp->if_hw_tsomaxsegsize == 0) {
726 * The TSO defaults needs to be such that an
727 * NFS mbuf list of 35 mbufs totalling just
728 * below 64K works and that a chain of mbufs
729 * can be defragged into at most 32 segments:
731 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
732 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
733 ifp->if_hw_tsomaxsegcount = 35;
734 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
736 /* XXX some drivers set IFCAP_TSO after ethernet attach */
737 if (ifp->if_capabilities & IFCAP_TSO) {
738 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
740 ifp->if_hw_tsomaxsegcount,
741 ifp->if_hw_tsomaxsegsize);
749 * Update the interface index in the link layer address
752 for (ifa = ifp->if_addr; ifa != NULL;
753 ifa = TAILQ_NEXT(ifa, ifa_link)) {
754 if (ifa->ifa_addr->sa_family == AF_LINK) {
755 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
756 sdl->sdl_index = ifp->if_index;
763 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
765 curvnet->vnet_ifcnt++;
769 if (domain_init_status >= 2)
770 if_attachdomain1(ifp);
772 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
773 if (IS_DEFAULT_VNET(curvnet))
774 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
776 /* Announce the interface. */
777 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
781 if_attachdomain(void *dummy)
785 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
786 if_attachdomain1(ifp);
788 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
789 if_attachdomain, NULL);
792 if_attachdomain1(struct ifnet *ifp)
797 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
798 * cannot lock ifp->if_afdata initialization, entirely.
800 if (IF_AFDATA_TRYLOCK(ifp) == 0)
802 if (ifp->if_afdata_initialized >= domain_init_status) {
803 IF_AFDATA_UNLOCK(ifp);
804 log(LOG_WARNING, "%s called more than once on %s\n",
805 __func__, ifp->if_xname);
808 ifp->if_afdata_initialized = domain_init_status;
809 IF_AFDATA_UNLOCK(ifp);
811 /* address family dependent data region */
812 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
813 for (dp = domains; dp; dp = dp->dom_next) {
814 if (dp->dom_ifattach)
815 ifp->if_afdata[dp->dom_family] =
816 (*dp->dom_ifattach)(ifp);
821 * Remove any unicast or broadcast network addresses from an interface.
824 if_purgeaddrs(struct ifnet *ifp)
826 struct ifaddr *ifa, *next;
828 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
829 if (ifa->ifa_addr->sa_family == AF_LINK)
832 /* XXX: Ugly!! ad hoc just for INET */
833 if (ifa->ifa_addr->sa_family == AF_INET) {
834 struct ifaliasreq ifr;
836 bzero(&ifr, sizeof(ifr));
837 ifr.ifra_addr = *ifa->ifa_addr;
838 if (ifa->ifa_dstaddr)
839 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
840 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
846 if (ifa->ifa_addr->sa_family == AF_INET6) {
848 /* ifp_addrhead is already updated */
852 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
858 * Remove any multicast network addresses from an interface when an ifnet
862 if_purgemaddrs(struct ifnet *ifp)
864 struct ifmultiaddr *ifma;
865 struct ifmultiaddr *next;
868 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
869 if_delmulti_locked(ifp, ifma, 1);
870 IF_ADDR_WUNLOCK(ifp);
874 * Detach an interface, removing it from the list of "active" interfaces.
875 * If vmove flag is set on entry to if_detach_internal(), perform only a
876 * limited subset of cleanup tasks, given that we are moving an ifnet from
877 * one vnet to another, where it must be fully operational.
879 * XXXRW: There are some significant questions about event ordering, and
880 * how to prevent things from starting to use the interface during detach.
883 if_detach(struct ifnet *ifp)
886 CURVNET_SET_QUIET(ifp->if_vnet);
887 if_detach_internal(ifp, 0, NULL);
892 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
901 TAILQ_FOREACH(iter, &V_ifnet, if_link)
903 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
909 curvnet->vnet_ifcnt--;
914 * While we would want to panic here, we cannot
915 * guarantee that the interface is indeed still on
916 * the list given we don't hold locks all the way.
921 panic("%s: ifp=%p not on the ifnet tailq %p",
922 __func__, ifp, &V_ifnet);
924 return; /* XXX this should panic as well? */
928 /* Check if this is a cloned interface or not. */
929 if (vmove && ifcp != NULL)
930 *ifcp = if_clone_findifc(ifp);
933 * Remove/wait for pending events.
935 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
938 * Remove routes and flush queues.
942 if (ALTQ_IS_ENABLED(&ifp->if_snd))
943 altq_disable(&ifp->if_snd);
944 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
945 altq_detach(&ifp->if_snd);
956 * Remove all IPv6 kernel structs related to ifp. This should be done
957 * before removing routing entries below, since IPv6 interface direct
958 * routes are expected to be removed by the IPv6-specific kernel API.
959 * Otherwise, the kernel will detect some inconsistency and bark it.
965 /* Announce that the interface is gone. */
966 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
967 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
968 if (IS_DEFAULT_VNET(curvnet))
969 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
973 * Prevent further calls into the device driver via ifnet.
978 * Remove link ifaddr pointer and maybe decrement if_index.
979 * Clean up all addresses.
983 /* We can now free link ifaddr. */
984 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
985 ifa = TAILQ_FIRST(&ifp->if_addrhead);
986 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
991 rt_flushifroutes(ifp);
995 * We cannot hold the lock over dom_ifdetach calls as they might
996 * sleep, for example trying to drain a callout, thus open up the
997 * theoretical race with re-attaching.
1000 i = ifp->if_afdata_initialized;
1001 ifp->if_afdata_initialized = 0;
1002 IF_AFDATA_UNLOCK(ifp);
1003 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1004 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
1005 (*dp->dom_ifdetach)(ifp,
1006 ifp->if_afdata[dp->dom_family]);
1014 * if_vmove() performs a limited version of if_detach() in current
1015 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1016 * An attempt is made to shrink if_index in current vnet, find an
1017 * unused if_index in target vnet and calls if_grow() if necessary,
1018 * and finally find an unused if_xname for the target vnet.
1021 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1023 struct if_clone *ifc;
1027 * Detach from current vnet, but preserve LLADDR info, do not
1028 * mark as dead etc. so that the ifnet can be reattached later.
1029 * If we cannot find it, we lost the race to someone else.
1031 rc = if_detach_internal(ifp, 1, &ifc);
1036 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1037 * the if_index for that vnet if possible.
1039 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1040 * or we'd lock on one vnet and unlock on another.
1043 ifindex_free_locked(ifp->if_index);
1047 * Perform interface-specific reassignment tasks, if provided by
1050 if (ifp->if_reassign != NULL)
1051 ifp->if_reassign(ifp, new_vnet, NULL);
1054 * Switch to the context of the target vnet.
