xref: /freebsd/sys/net/if.c (revision ba7f47d47dc1a177e4d8f115f791ec25f3da0eab)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org>
5  * Copyright (c) 1980, 1986, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include "opt_bpf.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_ddb.h"
37 
38 #include <sys/param.h>
39 #include <sys/capsicum.h>
40 #include <sys/conf.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/domainset.h>
44 #include <sys/sbuf.h>
45 #include <sys/bus.h>
46 #include <sys/epoch.h>
47 #include <sys/mbuf.h>
48 #include <sys/systm.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/refcount.h>
57 #include <sys/module.h>
58 #include <sys/nv.h>
59 #include <sys/rwlock.h>
60 #include <sys/sockio.h>
61 #include <sys/stdarg.h>
62 #include <sys/syslog.h>
63 #include <sys/sysctl.h>
64 #include <sys/sysent.h>
65 #include <sys/taskqueue.h>
66 #include <sys/domain.h>
67 #include <sys/jail.h>
68 #include <sys/priv.h>
69 
70 #ifdef DDB
71 #include <ddb/ddb.h>
72 #endif
73 
74 #include <vm/uma.h>
75 
76 #include <net/bpf.h>
77 #include <net/if.h>
78 #include <net/if_arp.h>
79 #include <net/if_clone.h>
80 #include <net/if_dl.h>
81 #include <net/if_strings.h>
82 #include <net/if_types.h>
83 #include <net/if_var.h>
84 #include <net/if_media.h>
85 #include <net/if_mib.h>
86 #include <net/if_private.h>
87 #include <net/if_vlan_var.h>
88 #include <net/radix.h>
89 #include <net/route.h>
90 #include <net/route/route_ctl.h>
91 #include <net/vnet.h>
92 
93 #if defined(INET) || defined(INET6)
94 #include <net/ethernet.h>
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip.h>
98 #include <netinet/ip_carp.h>
99 #ifdef INET
100 #include <net/debugnet.h>
101 #include <netinet/if_ether.h>
102 #endif /* INET */
103 #ifdef INET6
104 #include <netinet6/in6_var.h>
105 #endif /* INET6 */
106 #endif /* INET || INET6 */
107 
108 #include <security/mac/mac_framework.h>
109 
110 /*
111  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
112  * and ifr_ifru when it is used in SIOCGIFCONF.
113  */
114 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
115     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
116 
117 __read_mostly epoch_t net_epoch_preempt;
118 #ifdef COMPAT_FREEBSD32
119 #include <sys/mount.h>
120 #include <compat/freebsd32/freebsd32.h>
121 
122 struct ifreq_buffer32 {
123 	uint32_t	length;		/* (size_t) */
124 	uint32_t	buffer;		/* (void *) */
125 };
126 
127 /*
128  * Interface request structure used for socket
129  * ioctl's.  All interface ioctl's must have parameter
130  * definitions which begin with ifr_name.  The
131  * remainder may be interface specific.
132  */
133 struct ifreq32 {
134 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
135 	union {
136 		struct sockaddr	ifru_addr;
137 		struct sockaddr	ifru_dstaddr;
138 		struct sockaddr	ifru_broadaddr;
139 		struct ifreq_buffer32 ifru_buffer;
140 		short		ifru_flags[2];
141 		short		ifru_index;
142 		int		ifru_jid;
143 		int		ifru_metric;
144 		int		ifru_mtu;
145 		int		ifru_phys;
146 		int		ifru_media;
147 		uint32_t	ifru_data;
148 		int		ifru_cap[2];
149 		u_int		ifru_fib;
150 		u_char		ifru_vlan_pcp;
151 	} ifr_ifru;
152 };
153 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
154 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
155     __offsetof(struct ifreq32, ifr_ifru));
156 
157 struct ifconf32 {
158 	int32_t	ifc_len;
159 	union {
160 		uint32_t	ifcu_buf;
161 		uint32_t	ifcu_req;
162 	} ifc_ifcu;
163 };
164 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
165 
166 struct ifdrv32 {
167 	char		ifd_name[IFNAMSIZ];
168 	uint32_t	ifd_cmd;
169 	uint32_t	ifd_len;
170 	uint32_t	ifd_data;
171 };
172 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
173 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
174 
175 struct ifgroupreq32 {
176 	char	ifgr_name[IFNAMSIZ];
177 	u_int	ifgr_len;
178 	union {
179 		char		ifgru_group[IFNAMSIZ];
180 		uint32_t	ifgru_groups;
181 	} ifgr_ifgru;
182 };
183 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
184 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
185 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
186 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
187 
188 struct ifmediareq32 {
189 	char		ifm_name[IFNAMSIZ];
190 	int		ifm_current;
191 	int		ifm_mask;
192 	int		ifm_status;
193 	int		ifm_active;
194 	int		ifm_count;
195 	uint32_t	ifm_ulist;	/* (int *) */
196 };
197 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
198 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
199 #endif /* COMPAT_FREEBSD32 */
200 
201 union ifreq_union {
202 	struct ifreq	ifr;
203 #ifdef COMPAT_FREEBSD32
204 	struct ifreq32	ifr32;
205 #endif
206 };
207 
208 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
209     "Link layers");
210 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
211     "Generic link-management");
212 
213 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
214     &ifqmaxlen, 0, "max send queue size");
215 
216 /* Log link state change events */
217 static int log_link_state_change = 1;
218 
219 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
220 	&log_link_state_change, 0,
221 	"log interface link state change events");
222 
223 /* Log promiscuous mode change events */
224 static int log_promisc_mode_change = 1;
225 
226 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
227 	&log_promisc_mode_change, 1,
228 	"log promiscuous mode change events");
229 
230 /* Interface description */
231 static unsigned int ifdescr_maxlen = 1024;
232 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
233 	&ifdescr_maxlen, 0,
234 	"administrative maximum length for interface description");
235 
236 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
237 
238 /* global sx for non-critical path ifdescr */
239 static struct sx ifdescr_sx;
240 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
241 
242 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
243 /* These are external hooks for CARP. */
244 void	(*carp_linkstate_p)(struct ifnet *ifp);
245 void	(*carp_demote_adj_p)(int, char *);
246 int	(*carp_master_p)(struct ifaddr *);
247 #if defined(INET) || defined(INET6)
248 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
249 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
250     const struct sockaddr *sa);
251 int	(*carp_attach_p)(struct ifaddr *, int);
252 void	(*carp_detach_p)(struct ifaddr *, bool);
253 #endif
254 #ifdef INET
255 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
256 #endif
257 #ifdef INET6
258 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
259 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
260     const struct in6_addr *taddr);
261 #endif
262 
263 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
264 
265 /*
266  * XXX: Style; these should be sorted alphabetically, and unprototyped
267  * static functions should be prototyped. Currently they are sorted by
268  * declaration order.
269  */
270 static int	ifconf(u_long, caddr_t);
271 static void	if_input_default(struct ifnet *, struct mbuf *);
272 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
273 static int	if_setflag(struct ifnet *, int, int, int *, int);
274 static int	if_transmit_default(struct ifnet *ifp, struct mbuf *m);
275 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
276 static void	do_link_state_change(void *, int);
277 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
278 static int	if_getgroupmembers(struct ifgroupreq *);
279 static void	if_delgroups(struct ifnet *);
280 static void	if_attach_internal(struct ifnet *, bool);
281 static void	if_detach_internal(struct ifnet *, bool);
282 static void	if_siocaddmulti(void *, int);
283 static void	if_link_ifnet(struct ifnet *);
284 static bool	if_unlink_ifnet(struct ifnet *, bool);
285 #ifdef VIMAGE
286 static void	if_vmove(struct ifnet *, struct vnet *);
287 #endif
288 
289 #ifdef INET6
290 /*
291  * XXX: declare here to avoid to include many inet6 related files..
292  * should be more generalized?
293  */
294 extern void	nd6_setmtu(struct ifnet *);
295 #endif
296 
297 /* ipsec helper hooks */
298 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
299 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
300 
301 int	ifqmaxlen = IFQ_MAXLEN;
302 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
303 VNET_DEFINE(struct ifgrouphead, ifg_head);
304 
305 /* Table of ifnet by index. */
306 static int if_index;
307 static int if_indexlim = 8;
308 static struct ifindex_entry {
309 	struct ifnet	*ife_ifnet;
310 	uint16_t	ife_gencnt;
311 } *ifindex_table;
312 
313 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
314     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
315     "Variables global to all interfaces");
316 static int
sysctl_ifcount(SYSCTL_HANDLER_ARGS)317 sysctl_ifcount(SYSCTL_HANDLER_ARGS)
318 {
319 	int rv = 0;
320 
321 	IFNET_RLOCK();
322 	for (int i = 1; i <= if_index; i++)
323 		if (ifindex_table[i].ife_ifnet != NULL &&
324 		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
325 			rv = i;
326 	IFNET_RUNLOCK();
327 
328 	return (sysctl_handle_int(oidp, &rv, 0, req));
329 }
330 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
331     CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
332     "Maximum known interface index");
333 
334 /*
335  * The global network interface list (V_ifnet) and related state (such as
336  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
337  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
338  */
339 struct sx ifnet_sxlock;
340 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
341 
342 struct sx ifnet_detach_sxlock;
343 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
344     SX_RECURSE);
345 
346 static	if_com_alloc_t *if_com_alloc[256];
347 static	if_com_free_t *if_com_free[256];
348 
349 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
350 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
351 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
352 
353 struct ifnet *
ifnet_byindex(u_int idx)354 ifnet_byindex(u_int idx)
355 {
356 	struct ifnet *ifp;
357 
358 	NET_EPOCH_ASSERT();
359 
360 	if (__predict_false(idx > if_index))
361 		return (NULL);
362 
363 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
364 
365 	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
366 		ifp = NULL;
367 
368 	return (ifp);
369 }
370 
371 struct ifnet *
ifnet_byindex_ref(u_int idx)372 ifnet_byindex_ref(u_int idx)
373 {
374 	struct ifnet *ifp;
375 
376 	ifp = ifnet_byindex(idx);
377 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
378 		return (NULL);
379 	if (!if_try_ref(ifp))
380 		return (NULL);
381 	return (ifp);
382 }
383 
384 struct ifnet *
ifnet_byindexgen(uint16_t idx,uint16_t gen)385 ifnet_byindexgen(uint16_t idx, uint16_t gen)
386 {
387 	struct ifnet *ifp;
388 
389 	NET_EPOCH_ASSERT();
390 
391 	if (__predict_false(idx > if_index))
392 		return (NULL);
393 
394 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
395 
396 	if (ifindex_table[idx].ife_gencnt == gen)
397 		return (ifp);
398 	else
399 		return (NULL);
400 }
401 
402 /*
403  * Network interface utility routines.
404  *
405  * Routines with ifa_ifwith* names take sockaddr *'s as
406  * parameters.
