2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1980, 1986, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)if.c 8.5 (Berkeley) 1/9/95
35 #include "opt_inet6.h"
38 #include <sys/param.h>
40 #include <sys/eventhandler.h>
41 #include <sys/malloc.h>
42 #include <sys/domainset.h>
45 #include <sys/epoch.h>
47 #include <sys/systm.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/protosw.h>
53 #include <sys/kernel.h>
55 #include <sys/refcount.h>
56 #include <sys/module.h>
57 #include <sys/rwlock.h>
58 #include <sys/sockio.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/taskqueue.h>
63 #include <sys/domain.h>
67 #include <machine/stdarg.h>
71 #include <net/ethernet.h>
73 #include <net/if_arp.h>
74 #include <net/if_clone.h>
75 #include <net/if_dl.h>
76 #include <net/if_types.h>
77 #include <net/if_var.h>
78 #include <net/if_media.h>
79 #include <net/if_vlan_var.h>
80 #include <net/radix.h>
81 #include <net/route.h>
84 #if defined(INET) || defined(INET6)
85 #include <net/ethernet.h>
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_carp.h>
91 #include <netinet/if_ether.h>
92 #include <netinet/netdump/netdump.h>
95 #include <netinet6/in6_var.h>
96 #include <netinet6/in6_ifattach.h>
98 #endif /* INET || INET6 */
100 #include <security/mac/mac_framework.h>
103 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
104 * and ifr_ifru when it is used in SIOCGIFCONF.
106 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
107 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
109 __read_mostly epoch_t net_epoch_preempt;
110 __read_mostly epoch_t net_epoch;
111 #ifdef COMPAT_FREEBSD32
112 #include <sys/mount.h>
113 #include <compat/freebsd32/freebsd32.h>
115 struct ifreq_buffer32 {
116 uint32_t length; /* (size_t) */
117 uint32_t buffer; /* (void *) */
121 * Interface request structure used for socket
122 * ioctl's. All interface ioctl's must have parameter
123 * definitions which begin with ifr_name. The
124 * remainder may be interface specific.
127 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */
129 struct sockaddr ifru_addr;
130 struct sockaddr ifru_dstaddr;
131 struct sockaddr ifru_broadaddr;
132 struct ifreq_buffer32 ifru_buffer;
143 u_char ifru_vlan_pcp;
146 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
147 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
148 __offsetof(struct ifreq32, ifr_ifru));
150 struct ifgroupreq32 {
151 char ifgr_name[IFNAMSIZ];
154 char ifgru_group[IFNAMSIZ];
155 uint32_t ifgru_groups;
159 struct ifmediareq32 {
160 char ifm_name[IFNAMSIZ];
166 uint32_t ifm_ulist; /* (int *) */
168 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
169 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
171 #define _CASE_IOC_IFGROUPREQ_32(cmd) \
172 _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
173 #else /* !COMPAT_FREEBSD32 */
174 #define _CASE_IOC_IFGROUPREQ_32(cmd)
175 #endif /* !COMPAT_FREEBSD32 */
177 #define CASE_IOC_IFGROUPREQ(cmd) \
178 _CASE_IOC_IFGROUPREQ_32(cmd) \
183 #ifdef COMPAT_FREEBSD32
184 struct ifreq32 ifr32;
188 union ifgroupreq_union {
189 struct ifgroupreq ifgr;
190 #ifdef COMPAT_FREEBSD32
191 struct ifgroupreq32 ifgr32;
195 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
196 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
198 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
199 &ifqmaxlen, 0, "max send queue size");
201 /* Log link state change events */
202 static int log_link_state_change = 1;
204 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
205 &log_link_state_change, 0,
206 "log interface link state change events");
208 /* Log promiscuous mode change events */
209 static int log_promisc_mode_change = 1;
211 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
212 &log_promisc_mode_change, 1,
213 "log promiscuous mode change events");
215 /* Interface description */
216 static unsigned int ifdescr_maxlen = 1024;
217 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
219 "administrative maximum length for interface description");
221 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
223 /* global sx for non-critical path ifdescr */
224 static struct sx ifdescr_sx;
225 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
227 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
228 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
229 /* These are external hooks for CARP. */
230 void (*carp_linkstate_p)(struct ifnet *ifp);
231 void (*carp_demote_adj_p)(int, char *);
232 int (*carp_master_p)(struct ifaddr *);
233 #if defined(INET) || defined(INET6)
234 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
235 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
236 const struct sockaddr *sa);
237 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
238 int (*carp_attach_p)(struct ifaddr *, int);
239 void (*carp_detach_p)(struct ifaddr *, bool);
242 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
245 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
246 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
247 const struct in6_addr *taddr);
250 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
253 * XXX: Style; these should be sorted alphabetically, and unprototyped
254 * static functions should be prototyped. Currently they are sorted by
257 static void if_attachdomain(void *);
258 static void if_attachdomain1(struct ifnet *);
259 static int ifconf(u_long, caddr_t);
260 static void *if_grow(void);
261 static void if_input_default(struct ifnet *, struct mbuf *);
262 static int if_requestencap_default(struct ifnet *, struct if_encap_req *);
263 static void if_route(struct ifnet *, int flag, int fam);
264 static int if_setflag(struct ifnet *, int, int, int *, int);
265 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
266 static void if_unroute(struct ifnet *, int flag, int fam);
267 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void do_link_state_change(void *, int);
269 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int if_getgroupmembers(struct ifgroupreq *);
271 static void if_delgroups(struct ifnet *);
272 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
275 static void if_vmove(struct ifnet *, struct vnet *);
280 * XXX: declare here to avoid to include many inet6 related files..
281 * should be more generalized?
283 extern void nd6_setmtu(struct ifnet *);
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
290 VNET_DEFINE(int, if_index);
291 int ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
300 #define V_if_indexlim VNET(if_indexlim)
301 #define V_ifindex_table VNET(ifindex_table)
304 * The global network interface list (V_ifnet) and related state (such as
305 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306 * an rwlock. Either may be acquired shared to stablize the list, but both
307 * must be acquired writable to modify the list. This model allows us to
308 * both stablize the interface list during interrupt thread processing, but
309 * also to stablize it over long-running ioctls, without introducing priority
310 * inversions and deadlocks.
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
318 * The allocation of network interfaces is a rather non-atomic affair; we
319 * need to select an index before we are ready to expose the interface for
320 * use, so will use this pointer value to indicate reservation.
322 #define IFNET_HOLD (void *)(uintptr_t)(-1)
324 static if_com_alloc_t *if_com_alloc[256];
325 static if_com_free_t *if_com_free[256];
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
332 ifnet_byindex(u_short idx)
336 if (__predict_false(idx > V_if_index))
339 ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx);
340 return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp);
344 ifnet_byindex_ref(u_short idx)
350 ifp = ifnet_byindex(idx);
351 if (ifp == NULL || (ifp->if_flags & IFF_DYING))
358 * Allocate an ifindex array entry; return 0 on success or an error on
362 ifindex_alloc(void **old)
366 IFNET_WLOCK_ASSERT();
368 * Try to find an empty slot below V_if_index. If we fail, take the
371 for (idx = 1; idx <= V_if_index; idx++) {
372 if (V_ifindex_table[idx] == NULL)
376 /* Catch if_index overflow. */
377 if (idx >= V_if_indexlim) {
381 if (idx > V_if_index)
387 ifindex_free_locked(u_short idx)
390 IFNET_WLOCK_ASSERT();
392 V_ifindex_table[idx] = NULL;
393 while (V_if_index > 0 &&
394 V_ifindex_table[V_if_index] == NULL)
399 ifindex_free(u_short idx)
403 ifindex_free_locked(idx);
408 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
411 V_ifindex_table[idx] = ifp;
415 ifaddr_byindex(u_short idx)
418 struct ifaddr *ifa = NULL;
422 ifp = ifnet_byindex(idx);
423 if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
429 * Network interface utility routines.
431 * Routines with ifa_ifwith* names take sockaddr *'s as
436 vnet_if_init(const void *unused __unused)
440 CK_STAILQ_INIT(&V_ifnet);
441 CK_STAILQ_INIT(&V_ifg_head);
443 old = if_grow(); /* create initial table */
445 epoch_wait_preempt(net_epoch_preempt);
447 vnet_if_clone_init();
449 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
454 vnet_if_uninit(const void *unused __unused)
457 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
458 "not empty", __func__, __LINE__, &V_ifnet));
459 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
460 "not empty", __func__, __LINE__, &V_ifg_head));
462 free((caddr_t)V_ifindex_table, M_IFNET);
464 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
465 vnet_if_uninit, NULL);
468 vnet_if_return(const void *unused __unused)
470 struct ifnet *ifp, *nifp;
472 /* Return all inherited interfaces to their parent vnets. */
473 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
474 if (ifp->if_home_vnet != ifp->if_vnet)
475 if_vmove(ifp, ifp->if_home_vnet);
478 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
479 vnet_if_return, NULL);
492 IFNET_WLOCK_ASSERT();
493 oldlim = V_if_indexlim;
495 n = (oldlim << 1) * sizeof(*e);
496 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
498 if (V_if_indexlim != oldlim) {
502 if (V_ifindex_table != NULL) {
503 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
504 old = V_ifindex_table;
512 * Allocate a struct ifnet and an index for an interface. A layer 2
513 * common structure will also be allocated if an allocation routine is
514 * registered for the passed type.
517 if_alloc_domain(u_char type, int numa_domain)
523 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
524 if (numa_domain == IF_NODOM)
525 ifp = malloc(sizeof(struct ifnet), M_IFNET,
528 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
529 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
532 idx = ifindex_alloc(&old);
533 if (__predict_false(idx == USHRT_MAX)) {
535 epoch_wait_preempt(net_epoch_preempt);
539 ifnet_setbyindex(idx, IFNET_HOLD);
543 ifp->if_alloctype = type;
544 ifp->if_numa_domain = numa_domain;
546 ifp->if_vnet = curvnet;
548 if (if_com_alloc[type] != NULL) {
549 ifp->if_l2com = if_com_alloc[type](type, ifp);
550 if (ifp->if_l2com == NULL) {
557 IF_ADDR_LOCK_INIT(ifp);
558 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
559 ifp->if_afdata_initialized = 0;
560 IF_AFDATA_LOCK_INIT(ifp);
561 CK_STAILQ_INIT(&ifp->if_addrhead);
562 CK_STAILQ_INIT(&ifp->if_multiaddrs);
563 CK_STAILQ_INIT(&ifp->if_groups);
567 ifq_init(&ifp->if_snd, ifp);
569 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
570 for (int i = 0; i < IFCOUNTERS; i++)
571 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
572 ifp->if_get_counter = if_get_counter_default;
573 ifp->if_pcp = IFNET_PCP_NONE;
574 ifnet_setbyindex(ifp->if_index, ifp);
579 if_alloc_dev(u_char type, device_t dev)
583 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
584 return (if_alloc_domain(type, IF_NODOM));
585 return (if_alloc_domain(type, numa_domain));
589 if_alloc(u_char type)
592 return (if_alloc_domain(type, IF_NODOM));
595 * Do the actual work of freeing a struct ifnet, and layer 2 common
596 * structure. This call is made when the last reference to an
597 * interface is released.