1056 CURVNET_SET_QUIET(new_vnet);
1059 ifp->if_index = ifindex_alloc();
1060 ifnet_setbyindex_locked(ifp->if_index, ifp);
1063 if_attach_internal(ifp, 1, ifc);
1069 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1072 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1077 /* Try to find the prison within our visibility. */
1078 sx_slock(&allprison_lock);
1079 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1080 sx_sunlock(&allprison_lock);
1083 prison_hold_locked(pr);
1084 mtx_unlock(&pr->pr_mtx);
1086 /* Do not try to move the iface from and to the same prison. */
1087 if (pr->pr_vnet == ifp->if_vnet) {
1092 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1093 /* XXX Lock interfaces to avoid races. */
1094 CURVNET_SET_QUIET(pr->pr_vnet);
1095 difp = ifunit(ifname);
1102 /* Move the interface into the child jail/vnet. */
1103 if_vmove(ifp, pr->pr_vnet);
1105 /* Report the new if_xname back to the userland. */
1106 sprintf(ifname, "%s", ifp->if_xname);
1113 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1116 struct vnet *vnet_dst;
1119 /* Try to find the prison within our visibility. */
1120 sx_slock(&allprison_lock);
1121 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1122 sx_sunlock(&allprison_lock);
1125 prison_hold_locked(pr);
1126 mtx_unlock(&pr->pr_mtx);
1128 /* Make sure the named iface exists in the source prison/vnet. */
1129 CURVNET_SET(pr->pr_vnet);
1130 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1137 /* Do not try to move the iface from and to the same prison. */
1138 vnet_dst = TD_TO_VNET(td);
1139 if (vnet_dst == ifp->if_vnet) {
1145 /* Get interface back from child jail/vnet. */
1146 if_vmove(ifp, vnet_dst);
1149 /* Report the new if_xname back to the userland. */
1150 sprintf(ifname, "%s", ifp->if_xname);
1158 * Add a group to an interface
1161 if_addgroup(struct ifnet *ifp, const char *groupname)
1163 struct ifg_list *ifgl;
1164 struct ifg_group *ifg = NULL;
1165 struct ifg_member *ifgm;
1168 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1169 groupname[strlen(groupname) - 1] <= '9')
1173 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1174 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1179 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1180 M_NOWAIT)) == NULL) {
1185 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1186 M_TEMP, M_NOWAIT)) == NULL) {
1192 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1193 if (!strcmp(ifg->ifg_group, groupname))
1197 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1198 M_TEMP, M_NOWAIT)) == NULL) {
1204 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1205 ifg->ifg_refcnt = 0;
1206 TAILQ_INIT(&ifg->ifg_members);
1207 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1212 ifgl->ifgl_group = ifg;
1213 ifgm->ifgm_ifp = ifp;
1216 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1217 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1218 IF_ADDR_WUNLOCK(ifp);
1223 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1224 EVENTHANDLER_INVOKE(group_change_event, groupname);
1230 * Remove a group from an interface
1233 if_delgroup(struct ifnet *ifp, const char *groupname)
1235 struct ifg_list *ifgl;
1236 struct ifg_member *ifgm;
1239 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1240 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1248 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1249 IF_ADDR_WUNLOCK(ifp);
1251 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1252 if (ifgm->ifgm_ifp == ifp)
1256 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1260 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1261 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1263 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1264 free(ifgl->ifgl_group, M_TEMP);
1270 EVENTHANDLER_INVOKE(group_change_event, groupname);
1276 * Remove an interface from all groups
1279 if_delgroups(struct ifnet *ifp)
1281 struct ifg_list *ifgl;
1282 struct ifg_member *ifgm;
1283 char groupname[IFNAMSIZ];
1286 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1287 ifgl = TAILQ_FIRST(&ifp->if_groups);
1289 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1292 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1293 IF_ADDR_WUNLOCK(ifp);
1295 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1296 if (ifgm->ifgm_ifp == ifp)
1300 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1305 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1306 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1308 EVENTHANDLER_INVOKE(group_detach_event,
1310 free(ifgl->ifgl_group, M_TEMP);
1316 EVENTHANDLER_INVOKE(group_change_event, groupname);
1324 * Stores all groups from an interface in memory pointed
1328 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1331 struct ifg_list *ifgl;
1332 struct ifg_req ifgrq, *ifgp;
1333 struct ifgroupreq *ifgr = data;
1335 if (ifgr->ifgr_len == 0) {
1337 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1338 ifgr->ifgr_len += sizeof(struct ifg_req);
1339 IF_ADDR_RUNLOCK(ifp);
1343 len = ifgr->ifgr_len;
1344 ifgp = ifgr->ifgr_groups;
1347 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1348 if (len < sizeof(ifgrq)) {
1349 IF_ADDR_RUNLOCK(ifp);
1352 bzero(&ifgrq, sizeof ifgrq);
1353 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1354 sizeof(ifgrq.ifgrq_group));
1355 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1356 IF_ADDR_RUNLOCK(ifp);
1359 len -= sizeof(ifgrq);
1362 IF_ADDR_RUNLOCK(ifp);
1368 * Stores all members of a group in memory pointed to by data
1371 if_getgroupmembers(struct ifgroupreq *data)
1373 struct ifgroupreq *ifgr = data;
1374 struct ifg_group *ifg;
1375 struct ifg_member *ifgm;
1376 struct ifg_req ifgrq, *ifgp;
1380 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1381 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1388 if (ifgr->ifgr_len == 0) {
1389 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1390 ifgr->ifgr_len += sizeof(ifgrq);
1395 len = ifgr->ifgr_len;
1396 ifgp = ifgr->ifgr_groups;
1397 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1398 if (len < sizeof(ifgrq)) {
1402 bzero(&ifgrq, sizeof ifgrq);
1403 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1404 sizeof(ifgrq.ifgrq_member));
1405 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1409 len -= sizeof(ifgrq);
1418 * Return counter values from counter(9)s stored in ifnet.
1421 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1424 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1426 return (counter_u64_fetch(ifp->if_counters[cnt]));
1430 * Increase an ifnet counter. Usually used for counters shared
1431 * between the stack and a driver, but function supports them all.
1434 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1437 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1439 counter_u64_add(ifp->if_counters[cnt], inc);
1443 * Copy data from ifnet to userland API structure if_data.
1446 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1449 ifd->ifi_type = ifp->if_type;
1450 ifd->ifi_physical = 0;
1451 ifd->ifi_addrlen = ifp->if_addrlen;
1452 ifd->ifi_hdrlen = ifp->if_hdrlen;
1453 ifd->ifi_link_state = ifp->if_link_state;
1455 ifd->ifi_datalen = sizeof(struct if_data);
1456 ifd->ifi_mtu = ifp->if_mtu;
1457 ifd->ifi_metric = ifp->if_metric;
1458 ifd->ifi_baudrate = ifp->if_baudrate;
1459 ifd->ifi_hwassist = ifp->if_hwassist;
1460 ifd->ifi_epoch = ifp->if_epoch;
1461 ifd->ifi_lastchange = ifp->if_lastchange;
1463 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1464 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1465 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1466 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1467 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1468 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1469 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1470 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1471 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1472 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1473 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1474 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1478 * Wrapper functions for struct ifnet address list locking macros. These are
1479 * used by kernel modules to avoid encoding programming interface or binary
1480 * interface assumptions that may be violated when kernel-internal locking
1481 * approaches change.
1484 if_addr_rlock(struct ifnet *ifp)
1491 if_addr_runlock(struct ifnet *ifp)
1494 IF_ADDR_RUNLOCK(ifp);
1498 if_maddr_rlock(if_t ifp)
1501 IF_ADDR_RLOCK((struct ifnet *)ifp);
1505 if_maddr_runlock(if_t ifp)
1508 IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1512 * Initialization, destruction and refcounting functions for ifaddrs.
1515 ifa_alloc(size_t size, int flags)
1519 KASSERT(size >= sizeof(struct ifaddr),
1520 ("%s: invalid size %zu", __func__, size));
1522 ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1526 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1528 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1530 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1532 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1535 refcount_init(&ifa->ifa_refcnt, 1);
1540 /* free(NULL) is okay */
1541 counter_u64_free(ifa->ifa_opackets);
1542 counter_u64_free(ifa->ifa_ipackets);
1543 counter_u64_free(ifa->ifa_obytes);
1544 counter_u64_free(ifa->ifa_ibytes);
1545 free(ifa, M_IFADDR);
1551 ifa_ref(struct ifaddr *ifa)
1554 refcount_acquire(&ifa->ifa_refcnt);
1558 ifa_free(struct ifaddr *ifa)
1561 if (refcount_release(&ifa->ifa_refcnt)) {
1562 counter_u64_free(ifa->ifa_opackets);
1563 counter_u64_free(ifa->ifa_ipackets);
1564 counter_u64_free(ifa->ifa_obytes);
1565 counter_u64_free(ifa->ifa_ibytes);
1566 free(ifa, M_IFADDR);
1571 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1572 struct sockaddr *ia)
1575 struct rt_addrinfo info;
1576 struct sockaddr_dl null_sdl;
1581 bzero(&info, sizeof(info));
1582 if (cmd != RTM_DELETE)
1583 info.rti_ifp = V_loif;
1584 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1585 info.rti_info[RTAX_DST] = ia;
1586 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1587 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1589 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1592 log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1593 __func__, otype, if_name(ifp), error);
1599 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1602 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1606 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1609 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1613 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1616 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1620 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1621 * structs used to represent other address families, it is necessary
1622 * to perform a different comparison.