407  */
408 
409 static void
if_init_idxtable(void * arg __unused)410 if_init_idxtable(void *arg __unused)
411 {
412 
413 	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
414 	    M_IFNET, M_WAITOK | M_ZERO);
415 }
416 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL);
417 
418 static void
vnet_if_init(const void * unused __unused)419 vnet_if_init(const void *unused __unused)
420 {
421 
422 	CK_STAILQ_INIT(&V_ifnet);
423 	CK_STAILQ_INIT(&V_ifg_head);
424 }
425 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
426     NULL);
427 
428 static void
if_link_ifnet(struct ifnet * ifp)429 if_link_ifnet(struct ifnet *ifp)
430 {
431 	IFNET_WLOCK();
432 
433 	MPASS(refcount_load(&ifp->if_refcount) > 0);
434 	MPASS(ifp->if_vnet == curvnet);
435 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
436 
437 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
438 #ifdef VIMAGE
439 	curvnet->vnet_ifcnt++;
440 #endif
441 	IFNET_WUNLOCK();
442 }
443 
444 static bool
if_unlink_ifnet(struct ifnet * ifp,bool vmove)445 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
446 {
447 	struct ifnet *iter;
448 	int found = 0;
449 
450 	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
451 	IFNET_WLOCK();
452 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
453 		if (iter == ifp) {
454 			MPASS(refcount_load(&ifp->if_refcount) > 0);
455 			MPASS(ifp->if_vnet == curvnet);
456 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
457 
458 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
459 #ifdef VIMAGE
460 			curvnet->vnet_ifcnt--;
461 #endif
462 			if (!vmove)
463 				ifp->if_flags |= IFF_DYING;
464 			found = 1;
465 			break;
466 		}
467 	IFNET_WUNLOCK();
468 
469 	return (found);
470 }
471 
472 #ifdef VIMAGE
473 static void
vnet_if_return(const void * unused __unused)474 vnet_if_return(const void *unused __unused)
475 {
476 	struct ifnet *ifp, *nifp;
477 	struct ifnet **pending;
478 	int found __diagused;
479 	int i;
480 
481 	i = 0;
482 
483 	/* The lock has already been aquired in vnet_destroy() */
484 	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
485 	/*
486 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
487 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
488 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
489 	 * complete. We can't do that from within NET_EPOCH.
490 	 *
491 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
492 	 * read/write lock. We cannot hold the lock as we call if_vmove()
493 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
494 	 * ctx lock.
495 	 */
496 	IFNET_WLOCK();
497 
498 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
499 	    M_IFNET, M_WAITOK | M_ZERO);
500 
501 	/* Return all inherited interfaces to their parent vnets. */
502 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
503 		if (ifp->if_home_vnet != ifp->if_vnet) {
504 			found = if_unlink_ifnet(ifp, true);
505 			MPASS(found);
506 
507 			pending[i++] = ifp;
508 		}
509 	}
510 	IFNET_WUNLOCK();
511 
512 	for (int j = 0; j < i; j++) {
513 		if_vmove(pending[j], pending[j]->if_home_vnet);
514 	}
515 
516 	free(pending, M_IFNET);
517 }
518 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
519     vnet_if_return, NULL);
520 #endif
521 
522 /*
523  * Allocate a struct ifnet and an index for an interface.  A layer 2
524  * common structure will also be allocated if an allocation routine is
525  * registered for the passed type.
526  */
527 static struct ifnet *
if_alloc_domain(u_char type,int numa_domain)528 if_alloc_domain(u_char type, int numa_domain)
529 {
530 	struct ifnet *ifp;
531 	u_short idx;
532 
533 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
534 	if (numa_domain == IF_NODOM)
535 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
536 		    M_WAITOK | M_ZERO);
537 	else
538 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
539 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
540 	ifp->if_type = type;
541 	ifp->if_alloctype = type;
542 	ifp->if_numa_domain = numa_domain;
543 #ifdef VIMAGE
544 	ifp->if_vnet = curvnet;
545 #endif
546 	if (if_com_alloc[type] != NULL) {
547 		ifp->if_l2com = if_com_alloc[type](type, ifp);
548 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
549 		    type));
550 	}
551 
552 	IF_ADDR_LOCK_INIT(ifp);
553 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
554 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
555 	CK_STAILQ_INIT(&ifp->if_addrhead);
556 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
557 	CK_STAILQ_INIT(&ifp->if_groups);
558 #ifdef MAC
559 	mac_ifnet_init(ifp);
560 #endif
561 	ifq_init(&ifp->if_snd, ifp);
562 
563 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
564 	for (int i = 0; i < IFCOUNTERS; i++)
565 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
566 	ifp->if_get_counter = if_get_counter_default;
567 	ifp->if_pcp = IFNET_PCP_NONE;
568 
569 	/* Allocate an ifindex array entry. */
570 	IFNET_WLOCK();
571 	/*
572 	 * Try to find an empty slot below if_index.  If we fail, take the
573 	 * next slot.
574 	 */
575 	for (idx = 1; idx <= if_index; idx++) {
576 		if (ifindex_table[idx].ife_ifnet == NULL)
577 			break;
578 	}
579 
580 	/* Catch if_index overflow. */
581 	if (idx >= if_indexlim) {
582 		struct ifindex_entry *new, *old;
583 		int newlim;
584 
585 		newlim = if_indexlim * 2;
586 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
587 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
588 		old = ifindex_table;
589 		ck_pr_store_ptr(&ifindex_table, new);
590 		if_indexlim = newlim;
591 		NET_EPOCH_WAIT();
592 		free(old, M_IFNET);
593 	}
594 	if (idx > if_index)
595 		if_index = idx;
596 
597 	ifp->if_index = idx;
598 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
599 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
600 	IFNET_WUNLOCK();
601 
602 	return (ifp);
603 }
604 
605 struct ifnet *
if_alloc_dev(u_char type,device_t dev)606 if_alloc_dev(u_char type, device_t dev)
607 {
608 	int numa_domain;
609 
610 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
611 		return (if_alloc_domain(type, IF_NODOM));
612 	return (if_alloc_domain(type, numa_domain));
613 }
614 
615 struct ifnet *
if_alloc(u_char type)616 if_alloc(u_char type)
617 {
618 
619 	return (if_alloc_domain(type, IF_NODOM));
620 }
621 /*
622  * Do the actual work of freeing a struct ifnet, and layer 2 common
623  * structure.  This call is made when the network epoch guarantees
624  * us that nobody holds a pointer to the interface.
625  */
626 static void
if_free_deferred(epoch_context_t ctx)627 if_free_deferred(epoch_context_t ctx)
628 {
629 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
630 
631 	KASSERT((ifp->if_flags & IFF_DYING),
632 	    ("%s: interface not dying", __func__));
633 
634 	if (if_com_free[ifp->if_alloctype] != NULL)
635 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
636 		    ifp->if_alloctype);
637 
638 #ifdef MAC
639 	mac_ifnet_destroy(ifp);
640 #endif /* MAC */
641 	IF_ADDR_LOCK_DESTROY(ifp);
642 	ifq_delete(&ifp->if_snd);
643 
644 	for (int i = 0; i < IFCOUNTERS; i++)
645 		counter_u64_free(ifp->if_counters[i]);
646 
647 	if_freedescr(ifp->if_description);
648 	free(ifp->if_hw_addr, M_IFADDR);
649 	free(ifp, M_IFNET);
650 }
651 
652 /*
653  * Deregister an interface and free the associated storage.
654  */
655 void
if_free(struct ifnet * ifp)656 if_free(struct ifnet *ifp)
657 {
658 
659 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
660 
661 	/*
662 	 * XXXGL: An interface index is really an alias to ifp pointer.
663 	 * Why would we clear the alias now, and not in the deferred
664 	 * context?  Indeed there is nothing wrong with some network
665 	 * thread obtaining ifp via ifnet_byindex() inside the network
666 	 * epoch and then dereferencing ifp while we perform if_free(),
667 	 * and after if_free() finished, too.
668 	 *
669 	 * This early index freeing was important back when ifindex was
670 	 * virtualized and interface would outlive the vnet.
671 	 */
672 	IFNET_WLOCK();
673 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
674 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
675 	ifindex_table[ifp->if_index].ife_gencnt++;
676 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
677 		if_index--;
678 	IFNET_WUNLOCK();
679 
680 	if (refcount_release(&ifp->if_refcount))
681 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
682 }
683 
684 /*
685  * Interfaces to keep an ifnet type-stable despite the possibility of the
686  * driver calling if_free().  If there are additional references, we defer
687  * freeing the underlying data structure.
688  */
689 void
if_ref(struct ifnet * ifp)690 if_ref(struct ifnet *ifp)
691 {
692 	u_int old __diagused;
693 
694 	/* We don't assert the ifnet list lock here, but arguably should. */
695 	old = refcount_acquire(&ifp->if_refcount);
696 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
697 }
698 
699 bool
if_try_ref(struct ifnet * ifp)700 if_try_ref(struct ifnet *ifp)
701 {
702 	NET_EPOCH_ASSERT();
703 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
704 }
705 
706 void
if_rele(struct ifnet * ifp)707 if_rele(struct ifnet *ifp)
708 {
709 
710 	if (!refcount_release(&ifp->if_refcount))
711 		return;
712 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
713 }
714 
715 void
ifq_init(struct ifaltq * ifq,struct ifnet * ifp)716 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
717 {
718 
719 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
720 
721 	if (ifq->ifq_maxlen == 0)
722 		ifq->ifq_maxlen = ifqmaxlen;
723 
724 	ifq->altq_type = 0;
725 	ifq->altq_disc = NULL;
726 	ifq->altq_flags &= ALTQF_CANTCHANGE;
727 	ifq->altq_tbr  = NULL;
728 	ifq->altq_ifp  = ifp;
729 }
730 
731 void
ifq_delete(struct ifaltq * ifq)732 ifq_delete(struct ifaltq *ifq)
733 {
734 	mtx_destroy(&ifq->ifq_mtx);
735 }
736 
737 /*
738  * Perform generic interface initialization tasks and attach the interface
739  * to the list of "active" interfaces.  If vmove flag is set on entry
740  * to if_attach_internal(), perform only a limited subset of initialization
741  * tasks, given that we are moving from one vnet to another an ifnet which
742  * has already been fully initialized.
743  *
744  * Note that if_detach_internal() removes group membership unconditionally
745  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
746  * Thus, when if_vmove() is applied to a cloned interface, group membership
747  * is lost while a cloned one always joins a group whose name is
748  * ifc->ifc_name.  To recover this after if_detach_internal() and
749  * if_attach_internal(), the cloner should be specified to
750  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
751  * attempts to join a group whose name is ifc->ifc_name.
752  *
753  * XXX:
754  *  - The decision to return void and thus require this function to
755  *    succeed is questionable.
756  *  - We should probably do more sanity checking.  For instance we don't
757  *    do anything to insure if_xname is unique or non-empty.
