xref: /freebsd/sys/net/if.c (revision 00d96da231d007673a1672452748d8ea4f6788ae)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org>
5  * Copyright (c) 1980, 1986, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include "opt_bpf.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_ddb.h"
37 
38 #include <sys/param.h>
39 #include <sys/capsicum.h>
40 #include <sys/conf.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/domainset.h>
44 #include <sys/sbuf.h>
45 #include <sys/bus.h>
46 #include <sys/epoch.h>
47 #include <sys/mbuf.h>
48 #include <sys/systm.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/refcount.h>
57 #include <sys/module.h>
58 #include <sys/nv.h>
59 #include <sys/rwlock.h>
60 #include <sys/sockio.h>
61 #include <sys/stdarg.h>
62 #include <sys/syslog.h>
63 #include <sys/sysctl.h>
64 #include <sys/sysent.h>
65 #include <sys/taskqueue.h>
66 #include <sys/domain.h>
67 #include <sys/jail.h>
68 #include <sys/priv.h>
69 
70 #ifdef DDB
71 #include <ddb/ddb.h>
72 #endif
73 
74 #include <vm/uma.h>
75 
76 #include <net/bpf.h>
77 #include <net/if.h>
78 #include <net/if_arp.h>
79 #include <net/if_clone.h>
80 #include <net/if_dl.h>
81 #include <net/if_strings.h>
82 #include <net/if_types.h>
83 #include <net/if_var.h>
84 #include <net/if_media.h>
85 #include <net/if_mib.h>
86 #include <net/if_private.h>
87 #include <net/if_vlan_var.h>
88 #include <net/radix.h>
89 #include <net/route.h>
90 #include <net/route/route_ctl.h>
91 #include <net/vnet.h>
92 
93 #if defined(INET) || defined(INET6)
94 #include <net/ethernet.h>
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip.h>
98 #include <netinet/ip_carp.h>
99 #ifdef INET
100 #include <net/debugnet.h>
101 #include <netinet/if_ether.h>
102 #endif /* INET */
103 #ifdef INET6
104 #include <netinet6/in6_var.h>
105 #endif /* INET6 */
106 #endif /* INET || INET6 */
107 
108 #include <security/mac/mac_framework.h>
109 
110 /*
111  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
112  * and ifr_ifru when it is used in SIOCGIFCONF.
113  */
114 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
115     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
116 
117 __read_mostly epoch_t net_epoch_preempt;
118 #ifdef COMPAT_FREEBSD32
119 #include <sys/mount.h>
120 #include <compat/freebsd32/freebsd32.h>
121 
122 struct ifreq_buffer32 {
123 	uint32_t	length;		/* (size_t) */
124 	uint32_t	buffer;		/* (void *) */
125 };
126 
127 /*
128  * Interface request structure used for socket
129  * ioctl's.  All interface ioctl's must have parameter
130  * definitions which begin with ifr_name.  The
131  * remainder may be interface specific.
132  */
133 struct ifreq32 {
134 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
135 	union {
136 		struct sockaddr	ifru_addr;
137 		struct sockaddr	ifru_dstaddr;
138 		struct sockaddr	ifru_broadaddr;
139 		struct ifreq_buffer32 ifru_buffer;
140 		short		ifru_flags[2];
141 		short		ifru_index;
142 		int		ifru_jid;
143 		int		ifru_metric;
144 		int		ifru_mtu;
145 		int		ifru_phys;
146 		int		ifru_media;
147 		uint32_t	ifru_data;
148 		int		ifru_cap[2];
149 		u_int		ifru_fib;
150 		u_char		ifru_vlan_pcp;
151 	} ifr_ifru;
152 };
153 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
154 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
155     __offsetof(struct ifreq32, ifr_ifru));
156 
157 struct ifconf32 {
158 	int32_t	ifc_len;
159 	union {
160 		uint32_t	ifcu_buf;
161 		uint32_t	ifcu_req;
162 	} ifc_ifcu;
163 };
164 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
165 
166 struct ifdrv32 {
167 	char		ifd_name[IFNAMSIZ];
168 	uint32_t	ifd_cmd;
169 	uint32_t	ifd_len;
170 	uint32_t	ifd_data;
171 };
172 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
173 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
174 
175 struct ifgroupreq32 {
176 	char	ifgr_name[IFNAMSIZ];
177 	u_int	ifgr_len;
178 	union {
179 		char		ifgru_group[IFNAMSIZ];
180 		uint32_t	ifgru_groups;
181 	} ifgr_ifgru;
182 };
183 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
184 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
185 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
186 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
187 
188 struct ifmediareq32 {
189 	char		ifm_name[IFNAMSIZ];
190 	int		ifm_current;
191 	int		ifm_mask;
192 	int		ifm_status;
193 	int		ifm_active;
194 	int		ifm_count;
195 	uint32_t	ifm_ulist;	/* (int *) */
196 };
197 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
198 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
199 #endif /* COMPAT_FREEBSD32 */
200 
201 union ifreq_union {
202 	struct ifreq	ifr;
203 #ifdef COMPAT_FREEBSD32
204 	struct ifreq32	ifr32;
205 #endif
206 };
207 
208 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
209     "Link layers");
210 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
211     "Generic link-management");
212 
213 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
214     &ifqmaxlen, 0, "max send queue size");
215 
216 /* Log link state change events */
217 static int log_link_state_change = 1;
218 
219 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
220 	&log_link_state_change, 0,
221 	"log interface link state change events");
222 
223 /* Log promiscuous mode change events */
224 static int log_promisc_mode_change = 1;
225 
226 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
227 	&log_promisc_mode_change, 1,
228 	"log promiscuous mode change events");
229 
230 /* Interface description */
231 static unsigned int ifdescr_maxlen = 1024;
232 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
233 	&ifdescr_maxlen, 0,
234 	"administrative maximum length for interface description");
235 
236 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
237 
238 /* global sx for non-critical path ifdescr */
239 static struct sx ifdescr_sx;
240 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
241 
242 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
243 /* These are external hooks for CARP. */
244 void	(*carp_linkstate_p)(struct ifnet *ifp);
245 void	(*carp_demote_adj_p)(int, char *);
246 int	(*carp_master_p)(struct ifaddr *);
247 #if defined(INET) || defined(INET6)
248 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
249 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
250     const struct sockaddr *sa);
251 int	(*carp_attach_p)(struct ifaddr *, int);
252 void	(*carp_detach_p)(struct ifaddr *, bool);
253 #endif
254 #ifdef INET
255 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
256 #endif
257 #ifdef INET6
258 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
259 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
260     const struct in6_addr *taddr);
261 #endif
262 
263 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
264 
265 /*
266  * XXX: Style; these should be sorted alphabetically, and unprototyped
267  * static functions should be prototyped. Currently they are sorted by
268  * declaration order.
269  */
270 static int	ifconf(u_long, caddr_t);
271 static void	if_input_default(struct ifnet *, struct mbuf *);
272 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
273 static int	if_setflag(struct ifnet *, int, int, int *, int);
274 static int	if_transmit_default(struct ifnet *ifp, struct mbuf *m);
275 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
276 static void	do_link_state_change(void *, int);
277 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
278 static int	if_getgroupmembers(struct ifgroupreq *);
279 static void	if_delgroups(struct ifnet *);
280 static void	if_attach_internal(struct ifnet *, bool);
281 static void	if_detach_internal(struct ifnet *, bool);
282 static void	if_siocaddmulti(void *, int);
283 static void	if_link_ifnet(struct ifnet *);
284 static bool	if_unlink_ifnet(struct ifnet *, bool);
285 #ifdef VIMAGE
286 static void	if_vmove(struct ifnet *, struct vnet *);
287 #endif
288 
289 #ifdef INET6
290 /*
291  * XXX: declare here to avoid to include many inet6 related files..
292  * should be more generalized?
293  */
294 extern void	nd6_setmtu(struct ifnet *);
295 #endif
296 
297 /* ipsec helper hooks */
298 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
299 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
300 
301 int	ifqmaxlen = IFQ_MAXLEN;
302 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
303 VNET_DEFINE(struct ifgrouphead, ifg_head);
304 
305 /* Table of ifnet by index. */
306 static int if_index;
307 static int if_indexlim = 8;
308 static struct ifindex_entry {
309 	struct ifnet	*ife_ifnet;
310 	uint16_t	ife_gencnt;
311 } *ifindex_table;
312 
313 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
314     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
315     "Variables global to all interfaces");
316 static int
sysctl_ifcount(SYSCTL_HANDLER_ARGS)317 sysctl_ifcount(SYSCTL_HANDLER_ARGS)
318 {
319 	int rv = 0;
320 
321 	IFNET_RLOCK();
322 	for (int i = 1; i <= if_index; i++)
323 		if (ifindex_table[i].ife_ifnet != NULL &&
324 		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
325 			rv = i;
326 	IFNET_RUNLOCK();
327 
328 	return (sysctl_handle_int(oidp, &rv, 0, req));
329 }
330 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
331     CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
332     "Maximum known interface index");
333 
334 /*
335  * The global network interface list (V_ifnet) and related state (such as
336  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
337  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
338  */
339 struct sx ifnet_sxlock;
340 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
341 
342 struct sx ifnet_detach_sxlock;
343 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
344     SX_RECURSE);
345 
346 static	if_com_alloc_t *if_com_alloc[256];
347 static	if_com_free_t *if_com_free[256];
348 
349 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
350 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
351 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
352 
353 struct ifnet *
ifnet_byindex(u_int idx)354 ifnet_byindex(u_int idx)
355 {
356 	struct ifnet *ifp;
357 
358 	NET_EPOCH_ASSERT();
359 
360 	if (__predict_false(idx > if_index))
361 		return (NULL);
362 
363 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
364 
365 	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
366 		ifp = NULL;
367 
368 	return (ifp);
369 }
370 
371 struct ifnet *
ifnet_byindex_ref(u_int idx)372 ifnet_byindex_ref(u_int idx)
373 {
374 	struct ifnet *ifp;
375 
376 	ifp = ifnet_byindex(idx);
377 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
378 		return (NULL);
379 	if (!if_try_ref(ifp))
380 		return (NULL);
381 	return (ifp);
382 }
383 
384 struct ifnet *
ifnet_byindexgen(uint16_t idx,uint16_t gen)385 ifnet_byindexgen(uint16_t idx, uint16_t gen)
386 {
387 	struct ifnet *ifp;
388 
389 	NET_EPOCH_ASSERT();
390 
391 	if (__predict_false(idx > if_index))
392 		return (NULL);
393 
394 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
395 
396 	if (ifindex_table[idx].ife_gencnt == gen)
397 		return (ifp);
398 	else
399 		return (NULL);
400 }
401 
402 /*
403  * Network interface utility routines.
404  *
405  * Routines with ifa_ifwith* names take sockaddr *'s as
406  * parameters.
407  */
408 
409 static void
if_init_idxtable(void * arg __unused)410 if_init_idxtable(void *arg __unused)
411 {
412 
413 	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
414 	    M_IFNET, M_WAITOK | M_ZERO);
415 }
416 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL);
417 
418 static void
vnet_if_init(const void * unused __unused)419 vnet_if_init(const void *unused __unused)
420 {
421 
422 	CK_STAILQ_INIT(&V_ifnet);
423 	CK_STAILQ_INIT(&V_ifg_head);
424 }
425 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
426     NULL);
427 
428 static void
if_link_ifnet(struct ifnet * ifp)429 if_link_ifnet(struct ifnet *ifp)
430 {
431 	IFNET_WLOCK();
432 
433 	MPASS(refcount_load(&ifp->if_refcount) > 0);
434 	MPASS(ifp->if_vnet == curvnet);
435 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
436 
437 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
438 #ifdef VIMAGE
439 	curvnet->vnet_ifcnt++;
440 #endif
441 	IFNET_WUNLOCK();
442 }
443 
444 static bool
if_unlink_ifnet(struct ifnet * ifp,bool vmove)445 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
446 {
447 	struct ifnet *iter;
448 	int found = 0;
449 
450 	IFNET_WLOCK();
451 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
452 		if (iter == ifp) {
453 			MPASS(refcount_load(&ifp->if_refcount) > 0);
454 			MPASS(ifp->if_vnet == curvnet);
455 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
456 
457 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
458 #ifdef VIMAGE
459 			curvnet->vnet_ifcnt--;
460 #endif
461 			if (!vmove)
462 				ifp->if_flags |= IFF_DYING;
463 			found = 1;
464 			break;
465 		}
466 	IFNET_WUNLOCK();
467 
468 	return (found);
469 }
470 
471 #ifdef VIMAGE
472 static void
vnet_if_return(const void * unused __unused)473 vnet_if_return(const void *unused __unused)
474 {
475 	struct ifnet *ifp, *nifp;
476 	struct ifnet **pending;
477 	int found __diagused;
478 	int i;
479 
480 	i = 0;
481 
482 	/*
483 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
484 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
485 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
486 	 * complete. We can't do that from within NET_EPOCH.
487 	 *
488 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
489 	 * read/write lock. We cannot hold the lock as we call if_vmove()
490 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
491 	 * ctx lock.
492 	 */
493 	IFNET_WLOCK();
494 
495 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
496 	    M_IFNET, M_WAITOK | M_ZERO);
497 
498 	/* Return all inherited interfaces to their parent vnets. */
499 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
500 		if (ifp->if_home_vnet != ifp->if_vnet) {
501 			found = if_unlink_ifnet(ifp, true);
502 			MPASS(found);
503 
504 			pending[i++] = ifp;
505 		}
506 	}
507 	IFNET_WUNLOCK();
508 
509 	for (int j = 0; j < i; j++) {
510 		sx_xlock(&ifnet_detach_sxlock);
511 		if_vmove(pending[j], pending[j]->if_home_vnet);
512 		sx_xunlock(&ifnet_detach_sxlock);
513 	}
514 
515 	free(pending, M_IFNET);
516 }
517 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
518     vnet_if_return, NULL);
519 #endif
520 
521 /*
522  * Allocate a struct ifnet and an index for an interface.  A layer 2
523  * common structure will also be allocated if an allocation routine is
524  * registered for the passed type.
525  */
526 static struct ifnet *
if_alloc_domain(u_char type,int numa_domain)527 if_alloc_domain(u_char type, int numa_domain)
528 {
529 	struct ifnet *ifp;
530 	u_short idx;
531 
532 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
533 	if (numa_domain == IF_NODOM)
534 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
535 		    M_WAITOK | M_ZERO);
536 	else
537 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
538 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
539 	ifp->if_type = type;
540 	ifp->if_alloctype = type;
541 	ifp->if_numa_domain = numa_domain;
542 #ifdef VIMAGE
543 	ifp->if_vnet = curvnet;
544 #endif
545 	if (if_com_alloc[type] != NULL) {
546 		ifp->if_l2com = if_com_alloc[type](type, ifp);
547 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
548 		    type));
549 	}
550 
551 	IF_ADDR_LOCK_INIT(ifp);
552 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
553 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
554 	CK_STAILQ_INIT(&ifp->if_addrhead);
555 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
556 	CK_STAILQ_INIT(&ifp->if_groups);
557 #ifdef MAC
558 	mac_ifnet_init(ifp);
559 #endif
560 	ifq_init(&ifp->if_snd, ifp);
561 
562 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
563 	for (int i = 0; i < IFCOUNTERS; i++)
564 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
565 	ifp->if_get_counter = if_get_counter_default;
566 	ifp->if_pcp = IFNET_PCP_NONE;
567 
568 	/* Allocate an ifindex array entry. */
569 	IFNET_WLOCK();
570 	/*
571 	 * Try to find an empty slot below if_index.  If we fail, take the
572 	 * next slot.
573 	 */
574 	for (idx = 1; idx <= if_index; idx++) {
575 		if (ifindex_table[idx].ife_ifnet == NULL)
576 			break;
577 	}
578 
579 	/* Catch if_index overflow. */
580 	if (idx >= if_indexlim) {
581 		struct ifindex_entry *new, *old;
582 		int newlim;
583 
584 		newlim = if_indexlim * 2;
585 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
586 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
587 		old = ifindex_table;
588 		ck_pr_store_ptr(&ifindex_table, new);
589 		if_indexlim = newlim;
590 		NET_EPOCH_WAIT();
591 		free(old, M_IFNET);
592 	}
593 	if (idx > if_index)
594 		if_index = idx;
595 
596 	ifp->if_index = idx;
597 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
598 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
599 	IFNET_WUNLOCK();
600 
601 	return (ifp);
602 }
603 
604 struct ifnet *
if_alloc_dev(u_char type,device_t dev)605 if_alloc_dev(u_char type, device_t dev)
606 {
607 	int numa_domain;
608 
609 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
610 		return (if_alloc_domain(type, IF_NODOM));
611 	return (if_alloc_domain(type, numa_domain));
612 }
613 
614 struct ifnet *
if_alloc(u_char type)615 if_alloc(u_char type)
616 {
617 
618 	return (if_alloc_domain(type, IF_NODOM));
619 }
620 /*
621  * Do the actual work of freeing a struct ifnet, and layer 2 common
622  * structure.  This call is made when the network epoch guarantees
623  * us that nobody holds a pointer to the interface.
624  */
625 static void
if_free_deferred(epoch_context_t ctx)626 if_free_deferred(epoch_context_t ctx)
627 {
628 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
629 
630 	KASSERT((ifp->if_flags & IFF_DYING),
631 	    ("%s: interface not dying", __func__));
632 
633 	if (if_com_free[ifp->if_alloctype] != NULL)
634 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
635 		    ifp->if_alloctype);
636 
637 #ifdef MAC
638 	mac_ifnet_destroy(ifp);
639 #endif /* MAC */
640 	IF_ADDR_LOCK_DESTROY(ifp);
641 	ifq_delete(&ifp->if_snd);
642 
643 	for (int i = 0; i < IFCOUNTERS; i++)
644 		counter_u64_free(ifp->if_counters[i]);
645 
646 	if_freedescr(ifp->if_description);
647 	free(ifp->if_hw_addr, M_IFADDR);
648 	free(ifp, M_IFNET);
649 }
650 
651 /*
652  * Deregister an interface and free the associated storage.
653  */
654 void
if_free(struct ifnet * ifp)655 if_free(struct ifnet *ifp)
656 {
657 
658 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
659 
660 	/*
661 	 * XXXGL: An interface index is really an alias to ifp pointer.
662 	 * Why would we clear the alias now, and not in the deferred
663 	 * context?  Indeed there is nothing wrong with some network
664 	 * thread obtaining ifp via ifnet_byindex() inside the network
665 	 * epoch and then dereferencing ifp while we perform if_free(),
666 	 * and after if_free() finished, too.
667 	 *
668 	 * This early index freeing was important back when ifindex was
669 	 * virtualized and interface would outlive the vnet.
670 	 */
671 	IFNET_WLOCK();
672 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
673 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
674 	ifindex_table[ifp->if_index].ife_gencnt++;
675 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
676 		if_index--;
677 	IFNET_WUNLOCK();
678 
679 	if (refcount_release(&ifp->if_refcount))
680 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
681 }
682 
683 /*
684  * Interfaces to keep an ifnet type-stable despite the possibility of the
685  * driver calling if_free().  If there are additional references, we defer
686  * freeing the underlying data structure.
687  */
688 void
if_ref(struct ifnet * ifp)689 if_ref(struct ifnet *ifp)
690 {
691 	u_int old __diagused;
692 
693 	/* We don't assert the ifnet list lock here, but arguably should. */
694 	old = refcount_acquire(&ifp->if_refcount);
695 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
696 }
697 
698 bool
if_try_ref(struct ifnet * ifp)699 if_try_ref(struct ifnet *ifp)
700 {
701 	NET_EPOCH_ASSERT();
702 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
703 }
704 
705 void
if_rele(struct ifnet * ifp)706 if_rele(struct ifnet *ifp)
707 {
708 
709 	if (!refcount_release(&ifp->if_refcount))
710 		return;
711 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
712 }
713 
714 void
ifq_init(struct ifaltq * ifq,struct ifnet * ifp)715 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
716 {
717 
718 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
719 
720 	if (ifq->ifq_maxlen == 0)
721 		ifq->ifq_maxlen = ifqmaxlen;
722 
723 	ifq->altq_type = 0;
724 	ifq->altq_disc = NULL;
725 	ifq->altq_flags &= ALTQF_CANTCHANGE;
726 	ifq->altq_tbr  = NULL;
727 	ifq->altq_ifp  = ifp;
728 }
729 
730 void
ifq_delete(struct ifaltq * ifq)731 ifq_delete(struct ifaltq *ifq)
732 {
733 	mtx_destroy(&ifq->ifq_mtx);
734 }
735 
736 /*
737  * Perform generic interface initialization tasks and attach the interface
738  * to the list of "active" interfaces.  If vmove flag is set on entry
739  * to if_attach_internal(), perform only a limited subset of initialization
740  * tasks, given that we are moving from one vnet to another an ifnet which
741  * has already been fully initialized.