1625 #define sa_dl_equal(a1, a2) \
1626 ((((const struct sockaddr_dl *)(a1))->sdl_len == \
1627 ((const struct sockaddr_dl *)(a2))->sdl_len) && \
1628 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \
1629 CLLADDR((const struct sockaddr_dl *)(a2)), \
1630 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1633 * Locate an interface based on a complete address.
1636 static struct ifaddr *
1637 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1642 IFNET_RLOCK_NOSLEEP();
1643 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1645 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1646 if (ifa->ifa_addr->sa_family != addr->sa_family)
1648 if (sa_equal(addr, ifa->ifa_addr)) {
1651 IF_ADDR_RUNLOCK(ifp);
1654 /* IP6 doesn't have broadcast */
1655 if ((ifp->if_flags & IFF_BROADCAST) &&
1656 ifa->ifa_broadaddr &&
1657 ifa->ifa_broadaddr->sa_len != 0 &&
1658 sa_equal(ifa->ifa_broadaddr, addr)) {
1661 IF_ADDR_RUNLOCK(ifp);
1665 IF_ADDR_RUNLOCK(ifp);
1669 IFNET_RUNLOCK_NOSLEEP();
1674 ifa_ifwithaddr(const struct sockaddr *addr)
1677 return (ifa_ifwithaddr_internal(addr, 1));
1681 ifa_ifwithaddr_check(const struct sockaddr *addr)
1684 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1688 * Locate an interface based on the broadcast address.
1692 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1697 IFNET_RLOCK_NOSLEEP();
1698 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1699 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1702 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1703 if (ifa->ifa_addr->sa_family != addr->sa_family)
1705 if ((ifp->if_flags & IFF_BROADCAST) &&
1706 ifa->ifa_broadaddr &&
1707 ifa->ifa_broadaddr->sa_len != 0 &&
1708 sa_equal(ifa->ifa_broadaddr, addr)) {
1710 IF_ADDR_RUNLOCK(ifp);
1714 IF_ADDR_RUNLOCK(ifp);
1718 IFNET_RUNLOCK_NOSLEEP();
1723 * Locate the point to point interface with a given destination address.
1727 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1732 IFNET_RLOCK_NOSLEEP();
1733 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1734 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1736 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1739 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1740 if (ifa->ifa_addr->sa_family != addr->sa_family)
1742 if (ifa->ifa_dstaddr != NULL &&
1743 sa_equal(addr, ifa->ifa_dstaddr)) {
1745 IF_ADDR_RUNLOCK(ifp);
1749 IF_ADDR_RUNLOCK(ifp);
1753 IFNET_RUNLOCK_NOSLEEP();
1758 * Find an interface on a specific network. If many, choice
1759 * is most specific found.
1762 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1766 struct ifaddr *ifa_maybe = NULL;
1767 u_int af = addr->sa_family;
1768 const char *addr_data = addr->sa_data, *cplim;
1771 * AF_LINK addresses can be looked up directly by their index number,
1772 * so do that if we can.
1774 if (af == AF_LINK) {
1775 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
1776 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1777 return (ifaddr_byindex(sdl->sdl_index));
1781 * Scan though each interface, looking for ones that have addresses
1782 * in this address family and the requested fib. Maintain a reference
1783 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
1784 * kept it stable when we move onto the next interface.
1786 IFNET_RLOCK_NOSLEEP();
1787 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1788 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1791 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1792 const char *cp, *cp2, *cp3;
1794 if (ifa->ifa_addr->sa_family != af)
1796 if (af == AF_INET &&
1797 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1799 * This is a bit broken as it doesn't
1800 * take into account that the remote end may
1801 * be a single node in the network we are
1803 * The trouble is that we don't know the
1804 * netmask for the remote end.
1806 if (ifa->ifa_dstaddr != NULL &&
1807 sa_equal(addr, ifa->ifa_dstaddr)) {
1809 IF_ADDR_RUNLOCK(ifp);
1814 * Scan all the bits in the ifa's address.
1815 * If a bit dissagrees with what we are
1816 * looking for, mask it with the netmask
1817 * to see if it really matters.
1818 * (A byte at a time)
1820 if (ifa->ifa_netmask == 0)
1823 cp2 = ifa->ifa_addr->sa_data;
1824 cp3 = ifa->ifa_netmask->sa_data;
1825 cplim = ifa->ifa_netmask->sa_len
1826 + (char *)ifa->ifa_netmask;
1828 if ((*cp++ ^ *cp2++) & *cp3++)
1829 goto next; /* next address! */
1831 * If the netmask of what we just found
1832 * is more specific than what we had before
1833 * (if we had one), or if the virtual status
1834 * of new prefix is better than of the old one,
1835 * then remember the new one before continuing
1836 * to search for an even better one.
1838 if (ifa_maybe == NULL ||
1839 ifa_preferred(ifa_maybe, ifa) ||
1840 rn_refines((caddr_t)ifa->ifa_netmask,
1841 (caddr_t)ifa_maybe->ifa_netmask)) {
1842 if (ifa_maybe != NULL)
1843 ifa_free(ifa_maybe);
1849 IF_ADDR_RUNLOCK(ifp);
1854 IFNET_RUNLOCK_NOSLEEP();
1855 if (ifa_maybe != NULL)
1856 ifa_free(ifa_maybe);
1861 * Find an interface address specific to an interface best matching
1865 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1868 const char *cp, *cp2, *cp3;
1870 struct ifaddr *ifa_maybe = NULL;
1871 u_int af = addr->sa_family;
1876 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1877 if (ifa->ifa_addr->sa_family != af)
1879 if (ifa_maybe == NULL)
1881 if (ifa->ifa_netmask == 0) {
1882 if (sa_equal(addr, ifa->ifa_addr) ||
1883 (ifa->ifa_dstaddr &&
1884 sa_equal(addr, ifa->ifa_dstaddr)))
1888 if (ifp->if_flags & IFF_POINTOPOINT) {
1889 if (sa_equal(addr, ifa->ifa_dstaddr))
1893 cp2 = ifa->ifa_addr->sa_data;
1894 cp3 = ifa->ifa_netmask->sa_data;
1895 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1896 for (; cp3 < cplim; cp3++)
1897 if ((*cp++ ^ *cp2++) & *cp3)
1907 IF_ADDR_RUNLOCK(ifp);
1912 * See whether new ifa is better than current one:
1913 * 1) A non-virtual one is preferred over virtual.
1914 * 2) A virtual in master state preferred over any other state.
1916 * Used in several address selecting functions.
1919 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1922 return (cur->ifa_carp && (!next->ifa_carp ||
1923 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1926 #include <net/if_llatbl.h>
1929 * Default action when installing a route with a Link Level gateway.
1930 * Lookup an appropriate real ifa to point to.
1931 * This should be moved to /sys/net/link.c eventually.