758  */
759 void
if_attach(struct ifnet * ifp)760 if_attach(struct ifnet *ifp)
761 {
762 
763 	if_attach_internal(ifp, false);
764 }
765 
766 /*
767  * Compute the least common TSO limit.
768  */
769 void
if_hw_tsomax_common(if_t ifp,struct ifnet_hw_tsomax * pmax)770 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
771 {
772 	/*
773 	 * 1) If there is no limit currently, take the limit from
774 	 * the network adapter.
775 	 *
776 	 * 2) If the network adapter has a limit below the current
777 	 * limit, apply it.
778 	 */
779 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
780 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
781 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
782 	}
783 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
784 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
785 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
786 	}
787 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
788 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
789 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
790 	}
791 }
792 
793 /*
794  * Update TSO limit of a network adapter.
795  *
796  * Returns zero if no change. Else non-zero.
797  */
798 int
if_hw_tsomax_update(if_t ifp,struct ifnet_hw_tsomax * pmax)799 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
800 {
801 	int retval = 0;
802 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
803 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
804 		retval++;
805 	}
806 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
807 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
808 		retval++;
809 	}
810 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
811 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
812 		retval++;
813 	}
814 	return (retval);
815 }
816 
817 static void
if_attach_internal(struct ifnet * ifp,bool vmove)818 if_attach_internal(struct ifnet *ifp, bool vmove)
819 {
820 	unsigned socksize, ifasize;
821 	int namelen, masklen;
822 	struct sockaddr_dl *sdl;
823 	struct ifaddr *ifa;
824 
825 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
826 
827 #ifdef VIMAGE
828 	CURVNET_ASSERT_SET();
829 	ifp->if_vnet = curvnet;
830 	if (ifp->if_home_vnet == NULL)
831 		ifp->if_home_vnet = curvnet;
832 #endif
833 
834 	if_addgroup(ifp, IFG_ALL);
835 
836 #ifdef VIMAGE
837 	/* Restore group membership for cloned interface. */
838 	if (vmove)
839 		if_clone_restoregroup(ifp);
840 #endif
841 
842 	getmicrotime(&ifp->if_lastchange);
843 	ifp->if_epoch = time_uptime;
844 
845 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
846 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
847 	    ("transmit and qflush must both either be set or both be NULL"));
848 	if (ifp->if_transmit == NULL) {
849 		ifp->if_transmit = if_transmit_default;
850 		ifp->if_qflush = if_qflush;
851 	}
852 	if (ifp->if_input == NULL)
853 		ifp->if_input = if_input_default;
854 
855 	if (ifp->if_requestencap == NULL)
856 		ifp->if_requestencap = if_requestencap_default;
857 
858 	if (!vmove) {
859 #ifdef MAC
860 		mac_ifnet_create(ifp);
861 #endif
862 
863 		/*
864 		 * Create a Link Level name for this device.
865 		 */
866 		namelen = strlen(ifp->if_xname);
867 		/*
868 		 * Always save enough space for any possible name so we
869 		 * can do a rename in place later.
870 		 */
871 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
872 		socksize = masklen + ifp->if_addrlen;
873 		if (socksize < sizeof(*sdl))
874 			socksize = sizeof(*sdl);
875 		socksize = roundup2(socksize, sizeof(long));
876 		ifasize = sizeof(*ifa) + 2 * socksize;
877 		ifa = ifa_alloc(ifasize, M_WAITOK);
878 		sdl = (struct sockaddr_dl *)(ifa + 1);
879 		sdl->sdl_len = socksize;
880 		sdl->sdl_family = AF_LINK;
881 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
882 		sdl->sdl_nlen = namelen;
883 		sdl->sdl_index = ifp->if_index;
884 		sdl->sdl_type = ifp->if_type;
885 		ifp->if_addr = ifa;
886 		ifa->ifa_ifp = ifp;
887 		ifa->ifa_addr = (struct sockaddr *)sdl;
888 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
889 		ifa->ifa_netmask = (struct sockaddr *)sdl;
890 		sdl->sdl_len = masklen;
891 		while (namelen != 0)
892 			sdl->sdl_data[--namelen] = 0xff;
893 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
894 		/* Reliably crash if used uninitialized. */
895 		ifp->if_broadcastaddr = NULL;
896 
897 		if (ifp->if_type == IFT_ETHER) {
898 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
899 			    M_WAITOK | M_ZERO);
900 		}
901 
902 #if defined(INET) || defined(INET6)
903 		/* Use defaults for TSO, if nothing is set */
904 		if (ifp->if_hw_tsomax == 0 &&
905 		    ifp->if_hw_tsomaxsegcount == 0 &&
906 		    ifp->if_hw_tsomaxsegsize == 0) {
907 			/*
908 			 * The TSO defaults needs to be such that an
909 			 * NFS mbuf list of 35 mbufs totalling just
910 			 * below 64K works and that a chain of mbufs
911 			 * can be defragged into at most 32 segments:
912 			 */
913 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
914 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
915 			ifp->if_hw_tsomaxsegcount = 35;
916 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
917 
918 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
919 			if (ifp->if_capabilities & IFCAP_TSO) {
920 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
921 				    ifp->if_hw_tsomax,
922 				    ifp->if_hw_tsomaxsegcount,
923 				    ifp->if_hw_tsomaxsegsize);
924 			}
925 		}
926 #endif
927 	}
928 
929 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
930 	if_link_ifnet(ifp);
931 	EVENTHANDLER_INVOKE(ifnet_attached_event, ifp);
932 	if (IS_DEFAULT_VNET(curvnet))
933 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
934 }
935 
936 static void
if_epochalloc(void * dummy __unused)937 if_epochalloc(void *dummy __unused)
938 {
939 
940 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
941 }
942 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
943 
944 /*
945  * Remove any unicast or broadcast network addresses from an interface.
946  */
947 void
if_purgeaddrs(struct ifnet * ifp)948 if_purgeaddrs(struct ifnet *ifp)
949 {
950 	struct ifaddr *ifa;
951 
952 #ifdef INET6
953 	/*
954 	 * Need to leave multicast addresses of proxy NDP llentries
955 	 * before in6_purgeifaddr() because the llentries are keys
956 	 * for in6_multi objects of proxy NDP entries.
957 	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
958 	 * then we would lose the keys if they are called earlier.
959 	 */
960 	in6_purge_proxy_ndp(ifp);
961 #endif
962 	while (1) {
963 		struct epoch_tracker et;
964 
965 		NET_EPOCH_ENTER(et);
966 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
967 			if (ifa->ifa_addr->sa_family != AF_LINK)
968 				break;
969 		}
970 		NET_EPOCH_EXIT(et);
971 
972 		if (ifa == NULL)
973 			break;
974 #ifdef INET
975 		/* XXX: Ugly!! ad hoc just for INET */
976 		if (ifa->ifa_addr->sa_family == AF_INET) {
977 			struct ifreq ifr;
978 
979 			bzero(&ifr, sizeof(ifr));
980 			ifr.ifr_addr = *ifa->ifa_addr;
981 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
982 			    NULL) == 0)
983 				continue;
984 		}
985 #endif /* INET */
986 #ifdef INET6
987 		if (ifa->ifa_addr->sa_family == AF_INET6) {
988 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
989 			/* ifp_addrhead is already updated */
990 			continue;
991 		}
992 #endif /* INET6 */
993 		IF_ADDR_WLOCK(ifp);
994 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
995 		IF_ADDR_WUNLOCK(ifp);
996 		ifa_free(ifa);
997 	}
998 }
999 
1000 /*
1001  * Remove any multicast network addresses from an interface when an ifnet
1002  * is going away.
1003  */
1004 static void
if_purgemaddrs(struct ifnet * ifp)1005 if_purgemaddrs(struct ifnet *ifp)
1006 {
1007 	struct ifmultiaddr *ifma;
1008 
1009 	IF_ADDR_WLOCK(ifp);
1010 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1011 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1012 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1013 		if_delmulti_locked(ifp, ifma, 1);
1014 	}
1015 	IF_ADDR_WUNLOCK(ifp);
1016 }
1017 
1018 /*
1019  * Detach an interface, removing it from the list of "active" interfaces.
1020  * If vmove flag is set on entry to if_detach_internal(), perform only a
1021  * limited subset of cleanup tasks, given that we are moving an ifnet from
1022  * one vnet to another, where it must be fully operational.
1023  *
1024  * XXXRW: There are some significant questions about event ordering, and
1025  * how to prevent things from starting to use the interface during detach.
1026  */
1027 void
if_detach(struct ifnet * ifp)1028 if_detach(struct ifnet *ifp)
1029 {
1030 	bool found;
1031 
1032 	/*
1033 	 * The driver private data holds a strong reference to the ifnet, and
1034 	 * it is actually the "owner", hence this routine shall never fail.
1035 	 *
1036 	 * Ideally we can loop retrying when we lose race with other threads
1037 	 * those run if_unlink_ifnet(). For simplicity, use ifnet_detach_sxlock
1038 	 * to serialize all the detach / vmove operations.
1039 	 */
1040 	sx_xlock(&ifnet_detach_sxlock);
1041 	CURVNET_SET_QUIET(ifp->if_vnet);
1042 	found = if_unlink_ifnet(ifp, false);
1043 	if (! found)
1044 		panic("%s: interface is not on the active list",
1045 		    ifp->if_xname);
1046 	if_detach_internal(ifp, false);
1047 	CURVNET_RESTORE();
1048 	sx_xunlock(&ifnet_detach_sxlock);
1049 }
1050 
1051 /*
1052  * The vmove flag, if set, indicates that we are called from a callpath
1053  * that is moving an interface to a different vnet instance.
1054  *
1055  * The shutdown flag, if set, indicates that we are called in the
1056  * process of shutting down a vnet instance.  Currently only the
1057  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1058  * on a vnet instance shutdown without this flag being set, e.g., when
1059  * the cloned interfaces are destoyed as first thing of teardown.
1060  */
1061 static void
if_detach_internal(struct ifnet * ifp,bool vmove)1062 if_detach_internal(struct ifnet *ifp, bool vmove)
1063 {
1064 	struct ifaddr *ifa;
1065 #ifdef VIMAGE
1066 	bool shutdown;
1067 
1068 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1069 #endif
1070 
1071 	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
1072 
1073 	/*
1074 	 * At this point we know the interface still was on the ifnet list
1075 	 * and we removed it so we are in a stable state.
1076 	 */
1077 	NET_EPOCH_WAIT();
1078 
1079 	/*
1080 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1081 	 * fixes issues about late destruction of multicast options
1082 	 * which lead to leave group calls, which in turn access the
1083 	 * belonging ifnet structure:
1084 	 */
1085 	NET_EPOCH_DRAIN_CALLBACKS();
1086 
1087 	/*
1088 	 * In any case (destroy or vmove) detach us from the groups
1089 	 * and remove/wait for pending events on the taskq.