742  *
743  * Note that if_detach_internal() removes group membership unconditionally
744  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
745  * Thus, when if_vmove() is applied to a cloned interface, group membership
746  * is lost while a cloned one always joins a group whose name is
747  * ifc->ifc_name.  To recover this after if_detach_internal() and
748  * if_attach_internal(), the cloner should be specified to
749  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
750  * attempts to join a group whose name is ifc->ifc_name.
751  *
752  * XXX:
753  *  - The decision to return void and thus require this function to
754  *    succeed is questionable.
755  *  - We should probably do more sanity checking.  For instance we don't
756  *    do anything to insure if_xname is unique or non-empty.
757  */
758 void
if_attach(struct ifnet * ifp)759 if_attach(struct ifnet *ifp)
760 {
761 
762 	if_attach_internal(ifp, false);
763 }
764 
765 /*
766  * Compute the least common TSO limit.
767  */
768 void
if_hw_tsomax_common(if_t ifp,struct ifnet_hw_tsomax * pmax)769 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
770 {
771 	/*
772 	 * 1) If there is no limit currently, take the limit from
773 	 * the network adapter.
774 	 *
775 	 * 2) If the network adapter has a limit below the current
776 	 * limit, apply it.
777 	 */
778 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
779 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
780 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
781 	}
782 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
783 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
784 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
785 	}
786 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
787 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
788 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
789 	}
790 }
791 
792 /*
793  * Update TSO limit of a network adapter.
794  *
795  * Returns zero if no change. Else non-zero.
796  */
797 int
if_hw_tsomax_update(if_t ifp,struct ifnet_hw_tsomax * pmax)798 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
799 {
800 	int retval = 0;
801 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
802 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
803 		retval++;
804 	}
805 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
806 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
807 		retval++;
808 	}
809 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
810 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
811 		retval++;
812 	}
813 	return (retval);
814 }
815 
816 static void
if_attach_internal(struct ifnet * ifp,bool vmove)817 if_attach_internal(struct ifnet *ifp, bool vmove)
818 {
819 	unsigned socksize, ifasize;
820 	int namelen, masklen;
821 	struct sockaddr_dl *sdl;
822 	struct ifaddr *ifa;
823 
824 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
825 
826 #ifdef VIMAGE
827 	CURVNET_ASSERT_SET();
828 	ifp->if_vnet = curvnet;
829 	if (ifp->if_home_vnet == NULL)
830 		ifp->if_home_vnet = curvnet;
831 #endif
832 
833 	if_addgroup(ifp, IFG_ALL);
834 
835 #ifdef VIMAGE
836 	/* Restore group membership for cloned interface. */
837 	if (vmove)
838 		if_clone_restoregroup(ifp);
839 #endif
840 
841 	getmicrotime(&ifp->if_lastchange);
842 	ifp->if_epoch = time_uptime;
843 
844 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
845 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
846 	    ("transmit and qflush must both either be set or both be NULL"));
847 	if (ifp->if_transmit == NULL) {
848 		ifp->if_transmit = if_transmit_default;
849 		ifp->if_qflush = if_qflush;
850 	}
851 	if (ifp->if_input == NULL)
852 		ifp->if_input = if_input_default;
853 
854 	if (ifp->if_requestencap == NULL)
855 		ifp->if_requestencap = if_requestencap_default;
856 
857 	if (!vmove) {
858 #ifdef MAC
859 		mac_ifnet_create(ifp);
860 #endif
861 
862 		/*
863 		 * Create a Link Level name for this device.
864 		 */
865 		namelen = strlen(ifp->if_xname);
866 		/*
867 		 * Always save enough space for any possible name so we
868 		 * can do a rename in place later.
869 		 */
870 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
871 		socksize = masklen + ifp->if_addrlen;
872 		if (socksize < sizeof(*sdl))
873 			socksize = sizeof(*sdl);
874 		socksize = roundup2(socksize, sizeof(long));
875 		ifasize = sizeof(*ifa) + 2 * socksize;
876 		ifa = ifa_alloc(ifasize, M_WAITOK);
877 		sdl = (struct sockaddr_dl *)(ifa + 1);
878 		sdl->sdl_len = socksize;
879 		sdl->sdl_family = AF_LINK;
880 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
881 		sdl->sdl_nlen = namelen;
882 		sdl->sdl_index = ifp->if_index;
883 		sdl->sdl_type = ifp->if_type;
884 		ifp->if_addr = ifa;
885 		ifa->ifa_ifp = ifp;
886 		ifa->ifa_addr = (struct sockaddr *)sdl;
887 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
888 		ifa->ifa_netmask = (struct sockaddr *)sdl;
889 		sdl->sdl_len = masklen;
890 		while (namelen != 0)
891 			sdl->sdl_data[--namelen] = 0xff;
892 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
893 		/* Reliably crash if used uninitialized. */
894 		ifp->if_broadcastaddr = NULL;
895 
896 		if (ifp->if_type == IFT_ETHER) {
897 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
898 			    M_WAITOK | M_ZERO);
899 		}
900 
901 #if defined(INET) || defined(INET6)
902 		/* Use defaults for TSO, if nothing is set */
903 		if (ifp->if_hw_tsomax == 0 &&
904 		    ifp->if_hw_tsomaxsegcount == 0 &&
905 		    ifp->if_hw_tsomaxsegsize == 0) {
906 			/*
907 			 * The TSO defaults needs to be such that an
908 			 * NFS mbuf list of 35 mbufs totalling just
909 			 * below 64K works and that a chain of mbufs
910 			 * can be defragged into at most 32 segments:
911 			 */
912 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
913 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
914 			ifp->if_hw_tsomaxsegcount = 35;
915 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
916 
917 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
918 			if (ifp->if_capabilities & IFCAP_TSO) {
919 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
920 				    ifp->if_hw_tsomax,
921 				    ifp->if_hw_tsomaxsegcount,
922 				    ifp->if_hw_tsomaxsegsize);
923 			}
924 		}
925 #endif
926 	}
927 
928 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
929 	if_link_ifnet(ifp);
930 	EVENTHANDLER_INVOKE(ifnet_attached_event, ifp);
931 	if (IS_DEFAULT_VNET(curvnet))
932 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
933 }
934 
935 static void
if_epochalloc(void * dummy __unused)936 if_epochalloc(void *dummy __unused)
937 {
938 
939 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
940 }
941 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
942 
943 /*
944  * Remove any unicast or broadcast network addresses from an interface.
945  */
946 void
if_purgeaddrs(struct ifnet * ifp)947 if_purgeaddrs(struct ifnet *ifp)
948 {
949 	struct ifaddr *ifa;
950 
951 #ifdef INET6
952 	/*
953 	 * Need to leave multicast addresses of proxy NDP llentries
954 	 * before in6_purgeifaddr() because the llentries are keys
955 	 * for in6_multi objects of proxy NDP entries.
956 	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
957 	 * then we would lose the keys if they are called earlier.
958 	 */
959 	in6_purge_proxy_ndp(ifp);
960 #endif
961 	while (1) {
962 		struct epoch_tracker et;
963 
964 		NET_EPOCH_ENTER(et);
965 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
966 			if (ifa->ifa_addr->sa_family != AF_LINK)
967 				break;
968 		}
969 		NET_EPOCH_EXIT(et);
970 
971 		if (ifa == NULL)
972 			break;
973 #ifdef INET
974 		/* XXX: Ugly!! ad hoc just for INET */
975 		if (ifa->ifa_addr->sa_family == AF_INET) {
976 			struct ifreq ifr;
977 
978 			bzero(&ifr, sizeof(ifr));
979 			ifr.ifr_addr = *ifa->ifa_addr;
980 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
981 			    NULL) == 0)
982 				continue;
983 		}
984 #endif /* INET */
985 #ifdef INET6
986 		if (ifa->ifa_addr->sa_family == AF_INET6) {
987 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
988 			/* ifp_addrhead is already updated */
989 			continue;
990 		}
991 #endif /* INET6 */
992 		IF_ADDR_WLOCK(ifp);
993 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
994 		IF_ADDR_WUNLOCK(ifp);
995 		ifa_free(ifa);
996 	}
997 }
998 
999 /*
1000  * Remove any multicast network addresses from an interface when an ifnet
1001  * is going away.
1002  */
1003 static void
if_purgemaddrs(struct ifnet * ifp)1004 if_purgemaddrs(struct ifnet *ifp)
1005 {
1006 	struct ifmultiaddr *ifma;
1007 
1008 	IF_ADDR_WLOCK(ifp);
1009 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1010 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1011 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1012 		if_delmulti_locked(ifp, ifma, 1);
1013 	}
1014 	IF_ADDR_WUNLOCK(ifp);
1015 }
1016 
1017 /*
1018  * Detach an interface, removing it from the list of "active" interfaces.
1019  * If vmove flag is set on entry to if_detach_internal(), perform only a
1020  * limited subset of cleanup tasks, given that we are moving an ifnet from
1021  * one vnet to another, where it must be fully operational.
1022  *
1023  * XXXRW: There are some significant questions about event ordering, and
1024  * how to prevent things from starting to use the interface during detach.
1025  */
1026 void
if_detach(struct ifnet * ifp)1027 if_detach(struct ifnet *ifp)
1028 {
1029 	bool found;
1030 
1031 	CURVNET_SET_QUIET(ifp->if_vnet);
1032 	found = if_unlink_ifnet(ifp, false);
1033 	if (found) {
1034 		sx_xlock(&ifnet_detach_sxlock);
1035 		if_detach_internal(ifp, false);
1036 		sx_xunlock(&ifnet_detach_sxlock);
1037 	}
1038 	CURVNET_RESTORE();
1039 }
1040 
1041 /*
1042  * The vmove flag, if set, indicates that we are called from a callpath
1043  * that is moving an interface to a different vnet instance.
1044  *
1045  * The shutdown flag, if set, indicates that we are called in the
1046  * process of shutting down a vnet instance.  Currently only the
1047  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1048  * on a vnet instance shutdown without this flag being set, e.g., when
1049  * the cloned interfaces are destoyed as first thing of teardown.
1050  */
1051 static void
if_detach_internal(struct ifnet * ifp,bool vmove)1052 if_detach_internal(struct ifnet *ifp, bool vmove)
1053 {
1054 	struct ifaddr *ifa;
1055 #ifdef VIMAGE
1056 	bool shutdown;
1057 
1058 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1059 #endif
1060 
1061 	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
1062 
1063 	/*
1064 	 * At this point we know the interface still was on the ifnet list
1065 	 * and we removed it so we are in a stable state.
1066 	 */
1067 	NET_EPOCH_WAIT();
1068 
1069 	/*
1070 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1071 	 * fixes issues about late destruction of multicast options
1072 	 * which lead to leave group calls, which in turn access the
1073 	 * belonging ifnet structure:
1074 	 */
1075 	NET_EPOCH_DRAIN_CALLBACKS();
1076 
1077 	/*
1078 	 * In any case (destroy or vmove) detach us from the groups
1079 	 * and remove/wait for pending events on the taskq.
1080 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1081 	 */
1082 	if_delgroups(ifp);
1083 
1084 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1085 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1086 
1087 	if_down(ifp);
1088 
1089 #ifdef VIMAGE
1090 	/*
1091 	 * On VNET shutdown abort here as the stack teardown will do all
1092 	 * the work top-down for us.
1093 	 */
1094 	if (shutdown) {
1095 		/* Give interface users the chance to clean up. */
1096 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1097 
1098 		/*
1099 		 * In case of a vmove we are done here without error.
1100 		 * If we would signal an error it would lead to the same
1101 		 * abort as if we did not find the ifnet anymore.
1102 		 * if_detach() calls us in void context and does not care
1103 		 * about an early abort notification, so life is splendid :)
1104 		 */
1105 		return;
1106 	}
1107 #endif
1108 
1109 	/*
1110 	 * At this point we are not tearing down a VNET and are either
1111 	 * going to destroy or vmove the interface and have to cleanup
1112 	 * accordingly.
1113 	 */
1114 
1115 	/*
1116 	 * Remove routes and flush queues.
1117 	 */
1118 #ifdef ALTQ
1119 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1120 		altq_disable(&ifp->if_snd);
1121 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1122 		altq_detach(&ifp->if_snd);
1123 #endif
1124 
1125 	rt_flushifroutes(ifp);
1126 
1127 	if_purgeaddrs(ifp);
1128 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1129 	if_purgemaddrs(ifp);
1130 	if (IS_DEFAULT_VNET(curvnet))
1131 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1132 
1133 	if (!vmove) {
1134 		/*
1135 		 * Prevent further calls into the device driver via ifnet.
1136 		 */
1137 		if_dead(ifp);
1138 
1139 		/*
1140 		 * Clean up all addresses.
1141 		 */
1142 		IF_ADDR_WLOCK(ifp);
1143 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1144 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1145 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1146 			IF_ADDR_WUNLOCK(ifp);
1147 			ifa_free(ifa);
1148 		} else
1149 			IF_ADDR_WUNLOCK(ifp);
1150 	}
1151 }
1152 
1153 #ifdef VIMAGE
1154 /*
1155  * if_vmove() performs a limited version of if_detach() in current
1156  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1157  */
1158 static void
if_vmove(struct ifnet * ifp,struct vnet * new_vnet)1159 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1160 {
1161 #ifdef DEV_BPF
1162 	/*
1163 	 * Detach BPF file descriptors from its interface.
1164 	 */
1165 	bpf_ifdetach(ifp);
1166 #endif
1167 
1168 	/*
1169 	 * Detach from current vnet, but preserve LLADDR info, do not
1170 	 * mark as dead etc. so that the ifnet can be reattached later.
1171 	 */
1172 	if_detach_internal(ifp, true);
1173 
1174 	/*
1175 	 * Perform interface-specific reassignment tasks, if provided by
1176 	 * the driver.
1177 	 */
1178 	if (ifp->if_reassign != NULL)
1179 		ifp->if_reassign(ifp, new_vnet, NULL);
1180 
1181 	/*
1182 	 * Switch to the context of the target vnet.
1183 	 */
1184 	CURVNET_SET_QUIET(new_vnet);
1185 	if_attach_internal(ifp, true);
1186 	bpf_vmove(ifp->if_bpf);
1187 	CURVNET_RESTORE();
1188 }
1189 
1190 /*
1191  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1192  */
1193 static int
if_vmove_loan(struct thread * td,struct ifnet * ifp,char * ifname,int jid)1194 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1195 {
1196 	struct prison *pr;
1197 	struct ifnet *difp;
1198 	bool found;
1199 
1200 	/* Try to find the prison within our visibility. */
1201 	sx_slock(&allprison_lock);
1202 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1203 	sx_sunlock(&allprison_lock);
1204 	if (pr == NULL)
1205 		return (ENXIO);
1206 	prison_hold_locked(pr);
1207 	mtx_unlock(&pr->pr_mtx);
1208 
1209 	/* Do not try to move the iface from and to the same prison. */
1210 	if (pr->pr_vnet == ifp->if_vnet) {
1211 		prison_free(pr);
1212 		return (EEXIST);
1213 	}
1214 
1215 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1216 	/* XXX Lock interfaces to avoid races. */
1217 	CURVNET_SET_QUIET(pr->pr_vnet);
1218 	difp = ifunit(ifname);
1219 	CURVNET_RESTORE();
1220 	if (difp != NULL) {
1221 		prison_free(pr);
1222 		return (EEXIST);
1223 	}
1224 	sx_xlock(&ifnet_detach_sxlock);
1225 
1226 	found = if_unlink_ifnet(ifp, true);
1227 	if (! found) {
1228 		sx_xunlock(&ifnet_detach_sxlock);
1229 		prison_free(pr);
1230 		return (ENODEV);
1231 	}
1232 
1233 	/* Move the interface into the child jail/vnet. */
1234 	if_vmove(ifp, pr->pr_vnet);
1235 
1236 	/* Report the new if_xname back to the userland. */
1237 	sprintf(ifname, "%s", ifp->if_xname);
1238 
1239 	sx_xunlock(&ifnet_detach_sxlock);
1240 
1241 	prison_free(pr);
1242 	return (0);
1243 }
1244 
1245 static int
if_vmove_reclaim(struct thread * td,char * ifname,int jid)1246 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1247 {
1248 	struct prison *pr;
1249 	struct vnet *vnet_dst;
1250 	struct ifnet *ifp;
1251 	int found;
1252 
1253 	/* Try to find the prison within our visibility. */
1254 	sx_slock(&allprison_lock);
1255 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1256 	sx_sunlock(&allprison_lock);
1257 	if (pr == NULL)
1258 		return (ENXIO);
1259 	prison_hold_locked(pr);
1260 	mtx_unlock(&pr->pr_mtx);
1261 
1262 	/* Make sure the named iface exists in the source prison/vnet. */
1263 	CURVNET_SET(pr->pr_vnet);
1264 	ifp = ifunit(ifname);
1265 	if (ifp == NULL) {
1266 		CURVNET_RESTORE();
1267 		prison_free(pr);
1268 		return (ENXIO);
1269 	}
1270 
1271 	/* Do not try to move the iface from and to the same vnet. */
1272 	vnet_dst = TD_TO_VNET(td);
1273 	if (vnet_dst == pr->pr_vnet) {
1274 		CURVNET_RESTORE();
1275 		prison_free(pr);
1276 		return (EEXIST);
1277 	}
1278 
1279 	/* Get interface back from child jail/vnet. */
1280 	found = if_unlink_ifnet(ifp, true);
1281 	if (! found) {
1282 		CURVNET_RESTORE();
1283 		prison_free(pr);
1284 		return (ENODEV);
1285 	}
1286 	sx_xlock(&ifnet_detach_sxlock);
1287 	if_vmove(ifp, vnet_dst);
1288 	sx_xunlock(&ifnet_detach_sxlock);
1289 	CURVNET_RESTORE();
1290 
1291 	/* Report the new if_xname back to the userland. */
1292 	sprintf(ifname, "%s", ifp->if_xname);
1293 
1294 	prison_free(pr);
1295 	return (0);
1296 }
1297 #endif /* VIMAGE */
1298 
1299 /*
1300  * Add a group to an interface
1301  */
1302 int
if_addgroup(struct ifnet * ifp,const char * groupname)1303 if_addgroup(struct ifnet *ifp, const char *groupname)
1304 {
1305 	struct ifg_list		*ifgl;
1306 	struct ifg_group	*ifg = NULL;
1307 	struct ifg_member	*ifgm;
1308 	int 			 new = 0;
1309 
1310 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1311 	    groupname[strlen(groupname) - 1] <= '9')
1312 		return (EINVAL);
1313 
1314 	IFNET_WLOCK();
1315 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1316 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1317 			IFNET_WUNLOCK();
1318 			return (EEXIST);
1319 		}
1320 
1321 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1322 	    	IFNET_WUNLOCK();
1323 		return (ENOMEM);
1324 	}
1325 
1326 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1327 		free(ifgl, M_TEMP);
1328 		IFNET_WUNLOCK();
1329 		return (ENOMEM);
1330 	}
1331 
1332 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1333 		if (!strcmp(ifg->ifg_group, groupname))
1334 			break;
1335 
1336 	if (ifg == NULL) {
1337 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1338 			free(ifgl, M_TEMP);
1339 			free(ifgm, M_TEMP);
1340 			IFNET_WUNLOCK();
1341 			return (ENOMEM);
1342 		}
1343 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1344 		ifg->ifg_refcnt = 0;
1345 		CK_STAILQ_INIT(&ifg->ifg_members);
1346 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1347 		new = 1;
1348 	}
1349 
1350 	ifg->ifg_refcnt++;
1351 	ifgl->ifgl_group = ifg;
1352 	ifgm->ifgm_ifp = ifp;
1353 
1354 	IF_ADDR_WLOCK(ifp);
1355 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1356 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1357 	IF_ADDR_WUNLOCK(ifp);
1358 
1359 	IFNET_WUNLOCK();
1360 
1361 	if (new)
1362 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1363 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1364 
1365 	return (0);
1366 }
1367 
1368 /*
1369  * Helper function to remove a group out of an interface.  Expects the global
1370  * ifnet lock to be write-locked, and drops it before returning.