1934 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1936 struct ifaddr *ifa, *oifa;
1937 struct sockaddr *dst;
1940 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1941 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1943 ifa = ifaof_ifpforaddr(dst, ifp);
1948 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1949 ifa->ifa_rtrequest(cmd, rt, info);
1953 struct sockaddr_dl *
1954 link_alloc_sdl(size_t size, int flags)
1957 return (malloc(size, M_TEMP, flags));
1961 link_free_sdl(struct sockaddr *sa)
1967 * Fills in given sdl with interface basic info.
1968 * Returns pointer to filled sdl.
1970 struct sockaddr_dl *
1971 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1973 struct sockaddr_dl *sdl;
1975 sdl = (struct sockaddr_dl *)paddr;
1976 memset(sdl, 0, sizeof(struct sockaddr_dl));
1977 sdl->sdl_len = sizeof(struct sockaddr_dl);
1978 sdl->sdl_family = AF_LINK;
1979 sdl->sdl_index = ifp->if_index;
1980 sdl->sdl_type = iftype;
1986 * Mark an interface down and notify protocols of
1990 if_unroute(struct ifnet *ifp, int flag, int fam)
1994 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1996 ifp->if_flags &= ~flag;
1997 getmicrotime(&ifp->if_lastchange);
1998 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1999 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2000 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2001 ifp->if_qflush(ifp);
2004 (*carp_linkstate_p)(ifp);
2009 * Mark an interface up and notify protocols of
2013 if_route(struct ifnet *ifp, int flag, int fam)
2017 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2019 ifp->if_flags |= flag;
2020 getmicrotime(&ifp->if_lastchange);
2021 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2022 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2023 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2025 (*carp_linkstate_p)(ifp);
2032 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2033 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2034 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2035 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2036 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2037 int (*vlan_setcookie_p)(struct ifnet *, void *);
2038 void *(*vlan_cookie_p)(struct ifnet *);
2041 * Handle a change in the interface link state. To avoid LORs
2042 * between driver lock and upper layer locks, as well as possible
2043 * recursions, we post event to taskqueue, and all job
2044 * is done in static do_link_state_change().
2047 if_link_state_change(struct ifnet *ifp, int link_state)
2049 /* Return if state hasn't changed. */
2050 if (ifp->if_link_state == link_state)
2053 ifp->if_link_state = link_state;
2055 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2059 do_link_state_change(void *arg, int pending)
2061 struct ifnet *ifp = (struct ifnet *)arg;
2062 int link_state = ifp->if_link_state;
2063 CURVNET_SET(ifp->if_vnet);
2065 /* Notify that the link state has changed. */
2067 if (ifp->if_vlantrunk != NULL)
2068 (*vlan_link_state_p)(ifp);
2070 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2071 ifp->if_l2com != NULL)
2072 (*ng_ether_link_state_p)(ifp, link_state);
2074 (*carp_linkstate_p)(ifp);
2076 (*bridge_linkstate_p)(ifp);
2078 (*lagg_linkstate_p)(ifp, link_state);
2080 if (IS_DEFAULT_VNET(curvnet))
2081 devctl_notify("IFNET", ifp->if_xname,
2082 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2085 if_printf(ifp, "%d link states coalesced\n", pending);
2086 if (log_link_state_change)
2087 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2088 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2089 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
2094 * Mark an interface down and notify protocols of
2098 if_down(struct ifnet *ifp)
2101 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2105 * Mark an interface up and notify protocols of
2109 if_up(struct ifnet *ifp)
2112 if_route(ifp, IFF_UP, AF_UNSPEC);
2116 * Flush an interface queue.
2119 if_qflush(struct ifnet *ifp)
2127 if (ALTQ_IS_ENABLED(ifq))
2131 while ((m = n) != 0) {
2142 * Map interface name to interface structure pointer, with or without
2143 * returning a reference.
2146 ifunit_ref(const char *name)
2150 IFNET_RLOCK_NOSLEEP();
2151 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2152 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2153 !(ifp->if_flags & IFF_DYING))
2158 IFNET_RUNLOCK_NOSLEEP();
2163 ifunit(const char *name)
2167 IFNET_RLOCK_NOSLEEP();
2168 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2169 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2172 IFNET_RUNLOCK_NOSLEEP();
2177 * Hardware specific interface ioctls.
2180 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2184 int new_flags, temp_flags;
2185 size_t namelen, onamelen;
2187 char *descrbuf, *odescrbuf;
2188 char new_name[IFNAMSIZ];
2190 struct sockaddr_dl *sdl;
2192 ifr = (struct ifreq *)data;
2195 ifr->ifr_index = ifp->if_index;
2199 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2200 ifr->ifr_flags = temp_flags & 0xffff;
2201 ifr->ifr_flagshigh = temp_flags >> 16;
2205 ifr->ifr_reqcap = ifp->if_capabilities;
2206 ifr->ifr_curcap = ifp->if_capenable;
2211 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2216 ifr->ifr_metric = ifp->if_metric;
2220 ifr->ifr_mtu = ifp->if_mtu;
2224 /* XXXGL: did this ever worked? */
2230 sx_slock(&ifdescr_sx);
2231 if (ifp->if_description == NULL)
2234 /* space for terminating nul */
2235 descrlen = strlen(ifp->if_description) + 1;
2236 if (ifr->ifr_buffer.length < descrlen)
2237 ifr->ifr_buffer.buffer = NULL;
2239 error = copyout(ifp->if_description,
2240 ifr->ifr_buffer.buffer, descrlen);
2241 ifr->ifr_buffer.length = descrlen;
2243 sx_sunlock(&ifdescr_sx);
2247 error = priv_check(td, PRIV_NET_SETIFDESCR);
2252 * Copy only (length-1) bytes to make sure that
2253 * if_description is always nul terminated. The
2254 * length parameter is supposed to count the
2255 * terminating nul in.
2257 if (ifr->ifr_buffer.length > ifdescr_maxlen)
2258 return (ENAMETOOLONG);
2259 else if (ifr->ifr_buffer.length == 0)
2262 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2264 error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2265 ifr->ifr_buffer.length - 1);
2267 free(descrbuf, M_IFDESCR);
2272 sx_xlock(&ifdescr_sx);
2273 odescrbuf = ifp->if_description;
2274 ifp->if_description = descrbuf;
2275 sx_xunlock(&ifdescr_sx);
2277 getmicrotime(&ifp->if_lastchange);
2278 free(odescrbuf, M_IFDESCR);
2282 ifr->ifr_fib = ifp->if_fib;
2286 error = priv_check(td, PRIV_NET_SETIFFIB);
2289 if (ifr->ifr_fib >= rt_numfibs)
2292 ifp->if_fib = ifr->ifr_fib;
2296 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2300 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2301 * check, so we don't need special handling here yet.
2303 new_flags = (ifr->ifr_flags & 0xffff) |
2304 (ifr->ifr_flagshigh << 16);
2305 if (ifp->if_flags & IFF_UP &&
2306 (new_flags & IFF_UP) == 0) {
2308 } else if (new_flags & IFF_UP &&
2309 (ifp->if_flags & IFF_UP) == 0) {
2312 /* See if permanently promiscuous mode bit is about to flip */
2313 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2314 if (new_flags & IFF_PPROMISC)
2315 ifp->if_flags |= IFF_PROMISC;
2316 else if (ifp->if_pcount == 0)
2317 ifp->if_flags &= ~IFF_PROMISC;
2318 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2320 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2322 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2323 (new_flags &~ IFF_CANTCHANGE);
2324 if (ifp->if_ioctl) {
2325 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2327 getmicrotime(&ifp->if_lastchange);
2331 error = priv_check(td, PRIV_NET_SETIFCAP);
2334 if (ifp->if_ioctl == NULL)
2335 return (EOPNOTSUPP);
2336 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2338 error = (*ifp->if_ioctl)(ifp, cmd, data);
2340 getmicrotime(&ifp->if_lastchange);
2345 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2350 error = priv_check(td, PRIV_NET_SETIFNAME);
2353 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2356 if (new_name[0] == '\0')
2358 if (ifunit(new_name) != NULL)
2362 * XXX: Locking. Nothing else seems to lock if_flags,
2363 * and there are numerous other races with the
2364 * ifunit() checks not being atomic with namespace
2365 * changes (renames, vmoves, if_attach, etc).
