xref: /freebsd/sys/net/if.c (revision ee9456ce37539da5b651945eea18502f290eb133)
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 
432 	IFNET_WLOCK();
433 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
434 #ifdef VIMAGE
435 	curvnet->vnet_ifcnt++;
436 #endif
437 	IFNET_WUNLOCK();
438 }
439 
440 static bool
if_unlink_ifnet(struct ifnet * ifp,bool vmove)441 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
442 {
443 	struct ifnet *iter;
444 	int found = 0;
445 
446 	IFNET_WLOCK();
447 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
448 		if (iter == ifp) {
449 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
450 #ifdef VIMAGE
451 			curvnet->vnet_ifcnt--;
452 #endif
453 			if (!vmove)
454 				ifp->if_flags |= IFF_DYING;
455 			found = 1;
456 			break;
457 		}
458 	IFNET_WUNLOCK();
459 
460 	return (found);
461 }
462 
463 #ifdef VIMAGE
464 static void
vnet_if_return(const void * unused __unused)465 vnet_if_return(const void *unused __unused)
466 {
467 	struct ifnet *ifp, *nifp;
468 	struct ifnet **pending;
469 	int found __diagused;
470 	int i;
471 
472 	i = 0;
473 
474 	/*
475 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
476 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
477 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
478 	 * complete. We can't do that from within NET_EPOCH.
479 	 *
480 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
481 	 * read/write lock. We cannot hold the lock as we call if_vmove()
482 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
483 	 * ctx lock.
484 	 */
485 	IFNET_WLOCK();
486 
487 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
488 	    M_IFNET, M_WAITOK | M_ZERO);
489 
490 	/* Return all inherited interfaces to their parent vnets. */
491 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
492 		if (ifp->if_home_vnet != ifp->if_vnet) {
493 			found = if_unlink_ifnet(ifp, true);
494 			MPASS(found);
495 
496 			pending[i++] = ifp;
497 		}
498 	}
499 	IFNET_WUNLOCK();
500 
501 	for (int j = 0; j < i; j++) {
502 		sx_xlock(&ifnet_detach_sxlock);
503 		if_vmove(pending[j], pending[j]->if_home_vnet);
504 		sx_xunlock(&ifnet_detach_sxlock);
505 	}
506 
507 	free(pending, M_IFNET);
508 }
509 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
510     vnet_if_return, NULL);
511 #endif
512 
513 /*
514  * Allocate a struct ifnet and an index for an interface.  A layer 2
515  * common structure will also be allocated if an allocation routine is
516  * registered for the passed type.
517  */
518 static struct ifnet *
if_alloc_domain(u_char type,int numa_domain)519 if_alloc_domain(u_char type, int numa_domain)
520 {
521 	struct ifnet *ifp;
522 	u_short idx;
523 
524 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
525 	if (numa_domain == IF_NODOM)
526 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
527 		    M_WAITOK | M_ZERO);
528 	else
529 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
530 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
531 	ifp->if_type = type;
532 	ifp->if_alloctype = type;
533 	ifp->if_numa_domain = numa_domain;
534 #ifdef VIMAGE
535 	ifp->if_vnet = curvnet;
536 #endif
537 	if (if_com_alloc[type] != NULL) {
538 		ifp->if_l2com = if_com_alloc[type](type, ifp);
539 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
540 		    type));
541 	}
542 
543 	IF_ADDR_LOCK_INIT(ifp);
544 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
545 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
546 	CK_STAILQ_INIT(&ifp->if_addrhead);
547 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
548 	CK_STAILQ_INIT(&ifp->if_groups);
549 #ifdef MAC
550 	mac_ifnet_init(ifp);
551 #endif
552 	ifq_init(&ifp->if_snd, ifp);
553 
554 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
555 	for (int i = 0; i < IFCOUNTERS; i++)
556 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
557 	ifp->if_get_counter = if_get_counter_default;
558 	ifp->if_pcp = IFNET_PCP_NONE;
559 
560 	/* Allocate an ifindex array entry. */
561 	IFNET_WLOCK();
562 	/*
563 	 * Try to find an empty slot below if_index.  If we fail, take the
564 	 * next slot.
565 	 */
566 	for (idx = 1; idx <= if_index; idx++) {
567 		if (ifindex_table[idx].ife_ifnet == NULL)
568 			break;
569 	}
570 
571 	/* Catch if_index overflow. */
572 	if (idx >= if_indexlim) {
573 		struct ifindex_entry *new, *old;
574 		int newlim;
575 
576 		newlim = if_indexlim * 2;
577 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
578 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
579 		old = ifindex_table;
580 		ck_pr_store_ptr(&ifindex_table, new);
581 		if_indexlim = newlim;
582 		NET_EPOCH_WAIT();
583 		free(old, M_IFNET);
584 	}
585 	if (idx > if_index)
586 		if_index = idx;
587 
588 	ifp->if_index = idx;
589 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
590 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
591 	IFNET_WUNLOCK();
592 
593 	return (ifp);
594 }
595 
596 struct ifnet *
if_alloc_dev(u_char type,device_t dev)597 if_alloc_dev(u_char type, device_t dev)
598 {
599 	int numa_domain;
600 
601 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
602 		return (if_alloc_domain(type, IF_NODOM));
603 	return (if_alloc_domain(type, numa_domain));
604 }
605 
606 struct ifnet *
if_alloc(u_char type)607 if_alloc(u_char type)
608 {
609 
610 	return (if_alloc_domain(type, IF_NODOM));
611 }
612 /*
613  * Do the actual work of freeing a struct ifnet, and layer 2 common
614  * structure.  This call is made when the network epoch guarantees
615  * us that nobody holds a pointer to the interface.
616  */
617 static void
if_free_deferred(epoch_context_t ctx)618 if_free_deferred(epoch_context_t ctx)
619 {
620 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
621 
622 	KASSERT((ifp->if_flags & IFF_DYING),
623 	    ("%s: interface not dying", __func__));
624 
625 	if (if_com_free[ifp->if_alloctype] != NULL)
626 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
627 		    ifp->if_alloctype);
628 
629 #ifdef MAC
630 	mac_ifnet_destroy(ifp);
631 #endif /* MAC */
632 	IF_ADDR_LOCK_DESTROY(ifp);
633 	ifq_delete(&ifp->if_snd);
634 
635 	for (int i = 0; i < IFCOUNTERS; i++)
636 		counter_u64_free(ifp->if_counters[i]);
637 
638 	if_freedescr(ifp->if_description);
639 	free(ifp->if_hw_addr, M_IFADDR);
640 	free(ifp, M_IFNET);
641 }
642 
643 /*
644  * Deregister an interface and free the associated storage.
645  */
646 void
if_free(struct ifnet * ifp)647 if_free(struct ifnet *ifp)
648 {
649 
650 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
651 
652 	/*
653 	 * XXXGL: An interface index is really an alias to ifp pointer.
654 	 * Why would we clear the alias now, and not in the deferred
655 	 * context?  Indeed there is nothing wrong with some network
656 	 * thread obtaining ifp via ifnet_byindex() inside the network
657 	 * epoch and then dereferencing ifp while we perform if_free(),
658 	 * and after if_free() finished, too.
659 	 *
660 	 * This early index freeing was important back when ifindex was
661 	 * virtualized and interface would outlive the vnet.
662 	 */
663 	IFNET_WLOCK();
664 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
665 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
666 	ifindex_table[ifp->if_index].ife_gencnt++;
667 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
668 		if_index--;
669 	IFNET_WUNLOCK();
670 
671 	if (refcount_release(&ifp->if_refcount))
672 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
673 }
674 
675 /*
676  * Interfaces to keep an ifnet type-stable despite the possibility of the
677  * driver calling if_free().  If there are additional references, we defer
678  * freeing the underlying data structure.
679  */
680 void
if_ref(struct ifnet * ifp)681 if_ref(struct ifnet *ifp)
682 {
683 	u_int old __diagused;
684 
685 	/* We don't assert the ifnet list lock here, but arguably should. */
686 	old = refcount_acquire(&ifp->if_refcount);
687 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
688 }
689 
690 bool
if_try_ref(struct ifnet * ifp)691 if_try_ref(struct ifnet *ifp)
692 {
693 	NET_EPOCH_ASSERT();
694 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
695 }
696 
697 void
if_rele(struct ifnet * ifp)698 if_rele(struct ifnet *ifp)
699 {
700 
701 	if (!refcount_release(&ifp->if_refcount))
702 		return;
703 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
704 }
705 
706 void
ifq_init(struct ifaltq * ifq,struct ifnet * ifp)707 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
708 {
709 
710 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
711 
712 	if (ifq->ifq_maxlen == 0)
713 		ifq->ifq_maxlen = ifqmaxlen;
714 
715 	ifq->altq_type = 0;
716 	ifq->altq_disc = NULL;
717 	ifq->altq_flags &= ALTQF_CANTCHANGE;
718 	ifq->altq_tbr  = NULL;
719 	ifq->altq_ifp  = ifp;
720 }
721 
722 void
ifq_delete(struct ifaltq * ifq)723 ifq_delete(struct ifaltq *ifq)
724 {
725 	mtx_destroy(&ifq->ifq_mtx);
726 }
727 
728 /*
729  * Perform generic interface initialization tasks and attach the interface
730  * to the list of "active" interfaces.  If vmove flag is set on entry
731  * to if_attach_internal(), perform only a limited subset of initialization
732  * tasks, given that we are moving from one vnet to another an ifnet which
733  * has already been fully initialized.
734  *
735  * Note that if_detach_internal() removes group membership unconditionally
736  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
737  * Thus, when if_vmove() is applied to a cloned interface, group membership
738  * is lost while a cloned one always joins a group whose name is
739  * ifc->ifc_name.  To recover this after if_detach_internal() and
740  * if_attach_internal(), the cloner should be specified to
741  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
742  * attempts to join a group whose name is ifc->ifc_name.
743  *
744  * XXX:
745  *  - The decision to return void and thus require this function to
746  *    succeed is questionable.
747  *  - We should probably do more sanity checking.  For instance we don't
748  *    do anything to insure if_xname is unique or non-empty.
749  */
750 void
if_attach(struct ifnet * ifp)751 if_attach(struct ifnet *ifp)
752 {
753 
754 	if_attach_internal(ifp, false);
755 }
756 
757 /*
758  * Compute the least common TSO limit.
759  */
760 void
if_hw_tsomax_common(if_t ifp,struct ifnet_hw_tsomax * pmax)761 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
762 {
763 	/*
764 	 * 1) If there is no limit currently, take the limit from
765 	 * the network adapter.
766 	 *
767 	 * 2) If the network adapter has a limit below the current
768 	 * limit, apply it.
769 	 */
770 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
771 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
772 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
773 	}
774 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
775 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
776 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
777 	}
778 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
779 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
780 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
781 	}
782 }
783 
784 /*
785  * Update TSO limit of a network adapter.
786  *
787  * Returns zero if no change. Else non-zero.
788  */
789 int
if_hw_tsomax_update(if_t ifp,struct ifnet_hw_tsomax * pmax)790 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
791 {
792 	int retval = 0;
793 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
794 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
795 		retval++;
796 	}
797 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
798 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
799 		retval++;
800 	}
801 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
802 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
803 		retval++;
804 	}
805 	return (retval);
806 }
807 
808 static void
if_attach_internal(struct ifnet * ifp,bool vmove)809 if_attach_internal(struct ifnet *ifp, bool vmove)
810 {
811 	unsigned socksize, ifasize;
812 	int namelen, masklen;
813 	struct sockaddr_dl *sdl;
814 	struct ifaddr *ifa;
815 
816 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
817 
818 #ifdef VIMAGE
819 	CURVNET_ASSERT_SET();
820 	ifp->if_vnet = curvnet;
821 	if (ifp->if_home_vnet == NULL)
822 		ifp->if_home_vnet = curvnet;
823 #endif
824 
825 	if_addgroup(ifp, IFG_ALL);
826 
827 #ifdef VIMAGE
828 	/* Restore group membership for cloned interface. */
829 	if (vmove)
830 		if_clone_restoregroup(ifp);
831 #endif
832 
833 	getmicrotime(&ifp->if_lastchange);
834 	ifp->if_epoch = time_uptime;
835 
836 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
837 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
838 	    ("transmit and qflush must both either be set or both be NULL"));
839 	if (ifp->if_transmit == NULL) {
840 		ifp->if_transmit = if_transmit_default;
841 		ifp->if_qflush = if_qflush;
842 	}
843 	if (ifp->if_input == NULL)
844 		ifp->if_input = if_input_default;
845 
846 	if (ifp->if_requestencap == NULL)
847 		ifp->if_requestencap = if_requestencap_default;
848 
849 	if (!vmove) {
850 #ifdef MAC
851 		mac_ifnet_create(ifp);
852 #endif
853 
854 		/*
855 		 * Create a Link Level name for this device.
856 		 */
857 		namelen = strlen(ifp->if_xname);
858 		/*
859 		 * Always save enough space for any possible name so we
860 		 * can do a rename in place later.
861 		 */
862 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
863 		socksize = masklen + ifp->if_addrlen;
864 		if (socksize < sizeof(*sdl))
865 			socksize = sizeof(*sdl);
866 		socksize = roundup2(socksize, sizeof(long));
867 		ifasize = sizeof(*ifa) + 2 * socksize;
868 		ifa = ifa_alloc(ifasize, M_WAITOK);
869 		sdl = (struct sockaddr_dl *)(ifa + 1);
870 		sdl->sdl_len = socksize;
871 		sdl->sdl_family = AF_LINK;
872 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
873 		sdl->sdl_nlen = namelen;
874 		sdl->sdl_index = ifp->if_index;
875 		sdl->sdl_type = ifp->if_type;
876 		ifp->if_addr = ifa;
877 		ifa->ifa_ifp = ifp;
878 		ifa->ifa_addr = (struct sockaddr *)sdl;
879 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
880 		ifa->ifa_netmask = (struct sockaddr *)sdl;
881 		sdl->sdl_len = masklen;
882 		while (namelen != 0)
883 			sdl->sdl_data[--namelen] = 0xff;
884 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
885 		/* Reliably crash if used uninitialized. */
886 		ifp->if_broadcastaddr = NULL;
887 
888 		if (ifp->if_type == IFT_ETHER) {
889 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
890 			    M_WAITOK | M_ZERO);
891 		}
892 
893 #if defined(INET) || defined(INET6)
894 		/* Use defaults for TSO, if nothing is set */
895 		if (ifp->if_hw_tsomax == 0 &&
896 		    ifp->if_hw_tsomaxsegcount == 0 &&
897 		    ifp->if_hw_tsomaxsegsize == 0) {
898 			/*
899 			 * The TSO defaults needs to be such that an
900 			 * NFS mbuf list of 35 mbufs totalling just
901 			 * below 64K works and that a chain of mbufs
902 			 * can be defragged into at most 32 segments:
903 			 */
904 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
905 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
906 			ifp->if_hw_tsomaxsegcount = 35;
907 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
908 
909 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
910 			if (ifp->if_capabilities & IFCAP_TSO) {
911 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
912 				    ifp->if_hw_tsomax,
913 				    ifp->if_hw_tsomaxsegcount,
914 				    ifp->if_hw_tsomaxsegsize);
915 			}
916 		}
917 #endif
918 	}
919 
920 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
921 	if_link_ifnet(ifp);
922 	EVENTHANDLER_INVOKE(ifnet_attached_event, ifp);
923 	if (IS_DEFAULT_VNET(curvnet))
924 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
925 }
926 
927 static void
if_epochalloc(void * dummy __unused)928 if_epochalloc(void *dummy __unused)
929 {
930 
931 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
932 }
933 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
934 
935 /*
936  * Remove any unicast or broadcast network addresses from an interface.
937  */
938 void
if_purgeaddrs(struct ifnet * ifp)939 if_purgeaddrs(struct ifnet *ifp)
940 {
941 	struct ifaddr *ifa;
942 
943 #ifdef INET6
944 	/*
945 	 * Need to leave multicast addresses of proxy NDP llentries
946 	 * before in6_purgeifaddr() because the llentries are keys
947 	 * for in6_multi objects of proxy NDP entries.
948 	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
949 	 * then we would lose the keys if they are called earlier.
950 	 */
951 	in6_purge_proxy_ndp(ifp);
952 #endif
953 	while (1) {
954 		struct epoch_tracker et;
955 
956 		NET_EPOCH_ENTER(et);
957 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
958 			if (ifa->ifa_addr->sa_family != AF_LINK)
959 				break;
960 		}
961 		NET_EPOCH_EXIT(et);
962 
963 		if (ifa == NULL)
964 			break;
965 #ifdef INET
966 		/* XXX: Ugly!! ad hoc just for INET */
967 		if (ifa->ifa_addr->sa_family == AF_INET) {
968 			struct ifreq ifr;
969 
970 			bzero(&ifr, sizeof(ifr));
971 			ifr.ifr_addr = *ifa->ifa_addr;
972 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
973 			    NULL) == 0)
974 				continue;
975 		}
976 #endif /* INET */
977 #ifdef INET6
978 		if (ifa->ifa_addr->sa_family == AF_INET6) {
979 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
980 			/* ifp_addrhead is already updated */
981 			continue;
982 		}
983 #endif /* INET6 */
984 		IF_ADDR_WLOCK(ifp);
985 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
986 		IF_ADDR_WUNLOCK(ifp);
987 		ifa_free(ifa);
988 	}
989 }
990 
991 /*
992  * Remove any multicast network addresses from an interface when an ifnet
993  * is going away.
994  */
995 static void
if_purgemaddrs(struct ifnet * ifp)996 if_purgemaddrs(struct ifnet *ifp)
997 {
998 	struct ifmultiaddr *ifma;
999 
1000 	IF_ADDR_WLOCK(ifp);
1001 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1002 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1003 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1004 		if_delmulti_locked(ifp, ifma, 1);
1005 	}
1006 	IF_ADDR_WUNLOCK(ifp);
1007 }
1008 
1009 /*
1010  * Detach an interface, removing it from the list of "active" interfaces.
1011  * If vmove flag is set on entry to if_detach_internal(), perform only a
1012  * limited subset of cleanup tasks, given that we are moving an ifnet from
1013  * one vnet to another, where it must be fully operational.
1014  *
1015  * XXXRW: There are some significant questions about event ordering, and
1016  * how to prevent things from starting to use the interface during detach.
1017  */
1018 void
if_detach(struct ifnet * ifp)1019 if_detach(struct ifnet *ifp)
1020 {
1021 	bool found;
1022 
1023 	CURVNET_SET_QUIET(ifp->if_vnet);
1024 	found = if_unlink_ifnet(ifp, false);
1025 	if (found) {
1026 		sx_xlock(&ifnet_detach_sxlock);
1027 		if_detach_internal(ifp, false);
1028 		sx_xunlock(&ifnet_detach_sxlock);
1029 	}
1030 	CURVNET_RESTORE();
1031 }
1032 
1033 /*
1034  * The vmove flag, if set, indicates that we are called from a callpath
1035  * that is moving an interface to a different vnet instance.