1371  */
1372 static void
_if_delgroup_locked(struct ifnet * ifp,struct ifg_list * ifgl,const char * groupname)1373 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1374     const char *groupname)
1375 {
1376 	struct ifg_member *ifgm;
1377 	bool freeifgl;
1378 
1379 	IFNET_WLOCK_ASSERT();
1380 
1381 	IF_ADDR_WLOCK(ifp);
1382 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1383 	IF_ADDR_WUNLOCK(ifp);
1384 
1385 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1386 		if (ifgm->ifgm_ifp == ifp) {
1387 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1388 			    ifg_member, ifgm_next);
1389 			break;
1390 		}
1391 	}
1392 
1393 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1394 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1395 		    ifg_next);
1396 		freeifgl = true;
1397 	} else {
1398 		freeifgl = false;
1399 	}
1400 	IFNET_WUNLOCK();
1401 
1402 	NET_EPOCH_WAIT();
1403 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1404 	if (freeifgl) {
1405 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1406 		free(ifgl->ifgl_group, M_TEMP);
1407 	}
1408 	free(ifgm, M_TEMP);
1409 	free(ifgl, M_TEMP);
1410 }
1411 
1412 /*
1413  * Remove a group from an interface
1414  */
1415 int
if_delgroup(struct ifnet * ifp,const char * groupname)1416 if_delgroup(struct ifnet *ifp, const char *groupname)
1417 {
1418 	struct ifg_list *ifgl;
1419 
1420 	IFNET_WLOCK();
1421 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1422 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1423 			break;
1424 	if (ifgl == NULL) {
1425 		IFNET_WUNLOCK();
1426 		return (ENOENT);
1427 	}
1428 
1429 	_if_delgroup_locked(ifp, ifgl, groupname);
1430 
1431 	return (0);
1432 }
1433 
1434 /*
1435  * Remove an interface from all groups
1436  */
1437 static void
if_delgroups(struct ifnet * ifp)1438 if_delgroups(struct ifnet *ifp)
1439 {
1440 	struct ifg_list *ifgl;
1441 	char groupname[IFNAMSIZ];
1442 
1443 	IFNET_WLOCK();
1444 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1445 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1446 		_if_delgroup_locked(ifp, ifgl, groupname);
1447 		IFNET_WLOCK();
1448 	}
1449 	IFNET_WUNLOCK();
1450 }
1451 
1452 /*
1453  * Stores all groups from an interface in memory pointed to by ifgr.
1454  */
1455 static int
if_getgroup(struct ifgroupreq * ifgr,struct ifnet * ifp)1456 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1457 {
1458 	int			 len, error;
1459 	struct ifg_list		*ifgl;
1460 	struct ifg_req		 ifgrq, *ifgp;
1461 
1462 	NET_EPOCH_ASSERT();
1463 
1464 	if (ifgr->ifgr_len == 0) {
1465 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1466 			ifgr->ifgr_len += sizeof(struct ifg_req);
1467 		return (0);
1468 	}
1469 
1470 	len = ifgr->ifgr_len;
1471 	ifgp = ifgr->ifgr_groups;
1472 	/* XXX: wire */
1473 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1474 		if (len < sizeof(ifgrq))
1475 			return (EINVAL);
1476 		bzero(&ifgrq, sizeof ifgrq);
1477 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1478 		    sizeof(ifgrq.ifgrq_group));
1479 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1480 			return (error);
1481 		len -= sizeof(ifgrq);
1482 		ifgp++;
1483 	}
1484 
1485 	return (0);
1486 }
1487 
1488 /*
1489  * Stores all members of a group in memory pointed to by igfr
1490  */
1491 static int
if_getgroupmembers(struct ifgroupreq * ifgr)1492 if_getgroupmembers(struct ifgroupreq *ifgr)
1493 {
1494 	struct ifg_group	*ifg;
1495 	struct ifg_member	*ifgm;
1496 	struct ifg_req		 ifgrq, *ifgp;
1497 	int			 len, error;
1498 
1499 	IFNET_RLOCK();
1500 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1501 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1502 			break;
1503 	if (ifg == NULL) {
1504 		IFNET_RUNLOCK();
1505 		return (ENOENT);
1506 	}
1507 
1508 	if (ifgr->ifgr_len == 0) {
1509 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1510 			ifgr->ifgr_len += sizeof(ifgrq);
1511 		IFNET_RUNLOCK();
1512 		return (0);
1513 	}
1514 
1515 	len = ifgr->ifgr_len;
1516 	ifgp = ifgr->ifgr_groups;
1517 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1518 		if (len < sizeof(ifgrq)) {
1519 			IFNET_RUNLOCK();
1520 			return (EINVAL);
1521 		}
1522 		bzero(&ifgrq, sizeof ifgrq);
1523 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1524 		    sizeof(ifgrq.ifgrq_member));
1525 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1526 			IFNET_RUNLOCK();
1527 			return (error);
1528 		}
1529 		len -= sizeof(ifgrq);
1530 		ifgp++;
1531 	}
1532 	IFNET_RUNLOCK();
1533 
1534 	return (0);
1535 }
1536 
1537 /*
1538  * Return counter values from counter(9)s stored in ifnet.
1539  */
1540 uint64_t
if_get_counter_default(struct ifnet * ifp,ift_counter cnt)1541 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1542 {
1543 
1544 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1545 
1546 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1547 }
1548 
1549 /*
1550  * Increase an ifnet counter. Usually used for counters shared
1551  * between the stack and a driver, but function supports them all.
1552  */
1553 void
if_inc_counter(struct ifnet * ifp,ift_counter cnt,int64_t inc)1554 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1555 {
1556 
1557 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1558 
1559 	counter_u64_add(ifp->if_counters[cnt], inc);
1560 }
1561 
1562 /*
1563  * Copy data from ifnet to userland API structure if_data.
1564  */
1565 void
if_data_copy(struct ifnet * ifp,struct if_data * ifd)1566 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1567 {
1568 
1569 	ifd->ifi_type = ifp->if_type;
1570 	ifd->ifi_physical = 0;
1571 	ifd->ifi_addrlen = ifp->if_addrlen;
1572 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1573 	ifd->ifi_link_state = ifp->if_link_state;
1574 	ifd->ifi_vhid = 0;
1575 	ifd->ifi_datalen = sizeof(struct if_data);
1576 	ifd->ifi_mtu = ifp->if_mtu;
1577 	ifd->ifi_metric = ifp->if_metric;
1578 	ifd->ifi_baudrate = ifp->if_baudrate;
1579 	ifd->ifi_hwassist = ifp->if_hwassist;
1580 	ifd->ifi_epoch = ifp->if_epoch;
1581 	ifd->ifi_lastchange = ifp->if_lastchange;
1582 
1583 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1584 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1585 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1586 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1587 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1588 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1589 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1590 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1591 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1592 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1593 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1594 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1595 }
1596 
1597 /*
1598  * Initialization, destruction and refcounting functions for ifaddrs.
1599  */
1600 struct ifaddr *
ifa_alloc(size_t size,int flags)1601 ifa_alloc(size_t size, int flags)
1602 {
1603 	struct ifaddr *ifa;
1604 
1605 	KASSERT(size >= sizeof(struct ifaddr),
1606 	    ("%s: invalid size %zu", __func__, size));
1607 
1608 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1609 	if (ifa == NULL)
1610 		return (NULL);
1611 
1612 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1613 		goto fail;
1614 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1615 		goto fail;
1616 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1617 		goto fail;
1618 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1619 		goto fail;
1620 
1621 	refcount_init(&ifa->ifa_refcnt, 1);
1622 
1623 	return (ifa);
1624 
1625 fail:
1626 	/* free(NULL) is okay */
1627 	counter_u64_free(ifa->ifa_opackets);
1628 	counter_u64_free(ifa->ifa_ipackets);
1629 	counter_u64_free(ifa->ifa_obytes);
1630 	counter_u64_free(ifa->ifa_ibytes);
1631 	free(ifa, M_IFADDR);
1632 
1633 	return (NULL);
1634 }
1635 
1636 void
ifa_ref(struct ifaddr * ifa)1637 ifa_ref(struct ifaddr *ifa)
1638 {
1639 	u_int old __diagused;
1640 
1641 	old = refcount_acquire(&ifa->ifa_refcnt);
1642 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1643 }
1644 
1645 int
ifa_try_ref(struct ifaddr * ifa)1646 ifa_try_ref(struct ifaddr *ifa)
1647 {
1648 
1649 	NET_EPOCH_ASSERT();
1650 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1651 }
1652 
1653 static void
ifa_destroy(epoch_context_t ctx)1654 ifa_destroy(epoch_context_t ctx)
1655 {
1656 	struct ifaddr *ifa;
1657 
1658 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1659 	counter_u64_free(ifa->ifa_opackets);
1660 	counter_u64_free(ifa->ifa_ipackets);
1661 	counter_u64_free(ifa->ifa_obytes);
1662 	counter_u64_free(ifa->ifa_ibytes);
1663 	free(ifa, M_IFADDR);
1664 }
1665 
1666 void
ifa_free(struct ifaddr * ifa)1667 ifa_free(struct ifaddr *ifa)
1668 {
1669 
1670 	if (refcount_release(&ifa->ifa_refcnt))
1671 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1672 }
1673 
1674 /*
1675  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1676  * structs used to represent other address families, it is necessary
1677  * to perform a different comparison.
1678  */
1679 static bool
sa_dl_equal(const struct sockaddr * a,const struct sockaddr * b)1680 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b)
1681 {
1682 	const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a;
1683 	const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b;
1684 
1685 	return (sdl1->sdl_len == sdl2->sdl_len &&
1686 	    bcmp(sdl1->sdl_data + sdl1->sdl_nlen,
1687 	    sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0);
1688 }
1689 
1690 /*
1691  * Locate an interface based on a complete address.
1692  */
1693 /*ARGSUSED*/
1694 struct ifaddr *
ifa_ifwithaddr(const struct sockaddr * addr)1695 ifa_ifwithaddr(const struct sockaddr *addr)
1696 {
1697 	struct ifnet *ifp;
1698 	struct ifaddr *ifa;
1699 
1700 	NET_EPOCH_ASSERT();
1701 
1702 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1703 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1704 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1705 				continue;
1706 			if (sa_equal(addr, ifa->ifa_addr)) {
1707 				goto done;
1708 			}
1709 			/* IP6 doesn't have broadcast */
1710 			if ((ifp->if_flags & IFF_BROADCAST) &&
1711 			    ifa->ifa_broadaddr &&
1712 			    ifa->ifa_broadaddr->sa_len != 0 &&
1713 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1714 				goto done;
1715 			}
1716 		}
1717 	}
1718 	ifa = NULL;
1719 done:
1720 	return (ifa);
1721 }
1722 
1723 int
ifa_ifwithaddr_check(const struct sockaddr * addr)1724 ifa_ifwithaddr_check(const struct sockaddr *addr)
1725 {
1726 	struct epoch_tracker et;
1727 	int rc;
1728 
1729 	NET_EPOCH_ENTER(et);
1730 	rc = (ifa_ifwithaddr(addr) != NULL);
1731 	NET_EPOCH_EXIT(et);
1732 	return (rc);
1733 }
1734 
1735 /*
1736  * Locate an interface based on the broadcast address.
1737  */
1738 /* ARGSUSED */
1739 struct ifaddr *
ifa_ifwithbroadaddr(const struct sockaddr * addr,int fibnum)1740 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1741 {
1742 	struct ifnet *ifp;
1743 	struct ifaddr *ifa;
1744 
1745 	NET_EPOCH_ASSERT();
1746 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1747 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1748 			continue;
1749 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1750 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1751 				continue;
1752 			if ((ifp->if_flags & IFF_BROADCAST) &&
1753 			    ifa->ifa_broadaddr &&
1754 			    ifa->ifa_broadaddr->sa_len != 0 &&
1755 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1756 				goto done;
1757 			}
1758 		}
1759 	}
1760 	ifa = NULL;
1761 done:
1762 	return (ifa);
1763 }
1764 
1765 /*
1766  * Locate the point to point interface with a given destination address.
1767  */
1768 /*ARGSUSED*/
1769 struct ifaddr *
ifa_ifwithdstaddr(const struct sockaddr * addr,int fibnum)1770 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1771 {
1772 	struct ifnet *ifp;
1773 	struct ifaddr *ifa;
1774 
1775 	NET_EPOCH_ASSERT();
1776 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1777 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1778 			continue;
1779 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1780 			continue;
1781 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1782 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1783 				continue;
1784 			if (ifa->ifa_dstaddr != NULL &&
1785 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1786 				goto done;
1787 			}
1788 		}
1789 	}
1790 	ifa = NULL;
1791 done:
1792 	return (ifa);
1793 }
1794 
1795 /*
1796  * Find an interface on a specific network.  If many, choice
1797  * is most specific found.
1798  */
1799 struct ifaddr *
ifa_ifwithnet(const struct sockaddr * addr,int ignore_ptp,int fibnum)1800 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1801 {
1802 	struct ifnet *ifp;
1803 	struct ifaddr *ifa;
1804 	struct ifaddr *ifa_maybe = NULL;
1805 	u_int af = addr->sa_family;
1806 	const char *addr_data = addr->sa_data, *cplim;
1807 
1808 	NET_EPOCH_ASSERT();
1809 	/*
1810 	 * AF_LINK addresses can be looked up directly by their index number,
1811 	 * so do that if we can.
1812 	 */
1813 	if (af == AF_LINK) {
1814 		ifp = ifnet_byindex(
1815 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1816 		return (ifp ? ifp->if_addr : NULL);
1817 	}
1818 
1819 	/*
1820 	 * Scan though each interface, looking for ones that have addresses
1821 	 * in this address family and the requested fib.
1822 	 */
1823 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1824 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1825 			continue;
1826 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1827 			const char *cp, *cp2, *cp3;
1828 
1829 			if (ifa->ifa_addr->sa_family != af)
1830 next:				continue;
1831 			if (af == AF_INET &&
1832 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1833 				/*
1834 				 * This is a bit broken as it doesn't
1835 				 * take into account that the remote end may
1836 				 * be a single node in the network we are
1837 				 * looking for.
1838 				 * The trouble is that we don't know the
1839 				 * netmask for the remote end.
1840 				 */
1841 				if (ifa->ifa_dstaddr != NULL &&
1842 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1843 					goto done;
1844 				}
1845 			} else {
1846 				/*
1847 				 * Scan all the bits in the ifa's address.
1848 				 * If a bit dissagrees with what we are
1849 				 * looking for, mask it with the netmask
1850 				 * to see if it really matters.
1851 				 * (A byte at a time)
1852 				 */
1853 				if (ifa->ifa_netmask == 0)
1854 					continue;
1855 				cp = addr_data;
1856 				cp2 = ifa->ifa_addr->sa_data;
1857 				cp3 = ifa->ifa_netmask->sa_data;
1858 				cplim = ifa->ifa_netmask->sa_len
1859 					+ (char *)ifa->ifa_netmask;
1860 				while (cp3 < cplim)
1861 					if ((*cp++ ^ *cp2++) & *cp3++)
1862 						goto next; /* next address! */
1863 				/*
1864 				 * If the netmask of what we just found
1865 				 * is more specific than what we had before
1866 				 * (if we had one), or if the virtual status
1867 				 * of new prefix is better than of the old one,
1868 				 * then remember the new one before continuing
1869 				 * to search for an even better one.
1870 				 */
1871 				if (ifa_maybe == NULL ||
1872 				    ifa_preferred(ifa_maybe, ifa) ||
1873 				    rn_refines((caddr_t)ifa->ifa_netmask,
1874 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1875 					ifa_maybe = ifa;
1876 				}
1877 			}
1878 		}
1879 	}
1880 	ifa = ifa_maybe;
1881 	ifa_maybe = NULL;
1882 done:
1883 	return (ifa);
1884 }
1885 
1886 /*
1887  * Find an interface address specific to an interface best matching
1888  * a given address.
1889  */
1890 struct ifaddr *
ifaof_ifpforaddr(const struct sockaddr * addr,struct ifnet * ifp)1891 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1892 {
1893 	struct ifaddr *ifa;
1894 	const char *cp, *cp2, *cp3;
1895 	char *cplim;
1896 	struct ifaddr *ifa_maybe = NULL;
1897 	u_int af = addr->sa_family;
1898 
1899 	if (af >= AF_MAX)
1900 		return (NULL);
1901 
1902 	NET_EPOCH_ASSERT();
1903 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1904 		if (ifa->ifa_addr->sa_family != af)
1905 			continue;
1906 		if (ifa_maybe == NULL)
1907 			ifa_maybe = ifa;
1908 		if (ifa->ifa_netmask == 0) {
1909 			if (sa_equal(addr, ifa->ifa_addr) ||
1910 			    (ifa->ifa_dstaddr &&
1911 			    sa_equal(addr, ifa->ifa_dstaddr)))
1912 				goto done;
1913 			continue;
1914 		}
1915 		if (ifp->if_flags & IFF_POINTOPOINT) {
1916 			if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))
1917 				goto done;
1918 		} else {
1919 			cp = addr->sa_data;
1920 			cp2 = ifa->ifa_addr->sa_data;
1921 			cp3 = ifa->ifa_netmask->sa_data;
1922 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1923 			for (; cp3 < cplim; cp3++)
1924 				if ((*cp++ ^ *cp2++) & *cp3)
1925 					break;
1926 			if (cp3 == cplim)
1927 				goto done;
1928 		}
1929 	}
1930 	ifa = ifa_maybe;
1931 done:
1932 	return (ifa);
1933 }
1934 
1935 /*
1936  * See whether new ifa is better than current one:
1937  * 1) A non-virtual one is preferred over virtual.
1938  * 2) A virtual in master state preferred over any other state.
1939  *
1940  * Used in several address selecting functions.
1941  */
1942 int
ifa_preferred(struct ifaddr * cur,struct ifaddr * next)1943 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1944 {
1945 
1946 	return (cur->ifa_carp && (!next->ifa_carp ||
1947 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1948 }
1949 
1950 struct sockaddr_dl *
link_alloc_sdl(size_t size,int flags)1951 link_alloc_sdl(size_t size, int flags)
1952 {
1953 
1954 	return (malloc(size, M_TEMP, flags));
1955 }
1956 
1957 void
link_free_sdl(struct sockaddr * sa)1958 link_free_sdl(struct sockaddr *sa)
1959 {
1960 	free(sa, M_TEMP);
1961 }
1962 
1963 /*
1964  * Fills in given sdl with interface basic info.
1965  * Returns pointer to filled sdl.
1966  */
1967 struct sockaddr_dl *
link_init_sdl(struct ifnet * ifp,struct sockaddr * paddr,u_char iftype)1968 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1969 {
1970 	struct sockaddr_dl *sdl;
1971 
1972 	sdl = (struct sockaddr_dl *)paddr;
1973 	memset(sdl, 0, sizeof(struct sockaddr_dl));
1974 	sdl->sdl_len = sizeof(struct sockaddr_dl);
1975 	sdl->sdl_family = AF_LINK;
1976 	sdl->sdl_index = ifp->if_index;
1977 	sdl->sdl_type = iftype;
1978 
1979 	return (sdl);
1980 }
1981 
1982 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1983 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1984 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
1985 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
1986 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
1987 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
1988 int	(*vlan_setcookie_p)(struct ifnet *, void *);
1989 void	*(*vlan_cookie_p)(struct ifnet *);
1990 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
1991 
1992 /*
1993  * Handle a change in the interface link state. To avoid LORs
1994  * between driver lock and upper layer locks, as well as possible
1995  * recursions, we post event to taskqueue, and all job
1996  * is done in static do_link_state_change().