600 if_free_internal(struct ifnet *ifp)
603 KASSERT((ifp->if_flags & IFF_DYING),
604 ("if_free_internal: interface not dying"));
606 if (if_com_free[ifp->if_alloctype] != NULL)
607 if_com_free[ifp->if_alloctype](ifp->if_l2com,
611 mac_ifnet_destroy(ifp);
613 IF_AFDATA_DESTROY(ifp);
614 IF_ADDR_LOCK_DESTROY(ifp);
615 ifq_delete(&ifp->if_snd);
617 for (int i = 0; i < IFCOUNTERS; i++)
618 counter_u64_free(ifp->if_counters[i]);
620 free(ifp->if_description, M_IFDESCR);
621 free(ifp->if_hw_addr, M_IFADDR);
622 if (ifp->if_numa_domain == IF_NODOM)
625 free_domain(ifp, M_IFNET);
629 if_destroy(epoch_context_t ctx)
633 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
634 if_free_internal(ifp);
638 * Deregister an interface and free the associated storage.
641 if_free(struct ifnet *ifp)
644 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
646 CURVNET_SET_QUIET(ifp->if_vnet);
648 KASSERT(ifp == ifnet_byindex(ifp->if_index),
649 ("%s: freeing unallocated ifnet", ifp->if_xname));
651 ifindex_free_locked(ifp->if_index);
654 if (refcount_release(&ifp->if_refcount))
655 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
660 * Interfaces to keep an ifnet type-stable despite the possibility of the
661 * driver calling if_free(). If there are additional references, we defer
662 * freeing the underlying data structure.
665 if_ref(struct ifnet *ifp)
668 /* We don't assert the ifnet list lock here, but arguably should. */
669 refcount_acquire(&ifp->if_refcount);
673 if_rele(struct ifnet *ifp)
676 if (!refcount_release(&ifp->if_refcount))
678 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
682 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
685 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
687 if (ifq->ifq_maxlen == 0)
688 ifq->ifq_maxlen = ifqmaxlen;
691 ifq->altq_disc = NULL;
692 ifq->altq_flags &= ALTQF_CANTCHANGE;
693 ifq->altq_tbr = NULL;
698 ifq_delete(struct ifaltq *ifq)
700 mtx_destroy(&ifq->ifq_mtx);
704 * Perform generic interface initialization tasks and attach the interface
705 * to the list of "active" interfaces. If vmove flag is set on entry
706 * to if_attach_internal(), perform only a limited subset of initialization
707 * tasks, given that we are moving from one vnet to another an ifnet which
708 * has already been fully initialized.
710 * Note that if_detach_internal() removes group membership unconditionally
711 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
712 * Thus, when if_vmove() is applied to a cloned interface, group membership
713 * is lost while a cloned one always joins a group whose name is
714 * ifc->ifc_name. To recover this after if_detach_internal() and
715 * if_attach_internal(), the cloner should be specified to
716 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
717 * attempts to join a group whose name is ifc->ifc_name.
720 * - The decision to return void and thus require this function to
721 * succeed is questionable.
722 * - We should probably do more sanity checking. For instance we don't
723 * do anything to insure if_xname is unique or non-empty.
726 if_attach(struct ifnet *ifp)
729 if_attach_internal(ifp, 0, NULL);
733 * Compute the least common TSO limit.
736 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
739 * 1) If there is no limit currently, take the limit from
740 * the network adapter.
742 * 2) If the network adapter has a limit below the current
745 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
746 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
747 pmax->tsomaxbytes = ifp->if_hw_tsomax;
749 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
750 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
751 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
753 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
754 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
755 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
760 * Update TSO limit of a network adapter.
762 * Returns zero if no change. Else non-zero.
765 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
768 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
769 ifp->if_hw_tsomax = pmax->tsomaxbytes;
772 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
773 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
776 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
777 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
784 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
786 unsigned socksize, ifasize;
787 int namelen, masklen;
788 struct sockaddr_dl *sdl;
791 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
792 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
796 ifp->if_vnet = curvnet;
797 if (ifp->if_home_vnet == NULL)
798 ifp->if_home_vnet = curvnet;
801 if_addgroup(ifp, IFG_ALL);
803 /* Restore group membership for cloned interfaces. */
804 if (vmove && ifc != NULL)
805 if_clone_addgroup(ifp, ifc);
807 getmicrotime(&ifp->if_lastchange);
808 ifp->if_epoch = time_uptime;
810 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
811 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
812 ("transmit and qflush must both either be set or both be NULL"));
813 if (ifp->if_transmit == NULL) {
814 ifp->if_transmit = if_transmit;
815 ifp->if_qflush = if_qflush;
817 if (ifp->if_input == NULL)
818 ifp->if_input = if_input_default;
820 if (ifp->if_requestencap == NULL)
821 ifp->if_requestencap = if_requestencap_default;
825 mac_ifnet_create(ifp);
829 * Create a Link Level name for this device.
831 namelen = strlen(ifp->if_xname);
833 * Always save enough space for any possiable name so we
834 * can do a rename in place later.
836 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
837 socksize = masklen + ifp->if_addrlen;
838 if (socksize < sizeof(*sdl))
839 socksize = sizeof(*sdl);
840 socksize = roundup2(socksize, sizeof(long));
841 ifasize = sizeof(*ifa) + 2 * socksize;
842 ifa = ifa_alloc(ifasize, M_WAITOK);
843 sdl = (struct sockaddr_dl *)(ifa + 1);
844 sdl->sdl_len = socksize;
845 sdl->sdl_family = AF_LINK;
846 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
847 sdl->sdl_nlen = namelen;
848 sdl->sdl_index = ifp->if_index;
849 sdl->sdl_type = ifp->if_type;
852 ifa->ifa_addr = (struct sockaddr *)sdl;
853 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
854 ifa->ifa_netmask = (struct sockaddr *)sdl;
855 sdl->sdl_len = masklen;
857 sdl->sdl_data[--namelen] = 0xff;
858 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
859 /* Reliably crash if used uninitialized. */
860 ifp->if_broadcastaddr = NULL;
862 if (ifp->if_type == IFT_ETHER) {
863 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
867 #if defined(INET) || defined(INET6)
868 /* Use defaults for TSO, if nothing is set */
869 if (ifp->if_hw_tsomax == 0 &&
870 ifp->if_hw_tsomaxsegcount == 0 &&
871 ifp->if_hw_tsomaxsegsize == 0) {
873 * The TSO defaults needs to be such that an
874 * NFS mbuf list of 35 mbufs totalling just
875 * below 64K works and that a chain of mbufs
876 * can be defragged into at most 32 segments:
878 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
879 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
880 ifp->if_hw_tsomaxsegcount = 35;
881 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
883 /* XXX some drivers set IFCAP_TSO after ethernet attach */
884 if (ifp->if_capabilities & IFCAP_TSO) {
885 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
887 ifp->if_hw_tsomaxsegcount,
888 ifp->if_hw_tsomaxsegsize);
896 * Update the interface index in the link layer address
899 for (ifa = ifp->if_addr; ifa != NULL;
900 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
901 if (ifa->ifa_addr->sa_family == AF_LINK) {
902 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
903 sdl->sdl_index = ifp->if_index;
910 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
912 curvnet->vnet_ifcnt++;
916 if (domain_init_status >= 2)
917 if_attachdomain1(ifp);
919 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
920 if (IS_DEFAULT_VNET(curvnet))
921 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
923 /* Announce the interface. */
924 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
928 if_epochalloc(void *dummy __unused)
931 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
932 net_epoch = epoch_alloc("Net", 0);
934 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
935 if_epochalloc, NULL);
938 if_attachdomain(void *dummy)
942 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
943 if_attachdomain1(ifp);
945 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
946 if_attachdomain, NULL);
949 if_attachdomain1(struct ifnet *ifp)
954 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
955 * cannot lock ifp->if_afdata initialization, entirely.
958 if (ifp->if_afdata_initialized >= domain_init_status) {
959 IF_AFDATA_UNLOCK(ifp);
960 log(LOG_WARNING, "%s called more than once on %s\n",
961 __func__, ifp->if_xname);
964 ifp->if_afdata_initialized = domain_init_status;
965 IF_AFDATA_UNLOCK(ifp);
967 /* address family dependent data region */
968 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
969 for (dp = domains; dp; dp = dp->dom_next) {
970 if (dp->dom_ifattach)
971 ifp->if_afdata[dp->dom_family] =
972 (*dp->dom_ifattach)(ifp);
977 * Remove any unicast or broadcast network addresses from an interface.
980 if_purgeaddrs(struct ifnet *ifp)
985 struct epoch_tracker et;
988 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
989 if (ifa->ifa_addr->sa_family != AF_LINK)
997 /* XXX: Ugly!! ad hoc just for INET */
998 if (ifa->ifa_addr->sa_family == AF_INET) {
999 struct ifaliasreq ifr;
1001 bzero(&ifr, sizeof(ifr));
1002 ifr.ifra_addr = *ifa->ifa_addr;
1003 if (ifa->ifa_dstaddr)
1004 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1005 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1011 if (ifa->ifa_addr->sa_family == AF_INET6) {
1013 /* ifp_addrhead is already updated */
1018 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1019 IF_ADDR_WUNLOCK(ifp);
1025 * Remove any multicast network addresses from an interface when an ifnet
1029 if_purgemaddrs(struct ifnet *ifp)
1031 struct ifmultiaddr *ifma;
1034 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1035 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1036 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1037 if_delmulti_locked(ifp, ifma, 1);
1039 IF_ADDR_WUNLOCK(ifp);
1043 * Detach an interface, removing it from the list of "active" interfaces.
1044 * If vmove flag is set on entry to if_detach_internal(), perform only a
1045 * limited subset of cleanup tasks, given that we are moving an ifnet from
1046 * one vnet to another, where it must be fully operational.
1048 * XXXRW: There are some significant questions about event ordering, and
1049 * how to prevent things from starting to use the interface during detach.
1052 if_detach(struct ifnet *ifp)
1055 CURVNET_SET_QUIET(ifp->if_vnet);
1056 if_detach_internal(ifp, 0, NULL);
1061 * The vmove flag, if set, indicates that we are called from a callpath
1062 * that is moving an interface to a different vnet instance.
1064 * The shutdown flag, if set, indicates that we are called in the
1065 * process of shutting down a vnet instance. Currently only the
1066 * vnet_if_return SYSUNINIT function sets it. Note: we can be called
1067 * on a vnet instance shutdown without this flag being set, e.g., when
1068 * the cloned interfaces are destoyed as first thing of teardown.
1071 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1081 shutdown = ifp->if_vnet->vnet_shutdown;
1084 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1086 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1088 ifp->if_flags |= IFF_DYING;
1095 * While we would want to panic here, we cannot
1096 * guarantee that the interface is indeed still on
1097 * the list given we don't hold locks all the way.
1102 panic("%s: ifp=%p not on the ifnet tailq %p",
1103 __func__, ifp, &V_ifnet);
1105 return; /* XXX this should panic as well? */
1110 * At this point we know the interface still was on the ifnet list
1111 * and we removed it so we are in a stable state.