1036  *
1037  * The shutdown flag, if set, indicates that we are called in the
1038  * process of shutting down a vnet instance.  Currently only the
1039  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1040  * on a vnet instance shutdown without this flag being set, e.g., when
1041  * the cloned interfaces are destoyed as first thing of teardown.
1042  */
1043 static void
if_detach_internal(struct ifnet * ifp,bool vmove)1044 if_detach_internal(struct ifnet *ifp, bool vmove)
1045 {
1046 	struct ifaddr *ifa;
1047 #ifdef VIMAGE
1048 	bool shutdown;
1049 
1050 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1051 #endif
1052 
1053 	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
1054 
1055 	/*
1056 	 * At this point we know the interface still was on the ifnet list
1057 	 * and we removed it so we are in a stable state.
1058 	 */
1059 	NET_EPOCH_WAIT();
1060 
1061 	/*
1062 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1063 	 * fixes issues about late destruction of multicast options
1064 	 * which lead to leave group calls, which in turn access the
1065 	 * belonging ifnet structure:
1066 	 */
1067 	NET_EPOCH_DRAIN_CALLBACKS();
1068 
1069 	/*
1070 	 * In any case (destroy or vmove) detach us from the groups
1071 	 * and remove/wait for pending events on the taskq.
1072 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1073 	 */
1074 	if_delgroups(ifp);
1075 
1076 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1077 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1078 
1079 	if_down(ifp);
1080 
1081 #ifdef VIMAGE
1082 	/*
1083 	 * On VNET shutdown abort here as the stack teardown will do all
1084 	 * the work top-down for us.
1085 	 */
1086 	if (shutdown) {
1087 		/* Give interface users the chance to clean up. */
1088 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1089 
1090 		/*
1091 		 * In case of a vmove we are done here without error.
1092 		 * If we would signal an error it would lead to the same
1093 		 * abort as if we did not find the ifnet anymore.
1094 		 * if_detach() calls us in void context and does not care
1095 		 * about an early abort notification, so life is splendid :)
1096 		 */
1097 		return;
1098 	}
1099 #endif
1100 
1101 	/*
1102 	 * At this point we are not tearing down a VNET and are either
1103 	 * going to destroy or vmove the interface and have to cleanup
1104 	 * accordingly.
1105 	 */
1106 
1107 	/*
1108 	 * Remove routes and flush queues.
1109 	 */
1110 #ifdef ALTQ
1111 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1112 		altq_disable(&ifp->if_snd);
1113 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1114 		altq_detach(&ifp->if_snd);
1115 #endif
1116 
1117 	rt_flushifroutes(ifp);
1118 
1119 	if_purgeaddrs(ifp);
1120 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1121 	if_purgemaddrs(ifp);
1122 	if (IS_DEFAULT_VNET(curvnet))
1123 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1124 
1125 	if (!vmove) {
1126 		/*
1127 		 * Prevent further calls into the device driver via ifnet.
1128 		 */
1129 		if_dead(ifp);
1130 
1131 		/*
1132 		 * Clean up all addresses.
1133 		 */
1134 		IF_ADDR_WLOCK(ifp);
1135 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1136 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1137 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1138 			IF_ADDR_WUNLOCK(ifp);
1139 			ifa_free(ifa);
1140 		} else
1141 			IF_ADDR_WUNLOCK(ifp);
1142 	}
1143 }
1144 
1145 #ifdef VIMAGE
1146 /*
1147  * if_vmove() performs a limited version of if_detach() in current
1148  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1149  */
1150 static void
if_vmove(struct ifnet * ifp,struct vnet * new_vnet)1151 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1152 {
1153 #ifdef DEV_BPF
1154 	/*
1155 	 * Detach BPF file descriptors from its interface.
1156 	 */
1157 	bpf_ifdetach(ifp);
1158 #endif
1159 
1160 	/*
1161 	 * Detach from current vnet, but preserve LLADDR info, do not
1162 	 * mark as dead etc. so that the ifnet can be reattached later.
1163 	 */
1164 	if_detach_internal(ifp, true);
1165 
1166 	/*
1167 	 * Perform interface-specific reassignment tasks, if provided by
1168 	 * the driver.
1169 	 */
1170 	if (ifp->if_reassign != NULL)
1171 		ifp->if_reassign(ifp, new_vnet, NULL);
1172 
1173 	/*
1174 	 * Switch to the context of the target vnet.
1175 	 */
1176 	CURVNET_SET_QUIET(new_vnet);
1177 	if_attach_internal(ifp, true);
1178 	bpf_vmove(ifp->if_bpf);
1179 	CURVNET_RESTORE();
1180 }
1181 
1182 /*
1183  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1184  */
1185 static int
if_vmove_loan(struct thread * td,struct ifnet * ifp,char * ifname,int jid)1186 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1187 {
1188 	struct prison *pr;
1189 	struct ifnet *difp;
1190 	bool found;
1191 
1192 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
1193 
1194 	/* Try to find the prison within our visibility. */
1195 	sx_slock(&allprison_lock);
1196 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1197 	sx_sunlock(&allprison_lock);
1198 	if (pr == NULL)
1199 		return (ENXIO);
1200 	prison_hold_locked(pr);
1201 	mtx_unlock(&pr->pr_mtx);
1202 
1203 	/* Do not try to move the iface from and to the same prison. */
1204 	if (pr->pr_vnet == ifp->if_vnet) {
1205 		prison_free(pr);
1206 		return (EEXIST);
1207 	}
1208 
1209 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1210 	/* XXX Lock interfaces to avoid races. */
1211 	CURVNET_SET_QUIET(pr->pr_vnet);
1212 	difp = ifunit(ifname);
1213 	CURVNET_RESTORE();
1214 	if (difp != NULL) {
1215 		prison_free(pr);
1216 		return (EEXIST);
1217 	}
1218 	sx_xlock(&ifnet_detach_sxlock);
1219 
1220 	found = if_unlink_ifnet(ifp, true);
1221 	if (! found) {
1222 		sx_xunlock(&ifnet_detach_sxlock);
1223 		prison_free(pr);
1224 		return (ENODEV);
1225 	}
1226 
1227 	/* Move the interface into the child jail/vnet. */
1228 	if_vmove(ifp, pr->pr_vnet);
1229 
1230 	/* Report the new if_xname back to the userland. */
1231 	sprintf(ifname, "%s", ifp->if_xname);
1232 
1233 	sx_xunlock(&ifnet_detach_sxlock);
1234 
1235 	prison_free(pr);
1236 	return (0);
1237 }
1238 
1239 static int
if_vmove_reclaim(struct thread * td,char * ifname,int jid)1240 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1241 {
1242 	struct prison *pr;
1243 	struct vnet *vnet_dst;
1244 	struct ifnet *ifp;
1245 	int found;
1246 
1247 	/* Try to find the prison within our visibility. */
1248 	sx_slock(&allprison_lock);
1249 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1250 	sx_sunlock(&allprison_lock);
1251 	if (pr == NULL)
1252 		return (ENXIO);
1253 	prison_hold_locked(pr);
1254 	mtx_unlock(&pr->pr_mtx);
1255 
1256 	/* Make sure the named iface exists in the source prison/vnet. */
1257 	CURVNET_SET(pr->pr_vnet);
1258 	ifp = ifunit(ifname);
1259 	if (ifp == NULL) {
1260 		CURVNET_RESTORE();
1261 		prison_free(pr);
1262 		return (ENXIO);
1263 	}
1264 
1265 	/* Do not try to move the iface from and to the same vnet. */
1266 	vnet_dst = TD_TO_VNET(td);
1267 	if (vnet_dst == pr->pr_vnet) {
1268 		CURVNET_RESTORE();
1269 		prison_free(pr);
1270 		return (EEXIST);
1271 	}
1272 
1273 	/* Get interface back from child jail/vnet. */
1274 	found = if_unlink_ifnet(ifp, true);
1275 	if (! found) {
1276 		CURVNET_RESTORE();
1277 		prison_free(pr);
1278 		return (ENODEV);
1279 	}
1280 	sx_xlock(&ifnet_detach_sxlock);
1281 	if_vmove(ifp, vnet_dst);
1282 	sx_xunlock(&ifnet_detach_sxlock);
1283 	CURVNET_RESTORE();
1284 
1285 	/* Report the new if_xname back to the userland. */
1286 	sprintf(ifname, "%s", ifp->if_xname);
1287 
1288 	prison_free(pr);
1289 	return (0);
1290 }
1291 #endif /* VIMAGE */
1292 
1293 /*
1294  * Add a group to an interface
1295  */
1296 int
if_addgroup(struct ifnet * ifp,const char * groupname)1297 if_addgroup(struct ifnet *ifp, const char *groupname)
1298 {
1299 	struct ifg_list		*ifgl;
1300 	struct ifg_group	*ifg = NULL;
1301 	struct ifg_member	*ifgm;
1302 	int 			 new = 0;
1303 
1304 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1305 	    groupname[strlen(groupname) - 1] <= '9')
1306 		return (EINVAL);
1307 
1308 	IFNET_WLOCK();
1309 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1310 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1311 			IFNET_WUNLOCK();
1312 			return (EEXIST);
1313 		}
1314 
1315 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1316 	    	IFNET_WUNLOCK();
1317 		return (ENOMEM);
1318 	}
1319 
1320 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1321 		free(ifgl, M_TEMP);
1322 		IFNET_WUNLOCK();
1323 		return (ENOMEM);
1324 	}
1325 
1326 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1327 		if (!strcmp(ifg->ifg_group, groupname))
1328 			break;
1329 
1330 	if (ifg == NULL) {
1331 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1332 			free(ifgl, M_TEMP);
1333 			free(ifgm, M_TEMP);
1334 			IFNET_WUNLOCK();
1335 			return (ENOMEM);
1336 		}
1337 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1338 		ifg->ifg_refcnt = 0;
1339 		CK_STAILQ_INIT(&ifg->ifg_members);
1340 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1341 		new = 1;
1342 	}
1343 
1344 	ifg->ifg_refcnt++;
1345 	ifgl->ifgl_group = ifg;
1346 	ifgm->ifgm_ifp = ifp;
1347 
1348 	IF_ADDR_WLOCK(ifp);
1349 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1350 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1351 	IF_ADDR_WUNLOCK(ifp);
1352 
1353 	IFNET_WUNLOCK();
1354 
1355 	if (new)
1356 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1357 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1358 
1359 	return (0);
1360 }
1361 
1362 /*
1363  * Helper function to remove a group out of an interface.  Expects the global
1364  * ifnet lock to be write-locked, and drops it before returning.
1365  */
1366 static void
_if_delgroup_locked(struct ifnet * ifp,struct ifg_list * ifgl,const char * groupname)1367 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1368     const char *groupname)
1369 {
1370 	struct ifg_member *ifgm;
1371 	bool freeifgl;
1372 
1373 	IFNET_WLOCK_ASSERT();
1374 
1375 	IF_ADDR_WLOCK(ifp);
1376 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1377 	IF_ADDR_WUNLOCK(ifp);
1378 
1379 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1380 		if (ifgm->ifgm_ifp == ifp) {
1381 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1382 			    ifg_member, ifgm_next);
1383 			break;
1384 		}
1385 	}
1386 
1387 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1388 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1389 		    ifg_next);
1390 		freeifgl = true;
1391 	} else {
1392 		freeifgl = false;
1393 	}
1394 	IFNET_WUNLOCK();
1395 
1396 	NET_EPOCH_WAIT();
1397 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1398 	if (freeifgl) {
1399 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1400 		free(ifgl->ifgl_group, M_TEMP);
1401 	}
1402 	free(ifgm, M_TEMP);
1403 	free(ifgl, M_TEMP);
1404 }
1405 
1406 /*
1407  * Remove a group from an interface
1408  */
1409 int
if_delgroup(struct ifnet * ifp,const char * groupname)1410 if_delgroup(struct ifnet *ifp, const char *groupname)
1411 {
1412 	struct ifg_list *ifgl;
1413 
1414 	IFNET_WLOCK();
1415 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1416 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1417 			break;
1418 	if (ifgl == NULL) {
1419 		IFNET_WUNLOCK();
1420 		return (ENOENT);
1421 	}
1422 
1423 	_if_delgroup_locked(ifp, ifgl, groupname);
1424 
1425 	return (0);
1426 }
1427 
1428 /*
1429  * Remove an interface from all groups
1430  */
1431 static void
if_delgroups(struct ifnet * ifp)1432 if_delgroups(struct ifnet *ifp)
1433 {
1434 	struct ifg_list *ifgl;
1435 	char groupname[IFNAMSIZ];
1436 
1437 	IFNET_WLOCK();
1438 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1439 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1440 		_if_delgroup_locked(ifp, ifgl, groupname);
1441 		IFNET_WLOCK();
1442 	}
1443 	IFNET_WUNLOCK();
1444 }
1445 
1446 /*
1447  * Stores all groups from an interface in memory pointed to by ifgr.
1448  */
1449 static int
if_getgroup(struct ifgroupreq * ifgr,struct ifnet * ifp)1450 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1451 {
1452 	int			 len, error;
1453 	struct ifg_list		*ifgl;
1454 	struct ifg_req		 ifgrq, *ifgp;
1455 
1456 	NET_EPOCH_ASSERT();
1457 
1458 	if (ifgr->ifgr_len == 0) {
1459 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1460 			ifgr->ifgr_len += sizeof(struct ifg_req);
1461 		return (0);
1462 	}
1463 
1464 	len = ifgr->ifgr_len;
1465 	ifgp = ifgr->ifgr_groups;
1466 	/* XXX: wire */
1467 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1468 		if (len < sizeof(ifgrq))
1469 			return (EINVAL);
1470 		bzero(&ifgrq, sizeof ifgrq);
1471 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1472 		    sizeof(ifgrq.ifgrq_group));
1473 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1474 			return (error);
1475 		len -= sizeof(ifgrq);
1476 		ifgp++;
1477 	}
1478 
1479 	return (0);
1480 }
1481 
1482 /*
1483  * Stores all members of a group in memory pointed to by igfr
1484  */
1485 static int
if_getgroupmembers(struct ifgroupreq * ifgr)1486 if_getgroupmembers(struct ifgroupreq *ifgr)
1487 {
1488 	struct ifg_group	*ifg;
1489 	struct ifg_member	*ifgm;
1490 	struct ifg_req		 ifgrq, *ifgp;
1491 	int			 len, error;
1492 
1493 	IFNET_RLOCK();
1494 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1495 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1496 			break;
1497 	if (ifg == NULL) {
1498 		IFNET_RUNLOCK();
1499 		return (ENOENT);
1500 	}
1501 
1502 	if (ifgr->ifgr_len == 0) {
1503 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1504 			ifgr->ifgr_len += sizeof(ifgrq);
1505 		IFNET_RUNLOCK();
1506 		return (0);
1507 	}
1508 
1509 	len = ifgr->ifgr_len;
1510 	ifgp = ifgr->ifgr_groups;
1511 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1512 		if (len < sizeof(ifgrq)) {
1513 			IFNET_RUNLOCK();
1514 			return (EINVAL);
1515 		}
1516 		bzero(&ifgrq, sizeof ifgrq);
1517 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1518 		    sizeof(ifgrq.ifgrq_member));
1519 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1520 			IFNET_RUNLOCK();
1521 			return (error);
1522 		}
1523 		len -= sizeof(ifgrq);
1524 		ifgp++;
1525 	}
1526 	IFNET_RUNLOCK();
1527 
1528 	return (0);
1529 }
1530 
1531 /*
1532  * Return counter values from counter(9)s stored in ifnet.
1533  */
1534 uint64_t
if_get_counter_default(struct ifnet * ifp,ift_counter cnt)1535 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1536 {
1537 
1538 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1539 
1540 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1541 }
1542 
1543 /*
1544  * Increase an ifnet counter. Usually used for counters shared
1545  * between the stack and a driver, but function supports them all.
1546  */
1547 void
if_inc_counter(struct ifnet * ifp,ift_counter cnt,int64_t inc)1548 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1549 {
1550 
1551 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1552 
1553 	counter_u64_add(ifp->if_counters[cnt], inc);
1554 }
1555 
1556 /*
1557  * Copy data from ifnet to userland API structure if_data.
1558  */
1559 void
if_data_copy(struct ifnet * ifp,struct if_data * ifd)1560 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1561 {
1562 
1563 	ifd->ifi_type = ifp->if_type;
1564 	ifd->ifi_physical = 0;
1565 	ifd->ifi_addrlen = ifp->if_addrlen;
1566 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1567 	ifd->ifi_link_state = ifp->if_link_state;
1568 	ifd->ifi_vhid = 0;
1569 	ifd->ifi_datalen = sizeof(struct if_data);
1570 	ifd->ifi_mtu = ifp->if_mtu;
1571 	ifd->ifi_metric = ifp->if_metric;
1572 	ifd->ifi_baudrate = ifp->if_baudrate;
1573 	ifd->ifi_hwassist = ifp->if_hwassist;
1574 	ifd->ifi_epoch = ifp->if_epoch;
1575 	ifd->ifi_lastchange = ifp->if_lastchange;
1576 
1577 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1578 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1579 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1580 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1581 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1582 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1583 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1584 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1585 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1586 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1587 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1588 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1589 }
1590 
1591 /*
1592  * Initialization, destruction and refcounting functions for ifaddrs.
1593  */
1594 struct ifaddr *
ifa_alloc(size_t size,int flags)1595 ifa_alloc(size_t size, int flags)
1596 {
1597 	struct ifaddr *ifa;
1598 
1599 	KASSERT(size >= sizeof(struct ifaddr),
1600 	    ("%s: invalid size %zu", __func__, size));
1601 
1602 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1603 	if (ifa == NULL)
1604 		return (NULL);
1605 
1606 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1607 		goto fail;
1608 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1609 		goto fail;
1610 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1611 		goto fail;
1612 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1613 		goto fail;
1614 
1615 	refcount_init(&ifa->ifa_refcnt, 1);
1616 
1617 	return (ifa);
1618 
1619 fail:
1620 	/* free(NULL) is okay */
1621 	counter_u64_free(ifa->ifa_opackets);
1622 	counter_u64_free(ifa->ifa_ipackets);
1623 	counter_u64_free(ifa->ifa_obytes);
1624 	counter_u64_free(ifa->ifa_ibytes);
1625 	free(ifa, M_IFADDR);
1626 
1627 	return (NULL);
1628 }
1629 
1630 void
ifa_ref(struct ifaddr * ifa)1631 ifa_ref(struct ifaddr *ifa)
1632 {
1633 	u_int old __diagused;
1634 
1635 	old = refcount_acquire(&ifa->ifa_refcnt);
1636 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1637 }
1638 
1639 int
ifa_try_ref(struct ifaddr * ifa)1640 ifa_try_ref(struct ifaddr *ifa)
1641 {
1642 
1643 	NET_EPOCH_ASSERT();
1644 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1645 }
1646 
1647 static void
ifa_destroy(epoch_context_t ctx)1648 ifa_destroy(epoch_context_t ctx)
1649 {
1650 	struct ifaddr *ifa;
1651 
1652 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1653 	counter_u64_free(ifa->ifa_opackets);
1654 	counter_u64_free(ifa->ifa_ipackets);
1655 	counter_u64_free(ifa->ifa_obytes);
1656 	counter_u64_free(ifa->ifa_ibytes);
1657 	free(ifa, M_IFADDR);
1658 }
1659 
1660 void
ifa_free(struct ifaddr * ifa)1661 ifa_free(struct ifaddr *ifa)
1662 {
1663 
1664 	if (refcount_release(&ifa->ifa_refcnt))
1665 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1666 }
1667 
1668 /*
1669  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1670  * structs used to represent other address families, it is necessary
1671  * to perform a different comparison.