1997  */
1998 void
if_link_state_change(struct ifnet * ifp,int link_state)1999 if_link_state_change(struct ifnet *ifp, int link_state)
2000 {
2001 	/* Return if state hasn't changed. */
2002 	if (ifp->if_link_state == link_state)
2003 		return;
2004 
2005 	ifp->if_link_state = link_state;
2006 
2007 	/* XXXGL: reference ifp? */
2008 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2009 }
2010 
2011 static void
do_link_state_change(void * arg,int pending)2012 do_link_state_change(void *arg, int pending)
2013 {
2014 	struct ifnet *ifp;
2015 	int link_state;
2016 
2017 	ifp = arg;
2018 	link_state = ifp->if_link_state;
2019 
2020 	CURVNET_SET(ifp->if_vnet);
2021 	rt_ifmsg(ifp, 0);
2022 	if (ifp->if_vlantrunk != NULL)
2023 		(*vlan_link_state_p)(ifp);
2024 	if (ifp->if_carp)
2025 		(*carp_linkstate_p)(ifp);
2026 	if (ifp->if_bridge)
2027 		ifp->if_bridge_linkstate(ifp);
2028 	if (ifp->if_lagg)
2029 		(*lagg_linkstate_p)(ifp, link_state);
2030 
2031 	if (IS_DEFAULT_VNET(curvnet))
2032 		devctl_notify("IFNET", ifp->if_xname,
2033 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2034 		    NULL);
2035 	if (pending > 1)
2036 		if_printf(ifp, "%d link states coalesced\n", pending);
2037 	if (log_link_state_change)
2038 		if_printf(ifp, "link state changed to %s\n",
2039 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2040 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2041 	CURVNET_RESTORE();
2042 }
2043 
2044 /*
2045  * Mark an interface down and notify protocols of
2046  * the transition.
2047  */
2048 void
if_down(struct ifnet * ifp)2049 if_down(struct ifnet *ifp)
2050 {
2051 
2052 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2053 
2054 	ifp->if_flags &= ~IFF_UP;
2055 	getmicrotime(&ifp->if_lastchange);
2056 	ifp->if_qflush(ifp);
2057 
2058 	if (ifp->if_carp)
2059 		(*carp_linkstate_p)(ifp);
2060 	rt_ifmsg(ifp, IFF_UP);
2061 }
2062 
2063 /*
2064  * Mark an interface up and notify protocols of
2065  * the transition.
2066  */
2067 void
if_up(struct ifnet * ifp)2068 if_up(struct ifnet *ifp)
2069 {
2070 
2071 	ifp->if_flags |= IFF_UP;
2072 	getmicrotime(&ifp->if_lastchange);
2073 	if (ifp->if_carp)
2074 		(*carp_linkstate_p)(ifp);
2075 	rt_ifmsg(ifp, IFF_UP);
2076 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2077 }
2078 
2079 /*
2080  * Flush an interface queue.
2081  */
2082 void
if_qflush(struct ifnet * ifp)2083 if_qflush(struct ifnet *ifp)
2084 {
2085 	struct mbuf *m, *n;
2086 	struct ifaltq *ifq;
2087 
2088 	ifq = &ifp->if_snd;
2089 	IFQ_LOCK(ifq);
2090 #ifdef ALTQ
2091 	if (ALTQ_IS_ENABLED(ifq))
2092 		ALTQ_PURGE(ifq);
2093 #endif
2094 	n = ifq->ifq_head;
2095 	while ((m = n) != NULL) {
2096 		n = m->m_nextpkt;
2097 		m_freem(m);
2098 	}
2099 	ifq->ifq_head = 0;
2100 	ifq->ifq_tail = 0;
2101 	ifq->ifq_len = 0;
2102 	IFQ_UNLOCK(ifq);
2103 }
2104 
2105 /*
2106  * Map interface name to interface structure pointer, with or without
2107  * returning a reference.
2108  */
2109 struct ifnet *
ifunit_ref(const char * name)2110 ifunit_ref(const char *name)
2111 {
2112 	struct epoch_tracker et;
2113 	struct ifnet *ifp;
2114 
2115 	NET_EPOCH_ENTER(et);
2116 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2117 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2118 		    !(ifp->if_flags & IFF_DYING)) {
2119 			MPASS(ifp->if_vnet == curvnet);
2120 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2121 			if_ref(ifp);
2122 			break;
2123 		}
2124 	}
2125 	NET_EPOCH_EXIT(et);
2126 	return (ifp);
2127 }
2128 
2129 struct ifnet *
ifunit(const char * name)2130 ifunit(const char *name)
2131 {
2132 	struct epoch_tracker et;
2133 	struct ifnet *ifp;
2134 
2135 	NET_EPOCH_ENTER(et);
2136 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2137 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) {
2138 			MPASS(refcount_load(&ifp->if_refcount) > 0);
2139 			MPASS(ifp->if_vnet == curvnet);
2140 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2141 			break;
2142 		}
2143 	}
2144 	NET_EPOCH_EXIT(et);
2145 	return (ifp);
2146 }
2147 
2148 void *
ifr_buffer_get_buffer(void * data)2149 ifr_buffer_get_buffer(void *data)
2150 {
2151 	union ifreq_union *ifrup;
2152 
2153 	ifrup = data;
2154 #ifdef COMPAT_FREEBSD32
2155 	if (SV_CURPROC_FLAG(SV_ILP32))
2156 		return ((void *)(uintptr_t)
2157 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2158 #endif
2159 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2160 }
2161 
2162 static void
ifr_buffer_set_buffer_null(void * data)2163 ifr_buffer_set_buffer_null(void *data)
2164 {
2165 	union ifreq_union *ifrup;
2166 
2167 	ifrup = data;
2168 #ifdef COMPAT_FREEBSD32
2169 	if (SV_CURPROC_FLAG(SV_ILP32))
2170 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2171 	else
2172 #endif
2173 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2174 }
2175 
2176 size_t
ifr_buffer_get_length(void * data)2177 ifr_buffer_get_length(void *data)
2178 {
2179 	union ifreq_union *ifrup;
2180 
2181 	ifrup = data;
2182 #ifdef COMPAT_FREEBSD32
2183 	if (SV_CURPROC_FLAG(SV_ILP32))
2184 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2185 #endif
2186 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2187 }
2188 
2189 static void
ifr_buffer_set_length(void * data,size_t len)2190 ifr_buffer_set_length(void *data, size_t len)
2191 {
2192 	union ifreq_union *ifrup;
2193 
2194 	ifrup = data;
2195 #ifdef COMPAT_FREEBSD32
2196 	if (SV_CURPROC_FLAG(SV_ILP32))
2197 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2198 	else
2199 #endif
2200 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2201 }
2202 
2203 void *
ifr_data_get_ptr(void * ifrp)2204 ifr_data_get_ptr(void *ifrp)
2205 {
2206 	union ifreq_union *ifrup;
2207 
2208 	ifrup = ifrp;
2209 #ifdef COMPAT_FREEBSD32
2210 	if (SV_CURPROC_FLAG(SV_ILP32))
2211 		return ((void *)(uintptr_t)
2212 		    ifrup->ifr32.ifr_ifru.ifru_data);
2213 #endif
2214 		return (ifrup->ifr.ifr_ifru.ifru_data);
2215 }
2216 
2217 struct ifcap_nv_bit_name {
2218 	uint64_t cap_bit;
2219 	const char *cap_name;
2220 };
2221 #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2222     .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2223 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2224 	CAPNV(RXCSUM),
2225 	CAPNV(TXCSUM),
2226 	CAPNV(NETCONS),
2227 	CAPNV(VLAN_MTU),
2228 	CAPNV(VLAN_HWTAGGING),
2229 	CAPNV(JUMBO_MTU),
2230 	CAPNV(POLLING),
2231 	CAPNV(VLAN_HWCSUM),
2232 	CAPNV(TSO4),
2233 	CAPNV(TSO6),
2234 	CAPNV(LRO),
2235 	CAPNV(WOL_UCAST),
2236 	CAPNV(WOL_MCAST),
2237 	CAPNV(WOL_MAGIC),
2238 	CAPNV(TOE4),
2239 	CAPNV(TOE6),
2240 	CAPNV(VLAN_HWFILTER),
2241 	CAPNV(VLAN_HWTSO),
2242 	CAPNV(LINKSTATE),
2243 	CAPNV(NETMAP),
2244 	CAPNV(RXCSUM_IPV6),
2245 	CAPNV(TXCSUM_IPV6),
2246 	CAPNV(HWSTATS),
2247 	CAPNV(TXRTLMT),
2248 	CAPNV(HWRXTSTMP),
2249 	CAPNV(MEXTPG),
2250 	CAPNV(TXTLS4),
2251 	CAPNV(TXTLS6),
2252 	CAPNV(VXLAN_HWCSUM),
2253 	CAPNV(VXLAN_HWTSO),
2254 	CAPNV(TXTLS_RTLMT),
2255 	{0, NULL}
2256 };
2257 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \
2258     .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2259 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2260 	CAP2NV(RXTLS4),
2261 	CAP2NV(RXTLS6),
2262 	CAP2NV(IPSEC_OFFLOAD),
2263 	{0, NULL}
2264 };
2265 #undef CAPNV
2266 #undef CAP2NV
2267 
2268 int
if_capnv_to_capint(const nvlist_t * nv,int * old_cap,const struct ifcap_nv_bit_name * nn,bool all)2269 if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2270     const struct ifcap_nv_bit_name *nn, bool all)
2271 {
2272 	int i, res;
2273 
2274 	res = 0;
2275 	for (i = 0; nn[i].cap_name != NULL; i++) {
2276 		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2277 			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2278 				res |= nn[i].cap_bit;
2279 		} else {
2280 			res |= *old_cap & nn[i].cap_bit;
2281 		}
2282 	}
2283 	return (res);
2284 }
2285 
2286 void
if_capint_to_capnv(nvlist_t * nv,const struct ifcap_nv_bit_name * nn,int ifr_cap,int ifr_req)2287 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2288     int ifr_cap, int ifr_req)
2289 {
2290 	int i;
2291 
2292 	for (i = 0; nn[i].cap_name != NULL; i++) {
2293 		if ((nn[i].cap_bit & ifr_cap) != 0) {
2294 			nvlist_add_bool(nv, nn[i].cap_name,
2295 			    (nn[i].cap_bit & ifr_req) != 0);
2296 		}
2297 	}
2298 }
2299 
2300 /*
2301  * Hardware specific interface ioctls.
2302  */
2303 int
ifhwioctl(u_long cmd,struct ifnet * ifp,caddr_t data,struct thread * td)2304 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2305 {
2306 	struct ifreq *ifr;
2307 	int error = 0, do_ifup = 0;
2308 	int new_flags, temp_flags;
2309 	size_t descrlen, nvbuflen;
2310 	char *descrbuf;
2311 	char new_name[IFNAMSIZ];
2312 	void *buf;
2313 	nvlist_t *nvcap;
2314 	struct siocsifcapnv_driver_data drv_ioctl_data;
2315 
2316 	ifr = (struct ifreq *)data;
2317 	switch (cmd) {
2318 	case SIOCGIFINDEX:
2319 		ifr->ifr_index = ifp->if_index;
2320 		break;
2321 
2322 	case SIOCGIFFLAGS:
2323 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2324 		ifr->ifr_flags = temp_flags & 0xffff;
2325 		ifr->ifr_flagshigh = temp_flags >> 16;
2326 		break;
2327 
2328 	case SIOCGIFCAP:
2329 		ifr->ifr_reqcap = ifp->if_capabilities;
2330 		ifr->ifr_curcap = ifp->if_capenable;
2331 		break;
2332 
2333 	case SIOCGIFCAPNV:
2334 		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2335 			error = EINVAL;
2336 			break;
2337 		}
2338 		buf = NULL;
2339 		nvcap = nvlist_create(0);
2340 		for (;;) {
2341 			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2342 			    ifp->if_capabilities, ifp->if_capenable);
2343 			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2344 			    ifp->if_capabilities2, ifp->if_capenable2);
2345 			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2346 			    __DECONST(caddr_t, nvcap));
2347 			if (error != 0) {
2348 				if_printf(ifp,
2349 			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2350 				    error);
2351 				break;
2352 			}
2353 			buf = nvlist_pack(nvcap, &nvbuflen);
2354 			if (buf == NULL) {
2355 				error = nvlist_error(nvcap);
2356 				if (error == 0)
2357 					error = EDOOFUS;
2358 				break;
2359 			}
2360 			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2361 				ifr->ifr_cap_nv.length = nvbuflen;
2362 				ifr->ifr_cap_nv.buffer = NULL;
2363 				error = EFBIG;
2364 				break;
2365 			}
2366 			ifr->ifr_cap_nv.length = nvbuflen;
2367 			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2368 			break;
2369 		}
2370 		free(buf, M_NVLIST);
2371 		nvlist_destroy(nvcap);
2372 		break;
2373 
2374 	case SIOCGIFDATA:
2375 	{
2376 		struct if_data ifd;
2377 
2378 		/* Ensure uninitialised padding is not leaked. */
2379 		memset(&ifd, 0, sizeof(ifd));
2380 
2381 		if_data_copy(ifp, &ifd);
2382 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2383 		break;
2384 	}
2385 
2386 #ifdef MAC
2387 	case SIOCGIFMAC:
2388 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2389 		break;
2390 #endif
2391 
2392 	case SIOCGIFMETRIC:
2393 		ifr->ifr_metric = ifp->if_metric;
2394 		break;
2395 
2396 	case SIOCGIFMTU:
2397 		ifr->ifr_mtu = ifp->if_mtu;
2398 		break;
2399 
2400 	case SIOCGIFPHYS:
2401 		/* XXXGL: did this ever worked? */
2402 		ifr->ifr_phys = 0;
2403 		break;
2404 
2405 	case SIOCGIFDESCR:
2406 		error = 0;
2407 		sx_slock(&ifdescr_sx);
2408 		if (ifp->if_description == NULL)
2409 			error = ENOMSG;
2410 		else {
2411 			/* space for terminating nul */
2412 			descrlen = strlen(ifp->if_description) + 1;
2413 			if (ifr_buffer_get_length(ifr) < descrlen)
2414 				ifr_buffer_set_buffer_null(ifr);
2415 			else
2416 				error = copyout(ifp->if_description,
2417 				    ifr_buffer_get_buffer(ifr), descrlen);
2418 			ifr_buffer_set_length(ifr, descrlen);
2419 		}
2420 		sx_sunlock(&ifdescr_sx);
2421 		break;
2422 
2423 	case SIOCSIFDESCR:
2424 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2425 		if (error)
2426 			return (error);
2427 
2428 		/*
2429 		 * Copy only (length-1) bytes to make sure that
2430 		 * if_description is always nul terminated.  The
2431 		 * length parameter is supposed to count the
2432 		 * terminating nul in.
2433 		 */
2434 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2435 			return (ENAMETOOLONG);
2436 		else if (ifr_buffer_get_length(ifr) == 0)
2437 			descrbuf = NULL;
2438 		else {
2439 			descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK);
2440 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2441 			    ifr_buffer_get_length(ifr) - 1);
2442 			if (error) {
2443 				if_freedescr(descrbuf);
2444 				break;
2445 			}
2446 		}
2447 
2448 		if_setdescr(ifp, descrbuf);
2449 		getmicrotime(&ifp->if_lastchange);
2450 		break;
2451 
2452 	case SIOCGIFFIB:
2453 		ifr->ifr_fib = ifp->if_fib;
2454 		break;
2455 
2456 	case SIOCSIFFIB:
2457 		error = priv_check(td, PRIV_NET_SETIFFIB);
2458 		if (error)
2459 			return (error);
2460 		if (ifr->ifr_fib >= rt_numfibs)
2461 			return (EINVAL);
2462 
2463 		ifp->if_fib = ifr->ifr_fib;
2464 		break;
2465 
2466 	case SIOCSIFFLAGS:
2467 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2468 		if (error)
2469 			return (error);
2470 		/*
2471 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2472 		 * check, so we don't need special handling here yet.
2473 		 */
2474 		new_flags = (ifr->ifr_flags & 0xffff) |
2475 		    (ifr->ifr_flagshigh << 16);
2476 		if (ifp->if_flags & IFF_UP &&
2477 		    (new_flags & IFF_UP) == 0) {
2478 			if_down(ifp);
2479 		} else if (new_flags & IFF_UP &&
2480 		    (ifp->if_flags & IFF_UP) == 0) {
2481 			do_ifup = 1;
2482 		}
2483 
2484 		/*
2485 		 * See if the promiscuous mode or allmulti bits are about to
2486 		 * flip.  They require special handling because in-kernel
2487 		 * consumers may indepdently toggle them.
2488 		 */
2489 		if_setppromisc(ifp, new_flags & IFF_PPROMISC);
2490 		if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) {
2491 			if (new_flags & IFF_PALLMULTI)
2492 				ifp->if_flags |= IFF_ALLMULTI;
2493 			else if (ifp->if_amcount == 0)
2494 				ifp->if_flags &= ~IFF_ALLMULTI;
2495 		}
2496 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2497 			(new_flags &~ IFF_CANTCHANGE);
2498 		if (ifp->if_ioctl) {
2499 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2500 		}
2501 		if (do_ifup)
2502 			if_up(ifp);
2503 		getmicrotime(&ifp->if_lastchange);
2504 		break;
2505 
2506 	case SIOCSIFCAP:
2507 		error = priv_check(td, PRIV_NET_SETIFCAP);
2508 		if (error != 0)
2509 			return (error);
2510 		if (ifp->if_ioctl == NULL)
2511 			return (EOPNOTSUPP);
2512 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2513 			return (EINVAL);
2514 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2515 		if (error == 0)
2516 			getmicrotime(&ifp->if_lastchange);
2517 		break;
2518 
2519 	case SIOCSIFCAPNV:
2520 		error = priv_check(td, PRIV_NET_SETIFCAP);
2521 		if (error != 0)
2522 			return (error);
2523 		if (ifp->if_ioctl == NULL)
2524 			return (EOPNOTSUPP);
2525 		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2526 			return (EINVAL);
2527 		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2528 			return (EINVAL);
2529 		nvcap = NULL;
2530 		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2531 		for (;;) {
2532 			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2533 			    ifr->ifr_cap_nv.length);
2534 			if (error != 0)
2535 				break;
2536 			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2537 			if (nvcap == NULL) {
2538 				error = EINVAL;
2539 				break;
2540 			}
2541 			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2542 			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2543 			if ((drv_ioctl_data.reqcap &
2544 			    ~ifp->if_capabilities) != 0) {
2545 				error = EINVAL;
2546 				break;
2547 			}
2548 			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2549 			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2550 			if ((drv_ioctl_data.reqcap2 &
2551 			    ~ifp->if_capabilities2) != 0) {
2552 				error = EINVAL;
2553 				break;
2554 			}
2555 			drv_ioctl_data.nvcap = nvcap;
2556 			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2557 			    (caddr_t)&drv_ioctl_data);
2558 			break;
2559 		}
2560 		nvlist_destroy(nvcap);
2561 		free(buf, M_TEMP);
2562 		if (error == 0)
2563 			getmicrotime(&ifp->if_lastchange);
2564 		break;
2565 
2566 #ifdef MAC
2567 	case SIOCSIFMAC:
2568 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2569 		break;
2570 #endif
2571 
2572 	case SIOCSIFNAME:
2573 		error = priv_check(td, PRIV_NET_SETIFNAME);
2574 		if (error)
2575 			return (error);
2576 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2577 		    NULL);
2578 		if (error != 0)
2579 			return (error);
2580 		error = if_rename(ifp, new_name);
2581 		break;
2582 
2583 #ifdef VIMAGE
2584 	case SIOCSIFVNET:
2585 		error = priv_check(td, PRIV_NET_SETIFVNET);
2586 		if (error)
2587 			return (error);
2588 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2589 		break;
2590 #endif
2591 
2592 	case SIOCSIFMETRIC:
2593 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2594 		if (error)
2595 			return (error);
2596 		ifp->if_metric = ifr->ifr_metric;
2597 		getmicrotime(&ifp->if_lastchange);
2598 		break;
2599 
2600 	case SIOCSIFPHYS:
2601 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2602 		if (error)
2603 			return (error);
2604 		if (ifp->if_ioctl == NULL)
2605 			return (EOPNOTSUPP);
2606 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2607 		if (error == 0)
2608 			getmicrotime(&ifp->if_lastchange);
2609 		break;
2610 
2611 	case SIOCSIFMTU:
2612 	{
2613 		u_long oldmtu = ifp->if_mtu;
2614 
2615 		error = priv_check(td, PRIV_NET_SETIFMTU);
2616 		if (error)
2617 			return (error);
2618 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2619 			return (EINVAL);
2620 		if (ifp->if_ioctl == NULL)
2621 			return (EOPNOTSUPP);
2622 		/* Disallow MTU changes on bridge member interfaces. */
2623 		if (ifp->if_bridge)
2624 			return (EOPNOTSUPP);
2625 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2626 		if (error == 0) {
2627 			getmicrotime(&ifp->if_lastchange);
2628 			rt_ifmsg(ifp, 0);
2629 #ifdef INET
2630 			DEBUGNET_NOTIFY_MTU(ifp);
2631 #endif
2632 		}
2633 		/*
2634 		 * If the link MTU changed, do network layer specific procedure.