1090 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1091 	 */
1092 	if_delgroups(ifp);
1093 
1094 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1095 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1096 
1097 	if_down(ifp);
1098 
1099 #ifdef VIMAGE
1100 	/*
1101 	 * On VNET shutdown abort here as the stack teardown will do all
1102 	 * the work top-down for us.
1103 	 */
1104 	if (shutdown) {
1105 		/* Give interface users the chance to clean up. */
1106 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1107 
1108 		/*
1109 		 * In case of a vmove we are done here without error.
1110 		 * If we would signal an error it would lead to the same
1111 		 * abort as if we did not find the ifnet anymore.
1112 		 * if_detach() calls us in void context and does not care
1113 		 * about an early abort notification, so life is splendid :)
1114 		 */
1115 		return;
1116 	}
1117 #endif
1118 
1119 	/*
1120 	 * At this point we are not tearing down a VNET and are either
1121 	 * going to destroy or vmove the interface and have to cleanup
1122 	 * accordingly.
1123 	 */
1124 
1125 	/*
1126 	 * Remove routes and flush queues.
1127 	 */
1128 #ifdef ALTQ
1129 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1130 		altq_disable(&ifp->if_snd);
1131 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1132 		altq_detach(&ifp->if_snd);
1133 #endif
1134 
1135 	rt_flushifroutes(ifp);
1136 
1137 	if_purgeaddrs(ifp);
1138 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1139 	if_purgemaddrs(ifp);
1140 	if (IS_DEFAULT_VNET(curvnet))
1141 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1142 
1143 	if (!vmove) {
1144 		/*
1145 		 * Prevent further calls into the device driver via ifnet.
1146 		 */
1147 		if_dead(ifp);
1148 
1149 		/*
1150 		 * Clean up all addresses.
1151 		 */
1152 		IF_ADDR_WLOCK(ifp);
1153 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1154 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1155 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1156 			IF_ADDR_WUNLOCK(ifp);
1157 			ifa_free(ifa);
1158 		} else
1159 			IF_ADDR_WUNLOCK(ifp);
1160 	}
1161 }
1162 
1163 #ifdef VIMAGE
1164 /*
1165  * if_vmove() performs a limited version of if_detach() in current
1166  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1167  */
1168 static void
if_vmove(struct ifnet * ifp,struct vnet * new_vnet)1169 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1170 {
1171 #ifdef DEV_BPF
1172 	/*
1173 	 * Detach BPF file descriptors from its interface.
1174 	 */
1175 	bpf_ifdetach(ifp);
1176 #endif
1177 
1178 	/*
1179 	 * Detach from current vnet, but preserve LLADDR info, do not
1180 	 * mark as dead etc. so that the ifnet can be reattached later.
1181 	 */
1182 	if_detach_internal(ifp, true);
1183 
1184 	/*
1185 	 * Perform interface-specific reassignment tasks, if provided by
1186 	 * the driver.
1187 	 */
1188 	if (ifp->if_reassign != NULL)
1189 		ifp->if_reassign(ifp, new_vnet, NULL);
1190 
1191 	/*
1192 	 * Switch to the context of the target vnet.
1193 	 */
1194 	CURVNET_SET_QUIET(new_vnet);
1195 	if_attach_internal(ifp, true);
1196 	bpf_vmove(ifp->if_bpf);
1197 	CURVNET_RESTORE();
1198 }
1199 
1200 /*
1201  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1202  */
1203 static int
if_vmove_loan(struct thread * td,char * ifname,int jid)1204 if_vmove_loan(struct thread *td, char *ifname, int jid)
1205 {
1206 	struct prison *pr;
1207 	struct ifnet *ifp, *difp;
1208 	bool found;
1209 
1210 	MPASS(curthread == td);
1211 	MPASS(curvnet == TD_TO_VNET(td));
1212 
1213 	/*
1214 	 * We check the existence of the interface, and will later try to
1215 	 * unlink it from the "active" list, so it is sufficient to only
1216 	 * hold a weak reference to it.
1217 	 * Be aware that it is unsafe to access any member of it, until it
1218 	 * is proven to be safe to ( say it was on the "active" list ).
1219 	 */
1220 	ifp = ifunit(ifname);
1221 	if (ifp == NULL)
1222 		return (ENXIO);
1223 
1224 	/* Try to find the prison within our visibility. */
1225 	sx_slock(&allprison_lock);
1226 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1227 	sx_sunlock(&allprison_lock);
1228 	if (pr == NULL)
1229 		return (ENXIO);
1230 	/* Do not try to move the iface from and to the same vnet. */
1231 	if (pr->pr_vnet == TD_TO_VNET(td)) {
1232 		mtx_unlock(&pr->pr_mtx);
1233 		return (EEXIST);
1234 	}
1235 	prison_hold_locked(pr);
1236 	mtx_unlock(&pr->pr_mtx);
1237 
1238 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1239 	/* XXX Lock interfaces to avoid races. */
1240 	CURVNET_SET_QUIET(pr->pr_vnet);
1241 	difp = ifunit(ifname);
1242 	CURVNET_RESTORE();
1243 	if (difp != NULL) {
1244 		prison_free(pr);
1245 		return (EEXIST);
1246 	}
1247 	sx_xlock(&ifnet_detach_sxlock);
1248 
1249 	found = if_unlink_ifnet(ifp, true);
1250 	if (! found) {
1251 		sx_xunlock(&ifnet_detach_sxlock);
1252 		prison_free(pr);
1253 		return (ENODEV);
1254 	}
1255 
1256 	/* Move the interface into the child jail/vnet. */
1257 	if_vmove(ifp, pr->pr_vnet);
1258 
1259 	/* Report the new if_xname back to the userland. */
1260 	sprintf(ifname, "%s", ifp->if_xname);
1261 
1262 	sx_xunlock(&ifnet_detach_sxlock);
1263 
1264 	prison_free(pr);
1265 	return (0);
1266 }
1267 
1268 static int
if_vmove_reclaim(struct thread * td,char * ifname,int jid)1269 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1270 {
1271 	struct prison *pr;
1272 	struct vnet *vnet_dst;
1273 	struct ifnet *ifp;
1274 	int found;
1275 
1276 	/* Try to find the prison within our visibility. */
1277 	sx_slock(&allprison_lock);
1278 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1279 	sx_sunlock(&allprison_lock);
1280 	if (pr == NULL)
1281 		return (ENXIO);
1282 	prison_hold_locked(pr);
1283 	mtx_unlock(&pr->pr_mtx);
1284 
1285 	/* Make sure the named iface exists in the source prison/vnet. */
1286 	CURVNET_SET(pr->pr_vnet);
1287 	ifp = ifunit(ifname);
1288 	if (ifp == NULL) {
1289 		CURVNET_RESTORE();
1290 		prison_free(pr);
1291 		return (ENXIO);
1292 	}
1293 
1294 	/* Do not try to move the iface from and to the same vnet. */
1295 	vnet_dst = TD_TO_VNET(td);
1296 	if (vnet_dst == pr->pr_vnet) {
1297 		CURVNET_RESTORE();
1298 		prison_free(pr);
1299 		return (EEXIST);
1300 	}
1301 
1302 	/* Get interface back from child jail/vnet. */
1303 	sx_xlock(&ifnet_detach_sxlock);
1304 	found = if_unlink_ifnet(ifp, true);
1305 	if (! found) {
1306 		sx_xunlock(&ifnet_detach_sxlock);
1307 		CURVNET_RESTORE();
1308 		prison_free(pr);
1309 		return (ENODEV);
1310 	}
1311 	if_vmove(ifp, vnet_dst);
1312 	sx_xunlock(&ifnet_detach_sxlock);
1313 	CURVNET_RESTORE();
1314 
1315 	/* Report the new if_xname back to the userland. */
1316 	sprintf(ifname, "%s", ifp->if_xname);
1317 
1318 	prison_free(pr);
1319 	return (0);
1320 }
1321 #endif /* VIMAGE */
1322 
1323 /*
1324  * Add a group to an interface
1325  */
1326 int
if_addgroup(struct ifnet * ifp,const char * groupname)1327 if_addgroup(struct ifnet *ifp, const char *groupname)
1328 {
1329 	struct ifg_list		*ifgl;
1330 	struct ifg_group	*ifg = NULL;
1331 	struct ifg_member	*ifgm;
1332 	int 			 new = 0;
1333 
1334 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1335 	    groupname[strlen(groupname) - 1] <= '9')
1336 		return (EINVAL);
1337 
1338 	IFNET_WLOCK();
1339 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1340 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1341 			IFNET_WUNLOCK();
1342 			return (EEXIST);
1343 		}
1344 
1345 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1346 	    	IFNET_WUNLOCK();
1347 		return (ENOMEM);
1348 	}
1349 
1350 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1351 		free(ifgl, M_TEMP);
1352 		IFNET_WUNLOCK();
1353 		return (ENOMEM);
1354 	}
1355 
1356 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1357 		if (!strcmp(ifg->ifg_group, groupname))
1358 			break;
1359 
1360 	if (ifg == NULL) {
1361 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1362 			free(ifgl, M_TEMP);
1363 			free(ifgm, M_TEMP);
1364 			IFNET_WUNLOCK();
1365 			return (ENOMEM);
1366 		}
1367 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1368 		ifg->ifg_refcnt = 0;
1369 		CK_STAILQ_INIT(&ifg->ifg_members);
1370 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1371 		new = 1;
1372 	}
1373 
1374 	ifg->ifg_refcnt++;
1375 	ifgl->ifgl_group = ifg;
1376 	ifgm->ifgm_ifp = ifp;
1377 
1378 	IF_ADDR_WLOCK(ifp);
1379 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1380 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1381 	IF_ADDR_WUNLOCK(ifp);
1382 
1383 	IFNET_WUNLOCK();
1384 
1385 	if (new)
1386 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1387 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1388 
1389 	return (0);
1390 }
1391 
1392 /*
1393  * Helper function to remove a group out of an interface.  Expects the global
1394  * ifnet lock to be write-locked, and drops it before returning.