1672  */
1673 static bool
sa_dl_equal(const struct sockaddr * a,const struct sockaddr * b)1674 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b)
1675 {
1676 	const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a;
1677 	const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b;
1678 
1679 	return (sdl1->sdl_len == sdl2->sdl_len &&
1680 	    bcmp(sdl1->sdl_data + sdl1->sdl_nlen,
1681 	    sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0);
1682 }
1683 
1684 /*
1685  * Locate an interface based on a complete address.
1686  */
1687 /*ARGSUSED*/
1688 struct ifaddr *
ifa_ifwithaddr(const struct sockaddr * addr)1689 ifa_ifwithaddr(const struct sockaddr *addr)
1690 {
1691 	struct ifnet *ifp;
1692 	struct ifaddr *ifa;
1693 
1694 	NET_EPOCH_ASSERT();
1695 
1696 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1697 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1698 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1699 				continue;
1700 			if (sa_equal(addr, ifa->ifa_addr)) {
1701 				goto done;
1702 			}
1703 			/* IP6 doesn't have broadcast */
1704 			if ((ifp->if_flags & IFF_BROADCAST) &&
1705 			    ifa->ifa_broadaddr &&
1706 			    ifa->ifa_broadaddr->sa_len != 0 &&
1707 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1708 				goto done;
1709 			}
1710 		}
1711 	}
1712 	ifa = NULL;
1713 done:
1714 	return (ifa);
1715 }
1716 
1717 int
ifa_ifwithaddr_check(const struct sockaddr * addr)1718 ifa_ifwithaddr_check(const struct sockaddr *addr)
1719 {
1720 	struct epoch_tracker et;
1721 	int rc;
1722 
1723 	NET_EPOCH_ENTER(et);
1724 	rc = (ifa_ifwithaddr(addr) != NULL);
1725 	NET_EPOCH_EXIT(et);
1726 	return (rc);
1727 }
1728 
1729 /*
1730  * Locate an interface based on the broadcast address.
1731  */
1732 /* ARGSUSED */
1733 struct ifaddr *
ifa_ifwithbroadaddr(const struct sockaddr * addr,int fibnum)1734 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1735 {
1736 	struct ifnet *ifp;
1737 	struct ifaddr *ifa;
1738 
1739 	NET_EPOCH_ASSERT();
1740 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1741 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1742 			continue;
1743 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1744 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1745 				continue;
1746 			if ((ifp->if_flags & IFF_BROADCAST) &&
1747 			    ifa->ifa_broadaddr &&
1748 			    ifa->ifa_broadaddr->sa_len != 0 &&
1749 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1750 				goto done;
1751 			}
1752 		}
1753 	}
1754 	ifa = NULL;
1755 done:
1756 	return (ifa);
1757 }
1758 
1759 /*
1760  * Locate the point to point interface with a given destination address.
1761  */
1762 /*ARGSUSED*/
1763 struct ifaddr *
ifa_ifwithdstaddr(const struct sockaddr * addr,int fibnum)1764 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1765 {
1766 	struct ifnet *ifp;
1767 	struct ifaddr *ifa;
1768 
1769 	NET_EPOCH_ASSERT();
1770 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1771 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1772 			continue;
1773 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1774 			continue;
1775 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1776 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1777 				continue;
1778 			if (ifa->ifa_dstaddr != NULL &&
1779 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1780 				goto done;
1781 			}
1782 		}
1783 	}
1784 	ifa = NULL;
1785 done:
1786 	return (ifa);
1787 }
1788 
1789 /*
1790  * Find an interface on a specific network.  If many, choice
1791  * is most specific found.
1792  */
1793 struct ifaddr *
ifa_ifwithnet(const struct sockaddr * addr,int ignore_ptp,int fibnum)1794 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1795 {
1796 	struct ifnet *ifp;
1797 	struct ifaddr *ifa;
1798 	struct ifaddr *ifa_maybe = NULL;
1799 	u_int af = addr->sa_family;
1800 	const char *addr_data = addr->sa_data, *cplim;
1801 
1802 	NET_EPOCH_ASSERT();
1803 	/*
1804 	 * AF_LINK addresses can be looked up directly by their index number,
1805 	 * so do that if we can.
1806 	 */
1807 	if (af == AF_LINK) {
1808 		ifp = ifnet_byindex(
1809 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1810 		return (ifp ? ifp->if_addr : NULL);
1811 	}
1812 
1813 	/*
1814 	 * Scan though each interface, looking for ones that have addresses
1815 	 * in this address family and the requested fib.
1816 	 */
1817 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1818 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1819 			continue;
1820 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1821 			const char *cp, *cp2, *cp3;
1822 
1823 			if (ifa->ifa_addr->sa_family != af)
1824 next:				continue;
1825 			if (af == AF_INET &&
1826 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1827 				/*
1828 				 * This is a bit broken as it doesn't
1829 				 * take into account that the remote end may
1830 				 * be a single node in the network we are
1831 				 * looking for.
1832 				 * The trouble is that we don't know the
1833 				 * netmask for the remote end.
1834 				 */
1835 				if (ifa->ifa_dstaddr != NULL &&
1836 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1837 					goto done;
1838 				}
1839 			} else {
1840 				/*
1841 				 * Scan all the bits in the ifa's address.
1842 				 * If a bit dissagrees with what we are
1843 				 * looking for, mask it with the netmask
1844 				 * to see if it really matters.
1845 				 * (A byte at a time)
1846 				 */
1847 				if (ifa->ifa_netmask == 0)
1848 					continue;
1849 				cp = addr_data;
1850 				cp2 = ifa->ifa_addr->sa_data;
1851 				cp3 = ifa->ifa_netmask->sa_data;
1852 				cplim = ifa->ifa_netmask->sa_len
1853 					+ (char *)ifa->ifa_netmask;
1854 				while (cp3 < cplim)
1855 					if ((*cp++ ^ *cp2++) & *cp3++)
1856 						goto next; /* next address! */
1857 				/*
1858 				 * If the netmask of what we just found
1859 				 * is more specific than what we had before
1860 				 * (if we had one), or if the virtual status
1861 				 * of new prefix is better than of the old one,
1862 				 * then remember the new one before continuing
1863 				 * to search for an even better one.
1864 				 */
1865 				if (ifa_maybe == NULL ||
1866 				    ifa_preferred(ifa_maybe, ifa) ||
1867 				    rn_refines((caddr_t)ifa->ifa_netmask,
1868 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1869 					ifa_maybe = ifa;
1870 				}
1871 			}
1872 		}
1873 	}
1874 	ifa = ifa_maybe;
1875 	ifa_maybe = NULL;
1876 done:
1877 	return (ifa);
1878 }
1879 
1880 /*
1881  * Find an interface address specific to an interface best matching
1882  * a given address.
1883  */
1884 struct ifaddr *
ifaof_ifpforaddr(const struct sockaddr * addr,struct ifnet * ifp)1885 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1886 {
1887 	struct ifaddr *ifa;
1888 	const char *cp, *cp2, *cp3;
1889 	char *cplim;
1890 	struct ifaddr *ifa_maybe = NULL;
1891 	u_int af = addr->sa_family;
1892 
1893 	if (af >= AF_MAX)
1894 		return (NULL);
1895 
1896 	NET_EPOCH_ASSERT();
1897 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1898 		if (ifa->ifa_addr->sa_family != af)
1899 			continue;
1900 		if (ifa_maybe == NULL)
1901 			ifa_maybe = ifa;
1902 		if (ifa->ifa_netmask == 0) {
1903 			if (sa_equal(addr, ifa->ifa_addr) ||
1904 			    (ifa->ifa_dstaddr &&
1905 			    sa_equal(addr, ifa->ifa_dstaddr)))
1906 				goto done;
1907 			continue;
1908 		}
1909 		if (ifp->if_flags & IFF_POINTOPOINT) {
1910 			if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))
1911 				goto done;
1912 		} else {
1913 			cp = addr->sa_data;
1914 			cp2 = ifa->ifa_addr->sa_data;
1915 			cp3 = ifa->ifa_netmask->sa_data;
1916 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1917 			for (; cp3 < cplim; cp3++)
1918 				if ((*cp++ ^ *cp2++) & *cp3)
1919 					break;
1920 			if (cp3 == cplim)
1921 				goto done;
1922 		}
1923 	}
1924 	ifa = ifa_maybe;
1925 done:
1926 	return (ifa);
1927 }
1928 
1929 /*
1930  * See whether new ifa is better than current one:
1931  * 1) A non-virtual one is preferred over virtual.
1932  * 2) A virtual in master state preferred over any other state.
1933  *
1934  * Used in several address selecting functions.
1935  */
1936 int
ifa_preferred(struct ifaddr * cur,struct ifaddr * next)1937 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1938 {
1939 
1940 	return (cur->ifa_carp && (!next->ifa_carp ||
1941 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1942 }
1943 
1944 struct sockaddr_dl *
link_alloc_sdl(size_t size,int flags)1945 link_alloc_sdl(size_t size, int flags)
1946 {
1947 
1948 	return (malloc(size, M_TEMP, flags));
1949 }
1950 
1951 void
link_free_sdl(struct sockaddr * sa)1952 link_free_sdl(struct sockaddr *sa)
1953 {
1954 	free(sa, M_TEMP);
1955 }
1956 
1957 /*
1958  * Fills in given sdl with interface basic info.
1959  * Returns pointer to filled sdl.
1960  */
1961 struct sockaddr_dl *
link_init_sdl(struct ifnet * ifp,struct sockaddr * paddr,u_char iftype)1962 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1963 {
1964 	struct sockaddr_dl *sdl;
1965 
1966 	sdl = (struct sockaddr_dl *)paddr;
1967 	memset(sdl, 0, sizeof(struct sockaddr_dl));
1968 	sdl->sdl_len = sizeof(struct sockaddr_dl);
1969 	sdl->sdl_family = AF_LINK;
1970 	sdl->sdl_index = ifp->if_index;
1971 	sdl->sdl_type = iftype;
1972 
1973 	return (sdl);
1974 }
1975 
1976 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1977 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1978 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
1979 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
1980 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
1981 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
1982 int	(*vlan_setcookie_p)(struct ifnet *, void *);
1983 void	*(*vlan_cookie_p)(struct ifnet *);
1984 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
1985 
1986 /*
1987  * Handle a change in the interface link state. To avoid LORs
1988  * between driver lock and upper layer locks, as well as possible
1989  * recursions, we post event to taskqueue, and all job
1990  * is done in static do_link_state_change().
1991  */
1992 void
if_link_state_change(struct ifnet * ifp,int link_state)1993 if_link_state_change(struct ifnet *ifp, int link_state)
1994 {
1995 	/* Return if state hasn't changed. */
1996 	if (ifp->if_link_state == link_state)
1997 		return;
1998 
1999 	ifp->if_link_state = link_state;
2000 
2001 	/* XXXGL: reference ifp? */
2002 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2003 }
2004 
2005 static void
do_link_state_change(void * arg,int pending)2006 do_link_state_change(void *arg, int pending)
2007 {
2008 	struct ifnet *ifp;
2009 	int link_state;
2010 
2011 	ifp = arg;
2012 	link_state = ifp->if_link_state;
2013 
2014 	CURVNET_SET(ifp->if_vnet);
2015 	rt_ifmsg(ifp, 0);
2016 	if (ifp->if_vlantrunk != NULL)
2017 		(*vlan_link_state_p)(ifp);
2018 	if (ifp->if_carp)
2019 		(*carp_linkstate_p)(ifp);
2020 	if (ifp->if_bridge)
2021 		ifp->if_bridge_linkstate(ifp);
2022 	if (ifp->if_lagg)
2023 		(*lagg_linkstate_p)(ifp, link_state);
2024 
2025 	if (IS_DEFAULT_VNET(curvnet))
2026 		devctl_notify("IFNET", ifp->if_xname,
2027 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2028 		    NULL);
2029 	if (pending > 1)
2030 		if_printf(ifp, "%d link states coalesced\n", pending);
2031 	if (log_link_state_change)
2032 		if_printf(ifp, "link state changed to %s\n",
2033 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2034 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2035 	CURVNET_RESTORE();
2036 }
2037 
2038 /*
2039  * Mark an interface down and notify protocols of
2040  * the transition.
2041  */
2042 void
if_down(struct ifnet * ifp)2043 if_down(struct ifnet *ifp)
2044 {
2045 
2046 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2047 
2048 	ifp->if_flags &= ~IFF_UP;
2049 	getmicrotime(&ifp->if_lastchange);
2050 	ifp->if_qflush(ifp);
2051 
2052 	if (ifp->if_carp)
2053 		(*carp_linkstate_p)(ifp);
2054 	rt_ifmsg(ifp, IFF_UP);
2055 }
2056 
2057 /*
2058  * Mark an interface up and notify protocols of
2059  * the transition.
2060  */
2061 void
if_up(struct ifnet * ifp)2062 if_up(struct ifnet *ifp)
2063 {
2064 
2065 	ifp->if_flags |= IFF_UP;
2066 	getmicrotime(&ifp->if_lastchange);
2067 	if (ifp->if_carp)
2068 		(*carp_linkstate_p)(ifp);
2069 	rt_ifmsg(ifp, IFF_UP);
2070 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2071 }
2072 
2073 /*
2074  * Flush an interface queue.
2075  */
2076 void
if_qflush(struct ifnet * ifp)2077 if_qflush(struct ifnet *ifp)
2078 {
2079 	struct mbuf *m, *n;
2080 	struct ifaltq *ifq;
2081 
2082 	ifq = &ifp->if_snd;
2083 	IFQ_LOCK(ifq);
2084 #ifdef ALTQ
2085 	if (ALTQ_IS_ENABLED(ifq))
2086 		ALTQ_PURGE(ifq);
2087 #endif
2088 	n = ifq->ifq_head;
2089 	while ((m = n) != NULL) {
2090 		n = m->m_nextpkt;
2091 		m_freem(m);
2092 	}
2093 	ifq->ifq_head = 0;
2094 	ifq->ifq_tail = 0;
2095 	ifq->ifq_len = 0;
2096 	IFQ_UNLOCK(ifq);
2097 }
2098 
2099 /*
2100  * Map interface name to interface structure pointer, with or without
2101  * returning a reference.
2102  */
2103 struct ifnet *
ifunit_ref(const char * name)2104 ifunit_ref(const char *name)
2105 {
2106 	struct epoch_tracker et;
2107 	struct ifnet *ifp;
2108 
2109 	NET_EPOCH_ENTER(et);
2110 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2111 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2112 		    !(ifp->if_flags & IFF_DYING))
2113 			break;
2114 	}
2115 	if (ifp != NULL) {
2116 		if_ref(ifp);
2117 		MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2118 	}
2119 
2120 	NET_EPOCH_EXIT(et);
2121 	return (ifp);
2122 }
2123 
2124 struct ifnet *
ifunit(const char * name)2125 ifunit(const char *name)
2126 {
2127 	struct epoch_tracker et;
2128 	struct ifnet *ifp;
2129 
2130 	NET_EPOCH_ENTER(et);
2131 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2132 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2133 			break;
2134 	}
2135 	NET_EPOCH_EXIT(et);
2136 	return (ifp);
2137 }
2138 
2139 void *
ifr_buffer_get_buffer(void * data)2140 ifr_buffer_get_buffer(void *data)
2141 {
2142 	union ifreq_union *ifrup;
2143 
2144 	ifrup = data;
2145 #ifdef COMPAT_FREEBSD32
2146 	if (SV_CURPROC_FLAG(SV_ILP32))
2147 		return ((void *)(uintptr_t)
2148 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2149 #endif
2150 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2151 }
2152 
2153 static void
ifr_buffer_set_buffer_null(void * data)2154 ifr_buffer_set_buffer_null(void *data)
2155 {
2156 	union ifreq_union *ifrup;
2157 
2158 	ifrup = data;
2159 #ifdef COMPAT_FREEBSD32
2160 	if (SV_CURPROC_FLAG(SV_ILP32))
2161 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2162 	else
2163 #endif
2164 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2165 }
2166 
2167 size_t
ifr_buffer_get_length(void * data)2168 ifr_buffer_get_length(void *data)
2169 {
2170 	union ifreq_union *ifrup;
2171 
2172 	ifrup = data;
2173 #ifdef COMPAT_FREEBSD32
2174 	if (SV_CURPROC_FLAG(SV_ILP32))
2175 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2176 #endif
2177 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2178 }
2179 
2180 static void
ifr_buffer_set_length(void * data,size_t len)2181 ifr_buffer_set_length(void *data, size_t len)
2182 {
2183 	union ifreq_union *ifrup;
2184 
2185 	ifrup = data;
2186 #ifdef COMPAT_FREEBSD32
2187 	if (SV_CURPROC_FLAG(SV_ILP32))
2188 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2189 	else
2190 #endif
2191 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2192 }
2193 
2194 void *
ifr_data_get_ptr(void * ifrp)2195 ifr_data_get_ptr(void *ifrp)
2196 {
2197 	union ifreq_union *ifrup;
2198 
2199 	ifrup = ifrp;
2200 #ifdef COMPAT_FREEBSD32
2201 	if (SV_CURPROC_FLAG(SV_ILP32))
2202 		return ((void *)(uintptr_t)
2203 		    ifrup->ifr32.ifr_ifru.ifru_data);
2204 #endif
2205 		return (ifrup->ifr.ifr_ifru.ifru_data);
2206 }
2207 
2208 struct ifcap_nv_bit_name {
2209 	uint64_t cap_bit;
2210 	const char *cap_name;
2211 };
2212 #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2213     .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2214 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2215 	CAPNV(RXCSUM),
2216 	CAPNV(TXCSUM),
2217 	CAPNV(NETCONS),
2218 	CAPNV(VLAN_MTU),
2219 	CAPNV(VLAN_HWTAGGING),
2220 	CAPNV(JUMBO_MTU),
2221 	CAPNV(POLLING),
2222 	CAPNV(VLAN_HWCSUM),
2223 	CAPNV(TSO4),
2224 	CAPNV(TSO6),
2225 	CAPNV(LRO),
2226 	CAPNV(WOL_UCAST),
2227 	CAPNV(WOL_MCAST),
2228 	CAPNV(WOL_MAGIC),
2229 	CAPNV(TOE4),
2230 	CAPNV(TOE6),
2231 	CAPNV(VLAN_HWFILTER),
2232 	CAPNV(VLAN_HWTSO),
2233 	CAPNV(LINKSTATE),
2234 	CAPNV(NETMAP),
2235 	CAPNV(RXCSUM_IPV6),
2236 	CAPNV(TXCSUM_IPV6),
2237 	CAPNV(HWSTATS),
2238 	CAPNV(TXRTLMT),
2239 	CAPNV(HWRXTSTMP),
2240 	CAPNV(MEXTPG),
2241 	CAPNV(TXTLS4),
2242 	CAPNV(TXTLS6),
2243 	CAPNV(VXLAN_HWCSUM),
2244 	CAPNV(VXLAN_HWTSO),
2245 	CAPNV(TXTLS_RTLMT),
2246 	{0, NULL}
2247 };
2248 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \
2249     .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2250 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2251 	CAP2NV(RXTLS4),
2252 	CAP2NV(RXTLS6),
2253 	CAP2NV(IPSEC_OFFLOAD),
2254 	{0, NULL}
2255 };
2256 #undef CAPNV
2257 #undef CAP2NV
2258 
2259 int
if_capnv_to_capint(const nvlist_t * nv,int * old_cap,const struct ifcap_nv_bit_name * nn,bool all)2260 if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2261     const struct ifcap_nv_bit_name *nn, bool all)
2262 {
2263 	int i, res;
2264 
2265 	res = 0;
2266 	for (i = 0; nn[i].cap_name != NULL; i++) {
2267 		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2268 			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2269 				res |= nn[i].cap_bit;
2270 		} else {
2271 			res |= *old_cap & nn[i].cap_bit;
2272 		}
2273 	}
2274 	return (res);
2275 }
2276 
2277 void
if_capint_to_capnv(nvlist_t * nv,const struct ifcap_nv_bit_name * nn,int ifr_cap,int ifr_req)2278 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2279     int ifr_cap, int ifr_req)
2280 {
2281 	int i;
2282 
2283 	for (i = 0; nn[i].cap_name != NULL; i++) {
2284 		if ((nn[i].cap_bit & ifr_cap) != 0) {
2285 			nvlist_add_bool(nv, nn[i].cap_name,
2286 			    (nn[i].cap_bit & ifr_req) != 0);
2287 		}
2288 	}
2289 }
2290 
2291 /*
2292  * Hardware specific interface ioctls.