2635 		 */
2636 		if (ifp->if_mtu != oldmtu)
2637 			if_notifymtu(ifp);
2638 		break;
2639 	}
2640 
2641 	case SIOCADDMULTI:
2642 	case SIOCDELMULTI:
2643 		if (cmd == SIOCADDMULTI)
2644 			error = priv_check(td, PRIV_NET_ADDMULTI);
2645 		else
2646 			error = priv_check(td, PRIV_NET_DELMULTI);
2647 		if (error)
2648 			return (error);
2649 
2650 		/* Don't allow group membership on non-multicast interfaces. */
2651 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2652 			return (EOPNOTSUPP);
2653 
2654 		/* Don't let users screw up protocols' entries. */
2655 		if (ifr->ifr_addr.sa_family != AF_LINK)
2656 			return (EINVAL);
2657 
2658 		if (cmd == SIOCADDMULTI) {
2659 			struct epoch_tracker et;
2660 			struct ifmultiaddr *ifma;
2661 
2662 			/*
2663 			 * Userland is only permitted to join groups once
2664 			 * via the if_addmulti() KPI, because it cannot hold
2665 			 * struct ifmultiaddr * between calls. It may also
2666 			 * lose a race while we check if the membership
2667 			 * already exists.
2668 			 */
2669 			NET_EPOCH_ENTER(et);
2670 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2671 			NET_EPOCH_EXIT(et);
2672 			if (ifma != NULL)
2673 				error = EADDRINUSE;
2674 			else
2675 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2676 		} else {
2677 			error = if_delmulti(ifp, &ifr->ifr_addr);
2678 		}
2679 		if (error == 0)
2680 			getmicrotime(&ifp->if_lastchange);
2681 		break;
2682 
2683 	case SIOCSIFPHYADDR:
2684 	case SIOCDIFPHYADDR:
2685 #ifdef INET6
2686 	case SIOCSIFPHYADDR_IN6:
2687 #endif
2688 	case SIOCSIFMEDIA:
2689 	case SIOCSIFGENERIC:
2690 		error = priv_check(td, PRIV_NET_HWIOCTL);
2691 		if (error)
2692 			return (error);
2693 		if (ifp->if_ioctl == NULL)
2694 			return (EOPNOTSUPP);
2695 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2696 		if (error == 0)
2697 			getmicrotime(&ifp->if_lastchange);
2698 		break;
2699 
2700 	case SIOCGIFSTATUS:
2701 	case SIOCGIFPSRCADDR:
2702 	case SIOCGIFPDSTADDR:
2703 	case SIOCGIFMEDIA:
2704 	case SIOCGIFXMEDIA:
2705 	case SIOCGIFGENERIC:
2706 	case SIOCGIFRSSKEY:
2707 	case SIOCGIFRSSHASH:
2708 	case SIOCGIFDOWNREASON:
2709 		if (ifp->if_ioctl == NULL)
2710 			return (EOPNOTSUPP);
2711 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2712 		break;
2713 
2714 	case SIOCSIFLLADDR:
2715 		error = priv_check(td, PRIV_NET_SETLLADDR);
2716 		if (error)
2717 			return (error);
2718 		error = if_setlladdr(ifp,
2719 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2720 		break;
2721 
2722 	case SIOCGHWADDR:
2723 		error = if_gethwaddr(ifp, ifr);
2724 		break;
2725 
2726 	case SIOCAIFGROUP:
2727 	{
2728 		const char *groupname;
2729 
2730 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2731 		if (error)
2732 			return (error);
2733 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2734 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2735 			return (EINVAL);
2736 		error = if_addgroup(ifp, groupname);
2737 		if (error != 0)
2738 			return (error);
2739 		break;
2740 	}
2741 	case SIOCGIFGROUP:
2742 	{
2743 		struct epoch_tracker et;
2744 
2745 		NET_EPOCH_ENTER(et);
2746 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2747 		NET_EPOCH_EXIT(et);
2748 		break;
2749 	}
2750 
2751 	case SIOCDIFGROUP:
2752 	{
2753 		const char *groupname;
2754 
2755 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2756 		if (error)
2757 			return (error);
2758 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2759 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2760 			return (EINVAL);
2761 		error = if_delgroup(ifp, groupname);
2762 		if (error != 0)
2763 			return (error);
2764 		break;
2765 	}
2766 	default:
2767 		error = ENOIOCTL;
2768 		break;
2769 	}
2770 	return (error);
2771 }
2772 
2773 /*
2774  * Interface ioctls.
2775  */
2776 int
ifioctl(struct socket * so,u_long cmd,caddr_t data,struct thread * td)2777 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2778 {
2779 #ifdef COMPAT_FREEBSD32
2780 	union {
2781 		struct ifconf ifc;
2782 		struct ifdrv ifd;
2783 		struct ifgroupreq ifgr;
2784 		struct ifmediareq ifmr;
2785 	} thunk;
2786 	u_long saved_cmd;
2787 	struct ifconf32 *ifc32;
2788 	struct ifdrv32 *ifd32;
2789 	struct ifgroupreq32 *ifgr32;
2790 	struct ifmediareq32 *ifmr32;
2791 #endif
2792 	struct ifnet *ifp;
2793 	struct ifreq *ifr;
2794 	int error;
2795 	int oif_flags;
2796 #ifdef VIMAGE
2797 	bool shutdown;
2798 #endif
2799 
2800 	CURVNET_SET(so->so_vnet);
2801 #ifdef VIMAGE
2802 	/* Make sure the VNET is stable. */
2803 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2804 	if (shutdown) {
2805 		CURVNET_RESTORE();
2806 		return (EBUSY);
2807 	}
2808 #endif
2809 
2810 #ifdef COMPAT_FREEBSD32
2811 	saved_cmd = cmd;
2812 	switch (cmd) {
2813 	case SIOCGIFCONF32:
2814 		ifc32 = (struct ifconf32 *)data;
2815 		thunk.ifc.ifc_len = ifc32->ifc_len;
2816 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2817 		data = (caddr_t)&thunk.ifc;
2818 		cmd = SIOCGIFCONF;
2819 		break;
2820 	case SIOCGDRVSPEC32:
2821 	case SIOCSDRVSPEC32:
2822 		ifd32 = (struct ifdrv32 *)data;
2823 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2824 		    sizeof(thunk.ifd.ifd_name));
2825 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2826 		thunk.ifd.ifd_len = ifd32->ifd_len;
2827 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2828 		data = (caddr_t)&thunk.ifd;
2829 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2830 		break;
2831 	case SIOCAIFGROUP32:
2832 	case SIOCGIFGROUP32:
2833 	case SIOCDIFGROUP32:
2834 	case SIOCGIFGMEMB32:
2835 		ifgr32 = (struct ifgroupreq32 *)data;
2836 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2837 		    sizeof(thunk.ifgr.ifgr_name));
2838 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2839 		switch (cmd) {
2840 		case SIOCAIFGROUP32:
2841 		case SIOCDIFGROUP32:
2842 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2843 			    sizeof(thunk.ifgr.ifgr_group));
2844 			break;
2845 		case SIOCGIFGROUP32:
2846 		case SIOCGIFGMEMB32:
2847 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2848 			break;
2849 		}
2850 		data = (caddr_t)&thunk.ifgr;
2851 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2852 		break;
2853 	case SIOCGIFMEDIA32:
2854 	case SIOCGIFXMEDIA32:
2855 		ifmr32 = (struct ifmediareq32 *)data;
2856 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2857 		    sizeof(thunk.ifmr.ifm_name));
2858 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2859 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2860 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2861 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2862 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2863 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2864 		data = (caddr_t)&thunk.ifmr;
2865 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2866 		break;
2867 	}
2868 #endif
2869 
2870 	switch (cmd) {
2871 	case SIOCGIFCONF:
2872 		error = ifconf(cmd, data);
2873 		goto out_noref;
2874 	}
2875 
2876 	ifr = (struct ifreq *)data;
2877 	switch (cmd) {
2878 #ifdef VIMAGE
2879 	case SIOCSIFRVNET:
2880 		error = priv_check(td, PRIV_NET_SETIFVNET);
2881 		if (error == 0)
2882 			error = if_vmove_reclaim(td, ifr->ifr_name,
2883 			    ifr->ifr_jid);
2884 		goto out_noref;
2885 #endif
2886 	case SIOCIFCREATE:
2887 	case SIOCIFCREATE2:
2888 		error = priv_check(td, PRIV_NET_IFCREATE);
2889 		if (error == 0)
2890 			error = if_clone_create(ifr->ifr_name,
2891 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
2892 			    ifr_data_get_ptr(ifr) : NULL);
2893 		goto out_noref;
2894 	case SIOCIFDESTROY:
2895 		error = priv_check(td, PRIV_NET_IFDESTROY);
2896 
2897 		if (error == 0) {
2898 			sx_xlock(&ifnet_detach_sxlock);
2899 			error = if_clone_destroy(ifr->ifr_name);
2900 			sx_xunlock(&ifnet_detach_sxlock);
2901 		}
2902 		goto out_noref;
2903 
2904 	case SIOCIFGCLONERS:
2905 		error = if_clone_list((struct if_clonereq *)data);
2906 		goto out_noref;
2907 
2908 	case SIOCGIFGMEMB:
2909 	{
2910 		struct ifgroupreq *req;
2911 
2912 		req = (struct ifgroupreq *)data;
2913 		if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) {
2914 			error = EINVAL;
2915 			goto out_noref;
2916 		}
2917 		error = if_getgroupmembers(req);
2918 		goto out_noref;
2919 	}
2920 	}
2921 
2922 	ifp = ifunit_ref(ifr->ifr_name);
2923 	if (ifp == NULL) {
2924 		error = ENXIO;
2925 		goto out_noref;
2926 	}
2927 
2928 	error = ifhwioctl(cmd, ifp, data, td);
2929 	if (error != ENOIOCTL)
2930 		goto out_ref;
2931 
2932 	oif_flags = ifp->if_flags;
2933 	if (so->so_proto == NULL) {
2934 		error = EOPNOTSUPP;
2935 		goto out_ref;
2936 	}
2937 
2938 	/*
2939 	 * Pass the request on to the socket control method, and if the
2940 	 * latter returns EOPNOTSUPP, directly to the interface.
2941 	 *
2942 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2943 	 * trust SIOCSIFADDR et al to come from an already privileged
2944 	 * layer, and do not perform any credentials checks or input
2945 	 * validation.
2946 	 */
2947 	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
2948 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2949 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2950 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2951 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2952 
2953 	if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP))
2954 		if_up(ifp);
2955 out_ref:
2956 	if_rele(ifp);
2957 out_noref:
2958 	CURVNET_RESTORE();
2959 #ifdef COMPAT_FREEBSD32
2960 	if (error != 0)
2961 		return (error);
2962 	switch (saved_cmd) {
2963 	case SIOCGIFCONF32:
2964 		ifc32->ifc_len = thunk.ifc.ifc_len;
2965 		break;
2966 	case SIOCGDRVSPEC32:
2967 		/*
2968 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
2969 		 * the struct so just assert that ifd_len (the only
2970 		 * field it might make sense to update) hasn't
2971 		 * changed.
2972 		 */
2973 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
2974 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
2975 			thunk.ifd.ifd_len));
2976 		break;
2977 	case SIOCGIFGROUP32:
2978 	case SIOCGIFGMEMB32:
2979 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
2980 		break;
2981 	case SIOCGIFMEDIA32:
2982 	case SIOCGIFXMEDIA32:
2983 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
2984 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
2985 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
2986 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
2987 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
2988 		break;
2989 	}
2990 #endif
2991 	return (error);
2992 }
2993 
2994 int
if_rename(struct ifnet * ifp,char * new_name)2995 if_rename(struct ifnet *ifp, char *new_name)
2996 {
2997 	struct ifaddr *ifa;
2998 	struct sockaddr_dl *sdl;
2999 	size_t namelen, onamelen;
3000 	char old_name[IFNAMSIZ];
3001 	char strbuf[IFNAMSIZ + 8];
3002 
3003 	if (new_name[0] == '\0')
3004 		return (EINVAL);
3005 	if (strcmp(new_name, ifp->if_xname) == 0)
3006 		return (0);
3007 	if (ifunit(new_name) != NULL)
3008 		return (EEXIST);
3009 
3010 	if_printf(ifp, "changing name to '%s'\n", new_name);
3011 
3012 	IF_ADDR_WLOCK(ifp);
3013 	strlcpy(old_name, ifp->if_xname, sizeof(old_name));
3014 	strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
3015 	ifa = ifp->if_addr;
3016 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3017 	namelen = strlen(new_name);
3018 	onamelen = sdl->sdl_nlen;
3019 	/*
3020 	 * Move the address if needed.  This is safe because we
3021 	 * allocate space for a name of length IFNAMSIZ when we
3022 	 * create this in if_attach().
3023 	 */
3024 	if (namelen != onamelen) {
3025 		bcopy(sdl->sdl_data + onamelen,
3026 		    sdl->sdl_data + namelen, sdl->sdl_alen);
3027 	}
3028 	bcopy(new_name, sdl->sdl_data, namelen);
3029 	sdl->sdl_nlen = namelen;
3030 	sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
3031 	bzero(sdl->sdl_data, onamelen);
3032 	while (namelen != 0)
3033 		sdl->sdl_data[--namelen] = 0xff;
3034 	IF_ADDR_WUNLOCK(ifp);
3035 
3036 	EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name);
3037 
3038 	snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
3039 	devctl_notify("IFNET", old_name, "RENAME", strbuf);
3040 
3041 	return (0);
3042 }
3043 
3044 /*
3045  * The code common to handling reference counted flags,
3046  * e.g., in ifpromisc() and if_allmulti().
3047  * The "pflag" argument can specify a permanent mode flag to check,
3048  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3049  *
3050  * Only to be used on stack-owned flags, not driver-owned flags.
3051  */
3052 static int
if_setflag(struct ifnet * ifp,int flag,int pflag,int * refcount,int onswitch)3053 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3054 {
3055 	struct ifreq ifr;
3056 	int error;
3057 	int oldflags, oldcount;
3058 
3059 	/* Sanity checks to catch programming errors */
3060 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3061 	    ("%s: setting driver-owned flag %d", __func__, flag));
3062 
3063 	if (onswitch)
3064 		KASSERT(*refcount >= 0,
3065 		    ("%s: increment negative refcount %d for flag %d",
3066 		    __func__, *refcount, flag));
3067 	else
3068 		KASSERT(*refcount > 0,
3069 		    ("%s: decrement non-positive refcount %d for flag %d",
3070 		    __func__, *refcount, flag));
3071 
3072 	/* In case this mode is permanent, just touch refcount */
3073 	if (ifp->if_flags & pflag) {
3074 		*refcount += onswitch ? 1 : -1;
3075 		return (0);
3076 	}
3077 
3078 	/* Save ifnet parameters for if_ioctl() may fail */
3079 	oldcount = *refcount;
3080 	oldflags = ifp->if_flags;
3081 
3082 	/*
3083 	 * See if we aren't the only and touching refcount is enough.
3084 	 * Actually toggle interface flag if we are the first or last.
3085 	 */
3086 	if (onswitch) {
3087 		if ((*refcount)++)
3088 			return (0);
3089 		ifp->if_flags |= flag;
3090 	} else {
3091 		if (--(*refcount))
3092 			return (0);
3093 		ifp->if_flags &= ~flag;
3094 	}
3095 
3096 	/* Call down the driver since we've changed interface flags */
3097 	if (ifp->if_ioctl == NULL) {
3098 		error = EOPNOTSUPP;
3099 		goto recover;
3100 	}
3101 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3102 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3103 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3104 	if (error)
3105 		goto recover;
3106 	/* Notify userland that interface flags have changed */
3107 	rt_ifmsg(ifp, flag);
3108 	return (0);
3109 
3110 recover:
3111 	/* Recover after driver error */
3112 	*refcount = oldcount;
3113 	ifp->if_flags = oldflags;
3114 	return (error);
3115 }
3116 
3117 /*
3118  * Set/clear promiscuous mode on interface ifp based on the truth value
3119  * of pswitch.  The calls are reference counted so that only the first
3120  * "on" request actually has an effect, as does the final "off" request.
3121  * Results are undefined if the "off" and "on" requests are not matched.
3122  */
3123 int
ifpromisc(struct ifnet * ifp,int pswitch)3124 ifpromisc(struct ifnet *ifp, int pswitch)
3125 {
3126 	int error;
3127 	int oldflags = ifp->if_flags;
3128 
3129 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3130 			   &ifp->if_pcount, pswitch);
3131 	/* If promiscuous mode status has changed, log a message */
3132 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3133             log_promisc_mode_change)
3134 		if_printf(ifp, "promiscuous mode %s\n",
3135 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3136 	return (error);
3137 }
3138 
3139 /*
3140  * Return interface configuration
3141  * of system.  List may be used
3142  * in later ioctl's (above) to get
3143  * other information.
3144  */
3145 /*ARGSUSED*/
3146 static int
ifconf(u_long cmd,caddr_t data)3147 ifconf(u_long cmd, caddr_t data)
3148 {
3149 	struct ifconf *ifc = (struct ifconf *)data;
3150 	struct ifnet *ifp;
3151 	struct ifaddr *ifa;
3152 	struct ifreq ifr;
3153 	struct sbuf *sb;
3154 	int error, full = 0, valid_len, max_len;
3155 
3156 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3157 	max_len = maxphys - 1;
3158 
3159 	/* Prevent hostile input from being able to crash the system */
3160 	if (ifc->ifc_len <= 0)
3161 		return (EINVAL);
3162 
3163 again:
3164 	if (ifc->ifc_len <= max_len) {
3165 		max_len = ifc->ifc_len;
3166 		full = 1;
3167 	}
3168 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3169 	max_len = 0;
3170 	valid_len = 0;
3171 
3172 	IFNET_RLOCK();
3173 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3174 		struct epoch_tracker et;
3175 		int addrs;
3176 
3177 		/*
3178 		 * Zero the ifr to make sure we don't disclose the contents
3179 		 * of the stack.
3180 		 */
3181 		memset(&ifr, 0, sizeof(ifr));
3182 
3183 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3184 		    >= sizeof(ifr.ifr_name)) {
3185 			sbuf_delete(sb);
3186 			IFNET_RUNLOCK();
3187 			return (ENAMETOOLONG);
3188 		}
3189 
3190 		addrs = 0;
3191 		NET_EPOCH_ENTER(et);
3192 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3193 			struct sockaddr *sa = ifa->ifa_addr;
3194 
3195 			if (prison_if(curthread->td_ucred, sa) != 0)
3196 				continue;
3197 			addrs++;
3198 			if (sa->sa_len <= sizeof(*sa)) {
3199 				if (sa->sa_len < sizeof(*sa)) {
3200 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3201 					    sizeof(ifr.ifr_ifru.ifru_addr));
3202 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3203 					    sa->sa_len);
3204 				} else
3205 					ifr.ifr_ifru.ifru_addr = *sa;
3206 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3207 				max_len += sizeof(ifr);
3208 			} else {
3209 				sbuf_bcat(sb, &ifr,
3210 				    offsetof(struct ifreq, ifr_addr));
3211 				max_len += offsetof(struct ifreq, ifr_addr);
3212 				sbuf_bcat(sb, sa, sa->sa_len);
3213 				max_len += sa->sa_len;
3214 			}
3215 
3216 			if (sbuf_error(sb) == 0)
3217 				valid_len = sbuf_len(sb);
3218 		}
3219 		NET_EPOCH_EXIT(et);
3220 		if (addrs == 0) {
3221 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3222 			max_len += sizeof(ifr);
3223 
3224 			if (sbuf_error(sb) == 0)
3225 				valid_len = sbuf_len(sb);
3226 		}
3227 	}
3228 	IFNET_RUNLOCK();
3229 
3230 	/*
3231 	 * If we didn't allocate enough space (uncommon), try again.  If
3232 	 * we have already allocated as much space as we are allowed,
3233 	 * return what we've got.
3234 	 */
3235 	if (valid_len != max_len && !full) {
3236 		sbuf_delete(sb);
3237 		goto again;
3238 	}
3239 
3240 	ifc->ifc_len = valid_len;
3241 	sbuf_finish(sb);
3242 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3243 	sbuf_delete(sb);
3244 	return (error);
3245 }
3246 
3247 /*
3248  * Just like ifpromisc(), but for all-multicast-reception mode.