1395  */
1396 static void
_if_delgroup_locked(struct ifnet * ifp,struct ifg_list * ifgl,const char * groupname)1397 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1398     const char *groupname)
1399 {
1400 	struct ifg_member *ifgm;
1401 	bool freeifgl;
1402 
1403 	IFNET_WLOCK_ASSERT();
1404 
1405 	IF_ADDR_WLOCK(ifp);
1406 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1407 	IF_ADDR_WUNLOCK(ifp);
1408 
1409 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1410 		if (ifgm->ifgm_ifp == ifp) {
1411 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1412 			    ifg_member, ifgm_next);
1413 			break;
1414 		}
1415 	}
1416 
1417 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1418 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1419 		    ifg_next);
1420 		freeifgl = true;
1421 	} else {
1422 		freeifgl = false;
1423 	}
1424 	IFNET_WUNLOCK();
1425 
1426 	NET_EPOCH_WAIT();
1427 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1428 	if (freeifgl) {
1429 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1430 		free(ifgl->ifgl_group, M_TEMP);
1431 	}
1432 	free(ifgm, M_TEMP);
1433 	free(ifgl, M_TEMP);
1434 }
1435 
1436 /*
1437  * Remove a group from an interface
1438  */
1439 int
if_delgroup(struct ifnet * ifp,const char * groupname)1440 if_delgroup(struct ifnet *ifp, const char *groupname)
1441 {
1442 	struct ifg_list *ifgl;
1443 
1444 	IFNET_WLOCK();
1445 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1446 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1447 			break;
1448 	if (ifgl == NULL) {
1449 		IFNET_WUNLOCK();
1450 		return (ENOENT);
1451 	}
1452 
1453 	_if_delgroup_locked(ifp, ifgl, groupname);
1454 
1455 	return (0);
1456 }
1457 
1458 /*
1459  * Remove an interface from all groups
1460  */
1461 static void
if_delgroups(struct ifnet * ifp)1462 if_delgroups(struct ifnet *ifp)
1463 {
1464 	struct ifg_list *ifgl;
1465 	char groupname[IFNAMSIZ];
1466 
1467 	IFNET_WLOCK();
1468 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1469 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1470 		_if_delgroup_locked(ifp, ifgl, groupname);
1471 		IFNET_WLOCK();
1472 	}
1473 	IFNET_WUNLOCK();
1474 }
1475 
1476 /*
1477  * Stores all groups from an interface in memory pointed to by ifgr.
1478  */
1479 static int
if_getgroup(struct ifgroupreq * ifgr,struct ifnet * ifp)1480 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1481 {
1482 	int			 len, error;
1483 	struct ifg_list		*ifgl;
1484 	struct ifg_req		 ifgrq, *ifgp;
1485 
1486 	NET_EPOCH_ASSERT();
1487 
1488 	if (ifgr->ifgr_len == 0) {
1489 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1490 			ifgr->ifgr_len += sizeof(struct ifg_req);
1491 		return (0);
1492 	}
1493 
1494 	len = ifgr->ifgr_len;
1495 	ifgp = ifgr->ifgr_groups;
1496 	/* XXX: wire */
1497 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1498 		if (len < sizeof(ifgrq))
1499 			return (EINVAL);
1500 		bzero(&ifgrq, sizeof ifgrq);
1501 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1502 		    sizeof(ifgrq.ifgrq_group));
1503 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1504 			return (error);
1505 		len -= sizeof(ifgrq);
1506 		ifgp++;
1507 	}
1508 
1509 	return (0);
1510 }
1511 
1512 /*
1513  * Stores all members of a group in memory pointed to by igfr
1514  */
1515 static int
if_getgroupmembers(struct ifgroupreq * ifgr)1516 if_getgroupmembers(struct ifgroupreq *ifgr)
1517 {
1518 	struct ifg_group	*ifg;
1519 	struct ifg_member	*ifgm;
1520 	struct ifg_req		 ifgrq, *ifgp;
1521 	int			 len, error;
1522 
1523 	IFNET_RLOCK();
1524 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1525 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1526 			break;
1527 	if (ifg == NULL) {
1528 		IFNET_RUNLOCK();
1529 		return (ENOENT);
1530 	}
1531 
1532 	if (ifgr->ifgr_len == 0) {
1533 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1534 			ifgr->ifgr_len += sizeof(ifgrq);
1535 		IFNET_RUNLOCK();
1536 		return (0);
1537 	}
1538 
1539 	len = ifgr->ifgr_len;
1540 	ifgp = ifgr->ifgr_groups;
1541 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1542 		if (len < sizeof(ifgrq)) {
1543 			IFNET_RUNLOCK();
1544 			return (EINVAL);
1545 		}
1546 		bzero(&ifgrq, sizeof ifgrq);
1547 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1548 		    sizeof(ifgrq.ifgrq_member));
1549 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1550 			IFNET_RUNLOCK();
1551 			return (error);
1552 		}
1553 		len -= sizeof(ifgrq);
1554 		ifgp++;
1555 	}
1556 	IFNET_RUNLOCK();
1557 
1558 	return (0);
1559 }
1560 
1561 /*
1562  * Return counter values from counter(9)s stored in ifnet.
1563  */
1564 uint64_t
if_get_counter_default(struct ifnet * ifp,ift_counter cnt)1565 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1566 {
1567 
1568 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1569 
1570 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1571 }
1572 
1573 /*
1574  * Increase an ifnet counter. Usually used for counters shared
1575  * between the stack and a driver, but function supports them all.
1576  */
1577 void
if_inc_counter(struct ifnet * ifp,ift_counter cnt,int64_t inc)1578 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1579 {
1580 
1581 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1582 
1583 	counter_u64_add(ifp->if_counters[cnt], inc);
1584 }
1585 
1586 /*
1587  * Copy data from ifnet to userland API structure if_data.
1588  */
1589 void
if_data_copy(struct ifnet * ifp,struct if_data * ifd)1590 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1591 {
1592 
1593 	ifd->ifi_type = ifp->if_type;
1594 	ifd->ifi_physical = 0;
1595 	ifd->ifi_addrlen = ifp->if_addrlen;
1596 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1597 	ifd->ifi_link_state = ifp->if_link_state;
1598 	ifd->ifi_vhid = 0;
1599 	ifd->ifi_datalen = sizeof(struct if_data);
1600 	ifd->ifi_mtu = ifp->if_mtu;
1601 	ifd->ifi_metric = ifp->if_metric;
1602 	ifd->ifi_baudrate = ifp->if_baudrate;
1603 	ifd->ifi_hwassist = ifp->if_hwassist;
1604 	ifd->ifi_epoch = ifp->if_epoch;
1605 	ifd->ifi_lastchange = ifp->if_lastchange;
1606 
1607 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1608 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1609 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1610 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1611 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1612 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1613 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1614 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1615 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1616 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1617 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1618 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1619 }
1620 
1621 /*
1622  * Initialization, destruction and refcounting functions for ifaddrs.
1623  */
1624 struct ifaddr *
ifa_alloc(size_t size,int flags)1625 ifa_alloc(size_t size, int flags)
1626 {
1627 	struct ifaddr *ifa;
1628 
1629 	KASSERT(size >= sizeof(struct ifaddr),
1630 	    ("%s: invalid size %zu", __func__, size));
1631 
1632 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1633 	if (ifa == NULL)
1634 		return (NULL);
1635 
1636 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1637 		goto fail;
1638 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1639 		goto fail;
1640 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1641 		goto fail;
1642 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1643 		goto fail;
1644 
1645 	refcount_init(&ifa->ifa_refcnt, 1);
1646 
1647 	return (ifa);
1648 
1649 fail:
1650 	/* free(NULL) is okay */
1651 	counter_u64_free(ifa->ifa_opackets);
1652 	counter_u64_free(ifa->ifa_ipackets);
1653 	counter_u64_free(ifa->ifa_obytes);
1654 	counter_u64_free(ifa->ifa_ibytes);
1655 	free(ifa, M_IFADDR);
1656 
1657 	return (NULL);
1658 }
1659 
1660 void
ifa_ref(struct ifaddr * ifa)1661 ifa_ref(struct ifaddr *ifa)
1662 {
1663 	u_int old __diagused;
1664 
1665 	old = refcount_acquire(&ifa->ifa_refcnt);
1666 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1667 }
1668 
1669 int
ifa_try_ref(struct ifaddr * ifa)1670 ifa_try_ref(struct ifaddr *ifa)
1671 {
1672 
1673 	NET_EPOCH_ASSERT();
1674 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1675 }
1676 
1677 static void
ifa_destroy(epoch_context_t ctx)1678 ifa_destroy(epoch_context_t ctx)
1679 {
1680 	struct ifaddr *ifa;
1681 
1682 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1683 	counter_u64_free(ifa->ifa_opackets);
1684 	counter_u64_free(ifa->ifa_ipackets);
1685 	counter_u64_free(ifa->ifa_obytes);
1686 	counter_u64_free(ifa->ifa_ibytes);
1687 	free(ifa, M_IFADDR);
1688 }
1689 
1690 void
ifa_free(struct ifaddr * ifa)1691 ifa_free(struct ifaddr *ifa)
1692 {
1693 
1694 	if (refcount_release(&ifa->ifa_refcnt))
1695 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1696 }
1697 
1698 /*
1699  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1700  * structs used to represent other address families, it is necessary
1701  * to perform a different comparison.
1702  */
1703 static bool
sa_dl_equal(const struct sockaddr * a,const struct sockaddr * b)1704 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b)
1705 {
1706 	const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a;
1707 	const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b;
1708 
1709 	return (sdl1->sdl_len == sdl2->sdl_len &&
1710 	    bcmp(sdl1->sdl_data + sdl1->sdl_nlen,
1711 	    sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0);
1712 }
1713 
1714 /*
1715  * Locate an interface based on a complete address.
1716  */
1717 /*ARGSUSED*/
1718 struct ifaddr *
ifa_ifwithaddr(const struct sockaddr * addr)1719 ifa_ifwithaddr(const struct sockaddr *addr)
1720 {
1721 	struct ifnet *ifp;
1722 	struct ifaddr *ifa;
1723 
1724 	NET_EPOCH_ASSERT();
1725 
1726 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1727 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1728 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1729 				continue;
1730 			if (sa_equal(addr, ifa->ifa_addr)) {
1731 				goto done;
1732 			}
1733 			/* IP6 doesn't have broadcast */
1734 			if ((ifp->if_flags & IFF_BROADCAST) &&
1735 			    ifa->ifa_broadaddr &&
1736 			    ifa->ifa_broadaddr->sa_len != 0 &&
1737 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1738 				goto done;
1739 			}
1740 		}
1741 	}
1742 	ifa = NULL;
1743 done:
1744 	return (ifa);
1745 }
1746 
1747 int
ifa_ifwithaddr_check(const struct sockaddr * addr)1748 ifa_ifwithaddr_check(const struct sockaddr *addr)
1749 {
1750 	struct epoch_tracker et;
1751 	int rc;
1752 
1753 	NET_EPOCH_ENTER(et);
1754 	rc = (ifa_ifwithaddr(addr) != NULL);
1755 	NET_EPOCH_EXIT(et);
1756 	return (rc);
1757 }
1758 
1759 /*
1760  * Locate an interface based on the broadcast address.
1761  */
1762 /* ARGSUSED */
1763 struct ifaddr *
ifa_ifwithbroadaddr(const struct sockaddr * addr,int fibnum)1764 ifa_ifwithbroadaddr(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 ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1772 			continue;
1773 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1774 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1775 				continue;
1776 			if ((ifp->if_flags & IFF_BROADCAST) &&
1777 			    ifa->ifa_broadaddr &&
1778 			    ifa->ifa_broadaddr->sa_len != 0 &&
1779 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1780 				goto done;
1781 			}
1782 		}
1783 	}
1784 	ifa = NULL;
1785 done:
1786 	return (ifa);
1787 }
1788 
1789 /*
1790  * Locate the point to point interface with a given destination address.