1114 curvnet->vnet_ifcnt--;
1116 epoch_wait_preempt(net_epoch_preempt);
1119 * Ensure all pending EPOCH(9) callbacks have been executed. This
1120 * fixes issues about late destruction of multicast options
1121 * which lead to leave group calls, which in turn access the
1122 * belonging ifnet structure:
1124 epoch_drain_callbacks(net_epoch_preempt);
1127 * In any case (destroy or vmove) detach us from the groups
1128 * and remove/wait for pending events on the taskq.
1129 * XXX-BZ in theory an interface could still enqueue a taskq change?
1133 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1136 * Check if this is a cloned interface or not. Must do even if
1137 * shutting down as a if_vmove_reclaim() would move the ifp and
1138 * the if_clone_addgroup() will have a corrupted string overwise
1139 * from a gibberish pointer.
1141 if (vmove && ifcp != NULL)
1142 *ifcp = if_clone_findifc(ifp);
1148 * On VNET shutdown abort here as the stack teardown will do all
1149 * the work top-down for us.
1152 /* Give interface users the chance to clean up. */
1153 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1156 * In case of a vmove we are done here without error.
1157 * If we would signal an error it would lead to the same
1158 * abort as if we did not find the ifnet anymore.
1159 * if_detach() calls us in void context and does not care
1160 * about an early abort notification, so life is splendid :)
1162 goto finish_vnet_shutdown;
1167 * At this point we are not tearing down a VNET and are either
1168 * going to destroy or vmove the interface and have to cleanup
1173 * Remove routes and flush queues.
1176 if (ALTQ_IS_ENABLED(&ifp->if_snd))
1177 altq_disable(&ifp->if_snd);
1178 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1179 altq_detach(&ifp->if_snd);
1190 * Remove all IPv6 kernel structs related to ifp. This should be done
1191 * before removing routing entries below, since IPv6 interface direct
1192 * routes are expected to be removed by the IPv6-specific kernel API.
1193 * Otherwise, the kernel will detect some inconsistency and bark it.
1197 if_purgemaddrs(ifp);
1199 /* Announce that the interface is gone. */
1200 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1201 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1202 if (IS_DEFAULT_VNET(curvnet))
1203 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1207 * Prevent further calls into the device driver via ifnet.
1212 * Clean up all addresses.
1215 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1216 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1217 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1218 IF_ADDR_WUNLOCK(ifp);
1221 IF_ADDR_WUNLOCK(ifp);
1224 rt_flushifroutes(ifp);
1227 finish_vnet_shutdown:
1230 * We cannot hold the lock over dom_ifdetach calls as they might
1231 * sleep, for example trying to drain a callout, thus open up the
1232 * theoretical race with re-attaching.
1234 IF_AFDATA_LOCK(ifp);
1235 i = ifp->if_afdata_initialized;
1236 ifp->if_afdata_initialized = 0;
1237 IF_AFDATA_UNLOCK(ifp);
1238 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1239 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1240 (*dp->dom_ifdetach)(ifp,
1241 ifp->if_afdata[dp->dom_family]);
1242 ifp->if_afdata[dp->dom_family] = NULL;
1251 * if_vmove() performs a limited version of if_detach() in current
1252 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1253 * An attempt is made to shrink if_index in current vnet, find an
1254 * unused if_index in target vnet and calls if_grow() if necessary,
1255 * and finally find an unused if_xname for the target vnet.
1258 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1260 struct if_clone *ifc;
1261 u_int bif_dlt, bif_hdrlen;
1266 * if_detach_internal() will call the eventhandler to notify
1267 * interface departure. That will detach if_bpf. We need to
1268 * safe the dlt and hdrlen so we can re-attach it later.
1270 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1273 * Detach from current vnet, but preserve LLADDR info, do not
1274 * mark as dead etc. so that the ifnet can be reattached later.
1275 * If we cannot find it, we lost the race to someone else.
1277 rc = if_detach_internal(ifp, 1, &ifc);
1282 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1283 * the if_index for that vnet if possible.
1285 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1286 * or we'd lock on one vnet and unlock on another.
1289 ifindex_free_locked(ifp->if_index);
1293 * Perform interface-specific reassignment tasks, if provided by
1296 if (ifp->if_reassign != NULL)
1297 ifp->if_reassign(ifp, new_vnet, NULL);
1300 * Switch to the context of the target vnet.
1302 CURVNET_SET_QUIET(new_vnet);
1305 ifp->if_index = ifindex_alloc(&old);
1306 if (__predict_false(ifp->if_index == USHRT_MAX)) {
1308 epoch_wait_preempt(net_epoch_preempt);
1312 ifnet_setbyindex(ifp->if_index, ifp);
1315 if_attach_internal(ifp, 1, ifc);
1317 if (ifp->if_bpf == NULL)
1318 bpfattach(ifp, bif_dlt, bif_hdrlen);
1324 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1327 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1332 /* Try to find the prison within our visibility. */
1333 sx_slock(&allprison_lock);
1334 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1335 sx_sunlock(&allprison_lock);
1338 prison_hold_locked(pr);
1339 mtx_unlock(&pr->pr_mtx);
1341 /* Do not try to move the iface from and to the same prison. */
1342 if (pr->pr_vnet == ifp->if_vnet) {
1347 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1348 /* XXX Lock interfaces to avoid races. */
1349 CURVNET_SET_QUIET(pr->pr_vnet);
1350 difp = ifunit(ifname);
1357 /* Make sure the VNET is stable. */
1358 if (ifp->if_vnet->vnet_shutdown) {
1365 /* Move the interface into the child jail/vnet. */
1366 if_vmove(ifp, pr->pr_vnet);
1368 /* Report the new if_xname back to the userland. */
1369 sprintf(ifname, "%s", ifp->if_xname);
1376 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1379 struct vnet *vnet_dst;
1382 /* Try to find the prison within our visibility. */
1383 sx_slock(&allprison_lock);
1384 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1385 sx_sunlock(&allprison_lock);
1388 prison_hold_locked(pr);
1389 mtx_unlock(&pr->pr_mtx);
1391 /* Make sure the named iface exists in the source prison/vnet. */
1392 CURVNET_SET(pr->pr_vnet);
1393 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1400 /* Do not try to move the iface from and to the same prison. */
1401 vnet_dst = TD_TO_VNET(td);
1402 if (vnet_dst == ifp->if_vnet) {
1408 /* Make sure the VNET is stable. */
1409 if (ifp->if_vnet->vnet_shutdown) {
1415 /* Get interface back from child jail/vnet. */
1416 if_vmove(ifp, vnet_dst);
1419 /* Report the new if_xname back to the userland. */
1420 sprintf(ifname, "%s", ifp->if_xname);
1428 * Add a group to an interface
1431 if_addgroup(struct ifnet *ifp, const char *groupname)
1433 struct ifg_list *ifgl;
1434 struct ifg_group *ifg = NULL;
1435 struct ifg_member *ifgm;
1438 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1439 groupname[strlen(groupname) - 1] <= '9')
1443 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1444 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1449 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1450 M_NOWAIT)) == NULL) {
1455 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1456 M_TEMP, M_NOWAIT)) == NULL) {
1462 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1463 if (!strcmp(ifg->ifg_group, groupname))
1467 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1468 M_TEMP, M_NOWAIT)) == NULL) {
1474 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1475 ifg->ifg_refcnt = 0;
1476 CK_STAILQ_INIT(&ifg->ifg_members);
1477 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1482 ifgl->ifgl_group = ifg;
1483 ifgm->ifgm_ifp = ifp;
1486 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1487 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1488 IF_ADDR_WUNLOCK(ifp);
1493 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1494 EVENTHANDLER_INVOKE(group_change_event, groupname);
1500 * Remove a group from an interface
1503 if_delgroup(struct ifnet *ifp, const char *groupname)
1505 struct ifg_list *ifgl;
1506 struct ifg_member *ifgm;
1510 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1511 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1520 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1521 IF_ADDR_WUNLOCK(ifp);
1523 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1524 if (ifgm->ifgm_ifp == ifp)
1528 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1530 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1531 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1536 epoch_wait_preempt(net_epoch_preempt);
1538 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1539 free(ifgl->ifgl_group, M_TEMP);
1544 EVENTHANDLER_INVOKE(group_change_event, groupname);
1550 * Remove an interface from all groups
1553 if_delgroups(struct ifnet *ifp)
1555 struct ifg_list *ifgl;
1556 struct ifg_member *ifgm;
1557 char groupname[IFNAMSIZ];
1561 while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1562 ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1564 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1567 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1568 IF_ADDR_WUNLOCK(ifp);
1570 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1571 if (ifgm->ifgm_ifp == ifp)
1575 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1578 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1579 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1584 epoch_wait_preempt(net_epoch_preempt);
1587 EVENTHANDLER_INVOKE(group_detach_event,
1589 free(ifgl->ifgl_group, M_TEMP);
1591 EVENTHANDLER_INVOKE(group_change_event, groupname);
1599 ifgr_group_get(void *ifgrp)
1601 union ifgroupreq_union *ifgrup;
1604 #ifdef COMPAT_FREEBSD32
1605 if (SV_CURPROC_FLAG(SV_ILP32))
1606 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1608 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1611 static struct ifg_req *
1612 ifgr_groups_get(void *ifgrp)
1614 union ifgroupreq_union *ifgrup;
1617 #ifdef COMPAT_FREEBSD32
1618 if (SV_CURPROC_FLAG(SV_ILP32))
1619 return ((struct ifg_req *)(uintptr_t)
1620 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1622 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1626 * Stores all groups from an interface in memory pointed to by ifgr.
1629 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1632 struct ifg_list *ifgl;
1633 struct ifg_req ifgrq, *ifgp;
1637 if (ifgr->ifgr_len == 0) {
1638 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1639 ifgr->ifgr_len += sizeof(struct ifg_req);
1643 len = ifgr->ifgr_len;
1644 ifgp = ifgr_groups_get(ifgr);
1646 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1647 if (len < sizeof(ifgrq))
1649 bzero(&ifgrq, sizeof ifgrq);
1650 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1651 sizeof(ifgrq.ifgrq_group));
1652 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1654 len -= sizeof(ifgrq);
1662 * Stores all members of a group in memory pointed to by igfr
1665 if_getgroupmembers(struct ifgroupreq *ifgr)
1667 struct ifg_group *ifg;
1668 struct ifg_member *ifgm;
1669 struct ifg_req ifgrq, *ifgp;
1673 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1674 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1681 if (ifgr->ifgr_len == 0) {
1682 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1683 ifgr->ifgr_len += sizeof(ifgrq);
1688 len = ifgr->ifgr_len;
1689 ifgp = ifgr_groups_get(ifgr);
1690 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1691 if (len < sizeof(ifgrq)) {
1695 bzero(&ifgrq, sizeof ifgrq);
1696 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1697 sizeof(ifgrq.ifgrq_member));
1698 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1702 len -= sizeof(ifgrq);
1711 * Return counter values from counter(9)s stored in ifnet.