2293  */
2294 int
ifhwioctl(u_long cmd,struct ifnet * ifp,caddr_t data,struct thread * td)2295 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2296 {
2297 	struct ifreq *ifr;
2298 	int error = 0, do_ifup = 0;
2299 	int new_flags, temp_flags;
2300 	size_t descrlen, nvbuflen;
2301 	char *descrbuf;
2302 	char new_name[IFNAMSIZ];
2303 	void *buf;
2304 	nvlist_t *nvcap;
2305 	struct siocsifcapnv_driver_data drv_ioctl_data;
2306 
2307 	ifr = (struct ifreq *)data;
2308 	switch (cmd) {
2309 	case SIOCGIFINDEX:
2310 		ifr->ifr_index = ifp->if_index;
2311 		break;
2312 
2313 	case SIOCGIFFLAGS:
2314 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2315 		ifr->ifr_flags = temp_flags & 0xffff;
2316 		ifr->ifr_flagshigh = temp_flags >> 16;
2317 		break;
2318 
2319 	case SIOCGIFCAP:
2320 		ifr->ifr_reqcap = ifp->if_capabilities;
2321 		ifr->ifr_curcap = ifp->if_capenable;
2322 		break;
2323 
2324 	case SIOCGIFCAPNV:
2325 		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2326 			error = EINVAL;
2327 			break;
2328 		}
2329 		buf = NULL;
2330 		nvcap = nvlist_create(0);
2331 		for (;;) {
2332 			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2333 			    ifp->if_capabilities, ifp->if_capenable);
2334 			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2335 			    ifp->if_capabilities2, ifp->if_capenable2);
2336 			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2337 			    __DECONST(caddr_t, nvcap));
2338 			if (error != 0) {
2339 				if_printf(ifp,
2340 			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2341 				    error);
2342 				break;
2343 			}
2344 			buf = nvlist_pack(nvcap, &nvbuflen);
2345 			if (buf == NULL) {
2346 				error = nvlist_error(nvcap);
2347 				if (error == 0)
2348 					error = EDOOFUS;
2349 				break;
2350 			}
2351 			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2352 				ifr->ifr_cap_nv.length = nvbuflen;
2353 				ifr->ifr_cap_nv.buffer = NULL;
2354 				error = EFBIG;
2355 				break;
2356 			}
2357 			ifr->ifr_cap_nv.length = nvbuflen;
2358 			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2359 			break;
2360 		}
2361 		free(buf, M_NVLIST);
2362 		nvlist_destroy(nvcap);
2363 		break;
2364 
2365 	case SIOCGIFDATA:
2366 	{
2367 		struct if_data ifd;
2368 
2369 		/* Ensure uninitialised padding is not leaked. */
2370 		memset(&ifd, 0, sizeof(ifd));
2371 
2372 		if_data_copy(ifp, &ifd);
2373 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2374 		break;
2375 	}
2376 
2377 #ifdef MAC
2378 	case SIOCGIFMAC:
2379 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2380 		break;
2381 #endif
2382 
2383 	case SIOCGIFMETRIC:
2384 		ifr->ifr_metric = ifp->if_metric;
2385 		break;
2386 
2387 	case SIOCGIFMTU:
2388 		ifr->ifr_mtu = ifp->if_mtu;
2389 		break;
2390 
2391 	case SIOCGIFPHYS:
2392 		/* XXXGL: did this ever worked? */
2393 		ifr->ifr_phys = 0;
2394 		break;
2395 
2396 	case SIOCGIFDESCR:
2397 		error = 0;
2398 		sx_slock(&ifdescr_sx);
2399 		if (ifp->if_description == NULL)
2400 			error = ENOMSG;
2401 		else {
2402 			/* space for terminating nul */
2403 			descrlen = strlen(ifp->if_description) + 1;
2404 			if (ifr_buffer_get_length(ifr) < descrlen)
2405 				ifr_buffer_set_buffer_null(ifr);
2406 			else
2407 				error = copyout(ifp->if_description,
2408 				    ifr_buffer_get_buffer(ifr), descrlen);
2409 			ifr_buffer_set_length(ifr, descrlen);
2410 		}
2411 		sx_sunlock(&ifdescr_sx);
2412 		break;
2413 
2414 	case SIOCSIFDESCR:
2415 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2416 		if (error)
2417 			return (error);
2418 
2419 		/*
2420 		 * Copy only (length-1) bytes to make sure that
2421 		 * if_description is always nul terminated.  The
2422 		 * length parameter is supposed to count the
2423 		 * terminating nul in.
2424 		 */
2425 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2426 			return (ENAMETOOLONG);
2427 		else if (ifr_buffer_get_length(ifr) == 0)
2428 			descrbuf = NULL;
2429 		else {
2430 			descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK);
2431 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2432 			    ifr_buffer_get_length(ifr) - 1);
2433 			if (error) {
2434 				if_freedescr(descrbuf);
2435 				break;
2436 			}
2437 		}
2438 
2439 		if_setdescr(ifp, descrbuf);
2440 		getmicrotime(&ifp->if_lastchange);
2441 		break;
2442 
2443 	case SIOCGIFFIB:
2444 		ifr->ifr_fib = ifp->if_fib;
2445 		break;
2446 
2447 	case SIOCSIFFIB:
2448 		error = priv_check(td, PRIV_NET_SETIFFIB);
2449 		if (error)
2450 			return (error);
2451 		if (ifr->ifr_fib >= rt_numfibs)
2452 			return (EINVAL);
2453 
2454 		ifp->if_fib = ifr->ifr_fib;
2455 		break;
2456 
2457 	case SIOCSIFFLAGS:
2458 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2459 		if (error)
2460 			return (error);
2461 		/*
2462 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2463 		 * check, so we don't need special handling here yet.
2464 		 */
2465 		new_flags = (ifr->ifr_flags & 0xffff) |
2466 		    (ifr->ifr_flagshigh << 16);
2467 		if (ifp->if_flags & IFF_UP &&
2468 		    (new_flags & IFF_UP) == 0) {
2469 			if_down(ifp);
2470 		} else if (new_flags & IFF_UP &&
2471 		    (ifp->if_flags & IFF_UP) == 0) {
2472 			do_ifup = 1;
2473 		}
2474 
2475 		/*
2476 		 * See if the promiscuous mode or allmulti bits are about to
2477 		 * flip.  They require special handling because in-kernel
2478 		 * consumers may indepdently toggle them.
2479 		 */
2480 		if_setppromisc(ifp, new_flags & IFF_PPROMISC);
2481 		if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) {
2482 			if (new_flags & IFF_PALLMULTI)
2483 				ifp->if_flags |= IFF_ALLMULTI;
2484 			else if (ifp->if_amcount == 0)
2485 				ifp->if_flags &= ~IFF_ALLMULTI;
2486 		}
2487 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2488 			(new_flags &~ IFF_CANTCHANGE);
2489 		if (ifp->if_ioctl) {
2490 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2491 		}
2492 		if (do_ifup)
2493 			if_up(ifp);
2494 		getmicrotime(&ifp->if_lastchange);
2495 		break;
2496 
2497 	case SIOCSIFCAP:
2498 		error = priv_check(td, PRIV_NET_SETIFCAP);
2499 		if (error != 0)
2500 			return (error);
2501 		if (ifp->if_ioctl == NULL)
2502 			return (EOPNOTSUPP);
2503 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2504 			return (EINVAL);
2505 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2506 		if (error == 0)
2507 			getmicrotime(&ifp->if_lastchange);
2508 		break;
2509 
2510 	case SIOCSIFCAPNV:
2511 		error = priv_check(td, PRIV_NET_SETIFCAP);
2512 		if (error != 0)
2513 			return (error);
2514 		if (ifp->if_ioctl == NULL)
2515 			return (EOPNOTSUPP);
2516 		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2517 			return (EINVAL);
2518 		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2519 			return (EINVAL);
2520 		nvcap = NULL;
2521 		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2522 		for (;;) {
2523 			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2524 			    ifr->ifr_cap_nv.length);
2525 			if (error != 0)
2526 				break;
2527 			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2528 			if (nvcap == NULL) {
2529 				error = EINVAL;
2530 				break;
2531 			}
2532 			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2533 			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2534 			if ((drv_ioctl_data.reqcap &
2535 			    ~ifp->if_capabilities) != 0) {
2536 				error = EINVAL;
2537 				break;
2538 			}
2539 			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2540 			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2541 			if ((drv_ioctl_data.reqcap2 &
2542 			    ~ifp->if_capabilities2) != 0) {
2543 				error = EINVAL;
2544 				break;
2545 			}
2546 			drv_ioctl_data.nvcap = nvcap;
2547 			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2548 			    (caddr_t)&drv_ioctl_data);
2549 			break;
2550 		}
2551 		nvlist_destroy(nvcap);
2552 		free(buf, M_TEMP);
2553 		if (error == 0)
2554 			getmicrotime(&ifp->if_lastchange);
2555 		break;
2556 
2557 #ifdef MAC
2558 	case SIOCSIFMAC:
2559 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2560 		break;
2561 #endif
2562 
2563 	case SIOCSIFNAME:
2564 		error = priv_check(td, PRIV_NET_SETIFNAME);
2565 		if (error)
2566 			return (error);
2567 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2568 		    NULL);
2569 		if (error != 0)
2570 			return (error);
2571 		error = if_rename(ifp, new_name);
2572 		break;
2573 
2574 #ifdef VIMAGE
2575 	case SIOCSIFVNET:
2576 		error = priv_check(td, PRIV_NET_SETIFVNET);
2577 		if (error)
2578 			return (error);
2579 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2580 		break;
2581 #endif
2582 
2583 	case SIOCSIFMETRIC:
2584 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2585 		if (error)
2586 			return (error);
2587 		ifp->if_metric = ifr->ifr_metric;
2588 		getmicrotime(&ifp->if_lastchange);
2589 		break;
2590 
2591 	case SIOCSIFPHYS:
2592 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2593 		if (error)
2594 			return (error);
2595 		if (ifp->if_ioctl == NULL)
2596 			return (EOPNOTSUPP);
2597 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2598 		if (error == 0)
2599 			getmicrotime(&ifp->if_lastchange);
2600 		break;
2601 
2602 	case SIOCSIFMTU:
2603 	{
2604 		u_long oldmtu = ifp->if_mtu;
2605 
2606 		error = priv_check(td, PRIV_NET_SETIFMTU);
2607 		if (error)
2608 			return (error);
2609 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2610 			return (EINVAL);
2611 		if (ifp->if_ioctl == NULL)
2612 			return (EOPNOTSUPP);
2613 		/* Disallow MTU changes on bridge member interfaces. */
2614 		if (ifp->if_bridge)
2615 			return (EOPNOTSUPP);
2616 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2617 		if (error == 0) {
2618 			getmicrotime(&ifp->if_lastchange);
2619 			rt_ifmsg(ifp, 0);
2620 #ifdef INET
2621 			DEBUGNET_NOTIFY_MTU(ifp);
2622 #endif
2623 		}
2624 		/*
2625 		 * If the link MTU changed, do network layer specific procedure.
2626 		 */
2627 		if (ifp->if_mtu != oldmtu)
2628 			if_notifymtu(ifp);
2629 		break;
2630 	}
2631 
2632 	case SIOCADDMULTI:
2633 	case SIOCDELMULTI:
2634 		if (cmd == SIOCADDMULTI)
2635 			error = priv_check(td, PRIV_NET_ADDMULTI);
2636 		else
2637 			error = priv_check(td, PRIV_NET_DELMULTI);
2638 		if (error)
2639 			return (error);
2640 
2641 		/* Don't allow group membership on non-multicast interfaces. */
2642 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2643 			return (EOPNOTSUPP);
2644 
2645 		/* Don't let users screw up protocols' entries. */
2646 		if (ifr->ifr_addr.sa_family != AF_LINK)
2647 			return (EINVAL);
2648 
2649 		if (cmd == SIOCADDMULTI) {
2650 			struct epoch_tracker et;
2651 			struct ifmultiaddr *ifma;
2652 
2653 			/*
2654 			 * Userland is only permitted to join groups once
2655 			 * via the if_addmulti() KPI, because it cannot hold
2656 			 * struct ifmultiaddr * between calls. It may also
2657 			 * lose a race while we check if the membership
2658 			 * already exists.
2659 			 */
2660 			NET_EPOCH_ENTER(et);
2661 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2662 			NET_EPOCH_EXIT(et);
2663 			if (ifma != NULL)
2664 				error = EADDRINUSE;
2665 			else
2666 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2667 		} else {
2668 			error = if_delmulti(ifp, &ifr->ifr_addr);
2669 		}
2670 		if (error == 0)
2671 			getmicrotime(&ifp->if_lastchange);
2672 		break;
2673 
2674 	case SIOCSIFPHYADDR:
2675 	case SIOCDIFPHYADDR:
2676 #ifdef INET6
2677 	case SIOCSIFPHYADDR_IN6:
2678 #endif
2679 	case SIOCSIFMEDIA:
2680 	case SIOCSIFGENERIC:
2681 		error = priv_check(td, PRIV_NET_HWIOCTL);
2682 		if (error)
2683 			return (error);
2684 		if (ifp->if_ioctl == NULL)
2685 			return (EOPNOTSUPP);
2686 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2687 		if (error == 0)
2688 			getmicrotime(&ifp->if_lastchange);
2689 		break;
2690 
2691 	case SIOCGIFSTATUS:
2692 	case SIOCGIFPSRCADDR:
2693 	case SIOCGIFPDSTADDR:
2694 	case SIOCGIFMEDIA:
2695 	case SIOCGIFXMEDIA:
2696 	case SIOCGIFGENERIC:
2697 	case SIOCGIFRSSKEY:
2698 	case SIOCGIFRSSHASH:
2699 	case SIOCGIFDOWNREASON:
2700 		if (ifp->if_ioctl == NULL)
2701 			return (EOPNOTSUPP);
2702 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2703 		break;
2704 
2705 	case SIOCSIFLLADDR:
2706 		error = priv_check(td, PRIV_NET_SETLLADDR);
2707 		if (error)
2708 			return (error);
2709 		error = if_setlladdr(ifp,
2710 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2711 		break;
2712 
2713 	case SIOCGHWADDR:
2714 		error = if_gethwaddr(ifp, ifr);
2715 		break;
2716 
2717 	case SIOCAIFGROUP:
2718 	{
2719 		const char *groupname;
2720 
2721 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2722 		if (error)
2723 			return (error);
2724 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2725 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2726 			return (EINVAL);
2727 		error = if_addgroup(ifp, groupname);
2728 		if (error != 0)
2729 			return (error);
2730 		break;
2731 	}
2732 	case SIOCGIFGROUP:
2733 	{
2734 		struct epoch_tracker et;
2735 
2736 		NET_EPOCH_ENTER(et);
2737 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2738 		NET_EPOCH_EXIT(et);
2739 		break;
2740 	}
2741 
2742 	case SIOCDIFGROUP:
2743 	{
2744 		const char *groupname;
2745 
2746 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2747 		if (error)
2748 			return (error);
2749 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2750 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2751 			return (EINVAL);
2752 		error = if_delgroup(ifp, groupname);
2753 		if (error != 0)
2754 			return (error);
2755 		break;
2756 	}
2757 	default:
2758 		error = ENOIOCTL;
2759 		break;
2760 	}
2761 	return (error);
2762 }
2763 
2764 /*
2765  * Interface ioctls.