1791  */
1792 /*ARGSUSED*/
1793 struct ifaddr *
ifa_ifwithdstaddr(const struct sockaddr * addr,int fibnum)1794 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1795 {
1796 	struct ifnet *ifp;
1797 	struct ifaddr *ifa;
1798 
1799 	NET_EPOCH_ASSERT();
1800 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1801 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1802 			continue;
1803 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1804 			continue;
1805 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1806 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1807 				continue;
1808 			if (ifa->ifa_dstaddr != NULL &&
1809 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1810 				goto done;
1811 			}
1812 		}
1813 	}
1814 	ifa = NULL;
1815 done:
1816 	return (ifa);
1817 }
1818 
1819 /*
1820  * Find an interface on a specific network.  If many, choice
1821  * is most specific found.
1822  */
1823 struct ifaddr *
ifa_ifwithnet(const struct sockaddr * addr,int ignore_ptp,int fibnum)1824 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1825 {
1826 	struct ifnet *ifp;
1827 	struct ifaddr *ifa;
1828 	struct ifaddr *ifa_maybe = NULL;
1829 	u_int af = addr->sa_family;
1830 	const char *addr_data = addr->sa_data, *cplim;
1831 
1832 	NET_EPOCH_ASSERT();
1833 	/*
1834 	 * AF_LINK addresses can be looked up directly by their index number,
1835 	 * so do that if we can.
1836 	 */
1837 	if (af == AF_LINK) {
1838 		ifp = ifnet_byindex(
1839 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1840 		return (ifp ? ifp->if_addr : NULL);
1841 	}
1842 
1843 	/*
1844 	 * Scan though each interface, looking for ones that have addresses
1845 	 * in this address family and the requested fib.
1846 	 */
1847 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1848 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1849 			continue;
1850 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1851 			const char *cp, *cp2, *cp3;
1852 
1853 			if (ifa->ifa_addr->sa_family != af)
1854 next:				continue;
1855 			if (af == AF_INET &&
1856 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1857 				/*
1858 				 * This is a bit broken as it doesn't
1859 				 * take into account that the remote end may
1860 				 * be a single node in the network we are
1861 				 * looking for.
1862 				 * The trouble is that we don't know the
1863 				 * netmask for the remote end.
1864 				 */
1865 				if (ifa->ifa_dstaddr != NULL &&
1866 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1867 					goto done;
1868 				}
1869 			} else {
1870 				/*
1871 				 * Scan all the bits in the ifa's address.
1872 				 * If a bit dissagrees with what we are
1873 				 * looking for, mask it with the netmask
1874 				 * to see if it really matters.
1875 				 * (A byte at a time)
1876 				 */
1877 				if (ifa->ifa_netmask == 0)
1878 					continue;
1879 				cp = addr_data;
1880 				cp2 = ifa->ifa_addr->sa_data;
1881 				cp3 = ifa->ifa_netmask->sa_data;
1882 				cplim = ifa->ifa_netmask->sa_len
1883 					+ (char *)ifa->ifa_netmask;
1884 				while (cp3 < cplim)
1885 					if ((*cp++ ^ *cp2++) & *cp3++)
1886 						goto next; /* next address! */
1887 				/*
1888 				 * If the netmask of what we just found
1889 				 * is more specific than what we had before
1890 				 * (if we had one), or if the virtual status
1891 				 * of new prefix is better than of the old one,
1892 				 * then remember the new one before continuing
1893 				 * to search for an even better one.
1894 				 */
1895 				if (ifa_maybe == NULL ||
1896 				    ifa_preferred(ifa_maybe, ifa) ||
1897 				    rn_refines((caddr_t)ifa->ifa_netmask,
1898 				    (caddr_t)ifa_maybe->ifa_netmask)) {
1899 					ifa_maybe = ifa;
1900 				}
1901 			}
1902 		}
1903 	}
1904 	ifa = ifa_maybe;
1905 	ifa_maybe = NULL;
1906 done:
1907 	return (ifa);
1908 }
1909 
1910 /*
1911  * Find an interface address specific to an interface best matching
1912  * a given address.
1913  */
1914 struct ifaddr *
ifaof_ifpforaddr(const struct sockaddr * addr,struct ifnet * ifp)1915 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1916 {
1917 	struct ifaddr *ifa;
1918 	const char *cp, *cp2, *cp3;
1919 	char *cplim;
1920 	struct ifaddr *ifa_maybe = NULL;
1921 	u_int af = addr->sa_family;
1922 
1923 	if (af >= AF_MAX)
1924 		return (NULL);
1925 
1926 	NET_EPOCH_ASSERT();
1927 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1928 		if (ifa->ifa_addr->sa_family != af)
1929 			continue;
1930 		if (ifa_maybe == NULL)
1931 			ifa_maybe = ifa;
1932 		if (ifa->ifa_netmask == 0) {
1933 			if (sa_equal(addr, ifa->ifa_addr) ||
1934 			    (ifa->ifa_dstaddr &&
1935 			    sa_equal(addr, ifa->ifa_dstaddr)))
1936 				goto done;
1937 			continue;
1938 		}
1939 		if (ifp->if_flags & IFF_POINTOPOINT) {
1940 			if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))
1941 				goto done;
1942 		} else {
1943 			cp = addr->sa_data;
1944 			cp2 = ifa->ifa_addr->sa_data;
1945 			cp3 = ifa->ifa_netmask->sa_data;
1946 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1947 			for (; cp3 < cplim; cp3++)
1948 				if ((*cp++ ^ *cp2++) & *cp3)
1949 					break;
1950 			if (cp3 == cplim)
1951 				goto done;
1952 		}
1953 	}
1954 	ifa = ifa_maybe;
1955 done:
1956 	return (ifa);
1957 }
1958 
1959 /*
1960  * See whether new ifa is better than current one:
1961  * 1) A non-virtual one is preferred over virtual.
1962  * 2) A virtual in master state preferred over any other state.
1963  *
1964  * Used in several address selecting functions.
1965  */
1966 int
ifa_preferred(struct ifaddr * cur,struct ifaddr * next)1967 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
1968 {
1969 
1970 	return (cur->ifa_carp && (!next->ifa_carp ||
1971 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
1972 }
1973 
1974 struct sockaddr_dl *
link_alloc_sdl(size_t size,int flags)1975 link_alloc_sdl(size_t size, int flags)
1976 {
1977 
1978 	return (malloc(size, M_TEMP, flags));
1979 }
1980 
1981 void
link_free_sdl(struct sockaddr * sa)1982 link_free_sdl(struct sockaddr *sa)
1983 {
1984 	free(sa, M_TEMP);
1985 }
1986 
1987 /*
1988  * Fills in given sdl with interface basic info.
1989  * Returns pointer to filled sdl.
1990  */
1991 struct sockaddr_dl *
link_init_sdl(struct ifnet * ifp,struct sockaddr * paddr,u_char iftype)1992 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
1993 {
1994 	struct sockaddr_dl *sdl;
1995 
1996 	sdl = (struct sockaddr_dl *)paddr;
1997 	memset(sdl, 0, sizeof(struct sockaddr_dl));
1998 	sdl->sdl_len = sizeof(struct sockaddr_dl);
1999 	sdl->sdl_family = AF_LINK;
2000 	sdl->sdl_index = ifp->if_index;
2001 	sdl->sdl_type = iftype;
2002 
2003 	return (sdl);
2004 }
2005 
2006 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2007 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2008 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2009 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2010 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2011 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2012 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2013 void	*(*vlan_cookie_p)(struct ifnet *);
2014 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
2015 
2016 /*
2017  * Handle a change in the interface link state. To avoid LORs
2018  * between driver lock and upper layer locks, as well as possible
2019  * recursions, we post event to taskqueue, and all job
2020  * is done in static do_link_state_change().
2021  */
2022 void
if_link_state_change(struct ifnet * ifp,int link_state)2023 if_link_state_change(struct ifnet *ifp, int link_state)
2024 {
2025 	/* Return if state hasn't changed. */
2026 	if (ifp->if_link_state == link_state)
2027 		return;
2028 
2029 	ifp->if_link_state = link_state;
2030 
2031 	/* XXXGL: reference ifp? */
2032 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2033 }
2034 
2035 static void
do_link_state_change(void * arg,int pending)2036 do_link_state_change(void *arg, int pending)
2037 {
2038 	struct ifnet *ifp;
2039 	int link_state;
2040 
2041 	ifp = arg;
2042 	link_state = ifp->if_link_state;
2043 
2044 	CURVNET_SET(ifp->if_vnet);
2045 	rt_ifmsg(ifp, 0);
2046 	if (ifp->if_vlantrunk != NULL)
2047 		(*vlan_link_state_p)(ifp);
2048 	if (ifp->if_carp)
2049 		(*carp_linkstate_p)(ifp);
2050 	if (ifp->if_bridge)
2051 		ifp->if_bridge_linkstate(ifp);
2052 	if (ifp->if_lagg)
2053 		(*lagg_linkstate_p)(ifp, link_state);
2054 
2055 	if (IS_DEFAULT_VNET(curvnet))
2056 		devctl_notify("IFNET", ifp->if_xname,
2057 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2058 		    NULL);
2059 	if (pending > 1)
2060 		if_printf(ifp, "%d link states coalesced\n", pending);
2061 	if (log_link_state_change)
2062 		if_printf(ifp, "link state changed to %s\n",
2063 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2064 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2065 	CURVNET_RESTORE();
2066 }
2067 
2068 /*
2069  * Mark an interface down and notify protocols of
2070  * the transition.
2071  */
2072 void
if_down(struct ifnet * ifp)2073 if_down(struct ifnet *ifp)
2074 {
2075 
2076 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2077 
2078 	ifp->if_flags &= ~IFF_UP;
2079 	getmicrotime(&ifp->if_lastchange);
2080 	ifp->if_qflush(ifp);
2081 
2082 	if (ifp->if_carp)
2083 		(*carp_linkstate_p)(ifp);
2084 	rt_ifmsg(ifp, IFF_UP);
2085 }
2086 
2087 /*
2088  * Mark an interface up and notify protocols of
2089  * the transition.
2090  */
2091 void
if_up(struct ifnet * ifp)2092 if_up(struct ifnet *ifp)
2093 {
2094 
2095 	ifp->if_flags |= IFF_UP;
2096 	getmicrotime(&ifp->if_lastchange);
2097 	if (ifp->if_carp)
2098 		(*carp_linkstate_p)(ifp);
2099 	rt_ifmsg(ifp, IFF_UP);
2100 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2101 }
2102 
2103 /*
2104  * Flush an interface queue.