1714 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1717 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1719 return (counter_u64_fetch(ifp->if_counters[cnt]));
1723 * Increase an ifnet counter. Usually used for counters shared
1724 * between the stack and a driver, but function supports them all.
1727 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1730 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1732 counter_u64_add(ifp->if_counters[cnt], inc);
1736 * Copy data from ifnet to userland API structure if_data.
1739 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1742 ifd->ifi_type = ifp->if_type;
1743 ifd->ifi_physical = 0;
1744 ifd->ifi_addrlen = ifp->if_addrlen;
1745 ifd->ifi_hdrlen = ifp->if_hdrlen;
1746 ifd->ifi_link_state = ifp->if_link_state;
1748 ifd->ifi_datalen = sizeof(struct if_data);
1749 ifd->ifi_mtu = ifp->if_mtu;
1750 ifd->ifi_metric = ifp->if_metric;
1751 ifd->ifi_baudrate = ifp->if_baudrate;
1752 ifd->ifi_hwassist = ifp->if_hwassist;
1753 ifd->ifi_epoch = ifp->if_epoch;
1754 ifd->ifi_lastchange = ifp->if_lastchange;
1756 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1757 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1758 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1759 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1760 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1761 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1762 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1763 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1764 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1765 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1766 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1767 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1771 * Wrapper functions for struct ifnet address list locking macros. These are
1772 * used by kernel modules to avoid encoding programming interface or binary
1773 * interface assumptions that may be violated when kernel-internal locking
1774 * approaches change.
1777 if_addr_rlock(struct ifnet *ifp)
1780 epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1784 if_addr_runlock(struct ifnet *ifp)
1787 epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1791 if_maddr_rlock(if_t ifp)
1794 epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1798 if_maddr_runlock(if_t ifp)
1801 epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1805 * Initialization, destruction and refcounting functions for ifaddrs.
1808 ifa_alloc(size_t size, int flags)
1812 KASSERT(size >= sizeof(struct ifaddr),
1813 ("%s: invalid size %zu", __func__, size));
1815 ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1819 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1821 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1823 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1825 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1828 refcount_init(&ifa->ifa_refcnt, 1);
1833 /* free(NULL) is okay */
1834 counter_u64_free(ifa->ifa_opackets);
1835 counter_u64_free(ifa->ifa_ipackets);
1836 counter_u64_free(ifa->ifa_obytes);
1837 counter_u64_free(ifa->ifa_ibytes);
1838 free(ifa, M_IFADDR);
1844 ifa_ref(struct ifaddr *ifa)
1847 refcount_acquire(&ifa->ifa_refcnt);
1851 ifa_destroy(epoch_context_t ctx)
1855 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1856 counter_u64_free(ifa->ifa_opackets);
1857 counter_u64_free(ifa->ifa_ipackets);
1858 counter_u64_free(ifa->ifa_obytes);
1859 counter_u64_free(ifa->ifa_ibytes);
1860 free(ifa, M_IFADDR);
1864 ifa_free(struct ifaddr *ifa)
1867 if (refcount_release(&ifa->ifa_refcnt))
1868 epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1873 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1874 struct sockaddr *ia)
1876 struct epoch_tracker et;
1878 struct rt_addrinfo info;
1879 struct sockaddr_dl null_sdl;
1884 bzero(&info, sizeof(info));
1885 if (cmd != RTM_DELETE)
1886 info.rti_ifp = V_loif;
1887 if (cmd == RTM_ADD) {
1888 /* explicitly specify (loopback) ifa */
1889 if (info.rti_ifp != NULL) {
1890 NET_EPOCH_ENTER(et);
1891 info.rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1892 if (info.rti_ifa != NULL)
1893 ifa_ref(info.rti_ifa);
1897 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1898 info.rti_info[RTAX_DST] = ia;
1899 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1900 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1902 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1905 !(cmd == RTM_ADD && error == EEXIST) &&
1906 !(cmd == RTM_DELETE && error == ENOENT))
1907 if_printf(ifp, "%s failed: %d\n", otype, error);
1913 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1916 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1920 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1923 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1927 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1930 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1934 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1935 * structs used to represent other address families, it is necessary
1936 * to perform a different comparison.
1939 #define sa_dl_equal(a1, a2) \
1940 ((((const struct sockaddr_dl *)(a1))->sdl_len == \
1941 ((const struct sockaddr_dl *)(a2))->sdl_len) && \
1942 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \
1943 CLLADDR((const struct sockaddr_dl *)(a2)), \
1944 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1947 * Locate an interface based on a complete address.
1951 ifa_ifwithaddr(const struct sockaddr *addr)
1958 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1959 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1960 if (ifa->ifa_addr->sa_family != addr->sa_family)
1962 if (sa_equal(addr, ifa->ifa_addr)) {
1965 /* IP6 doesn't have broadcast */
1966 if ((ifp->if_flags & IFF_BROADCAST) &&
1967 ifa->ifa_broadaddr &&
1968 ifa->ifa_broadaddr->sa_len != 0 &&
1969 sa_equal(ifa->ifa_broadaddr, addr)) {
1980 ifa_ifwithaddr_check(const struct sockaddr *addr)
1982 struct epoch_tracker et;
1985 NET_EPOCH_ENTER(et);
1986 rc = (ifa_ifwithaddr(addr) != NULL);
1992 * Locate an interface based on the broadcast address.
1996 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
2001 MPASS(in_epoch(net_epoch_preempt));
2002 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2003 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2005 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2006 if (ifa->ifa_addr->sa_family != addr->sa_family)
2008 if ((ifp->if_flags & IFF_BROADCAST) &&
2009 ifa->ifa_broadaddr &&
2010 ifa->ifa_broadaddr->sa_len != 0 &&
2011 sa_equal(ifa->ifa_broadaddr, addr)) {
2022 * Locate the point to point interface with a given destination address.
2026 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2031 MPASS(in_epoch(net_epoch_preempt));
2032 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2033 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2035 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2037 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2038 if (ifa->ifa_addr->sa_family != addr->sa_family)
2040 if (ifa->ifa_dstaddr != NULL &&
2041 sa_equal(addr, ifa->ifa_dstaddr)) {
2052 * Find an interface on a specific network. If many, choice
2053 * is most specific found.
2056 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2060 struct ifaddr *ifa_maybe = NULL;
2061 u_int af = addr->sa_family;
2062 const char *addr_data = addr->sa_data, *cplim;
2064 MPASS(in_epoch(net_epoch_preempt));
2066 * AF_LINK addresses can be looked up directly by their index number,
2067 * so do that if we can.
2069 if (af == AF_LINK) {
2070 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2071 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2072 return (ifaddr_byindex(sdl->sdl_index));
2076 * Scan though each interface, looking for ones that have addresses
2077 * in this address family and the requested fib.
2079 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2080 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2082 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2083 const char *cp, *cp2, *cp3;
2085 if (ifa->ifa_addr->sa_family != af)
2087 if (af == AF_INET &&
2088 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2090 * This is a bit broken as it doesn't
2091 * take into account that the remote end may
2092 * be a single node in the network we are
2094 * The trouble is that we don't know the
2095 * netmask for the remote end.
2097 if (ifa->ifa_dstaddr != NULL &&
2098 sa_equal(addr, ifa->ifa_dstaddr)) {
2103 * Scan all the bits in the ifa's address.
2104 * If a bit dissagrees with what we are
2105 * looking for, mask it with the netmask
2106 * to see if it really matters.
2107 * (A byte at a time)
2109 if (ifa->ifa_netmask == 0)
2112 cp2 = ifa->ifa_addr->sa_data;
2113 cp3 = ifa->ifa_netmask->sa_data;
2114 cplim = ifa->ifa_netmask->sa_len
2115 + (char *)ifa->ifa_netmask;
2117 if ((*cp++ ^ *cp2++) & *cp3++)
2118 goto next; /* next address! */
2120 * If the netmask of what we just found
2121 * is more specific than what we had before
2122 * (if we had one), or if the virtual status
2123 * of new prefix is better than of the old one,
2124 * then remember the new one before continuing
2125 * to search for an even better one.
2127 if (ifa_maybe == NULL ||
2128 ifa_preferred(ifa_maybe, ifa) ||
2129 rn_refines((caddr_t)ifa->ifa_netmask,
2130 (caddr_t)ifa_maybe->ifa_netmask)) {
2143 * Find an interface address specific to an interface best matching
2147 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2150 const char *cp, *cp2, *cp3;
2152 struct ifaddr *ifa_maybe = NULL;
2153 u_int af = addr->sa_family;
2158 MPASS(in_epoch(net_epoch_preempt));
2159 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2160 if (ifa->ifa_addr->sa_family != af)
2162 if (ifa_maybe == NULL)
2164 if (ifa->ifa_netmask == 0) {
2165 if (sa_equal(addr, ifa->ifa_addr) ||
2166 (ifa->ifa_dstaddr &&
2167 sa_equal(addr, ifa->ifa_dstaddr)))
2171 if (ifp->if_flags & IFF_POINTOPOINT) {
2172 if (sa_equal(addr, ifa->ifa_dstaddr))
2176 cp2 = ifa->ifa_addr->sa_data;
2177 cp3 = ifa->ifa_netmask->sa_data;
2178 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2179 for (; cp3 < cplim; cp3++)
2180 if ((*cp++ ^ *cp2++) & *cp3)
2192 * See whether new ifa is better than current one:
2193 * 1) A non-virtual one is preferred over virtual.
2194 * 2) A virtual in master state preferred over any other state.
2196 * Used in several address selecting functions.
2199 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2202 return (cur->ifa_carp && (!next->ifa_carp ||
2203 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2206 struct sockaddr_dl *
2207 link_alloc_sdl(size_t size, int flags)
2210 return (malloc(size, M_TEMP, flags));
2214 link_free_sdl(struct sockaddr *sa)
2220 * Fills in given sdl with interface basic info.
2221 * Returns pointer to filled sdl.
2223 struct sockaddr_dl *
2224 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2226 struct sockaddr_dl *sdl;
2228 sdl = (struct sockaddr_dl *)paddr;
2229 memset(sdl, 0, sizeof(struct sockaddr_dl));
2230 sdl->sdl_len = sizeof(struct sockaddr_dl);
2231 sdl->sdl_family = AF_LINK;
2232 sdl->sdl_index = ifp->if_index;
2233 sdl->sdl_type = iftype;
2239 * Mark an interface down and notify protocols of
2243 if_unroute(struct ifnet *ifp, int flag, int fam)
2247 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2249 ifp->if_flags &= ~flag;
2250 getmicrotime(&ifp->if_lastchange);
2251 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2252 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2253 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2254 ifp->if_qflush(ifp);
2257 (*carp_linkstate_p)(ifp);
2262 * Mark an interface up and notify protocols of
2266 if_route(struct ifnet *ifp, int flag, int fam)
2270 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2272 ifp->if_flags |= flag;
2273 getmicrotime(&ifp->if_lastchange);
2274 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2275 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2276 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2278 (*carp_linkstate_p)(ifp);
2285 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2286 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2287 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2288 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2289 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2290 int (*vlan_pcp_p)(struct ifnet *, uint16_t *);
2291 int (*vlan_setcookie_p)(struct ifnet *, void *);
2292 void *(*vlan_cookie_p)(struct ifnet *);
2295 * Handle a change in the interface link state. To avoid LORs
2296 * between driver lock and upper layer locks, as well as possible
2297 * recursions, we post event to taskqueue, and all job
2298 * is done in static do_link_state_change().