3249  */
3250 int
if_allmulti(struct ifnet * ifp,int onswitch)3251 if_allmulti(struct ifnet *ifp, int onswitch)
3252 {
3253 
3254 	return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount,
3255 	    onswitch));
3256 }
3257 
3258 struct ifmultiaddr *
if_findmulti(struct ifnet * ifp,const struct sockaddr * sa)3259 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3260 {
3261 	struct ifmultiaddr *ifma;
3262 
3263 	IF_ADDR_LOCK_ASSERT(ifp);
3264 
3265 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3266 		if (sa->sa_family == AF_LINK) {
3267 			if (sa_dl_equal(ifma->ifma_addr, sa))
3268 				break;
3269 		} else {
3270 			if (sa_equal(ifma->ifma_addr, sa))
3271 				break;
3272 		}
3273 	}
3274 
3275 	return ifma;
3276 }
3277 
3278 /*
3279  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3280  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3281  * the ifnet multicast address list here, so the caller must do that and
3282  * other setup work (such as notifying the device driver).  The reference
3283  * count is initialized to 1.
3284  */
3285 static struct ifmultiaddr *
if_allocmulti(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr * llsa,int mflags)3286 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3287     int mflags)
3288 {
3289 	struct ifmultiaddr *ifma;
3290 	struct sockaddr *dupsa;
3291 
3292 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3293 	    M_ZERO);
3294 	if (ifma == NULL)
3295 		return (NULL);
3296 
3297 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3298 	if (dupsa == NULL) {
3299 		free(ifma, M_IFMADDR);
3300 		return (NULL);
3301 	}
3302 	bcopy(sa, dupsa, sa->sa_len);
3303 	ifma->ifma_addr = dupsa;
3304 
3305 	ifma->ifma_ifp = ifp;
3306 	ifma->ifma_refcount = 1;
3307 	ifma->ifma_protospec = NULL;
3308 
3309 	if (llsa == NULL) {
3310 		ifma->ifma_lladdr = NULL;
3311 		return (ifma);
3312 	}
3313 
3314 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3315 	if (dupsa == NULL) {
3316 		free(ifma->ifma_addr, M_IFMADDR);
3317 		free(ifma, M_IFMADDR);
3318 		return (NULL);
3319 	}
3320 	bcopy(llsa, dupsa, llsa->sa_len);
3321 	ifma->ifma_lladdr = dupsa;
3322 
3323 	return (ifma);
3324 }
3325 
3326 /*
3327  * if_freemulti: free ifmultiaddr structure and possibly attached related
3328  * addresses.  The caller is responsible for implementing reference
3329  * counting, notifying the driver, handling routing messages, and releasing
3330  * any dependent link layer state.
3331  */
3332 #ifdef MCAST_VERBOSE
3333 extern void kdb_backtrace(void);
3334 #endif
3335 static void
if_freemulti_internal(struct ifmultiaddr * ifma)3336 if_freemulti_internal(struct ifmultiaddr *ifma)
3337 {
3338 
3339 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3340 	    ifma->ifma_refcount));
3341 
3342 	if (ifma->ifma_lladdr != NULL)
3343 		free(ifma->ifma_lladdr, M_IFMADDR);
3344 #ifdef MCAST_VERBOSE
3345 	kdb_backtrace();
3346 	printf("%s freeing ifma: %p\n", __func__, ifma);
3347 #endif
3348 	free(ifma->ifma_addr, M_IFMADDR);
3349 	free(ifma, M_IFMADDR);
3350 }
3351 
3352 static void
if_destroymulti(epoch_context_t ctx)3353 if_destroymulti(epoch_context_t ctx)
3354 {
3355 	struct ifmultiaddr *ifma;
3356 
3357 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3358 	if_freemulti_internal(ifma);
3359 }
3360 
3361 void
if_freemulti(struct ifmultiaddr * ifma)3362 if_freemulti(struct ifmultiaddr *ifma)
3363 {
3364 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3365 	    ifma->ifma_refcount));
3366 
3367 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3368 }
3369 
3370 /*
3371  * Register an additional multicast address with a network interface.
3372  *
3373  * - If the address is already present, bump the reference count on the
3374  *   address and return.
3375  * - If the address is not link-layer, look up a link layer address.
3376  * - Allocate address structures for one or both addresses, and attach to the
3377  *   multicast address list on the interface.  If automatically adding a link
3378  *   layer address, the protocol address will own a reference to the link
3379  *   layer address, to be freed when it is freed.
3380  * - Notify the network device driver of an addition to the multicast address
3381  *   list.
3382  *
3383  * 'sa' points to caller-owned memory with the desired multicast address.
3384  *
3385  * 'retifma' will be used to return a pointer to the resulting multicast
3386  * address reference, if desired.
3387  */
3388 int
if_addmulti(struct ifnet * ifp,struct sockaddr * sa,struct ifmultiaddr ** retifma)3389 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3390     struct ifmultiaddr **retifma)
3391 {
3392 	struct ifmultiaddr *ifma, *ll_ifma;
3393 	struct sockaddr *llsa;
3394 	struct sockaddr_dl sdl;
3395 	int error;
3396 
3397 #ifdef INET
3398 	IN_MULTI_LIST_UNLOCK_ASSERT();
3399 #endif
3400 #ifdef INET6
3401 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3402 #endif
3403 	/*
3404 	 * If the address is already present, return a new reference to it;
3405 	 * otherwise, allocate storage and set up a new address.
3406 	 */
3407 	IF_ADDR_WLOCK(ifp);
3408 	ifma = if_findmulti(ifp, sa);
3409 	if (ifma != NULL) {
3410 		ifma->ifma_refcount++;
3411 		if (retifma != NULL)
3412 			*retifma = ifma;
3413 		IF_ADDR_WUNLOCK(ifp);
3414 		return (0);
3415 	}
3416 
3417 	/*
3418 	 * The address isn't already present; resolve the protocol address
3419 	 * into a link layer address, and then look that up, bump its
3420 	 * refcount or allocate an ifma for that also.
3421 	 * Most link layer resolving functions returns address data which
3422 	 * fits inside default sockaddr_dl structure. However callback
3423 	 * can allocate another sockaddr structure, in that case we need to
3424 	 * free it later.
3425 	 */
3426 	llsa = NULL;
3427 	ll_ifma = NULL;
3428 	if (ifp->if_resolvemulti != NULL) {
3429 		/* Provide called function with buffer size information */
3430 		sdl.sdl_len = sizeof(sdl);
3431 		llsa = (struct sockaddr *)&sdl;
3432 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3433 		if (error)
3434 			goto unlock_out;
3435 	}
3436 
3437 	/*
3438 	 * Allocate the new address.  Don't hook it up yet, as we may also
3439 	 * need to allocate a link layer multicast address.
3440 	 */
3441 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3442 	if (ifma == NULL) {
3443 		error = ENOMEM;
3444 		goto free_llsa_out;
3445 	}
3446 
3447 	/*
3448 	 * If a link layer address is found, we'll need to see if it's
3449 	 * already present in the address list, or allocate is as well.
3450 	 * When this block finishes, the link layer address will be on the
3451 	 * list.
3452 	 */
3453 	if (llsa != NULL) {
3454 		ll_ifma = if_findmulti(ifp, llsa);
3455 		if (ll_ifma == NULL) {
3456 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3457 			if (ll_ifma == NULL) {
3458 				--ifma->ifma_refcount;
3459 				if_freemulti(ifma);
3460 				error = ENOMEM;
3461 				goto free_llsa_out;
3462 			}
3463 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3464 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3465 			    ifma_link);
3466 		} else
3467 			ll_ifma->ifma_refcount++;
3468 		ifma->ifma_llifma = ll_ifma;
3469 	}
3470 
3471 	/*
3472 	 * We now have a new multicast address, ifma, and possibly a new or
3473 	 * referenced link layer address.  Add the primary address to the
3474 	 * ifnet address list.
3475 	 */
3476 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3477 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3478 
3479 	if (retifma != NULL)
3480 		*retifma = ifma;
3481 
3482 	/*
3483 	 * Must generate the message while holding the lock so that 'ifma'
3484 	 * pointer is still valid.
3485 	 */
3486 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3487 	IF_ADDR_WUNLOCK(ifp);
3488 
3489 	/*
3490 	 * We are certain we have added something, so call down to the
3491 	 * interface to let them know about it.
3492 	 */
3493 	if (ifp->if_ioctl != NULL) {
3494 		if (THREAD_CAN_SLEEP())
3495 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3496 		else
3497 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3498 	}
3499 
3500 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3501 		link_free_sdl(llsa);
3502 
3503 	return (0);
3504 
3505 free_llsa_out:
3506 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3507 		link_free_sdl(llsa);
3508 
3509 unlock_out:
3510 	IF_ADDR_WUNLOCK(ifp);
3511 	return (error);
3512 }
3513 
3514 static void
if_siocaddmulti(void * arg,int pending)3515 if_siocaddmulti(void *arg, int pending)
3516 {
3517 	struct ifnet *ifp;
3518 
3519 	ifp = arg;
3520 #ifdef DIAGNOSTIC
3521 	if (pending > 1)
3522 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3523 #endif
3524 	CURVNET_SET(ifp->if_vnet);
3525 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3526 	CURVNET_RESTORE();
3527 }
3528 
3529 /*
3530  * Delete a multicast group membership by network-layer group address.
3531  *
3532  * Returns ENOENT if the entry could not be found. If ifp no longer
3533  * exists, results are undefined. This entry point should only be used
3534  * from subsystems which do appropriate locking to hold ifp for the
3535  * duration of the call.
3536  * Network-layer protocol domains must use if_delmulti_ifma().
3537  */
3538 int
if_delmulti(struct ifnet * ifp,struct sockaddr * sa)3539 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3540 {
3541 	struct ifmultiaddr *ifma;
3542 	int lastref;
3543 
3544 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3545 
3546 	IF_ADDR_WLOCK(ifp);
3547 	lastref = 0;
3548 	ifma = if_findmulti(ifp, sa);
3549 	if (ifma != NULL)
3550 		lastref = if_delmulti_locked(ifp, ifma, 0);
3551 	IF_ADDR_WUNLOCK(ifp);
3552 
3553 	if (ifma == NULL)
3554 		return (ENOENT);
3555 
3556 	if (lastref && ifp->if_ioctl != NULL) {
3557 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3558 	}
3559 
3560 	return (0);
3561 }
3562 
3563 /*
3564  * Delete all multicast group membership for an interface.
3565  * Should be used to quickly flush all multicast filters.
3566  */
3567 void
if_delallmulti(struct ifnet * ifp)3568 if_delallmulti(struct ifnet *ifp)
3569 {
3570 	struct ifmultiaddr *ifma;
3571 	struct ifmultiaddr *next;
3572 
3573 	IF_ADDR_WLOCK(ifp);
3574 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3575 		if_delmulti_locked(ifp, ifma, 0);
3576 	IF_ADDR_WUNLOCK(ifp);
3577 }
3578 
3579 void
if_delmulti_ifma(struct ifmultiaddr * ifma)3580 if_delmulti_ifma(struct ifmultiaddr *ifma)
3581 {
3582 	if_delmulti_ifma_flags(ifma, 0);
3583 }
3584 
3585 /*
3586  * Delete a multicast group membership by group membership pointer.
3587  * Network-layer protocol domains must use this routine.
3588  *
3589  * It is safe to call this routine if the ifp disappeared.
3590  */
3591 void
if_delmulti_ifma_flags(struct ifmultiaddr * ifma,int flags)3592 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3593 {
3594 	struct ifnet *ifp;
3595 	int lastref;
3596 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3597 #ifdef INET
3598 	IN_MULTI_LIST_UNLOCK_ASSERT();
3599 #endif
3600 	ifp = ifma->ifma_ifp;
3601 #ifdef DIAGNOSTIC
3602 	if (ifp == NULL) {
3603 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3604 	} else {
3605 		struct epoch_tracker et;
3606 		struct ifnet *oifp;
3607 
3608 		NET_EPOCH_ENTER(et);
3609 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3610 			if (ifp == oifp)
3611 				break;
3612 		NET_EPOCH_EXIT(et);
3613 		if (ifp != oifp)
3614 			ifp = NULL;
3615 	}
3616 #endif
3617 	/*
3618 	 * If and only if the ifnet instance exists: Acquire the address lock.
3619 	 */
3620 	if (ifp != NULL)
3621 		IF_ADDR_WLOCK(ifp);
3622 
3623 	lastref = if_delmulti_locked(ifp, ifma, flags);
3624 
3625 	if (ifp != NULL) {
3626 		/*
3627 		 * If and only if the ifnet instance exists:
3628 		 *  Release the address lock.
3629 		 *  If the group was left: update the hardware hash filter.
3630 		 */
3631 		IF_ADDR_WUNLOCK(ifp);
3632 		if (lastref && ifp->if_ioctl != NULL) {
3633 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3634 		}
3635 	}
3636 }
3637 
3638 /*
3639  * Perform deletion of network-layer and/or link-layer multicast address.
3640  *
3641  * Return 0 if the reference count was decremented.
3642  * Return 1 if the final reference was released, indicating that the
3643  * hardware hash filter should be reprogrammed.
3644  */
3645 static int
if_delmulti_locked(struct ifnet * ifp,struct ifmultiaddr * ifma,int detaching)3646 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3647 {
3648 	struct ifmultiaddr *ll_ifma;
3649 
3650 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3651 		KASSERT(ifma->ifma_ifp == ifp,
3652 		    ("%s: inconsistent ifp %p", __func__, ifp));
3653 		IF_ADDR_WLOCK_ASSERT(ifp);
3654 	}
3655 
3656 	ifp = ifma->ifma_ifp;
3657 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3658 
3659 	/*
3660 	 * If the ifnet is detaching, null out references to ifnet,
3661 	 * so that upper protocol layers will notice, and not attempt
3662 	 * to obtain locks for an ifnet which no longer exists. The
3663 	 * routing socket announcement must happen before the ifnet
3664 	 * instance is detached from the system.
3665 	 */
3666 	if (detaching) {
3667 #ifdef DIAGNOSTIC
3668 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3669 #endif
3670 		/*
3671 		 * ifp may already be nulled out if we are being reentered
3672 		 * to delete the ll_ifma.
3673 		 */
3674 		if (ifp != NULL) {
3675 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3676 			ifma->ifma_ifp = NULL;
3677 		}
3678 	}
3679 
3680 	if (--ifma->ifma_refcount > 0)
3681 		return 0;
3682 
3683 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3684 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3685 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3686 	}
3687 	/*
3688 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3689 	 * have been associated with it. Release it first if so.
3690 	 */
3691 	ll_ifma = ifma->ifma_llifma;
3692 	if (ll_ifma != NULL) {
3693 		KASSERT(ifma->ifma_lladdr != NULL,
3694 		    ("%s: llifma w/o lladdr", __func__));
3695 		if (detaching)
3696 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3697 		if (--ll_ifma->ifma_refcount == 0) {
3698 			if (ifp != NULL) {
3699 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3700 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3701 						ifma_link);
3702 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3703 				}
3704 			}
3705 			if_freemulti(ll_ifma);
3706 		}
3707 	}
3708 #ifdef INVARIANTS
3709 	if (ifp) {
3710 		struct ifmultiaddr *ifmatmp;
3711 
3712 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3713 			MPASS(ifma != ifmatmp);
3714 	}
3715 #endif
3716 	if_freemulti(ifma);
3717 	/*
3718 	 * The last reference to this instance of struct ifmultiaddr
3719 	 * was released; the hardware should be notified of this change.
3720 	 */
3721 	return 1;
3722 }
3723 
3724 /*
3725  * Set the link layer address on an interface.
3726  *
3727  * At this time we only support certain types of interfaces,
3728  * and we don't allow the length of the address to change.
3729  *
3730  * Set noinline to be dtrace-friendly
3731  */
3732 __noinline int
if_setlladdr(struct ifnet * ifp,const u_char * lladdr,int len)3733 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3734 {
3735 	struct sockaddr_dl *sdl;
3736 	struct ifaddr *ifa;
3737 	struct ifreq ifr;
3738 
3739 	ifa = ifp->if_addr;
3740 	if (ifa == NULL)
3741 		return (EINVAL);
3742 
3743 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3744 	if (sdl == NULL)
3745 		return (EINVAL);
3746 
3747 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3748 		return (EINVAL);
3749 
3750 	switch (ifp->if_type) {
3751 	case IFT_ETHER:
3752 	case IFT_XETHER:
3753 	case IFT_L2VLAN:
3754 	case IFT_BRIDGE:
3755 	case IFT_IEEE8023ADLAG:
3756 		bcopy(lladdr, LLADDR(sdl), len);
3757 		break;
3758 	default:
3759 		return (ENODEV);
3760 	}
3761 
3762 	/*
3763 	 * If the interface is already up, we need
3764 	 * to re-init it in order to reprogram its
3765 	 * address filter.
3766 	 */
3767 	if ((ifp->if_flags & IFF_UP) != 0) {
3768 		if (ifp->if_ioctl) {
3769 			ifp->if_flags &= ~IFF_UP;
3770 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3771 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3772 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3773 			ifp->if_flags |= IFF_UP;
3774 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3775 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3776 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3777 		}
3778 	}
3779 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3780 
3781 	return (0);
3782 }
3783 
3784 /*
3785  * Compat function for handling basic encapsulation requests.
3786  * Not converted stacks (FDDI, IB, ..) supports traditional
3787  * output model: ARP (and other similar L2 protocols) are handled
3788  * inside output routine, arpresolve/nd6_resolve() returns MAC
3789  * address instead of full prepend.
3790  *
3791  * This function creates calculated header==MAC for IPv4/IPv6 and
3792  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3793  * address families.
3794  */
3795 static int
if_requestencap_default(struct ifnet * ifp,struct if_encap_req * req)3796 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3797 {
3798 	if (req->rtype != IFENCAP_LL)
3799 		return (EOPNOTSUPP);
3800 
3801 	if (req->bufsize < req->lladdr_len)
3802 		return (ENOMEM);
3803 
3804 	switch (req->family) {
3805 	case AF_INET:
3806 	case AF_INET6:
3807 		break;
3808 	default:
3809 		return (EAFNOSUPPORT);
3810 	}
3811 
3812 	/* Copy lladdr to storage as is */
3813 	memmove(req->buf, req->lladdr, req->lladdr_len);
3814 	req->bufsize = req->lladdr_len;
3815 	req->lladdr_off = 0;
3816 
3817 	return (0);
3818 }
3819 
3820 /*
3821  * Tunnel interfaces can nest, also they may cause infinite recursion
3822  * calls when misconfigured. We'll prevent this by detecting loops.
3823  * High nesting level may cause stack exhaustion. We'll prevent this
3824  * by introducing upper limit.
3825  *
3826  * Return 0, if tunnel nesting count is equal or less than limit.
3827  */
3828 int
if_tunnel_check_nesting(struct ifnet * ifp,struct mbuf * m,uint32_t cookie,int limit)3829 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3830     int limit)
3831 {
3832 	struct m_tag *mtag;
3833 	int count;
3834 
3835 	count = 1;
3836 	mtag = NULL;
3837 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3838 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3839 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3840 			return (EIO);
3841 		}
3842 		count++;
3843 	}
3844 	if (count > limit) {
3845 		log(LOG_NOTICE,
3846 		    "%s: if_output recursively called too many times(%d)\n",
3847 		    if_name(ifp), count);
3848 		return (EIO);
3849 	}
3850 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3851 	if (mtag == NULL)
3852 		return (ENOMEM);
3853 	*(struct ifnet **)(mtag + 1) = ifp;
3854 	m_tag_prepend(m, mtag);
3855 	return (0);
3856 }
3857 
3858 /*
3859  * Get the link layer address that was read from the hardware at attach.
3860  *
3861  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3862  * their component interfaces as IFT_IEEE8023ADLAG.
3863  */
3864 int
if_gethwaddr(struct ifnet * ifp,struct ifreq * ifr)3865 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3866 {
3867 	if (ifp->if_hw_addr == NULL)
3868 		return (ENODEV);
3869 
3870 	switch (ifp->if_type) {
3871 	case IFT_ETHER:
3872 	case IFT_IEEE8023ADLAG:
3873 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3874 		return (0);
3875 	default:
3876 		return (ENODEV);
3877 	}
3878 }
3879 
3880 /*
3881  * The name argument must be a pointer to storage which will last as
3882  * long as the interface does.  For physical devices, the result of
3883  * device_get_name(dev) is a good choice and for pseudo-devices a
3884  * static string works well.