2105  */
2106 void
if_qflush(struct ifnet * ifp)2107 if_qflush(struct ifnet *ifp)
2108 {
2109 	struct mbuf *m, *n;
2110 	struct ifaltq *ifq;
2111 
2112 	ifq = &ifp->if_snd;
2113 	IFQ_LOCK(ifq);
2114 #ifdef ALTQ
2115 	if (ALTQ_IS_ENABLED(ifq))
2116 		ALTQ_PURGE(ifq);
2117 #endif
2118 	n = ifq->ifq_head;
2119 	while ((m = n) != NULL) {
2120 		n = m->m_nextpkt;
2121 		m_freem(m);
2122 	}
2123 	ifq->ifq_head = 0;
2124 	ifq->ifq_tail = 0;
2125 	ifq->ifq_len = 0;
2126 	IFQ_UNLOCK(ifq);
2127 }
2128 
2129 /*
2130  * Map interface name to interface structure pointer, with or without
2131  * returning a reference.
2132  */
2133 struct ifnet *
ifunit_ref(const char * name)2134 ifunit_ref(const char *name)
2135 {
2136 	struct epoch_tracker et;
2137 	struct ifnet *ifp;
2138 
2139 	NET_EPOCH_ENTER(et);
2140 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2141 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2142 		    !(ifp->if_flags & IFF_DYING)) {
2143 			MPASS(ifp->if_vnet == curvnet);
2144 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2145 			if_ref(ifp);
2146 			break;
2147 		}
2148 	}
2149 	NET_EPOCH_EXIT(et);
2150 	return (ifp);
2151 }
2152 
2153 struct ifnet *
ifunit(const char * name)2154 ifunit(const char *name)
2155 {
2156 	struct epoch_tracker et;
2157 	struct ifnet *ifp;
2158 
2159 	NET_EPOCH_ENTER(et);
2160 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2161 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) {
2162 			MPASS(refcount_load(&ifp->if_refcount) > 0);
2163 			MPASS(ifp->if_vnet == curvnet);
2164 			MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2165 			break;
2166 		}
2167 	}
2168 	NET_EPOCH_EXIT(et);
2169 	return (ifp);
2170 }
2171 
2172 void *
ifr_buffer_get_buffer(void * data)2173 ifr_buffer_get_buffer(void *data)
2174 {
2175 	union ifreq_union *ifrup;
2176 
2177 	ifrup = data;
2178 #ifdef COMPAT_FREEBSD32
2179 	if (SV_CURPROC_FLAG(SV_ILP32))
2180 		return ((void *)(uintptr_t)
2181 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2182 #endif
2183 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2184 }
2185 
2186 static void
ifr_buffer_set_buffer_null(void * data)2187 ifr_buffer_set_buffer_null(void *data)
2188 {
2189 	union ifreq_union *ifrup;
2190 
2191 	ifrup = data;
2192 #ifdef COMPAT_FREEBSD32
2193 	if (SV_CURPROC_FLAG(SV_ILP32))
2194 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2195 	else
2196 #endif
2197 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2198 }
2199 
2200 size_t
ifr_buffer_get_length(void * data)2201 ifr_buffer_get_length(void *data)
2202 {
2203 	union ifreq_union *ifrup;
2204 
2205 	ifrup = data;
2206 #ifdef COMPAT_FREEBSD32
2207 	if (SV_CURPROC_FLAG(SV_ILP32))
2208 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2209 #endif
2210 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2211 }
2212 
2213 static void
ifr_buffer_set_length(void * data,size_t len)2214 ifr_buffer_set_length(void *data, size_t len)
2215 {
2216 	union ifreq_union *ifrup;
2217 
2218 	ifrup = data;
2219 #ifdef COMPAT_FREEBSD32
2220 	if (SV_CURPROC_FLAG(SV_ILP32))
2221 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2222 	else
2223 #endif
2224 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2225 }
2226 
2227 void *
ifr_data_get_ptr(void * ifrp)2228 ifr_data_get_ptr(void *ifrp)
2229 {
2230 	union ifreq_union *ifrup;
2231 
2232 	ifrup = ifrp;
2233 #ifdef COMPAT_FREEBSD32
2234 	if (SV_CURPROC_FLAG(SV_ILP32))
2235 		return ((void *)(uintptr_t)
2236 		    ifrup->ifr32.ifr_ifru.ifru_data);
2237 #endif
2238 		return (ifrup->ifr.ifr_ifru.ifru_data);
2239 }
2240 
2241 struct ifcap_nv_bit_name {
2242 	uint64_t cap_bit;
2243 	const char *cap_name;
2244 };
2245 #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2246     .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2247 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2248 	CAPNV(RXCSUM),
2249 	CAPNV(TXCSUM),
2250 	CAPNV(NETCONS),
2251 	CAPNV(VLAN_MTU),
2252 	CAPNV(VLAN_HWTAGGING),
2253 	CAPNV(JUMBO_MTU),
2254 	CAPNV(POLLING),
2255 	CAPNV(VLAN_HWCSUM),
2256 	CAPNV(TSO4),
2257 	CAPNV(TSO6),
2258 	CAPNV(LRO),
2259 	CAPNV(WOL_UCAST),
2260 	CAPNV(WOL_MCAST),
2261 	CAPNV(WOL_MAGIC),
2262 	CAPNV(TOE4),
2263 	CAPNV(TOE6),
2264 	CAPNV(VLAN_HWFILTER),
2265 	CAPNV(VLAN_HWTSO),
2266 	CAPNV(LINKSTATE),
2267 	CAPNV(NETMAP),
2268 	CAPNV(RXCSUM_IPV6),
2269 	CAPNV(TXCSUM_IPV6),
2270 	CAPNV(HWSTATS),
2271 	CAPNV(TXRTLMT),
2272 	CAPNV(HWRXTSTMP),
2273 	CAPNV(MEXTPG),
2274 	CAPNV(TXTLS4),
2275 	CAPNV(TXTLS6),
2276 	CAPNV(VXLAN_HWCSUM),
2277 	CAPNV(VXLAN_HWTSO),
2278 	CAPNV(TXTLS_RTLMT),
2279 	{0, NULL}
2280 };
2281 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \
2282     .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2283 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2284 	CAP2NV(RXTLS4),
2285 	CAP2NV(RXTLS6),
2286 	CAP2NV(IPSEC_OFFLOAD),
2287 	CAP2NV(GENEVE_HWCSUM),
2288 	CAP2NV(GENEVE_HWTSO),
2289 	{0, NULL}
2290 };
2291 #undef CAPNV
2292 #undef CAP2NV
2293 
2294 int
if_capnv_to_capint(const nvlist_t * nv,int * old_cap,const struct ifcap_nv_bit_name * nn,bool all)2295 if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2296     const struct ifcap_nv_bit_name *nn, bool all)
2297 {
2298 	int i, res;
2299 
2300 	res = 0;
2301 	for (i = 0; nn[i].cap_name != NULL; i++) {
2302 		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2303 			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2304 				res |= nn[i].cap_bit;
2305 		} else {
2306 			res |= *old_cap & nn[i].cap_bit;
2307 		}
2308 	}
2309 	return (res);
2310 }
2311 
2312 void
if_capint_to_capnv(nvlist_t * nv,const struct ifcap_nv_bit_name * nn,int ifr_cap,int ifr_req)2313 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2314     int ifr_cap, int ifr_req)
2315 {
2316 	int i;
2317 
2318 	for (i = 0; nn[i].cap_name != NULL; i++) {
2319 		if ((nn[i].cap_bit & ifr_cap) != 0) {
2320 			nvlist_add_bool(nv, nn[i].cap_name,
2321 			    (nn[i].cap_bit & ifr_req) != 0);
2322 		}
2323 	}
2324 }
2325 
2326 /*
2327  * Hardware specific interface ioctls.
2328  */
2329 int
ifhwioctl(u_long cmd,struct ifnet * ifp,caddr_t data,struct thread * td)2330 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2331 {
2332 	struct ifreq *ifr;
2333 	int error = 0, do_ifup = 0;
2334 	int new_flags, temp_flags;
2335 	size_t descrlen, nvbuflen;
2336 	char *descrbuf;
2337 	char new_name[IFNAMSIZ];
2338 	void *buf;
2339 	nvlist_t *nvcap;
2340 	struct siocsifcapnv_driver_data drv_ioctl_data;
2341 
2342 	ifr = (struct ifreq *)data;
2343 	switch (cmd) {
2344 	case SIOCGIFINDEX:
2345 		ifr->ifr_index = ifp->if_index;
2346 		break;
2347 
2348 	case SIOCGIFFLAGS:
2349 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2350 		ifr->ifr_flags = temp_flags & 0xffff;
2351 		ifr->ifr_flagshigh = temp_flags >> 16;
2352 		break;
2353 
2354 	case SIOCGIFCAP:
2355 		ifr->ifr_reqcap = ifp->if_capabilities;
2356 		ifr->ifr_curcap = ifp->if_capenable;
2357 		break;
2358 
2359 	case SIOCGIFCAPNV:
2360 		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2361 			error = EINVAL;
2362 			break;
2363 		}
2364 		buf = NULL;
2365 		nvcap = nvlist_create(0);
2366 		for (;;) {
2367 			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2368 			    ifp->if_capabilities, ifp->if_capenable);
2369 			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2370 			    ifp->if_capabilities2, ifp->if_capenable2);
2371 			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2372 			    __DECONST(caddr_t, nvcap));
2373 			if (error != 0) {
2374 				if_printf(ifp,
2375 			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2376 				    error);
2377 				break;
2378 			}
2379 			buf = nvlist_pack(nvcap, &nvbuflen);
2380 			if (buf == NULL) {
2381 				error = nvlist_error(nvcap);
2382 				if (error == 0)
2383 					error = EDOOFUS;
2384 				break;
2385 			}
2386 			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2387 				ifr->ifr_cap_nv.length = nvbuflen;
2388 				ifr->ifr_cap_nv.buffer = NULL;
2389 				error = EFBIG;
2390 				break;
2391 			}
2392 			ifr->ifr_cap_nv.length = nvbuflen;
2393 			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2394 			break;
2395 		}
2396 		free(buf, M_NVLIST);
2397 		nvlist_destroy(nvcap);
2398 		break;
2399 
2400 	case SIOCGIFDATA:
2401 	{
2402 		struct if_data ifd;
2403 
2404 		/* Ensure uninitialised padding is not leaked. */
2405 		memset(&ifd, 0, sizeof(ifd));
2406 
2407 		if_data_copy(ifp, &ifd);
2408 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2409 		break;
2410 	}
2411 
2412 #ifdef MAC
2413 	case SIOCGIFMAC:
2414 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2415 		break;
2416 #endif
2417 
2418 	case SIOCGIFMETRIC:
2419 		ifr->ifr_metric = ifp->if_metric;
2420 		break;
2421 
2422 	case SIOCGIFMTU:
2423 		ifr->ifr_mtu = ifp->if_mtu;
2424 		break;
2425 
2426 	case SIOCGIFPHYS:
2427 		/* XXXGL: did this ever worked? */
2428 		ifr->ifr_phys = 0;
2429 		break;
2430 
2431 	case SIOCGIFDESCR:
2432 		error = 0;
2433 		sx_slock(&ifdescr_sx);
2434 		if (ifp->if_description == NULL)
2435 			error = ENOMSG;
2436 		else {
2437 			/* space for terminating nul */
2438 			descrlen = strlen(ifp->if_description) + 1;
2439 			if (ifr_buffer_get_length(ifr) < descrlen)
2440 				ifr_buffer_set_buffer_null(ifr);
2441 			else
2442 				error = copyout(ifp->if_description,
2443 				    ifr_buffer_get_buffer(ifr), descrlen);
2444 			ifr_buffer_set_length(ifr, descrlen);
2445 		}
2446 		sx_sunlock(&ifdescr_sx);
2447 		break;
2448 
2449 	case SIOCSIFDESCR:
2450 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2451 		if (error)
2452 			return (error);
2453 
2454 		/*
2455 		 * Copy only (length-1) bytes to make sure that
2456 		 * if_description is always nul terminated.  The
2457 		 * length parameter is supposed to count the
2458 		 * terminating nul in.