2766  */
2767 int
ifioctl(struct socket * so,u_long cmd,caddr_t data,struct thread * td)2768 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2769 {
2770 #ifdef COMPAT_FREEBSD32
2771 	union {
2772 		struct ifconf ifc;
2773 		struct ifdrv ifd;
2774 		struct ifgroupreq ifgr;
2775 		struct ifmediareq ifmr;
2776 	} thunk;
2777 	u_long saved_cmd;
2778 	struct ifconf32 *ifc32;
2779 	struct ifdrv32 *ifd32;
2780 	struct ifgroupreq32 *ifgr32;
2781 	struct ifmediareq32 *ifmr32;
2782 #endif
2783 	struct ifnet *ifp;
2784 	struct ifreq *ifr;
2785 	int error;
2786 	int oif_flags;
2787 #ifdef VIMAGE
2788 	bool shutdown;
2789 #endif
2790 
2791 	CURVNET_SET(so->so_vnet);
2792 #ifdef VIMAGE
2793 	/* Make sure the VNET is stable. */
2794 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2795 	if (shutdown) {
2796 		CURVNET_RESTORE();
2797 		return (EBUSY);
2798 	}
2799 #endif
2800 
2801 #ifdef COMPAT_FREEBSD32
2802 	saved_cmd = cmd;
2803 	switch (cmd) {
2804 	case SIOCGIFCONF32:
2805 		ifc32 = (struct ifconf32 *)data;
2806 		thunk.ifc.ifc_len = ifc32->ifc_len;
2807 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2808 		data = (caddr_t)&thunk.ifc;
2809 		cmd = SIOCGIFCONF;
2810 		break;
2811 	case SIOCGDRVSPEC32:
2812 	case SIOCSDRVSPEC32:
2813 		ifd32 = (struct ifdrv32 *)data;
2814 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2815 		    sizeof(thunk.ifd.ifd_name));
2816 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2817 		thunk.ifd.ifd_len = ifd32->ifd_len;
2818 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2819 		data = (caddr_t)&thunk.ifd;
2820 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2821 		break;
2822 	case SIOCAIFGROUP32:
2823 	case SIOCGIFGROUP32:
2824 	case SIOCDIFGROUP32:
2825 	case SIOCGIFGMEMB32:
2826 		ifgr32 = (struct ifgroupreq32 *)data;
2827 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2828 		    sizeof(thunk.ifgr.ifgr_name));
2829 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2830 		switch (cmd) {
2831 		case SIOCAIFGROUP32:
2832 		case SIOCDIFGROUP32:
2833 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2834 			    sizeof(thunk.ifgr.ifgr_group));
2835 			break;
2836 		case SIOCGIFGROUP32:
2837 		case SIOCGIFGMEMB32:
2838 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2839 			break;
2840 		}
2841 		data = (caddr_t)&thunk.ifgr;
2842 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2843 		break;
2844 	case SIOCGIFMEDIA32:
2845 	case SIOCGIFXMEDIA32:
2846 		ifmr32 = (struct ifmediareq32 *)data;
2847 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2848 		    sizeof(thunk.ifmr.ifm_name));
2849 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2850 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2851 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2852 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2853 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2854 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2855 		data = (caddr_t)&thunk.ifmr;
2856 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2857 		break;
2858 	}
2859 #endif
2860 
2861 	switch (cmd) {
2862 	case SIOCGIFCONF:
2863 		error = ifconf(cmd, data);
2864 		goto out_noref;
2865 	}
2866 
2867 	ifr = (struct ifreq *)data;
2868 	switch (cmd) {
2869 #ifdef VIMAGE
2870 	case SIOCSIFRVNET:
2871 		error = priv_check(td, PRIV_NET_SETIFVNET);
2872 		if (error == 0)
2873 			error = if_vmove_reclaim(td, ifr->ifr_name,
2874 			    ifr->ifr_jid);
2875 		goto out_noref;
2876 #endif
2877 	case SIOCIFCREATE:
2878 	case SIOCIFCREATE2:
2879 		error = priv_check(td, PRIV_NET_IFCREATE);
2880 		if (error == 0)
2881 			error = if_clone_create(ifr->ifr_name,
2882 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
2883 			    ifr_data_get_ptr(ifr) : NULL);
2884 		goto out_noref;
2885 	case SIOCIFDESTROY:
2886 		error = priv_check(td, PRIV_NET_IFDESTROY);
2887 
2888 		if (error == 0) {
2889 			sx_xlock(&ifnet_detach_sxlock);
2890 			error = if_clone_destroy(ifr->ifr_name);
2891 			sx_xunlock(&ifnet_detach_sxlock);
2892 		}
2893 		goto out_noref;
2894 
2895 	case SIOCIFGCLONERS:
2896 		error = if_clone_list((struct if_clonereq *)data);
2897 		goto out_noref;
2898 
2899 	case SIOCGIFGMEMB:
2900 	{
2901 		struct ifgroupreq *req;
2902 
2903 		req = (struct ifgroupreq *)data;
2904 		if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) {
2905 			error = EINVAL;
2906 			goto out_noref;
2907 		}
2908 		error = if_getgroupmembers(req);
2909 		goto out_noref;
2910 	}
2911 	}
2912 
2913 	ifp = ifunit_ref(ifr->ifr_name);
2914 	if (ifp == NULL) {
2915 		error = ENXIO;
2916 		goto out_noref;
2917 	}
2918 
2919 	error = ifhwioctl(cmd, ifp, data, td);
2920 	if (error != ENOIOCTL)
2921 		goto out_ref;
2922 
2923 	oif_flags = ifp->if_flags;
2924 	if (so->so_proto == NULL) {
2925 		error = EOPNOTSUPP;
2926 		goto out_ref;
2927 	}
2928 
2929 	/*
2930 	 * Pass the request on to the socket control method, and if the
2931 	 * latter returns EOPNOTSUPP, directly to the interface.
2932 	 *
2933 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2934 	 * trust SIOCSIFADDR et al to come from an already privileged
2935 	 * layer, and do not perform any credentials checks or input
2936 	 * validation.
2937 	 */
2938 	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
2939 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2940 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2941 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2942 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2943 
2944 	if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP))
2945 		if_up(ifp);
2946 out_ref:
2947 	if_rele(ifp);
2948 out_noref:
2949 	CURVNET_RESTORE();
2950 #ifdef COMPAT_FREEBSD32
2951 	if (error != 0)
2952 		return (error);
2953 	switch (saved_cmd) {
2954 	case SIOCGIFCONF32:
2955 		ifc32->ifc_len = thunk.ifc.ifc_len;
2956 		break;
2957 	case SIOCGDRVSPEC32:
2958 		/*
2959 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
2960 		 * the struct so just assert that ifd_len (the only
2961 		 * field it might make sense to update) hasn't
2962 		 * changed.
2963 		 */
2964 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
2965 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
2966 			thunk.ifd.ifd_len));
2967 		break;
2968 	case SIOCGIFGROUP32:
2969 	case SIOCGIFGMEMB32:
2970 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
2971 		break;
2972 	case SIOCGIFMEDIA32:
2973 	case SIOCGIFXMEDIA32:
2974 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
2975 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
2976 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
2977 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
2978 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
2979 		break;
2980 	}
2981 #endif
2982 	return (error);
2983 }
2984 
2985 int
if_rename(struct ifnet * ifp,char * new_name)2986 if_rename(struct ifnet *ifp, char *new_name)
2987 {
2988 	struct ifaddr *ifa;
2989 	struct sockaddr_dl *sdl;
2990 	size_t namelen, onamelen;
2991 	char old_name[IFNAMSIZ];
2992 	char strbuf[IFNAMSIZ + 8];
2993 
2994 	if (new_name[0] == '\0')
2995 		return (EINVAL);
2996 	if (strcmp(new_name, ifp->if_xname) == 0)
2997 		return (0);
2998 	if (ifunit(new_name) != NULL)
2999 		return (EEXIST);
3000 
3001 	if_printf(ifp, "changing name to '%s'\n", new_name);
3002 
3003 	IF_ADDR_WLOCK(ifp);
3004 	strlcpy(old_name, ifp->if_xname, sizeof(old_name));
3005 	strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
3006 	ifa = ifp->if_addr;
3007 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3008 	namelen = strlen(new_name);
3009 	onamelen = sdl->sdl_nlen;
3010 	/*
3011 	 * Move the address if needed.  This is safe because we
3012 	 * allocate space for a name of length IFNAMSIZ when we
3013 	 * create this in if_attach().
3014 	 */
3015 	if (namelen != onamelen) {
3016 		bcopy(sdl->sdl_data + onamelen,
3017 		    sdl->sdl_data + namelen, sdl->sdl_alen);
3018 	}
3019 	bcopy(new_name, sdl->sdl_data, namelen);
3020 	sdl->sdl_nlen = namelen;
3021 	sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
3022 	bzero(sdl->sdl_data, onamelen);
3023 	while (namelen != 0)
3024 		sdl->sdl_data[--namelen] = 0xff;
3025 	IF_ADDR_WUNLOCK(ifp);
3026 
3027 	EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name);
3028 
3029 	snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
3030 	devctl_notify("IFNET", old_name, "RENAME", strbuf);
3031 
3032 	return (0);
3033 }
3034 
3035 /*
3036  * The code common to handling reference counted flags,
3037  * e.g., in ifpromisc() and if_allmulti().
3038  * The "pflag" argument can specify a permanent mode flag to check,
3039  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3040  *
3041  * Only to be used on stack-owned flags, not driver-owned flags.
3042  */
3043 static int
if_setflag(struct ifnet * ifp,int flag,int pflag,int * refcount,int onswitch)3044 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3045 {
3046 	struct ifreq ifr;
3047 	int error;
3048 	int oldflags, oldcount;
3049 
3050 	/* Sanity checks to catch programming errors */
3051 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3052 	    ("%s: setting driver-owned flag %d", __func__, flag));
3053 
3054 	if (onswitch)
3055 		KASSERT(*refcount >= 0,
3056 		    ("%s: increment negative refcount %d for flag %d",
3057 		    __func__, *refcount, flag));
3058 	else
3059 		KASSERT(*refcount > 0,
3060 		    ("%s: decrement non-positive refcount %d for flag %d",
3061 		    __func__, *refcount, flag));
3062 
3063 	/* In case this mode is permanent, just touch refcount */
3064 	if (ifp->if_flags & pflag) {
3065 		*refcount += onswitch ? 1 : -1;
3066 		return (0);
3067 	}
3068 
3069 	/* Save ifnet parameters for if_ioctl() may fail */
3070 	oldcount = *refcount;
3071 	oldflags = ifp->if_flags;
3072 
3073 	/*
3074 	 * See if we aren't the only and touching refcount is enough.
3075 	 * Actually toggle interface flag if we are the first or last.
3076 	 */
3077 	if (onswitch) {
3078 		if ((*refcount)++)
3079 			return (0);
3080 		ifp->if_flags |= flag;
3081 	} else {
3082 		if (--(*refcount))
3083 			return (0);
3084 		ifp->if_flags &= ~flag;
3085 	}
3086 
3087 	/* Call down the driver since we've changed interface flags */
3088 	if (ifp->if_ioctl == NULL) {
3089 		error = EOPNOTSUPP;
3090 		goto recover;
3091 	}
3092 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3093 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3094 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3095 	if (error)
3096 		goto recover;
3097 	/* Notify userland that interface flags have changed */
3098 	rt_ifmsg(ifp, flag);
3099 	return (0);
3100 
3101 recover:
3102 	/* Recover after driver error */
3103 	*refcount = oldcount;
3104 	ifp->if_flags = oldflags;
3105 	return (error);
3106 }
3107 
3108 /*
3109  * Set/clear promiscuous mode on interface ifp based on the truth value
3110  * of pswitch.  The calls are reference counted so that only the first
3111  * "on" request actually has an effect, as does the final "off" request.
3112  * Results are undefined if the "off" and "on" requests are not matched.
3113  */
3114 int
ifpromisc(struct ifnet * ifp,int pswitch)3115 ifpromisc(struct ifnet *ifp, int pswitch)
3116 {
3117 	int error;
3118 	int oldflags = ifp->if_flags;
3119 
3120 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3121 			   &ifp->if_pcount, pswitch);
3122 	/* If promiscuous mode status has changed, log a message */
3123 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3124             log_promisc_mode_change)
3125 		if_printf(ifp, "promiscuous mode %s\n",
3126 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3127 	return (error);
3128 }
3129 
3130 /*
3131  * Return interface configuration
3132  * of system.  List may be used
3133  * in later ioctl's (above) to get
3134  * other information.
3135  */
3136 /*ARGSUSED*/
3137 static int
ifconf(u_long cmd,caddr_t data)3138 ifconf(u_long cmd, caddr_t data)
3139 {
3140 	struct ifconf *ifc = (struct ifconf *)data;
3141 	struct ifnet *ifp;
3142 	struct ifaddr *ifa;
3143 	struct ifreq ifr;
3144 	struct sbuf *sb;
3145 	int error, full = 0, valid_len, max_len;
3146 
3147 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3148 	max_len = maxphys - 1;
3149 
3150 	/* Prevent hostile input from being able to crash the system */
3151 	if (ifc->ifc_len <= 0)
3152 		return (EINVAL);
3153 
3154 again:
3155 	if (ifc->ifc_len <= max_len) {
3156 		max_len = ifc->ifc_len;
3157 		full = 1;
3158 	}
3159 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3160 	max_len = 0;
3161 	valid_len = 0;
3162 
3163 	IFNET_RLOCK();
3164 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3165 		struct epoch_tracker et;
3166 		int addrs;
3167 
3168 		/*
3169 		 * Zero the ifr to make sure we don't disclose the contents
3170 		 * of the stack.
3171 		 */
3172 		memset(&ifr, 0, sizeof(ifr));
3173 
3174 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3175 		    >= sizeof(ifr.ifr_name)) {
3176 			sbuf_delete(sb);
3177 			IFNET_RUNLOCK();
3178 			return (ENAMETOOLONG);
3179 		}
3180 
3181 		addrs = 0;
3182 		NET_EPOCH_ENTER(et);
3183 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3184 			struct sockaddr *sa = ifa->ifa_addr;
3185 
3186 			if (prison_if(curthread->td_ucred, sa) != 0)
3187 				continue;
3188 			addrs++;
3189 			if (sa->sa_len <= sizeof(*sa)) {
3190 				if (sa->sa_len < sizeof(*sa)) {
3191 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3192 					    sizeof(ifr.ifr_ifru.ifru_addr));
3193 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3194 					    sa->sa_len);
3195 				} else
3196 					ifr.ifr_ifru.ifru_addr = *sa;
3197 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3198 				max_len += sizeof(ifr);
3199 			} else {
3200 				sbuf_bcat(sb, &ifr,
3201 				    offsetof(struct ifreq, ifr_addr));
3202 				max_len += offsetof(struct ifreq, ifr_addr);
3203 				sbuf_bcat(sb, sa, sa->sa_len);
3204 				max_len += sa->sa_len;
3205 			}
3206 
3207 			if (sbuf_error(sb) == 0)
3208 				valid_len = sbuf_len(sb);
3209 		}
3210 		NET_EPOCH_EXIT(et);
3211 		if (addrs == 0) {
3212 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3213 			max_len += sizeof(ifr);
3214 
3215 			if (sbuf_error(sb) == 0)
3216 				valid_len = sbuf_len(sb);
3217 		}
3218 	}
3219 	IFNET_RUNLOCK();
3220 
3221 	/*
3222 	 * If we didn't allocate enough space (uncommon), try again.  If
3223 	 * we have already allocated as much space as we are allowed,
3224 	 * return what we've got.
3225 	 */
3226 	if (valid_len != max_len && !full) {
3227 		sbuf_delete(sb);
3228 		goto again;
3229 	}
3230 
3231 	ifc->ifc_len = valid_len;
3232 	sbuf_finish(sb);
3233 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3234 	sbuf_delete(sb);
3235 	return (error);
3236 }
3237 
3238 /*
3239  * Just like ifpromisc(), but for all-multicast-reception mode.
3240  */
3241 int
if_allmulti(struct ifnet * ifp,int onswitch)3242 if_allmulti(struct ifnet *ifp, int onswitch)
3243 {
3244 
3245 	return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount,
3246 	    onswitch));
3247 }
3248 
3249 struct ifmultiaddr *
if_findmulti(struct ifnet * ifp,const struct sockaddr * sa)3250 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3251 {
3252 	struct ifmultiaddr *ifma;
3253 
3254 	IF_ADDR_LOCK_ASSERT(ifp);
3255 
3256 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3257 		if (sa->sa_family == AF_LINK) {
3258 			if (sa_dl_equal(ifma->ifma_addr, sa))
3259 				break;
3260 		} else {
3261 			if (sa_equal(ifma->ifma_addr, sa))
3262 				break;
3263 		}
3264 	}
3265 
3266 	return ifma;
3267 }
3268 
3269 /*
3270  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3271  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3272  * the ifnet multicast address list here, so the caller must do that and
3273  * other setup work (such as notifying the device driver).  The reference
3274  * count is initialized to 1.
3275  */
3276 static struct ifmultiaddr *
if_allocmulti(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr * llsa,int mflags)3277 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3278     int mflags)
3279 {
3280 	struct ifmultiaddr *ifma;
3281 	struct sockaddr *dupsa;
3282 
3283 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3284 	    M_ZERO);
3285 	if (ifma == NULL)
3286 		return (NULL);
3287 
3288 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3289 	if (dupsa == NULL) {
3290 		free(ifma, M_IFMADDR);
3291 		return (NULL);
3292 	}
3293 	bcopy(sa, dupsa, sa->sa_len);
3294 	ifma->ifma_addr = dupsa;
3295 
3296 	ifma->ifma_ifp = ifp;
3297 	ifma->ifma_refcount = 1;
3298 	ifma->ifma_protospec = NULL;
3299 
3300 	if (llsa == NULL) {
3301 		ifma->ifma_lladdr = NULL;
3302 		return (ifma);
3303 	}
3304 
3305 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3306 	if (dupsa == NULL) {
3307 		free(ifma->ifma_addr, M_IFMADDR);
3308 		free(ifma, M_IFMADDR);
3309 		return (NULL);
3310 	}
3311 	bcopy(llsa, dupsa, llsa->sa_len);
3312 	ifma->ifma_lladdr = dupsa;
3313 
3314 	return (ifma);
3315 }
3316 
3317 /*
3318  * if_freemulti: free ifmultiaddr structure and possibly attached related
3319  * addresses.  The caller is responsible for implementing reference
3320  * counting, notifying the driver, handling routing messages, and releasing
3321  * any dependent link layer state.
3322  */
3323 #ifdef MCAST_VERBOSE
3324 extern void kdb_backtrace(void);
3325 #endif
3326 static void
if_freemulti_internal(struct ifmultiaddr * ifma)3327 if_freemulti_internal(struct ifmultiaddr *ifma)
3328 {
3329 
3330 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3331 	    ifma->ifma_refcount));
3332 
3333 	if (ifma->ifma_lladdr != NULL)
3334 		free(ifma->ifma_lladdr, M_IFMADDR);
3335 #ifdef MCAST_VERBOSE
3336 	kdb_backtrace();
3337 	printf("%s freeing ifma: %p\n", __func__, ifma);
3338 #endif
3339 	free(ifma->ifma_addr, M_IFMADDR);
3340 	free(ifma, M_IFMADDR);
3341 }
3342 
3343 static void
if_destroymulti(epoch_context_t ctx)3344 if_destroymulti(epoch_context_t ctx)
3345 {
3346 	struct ifmultiaddr *ifma;
3347 
3348 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3349 	if_freemulti_internal(ifma);
3350 }
3351 
3352 void
if_freemulti(struct ifmultiaddr * ifma)3353 if_freemulti(struct ifmultiaddr *ifma)
3354 {
3355 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3356 	    ifma->ifma_refcount));
3357 
3358 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3359 }
3360 
3361 /*
3362  * Register an additional multicast address with a network interface.
3363  *
3364  * - If the address is already present, bump the reference count on the
3365  *   address and return.
3366  * - If the address is not link-layer, look up a link layer address.
3367  * - Allocate address structures for one or both addresses, and attach to the
3368  *   multicast address list on the interface.  If automatically adding a link
3369  *   layer address, the protocol address will own a reference to the link
3370  *   layer address, to be freed when it is freed.
3371  * - Notify the network device driver of an addition to the multicast address
3372  *   list.