3885  */
3886 void
if_initname(struct ifnet * ifp,const char * name,int unit)3887 if_initname(struct ifnet *ifp, const char *name, int unit)
3888 {
3889 	ifp->if_dname = name;
3890 	ifp->if_dunit = unit;
3891 	if (unit != IF_DUNIT_NONE)
3892 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3893 	else
3894 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3895 }
3896 
3897 static int
if_vlog(struct ifnet * ifp,int pri,const char * fmt,va_list ap)3898 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
3899 {
3900 	char if_fmt[256];
3901 
3902 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3903 	vlog(pri, if_fmt, ap);
3904 	return (0);
3905 }
3906 
3907 
3908 int
if_printf(struct ifnet * ifp,const char * fmt,...)3909 if_printf(struct ifnet *ifp, const char *fmt, ...)
3910 {
3911 	va_list ap;
3912 
3913 	va_start(ap, fmt);
3914 	if_vlog(ifp, LOG_INFO, fmt, ap);
3915 	va_end(ap);
3916 	return (0);
3917 }
3918 
3919 int
if_log(struct ifnet * ifp,int pri,const char * fmt,...)3920 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
3921 {
3922 	va_list ap;
3923 
3924 	va_start(ap, fmt);
3925 	if_vlog(ifp, pri, fmt, ap);
3926 	va_end(ap);
3927 	return (0);
3928 }
3929 
3930 void
if_start(struct ifnet * ifp)3931 if_start(struct ifnet *ifp)
3932 {
3933 
3934 	(*(ifp)->if_start)(ifp);
3935 }
3936 
3937 /*
3938  * Backwards compatibility interface for drivers
3939  * that have not implemented it
3940  */
3941 static int
if_transmit_default(struct ifnet * ifp,struct mbuf * m)3942 if_transmit_default(struct ifnet *ifp, struct mbuf *m)
3943 {
3944 	int error;
3945 
3946 	IFQ_HANDOFF(ifp, m, error);
3947 	return (error);
3948 }
3949 
3950 static void
if_input_default(struct ifnet * ifp __unused,struct mbuf * m)3951 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3952 {
3953 	m_freem(m);
3954 }
3955 
3956 int
if_handoff(struct ifqueue * ifq,struct mbuf * m,struct ifnet * ifp,int adjust)3957 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3958 {
3959 	int active = 0;
3960 
3961 	IF_LOCK(ifq);
3962 	if (_IF_QFULL(ifq)) {
3963 		IF_UNLOCK(ifq);
3964 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3965 		m_freem(m);
3966 		return (0);
3967 	}
3968 	if (ifp != NULL) {
3969 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3970 		if (m->m_flags & (M_BCAST|M_MCAST))
3971 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3972 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3973 	}
3974 	_IF_ENQUEUE(ifq, m);
3975 	IF_UNLOCK(ifq);
3976 	if (ifp != NULL && !active)
3977 		(*(ifp)->if_start)(ifp);
3978 	return (1);
3979 }
3980 
3981 void
if_register_com_alloc(u_char type,if_com_alloc_t * a,if_com_free_t * f)3982 if_register_com_alloc(u_char type,
3983     if_com_alloc_t *a, if_com_free_t *f)
3984 {
3985 
3986 	KASSERT(if_com_alloc[type] == NULL,
3987 	    ("if_register_com_alloc: %d already registered", type));
3988 	KASSERT(if_com_free[type] == NULL,
3989 	    ("if_register_com_alloc: %d free already registered", type));
3990 
3991 	if_com_alloc[type] = a;
3992 	if_com_free[type] = f;
3993 }
3994 
3995 void
if_deregister_com_alloc(u_char type)3996 if_deregister_com_alloc(u_char type)
3997 {
3998 
3999 	KASSERT(if_com_alloc[type] != NULL,
4000 	    ("if_deregister_com_alloc: %d not registered", type));
4001 	KASSERT(if_com_free[type] != NULL,
4002 	    ("if_deregister_com_alloc: %d free not registered", type));
4003 
4004 	/*
4005 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4006 	 * fixes issues about late invocation of if_destroy(), which leads
4007 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4008 	 */
4009 	NET_EPOCH_DRAIN_CALLBACKS();
4010 
4011 	if_com_alloc[type] = NULL;
4012 	if_com_free[type] = NULL;
4013 }
4014 
4015 /* API for driver access to network stack owned ifnet.*/
4016 uint64_t
if_setbaudrate(struct ifnet * ifp,uint64_t baudrate)4017 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4018 {
4019 	uint64_t oldbrate;
4020 
4021 	oldbrate = ifp->if_baudrate;
4022 	ifp->if_baudrate = baudrate;
4023 	return (oldbrate);
4024 }
4025 
4026 uint64_t
if_getbaudrate(const if_t ifp)4027 if_getbaudrate(const if_t ifp)
4028 {
4029 	return (ifp->if_baudrate);
4030 }
4031 
4032 int
if_setcapabilities(if_t ifp,int capabilities)4033 if_setcapabilities(if_t ifp, int capabilities)
4034 {
4035 	ifp->if_capabilities = capabilities;
4036 	return (0);
4037 }
4038 
4039 int
if_setcapabilitiesbit(if_t ifp,int setbit,int clearbit)4040 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4041 {
4042 	ifp->if_capabilities &= ~clearbit;
4043 	ifp->if_capabilities |= setbit;
4044 	return (0);
4045 }
4046 
4047 int
if_getcapabilities(const if_t ifp)4048 if_getcapabilities(const if_t ifp)
4049 {
4050 	return (ifp->if_capabilities);
4051 }
4052 
4053 int
if_setcapenable(if_t ifp,int capabilities)4054 if_setcapenable(if_t ifp, int capabilities)
4055 {
4056 	ifp->if_capenable = capabilities;
4057 	return (0);
4058 }
4059 
4060 int
if_setcapenablebit(if_t ifp,int setcap,int clearcap)4061 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4062 {
4063 	ifp->if_capenable &= ~clearcap;
4064 	ifp->if_capenable |= setcap;
4065 	return (0);
4066 }
4067 
4068 int
if_setcapabilities2(if_t ifp,int capabilities)4069 if_setcapabilities2(if_t ifp, int capabilities)
4070 {
4071 	ifp->if_capabilities2 = capabilities;
4072 	return (0);
4073 }
4074 
4075 int
if_setcapabilities2bit(if_t ifp,int setbit,int clearbit)4076 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit)
4077 {
4078 	ifp->if_capabilities2 &= ~clearbit;
4079 	ifp->if_capabilities2 |= setbit;
4080 	return (0);
4081 }
4082 
4083 int
if_getcapabilities2(const if_t ifp)4084 if_getcapabilities2(const if_t ifp)
4085 {
4086 	return (ifp->if_capabilities2);
4087 }
4088 
4089 int
if_setcapenable2(if_t ifp,int capabilities2)4090 if_setcapenable2(if_t ifp, int capabilities2)
4091 {
4092 	ifp->if_capenable2 = capabilities2;
4093 	return (0);
4094 }
4095 
4096 int
if_setcapenable2bit(if_t ifp,int setcap,int clearcap)4097 if_setcapenable2bit(if_t ifp, int setcap, int clearcap)
4098 {
4099 	ifp->if_capenable2 &= ~clearcap;
4100 	ifp->if_capenable2 |= setcap;
4101 	return (0);
4102 }
4103 
4104 const char *
if_getdname(const if_t ifp)4105 if_getdname(const if_t ifp)
4106 {
4107 	return (ifp->if_dname);
4108 }
4109 
4110 void
if_setdname(if_t ifp,const char * dname)4111 if_setdname(if_t ifp, const char *dname)
4112 {
4113 	ifp->if_dname = dname;
4114 }
4115 
4116 const char *
if_name(if_t ifp)4117 if_name(if_t ifp)
4118 {
4119 	return (ifp->if_xname);
4120 }
4121 
4122 int
if_setname(if_t ifp,const char * name)4123 if_setname(if_t ifp, const char *name)
4124 {
4125 	if (strlen(name) > sizeof(ifp->if_xname) - 1)
4126 		return (ENAMETOOLONG);
4127 	strcpy(ifp->if_xname, name);
4128 
4129 	return (0);
4130 }
4131 
4132 int
if_togglecapenable(if_t ifp,int togglecap)4133 if_togglecapenable(if_t ifp, int togglecap)
4134 {
4135 	ifp->if_capenable ^= togglecap;
4136 	return (0);
4137 }
4138 
4139 int
if_getcapenable(const if_t ifp)4140 if_getcapenable(const if_t ifp)
4141 {
4142 	return (ifp->if_capenable);
4143 }
4144 
4145 int
if_togglecapenable2(if_t ifp,int togglecap)4146 if_togglecapenable2(if_t ifp, int togglecap)
4147 {
4148 	ifp->if_capenable2 ^= togglecap;
4149 	return (0);
4150 }
4151 
4152 int
if_getcapenable2(const if_t ifp)4153 if_getcapenable2(const if_t ifp)
4154 {
4155 	return (ifp->if_capenable2);
4156 }
4157 
4158 int
if_getdunit(const if_t ifp)4159 if_getdunit(const if_t ifp)
4160 {
4161 	return (ifp->if_dunit);
4162 }
4163 
4164 int
if_getindex(const if_t ifp)4165 if_getindex(const if_t ifp)
4166 {
4167 	return (ifp->if_index);
4168 }
4169 
4170 int
if_getidxgen(const if_t ifp)4171 if_getidxgen(const if_t ifp)
4172 {
4173 	return (ifp->if_idxgen);
4174 }
4175 
4176 const char *
if_getdescr(if_t ifp)4177 if_getdescr(if_t ifp)
4178 {
4179 	return (ifp->if_description);
4180 }
4181 
4182 void
if_setdescr(if_t ifp,char * descrbuf)4183 if_setdescr(if_t ifp, char *descrbuf)
4184 {
4185 	sx_xlock(&ifdescr_sx);
4186 	char *odescrbuf = ifp->if_description;
4187 	ifp->if_description = descrbuf;
4188 	sx_xunlock(&ifdescr_sx);
4189 
4190 	if_freedescr(odescrbuf);
4191 }
4192 
4193 char *
if_allocdescr(size_t sz,int malloc_flag)4194 if_allocdescr(size_t sz, int malloc_flag)
4195 {
4196 	malloc_flag &= (M_WAITOK | M_NOWAIT);
4197 	return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag));
4198 }
4199 
4200 void
if_freedescr(char * descrbuf)4201 if_freedescr(char *descrbuf)
4202 {
4203 	free(descrbuf, M_IFDESCR);
4204 }
4205 
4206 int
if_getalloctype(const if_t ifp)4207 if_getalloctype(const if_t ifp)
4208 {
4209 	return (ifp->if_alloctype);
4210 }
4211 
4212 void
if_setlastchange(if_t ifp)4213 if_setlastchange(if_t ifp)
4214 {
4215 	getmicrotime(&ifp->if_lastchange);
4216 }
4217 
4218 /*
4219  * This is largely undesirable because it ties ifnet to a device, but does
4220  * provide flexiblity for an embedded product vendor. Should be used with
4221  * the understanding that it violates the interface boundaries, and should be
4222  * a last resort only.
4223  */
4224 int
if_setdev(if_t ifp,void * dev)4225 if_setdev(if_t ifp, void *dev)
4226 {
4227 	return (0);
4228 }
4229 
4230 int
if_setdrvflagbits(if_t ifp,int set_flags,int clear_flags)4231 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4232 {
4233 	ifp->if_drv_flags &= ~clear_flags;
4234 	ifp->if_drv_flags |= set_flags;
4235 
4236 	return (0);
4237 }
4238 
4239 int
if_getdrvflags(const if_t ifp)4240 if_getdrvflags(const if_t ifp)
4241 {
4242 	return (ifp->if_drv_flags);
4243 }
4244 
4245 int
if_setdrvflags(if_t ifp,int flags)4246 if_setdrvflags(if_t ifp, int flags)
4247 {
4248 	ifp->if_drv_flags = flags;
4249 	return (0);
4250 }
4251 
4252 int
if_setflags(if_t ifp,int flags)4253 if_setflags(if_t ifp, int flags)
4254 {
4255 	ifp->if_flags = flags;
4256 	return (0);
4257 }
4258 
4259 int
if_setflagbits(if_t ifp,int set,int clear)4260 if_setflagbits(if_t ifp, int set, int clear)
4261 {
4262 	ifp->if_flags &= ~clear;
4263 	ifp->if_flags |= set;
4264 	return (0);
4265 }
4266 
4267 int
if_getflags(const if_t ifp)4268 if_getflags(const if_t ifp)
4269 {
4270 	return (ifp->if_flags);
4271 }
4272 
4273 int
if_clearhwassist(if_t ifp)4274 if_clearhwassist(if_t ifp)
4275 {
4276 	ifp->if_hwassist = 0;
4277 	return (0);
4278 }
4279 
4280 int
if_sethwassistbits(if_t ifp,int toset,int toclear)4281 if_sethwassistbits(if_t ifp, int toset, int toclear)
4282 {
4283 	ifp->if_hwassist &= ~toclear;
4284 	ifp->if_hwassist |= toset;
4285 
4286 	return (0);
4287 }
4288 
4289 int
if_sethwassist(if_t ifp,int hwassist_bit)4290 if_sethwassist(if_t ifp, int hwassist_bit)
4291 {
4292 	ifp->if_hwassist = hwassist_bit;
4293 	return (0);
4294 }
4295 
4296 int
if_gethwassist(const if_t ifp)4297 if_gethwassist(const if_t ifp)
4298 {
4299 	return (ifp->if_hwassist);
4300 }
4301 
4302 int
if_togglehwassist(if_t ifp,int toggle_bits)4303 if_togglehwassist(if_t ifp, int toggle_bits)
4304 {
4305 	ifp->if_hwassist ^= toggle_bits;
4306 	return (0);
4307 }
4308 
4309 int
if_setmtu(if_t ifp,int mtu)4310 if_setmtu(if_t ifp, int mtu)
4311 {
4312 	ifp->if_mtu = mtu;
4313 	return (0);
4314 }
4315 
4316 void
if_notifymtu(if_t ifp)4317 if_notifymtu(if_t ifp)
4318 {
4319 #ifdef INET6
4320 	nd6_setmtu(ifp);
4321 #endif
4322 	rt_updatemtu(ifp);
4323 }
4324 
4325 int
if_getmtu(const if_t ifp)4326 if_getmtu(const if_t ifp)
4327 {
4328 	return (ifp->if_mtu);
4329 }
4330 
4331 void
if_setppromisc(if_t ifp,bool ppromisc)4332 if_setppromisc(if_t ifp, bool ppromisc)
4333 {
4334 	int new_flags;
4335 
4336 	if (ppromisc)
4337 		new_flags = ifp->if_flags | IFF_PPROMISC;
4338 	else
4339 		new_flags = ifp->if_flags & ~IFF_PPROMISC;
4340 	if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
4341 		if (new_flags & IFF_PPROMISC)
4342 			new_flags |= IFF_PROMISC;
4343 		/*
4344 		 * Only unset IFF_PROMISC if there are no more consumers of
4345 		 * promiscuity, i.e. the ifp->if_pcount refcount is 0.
4346 		 */
4347 		else if (ifp->if_pcount == 0)
4348 			new_flags &= ~IFF_PROMISC;
4349 		if (log_promisc_mode_change)
4350                         if_printf(ifp, "permanently promiscuous mode %s\n",
4351                             ((new_flags & IFF_PPROMISC) ?
4352                              "enabled" : "disabled"));
4353 	}
4354 	ifp->if_flags = new_flags;
4355 }
4356 
4357 /*
4358  * Methods for drivers to access interface unicast and multicast
4359  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4360  * 'struct ifmultiaddr'.
4361  */
4362 u_int
if_lladdr_count(if_t ifp)4363 if_lladdr_count(if_t ifp)
4364 {
4365 	struct epoch_tracker et;
4366 	struct ifaddr *ifa;
4367 	u_int count;
4368 
4369 	count = 0;
4370 	NET_EPOCH_ENTER(et);
4371 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4372 		if (ifa->ifa_addr->sa_family == AF_LINK)
4373 			count++;
4374 	NET_EPOCH_EXIT(et);
4375 
4376 	return (count);
4377 }
4378 
4379 int
if_foreach(if_foreach_cb_t cb,void * cb_arg)4380 if_foreach(if_foreach_cb_t cb, void *cb_arg)
4381 {
4382 	if_t ifp;
4383 	int error;
4384 
4385 	NET_EPOCH_ASSERT();
4386 	MPASS(cb);
4387 
4388 	error = 0;
4389 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4390 		error = cb(ifp, cb_arg);
4391 		if (error != 0)
4392 			break;
4393 	}
4394 
4395 	return (error);
4396 }
4397 
4398 /*
4399  * Iterates over the list of interfaces, permitting callback function @cb to sleep.
4400  * Stops iteration if @cb returns non-zero error code.
4401  * Returns the last error code from @cb.
4402  * @match_cb: optional match callback limiting the iteration to only matched interfaces
4403  * @match_arg: argument to pass to @match_cb
4404  * @cb: iteration callback
4405  * @cb_arg: argument to pass to @cb
4406  */
4407 int
if_foreach_sleep(if_foreach_match_t match_cb,void * match_arg,if_foreach_cb_t cb,void * cb_arg)4408 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb,
4409     void *cb_arg)
4410 {
4411 	int match_count = 0, array_size = 16; /* 128 bytes for malloc */
4412 	struct ifnet **match_array = NULL;
4413 	int error = 0;
4414 
4415 	MPASS(cb);
4416 
4417 	while (true) {
4418 		struct ifnet **new_array;
4419 		int new_size = array_size;
4420 		struct epoch_tracker et;
4421 		struct ifnet *ifp;
4422 
4423 		while (new_size < match_count)
4424 			new_size *= 2;
4425 		new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK);
4426 		if (match_array != NULL)
4427 			memcpy(new_array, match_array, array_size * sizeof(void *));
4428 		free(match_array, M_TEMP);
4429 		match_array = new_array;
4430 		array_size = new_size;
4431 
4432 		match_count = 0;
4433 		NET_EPOCH_ENTER(et);
4434 		CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4435 			if (match_cb != NULL && !match_cb(ifp, match_arg))
4436 				continue;
4437 			if (match_count < array_size) {
4438 				if (if_try_ref(ifp))
4439 					match_array[match_count++] = ifp;
4440 			} else
4441 				match_count++;
4442 		}
4443 		NET_EPOCH_EXIT(et);
4444 
4445 		if (match_count > array_size) {
4446 			for (int i = 0; i < array_size; i++)
4447 				if_rele(match_array[i]);
4448 			continue;
4449 		} else {
4450 			for (int i = 0; i < match_count; i++) {
4451 				if (error == 0)
4452 					error = cb(match_array[i], cb_arg);
4453 				if_rele(match_array[i]);
4454 			}
4455 			free(match_array, M_TEMP);
4456 			break;
4457 		}
4458 	}
4459 
4460 	return (error);
4461 }
4462 
4463 
4464 /*
4465  * Uses just 1 pointer of the 4 available in the public struct.