2367 ifp->if_flags |= IFF_RENAMING;
2369 /* Announce the departure of the interface. */
2370 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2371 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2373 log(LOG_INFO, "%s: changing name to '%s'\n",
2374 ifp->if_xname, new_name);
2377 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2379 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2380 namelen = strlen(new_name);
2381 onamelen = sdl->sdl_nlen;
2383 * Move the address if needed. This is safe because we
2384 * allocate space for a name of length IFNAMSIZ when we
2385 * create this in if_attach().
2387 if (namelen != onamelen) {
2388 bcopy(sdl->sdl_data + onamelen,
2389 sdl->sdl_data + namelen, sdl->sdl_alen);
2391 bcopy(new_name, sdl->sdl_data, namelen);
2392 sdl->sdl_nlen = namelen;
2393 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2394 bzero(sdl->sdl_data, onamelen);
2395 while (namelen != 0)
2396 sdl->sdl_data[--namelen] = 0xff;
2397 IF_ADDR_WUNLOCK(ifp);
2399 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2400 /* Announce the return of the interface. */
2401 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2403 ifp->if_flags &= ~IFF_RENAMING;
2408 error = priv_check(td, PRIV_NET_SETIFVNET);
2411 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2416 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2419 ifp->if_metric = ifr->ifr_metric;
2420 getmicrotime(&ifp->if_lastchange);
2424 error = priv_check(td, PRIV_NET_SETIFPHYS);
2427 if (ifp->if_ioctl == NULL)
2428 return (EOPNOTSUPP);
2429 error = (*ifp->if_ioctl)(ifp, cmd, data);
2431 getmicrotime(&ifp->if_lastchange);
2436 u_long oldmtu = ifp->if_mtu;
2438 error = priv_check(td, PRIV_NET_SETIFMTU);
2441 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2443 if (ifp->if_ioctl == NULL)
2444 return (EOPNOTSUPP);
2445 error = (*ifp->if_ioctl)(ifp, cmd, data);
2447 getmicrotime(&ifp->if_lastchange);
2451 * If the link MTU changed, do network layer specific procedure.
2453 if (ifp->if_mtu != oldmtu) {
2464 if (cmd == SIOCADDMULTI)
2465 error = priv_check(td, PRIV_NET_ADDMULTI);
2467 error = priv_check(td, PRIV_NET_DELMULTI);
2471 /* Don't allow group membership on non-multicast interfaces. */
2472 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2473 return (EOPNOTSUPP);
2475 /* Don't let users screw up protocols' entries. */
2476 if (ifr->ifr_addr.sa_family != AF_LINK)
2479 if (cmd == SIOCADDMULTI) {
2480 struct ifmultiaddr *ifma;
2483 * Userland is only permitted to join groups once
2484 * via the if_addmulti() KPI, because it cannot hold
2485 * struct ifmultiaddr * between calls. It may also
2486 * lose a race while we check if the membership
2490 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2491 IF_ADDR_RUNLOCK(ifp);
2495 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2497 error = if_delmulti(ifp, &ifr->ifr_addr);
2500 getmicrotime(&ifp->if_lastchange);
2503 case SIOCSIFPHYADDR:
2504 case SIOCDIFPHYADDR:
2506 case SIOCSIFPHYADDR_IN6:
2509 case SIOCSIFGENERIC:
2510 error = priv_check(td, PRIV_NET_HWIOCTL);
2513 if (ifp->if_ioctl == NULL)
2514 return (EOPNOTSUPP);
2515 error = (*ifp->if_ioctl)(ifp, cmd, data);
2517 getmicrotime(&ifp->if_lastchange);
2521 case SIOCGIFPSRCADDR:
2522 case SIOCGIFPDSTADDR:
2525 case SIOCGIFGENERIC:
2526 if (ifp->if_ioctl == NULL)
2527 return (EOPNOTSUPP);
2528 error = (*ifp->if_ioctl)(ifp, cmd, data);
2532 error = priv_check(td, PRIV_NET_SETLLADDR);
2535 error = if_setlladdr(ifp,
2536 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2541 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2543 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2546 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2552 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2558 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2560 error = priv_check(td, PRIV_NET_DELIFGROUP);
2563 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2575 #ifdef COMPAT_FREEBSD32
2583 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2590 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2597 CURVNET_SET(so->so_vnet);
2600 error = ifconf(cmd, data);
2604 #ifdef COMPAT_FREEBSD32
2607 struct ifconf32 *ifc32;
2610 ifc32 = (struct ifconf32 *)data;
2611 ifc.ifc_len = ifc32->ifc_len;
2612 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2614 error = ifconf(SIOCGIFCONF, (void *)&ifc);
2617 ifc32->ifc_len = ifc.ifc_len;
2622 ifr = (struct ifreq *)data;
2627 error = priv_check(td, PRIV_NET_SETIFVNET);
2629 error = if_vmove_reclaim(td, ifr->ifr_name,
2636 error = priv_check(td, PRIV_NET_IFCREATE);
2638 error = if_clone_create(ifr->ifr_name,
2639 sizeof(ifr->ifr_name),
2640 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2644 error = priv_check(td, PRIV_NET_IFDESTROY);
2646 error = if_clone_destroy(ifr->ifr_name);
2650 case SIOCIFGCLONERS:
2651 error = if_clone_list((struct if_clonereq *)data);
2655 error = if_getgroupmembers((struct ifgroupreq *)data);
2658 #if defined(INET) || defined(INET6)
2661 if (carp_ioctl_p == NULL)
2662 error = EPROTONOSUPPORT;
2664 error = (*carp_ioctl_p)(ifr, cmd, td);
2670 ifp = ifunit_ref(ifr->ifr_name);
2676 error = ifhwioctl(cmd, ifp, data, td);
2677 if (error != ENOIOCTL) {
2683 oif_flags = ifp->if_flags;
2684 if (so->so_proto == NULL) {
2687 return (EOPNOTSUPP);
2691 * Pass the request on to the socket control method, and if the
2692 * latter returns EOPNOTSUPP, directly to the interface.
2694 * Make an exception for the legacy SIOCSIF* requests. Drivers
2695 * trust SIOCSIFADDR et al to come from an already privileged
2696 * layer, and do not perform any credentials checks or input
2699 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
2701 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2702 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2703 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2704 error = (*ifp->if_ioctl)(ifp, cmd, data);
2706 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2708 if (ifp->if_flags & IFF_UP)
2718 * The code common to handling reference counted flags,
2719 * e.g., in ifpromisc() and if_allmulti().
2720 * The "pflag" argument can specify a permanent mode flag to check,
2721 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2723 * Only to be used on stack-owned flags, not driver-owned flags.
2726 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2730 int oldflags, oldcount;
2732 /* Sanity checks to catch programming errors */
2733 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2734 ("%s: setting driver-owned flag %d", __func__, flag));
2737 KASSERT(*refcount >= 0,
2738 ("%s: increment negative refcount %d for flag %d",
2739 __func__, *refcount, flag));
2741 KASSERT(*refcount > 0,
2742 ("%s: decrement non-positive refcount %d for flag %d",
2743 __func__, *refcount, flag));
2745 /* In case this mode is permanent, just touch refcount */
2746 if (ifp->if_flags & pflag) {
2747 *refcount += onswitch ? 1 : -1;
2751 /* Save ifnet parameters for if_ioctl() may fail */
2752 oldcount = *refcount;
2753 oldflags = ifp->if_flags;
2756 * See if we aren't the only and touching refcount is enough.