2301 if_link_state_change(struct ifnet *ifp, int link_state)
2303 /* Return if state hasn't changed. */
2304 if (ifp->if_link_state == link_state)
2307 ifp->if_link_state = link_state;
2309 /* XXXGL: reference ifp? */
2310 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2314 do_link_state_change(void *arg, int pending)
2320 link_state = ifp->if_link_state;
2322 CURVNET_SET(ifp->if_vnet);
2324 if (ifp->if_vlantrunk != NULL)
2325 (*vlan_link_state_p)(ifp);
2327 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2328 ifp->if_l2com != NULL)
2329 (*ng_ether_link_state_p)(ifp, link_state);
2331 (*carp_linkstate_p)(ifp);
2333 ifp->if_bridge_linkstate(ifp);
2335 (*lagg_linkstate_p)(ifp, link_state);
2337 if (IS_DEFAULT_VNET(curvnet))
2338 devctl_notify("IFNET", ifp->if_xname,
2339 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2342 if_printf(ifp, "%d link states coalesced\n", pending);
2343 if (log_link_state_change)
2344 if_printf(ifp, "link state changed to %s\n",
2345 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2346 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2351 * Mark an interface down and notify protocols of
2355 if_down(struct ifnet *ifp)
2358 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2359 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2363 * Mark an interface up and notify protocols of
2367 if_up(struct ifnet *ifp)
2370 if_route(ifp, IFF_UP, AF_UNSPEC);
2371 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2375 * Flush an interface queue.
2378 if_qflush(struct ifnet *ifp)
2386 if (ALTQ_IS_ENABLED(ifq))
2390 while ((m = n) != NULL) {
2401 * Map interface name to interface structure pointer, with or without
2402 * returning a reference.
2405 ifunit_ref(const char *name)
2407 struct epoch_tracker et;
2410 NET_EPOCH_ENTER(et);
2411 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2412 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2413 !(ifp->if_flags & IFF_DYING))
2423 ifunit(const char *name)
2425 struct epoch_tracker et;
2428 NET_EPOCH_ENTER(et);
2429 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2430 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2438 ifr_buffer_get_buffer(void *data)
2440 union ifreq_union *ifrup;
2443 #ifdef COMPAT_FREEBSD32
2444 if (SV_CURPROC_FLAG(SV_ILP32))
2445 return ((void *)(uintptr_t)
2446 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2448 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2452 ifr_buffer_set_buffer_null(void *data)
2454 union ifreq_union *ifrup;
2457 #ifdef COMPAT_FREEBSD32
2458 if (SV_CURPROC_FLAG(SV_ILP32))
2459 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2462 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2466 ifr_buffer_get_length(void *data)
2468 union ifreq_union *ifrup;
2471 #ifdef COMPAT_FREEBSD32
2472 if (SV_CURPROC_FLAG(SV_ILP32))
2473 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2475 return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2479 ifr_buffer_set_length(void *data, size_t len)
2481 union ifreq_union *ifrup;
2484 #ifdef COMPAT_FREEBSD32
2485 if (SV_CURPROC_FLAG(SV_ILP32))
2486 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2489 ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2493 ifr_data_get_ptr(void *ifrp)
2495 union ifreq_union *ifrup;
2498 #ifdef COMPAT_FREEBSD32
2499 if (SV_CURPROC_FLAG(SV_ILP32))
2500 return ((void *)(uintptr_t)
2501 ifrup->ifr32.ifr_ifru.ifru_data);
2503 return (ifrup->ifr.ifr_ifru.ifru_data);
2507 * Hardware specific interface ioctls.
2510 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2513 int error = 0, do_ifup = 0;
2514 int new_flags, temp_flags;
2515 size_t namelen, onamelen;
2517 char *descrbuf, *odescrbuf;
2518 char new_name[IFNAMSIZ];
2520 struct sockaddr_dl *sdl;
2522 ifr = (struct ifreq *)data;
2525 ifr->ifr_index = ifp->if_index;
2529 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2530 ifr->ifr_flags = temp_flags & 0xffff;
2531 ifr->ifr_flagshigh = temp_flags >> 16;
2535 ifr->ifr_reqcap = ifp->if_capabilities;
2536 ifr->ifr_curcap = ifp->if_capenable;
2541 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2546 ifr->ifr_metric = ifp->if_metric;
2550 ifr->ifr_mtu = ifp->if_mtu;
2554 /* XXXGL: did this ever worked? */
2560 sx_slock(&ifdescr_sx);
2561 if (ifp->if_description == NULL)
2564 /* space for terminating nul */
2565 descrlen = strlen(ifp->if_description) + 1;
2566 if (ifr_buffer_get_length(ifr) < descrlen)
2567 ifr_buffer_set_buffer_null(ifr);
2569 error = copyout(ifp->if_description,
2570 ifr_buffer_get_buffer(ifr), descrlen);
2571 ifr_buffer_set_length(ifr, descrlen);
2573 sx_sunlock(&ifdescr_sx);
2577 error = priv_check(td, PRIV_NET_SETIFDESCR);
2582 * Copy only (length-1) bytes to make sure that
2583 * if_description is always nul terminated. The
2584 * length parameter is supposed to count the
2585 * terminating nul in.
2587 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2588 return (ENAMETOOLONG);
2589 else if (ifr_buffer_get_length(ifr) == 0)
2592 descrbuf = malloc(ifr_buffer_get_length(ifr),
2593 M_IFDESCR, M_WAITOK | M_ZERO);
2594 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2595 ifr_buffer_get_length(ifr) - 1);
2597 free(descrbuf, M_IFDESCR);
2602 sx_xlock(&ifdescr_sx);
2603 odescrbuf = ifp->if_description;
2604 ifp->if_description = descrbuf;
2605 sx_xunlock(&ifdescr_sx);
2607 getmicrotime(&ifp->if_lastchange);
2608 free(odescrbuf, M_IFDESCR);
2612 ifr->ifr_fib = ifp->if_fib;
2616 error = priv_check(td, PRIV_NET_SETIFFIB);
2619 if (ifr->ifr_fib >= rt_numfibs)
2622 ifp->if_fib = ifr->ifr_fib;
2626 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2630 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2631 * check, so we don't need special handling here yet.
2633 new_flags = (ifr->ifr_flags & 0xffff) |
2634 (ifr->ifr_flagshigh << 16);
2635 if (ifp->if_flags & IFF_UP &&
2636 (new_flags & IFF_UP) == 0) {
2638 } else if (new_flags & IFF_UP &&
2639 (ifp->if_flags & IFF_UP) == 0) {
2642 /* See if permanently promiscuous mode bit is about to flip */
2643 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2644 if (new_flags & IFF_PPROMISC)
2645 ifp->if_flags |= IFF_PROMISC;
2646 else if (ifp->if_pcount == 0)
2647 ifp->if_flags &= ~IFF_PROMISC;
2648 if (log_promisc_mode_change)
2649 if_printf(ifp, "permanently promiscuous mode %s\n",
2650 ((new_flags & IFF_PPROMISC) ?
2651 "enabled" : "disabled"));
2653 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2654 (new_flags &~ IFF_CANTCHANGE);
2655 if (ifp->if_ioctl) {
2656 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2660 getmicrotime(&ifp->if_lastchange);
2664 error = priv_check(td, PRIV_NET_SETIFCAP);
2667 if (ifp->if_ioctl == NULL)
2668 return (EOPNOTSUPP);
2669 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2671 error = (*ifp->if_ioctl)(ifp, cmd, data);
2673 getmicrotime(&ifp->if_lastchange);
2678 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2683 error = priv_check(td, PRIV_NET_SETIFNAME);
2686 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2690 if (new_name[0] == '\0')
2692 if (new_name[IFNAMSIZ-1] != '\0') {
2693 new_name[IFNAMSIZ-1] = '\0';
2694 if (strlen(new_name) == IFNAMSIZ-1)
2697 if (strcmp(new_name, ifp->if_xname) == 0)
2699 if (ifunit(new_name) != NULL)
2703 * XXX: Locking. Nothing else seems to lock if_flags,
2704 * and there are numerous other races with the
2705 * ifunit() checks not being atomic with namespace
2706 * changes (renames, vmoves, if_attach, etc).
2708 ifp->if_flags |= IFF_RENAMING;
2710 /* Announce the departure of the interface. */
2711 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2712 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2714 if_printf(ifp, "changing name to '%s'\n", new_name);
2717 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2719 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2720 namelen = strlen(new_name);
2721 onamelen = sdl->sdl_nlen;
2723 * Move the address if needed. This is safe because we
2724 * allocate space for a name of length IFNAMSIZ when we
2725 * create this in if_attach().
2727 if (namelen != onamelen) {
2728 bcopy(sdl->sdl_data + onamelen,
2729 sdl->sdl_data + namelen, sdl->sdl_alen);
2731 bcopy(new_name, sdl->sdl_data, namelen);
2732 sdl->sdl_nlen = namelen;
2733 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2734 bzero(sdl->sdl_data, onamelen);
2735 while (namelen != 0)
2736 sdl->sdl_data[--namelen] = 0xff;
2737 IF_ADDR_WUNLOCK(ifp);
2739 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2740 /* Announce the return of the interface. */
2741 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2743 ifp->if_flags &= ~IFF_RENAMING;
2748 error = priv_check(td, PRIV_NET_SETIFVNET);
2751 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2756 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2759 ifp->if_metric = ifr->ifr_metric;
2760 getmicrotime(&ifp->if_lastchange);
2764 error = priv_check(td, PRIV_NET_SETIFPHYS);
2767 if (ifp->if_ioctl == NULL)
2768 return (EOPNOTSUPP);
2769 error = (*ifp->if_ioctl)(ifp, cmd, data);
2771 getmicrotime(&ifp->if_lastchange);
2776 u_long oldmtu = ifp->if_mtu;
2778 error = priv_check(td, PRIV_NET_SETIFMTU);
2781 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2783 if (ifp->if_ioctl == NULL)
2784 return (EOPNOTSUPP);
2785 error = (*ifp->if_ioctl)(ifp, cmd, data);
2787 getmicrotime(&ifp->if_lastchange);
2790 NETDUMP_REINIT(ifp);
2794 * If the link MTU changed, do network layer specific procedure.