3373  *
3374  * 'sa' points to caller-owned memory with the desired multicast address.
3375  *
3376  * 'retifma' will be used to return a pointer to the resulting multicast
3377  * address reference, if desired.
3378  */
3379 int
if_addmulti(struct ifnet * ifp,struct sockaddr * sa,struct ifmultiaddr ** retifma)3380 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3381     struct ifmultiaddr **retifma)
3382 {
3383 	struct ifmultiaddr *ifma, *ll_ifma;
3384 	struct sockaddr *llsa;
3385 	struct sockaddr_dl sdl;
3386 	int error;
3387 
3388 #ifdef INET
3389 	IN_MULTI_LIST_UNLOCK_ASSERT();
3390 #endif
3391 #ifdef INET6
3392 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3393 #endif
3394 	/*
3395 	 * If the address is already present, return a new reference to it;
3396 	 * otherwise, allocate storage and set up a new address.
3397 	 */
3398 	IF_ADDR_WLOCK(ifp);
3399 	ifma = if_findmulti(ifp, sa);
3400 	if (ifma != NULL) {
3401 		ifma->ifma_refcount++;
3402 		if (retifma != NULL)
3403 			*retifma = ifma;
3404 		IF_ADDR_WUNLOCK(ifp);
3405 		return (0);
3406 	}
3407 
3408 	/*
3409 	 * The address isn't already present; resolve the protocol address
3410 	 * into a link layer address, and then look that up, bump its
3411 	 * refcount or allocate an ifma for that also.
3412 	 * Most link layer resolving functions returns address data which
3413 	 * fits inside default sockaddr_dl structure. However callback
3414 	 * can allocate another sockaddr structure, in that case we need to
3415 	 * free it later.
3416 	 */
3417 	llsa = NULL;
3418 	ll_ifma = NULL;
3419 	if (ifp->if_resolvemulti != NULL) {
3420 		/* Provide called function with buffer size information */
3421 		sdl.sdl_len = sizeof(sdl);
3422 		llsa = (struct sockaddr *)&sdl;
3423 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3424 		if (error)
3425 			goto unlock_out;
3426 	}
3427 
3428 	/*
3429 	 * Allocate the new address.  Don't hook it up yet, as we may also
3430 	 * need to allocate a link layer multicast address.
3431 	 */
3432 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3433 	if (ifma == NULL) {
3434 		error = ENOMEM;
3435 		goto free_llsa_out;
3436 	}
3437 
3438 	/*
3439 	 * If a link layer address is found, we'll need to see if it's
3440 	 * already present in the address list, or allocate is as well.
3441 	 * When this block finishes, the link layer address will be on the
3442 	 * list.
3443 	 */
3444 	if (llsa != NULL) {
3445 		ll_ifma = if_findmulti(ifp, llsa);
3446 		if (ll_ifma == NULL) {
3447 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3448 			if (ll_ifma == NULL) {
3449 				--ifma->ifma_refcount;
3450 				if_freemulti(ifma);
3451 				error = ENOMEM;
3452 				goto free_llsa_out;
3453 			}
3454 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3455 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3456 			    ifma_link);
3457 		} else
3458 			ll_ifma->ifma_refcount++;
3459 		ifma->ifma_llifma = ll_ifma;
3460 	}
3461 
3462 	/*
3463 	 * We now have a new multicast address, ifma, and possibly a new or
3464 	 * referenced link layer address.  Add the primary address to the
3465 	 * ifnet address list.
3466 	 */
3467 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3468 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3469 
3470 	if (retifma != NULL)
3471 		*retifma = ifma;
3472 
3473 	/*
3474 	 * Must generate the message while holding the lock so that 'ifma'
3475 	 * pointer is still valid.
3476 	 */
3477 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3478 	IF_ADDR_WUNLOCK(ifp);
3479 
3480 	/*
3481 	 * We are certain we have added something, so call down to the
3482 	 * interface to let them know about it.
3483 	 */
3484 	if (ifp->if_ioctl != NULL) {
3485 		if (THREAD_CAN_SLEEP())
3486 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3487 		else
3488 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3489 	}
3490 
3491 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3492 		link_free_sdl(llsa);
3493 
3494 	return (0);
3495 
3496 free_llsa_out:
3497 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3498 		link_free_sdl(llsa);
3499 
3500 unlock_out:
3501 	IF_ADDR_WUNLOCK(ifp);
3502 	return (error);
3503 }
3504 
3505 static void
if_siocaddmulti(void * arg,int pending)3506 if_siocaddmulti(void *arg, int pending)
3507 {
3508 	struct ifnet *ifp;
3509 
3510 	ifp = arg;
3511 #ifdef DIAGNOSTIC
3512 	if (pending > 1)
3513 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3514 #endif
3515 	CURVNET_SET(ifp->if_vnet);
3516 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3517 	CURVNET_RESTORE();
3518 }
3519 
3520 /*
3521  * Delete a multicast group membership by network-layer group address.
3522  *
3523  * Returns ENOENT if the entry could not be found. If ifp no longer
3524  * exists, results are undefined. This entry point should only be used
3525  * from subsystems which do appropriate locking to hold ifp for the
3526  * duration of the call.
3527  * Network-layer protocol domains must use if_delmulti_ifma().
3528  */
3529 int
if_delmulti(struct ifnet * ifp,struct sockaddr * sa)3530 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3531 {
3532 	struct ifmultiaddr *ifma;
3533 	int lastref;
3534 
3535 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3536 
3537 	IF_ADDR_WLOCK(ifp);
3538 	lastref = 0;
3539 	ifma = if_findmulti(ifp, sa);
3540 	if (ifma != NULL)
3541 		lastref = if_delmulti_locked(ifp, ifma, 0);
3542 	IF_ADDR_WUNLOCK(ifp);
3543 
3544 	if (ifma == NULL)
3545 		return (ENOENT);
3546 
3547 	if (lastref && ifp->if_ioctl != NULL) {
3548 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3549 	}
3550 
3551 	return (0);
3552 }
3553 
3554 /*
3555  * Delete all multicast group membership for an interface.
3556  * Should be used to quickly flush all multicast filters.
3557  */
3558 void
if_delallmulti(struct ifnet * ifp)3559 if_delallmulti(struct ifnet *ifp)
3560 {
3561 	struct ifmultiaddr *ifma;
3562 	struct ifmultiaddr *next;
3563 
3564 	IF_ADDR_WLOCK(ifp);
3565 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3566 		if_delmulti_locked(ifp, ifma, 0);
3567 	IF_ADDR_WUNLOCK(ifp);
3568 }
3569 
3570 void
if_delmulti_ifma(struct ifmultiaddr * ifma)3571 if_delmulti_ifma(struct ifmultiaddr *ifma)
3572 {
3573 	if_delmulti_ifma_flags(ifma, 0);
3574 }
3575 
3576 /*
3577  * Delete a multicast group membership by group membership pointer.
3578  * Network-layer protocol domains must use this routine.
3579  *
3580  * It is safe to call this routine if the ifp disappeared.
3581  */
3582 void
if_delmulti_ifma_flags(struct ifmultiaddr * ifma,int flags)3583 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3584 {
3585 	struct ifnet *ifp;
3586 	int lastref;
3587 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3588 #ifdef INET
3589 	IN_MULTI_LIST_UNLOCK_ASSERT();
3590 #endif
3591 	ifp = ifma->ifma_ifp;
3592 #ifdef DIAGNOSTIC
3593 	if (ifp == NULL) {
3594 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3595 	} else {
3596 		struct epoch_tracker et;
3597 		struct ifnet *oifp;
3598 
3599 		NET_EPOCH_ENTER(et);
3600 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3601 			if (ifp == oifp)
3602 				break;
3603 		NET_EPOCH_EXIT(et);
3604 		if (ifp != oifp)
3605 			ifp = NULL;
3606 	}
3607 #endif
3608 	/*
3609 	 * If and only if the ifnet instance exists: Acquire the address lock.
3610 	 */
3611 	if (ifp != NULL)
3612 		IF_ADDR_WLOCK(ifp);
3613 
3614 	lastref = if_delmulti_locked(ifp, ifma, flags);
3615 
3616 	if (ifp != NULL) {
3617 		/*
3618 		 * If and only if the ifnet instance exists:
3619 		 *  Release the address lock.
3620 		 *  If the group was left: update the hardware hash filter.
3621 		 */
3622 		IF_ADDR_WUNLOCK(ifp);
3623 		if (lastref && ifp->if_ioctl != NULL) {
3624 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3625 		}
3626 	}
3627 }
3628 
3629 /*
3630  * Perform deletion of network-layer and/or link-layer multicast address.
3631  *
3632  * Return 0 if the reference count was decremented.
3633  * Return 1 if the final reference was released, indicating that the
3634  * hardware hash filter should be reprogrammed.
3635  */
3636 static int
if_delmulti_locked(struct ifnet * ifp,struct ifmultiaddr * ifma,int detaching)3637 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3638 {
3639 	struct ifmultiaddr *ll_ifma;
3640 
3641 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3642 		KASSERT(ifma->ifma_ifp == ifp,
3643 		    ("%s: inconsistent ifp %p", __func__, ifp));
3644 		IF_ADDR_WLOCK_ASSERT(ifp);
3645 	}
3646 
3647 	ifp = ifma->ifma_ifp;
3648 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3649 
3650 	/*
3651 	 * If the ifnet is detaching, null out references to ifnet,
3652 	 * so that upper protocol layers will notice, and not attempt
3653 	 * to obtain locks for an ifnet which no longer exists. The
3654 	 * routing socket announcement must happen before the ifnet
3655 	 * instance is detached from the system.
3656 	 */
3657 	if (detaching) {
3658 #ifdef DIAGNOSTIC
3659 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3660 #endif
3661 		/*
3662 		 * ifp may already be nulled out if we are being reentered
3663 		 * to delete the ll_ifma.
3664 		 */
3665 		if (ifp != NULL) {
3666 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3667 			ifma->ifma_ifp = NULL;
3668 		}
3669 	}
3670 
3671 	if (--ifma->ifma_refcount > 0)
3672 		return 0;
3673 
3674 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3675 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3676 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3677 	}
3678 	/*
3679 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3680 	 * have been associated with it. Release it first if so.
3681 	 */
3682 	ll_ifma = ifma->ifma_llifma;
3683 	if (ll_ifma != NULL) {
3684 		KASSERT(ifma->ifma_lladdr != NULL,
3685 		    ("%s: llifma w/o lladdr", __func__));
3686 		if (detaching)
3687 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3688 		if (--ll_ifma->ifma_refcount == 0) {
3689 			if (ifp != NULL) {
3690 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3691 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3692 						ifma_link);
3693 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3694 				}
3695 			}
3696 			if_freemulti(ll_ifma);
3697 		}
3698 	}
3699 #ifdef INVARIANTS
3700 	if (ifp) {
3701 		struct ifmultiaddr *ifmatmp;
3702 
3703 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3704 			MPASS(ifma != ifmatmp);
3705 	}
3706 #endif
3707 	if_freemulti(ifma);
3708 	/*
3709 	 * The last reference to this instance of struct ifmultiaddr
3710 	 * was released; the hardware should be notified of this change.
3711 	 */
3712 	return 1;
3713 }
3714 
3715 /*
3716  * Set the link layer address on an interface.
3717  *
3718  * At this time we only support certain types of interfaces,
3719  * and we don't allow the length of the address to change.
3720  *
3721  * Set noinline to be dtrace-friendly
3722  */
3723 __noinline int
if_setlladdr(struct ifnet * ifp,const u_char * lladdr,int len)3724 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3725 {
3726 	struct sockaddr_dl *sdl;
3727 	struct ifaddr *ifa;
3728 	struct ifreq ifr;
3729 
3730 	ifa = ifp->if_addr;
3731 	if (ifa == NULL)
3732 		return (EINVAL);
3733 
3734 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3735 	if (sdl == NULL)
3736 		return (EINVAL);
3737 
3738 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3739 		return (EINVAL);
3740 
3741 	switch (ifp->if_type) {
3742 	case IFT_ETHER:
3743 	case IFT_XETHER:
3744 	case IFT_L2VLAN:
3745 	case IFT_BRIDGE:
3746 	case IFT_IEEE8023ADLAG:
3747 		bcopy(lladdr, LLADDR(sdl), len);
3748 		break;
3749 	default:
3750 		return (ENODEV);
3751 	}
3752 
3753 	/*
3754 	 * If the interface is already up, we need
3755 	 * to re-init it in order to reprogram its
3756 	 * address filter.
3757 	 */
3758 	if ((ifp->if_flags & IFF_UP) != 0) {
3759 		if (ifp->if_ioctl) {
3760 			ifp->if_flags &= ~IFF_UP;
3761 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3762 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3763 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3764 			ifp->if_flags |= IFF_UP;
3765 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3766 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3767 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3768 		}
3769 	}
3770 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3771 
3772 	return (0);
3773 }
3774 
3775 /*
3776  * Compat function for handling basic encapsulation requests.
3777  * Not converted stacks (FDDI, IB, ..) supports traditional
3778  * output model: ARP (and other similar L2 protocols) are handled
3779  * inside output routine, arpresolve/nd6_resolve() returns MAC
3780  * address instead of full prepend.
3781  *
3782  * This function creates calculated header==MAC for IPv4/IPv6 and
3783  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3784  * address families.
3785  */
3786 static int
if_requestencap_default(struct ifnet * ifp,struct if_encap_req * req)3787 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3788 {
3789 	if (req->rtype != IFENCAP_LL)
3790 		return (EOPNOTSUPP);
3791 
3792 	if (req->bufsize < req->lladdr_len)
3793 		return (ENOMEM);
3794 
3795 	switch (req->family) {
3796 	case AF_INET:
3797 	case AF_INET6:
3798 		break;
3799 	default:
3800 		return (EAFNOSUPPORT);
3801 	}
3802 
3803 	/* Copy lladdr to storage as is */
3804 	memmove(req->buf, req->lladdr, req->lladdr_len);
3805 	req->bufsize = req->lladdr_len;
3806 	req->lladdr_off = 0;
3807 
3808 	return (0);
3809 }
3810 
3811 /*
3812  * Tunnel interfaces can nest, also they may cause infinite recursion
3813  * calls when misconfigured. We'll prevent this by detecting loops.
3814  * High nesting level may cause stack exhaustion. We'll prevent this
3815  * by introducing upper limit.
3816  *
3817  * Return 0, if tunnel nesting count is equal or less than limit.
3818  */
3819 int
if_tunnel_check_nesting(struct ifnet * ifp,struct mbuf * m,uint32_t cookie,int limit)3820 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3821     int limit)
3822 {
3823 	struct m_tag *mtag;
3824 	int count;
3825 
3826 	count = 1;
3827 	mtag = NULL;
3828 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3829 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3830 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3831 			return (EIO);
3832 		}
3833 		count++;
3834 	}
3835 	if (count > limit) {
3836 		log(LOG_NOTICE,
3837 		    "%s: if_output recursively called too many times(%d)\n",
3838 		    if_name(ifp), count);
3839 		return (EIO);
3840 	}
3841 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3842 	if (mtag == NULL)
3843 		return (ENOMEM);
3844 	*(struct ifnet **)(mtag + 1) = ifp;
3845 	m_tag_prepend(m, mtag);
3846 	return (0);
3847 }
3848 
3849 /*
3850  * Get the link layer address that was read from the hardware at attach.
3851  *
3852  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3853  * their component interfaces as IFT_IEEE8023ADLAG.
3854  */
3855 int
if_gethwaddr(struct ifnet * ifp,struct ifreq * ifr)3856 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3857 {
3858 	if (ifp->if_hw_addr == NULL)
3859 		return (ENODEV);
3860 
3861 	switch (ifp->if_type) {
3862 	case IFT_ETHER:
3863 	case IFT_IEEE8023ADLAG:
3864 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3865 		return (0);
3866 	default:
3867 		return (ENODEV);
3868 	}
3869 }
3870 
3871 /*
3872  * The name argument must be a pointer to storage which will last as
3873  * long as the interface does.  For physical devices, the result of
3874  * device_get_name(dev) is a good choice and for pseudo-devices a
3875  * static string works well.
3876  */
3877 void
if_initname(struct ifnet * ifp,const char * name,int unit)3878 if_initname(struct ifnet *ifp, const char *name, int unit)
3879 {
3880 	ifp->if_dname = name;
3881 	ifp->if_dunit = unit;
3882 	if (unit != IF_DUNIT_NONE)
3883 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3884 	else
3885 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3886 }
3887 
3888 static int
if_vlog(struct ifnet * ifp,int pri,const char * fmt,va_list ap)3889 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
3890 {
3891 	char if_fmt[256];
3892 
3893 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3894 	vlog(pri, if_fmt, ap);
3895 	return (0);
3896 }
3897 
3898 
3899 int
if_printf(struct ifnet * ifp,const char * fmt,...)3900 if_printf(struct ifnet *ifp, const char *fmt, ...)