2757 * Actually toggle interface flag if we are the first or last.
2762 ifp->if_flags |= flag;
2766 ifp->if_flags &= ~flag;
2769 /* Call down the driver since we've changed interface flags */
2770 if (ifp->if_ioctl == NULL) {
2774 ifr.ifr_flags = ifp->if_flags & 0xffff;
2775 ifr.ifr_flagshigh = ifp->if_flags >> 16;
2776 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2779 /* Notify userland that interface flags have changed */
2784 /* Recover after driver error */
2785 *refcount = oldcount;
2786 ifp->if_flags = oldflags;
2791 * Set/clear promiscuous mode on interface ifp based on the truth value
2792 * of pswitch. The calls are reference counted so that only the first
2793 * "on" request actually has an effect, as does the final "off" request.
2794 * Results are undefined if the "off" and "on" requests are not matched.
2797 ifpromisc(struct ifnet *ifp, int pswitch)
2800 int oldflags = ifp->if_flags;
2802 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2803 &ifp->if_pcount, pswitch);
2804 /* If promiscuous mode status has changed, log a message */
2805 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2806 log(LOG_INFO, "%s: promiscuous mode %s\n",
2808 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2813 * Return interface configuration
2814 * of system. List may be used
2815 * in later ioctl's (above) to get
2816 * other information.
2820 ifconf(u_long cmd, caddr_t data)
2822 struct ifconf *ifc = (struct ifconf *)data;
2827 int error, full = 0, valid_len, max_len;
2829 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2830 max_len = MAXPHYS - 1;
2832 /* Prevent hostile input from being able to crash the system */
2833 if (ifc->ifc_len <= 0)
2837 if (ifc->ifc_len <= max_len) {
2838 max_len = ifc->ifc_len;
2841 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2846 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2850 * Zero the ifr_name buffer to make sure we don't
2851 * disclose the contents of the stack.
2853 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2855 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2856 >= sizeof(ifr.ifr_name)) {
2859 return (ENAMETOOLONG);
2864 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2865 struct sockaddr *sa = ifa->ifa_addr;
2867 if (prison_if(curthread->td_ucred, sa) != 0)
2870 if (sa->sa_len <= sizeof(*sa)) {
2872 sbuf_bcat(sb, &ifr, sizeof(ifr));
2873 max_len += sizeof(ifr);
2876 offsetof(struct ifreq, ifr_addr));
2877 max_len += offsetof(struct ifreq, ifr_addr);
2878 sbuf_bcat(sb, sa, sa->sa_len);
2879 max_len += sa->sa_len;
2882 if (sbuf_error(sb) == 0)
2883 valid_len = sbuf_len(sb);
2885 IF_ADDR_RUNLOCK(ifp);
2887 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2888 sbuf_bcat(sb, &ifr, sizeof(ifr));
2889 max_len += sizeof(ifr);
2891 if (sbuf_error(sb) == 0)
2892 valid_len = sbuf_len(sb);
2898 * If we didn't allocate enough space (uncommon), try again. If
2899 * we have already allocated as much space as we are allowed,
2900 * return what we've got.
2902 if (valid_len != max_len && !full) {
2907 ifc->ifc_len = valid_len;
2909 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2915 * Just like ifpromisc(), but for all-multicast-reception mode.
2918 if_allmulti(struct ifnet *ifp, int onswitch)
2921 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2924 struct ifmultiaddr *
2925 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
2927 struct ifmultiaddr *ifma;
2929 IF_ADDR_LOCK_ASSERT(ifp);
2931 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2932 if (sa->sa_family == AF_LINK) {
2933 if (sa_dl_equal(ifma->ifma_addr, sa))
2936 if (sa_equal(ifma->ifma_addr, sa))
2945 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
2946 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
2947 * the ifnet multicast address list here, so the caller must do that and
2948 * other setup work (such as notifying the device driver). The reference
2949 * count is initialized to 1.
2951 static struct ifmultiaddr *
2952 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2955 struct ifmultiaddr *ifma;
2956 struct sockaddr *dupsa;
2958 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2963 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2964 if (dupsa == NULL) {
2965 free(ifma, M_IFMADDR);
2968 bcopy(sa, dupsa, sa->sa_len);
2969 ifma->ifma_addr = dupsa;
2971 ifma->ifma_ifp = ifp;
2972 ifma->ifma_refcount = 1;
2973 ifma->ifma_protospec = NULL;
2976 ifma->ifma_lladdr = NULL;
2980 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2981 if (dupsa == NULL) {
2982 free(ifma->ifma_addr, M_IFMADDR);
2983 free(ifma, M_IFMADDR);
2986 bcopy(llsa, dupsa, llsa->sa_len);
2987 ifma->ifma_lladdr = dupsa;
2993 * if_freemulti: free ifmultiaddr structure and possibly attached related
2994 * addresses. The caller is responsible for implementing reference
2995 * counting, notifying the driver, handling routing messages, and releasing
2996 * any dependent link layer state.
2999 if_freemulti(struct ifmultiaddr *ifma)
3002 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3003 ifma->ifma_refcount));
3005 if (ifma->ifma_lladdr != NULL)
3006 free(ifma->ifma_lladdr, M_IFMADDR);
3007 free(ifma->ifma_addr, M_IFMADDR);
3008 free(ifma, M_IFMADDR);
3012 * Register an additional multicast address with a network interface.
3014 * - If the address is already present, bump the reference count on the
3015 * address and return.
3016 * - If the address is not link-layer, look up a link layer address.
3017 * - Allocate address structures for one or both addresses, and attach to the
3018 * multicast address list on the interface. If automatically adding a link
3019 * layer address, the protocol address will own a reference to the link
3020 * layer address, to be freed when it is freed.
3021 * - Notify the network device driver of an addition to the multicast address
3024 * 'sa' points to caller-owned memory with the desired multicast address.
3026 * 'retifma' will be used to return a pointer to the resulting multicast
3027 * address reference, if desired.
3030 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3031 struct ifmultiaddr **retifma)
3033 struct ifmultiaddr *ifma, *ll_ifma;
3034 struct sockaddr *llsa;
3035 struct sockaddr_dl sdl;
3039 * If the address is already present, return a new reference to it;
3040 * otherwise, allocate storage and set up a new address.
3043 ifma = if_findmulti(ifp, sa);
3045 ifma->ifma_refcount++;
3046 if (retifma != NULL)
3048 IF_ADDR_WUNLOCK(ifp);
3053 * The address isn't already present; resolve the protocol address
3054 * into a link layer address, and then look that up, bump its
3055 * refcount or allocate an ifma for that also.
3056 * Most link layer resolving functions returns address data which
3057 * fits inside default sockaddr_dl structure. However callback
3058 * can allocate another sockaddr structure, in that case we need to
3063 if (ifp->if_resolvemulti != NULL) {
3064 /* Provide called function with buffer size information */
3065 sdl.sdl_len = sizeof(sdl);
3066 llsa = (struct sockaddr *)&sdl;
3067 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3073 * Allocate the new address. Don't hook it up yet, as we may also
3074 * need to allocate a link layer multicast address.
3076 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3083 * If a link layer address is found, we'll need to see if it's
3084 * already present in the address list, or allocate is as well.
3085 * When this block finishes, the link layer address will be on the
3089 ll_ifma = if_findmulti(ifp, llsa);
3090 if (ll_ifma == NULL) {
3091 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3092 if (ll_ifma == NULL) {
3093 --ifma->ifma_refcount;
3098 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3101 ll_ifma->ifma_refcount++;
3102 ifma->ifma_llifma = ll_ifma;
3106 * We now have a new multicast address, ifma, and possibly a new or
3107 * referenced link layer address. Add the primary address to the
3108 * ifnet address list.