2796 if (ifp->if_mtu != oldmtu) {
2807 if (cmd == SIOCADDMULTI)
2808 error = priv_check(td, PRIV_NET_ADDMULTI);
2810 error = priv_check(td, PRIV_NET_DELMULTI);
2814 /* Don't allow group membership on non-multicast interfaces. */
2815 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2816 return (EOPNOTSUPP);
2818 /* Don't let users screw up protocols' entries. */
2819 if (ifr->ifr_addr.sa_family != AF_LINK)
2822 if (cmd == SIOCADDMULTI) {
2823 struct epoch_tracker et;
2824 struct ifmultiaddr *ifma;
2827 * Userland is only permitted to join groups once
2828 * via the if_addmulti() KPI, because it cannot hold
2829 * struct ifmultiaddr * between calls. It may also
2830 * lose a race while we check if the membership
2833 NET_EPOCH_ENTER(et);
2834 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2839 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2841 error = if_delmulti(ifp, &ifr->ifr_addr);
2844 getmicrotime(&ifp->if_lastchange);
2847 case SIOCSIFPHYADDR:
2848 case SIOCDIFPHYADDR:
2850 case SIOCSIFPHYADDR_IN6:
2853 case SIOCSIFGENERIC:
2854 error = priv_check(td, PRIV_NET_HWIOCTL);
2857 if (ifp->if_ioctl == NULL)
2858 return (EOPNOTSUPP);
2859 error = (*ifp->if_ioctl)(ifp, cmd, data);
2861 getmicrotime(&ifp->if_lastchange);
2865 case SIOCGIFPSRCADDR:
2866 case SIOCGIFPDSTADDR:
2869 case SIOCGIFGENERIC:
2871 case SIOCGIFRSSHASH:
2872 case SIOCGIFDOWNREASON:
2873 if (ifp->if_ioctl == NULL)
2874 return (EOPNOTSUPP);
2875 error = (*ifp->if_ioctl)(ifp, cmd, data);
2879 error = priv_check(td, PRIV_NET_SETLLADDR);
2882 error = if_setlladdr(ifp,
2883 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2887 error = if_gethwaddr(ifp, ifr);
2890 case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2891 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2894 if ((error = if_addgroup(ifp,
2895 ifgr_group_get((struct ifgroupreq *)data))))
2899 case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2901 struct epoch_tracker et;
2903 NET_EPOCH_ENTER(et);
2904 error = if_getgroup((struct ifgroupreq *)data, ifp);
2909 case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2910 error = priv_check(td, PRIV_NET_DELIFGROUP);
2913 if ((error = if_delgroup(ifp,
2914 ifgr_group_get((struct ifgroupreq *)data))))
2925 #ifdef COMPAT_FREEBSD32
2933 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2936 #ifdef COMPAT_FREEBSD32
2938 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2940 struct ifmediareq32 *ifmr32;
2942 ifmr32 = (struct ifmediareq32 *)data;
2943 memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2944 sizeof(ifmr->ifm_name));
2945 ifmr->ifm_current = ifmr32->ifm_current;
2946 ifmr->ifm_mask = ifmr32->ifm_mask;
2947 ifmr->ifm_status = ifmr32->ifm_status;
2948 ifmr->ifm_active = ifmr32->ifm_active;
2949 ifmr->ifm_count = ifmr32->ifm_count;
2950 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2954 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2956 struct ifmediareq32 *ifmr32;
2958 ifmr32 = (struct ifmediareq32 *)data;
2959 ifmr32->ifm_current = ifmr->ifm_current;
2960 ifmr32->ifm_mask = ifmr->ifm_mask;
2961 ifmr32->ifm_status = ifmr->ifm_status;
2962 ifmr32->ifm_active = ifmr->ifm_active;
2963 ifmr32->ifm_count = ifmr->ifm_count;
2971 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2973 #ifdef COMPAT_FREEBSD32
2974 caddr_t saved_data = NULL;
2975 struct ifmediareq ifmr;
2976 struct ifmediareq *ifmrp = NULL;
2983 CURVNET_SET(so->so_vnet);
2985 /* Make sure the VNET is stable. */
2986 if (so->so_vnet->vnet_shutdown) {
2994 error = ifconf(cmd, data);
2997 #ifdef COMPAT_FREEBSD32
3000 struct ifconf32 *ifc32;
3003 ifc32 = (struct ifconf32 *)data;
3004 ifc.ifc_len = ifc32->ifc_len;
3005 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3007 error = ifconf(SIOCGIFCONF, (void *)&ifc);
3009 ifc32->ifc_len = ifc.ifc_len;
3015 #ifdef COMPAT_FREEBSD32
3017 case SIOCGIFMEDIA32:
3018 case SIOCGIFXMEDIA32:
3020 ifmr_init(ifmrp, data);
3021 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3023 data = (caddr_t)ifmrp;
3027 ifr = (struct ifreq *)data;
3031 error = priv_check(td, PRIV_NET_SETIFVNET);
3033 error = if_vmove_reclaim(td, ifr->ifr_name,
3039 error = priv_check(td, PRIV_NET_IFCREATE);
3041 error = if_clone_create(ifr->ifr_name,
3042 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3043 ifr_data_get_ptr(ifr) : NULL);
3046 error = priv_check(td, PRIV_NET_IFDESTROY);
3048 error = if_clone_destroy(ifr->ifr_name);
3051 case SIOCIFGCLONERS:
3052 error = if_clone_list((struct if_clonereq *)data);
3055 case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3056 error = if_getgroupmembers((struct ifgroupreq *)data);
3059 #if defined(INET) || defined(INET6)
3062 if (carp_ioctl_p == NULL)
3063 error = EPROTONOSUPPORT;
3065 error = (*carp_ioctl_p)(ifr, cmd, td);
3070 ifp = ifunit_ref(ifr->ifr_name);
3076 error = ifhwioctl(cmd, ifp, data, td);
3077 if (error != ENOIOCTL)
3080 oif_flags = ifp->if_flags;
3081 if (so->so_proto == NULL) {
3087 * Pass the request on to the socket control method, and if the
3088 * latter returns EOPNOTSUPP, directly to the interface.
3090 * Make an exception for the legacy SIOCSIF* requests. Drivers
3091 * trust SIOCSIFADDR et al to come from an already privileged
3092 * layer, and do not perform any credentials checks or input
3095 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3097 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3098 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3099 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3100 error = (*ifp->if_ioctl)(ifp, cmd, data);
3102 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3104 if (ifp->if_flags & IFF_UP)
3112 #ifdef COMPAT_FREEBSD32
3113 if (ifmrp != NULL) {
3114 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3115 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3118 ifmr_update(ifmrp, data);
3126 * The code common to handling reference counted flags,
3127 * e.g., in ifpromisc() and if_allmulti().
3128 * The "pflag" argument can specify a permanent mode flag to check,
3129 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3131 * Only to be used on stack-owned flags, not driver-owned flags.
3134 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3138 int oldflags, oldcount;
3140 /* Sanity checks to catch programming errors */
3141 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3142 ("%s: setting driver-owned flag %d", __func__, flag));
3145 KASSERT(*refcount >= 0,
3146 ("%s: increment negative refcount %d for flag %d",
3147 __func__, *refcount, flag));
3149 KASSERT(*refcount > 0,
3150 ("%s: decrement non-positive refcount %d for flag %d",
3151 __func__, *refcount, flag));
3153 /* In case this mode is permanent, just touch refcount */
3154 if (ifp->if_flags & pflag) {
3155 *refcount += onswitch ? 1 : -1;
3159 /* Save ifnet parameters for if_ioctl() may fail */
3160 oldcount = *refcount;
3161 oldflags = ifp->if_flags;
3164 * See if we aren't the only and touching refcount is enough.
3165 * Actually toggle interface flag if we are the first or last.
3170 ifp->if_flags |= flag;
3174 ifp->if_flags &= ~flag;
3177 /* Call down the driver since we've changed interface flags */
3178 if (ifp->if_ioctl == NULL) {
3182 ifr.ifr_flags = ifp->if_flags & 0xffff;
3183 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3184 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3187 /* Notify userland that interface flags have changed */
3192 /* Recover after driver error */
3193 *refcount = oldcount;
3194 ifp->if_flags = oldflags;
3199 * Set/clear promiscuous mode on interface ifp based on the truth value
3200 * of pswitch. The calls are reference counted so that only the first
3201 * "on" request actually has an effect, as does the final "off" request.
3202 * Results are undefined if the "off" and "on" requests are not matched.
3205 ifpromisc(struct ifnet *ifp, int pswitch)
3208 int oldflags = ifp->if_flags;
3210 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3211 &ifp->if_pcount, pswitch);
3212 /* If promiscuous mode status has changed, log a message */
3213 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3214 log_promisc_mode_change)
3215 if_printf(ifp, "promiscuous mode %s\n",
3216 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3221 * Return interface configuration
3222 * of system. List may be used
3223 * in later ioctl's (above) to get
3224 * other information.
3228 ifconf(u_long cmd, caddr_t data)
3230 struct ifconf *ifc = (struct ifconf *)data;
3235 int error, full = 0, valid_len, max_len;
3237 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3238 max_len = MAXPHYS - 1;
3240 /* Prevent hostile input from being able to crash the system */
3241 if (ifc->ifc_len <= 0)
3245 if (ifc->ifc_len <= max_len) {
3246 max_len = ifc->ifc_len;
3249 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3254 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3255 struct epoch_tracker et;
3259 * Zero the ifr to make sure we don't disclose the contents
3262 memset(&ifr, 0, sizeof(ifr));
3264 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3265 >= sizeof(ifr.ifr_name)) {
3268 return (ENAMETOOLONG);
3272 NET_EPOCH_ENTER(et);
3273 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3274 struct sockaddr *sa = ifa->ifa_addr;
3276 if (prison_if(curthread->td_ucred, sa) != 0)
3279 if (sa->sa_len <= sizeof(*sa)) {
3280 if (sa->sa_len < sizeof(*sa)) {
3281 memset(&ifr.ifr_ifru.ifru_addr, 0,
3282 sizeof(ifr.ifr_ifru.ifru_addr));
3283 memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3286 ifr.ifr_ifru.ifru_addr = *sa;
3287 sbuf_bcat(sb, &ifr, sizeof(ifr));
3288 max_len += sizeof(ifr);
3291 offsetof(struct ifreq, ifr_addr));
3292 max_len += offsetof(struct ifreq, ifr_addr);
3293 sbuf_bcat(sb, sa, sa->sa_len);
3294 max_len += sa->sa_len;
3297 if (sbuf_error(sb) == 0)
3298 valid_len = sbuf_len(sb);
3302 sbuf_bcat(sb, &ifr, sizeof(ifr));
3303 max_len += sizeof(ifr);
3305 if (sbuf_error(sb) == 0)
3306 valid_len = sbuf_len(sb);
3312 * If we didn't allocate enough space (uncommon), try again. If
3313 * we have already allocated as much space as we are allowed,
3314 * return what we've got.
3316 if (valid_len != max_len && !full) {
3321 ifc->ifc_len = valid_len;
3323 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3329 * Just like ifpromisc(), but for all-multicast-reception mode.
3332 if_allmulti(struct ifnet *ifp, int onswitch)
3335 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3338 struct ifmultiaddr *
3339 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3341 struct ifmultiaddr *ifma;
3343 IF_ADDR_LOCK_ASSERT(ifp);
3345 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3346 if (sa->sa_family == AF_LINK) {
3347 if (sa_dl_equal(ifma->ifma_addr, sa))
3350 if (sa_equal(ifma->ifma_addr, sa))
3359 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3360 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3361 * the ifnet multicast address list here, so the caller must do that and
3362 * other setup work (such as notifying the device driver). The reference
3363 * count is initialized to 1.