3901 {
3902 	va_list ap;
3903 
3904 	va_start(ap, fmt);
3905 	if_vlog(ifp, LOG_INFO, fmt, ap);
3906 	va_end(ap);
3907 	return (0);
3908 }
3909 
3910 int
if_log(struct ifnet * ifp,int pri,const char * fmt,...)3911 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
3912 {
3913 	va_list ap;
3914 
3915 	va_start(ap, fmt);
3916 	if_vlog(ifp, pri, fmt, ap);
3917 	va_end(ap);
3918 	return (0);
3919 }
3920 
3921 void
if_start(struct ifnet * ifp)3922 if_start(struct ifnet *ifp)
3923 {
3924 
3925 	(*(ifp)->if_start)(ifp);
3926 }
3927 
3928 /*
3929  * Backwards compatibility interface for drivers
3930  * that have not implemented it
3931  */
3932 static int
if_transmit_default(struct ifnet * ifp,struct mbuf * m)3933 if_transmit_default(struct ifnet *ifp, struct mbuf *m)
3934 {
3935 	int error;
3936 
3937 	IFQ_HANDOFF(ifp, m, error);
3938 	return (error);
3939 }
3940 
3941 static void
if_input_default(struct ifnet * ifp __unused,struct mbuf * m)3942 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3943 {
3944 	m_freem(m);
3945 }
3946 
3947 int
if_handoff(struct ifqueue * ifq,struct mbuf * m,struct ifnet * ifp,int adjust)3948 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3949 {
3950 	int active = 0;
3951 
3952 	IF_LOCK(ifq);
3953 	if (_IF_QFULL(ifq)) {
3954 		IF_UNLOCK(ifq);
3955 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3956 		m_freem(m);
3957 		return (0);
3958 	}
3959 	if (ifp != NULL) {
3960 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3961 		if (m->m_flags & (M_BCAST|M_MCAST))
3962 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3963 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3964 	}
3965 	_IF_ENQUEUE(ifq, m);
3966 	IF_UNLOCK(ifq);
3967 	if (ifp != NULL && !active)
3968 		(*(ifp)->if_start)(ifp);
3969 	return (1);
3970 }
3971 
3972 void
if_register_com_alloc(u_char type,if_com_alloc_t * a,if_com_free_t * f)3973 if_register_com_alloc(u_char type,
3974     if_com_alloc_t *a, if_com_free_t *f)
3975 {
3976 
3977 	KASSERT(if_com_alloc[type] == NULL,
3978 	    ("if_register_com_alloc: %d already registered", type));
3979 	KASSERT(if_com_free[type] == NULL,
3980 	    ("if_register_com_alloc: %d free already registered", type));
3981 
3982 	if_com_alloc[type] = a;
3983 	if_com_free[type] = f;
3984 }
3985 
3986 void
if_deregister_com_alloc(u_char type)3987 if_deregister_com_alloc(u_char type)
3988 {
3989 
3990 	KASSERT(if_com_alloc[type] != NULL,
3991 	    ("if_deregister_com_alloc: %d not registered", type));
3992 	KASSERT(if_com_free[type] != NULL,
3993 	    ("if_deregister_com_alloc: %d free not registered", type));
3994 
3995 	/*
3996 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
3997 	 * fixes issues about late invocation of if_destroy(), which leads
3998 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
3999 	 */
4000 	NET_EPOCH_DRAIN_CALLBACKS();
4001 
4002 	if_com_alloc[type] = NULL;
4003 	if_com_free[type] = NULL;
4004 }
4005 
4006 /* API for driver access to network stack owned ifnet.*/
4007 uint64_t
if_setbaudrate(struct ifnet * ifp,uint64_t baudrate)4008 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4009 {
4010 	uint64_t oldbrate;
4011 
4012 	oldbrate = ifp->if_baudrate;
4013 	ifp->if_baudrate = baudrate;
4014 	return (oldbrate);
4015 }
4016 
4017 uint64_t
if_getbaudrate(const if_t ifp)4018 if_getbaudrate(const if_t ifp)
4019 {
4020 	return (ifp->if_baudrate);
4021 }
4022 
4023 int
if_setcapabilities(if_t ifp,int capabilities)4024 if_setcapabilities(if_t ifp, int capabilities)
4025 {
4026 	ifp->if_capabilities = capabilities;
4027 	return (0);
4028 }
4029 
4030 int
if_setcapabilitiesbit(if_t ifp,int setbit,int clearbit)4031 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4032 {
4033 	ifp->if_capabilities &= ~clearbit;
4034 	ifp->if_capabilities |= setbit;
4035 	return (0);
4036 }
4037 
4038 int
if_getcapabilities(const if_t ifp)4039 if_getcapabilities(const if_t ifp)
4040 {
4041 	return (ifp->if_capabilities);
4042 }
4043 
4044 int
if_setcapenable(if_t ifp,int capabilities)4045 if_setcapenable(if_t ifp, int capabilities)
4046 {
4047 	ifp->if_capenable = capabilities;
4048 	return (0);
4049 }
4050 
4051 int
if_setcapenablebit(if_t ifp,int setcap,int clearcap)4052 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4053 {
4054 	ifp->if_capenable &= ~clearcap;
4055 	ifp->if_capenable |= setcap;
4056 	return (0);
4057 }
4058 
4059 int
if_setcapabilities2(if_t ifp,int capabilities)4060 if_setcapabilities2(if_t ifp, int capabilities)
4061 {
4062 	ifp->if_capabilities2 = capabilities;
4063 	return (0);
4064 }
4065 
4066 int
if_setcapabilities2bit(if_t ifp,int setbit,int clearbit)4067 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit)
4068 {
4069 	ifp->if_capabilities2 &= ~clearbit;
4070 	ifp->if_capabilities2 |= setbit;
4071 	return (0);
4072 }
4073 
4074 int
if_getcapabilities2(const if_t ifp)4075 if_getcapabilities2(const if_t ifp)
4076 {
4077 	return (ifp->if_capabilities2);
4078 }
4079 
4080 int
if_setcapenable2(if_t ifp,int capabilities2)4081 if_setcapenable2(if_t ifp, int capabilities2)
4082 {
4083 	ifp->if_capenable2 = capabilities2;
4084 	return (0);
4085 }
4086 
4087 int
if_setcapenable2bit(if_t ifp,int setcap,int clearcap)4088 if_setcapenable2bit(if_t ifp, int setcap, int clearcap)
4089 {
4090 	ifp->if_capenable2 &= ~clearcap;
4091 	ifp->if_capenable2 |= setcap;
4092 	return (0);
4093 }
4094 
4095 const char *
if_getdname(const if_t ifp)4096 if_getdname(const if_t ifp)
4097 {
4098 	return (ifp->if_dname);
4099 }
4100 
4101 void
if_setdname(if_t ifp,const char * dname)4102 if_setdname(if_t ifp, const char *dname)
4103 {
4104 	ifp->if_dname = dname;
4105 }
4106 
4107 const char *
if_name(if_t ifp)4108 if_name(if_t ifp)
4109 {
4110 	return (ifp->if_xname);
4111 }
4112 
4113 int
if_setname(if_t ifp,const char * name)4114 if_setname(if_t ifp, const char *name)
4115 {
4116 	if (strlen(name) > sizeof(ifp->if_xname) - 1)
4117 		return (ENAMETOOLONG);
4118 	strcpy(ifp->if_xname, name);
4119 
4120 	return (0);
4121 }
4122 
4123 int
if_togglecapenable(if_t ifp,int togglecap)4124 if_togglecapenable(if_t ifp, int togglecap)
4125 {
4126 	ifp->if_capenable ^= togglecap;
4127 	return (0);
4128 }
4129 
4130 int
if_getcapenable(const if_t ifp)4131 if_getcapenable(const if_t ifp)
4132 {
4133 	return (ifp->if_capenable);
4134 }
4135 
4136 int
if_togglecapenable2(if_t ifp,int togglecap)4137 if_togglecapenable2(if_t ifp, int togglecap)
4138 {
4139 	ifp->if_capenable2 ^= togglecap;
4140 	return (0);
4141 }
4142 
4143 int
if_getcapenable2(const if_t ifp)4144 if_getcapenable2(const if_t ifp)
4145 {
4146 	return (ifp->if_capenable2);
4147 }
4148 
4149 int
if_getdunit(const if_t ifp)4150 if_getdunit(const if_t ifp)
4151 {
4152 	return (ifp->if_dunit);
4153 }
4154 
4155 int
if_getindex(const if_t ifp)4156 if_getindex(const if_t ifp)
4157 {
4158 	return (ifp->if_index);
4159 }
4160 
4161 int
if_getidxgen(const if_t ifp)4162 if_getidxgen(const if_t ifp)
4163 {
4164 	return (ifp->if_idxgen);
4165 }
4166 
4167 const char *
if_getdescr(if_t ifp)4168 if_getdescr(if_t ifp)
4169 {
4170 	return (ifp->if_description);
4171 }
4172 
4173 void
if_setdescr(if_t ifp,char * descrbuf)4174 if_setdescr(if_t ifp, char *descrbuf)
4175 {
4176 	sx_xlock(&ifdescr_sx);
4177 	char *odescrbuf = ifp->if_description;
4178 	ifp->if_description = descrbuf;
4179 	sx_xunlock(&ifdescr_sx);
4180 
4181 	if_freedescr(odescrbuf);
4182 }
4183 
4184 char *
if_allocdescr(size_t sz,int malloc_flag)4185 if_allocdescr(size_t sz, int malloc_flag)
4186 {
4187 	malloc_flag &= (M_WAITOK | M_NOWAIT);
4188 	return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag));
4189 }
4190 
4191 void
if_freedescr(char * descrbuf)4192 if_freedescr(char *descrbuf)
4193 {
4194 	free(descrbuf, M_IFDESCR);
4195 }
4196 
4197 int
if_getalloctype(const if_t ifp)4198 if_getalloctype(const if_t ifp)
4199 {
4200 	return (ifp->if_alloctype);
4201 }
4202 
4203 void
if_setlastchange(if_t ifp)4204 if_setlastchange(if_t ifp)
4205 {
4206 	getmicrotime(&ifp->if_lastchange);
4207 }
4208 
4209 /*
4210  * This is largely undesirable because it ties ifnet to a device, but does
4211  * provide flexiblity for an embedded product vendor. Should be used with
4212  * the understanding that it violates the interface boundaries, and should be
4213  * a last resort only.
4214  */
4215 int
if_setdev(if_t ifp,void * dev)4216 if_setdev(if_t ifp, void *dev)
4217 {
4218 	return (0);
4219 }
4220 
4221 int
if_setdrvflagbits(if_t ifp,int set_flags,int clear_flags)4222 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4223 {
4224 	ifp->if_drv_flags &= ~clear_flags;
4225 	ifp->if_drv_flags |= set_flags;
4226 
4227 	return (0);
4228 }
4229 
4230 int
if_getdrvflags(const if_t ifp)4231 if_getdrvflags(const if_t ifp)
4232 {
4233 	return (ifp->if_drv_flags);
4234 }
4235 
4236 int
if_setdrvflags(if_t ifp,int flags)4237 if_setdrvflags(if_t ifp, int flags)
4238 {
4239 	ifp->if_drv_flags = flags;
4240 	return (0);
4241 }
4242 
4243 int
if_setflags(if_t ifp,int flags)4244 if_setflags(if_t ifp, int flags)
4245 {
4246 	ifp->if_flags = flags;
4247 	return (0);
4248 }
4249 
4250 int
if_setflagbits(if_t ifp,int set,int clear)4251 if_setflagbits(if_t ifp, int set, int clear)
4252 {
4253 	ifp->if_flags &= ~clear;
4254 	ifp->if_flags |= set;
4255 	return (0);
4256 }
4257 
4258 int
if_getflags(const if_t ifp)4259 if_getflags(const if_t ifp)
4260 {
4261 	return (ifp->if_flags);
4262 }
4263 
4264 int
if_clearhwassist(if_t ifp)4265 if_clearhwassist(if_t ifp)
4266 {
4267 	ifp->if_hwassist = 0;
4268 	return (0);
4269 }
4270 
4271 int
if_sethwassistbits(if_t ifp,int toset,int toclear)4272 if_sethwassistbits(if_t ifp, int toset, int toclear)
4273 {
4274 	ifp->if_hwassist &= ~toclear;
4275 	ifp->if_hwassist |= toset;
4276 
4277 	return (0);
4278 }
4279 
4280 int
if_sethwassist(if_t ifp,int hwassist_bit)4281 if_sethwassist(if_t ifp, int hwassist_bit)
4282 {
4283 	ifp->if_hwassist = hwassist_bit;
4284 	return (0);
4285 }
4286 
4287 int
if_gethwassist(const if_t ifp)4288 if_gethwassist(const if_t ifp)
4289 {
4290 	return (ifp->if_hwassist);
4291 }
4292 
4293 int
if_togglehwassist(if_t ifp,int toggle_bits)4294 if_togglehwassist(if_t ifp, int toggle_bits)
4295 {
4296 	ifp->if_hwassist ^= toggle_bits;
4297 	return (0);
4298 }
4299 
4300 int
if_setmtu(if_t ifp,int mtu)4301 if_setmtu(if_t ifp, int mtu)
4302 {
4303 	ifp->if_mtu = mtu;
4304 	return (0);
4305 }
4306 
4307 void
if_notifymtu(if_t ifp)4308 if_notifymtu(if_t ifp)
4309 {
4310 #ifdef INET6
4311 	nd6_setmtu(ifp);
4312 #endif
4313 	rt_updatemtu(ifp);
4314 }
4315 
4316 int
if_getmtu(const if_t ifp)4317 if_getmtu(const if_t ifp)
4318 {
4319 	return (ifp->if_mtu);
4320 }
4321 
4322 void
if_setppromisc(if_t ifp,bool ppromisc)4323 if_setppromisc(if_t ifp, bool ppromisc)
4324 {
4325 	int new_flags;
4326 
4327 	if (ppromisc)
4328 		new_flags = ifp->if_flags | IFF_PPROMISC;
4329 	else
4330 		new_flags = ifp->if_flags & ~IFF_PPROMISC;
4331 	if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
4332 		if (new_flags & IFF_PPROMISC)
4333 			new_flags |= IFF_PROMISC;
4334 		/*
4335 		 * Only unset IFF_PROMISC if there are no more consumers of
4336 		 * promiscuity, i.e. the ifp->if_pcount refcount is 0.
4337 		 */
4338 		else if (ifp->if_pcount == 0)
4339 			new_flags &= ~IFF_PROMISC;
4340 		if (log_promisc_mode_change)
4341                         if_printf(ifp, "permanently promiscuous mode %s\n",
4342                             ((new_flags & IFF_PPROMISC) ?
4343                              "enabled" : "disabled"));
4344 	}
4345 	ifp->if_flags = new_flags;
4346 }
4347 
4348 /*
4349  * Methods for drivers to access interface unicast and multicast
4350  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4351  * 'struct ifmultiaddr'.
4352  */
4353 u_int
if_lladdr_count(if_t ifp)4354 if_lladdr_count(if_t ifp)
4355 {
4356 	struct epoch_tracker et;
4357 	struct ifaddr *ifa;
4358 	u_int count;
4359 
4360 	count = 0;
4361 	NET_EPOCH_ENTER(et);
4362 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4363 		if (ifa->ifa_addr->sa_family == AF_LINK)
4364 			count++;
4365 	NET_EPOCH_EXIT(et);
4366 
4367 	return (count);
4368 }
4369 
4370 int
if_foreach(if_foreach_cb_t cb,void * cb_arg)4371 if_foreach(if_foreach_cb_t cb, void *cb_arg)
4372 {
4373 	if_t ifp;
4374 	int error;
4375 
4376 	NET_EPOCH_ASSERT();
4377 	MPASS(cb);
4378 
4379 	error = 0;
4380 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4381 		error = cb(ifp, cb_arg);
4382 		if (error != 0)
4383 			break;
4384 	}
4385 
4386 	return (error);
4387 }
4388 
4389 /*
4390  * Iterates over the list of interfaces, permitting callback function @cb to sleep.
4391  * Stops iteration if @cb returns non-zero error code.
4392  * Returns the last error code from @cb.
4393  * @match_cb: optional match callback limiting the iteration to only matched interfaces
4394  * @match_arg: argument to pass to @match_cb
4395  * @cb: iteration callback
4396  * @cb_arg: argument to pass to @cb
4397  */
4398 int
if_foreach_sleep(if_foreach_match_t match_cb,void * match_arg,if_foreach_cb_t cb,void * cb_arg)4399 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb,
4400     void *cb_arg)
4401 {
4402 	int match_count = 0, array_size = 16; /* 128 bytes for malloc */
4403 	struct ifnet **match_array = NULL;
4404 	int error = 0;
4405 
4406 	MPASS(cb);
4407 
4408 	while (true) {
4409 		struct ifnet **new_array;
4410 		int new_size = array_size;
4411 		struct epoch_tracker et;
4412 		struct ifnet *ifp;
4413 
4414 		while (new_size < match_count)
4415 			new_size *= 2;
4416 		new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK);
4417 		if (match_array != NULL)
4418 			memcpy(new_array, match_array, array_size * sizeof(void *));
4419 		free(match_array, M_TEMP);
4420 		match_array = new_array;
4421 		array_size = new_size;
4422 
4423 		match_count = 0;
4424 		NET_EPOCH_ENTER(et);
4425 		CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4426 			if (match_cb != NULL && !match_cb(ifp, match_arg))
4427 				continue;
4428 			if (match_count < array_size) {
4429 				if (if_try_ref(ifp))
4430 					match_array[match_count++] = ifp;
4431 			} else
4432 				match_count++;
4433 		}
4434 		NET_EPOCH_EXIT(et);
4435 
4436 		if (match_count > array_size) {
4437 			for (int i = 0; i < array_size; i++)
4438 				if_rele(match_array[i]);
4439 			continue;
4440 		} else {
4441 			for (int i = 0; i < match_count; i++) {
4442 				if (error == 0)
4443 					error = cb(match_array[i], cb_arg);
4444 				if_rele(match_array[i]);
4445 			}
4446 			free(match_array, M_TEMP);
4447 			break;
4448 		}
4449 	}
4450 
4451 	return (error);
4452 }
4453 
4454 
4455 /*
4456  * Uses just 1 pointer of the 4 available in the public struct.