3110 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3112 if (retifma != NULL)
3116 * Must generate the message while holding the lock so that 'ifma'
3117 * pointer is still valid.
3119 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3120 IF_ADDR_WUNLOCK(ifp);
3123 * We are certain we have added something, so call down to the
3124 * interface to let them know about it.
3126 if (ifp->if_ioctl != NULL) {
3127 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3130 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3131 link_free_sdl(llsa);
3136 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3137 link_free_sdl(llsa);
3140 IF_ADDR_WUNLOCK(ifp);
3145 * Delete a multicast group membership by network-layer group address.
3147 * Returns ENOENT if the entry could not be found. If ifp no longer
3148 * exists, results are undefined. This entry point should only be used
3149 * from subsystems which do appropriate locking to hold ifp for the
3150 * duration of the call.
3151 * Network-layer protocol domains must use if_delmulti_ifma().
3154 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3156 struct ifmultiaddr *ifma;
3161 IFNET_RLOCK_NOSLEEP();
3162 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3167 IFNET_RUNLOCK_NOSLEEP();
3169 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3176 ifma = if_findmulti(ifp, sa);
3178 lastref = if_delmulti_locked(ifp, ifma, 0);
3179 IF_ADDR_WUNLOCK(ifp);
3184 if (lastref && ifp->if_ioctl != NULL) {
3185 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3192 * Delete all multicast group membership for an interface.
3193 * Should be used to quickly flush all multicast filters.
3196 if_delallmulti(struct ifnet *ifp)
3198 struct ifmultiaddr *ifma;
3199 struct ifmultiaddr *next;
3202 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3203 if_delmulti_locked(ifp, ifma, 0);
3204 IF_ADDR_WUNLOCK(ifp);
3208 * Delete a multicast group membership by group membership pointer.
3209 * Network-layer protocol domains must use this routine.
3211 * It is safe to call this routine if the ifp disappeared.
3214 if_delmulti_ifma(struct ifmultiaddr *ifma)
3219 ifp = ifma->ifma_ifp;
3222 printf("%s: ifma_ifp seems to be detached\n", __func__);
3226 IFNET_RLOCK_NOSLEEP();
3227 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3231 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3234 IFNET_RUNLOCK_NOSLEEP();
3238 * If and only if the ifnet instance exists: Acquire the address lock.
3243 lastref = if_delmulti_locked(ifp, ifma, 0);
3247 * If and only if the ifnet instance exists:
3248 * Release the address lock.
3249 * If the group was left: update the hardware hash filter.
3251 IF_ADDR_WUNLOCK(ifp);
3252 if (lastref && ifp->if_ioctl != NULL) {
3253 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3259 * Perform deletion of network-layer and/or link-layer multicast address.
3261 * Return 0 if the reference count was decremented.
3262 * Return 1 if the final reference was released, indicating that the
3263 * hardware hash filter should be reprogrammed.
3266 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3268 struct ifmultiaddr *ll_ifma;
3270 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3271 KASSERT(ifma->ifma_ifp == ifp,
3272 ("%s: inconsistent ifp %p", __func__, ifp));
3273 IF_ADDR_WLOCK_ASSERT(ifp);
3276 ifp = ifma->ifma_ifp;
3279 * If the ifnet is detaching, null out references to ifnet,
3280 * so that upper protocol layers will notice, and not attempt
3281 * to obtain locks for an ifnet which no longer exists. The
3282 * routing socket announcement must happen before the ifnet
3283 * instance is detached from the system.
3287 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3290 * ifp may already be nulled out if we are being reentered
3291 * to delete the ll_ifma.
3294 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3295 ifma->ifma_ifp = NULL;
3299 if (--ifma->ifma_refcount > 0)
3303 * If this ifma is a network-layer ifma, a link-layer ifma may
3304 * have been associated with it. Release it first if so.
3306 ll_ifma = ifma->ifma_llifma;
3307 if (ll_ifma != NULL) {
3308 KASSERT(ifma->ifma_lladdr != NULL,
3309 ("%s: llifma w/o lladdr", __func__));
3311 ll_ifma->ifma_ifp = NULL; /* XXX */
3312 if (--ll_ifma->ifma_refcount == 0) {
3314 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3317 if_freemulti(ll_ifma);
3322 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3327 * The last reference to this instance of struct ifmultiaddr
3328 * was released; the hardware should be notified of this change.
3334 * Set the link layer address on an interface.
3336 * At this time we only support certain types of interfaces,
3337 * and we don't allow the length of the address to change.
3339 * Set noinline to be dtrace-friendly
3342 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3344 struct sockaddr_dl *sdl;
3351 IF_ADDR_RUNLOCK(ifp);
3355 IF_ADDR_RUNLOCK(ifp);
3356 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3361 if (len != sdl->sdl_alen) { /* don't allow length to change */
3365 switch (ifp->if_type) {
3373 case IFT_IEEE8023ADLAG:
3375 bcopy(lladdr, LLADDR(sdl), len);
3384 * If the interface is already up, we need
3385 * to re-init it in order to reprogram its
3388 if ((ifp->if_flags & IFF_UP) != 0) {
3389 if (ifp->if_ioctl) {
3390 ifp->if_flags &= ~IFF_UP;
3391 ifr.ifr_flags = ifp->if_flags & 0xffff;
3392 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3393 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3394 ifp->if_flags |= IFF_UP;
3395 ifr.ifr_flags = ifp->if_flags & 0xffff;
3396 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3397 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3400 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3405 * Compat function for handling basic encapsulation requests.
3406 * Not converted stacks (FDDI, IB, ..) supports traditional
3407 * output model: ARP (and other similar L2 protocols) are handled
3408 * inside output routine, arpresolve/nd6_resolve() returns MAC
3409 * address instead of full prepend.
3411 * This function creates calculated header==MAC for IPv4/IPv6 and
3412 * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3416 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3419 if (req->rtype != IFENCAP_LL)
3420 return (EOPNOTSUPP);
3422 if (req->bufsize < req->lladdr_len)
3425 switch (req->family) {
3430 return (EAFNOSUPPORT);
3433 /* Copy lladdr to storage as is */
3434 memmove(req->buf, req->lladdr, req->lladdr_len);
3435 req->bufsize = req->lladdr_len;
3436 req->lladdr_off = 0;
3442 * The name argument must be a pointer to storage which will last as
3443 * long as the interface does. For physical devices, the result of
3444 * device_get_name(dev) is a good choice and for pseudo-devices a
3445 * static string works well.
3448 if_initname(struct ifnet *ifp, const char *name, int unit)
3450 ifp->if_dname = name;
3451 ifp->if_dunit = unit;
3452 if (unit != IF_DUNIT_NONE)
3453 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3455 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3459 if_printf(struct ifnet *ifp, const char * fmt, ...)