3365 static struct ifmultiaddr *
3366 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3369 struct ifmultiaddr *ifma;
3370 struct sockaddr *dupsa;
3372 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3377 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3378 if (dupsa == NULL) {
3379 free(ifma, M_IFMADDR);
3382 bcopy(sa, dupsa, sa->sa_len);
3383 ifma->ifma_addr = dupsa;
3385 ifma->ifma_ifp = ifp;
3386 ifma->ifma_refcount = 1;
3387 ifma->ifma_protospec = NULL;
3390 ifma->ifma_lladdr = NULL;
3394 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3395 if (dupsa == NULL) {
3396 free(ifma->ifma_addr, M_IFMADDR);
3397 free(ifma, M_IFMADDR);
3400 bcopy(llsa, dupsa, llsa->sa_len);
3401 ifma->ifma_lladdr = dupsa;
3407 * if_freemulti: free ifmultiaddr structure and possibly attached related
3408 * addresses. The caller is responsible for implementing reference
3409 * counting, notifying the driver, handling routing messages, and releasing
3410 * any dependent link layer state.
3412 #ifdef MCAST_VERBOSE
3413 extern void kdb_backtrace(void);
3416 if_freemulti_internal(struct ifmultiaddr *ifma)
3419 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3420 ifma->ifma_refcount));
3422 if (ifma->ifma_lladdr != NULL)
3423 free(ifma->ifma_lladdr, M_IFMADDR);
3424 #ifdef MCAST_VERBOSE
3426 printf("%s freeing ifma: %p\n", __func__, ifma);
3428 free(ifma->ifma_addr, M_IFMADDR);
3429 free(ifma, M_IFMADDR);
3433 if_destroymulti(epoch_context_t ctx)
3435 struct ifmultiaddr *ifma;
3437 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3438 if_freemulti_internal(ifma);
3442 if_freemulti(struct ifmultiaddr *ifma)
3444 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3445 ifma->ifma_refcount));
3447 epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3452 * Register an additional multicast address with a network interface.
3454 * - If the address is already present, bump the reference count on the
3455 * address and return.
3456 * - If the address is not link-layer, look up a link layer address.
3457 * - Allocate address structures for one or both addresses, and attach to the
3458 * multicast address list on the interface. If automatically adding a link
3459 * layer address, the protocol address will own a reference to the link
3460 * layer address, to be freed when it is freed.
3461 * - Notify the network device driver of an addition to the multicast address
3464 * 'sa' points to caller-owned memory with the desired multicast address.
3466 * 'retifma' will be used to return a pointer to the resulting multicast
3467 * address reference, if desired.
3470 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3471 struct ifmultiaddr **retifma)
3473 struct ifmultiaddr *ifma, *ll_ifma;
3474 struct sockaddr *llsa;
3475 struct sockaddr_dl sdl;
3479 IN_MULTI_LIST_UNLOCK_ASSERT();
3482 IN6_MULTI_LIST_UNLOCK_ASSERT();
3485 * If the address is already present, return a new reference to it;
3486 * otherwise, allocate storage and set up a new address.
3489 ifma = if_findmulti(ifp, sa);
3491 ifma->ifma_refcount++;
3492 if (retifma != NULL)
3494 IF_ADDR_WUNLOCK(ifp);
3499 * The address isn't already present; resolve the protocol address
3500 * into a link layer address, and then look that up, bump its
3501 * refcount or allocate an ifma for that also.
3502 * Most link layer resolving functions returns address data which
3503 * fits inside default sockaddr_dl structure. However callback
3504 * can allocate another sockaddr structure, in that case we need to
3509 if (ifp->if_resolvemulti != NULL) {
3510 /* Provide called function with buffer size information */
3511 sdl.sdl_len = sizeof(sdl);
3512 llsa = (struct sockaddr *)&sdl;
3513 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3519 * Allocate the new address. Don't hook it up yet, as we may also
3520 * need to allocate a link layer multicast address.
3522 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3529 * If a link layer address is found, we'll need to see if it's
3530 * already present in the address list, or allocate is as well.
3531 * When this block finishes, the link layer address will be on the
3535 ll_ifma = if_findmulti(ifp, llsa);
3536 if (ll_ifma == NULL) {
3537 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3538 if (ll_ifma == NULL) {
3539 --ifma->ifma_refcount;
3544 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3545 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3548 ll_ifma->ifma_refcount++;
3549 ifma->ifma_llifma = ll_ifma;
3553 * We now have a new multicast address, ifma, and possibly a new or
3554 * referenced link layer address. Add the primary address to the
3555 * ifnet address list.
3557 ifma->ifma_flags |= IFMA_F_ENQUEUED;
3558 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3560 if (retifma != NULL)
3564 * Must generate the message while holding the lock so that 'ifma'
3565 * pointer is still valid.
3567 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3568 IF_ADDR_WUNLOCK(ifp);
3571 * We are certain we have added something, so call down to the
3572 * interface to let them know about it.
3574 if (ifp->if_ioctl != NULL) {
3575 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3578 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3579 link_free_sdl(llsa);
3584 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3585 link_free_sdl(llsa);
3588 IF_ADDR_WUNLOCK(ifp);
3593 * Delete a multicast group membership by network-layer group address.
3595 * Returns ENOENT if the entry could not be found. If ifp no longer
3596 * exists, results are undefined. This entry point should only be used
3597 * from subsystems which do appropriate locking to hold ifp for the
3598 * duration of the call.
3599 * Network-layer protocol domains must use if_delmulti_ifma().
3602 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3604 struct ifmultiaddr *ifma;
3607 KASSERT(ifp, ("%s: NULL ifp", __func__));
3611 ifma = if_findmulti(ifp, sa);
3613 lastref = if_delmulti_locked(ifp, ifma, 0);
3614 IF_ADDR_WUNLOCK(ifp);
3619 if (lastref && ifp->if_ioctl != NULL) {
3620 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3627 * Delete all multicast group membership for an interface.
3628 * Should be used to quickly flush all multicast filters.
3631 if_delallmulti(struct ifnet *ifp)
3633 struct ifmultiaddr *ifma;
3634 struct ifmultiaddr *next;
3637 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3638 if_delmulti_locked(ifp, ifma, 0);
3639 IF_ADDR_WUNLOCK(ifp);
3643 if_delmulti_ifma(struct ifmultiaddr *ifma)
3645 if_delmulti_ifma_flags(ifma, 0);
3649 * Delete a multicast group membership by group membership pointer.
3650 * Network-layer protocol domains must use this routine.
3652 * It is safe to call this routine if the ifp disappeared.
3655 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3659 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3661 IN_MULTI_LIST_UNLOCK_ASSERT();
3663 ifp = ifma->ifma_ifp;
3666 printf("%s: ifma_ifp seems to be detached\n", __func__);
3668 struct epoch_tracker et;
3671 NET_EPOCH_ENTER(et);
3672 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3681 * If and only if the ifnet instance exists: Acquire the address lock.
3686 lastref = if_delmulti_locked(ifp, ifma, flags);
3690 * If and only if the ifnet instance exists:
3691 * Release the address lock.
3692 * If the group was left: update the hardware hash filter.
3694 IF_ADDR_WUNLOCK(ifp);
3695 if (lastref && ifp->if_ioctl != NULL) {
3696 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3702 * Perform deletion of network-layer and/or link-layer multicast address.
3704 * Return 0 if the reference count was decremented.
3705 * Return 1 if the final reference was released, indicating that the
3706 * hardware hash filter should be reprogrammed.
3709 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3711 struct ifmultiaddr *ll_ifma;
3713 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3714 KASSERT(ifma->ifma_ifp == ifp,
3715 ("%s: inconsistent ifp %p", __func__, ifp));
3716 IF_ADDR_WLOCK_ASSERT(ifp);
3719 ifp = ifma->ifma_ifp;
3720 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3723 * If the ifnet is detaching, null out references to ifnet,
3724 * so that upper protocol layers will notice, and not attempt
3725 * to obtain locks for an ifnet which no longer exists. The
3726 * routing socket announcement must happen before the ifnet
3727 * instance is detached from the system.
3731 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3734 * ifp may already be nulled out if we are being reentered
3735 * to delete the ll_ifma.
3738 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3739 ifma->ifma_ifp = NULL;
3743 if (--ifma->ifma_refcount > 0)
3746 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3747 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3748 ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3751 * If this ifma is a network-layer ifma, a link-layer ifma may
3752 * have been associated with it. Release it first if so.
3754 ll_ifma = ifma->ifma_llifma;
3755 if (ll_ifma != NULL) {
3756 KASSERT(ifma->ifma_lladdr != NULL,
3757 ("%s: llifma w/o lladdr", __func__));
3759 ll_ifma->ifma_ifp = NULL; /* XXX */
3760 if (--ll_ifma->ifma_refcount == 0) {
3762 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3763 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3765 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3768 if_freemulti(ll_ifma);
3773 struct ifmultiaddr *ifmatmp;
3775 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3776 MPASS(ifma != ifmatmp);
3781 * The last reference to this instance of struct ifmultiaddr
3782 * was released; the hardware should be notified of this change.
3788 * Set the link layer address on an interface.
3790 * At this time we only support certain types of interfaces,
3791 * and we don't allow the length of the address to change.
3793 * Set noinline to be dtrace-friendly
3796 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3798 struct sockaddr_dl *sdl;
3806 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3810 if (len != sdl->sdl_alen) /* don't allow length to change */
3813 switch (ifp->if_type) {
3818 case IFT_IEEE8023ADLAG:
3819 bcopy(lladdr, LLADDR(sdl), len);
3826 * If the interface is already up, we need
3827 * to re-init it in order to reprogram its
3830 if ((ifp->if_flags & IFF_UP) != 0) {
3831 if (ifp->if_ioctl) {
3832 ifp->if_flags &= ~IFF_UP;
3833 ifr.ifr_flags = ifp->if_flags & 0xffff;
3834 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3835 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3836 ifp->if_flags |= IFF_UP;
3837 ifr.ifr_flags = ifp->if_flags & 0xffff;
3838 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3839 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3842 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3848 * Compat function for handling basic encapsulation requests.
3849 * Not converted stacks (FDDI, IB, ..) supports traditional
3850 * output model: ARP (and other similar L2 protocols) are handled
3851 * inside output routine, arpresolve/nd6_resolve() returns MAC
3852 * address instead of full prepend.
3854 * This function creates calculated header==MAC for IPv4/IPv6 and
3855 * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3859 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3862 if (req->rtype != IFENCAP_LL)
3863 return (EOPNOTSUPP);
3865 if (req->bufsize < req->lladdr_len)
3868 switch (req->family) {
3873 return (EAFNOSUPPORT);
3876 /* Copy lladdr to storage as is */
3877 memmove(req->buf, req->lladdr, req->lladdr_len);
3878 req->bufsize = req->lladdr_len;
3879 req->lladdr_off = 0;
3885 * Tunnel interfaces can nest, also they may cause infinite recursion
3886 * calls when misconfigured. We'll prevent this by detecting loops.