4457  */
4458 if_t
if_iter_start(struct if_iter * iter)4459 if_iter_start(struct if_iter *iter)
4460 {
4461 	if_t ifp;
4462 
4463 	NET_EPOCH_ASSERT();
4464 
4465 	bzero(iter, sizeof(*iter));
4466 	ifp = CK_STAILQ_FIRST(&V_ifnet);
4467 	if (ifp != NULL)
4468 		iter->context[0] = CK_STAILQ_NEXT(ifp, if_link);
4469 	else
4470 		iter->context[0] = NULL;
4471 	return (ifp);
4472 }
4473 
4474 if_t
if_iter_next(struct if_iter * iter)4475 if_iter_next(struct if_iter *iter)
4476 {
4477 	if_t cur_ifp = iter->context[0];
4478 
4479 	if (cur_ifp != NULL)
4480 		iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link);
4481 	return (cur_ifp);
4482 }
4483 
4484 void
if_iter_finish(struct if_iter * iter)4485 if_iter_finish(struct if_iter *iter)
4486 {
4487 	/* Nothing to do here for now. */
4488 }
4489 
4490 u_int
if_foreach_lladdr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4491 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4492 {
4493 	struct epoch_tracker et;
4494 	struct ifaddr *ifa;
4495 	u_int count;
4496 
4497 	MPASS(cb);
4498 
4499 	count = 0;
4500 	NET_EPOCH_ENTER(et);
4501 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4502 		if (ifa->ifa_addr->sa_family != AF_LINK)
4503 			continue;
4504 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4505 		    count);
4506 	}
4507 	NET_EPOCH_EXIT(et);
4508 
4509 	return (count);
4510 }
4511 
4512 u_int
if_llmaddr_count(if_t ifp)4513 if_llmaddr_count(if_t ifp)
4514 {
4515 	struct epoch_tracker et;
4516 	struct ifmultiaddr *ifma;
4517 	int count;
4518 
4519 	count = 0;
4520 	NET_EPOCH_ENTER(et);
4521 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4522 		if (ifma->ifma_addr->sa_family == AF_LINK)
4523 			count++;
4524 	NET_EPOCH_EXIT(et);
4525 
4526 	return (count);
4527 }
4528 
4529 bool
if_maddr_empty(if_t ifp)4530 if_maddr_empty(if_t ifp)
4531 {
4532 
4533 	return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs));
4534 }
4535 
4536 u_int
if_foreach_llmaddr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4537 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4538 {
4539 	struct epoch_tracker et;
4540 	struct ifmultiaddr *ifma;
4541 	u_int count;
4542 
4543 	MPASS(cb);
4544 
4545 	count = 0;
4546 	NET_EPOCH_ENTER(et);
4547 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4548 		if (ifma->ifma_addr->sa_family != AF_LINK)
4549 			continue;
4550 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4551 		    count);
4552 	}
4553 	NET_EPOCH_EXIT(et);
4554 
4555 	return (count);
4556 }
4557 
4558 u_int
if_foreach_addr_type(if_t ifp,int type,if_addr_cb_t cb,void * cb_arg)4559 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg)
4560 {
4561 	struct epoch_tracker et;
4562 	struct ifaddr *ifa;
4563 	u_int count;
4564 
4565 	MPASS(cb);
4566 
4567 	count = 0;
4568 	NET_EPOCH_ENTER(et);
4569 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4570 		if (ifa->ifa_addr->sa_family != type)
4571 			continue;
4572 		count += (*cb)(cb_arg, ifa, count);
4573 	}
4574 	NET_EPOCH_EXIT(et);
4575 
4576 	return (count);
4577 }
4578 
4579 struct ifaddr *
ifa_iter_start(if_t ifp,struct ifa_iter * iter)4580 ifa_iter_start(if_t ifp, struct ifa_iter *iter)
4581 {
4582 	struct ifaddr *ifa;
4583 
4584 	NET_EPOCH_ASSERT();
4585 
4586 	bzero(iter, sizeof(*iter));
4587 	ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
4588 	if (ifa != NULL)
4589 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4590 	else
4591 		iter->context[0] = NULL;
4592 	return (ifa);
4593 }
4594 
4595 struct ifaddr *
ifa_iter_next(struct ifa_iter * iter)4596 ifa_iter_next(struct ifa_iter *iter)
4597 {
4598 	struct ifaddr *ifa = iter->context[0];
4599 
4600 	if (ifa != NULL)
4601 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4602 	return (ifa);
4603 }
4604 
4605 void
ifa_iter_finish(struct ifa_iter * iter)4606 ifa_iter_finish(struct ifa_iter *iter)
4607 {
4608 	/* Nothing to do here for now. */
4609 }
4610 
4611 int
if_setsoftc(if_t ifp,void * softc)4612 if_setsoftc(if_t ifp, void *softc)
4613 {
4614 	ifp->if_softc = softc;
4615 	return (0);
4616 }
4617 
4618 void *
if_getsoftc(const if_t ifp)4619 if_getsoftc(const if_t ifp)
4620 {
4621 	return (ifp->if_softc);
4622 }
4623 
4624 void
if_setrcvif(struct mbuf * m,if_t ifp)4625 if_setrcvif(struct mbuf *m, if_t ifp)
4626 {
4627 
4628 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4629 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4630 }
4631 
4632 void
if_setvtag(struct mbuf * m,uint16_t tag)4633 if_setvtag(struct mbuf *m, uint16_t tag)
4634 {
4635 	m->m_pkthdr.ether_vtag = tag;
4636 }
4637 
4638 uint16_t
if_getvtag(struct mbuf * m)4639 if_getvtag(struct mbuf *m)
4640 {
4641 	return (m->m_pkthdr.ether_vtag);
4642 }
4643 
4644 int
if_sendq_empty(if_t ifp)4645 if_sendq_empty(if_t ifp)
4646 {
4647 	return (IFQ_DRV_IS_EMPTY(&ifp->if_snd));
4648 }
4649 
4650 struct ifaddr *
if_getifaddr(const if_t ifp)4651 if_getifaddr(const if_t ifp)
4652 {
4653 	return (ifp->if_addr);
4654 }
4655 
4656 int
if_setsendqready(if_t ifp)4657 if_setsendqready(if_t ifp)
4658 {
4659 	IFQ_SET_READY(&ifp->if_snd);
4660 	return (0);
4661 }
4662 
4663 int
if_setsendqlen(if_t ifp,int tx_desc_count)4664 if_setsendqlen(if_t ifp, int tx_desc_count)
4665 {
4666 	IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count);
4667 	ifp->if_snd.ifq_drv_maxlen = tx_desc_count;
4668 	return (0);
4669 }
4670 
4671 void
if_setnetmapadapter(if_t ifp,struct netmap_adapter * na)4672 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na)
4673 {
4674 	ifp->if_netmap = na;
4675 }
4676 
4677 struct netmap_adapter *
if_getnetmapadapter(if_t ifp)4678 if_getnetmapadapter(if_t ifp)
4679 {
4680 	return (ifp->if_netmap);
4681 }
4682 
4683 int
if_vlantrunkinuse(if_t ifp)4684 if_vlantrunkinuse(if_t ifp)
4685 {
4686 	return (ifp->if_vlantrunk != NULL);
4687 }
4688 
4689 void
if_init(if_t ifp,void * ctx)4690 if_init(if_t ifp, void *ctx)
4691 {
4692 	(*ifp->if_init)(ctx);
4693 }
4694 
4695 void
if_input(if_t ifp,struct mbuf * sendmp)4696 if_input(if_t ifp, struct mbuf* sendmp)
4697 {
4698 	(*ifp->if_input)(ifp, sendmp);
4699 }
4700 
4701 int
if_transmit(if_t ifp,struct mbuf * m)4702 if_transmit(if_t ifp, struct mbuf *m)
4703 {
4704 	return ((*ifp->if_transmit)(ifp, m));
4705 }
4706 
4707 int
if_resolvemulti(if_t ifp,struct sockaddr ** srcs,struct sockaddr * dst)4708 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst)
4709 {
4710 	if (ifp->if_resolvemulti == NULL)
4711 		return (EOPNOTSUPP);
4712 
4713 	return (ifp->if_resolvemulti(ifp, srcs, dst));
4714 }
4715 
4716 int
if_ioctl(if_t ifp,u_long cmd,void * data)4717 if_ioctl(if_t ifp, u_long cmd, void *data)
4718 {
4719 	if (ifp->if_ioctl == NULL)
4720 		return (EOPNOTSUPP);
4721 
4722 	return (ifp->if_ioctl(ifp, cmd, data));
4723 }
4724 
4725 struct mbuf *
if_dequeue(if_t ifp)4726 if_dequeue(if_t ifp)
4727 {
4728 	struct mbuf *m;
4729 
4730 	IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
4731 	return (m);
4732 }
4733 
4734 int
if_sendq_prepend(if_t ifp,struct mbuf * m)4735 if_sendq_prepend(if_t ifp, struct mbuf *m)
4736 {
4737 	IFQ_DRV_PREPEND(&ifp->if_snd, m);
4738 	return (0);
4739 }
4740 
4741 int
if_setifheaderlen(if_t ifp,int len)4742 if_setifheaderlen(if_t ifp, int len)
4743 {
4744 	ifp->if_hdrlen = len;
4745 	return (0);
4746 }
4747 
4748 char *
if_getlladdr(const if_t ifp)4749 if_getlladdr(const if_t ifp)
4750 {
4751 	return (IF_LLADDR(ifp));
4752 }
4753 
4754 void *
if_gethandle(u_char type)4755 if_gethandle(u_char type)
4756 {
4757 	return (if_alloc(type));
4758 }
4759 
4760 void
if_vlancap(if_t ifp)4761 if_vlancap(if_t ifp)
4762 {
4763 	VLAN_CAPABILITIES(ifp);
4764 }
4765 
4766 int
if_sethwtsomax(if_t ifp,u_int if_hw_tsomax)4767 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4768 {
4769 	ifp->if_hw_tsomax = if_hw_tsomax;
4770         return (0);
4771 }
4772 
4773 int
if_sethwtsomaxsegcount(if_t ifp,u_int if_hw_tsomaxsegcount)4774 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4775 {
4776 	ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4777         return (0);
4778 }
4779 
4780 int
if_sethwtsomaxsegsize(if_t ifp,u_int if_hw_tsomaxsegsize)4781 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4782 {
4783 	ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4784         return (0);
4785 }
4786 
4787 u_int
if_gethwtsomax(const if_t ifp)4788 if_gethwtsomax(const if_t ifp)
4789 {
4790 	return (ifp->if_hw_tsomax);
4791 }
4792 
4793 u_int
if_gethwtsomaxsegcount(const if_t ifp)4794 if_gethwtsomaxsegcount(const if_t ifp)
4795 {
4796 	return (ifp->if_hw_tsomaxsegcount);
4797 }
4798 
4799 u_int
if_gethwtsomaxsegsize(const if_t ifp)4800 if_gethwtsomaxsegsize(const if_t ifp)
4801 {
4802 	return (ifp->if_hw_tsomaxsegsize);
4803 }
4804 
4805 void
if_setinitfn(if_t ifp,if_init_fn_t init_fn)4806 if_setinitfn(if_t ifp, if_init_fn_t init_fn)
4807 {
4808 	ifp->if_init = init_fn;
4809 }
4810 
4811 void
if_setinputfn(if_t ifp,if_input_fn_t input_fn)4812 if_setinputfn(if_t ifp, if_input_fn_t input_fn)
4813 {
4814 	ifp->if_input = input_fn;
4815 }
4816 
4817 if_input_fn_t
if_getinputfn(if_t ifp)4818 if_getinputfn(if_t ifp)
4819 {
4820 	return (ifp->if_input);
4821 }
4822 
4823 void
if_setioctlfn(if_t ifp,if_ioctl_fn_t ioctl_fn)4824 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn)
4825 {
4826 	ifp->if_ioctl = ioctl_fn;
4827 }
4828 
4829 void
if_setoutputfn(if_t ifp,if_output_fn_t output_fn)4830 if_setoutputfn(if_t ifp, if_output_fn_t output_fn)
4831 {
4832 	ifp->if_output = output_fn;
4833 }
4834 
4835 void
if_setstartfn(if_t ifp,if_start_fn_t start_fn)4836 if_setstartfn(if_t ifp, if_start_fn_t start_fn)
4837 {
4838 	ifp->if_start = start_fn;
4839 }
4840 
4841 if_start_fn_t
if_getstartfn(if_t ifp)4842 if_getstartfn(if_t ifp)
4843 {
4844 	return (ifp->if_start);
4845 }
4846 
4847 void
if_settransmitfn(if_t ifp,if_transmit_fn_t start_fn)4848 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4849 {
4850 	ifp->if_transmit = start_fn;
4851 }
4852 
4853 if_transmit_fn_t
if_gettransmitfn(if_t ifp)4854 if_gettransmitfn(if_t ifp)
4855 {
4856 	return (ifp->if_transmit);
4857 }
4858 
4859 void
if_setqflushfn(if_t ifp,if_qflush_fn_t flush_fn)4860 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4861 {
4862 	ifp->if_qflush = flush_fn;
4863 }
4864 
4865 void
if_setsndtagallocfn(if_t ifp,if_snd_tag_alloc_t alloc_fn)4866 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn)
4867 {
4868 	ifp->if_snd_tag_alloc = alloc_fn;
4869 }
4870 
4871 int
if_snd_tag_alloc(if_t ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** mstp)4872 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
4873     struct m_snd_tag **mstp)
4874 {
4875 	if (ifp->if_snd_tag_alloc == NULL)
4876 		return (EOPNOTSUPP);
4877 	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
4878 }
4879 
4880 void
if_setgetcounterfn(if_t ifp,if_get_counter_t fn)4881 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4882 {
4883 	ifp->if_get_counter = fn;
4884 }
4885 
4886 void
if_setreassignfn(if_t ifp,if_reassign_fn_t fn)4887 if_setreassignfn(if_t ifp, if_reassign_fn_t fn)
4888 {
4889 	ifp->if_reassign = fn;
4890 }
4891 
4892 void
if_setratelimitqueryfn(if_t ifp,if_ratelimit_query_t fn)4893 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn)
4894 {
4895 	ifp->if_ratelimit_query = fn;
4896 }
4897 
4898 void
if_setdebugnet_methods(if_t ifp,struct debugnet_methods * m)4899 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m)
4900 {
4901 	ifp->if_debugnet_methods = m;
4902 }
4903 
4904 struct label *
if_getmaclabel(if_t ifp)4905 if_getmaclabel(if_t ifp)
4906 {
4907 	return (ifp->if_label);
4908 }
4909 
4910 void
if_setmaclabel(if_t ifp,struct label * label)4911 if_setmaclabel(if_t ifp, struct label *label)
4912 {
4913 	ifp->if_label = label;
4914 }
4915 
4916 int
if_gettype(if_t ifp)4917 if_gettype(if_t ifp)
4918 {
4919 	return (ifp->if_type);
4920 }
4921 
4922 void *
if_getllsoftc(if_t ifp)4923 if_getllsoftc(if_t ifp)
4924 {
4925 	return (ifp->if_llsoftc);
4926 }
4927 
4928 void
if_setllsoftc(if_t ifp,void * llsoftc)4929 if_setllsoftc(if_t ifp, void *llsoftc)
4930 {
4931 	ifp->if_llsoftc = llsoftc;
4932 };
4933 
4934 int
if_getlinkstate(if_t ifp)4935 if_getlinkstate(if_t ifp)
4936 {
4937 	return (ifp->if_link_state);
4938 }
4939 
4940 const uint8_t *
if_getbroadcastaddr(if_t ifp)4941 if_getbroadcastaddr(if_t ifp)
4942 {
4943 	return (ifp->if_broadcastaddr);
4944 }
4945 
4946 void
if_setbroadcastaddr(if_t ifp,const uint8_t * addr)4947 if_setbroadcastaddr(if_t ifp, const uint8_t *addr)
4948 {
4949 	ifp->if_broadcastaddr = addr;
4950 }
4951 
4952 int
if_getnumadomain(if_t ifp)4953 if_getnumadomain(if_t ifp)
4954 {
4955 	return (ifp->if_numa_domain);
4956 }
4957 
4958 uint64_t
if_getcounter(if_t ifp,ift_counter counter)4959 if_getcounter(if_t ifp, ift_counter counter)
4960 {
4961 	return (ifp->if_get_counter(ifp, counter));
4962 }
4963 
4964 bool
if_altq_is_enabled(if_t ifp)4965 if_altq_is_enabled(if_t ifp)
4966 {
4967 	return (ALTQ_IS_ENABLED(&ifp->if_snd));
4968 }
4969 
4970 struct vnet *
if_getvnet(if_t ifp)4971 if_getvnet(if_t ifp)
4972 {
4973 	return (ifp->if_vnet);
4974 }
4975 
4976 struct in_ifinfo *
if_getinet(if_t ifp)4977 if_getinet(if_t ifp)
4978 {
4979 	return (ifp->if_inet);
4980 }
4981 
4982 struct in6_ifextra *
if_getinet6(if_t ifp)4983 if_getinet6(if_t ifp)
4984 {
4985 	return (ifp->if_inet6);
4986 }
4987 
4988 u_int
if_getfib(if_t ifp)4989 if_getfib(if_t ifp)
4990 {
4991 	return (ifp->if_fib);
4992 }
4993 
4994 uint8_t
if_getaddrlen(if_t ifp)4995 if_getaddrlen(if_t ifp)
4996 {
4997 	return (ifp->if_addrlen);
4998 }
4999 
5000 struct bpf_if *
if_getbpf(if_t ifp)5001 if_getbpf(if_t ifp)
5002 {
5003 	return (ifp->if_bpf);
5004 }
5005 
5006 struct ifvlantrunk *
if_getvlantrunk(if_t ifp)5007 if_getvlantrunk(if_t ifp)
5008 {
5009 	return (ifp->if_vlantrunk);
5010 }
5011 
5012 uint8_t
if_getpcp(if_t ifp)5013 if_getpcp(if_t ifp)
5014 {
5015 	return (ifp->if_pcp);
5016 }
5017 
5018 void *
if_getl2com(if_t ifp)5019 if_getl2com(if_t ifp)
5020 {
5021 	return (ifp->if_l2com);
5022 }
5023 
5024 void
if_setipsec_accel_methods(if_t ifp,const struct if_ipsec_accel_methods * m)5025 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m)
5026 {
5027 	ifp->if_ipsec_accel_m = m;
5028 }
5029 
5030 #ifdef DDB
5031 static void
if_show_ifnet(struct ifnet * ifp)5032 if_show_ifnet(struct ifnet *ifp)
5033 {
5034 	if (ifp == NULL)
5035 		return;
5036 	db_printf("%s:\n", ifp->if_xname);
5037 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
5038 	IF_DB_PRINTF("%s", if_dname);
5039 	IF_DB_PRINTF("%d", if_dunit);
5040 	IF_DB_PRINTF("%s", if_description);
5041 	IF_DB_PRINTF("%u", if_index);
5042 	IF_DB_PRINTF("%d", if_idxgen);
5043 	IF_DB_PRINTF("%u", if_refcount);
5044 	IF_DB_PRINTF("%p", if_softc);
5045 	IF_DB_PRINTF("%p", if_l2com);
5046 	IF_DB_PRINTF("%p", if_llsoftc);
5047 	IF_DB_PRINTF("%d", if_amcount);
5048 	IF_DB_PRINTF("%p", if_addr);
5049 	IF_DB_PRINTF("%p", if_broadcastaddr);
5050 	IF_DB_PRINTF("%u", if_fib);
5051 	IF_DB_PRINTF("%p", if_vnet);
5052 	IF_DB_PRINTF("%p", if_home_vnet);
5053 	IF_DB_PRINTF("%p", if_vlantrunk);
5054 	IF_DB_PRINTF("%p", if_bpf);
5055 	IF_DB_PRINTF("%u", if_pcount);
5056 	IF_DB_PRINTF("%p", if_bridge);
5057 	IF_DB_PRINTF("%p", if_lagg);
5058 	IF_DB_PRINTF("%p", if_pf_kif);
5059 	IF_DB_PRINTF("%p", if_carp);
5060 	IF_DB_PRINTF("%p", if_label);
5061 	IF_DB_PRINTF("%p", if_netmap);
5062 	IF_DB_PRINTF("0x%08x", if_flags);
5063 	IF_DB_PRINTF("0x%08x", if_drv_flags);
5064 	IF_DB_PRINTF("0x%08x", if_capabilities);
5065 	IF_DB_PRINTF("0x%08x", if_capenable);
5066 	IF_DB_PRINTF("%p", if_snd.ifq_head);
5067 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
5068 	IF_DB_PRINTF("%d", if_snd.ifq_len);
5069 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
5070 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
5071 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
5072 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
5073 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
5074 	IF_DB_PRINTF("%d", if_snd.altq_type);
5075 	IF_DB_PRINTF("%x", if_snd.altq_flags);
5076 #undef IF_DB_PRINTF
5077 }
5078 
DB_SHOW_COMMAND(ifnet,db_show_ifnet)5079 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
5080 {
5081 	if (!have_addr) {
5082 		db_printf("usage: show ifnet <struct ifnet *>\n");
5083 		return;
5084 	}
5085 
5086 	if_show_ifnet((struct ifnet *)addr);
5087 }
5088 
DB_SHOW_ALL_COMMAND(ifnets,db_show_all_ifnets)5089 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
5090 {
5091 	struct ifnet *ifp;
5092 	u_short idx;
5093 
5094 	for (idx = 1; idx <= if_index; idx++) {
5095 		ifp = ifindex_table[idx].ife_ifnet;
5096 		if (ifp == NULL)
5097 			continue;
5098 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
5099 		if (db_pager_quit)
5100 			break;
5101 	}
5102 }
5103 #endif	/* DDB */
5104