3464 retval = printf("%s: ", ifp->if_xname);
3466 retval += vprintf(fmt, ap);
3472 if_start(struct ifnet *ifp)
3475 (*(ifp)->if_start)(ifp);
3479 * Backwards compatibility interface for drivers
3480 * that have not implemented it
3483 if_transmit(struct ifnet *ifp, struct mbuf *m)
3487 IFQ_HANDOFF(ifp, m, error);
3492 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3499 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3504 if (_IF_QFULL(ifq)) {
3506 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3511 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3512 if (m->m_flags & (M_BCAST|M_MCAST))
3513 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3514 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3516 _IF_ENQUEUE(ifq, m);
3518 if (ifp != NULL && !active)
3519 (*(ifp)->if_start)(ifp);
3524 if_register_com_alloc(u_char type,
3525 if_com_alloc_t *a, if_com_free_t *f)
3528 KASSERT(if_com_alloc[type] == NULL,
3529 ("if_register_com_alloc: %d already registered", type));
3530 KASSERT(if_com_free[type] == NULL,
3531 ("if_register_com_alloc: %d free already registered", type));
3533 if_com_alloc[type] = a;
3534 if_com_free[type] = f;
3538 if_deregister_com_alloc(u_char type)
3541 KASSERT(if_com_alloc[type] != NULL,
3542 ("if_deregister_com_alloc: %d not registered", type));
3543 KASSERT(if_com_free[type] != NULL,
3544 ("if_deregister_com_alloc: %d free not registered", type));
3545 if_com_alloc[type] = NULL;
3546 if_com_free[type] = NULL;
3549 /* API for driver access to network stack owned ifnet.*/
3551 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3555 oldbrate = ifp->if_baudrate;
3556 ifp->if_baudrate = baudrate;
3561 if_getbaudrate(if_t ifp)
3564 return (((struct ifnet *)ifp)->if_baudrate);
3568 if_setcapabilities(if_t ifp, int capabilities)
3570 ((struct ifnet *)ifp)->if_capabilities = capabilities;
3575 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
3577 ((struct ifnet *)ifp)->if_capabilities |= setbit;
3578 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
3584 if_getcapabilities(if_t ifp)
3586 return ((struct ifnet *)ifp)->if_capabilities;
3590 if_setcapenable(if_t ifp, int capabilities)
3592 ((struct ifnet *)ifp)->if_capenable = capabilities;
3597 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
3600 ((struct ifnet *)ifp)->if_capenable |= setcap;
3602 ((struct ifnet *)ifp)->if_capenable &= ~clearcap;
3608 if_getdname(if_t ifp)
3610 return ((struct ifnet *)ifp)->if_dname;
3614 if_togglecapenable(if_t ifp, int togglecap)
3616 ((struct ifnet *)ifp)->if_capenable ^= togglecap;
3621 if_getcapenable(if_t ifp)
3623 return ((struct ifnet *)ifp)->if_capenable;
3627 * This is largely undesirable because it ties ifnet to a device, but does
3628 * provide flexiblity for an embedded product vendor. Should be used with
3629 * the understanding that it violates the interface boundaries, and should be
3630 * a last resort only.
3633 if_setdev(if_t ifp, void *dev)
3639 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
3641 ((struct ifnet *)ifp)->if_drv_flags |= set_flags;
3642 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
3648 if_getdrvflags(if_t ifp)
3650 return ((struct ifnet *)ifp)->if_drv_flags;
3654 if_setdrvflags(if_t ifp, int flags)
3656 ((struct ifnet *)ifp)->if_drv_flags = flags;
3662 if_setflags(if_t ifp, int flags)
3664 ((struct ifnet *)ifp)->if_flags = flags;
3669 if_setflagbits(if_t ifp, int set, int clear)
3671 ((struct ifnet *)ifp)->if_flags |= set;
3672 ((struct ifnet *)ifp)->if_flags &= ~clear;
3678 if_getflags(if_t ifp)
3680 return ((struct ifnet *)ifp)->if_flags;
3684 if_clearhwassist(if_t ifp)
3686 ((struct ifnet *)ifp)->if_hwassist = 0;
3691 if_sethwassistbits(if_t ifp, int toset, int toclear)
3693 ((struct ifnet *)ifp)->if_hwassist |= toset;
3694 ((struct ifnet *)ifp)->if_hwassist &= ~toclear;
3700 if_sethwassist(if_t ifp, int hwassist_bit)
3702 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
3707 if_gethwassist(if_t ifp)
3709 return ((struct ifnet *)ifp)->if_hwassist;
3713 if_setmtu(if_t ifp, int mtu)
3715 ((struct ifnet *)ifp)->if_mtu = mtu;
3722 return ((struct ifnet *)ifp)->if_mtu;
3726 if_getmtu_family(if_t ifp, int family)
3730 for (dp = domains; dp; dp = dp->dom_next) {
3731 if (dp->dom_family == family && dp->dom_ifmtu != NULL)
3732 return (dp->dom_ifmtu((struct ifnet *)ifp));
3735 return (((struct ifnet *)ifp)->if_mtu);
3739 if_setsoftc(if_t ifp, void *softc)
3741 ((struct ifnet *)ifp)->if_softc = softc;
3746 if_getsoftc(if_t ifp)
3748 return ((struct ifnet *)ifp)->if_softc;
3752 if_setrcvif(struct mbuf *m, if_t ifp)
3754 m->m_pkthdr.rcvif = (struct ifnet *)ifp;
3758 if_setvtag(struct mbuf *m, uint16_t tag)
3760 m->m_pkthdr.ether_vtag = tag;
3764 if_getvtag(struct mbuf *m)
3767 return (m->m_pkthdr.ether_vtag);
3771 if_sendq_empty(if_t ifp)
3773 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
3777 if_getifaddr(if_t ifp)
3779 return ((struct ifnet *)ifp)->if_addr;
3783 if_getamcount(if_t ifp)
3785 return ((struct ifnet *)ifp)->if_amcount;
3790 if_setsendqready(if_t ifp)
3792 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
3797 if_setsendqlen(if_t ifp, int tx_desc_count)
3799 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
3800 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
3806 if_vlantrunkinuse(if_t ifp)
3808 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
3812 if_input(if_t ifp, struct mbuf* sendmp)
3814 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
3820 #ifndef ETH_ADDR_LEN
3821 #define ETH_ADDR_LEN 6
3825 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
3827 struct ifmultiaddr *ifma;
3828 uint8_t *lmta = (uint8_t *)mta;
3831 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3832 if (ifma->ifma_addr->sa_family != AF_LINK)
3838 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
3839 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
3848 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
3852 if_maddr_rlock(ifp);
3853 error = if_setupmultiaddr(ifp, mta, cnt, max);
3854 if_maddr_runlock(ifp);
3859 if_multiaddr_count(if_t ifp, int max)
3861 struct ifmultiaddr *ifma;
3865 if_maddr_rlock(ifp);
3866 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
3867 if (ifma->ifma_addr->sa_family != AF_LINK)
3873 if_maddr_runlock(ifp);
3878 if_dequeue(if_t ifp)
3881 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
3887 if_sendq_prepend(if_t ifp, struct mbuf *m)
3889 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
3894 if_setifheaderlen(if_t ifp, int len)
3896 ((struct ifnet *)ifp)->if_hdrlen = len;
3901 if_getlladdr(if_t ifp)
3903 return (IF_LLADDR((struct ifnet *)ifp));
3907 if_gethandle(u_char type)
3909 return (if_alloc(type));
3913 if_bpfmtap(if_t ifh, struct mbuf *m)
3915 struct ifnet *ifp = (struct ifnet *)ifh;
3921 if_etherbpfmtap(if_t ifh, struct mbuf *m)
3923 struct ifnet *ifp = (struct ifnet *)ifh;
3925 ETHER_BPF_MTAP(ifp, m);
3929 if_vlancap(if_t ifh)
3931 struct ifnet *ifp = (struct ifnet *)ifh;
3932 VLAN_CAPABILITIES(ifp);
3936 if_setinitfn(if_t ifp, void (*init_fn)(void *))
3938 ((struct ifnet *)ifp)->if_init = init_fn;
3942 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
3944 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
3948 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
3950 ((struct ifnet *)ifp)->if_start = (void *)start_fn;
3954 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
3956 ((struct ifnet *)ifp)->if_transmit = start_fn;
3959 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
3961 ((struct ifnet *)ifp)->if_qflush = flush_fn;
3966 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
3969 ifp->if_get_counter = fn;
3972 /* Revisit these - These are inline functions originally. */
3974 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
3976 return drbr_inuse(ifh, br);
3980 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
3982 return drbr_dequeue(ifh, br);
3986 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
3988 return drbr_needs_enqueue(ifh, br);
3992 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
3994 return drbr_enqueue(ifh, br, m);