3887 * High nesting level may cause stack exhaustion. We'll prevent this
3888 * by introducing upper limit.
3890 * Return 0, if tunnel nesting count is equal or less than limit.
3893 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3901 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3902 if (*(struct ifnet **)(mtag + 1) == ifp) {
3903 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3908 if (count > limit) {
3910 "%s: if_output recursively called too many times(%d)\n",
3911 if_name(ifp), count);
3914 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3917 *(struct ifnet **)(mtag + 1) = ifp;
3918 m_tag_prepend(m, mtag);
3923 * Get the link layer address that was read from the hardware at attach.
3925 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3926 * their component interfaces as IFT_IEEE8023ADLAG.
3929 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3932 if (ifp->if_hw_addr == NULL)
3935 switch (ifp->if_type) {
3937 case IFT_IEEE8023ADLAG:
3938 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3946 * The name argument must be a pointer to storage which will last as
3947 * long as the interface does. For physical devices, the result of
3948 * device_get_name(dev) is a good choice and for pseudo-devices a
3949 * static string works well.
3952 if_initname(struct ifnet *ifp, const char *name, int unit)
3954 ifp->if_dname = name;
3955 ifp->if_dunit = unit;
3956 if (unit != IF_DUNIT_NONE)
3957 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3959 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3963 if_printf(struct ifnet *ifp, const char *fmt, ...)
3968 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3970 vlog(LOG_INFO, if_fmt, ap);
3976 if_start(struct ifnet *ifp)
3979 (*(ifp)->if_start)(ifp);
3983 * Backwards compatibility interface for drivers
3984 * that have not implemented it
3987 if_transmit(struct ifnet *ifp, struct mbuf *m)
3991 IFQ_HANDOFF(ifp, m, error);
3996 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4003 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4008 if (_IF_QFULL(ifq)) {
4010 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4015 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4016 if (m->m_flags & (M_BCAST|M_MCAST))
4017 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4018 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4020 _IF_ENQUEUE(ifq, m);
4022 if (ifp != NULL && !active)
4023 (*(ifp)->if_start)(ifp);
4028 if_register_com_alloc(u_char type,
4029 if_com_alloc_t *a, if_com_free_t *f)
4032 KASSERT(if_com_alloc[type] == NULL,
4033 ("if_register_com_alloc: %d already registered", type));
4034 KASSERT(if_com_free[type] == NULL,
4035 ("if_register_com_alloc: %d free already registered", type));
4037 if_com_alloc[type] = a;
4038 if_com_free[type] = f;
4042 if_deregister_com_alloc(u_char type)
4045 KASSERT(if_com_alloc[type] != NULL,
4046 ("if_deregister_com_alloc: %d not registered", type));
4047 KASSERT(if_com_free[type] != NULL,
4048 ("if_deregister_com_alloc: %d free not registered", type));
4049 if_com_alloc[type] = NULL;
4050 if_com_free[type] = NULL;
4053 /* API for driver access to network stack owned ifnet.*/
4055 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4059 oldbrate = ifp->if_baudrate;
4060 ifp->if_baudrate = baudrate;
4065 if_getbaudrate(if_t ifp)
4068 return (((struct ifnet *)ifp)->if_baudrate);
4072 if_setcapabilities(if_t ifp, int capabilities)
4074 ((struct ifnet *)ifp)->if_capabilities = capabilities;
4079 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4081 ((struct ifnet *)ifp)->if_capabilities |= setbit;
4082 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4088 if_getcapabilities(if_t ifp)
4090 return ((struct ifnet *)ifp)->if_capabilities;
4094 if_setcapenable(if_t ifp, int capabilities)
4096 ((struct ifnet *)ifp)->if_capenable = capabilities;
4101 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4104 ((struct ifnet *)ifp)->if_capenable |= setcap;
4106 ((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4112 if_getdname(if_t ifp)
4114 return ((struct ifnet *)ifp)->if_dname;
4118 if_togglecapenable(if_t ifp, int togglecap)
4120 ((struct ifnet *)ifp)->if_capenable ^= togglecap;
4125 if_getcapenable(if_t ifp)
4127 return ((struct ifnet *)ifp)->if_capenable;
4131 * This is largely undesirable because it ties ifnet to a device, but does
4132 * provide flexiblity for an embedded product vendor. Should be used with
4133 * the understanding that it violates the interface boundaries, and should be
4134 * a last resort only.
4137 if_setdev(if_t ifp, void *dev)
4143 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4145 ((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4146 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4152 if_getdrvflags(if_t ifp)
4154 return ((struct ifnet *)ifp)->if_drv_flags;
4158 if_setdrvflags(if_t ifp, int flags)
4160 ((struct ifnet *)ifp)->if_drv_flags = flags;
4166 if_setflags(if_t ifp, int flags)
4168 ((struct ifnet *)ifp)->if_flags = flags;
4173 if_setflagbits(if_t ifp, int set, int clear)
4175 ((struct ifnet *)ifp)->if_flags |= set;
4176 ((struct ifnet *)ifp)->if_flags &= ~clear;
4182 if_getflags(if_t ifp)
4184 return ((struct ifnet *)ifp)->if_flags;
4188 if_clearhwassist(if_t ifp)
4190 ((struct ifnet *)ifp)->if_hwassist = 0;
4195 if_sethwassistbits(if_t ifp, int toset, int toclear)
4197 ((struct ifnet *)ifp)->if_hwassist |= toset;
4198 ((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4204 if_sethwassist(if_t ifp, int hwassist_bit)
4206 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4211 if_gethwassist(if_t ifp)
4213 return ((struct ifnet *)ifp)->if_hwassist;
4217 if_setmtu(if_t ifp, int mtu)
4219 ((struct ifnet *)ifp)->if_mtu = mtu;
4226 return ((struct ifnet *)ifp)->if_mtu;
4230 if_getmtu_family(if_t ifp, int family)
4234 for (dp = domains; dp; dp = dp->dom_next) {
4235 if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4236 return (dp->dom_ifmtu((struct ifnet *)ifp));
4239 return (((struct ifnet *)ifp)->if_mtu);
4243 * Methods for drivers to access interface unicast and multicast
4244 * link level addresses. Driver shall not know 'struct ifaddr' neither
4245 * 'struct ifmultiaddr'.
4248 if_lladdr_count(if_t ifp)
4250 struct epoch_tracker et;
4255 NET_EPOCH_ENTER(et);
4256 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4257 if (ifa->ifa_addr->sa_family == AF_LINK)
4265 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4267 struct epoch_tracker et;
4274 NET_EPOCH_ENTER(et);
4275 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4276 if (ifa->ifa_addr->sa_family != AF_LINK)
4278 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4287 if_llmaddr_count(if_t ifp)
4289 struct epoch_tracker et;
4290 struct ifmultiaddr *ifma;
4294 NET_EPOCH_ENTER(et);
4295 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4296 if (ifma->ifma_addr->sa_family == AF_LINK)
4304 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4306 struct epoch_tracker et;
4307 struct ifmultiaddr *ifma;
4313 NET_EPOCH_ENTER(et);
4314 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4315 if (ifma->ifma_addr->sa_family != AF_LINK)
4317 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4326 if_setsoftc(if_t ifp, void *softc)
4328 ((struct ifnet *)ifp)->if_softc = softc;
4333 if_getsoftc(if_t ifp)
4335 return ((struct ifnet *)ifp)->if_softc;
4339 if_setrcvif(struct mbuf *m, if_t ifp)
4342 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4343 m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4347 if_setvtag(struct mbuf *m, uint16_t tag)
4349 m->m_pkthdr.ether_vtag = tag;
4353 if_getvtag(struct mbuf *m)
4356 return (m->m_pkthdr.ether_vtag);
4360 if_sendq_empty(if_t ifp)
4362 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4366 if_getifaddr(if_t ifp)
4368 return ((struct ifnet *)ifp)->if_addr;
4372 if_getamcount(if_t ifp)
4374 return ((struct ifnet *)ifp)->if_amcount;
4379 if_setsendqready(if_t ifp)
4381 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4386 if_setsendqlen(if_t ifp, int tx_desc_count)
4388 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4389 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4395 if_vlantrunkinuse(if_t ifp)
4397 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4401 if_input(if_t ifp, struct mbuf* sendmp)
4403 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4409 #ifndef ETH_ADDR_LEN
4410 #define ETH_ADDR_LEN 6
4414 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4416 struct ifmultiaddr *ifma;
4417 uint8_t *lmta = (uint8_t *)mta;
4420 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4421 if (ifma->ifma_addr->sa_family != AF_LINK)
4427 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4428 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4437 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4441 if_maddr_rlock(ifp);
4442 error = if_setupmultiaddr(ifp, mta, cnt, max);
4443 if_maddr_runlock(ifp);
4448 if_multiaddr_count(if_t ifp, int max)
4450 struct ifmultiaddr *ifma;
4454 if_maddr_rlock(ifp);
4455 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4456 if (ifma->ifma_addr->sa_family != AF_LINK)
4462 if_maddr_runlock(ifp);
4467 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4469 struct ifmultiaddr *ifma;
4472 if_maddr_rlock(ifp);
4473 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4474 cnt += filter(arg, ifma, cnt);
4475 if_maddr_runlock(ifp);
4480 if_dequeue(if_t ifp)
4483 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4489 if_sendq_prepend(if_t ifp, struct mbuf *m)
4491 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4496 if_setifheaderlen(if_t ifp, int len)
4498 ((struct ifnet *)ifp)->if_hdrlen = len;
4503 if_getlladdr(if_t ifp)
4505 return (IF_LLADDR((struct ifnet *)ifp));
4509 if_gethandle(u_char type)
4511 return (if_alloc(type));
4515 if_bpfmtap(if_t ifh, struct mbuf *m)
4517 struct ifnet *ifp = (struct ifnet *)ifh;
4523 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4525 struct ifnet *ifp = (struct ifnet *)ifh;
4527 ETHER_BPF_MTAP(ifp, m);
4531 if_vlancap(if_t ifh)
4533 struct ifnet *ifp = (struct ifnet *)ifh;
4534 VLAN_CAPABILITIES(ifp);
4538 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4541 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4546 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4549 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4554 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4557 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4562 if_gethwtsomax(if_t ifp)
4565 return (((struct ifnet *)ifp)->if_hw_tsomax);
4569 if_gethwtsomaxsegcount(if_t ifp)
4572 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4576 if_gethwtsomaxsegsize(if_t ifp)
4579 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4583 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4585 ((struct ifnet *)ifp)->if_init = init_fn;
4589 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4591 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4595 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4597 ((struct ifnet *)ifp)->if_start = (void *)start_fn;
4601 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4603 ((struct ifnet *)ifp)->if_transmit = start_fn;
4606 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4608 ((struct ifnet *)ifp)->if_qflush = flush_fn;
4613 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4616 ifp->if_get_counter = fn;
4619 /* Revisit these - These are inline functions originally. */
4621 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4623 return drbr_inuse(ifh, br);
4627 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4629 return drbr_dequeue(ifh, br);
4633 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4635 return drbr_needs_enqueue(ifh, br);
4639 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4641 return drbr_enqueue(ifh, br, m);