4466  */
4467 if_t
if_iter_start(struct if_iter * iter)4468 if_iter_start(struct if_iter *iter)
4469 {
4470 	if_t ifp;
4471 
4472 	NET_EPOCH_ASSERT();
4473 
4474 	bzero(iter, sizeof(*iter));
4475 	ifp = CK_STAILQ_FIRST(&V_ifnet);
4476 	if (ifp != NULL)
4477 		iter->context[0] = CK_STAILQ_NEXT(ifp, if_link);
4478 	else
4479 		iter->context[0] = NULL;
4480 	return (ifp);
4481 }
4482 
4483 if_t
if_iter_next(struct if_iter * iter)4484 if_iter_next(struct if_iter *iter)
4485 {
4486 	if_t cur_ifp = iter->context[0];
4487 
4488 	if (cur_ifp != NULL)
4489 		iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link);
4490 	return (cur_ifp);
4491 }
4492 
4493 void
if_iter_finish(struct if_iter * iter)4494 if_iter_finish(struct if_iter *iter)
4495 {
4496 	/* Nothing to do here for now. */
4497 }
4498 
4499 u_int
if_foreach_lladdr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4500 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4501 {
4502 	struct epoch_tracker et;
4503 	struct ifaddr *ifa;
4504 	u_int count;
4505 
4506 	MPASS(cb);
4507 
4508 	count = 0;
4509 	NET_EPOCH_ENTER(et);
4510 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4511 		if (ifa->ifa_addr->sa_family != AF_LINK)
4512 			continue;
4513 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4514 		    count);
4515 	}
4516 	NET_EPOCH_EXIT(et);
4517 
4518 	return (count);
4519 }
4520 
4521 u_int
if_llmaddr_count(if_t ifp)4522 if_llmaddr_count(if_t ifp)
4523 {
4524 	struct epoch_tracker et;
4525 	struct ifmultiaddr *ifma;
4526 	int count;
4527 
4528 	count = 0;
4529 	NET_EPOCH_ENTER(et);
4530 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4531 		if (ifma->ifma_addr->sa_family == AF_LINK)
4532 			count++;
4533 	NET_EPOCH_EXIT(et);
4534 
4535 	return (count);
4536 }
4537 
4538 bool
if_maddr_empty(if_t ifp)4539 if_maddr_empty(if_t ifp)
4540 {
4541 
4542 	return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs));
4543 }
4544 
4545 u_int
if_foreach_llmaddr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4546 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4547 {
4548 	struct epoch_tracker et;
4549 	struct ifmultiaddr *ifma;
4550 	u_int count;
4551 
4552 	MPASS(cb);
4553 
4554 	count = 0;
4555 	NET_EPOCH_ENTER(et);
4556 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4557 		if (ifma->ifma_addr->sa_family != AF_LINK)
4558 			continue;
4559 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4560 		    count);
4561 	}
4562 	NET_EPOCH_EXIT(et);
4563 
4564 	return (count);
4565 }
4566 
4567 u_int
if_foreach_addr_type(if_t ifp,int type,if_addr_cb_t cb,void * cb_arg)4568 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg)
4569 {
4570 	struct epoch_tracker et;
4571 	struct ifaddr *ifa;
4572 	u_int count;
4573 
4574 	MPASS(cb);
4575 
4576 	count = 0;
4577 	NET_EPOCH_ENTER(et);
4578 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4579 		if (ifa->ifa_addr->sa_family != type)
4580 			continue;
4581 		count += (*cb)(cb_arg, ifa, count);
4582 	}
4583 	NET_EPOCH_EXIT(et);
4584 
4585 	return (count);
4586 }
4587 
4588 struct ifaddr *
ifa_iter_start(if_t ifp,struct ifa_iter * iter)4589 ifa_iter_start(if_t ifp, struct ifa_iter *iter)
4590 {
4591 	struct ifaddr *ifa;
4592 
4593 	NET_EPOCH_ASSERT();
4594 
4595 	bzero(iter, sizeof(*iter));
4596 	ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
4597 	if (ifa != NULL)
4598 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4599 	else
4600 		iter->context[0] = NULL;
4601 	return (ifa);
4602 }
4603 
4604 struct ifaddr *
ifa_iter_next(struct ifa_iter * iter)4605 ifa_iter_next(struct ifa_iter *iter)
4606 {
4607 	struct ifaddr *ifa = iter->context[0];
4608 
4609 	if (ifa != NULL)
4610 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4611 	return (ifa);
4612 }
4613 
4614 void
ifa_iter_finish(struct ifa_iter * iter)4615 ifa_iter_finish(struct ifa_iter *iter)
4616 {
4617 	/* Nothing to do here for now. */
4618 }
4619 
4620 int
if_setsoftc(if_t ifp,void * softc)4621 if_setsoftc(if_t ifp, void *softc)
4622 {
4623 	ifp->if_softc = softc;
4624 	return (0);
4625 }
4626 
4627 void *
if_getsoftc(const if_t ifp)4628 if_getsoftc(const if_t ifp)
4629 {
4630 	return (ifp->if_softc);
4631 }
4632 
4633 void
if_setrcvif(struct mbuf * m,if_t ifp)4634 if_setrcvif(struct mbuf *m, if_t ifp)
4635 {
4636 
4637 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4638 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4639 }
4640 
4641 void
if_setvtag(struct mbuf * m,uint16_t tag)4642 if_setvtag(struct mbuf *m, uint16_t tag)
4643 {
4644 	m->m_pkthdr.ether_vtag = tag;
4645 }
4646 
4647 uint16_t
if_getvtag(struct mbuf * m)4648 if_getvtag(struct mbuf *m)
4649 {
4650 	return (m->m_pkthdr.ether_vtag);
4651 }
4652 
4653 int
if_sendq_empty(if_t ifp)4654 if_sendq_empty(if_t ifp)
4655 {
4656 	return (IFQ_DRV_IS_EMPTY(&ifp->if_snd));
4657 }
4658 
4659 struct ifaddr *
if_getifaddr(const if_t ifp)4660 if_getifaddr(const if_t ifp)
4661 {
4662 	return (ifp->if_addr);
4663 }
4664 
4665 int
if_setsendqready(if_t ifp)4666 if_setsendqready(if_t ifp)
4667 {
4668 	IFQ_SET_READY(&ifp->if_snd);
4669 	return (0);
4670 }
4671 
4672 int
if_setsendqlen(if_t ifp,int tx_desc_count)4673 if_setsendqlen(if_t ifp, int tx_desc_count)
4674 {
4675 	IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count);
4676 	ifp->if_snd.ifq_drv_maxlen = tx_desc_count;
4677 	return (0);
4678 }
4679 
4680 void
if_setnetmapadapter(if_t ifp,struct netmap_adapter * na)4681 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na)
4682 {
4683 	ifp->if_netmap = na;
4684 }
4685 
4686 struct netmap_adapter *
if_getnetmapadapter(if_t ifp)4687 if_getnetmapadapter(if_t ifp)
4688 {
4689 	return (ifp->if_netmap);
4690 }
4691 
4692 int
if_vlantrunkinuse(if_t ifp)4693 if_vlantrunkinuse(if_t ifp)
4694 {
4695 	return (ifp->if_vlantrunk != NULL);
4696 }
4697 
4698 void
if_init(if_t ifp,void * ctx)4699 if_init(if_t ifp, void *ctx)
4700 {
4701 	(*ifp->if_init)(ctx);
4702 }
4703 
4704 void
if_input(if_t ifp,struct mbuf * sendmp)4705 if_input(if_t ifp, struct mbuf* sendmp)
4706 {
4707 	(*ifp->if_input)(ifp, sendmp);
4708 }
4709 
4710 int
if_transmit(if_t ifp,struct mbuf * m)4711 if_transmit(if_t ifp, struct mbuf *m)
4712 {
4713 	return ((*ifp->if_transmit)(ifp, m));
4714 }
4715 
4716 int
if_resolvemulti(if_t ifp,struct sockaddr ** srcs,struct sockaddr * dst)4717 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst)
4718 {
4719 	if (ifp->if_resolvemulti == NULL)
4720 		return (EOPNOTSUPP);
4721 
4722 	return (ifp->if_resolvemulti(ifp, srcs, dst));
4723 }
4724 
4725 int
if_ioctl(if_t ifp,u_long cmd,void * data)4726 if_ioctl(if_t ifp, u_long cmd, void *data)
4727 {
4728 	if (ifp->if_ioctl == NULL)
4729 		return (EOPNOTSUPP);
4730 
4731 	return (ifp->if_ioctl(ifp, cmd, data));
4732 }
4733 
4734 struct mbuf *
if_dequeue(if_t ifp)4735 if_dequeue(if_t ifp)
4736 {
4737 	struct mbuf *m;
4738 
4739 	IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
4740 	return (m);
4741 }
4742 
4743 int
if_sendq_prepend(if_t ifp,struct mbuf * m)4744 if_sendq_prepend(if_t ifp, struct mbuf *m)
4745 {
4746 	IFQ_DRV_PREPEND(&ifp->if_snd, m);
4747 	return (0);
4748 }
4749 
4750 int
if_setifheaderlen(if_t ifp,int len)4751 if_setifheaderlen(if_t ifp, int len)
4752 {
4753 	ifp->if_hdrlen = len;
4754 	return (0);
4755 }
4756 
4757 char *
if_getlladdr(const if_t ifp)4758 if_getlladdr(const if_t ifp)
4759 {
4760 	return (IF_LLADDR(ifp));
4761 }
4762 
4763 void *
if_gethandle(u_char type)4764 if_gethandle(u_char type)
4765 {
4766 	return (if_alloc(type));
4767 }
4768 
4769 void
if_vlancap(if_t ifp)4770 if_vlancap(if_t ifp)
4771 {
4772 	VLAN_CAPABILITIES(ifp);
4773 }
4774 
4775 int
if_sethwtsomax(if_t ifp,u_int if_hw_tsomax)4776 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4777 {
4778 	ifp->if_hw_tsomax = if_hw_tsomax;
4779         return (0);
4780 }
4781 
4782 int
if_sethwtsomaxsegcount(if_t ifp,u_int if_hw_tsomaxsegcount)4783 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4784 {
4785 	ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4786         return (0);
4787 }
4788 
4789 int
if_sethwtsomaxsegsize(if_t ifp,u_int if_hw_tsomaxsegsize)4790 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4791 {
4792 	ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4793         return (0);
4794 }
4795 
4796 u_int
if_gethwtsomax(const if_t ifp)4797 if_gethwtsomax(const if_t ifp)
4798 {
4799 	return (ifp->if_hw_tsomax);
4800 }
4801 
4802 u_int
if_gethwtsomaxsegcount(const if_t ifp)4803 if_gethwtsomaxsegcount(const if_t ifp)
4804 {
4805 	return (ifp->if_hw_tsomaxsegcount);
4806 }
4807 
4808 u_int
if_gethwtsomaxsegsize(const if_t ifp)4809 if_gethwtsomaxsegsize(const if_t ifp)
4810 {
4811 	return (ifp->if_hw_tsomaxsegsize);
4812 }
4813 
4814 void
if_setinitfn(if_t ifp,if_init_fn_t init_fn)4815 if_setinitfn(if_t ifp, if_init_fn_t init_fn)
4816 {
4817 	ifp->if_init = init_fn;
4818 }
4819 
4820 void
if_setinputfn(if_t ifp,if_input_fn_t input_fn)4821 if_setinputfn(if_t ifp, if_input_fn_t input_fn)
4822 {
4823 	ifp->if_input = input_fn;
4824 }
4825 
4826 if_input_fn_t
if_getinputfn(if_t ifp)4827 if_getinputfn(if_t ifp)
4828 {
4829 	return (ifp->if_input);
4830 }
4831 
4832 void
if_setioctlfn(if_t ifp,if_ioctl_fn_t ioctl_fn)4833 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn)
4834 {
4835 	ifp->if_ioctl = ioctl_fn;
4836 }
4837 
4838 void
if_setoutputfn(if_t ifp,if_output_fn_t output_fn)4839 if_setoutputfn(if_t ifp, if_output_fn_t output_fn)
4840 {
4841 	ifp->if_output = output_fn;
4842 }
4843 
4844 void
if_setstartfn(if_t ifp,if_start_fn_t start_fn)4845 if_setstartfn(if_t ifp, if_start_fn_t start_fn)
4846 {
4847 	ifp->if_start = start_fn;
4848 }
4849 
4850 if_start_fn_t
if_getstartfn(if_t ifp)4851 if_getstartfn(if_t ifp)
4852 {
4853 	return (ifp->if_start);
4854 }
4855 
4856 void
if_settransmitfn(if_t ifp,if_transmit_fn_t start_fn)4857 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4858 {
4859 	ifp->if_transmit = start_fn;
4860 }
4861 
4862 if_transmit_fn_t
if_gettransmitfn(if_t ifp)4863 if_gettransmitfn(if_t ifp)
4864 {
4865 	return (ifp->if_transmit);
4866 }
4867 
4868 void
if_setqflushfn(if_t ifp,if_qflush_fn_t flush_fn)4869 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4870 {
4871 	ifp->if_qflush = flush_fn;
4872 }
4873 
4874 void
if_setsndtagallocfn(if_t ifp,if_snd_tag_alloc_t alloc_fn)4875 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn)
4876 {
4877 	ifp->if_snd_tag_alloc = alloc_fn;
4878 }
4879 
4880 int
if_snd_tag_alloc(if_t ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** mstp)4881 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
4882     struct m_snd_tag **mstp)
4883 {
4884 	if (ifp->if_snd_tag_alloc == NULL)
4885 		return (EOPNOTSUPP);
4886 	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
4887 }
4888 
4889 void
if_setgetcounterfn(if_t ifp,if_get_counter_t fn)4890 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4891 {
4892 	ifp->if_get_counter = fn;
4893 }
4894 
4895 void
if_setreassignfn(if_t ifp,if_reassign_fn_t fn)4896 if_setreassignfn(if_t ifp, if_reassign_fn_t fn)
4897 {
4898 	ifp->if_reassign = fn;
4899 }
4900 
4901 void
if_setratelimitqueryfn(if_t ifp,if_ratelimit_query_t fn)4902 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn)
4903 {
4904 	ifp->if_ratelimit_query = fn;
4905 }
4906 
4907 void
if_setdebugnet_methods(if_t ifp,struct debugnet_methods * m)4908 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m)
4909 {
4910 	ifp->if_debugnet_methods = m;
4911 }
4912 
4913 struct label *
if_getmaclabel(if_t ifp)4914 if_getmaclabel(if_t ifp)
4915 {
4916 	return (ifp->if_label);
4917 }
4918 
4919 void
if_setmaclabel(if_t ifp,struct label * label)4920 if_setmaclabel(if_t ifp, struct label *label)
4921 {
4922 	ifp->if_label = label;
4923 }
4924 
4925 int
if_gettype(if_t ifp)4926 if_gettype(if_t ifp)
4927 {
4928 	return (ifp->if_type);
4929 }
4930 
4931 void *
if_getllsoftc(if_t ifp)4932 if_getllsoftc(if_t ifp)
4933 {
4934 	return (ifp->if_llsoftc);
4935 }
4936 
4937 void
if_setllsoftc(if_t ifp,void * llsoftc)4938 if_setllsoftc(if_t ifp, void *llsoftc)
4939 {
4940 	ifp->if_llsoftc = llsoftc;
4941 };
4942 
4943 int
if_getlinkstate(if_t ifp)4944 if_getlinkstate(if_t ifp)
4945 {
4946 	return (ifp->if_link_state);
4947 }
4948 
4949 const uint8_t *
if_getbroadcastaddr(if_t ifp)4950 if_getbroadcastaddr(if_t ifp)
4951 {
4952 	return (ifp->if_broadcastaddr);
4953 }
4954 
4955 void
if_setbroadcastaddr(if_t ifp,const uint8_t * addr)4956 if_setbroadcastaddr(if_t ifp, const uint8_t *addr)
4957 {
4958 	ifp->if_broadcastaddr = addr;
4959 }
4960 
4961 int
if_getnumadomain(if_t ifp)4962 if_getnumadomain(if_t ifp)
4963 {
4964 	return (ifp->if_numa_domain);
4965 }
4966 
4967 uint64_t
if_getcounter(if_t ifp,ift_counter counter)4968 if_getcounter(if_t ifp, ift_counter counter)
4969 {
4970 	return (ifp->if_get_counter(ifp, counter));
4971 }
4972 
4973 bool
if_altq_is_enabled(if_t ifp)4974 if_altq_is_enabled(if_t ifp)
4975 {
4976 	return (ALTQ_IS_ENABLED(&ifp->if_snd));
4977 }
4978 
4979 struct vnet *
if_getvnet(if_t ifp)4980 if_getvnet(if_t ifp)
4981 {
4982 	return (ifp->if_vnet);
4983 }
4984 
4985 struct in_ifinfo *
if_getinet(if_t ifp)4986 if_getinet(if_t ifp)
4987 {
4988 	return (ifp->if_inet);
4989 }
4990 
4991 struct in6_ifextra *
if_getinet6(if_t ifp)4992 if_getinet6(if_t ifp)
4993 {
4994 	return (ifp->if_inet6);
4995 }
4996 
4997 u_int
if_getfib(if_t ifp)4998 if_getfib(if_t ifp)
4999 {
5000 	return (ifp->if_fib);
5001 }
5002 
5003 uint8_t
if_getaddrlen(if_t ifp)5004 if_getaddrlen(if_t ifp)
5005 {
5006 	return (ifp->if_addrlen);
5007 }
5008 
5009 struct bpf_if *
if_getbpf(if_t ifp)5010 if_getbpf(if_t ifp)
5011 {
5012 	return (ifp->if_bpf);
5013 }
5014 
5015 struct ifvlantrunk *
if_getvlantrunk(if_t ifp)5016 if_getvlantrunk(if_t ifp)
5017 {
5018 	return (ifp->if_vlantrunk);
5019 }
5020 
5021 uint8_t
if_getpcp(if_t ifp)5022 if_getpcp(if_t ifp)
5023 {
5024 	return (ifp->if_pcp);
5025 }
5026 
5027 void *
if_getl2com(if_t ifp)5028 if_getl2com(if_t ifp)
5029 {
5030 	return (ifp->if_l2com);
5031 }
5032 
5033 void
if_setipsec_accel_methods(if_t ifp,const struct if_ipsec_accel_methods * m)5034 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m)
5035 {
5036 	ifp->if_ipsec_accel_m = m;
5037 }
5038 
5039 #ifdef DDB
5040 static void
if_show_ifnet(struct ifnet * ifp)5041 if_show_ifnet(struct ifnet *ifp)
5042 {
5043 	if (ifp == NULL)
5044 		return;
5045 	db_printf("%s:\n", ifp->if_xname);
5046 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
5047 	IF_DB_PRINTF("%s", if_dname);
5048 	IF_DB_PRINTF("%d", if_dunit);
5049 	IF_DB_PRINTF("%s", if_description);
5050 	IF_DB_PRINTF("%u", if_index);
5051 	IF_DB_PRINTF("%d", if_idxgen);
5052 	IF_DB_PRINTF("%u", if_refcount);
5053 	IF_DB_PRINTF("%p", if_softc);
5054 	IF_DB_PRINTF("%p", if_l2com);
5055 	IF_DB_PRINTF("%p", if_llsoftc);
5056 	IF_DB_PRINTF("%d", if_amcount);
5057 	IF_DB_PRINTF("%p", if_addr);
5058 	IF_DB_PRINTF("%p", if_broadcastaddr);
5059 	IF_DB_PRINTF("%u", if_fib);
5060 	IF_DB_PRINTF("%p", if_vnet);
5061 	IF_DB_PRINTF("%p", if_home_vnet);
5062 	IF_DB_PRINTF("%p", if_vlantrunk);
5063 	IF_DB_PRINTF("%p", if_bpf);
5064 	IF_DB_PRINTF("%u", if_pcount);
5065 	IF_DB_PRINTF("%p", if_bridge);
5066 	IF_DB_PRINTF("%p", if_lagg);
5067 	IF_DB_PRINTF("%p", if_pf_kif);
5068 	IF_DB_PRINTF("%p", if_carp);
5069 	IF_DB_PRINTF("%p", if_label);
5070 	IF_DB_PRINTF("%p", if_netmap);
5071 	IF_DB_PRINTF("0x%08x", if_flags);
5072 	IF_DB_PRINTF("0x%08x", if_drv_flags);
5073 	IF_DB_PRINTF("0x%08x", if_capabilities);
5074 	IF_DB_PRINTF("0x%08x", if_capenable);
5075 	IF_DB_PRINTF("%p", if_snd.ifq_head);
5076 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
5077 	IF_DB_PRINTF("%d", if_snd.ifq_len);
5078 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
5079 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
5080 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
5081 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
5082 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
5083 	IF_DB_PRINTF("%d", if_snd.altq_type);
5084 	IF_DB_PRINTF("%x", if_snd.altq_flags);
5085 #undef IF_DB_PRINTF
5086 }
5087 
DB_SHOW_COMMAND(ifnet,db_show_ifnet)5088 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
5089 {
5090 	if (!have_addr) {
5091 		db_printf("usage: show ifnet <struct ifnet *>\n");
5092 		return;
5093 	}
5094 
5095 	if_show_ifnet((struct ifnet *)addr);
5096 }
5097 
DB_SHOW_ALL_COMMAND(ifnets,db_show_all_ifnets)5098 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
5099 {
5100 	struct ifnet *ifp;
5101 	u_short idx;
5102 
5103 	for (idx = 1; idx <= if_index; idx++) {
5104 		ifp = ifindex_table[idx].ife_ifnet;
5105 		if (ifp == NULL)
5106 			continue;
5107 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
5108 		if (db_pager_quit)
5109 			break;
5110 	}
5111 }
5112 #endif	/* DDB */
5113