2459 		 */
2460 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2461 			return (ENAMETOOLONG);
2462 		else if (ifr_buffer_get_length(ifr) == 0)
2463 			descrbuf = NULL;
2464 		else {
2465 			descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK);
2466 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2467 			    ifr_buffer_get_length(ifr) - 1);
2468 			if (error) {
2469 				if_freedescr(descrbuf);
2470 				break;
2471 			}
2472 		}
2473 
2474 		if_setdescr(ifp, descrbuf);
2475 		getmicrotime(&ifp->if_lastchange);
2476 		break;
2477 
2478 	case SIOCGIFFIB:
2479 		ifr->ifr_fib = ifp->if_fib;
2480 		break;
2481 
2482 	case SIOCSIFFIB:
2483 		error = priv_check(td, PRIV_NET_SETIFFIB);
2484 		if (error)
2485 			return (error);
2486 		if (ifr->ifr_fib >= rt_numfibs)
2487 			return (EINVAL);
2488 
2489 		ifp->if_fib = ifr->ifr_fib;
2490 		break;
2491 
2492 	case SIOCSIFFLAGS:
2493 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2494 		if (error)
2495 			return (error);
2496 		/*
2497 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2498 		 * check, so we don't need special handling here yet.
2499 		 */
2500 		new_flags = (ifr->ifr_flags & 0xffff) |
2501 		    (ifr->ifr_flagshigh << 16);
2502 		if (ifp->if_flags & IFF_UP &&
2503 		    (new_flags & IFF_UP) == 0) {
2504 			if_down(ifp);
2505 		} else if (new_flags & IFF_UP &&
2506 		    (ifp->if_flags & IFF_UP) == 0) {
2507 			do_ifup = 1;
2508 		}
2509 
2510 		/*
2511 		 * See if the promiscuous mode or allmulti bits are about to
2512 		 * flip.  They require special handling because in-kernel
2513 		 * consumers may indepdently toggle them.
2514 		 */
2515 		if_setppromisc(ifp, new_flags & IFF_PPROMISC);
2516 		if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) {
2517 			if (new_flags & IFF_PALLMULTI)
2518 				ifp->if_flags |= IFF_ALLMULTI;
2519 			else if (ifp->if_amcount == 0)
2520 				ifp->if_flags &= ~IFF_ALLMULTI;
2521 		}
2522 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2523 			(new_flags &~ IFF_CANTCHANGE);
2524 		if (ifp->if_ioctl) {
2525 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2526 		}
2527 		if (do_ifup)
2528 			if_up(ifp);
2529 		getmicrotime(&ifp->if_lastchange);
2530 		break;
2531 
2532 	case SIOCSIFCAP:
2533 		error = priv_check(td, PRIV_NET_SETIFCAP);
2534 		if (error != 0)
2535 			return (error);
2536 		if (ifp->if_ioctl == NULL)
2537 			return (EOPNOTSUPP);
2538 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2539 			return (EINVAL);
2540 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2541 		if (error == 0)
2542 			getmicrotime(&ifp->if_lastchange);
2543 		break;
2544 
2545 	case SIOCSIFCAPNV:
2546 		error = priv_check(td, PRIV_NET_SETIFCAP);
2547 		if (error != 0)
2548 			return (error);
2549 		if (ifp->if_ioctl == NULL)
2550 			return (EOPNOTSUPP);
2551 		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2552 			return (EINVAL);
2553 		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2554 			return (EINVAL);
2555 		nvcap = NULL;
2556 		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2557 		for (;;) {
2558 			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2559 			    ifr->ifr_cap_nv.length);
2560 			if (error != 0)
2561 				break;
2562 			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2563 			if (nvcap == NULL) {
2564 				error = EINVAL;
2565 				break;
2566 			}
2567 			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2568 			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2569 			if ((drv_ioctl_data.reqcap &
2570 			    ~ifp->if_capabilities) != 0) {
2571 				error = EINVAL;
2572 				break;
2573 			}
2574 			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2575 			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2576 			if ((drv_ioctl_data.reqcap2 &
2577 			    ~ifp->if_capabilities2) != 0) {
2578 				error = EINVAL;
2579 				break;
2580 			}
2581 			drv_ioctl_data.nvcap = nvcap;
2582 			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2583 			    (caddr_t)&drv_ioctl_data);
2584 			break;
2585 		}
2586 		nvlist_destroy(nvcap);
2587 		free(buf, M_TEMP);
2588 		if (error == 0)
2589 			getmicrotime(&ifp->if_lastchange);
2590 		break;
2591 
2592 #ifdef MAC
2593 	case SIOCSIFMAC:
2594 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2595 		break;
2596 #endif
2597 
2598 	case SIOCSIFNAME:
2599 		error = priv_check(td, PRIV_NET_SETIFNAME);
2600 		if (error)
2601 			return (error);
2602 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2603 		    NULL);
2604 		if (error != 0)
2605 			return (error);
2606 		error = if_rename(ifp, new_name);
2607 		break;
2608 
2609 	case SIOCSIFMETRIC:
2610 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2611 		if (error)
2612 			return (error);
2613 		ifp->if_metric = ifr->ifr_metric;
2614 		getmicrotime(&ifp->if_lastchange);
2615 		break;
2616 
2617 	case SIOCSIFPHYS:
2618 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2619 		if (error)
2620 			return (error);
2621 		if (ifp->if_ioctl == NULL)
2622 			return (EOPNOTSUPP);
2623 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2624 		if (error == 0)
2625 			getmicrotime(&ifp->if_lastchange);
2626 		break;
2627 
2628 	case SIOCSIFMTU:
2629 	{
2630 		u_long oldmtu = ifp->if_mtu;
2631 
2632 		error = priv_check(td, PRIV_NET_SETIFMTU);
2633 		if (error)
2634 			return (error);
2635 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2636 			return (EINVAL);
2637 		if (ifp->if_ioctl == NULL)
2638 			return (EOPNOTSUPP);
2639 		/* Disallow MTU changes on bridge member interfaces. */
2640 		if (ifp->if_bridge)
2641 			return (EOPNOTSUPP);
2642 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2643 		if (error == 0) {
2644 			getmicrotime(&ifp->if_lastchange);
2645 			rt_ifmsg(ifp, 0);
2646 #ifdef INET
2647 			DEBUGNET_NOTIFY_MTU(ifp);
2648 #endif
2649 		}
2650 		/*
2651 		 * If the link MTU changed, do network layer specific procedure.
2652 		 */
2653 		if (ifp->if_mtu != oldmtu)
2654 			if_notifymtu(ifp);
2655 		break;
2656 	}
2657 
2658 	case SIOCADDMULTI:
2659 	case SIOCDELMULTI:
2660 		if (cmd == SIOCADDMULTI)
2661 			error = priv_check(td, PRIV_NET_ADDMULTI);
2662 		else
2663 			error = priv_check(td, PRIV_NET_DELMULTI);
2664 		if (error)
2665 			return (error);
2666 
2667 		/* Don't allow group membership on non-multicast interfaces. */
2668 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2669 			return (EOPNOTSUPP);
2670 
2671 		/* Don't let users screw up protocols' entries. */
2672 		if (ifr->ifr_addr.sa_family != AF_LINK)
2673 			return (EINVAL);
2674 
2675 		if (cmd == SIOCADDMULTI) {
2676 			struct epoch_tracker et;
2677 			struct ifmultiaddr *ifma;
2678 
2679 			/*
2680 			 * Userland is only permitted to join groups once
2681 			 * via the if_addmulti() KPI, because it cannot hold
2682 			 * struct ifmultiaddr * between calls. It may also
2683 			 * lose a race while we check if the membership
2684 			 * already exists.
2685 			 */
2686 			NET_EPOCH_ENTER(et);
2687 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2688 			NET_EPOCH_EXIT(et);
2689 			if (ifma != NULL)
2690 				error = EADDRINUSE;
2691 			else
2692 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2693 		} else {
2694 			error = if_delmulti(ifp, &ifr->ifr_addr);
2695 		}
2696 		if (error == 0)
2697 			getmicrotime(&ifp->if_lastchange);
2698 		break;
2699 
2700 	case SIOCSIFPHYADDR:
2701 	case SIOCDIFPHYADDR:
2702 #ifdef INET6
2703 	case SIOCSIFPHYADDR_IN6:
2704 #endif
2705 	case SIOCSIFMEDIA:
2706 	case SIOCSIFGENERIC:
2707 		error = priv_check(td, PRIV_NET_HWIOCTL);
2708 		if (error)
2709 			return (error);
2710 		if (ifp->if_ioctl == NULL)
2711 			return (EOPNOTSUPP);
2712 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2713 		if (error == 0)
2714 			getmicrotime(&ifp->if_lastchange);
2715 		break;
2716 
2717 	case SIOCGIFSTATUS:
2718 	case SIOCGIFPSRCADDR:
2719 	case SIOCGIFPDSTADDR:
2720 	case SIOCGIFMEDIA:
2721 	case SIOCGIFXMEDIA:
2722 	case SIOCGIFGENERIC:
2723 	case SIOCGIFRSSKEY:
2724 	case SIOCGIFRSSHASH:
2725 	case SIOCGIFDOWNREASON:
2726 		if (ifp->if_ioctl == NULL)
2727 			return (EOPNOTSUPP);
2728 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2729 		break;
2730 
2731 	case SIOCSIFLLADDR:
2732 		error = priv_check(td, PRIV_NET_SETLLADDR);
2733 		if (error)
2734 			return (error);
2735 		error = if_setlladdr(ifp,
2736 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2737 		break;
2738 
2739 	case SIOCGHWADDR:
2740 		error = if_gethwaddr(ifp, ifr);
2741 		break;
2742 
2743 	case SIOCAIFGROUP:
2744 	{
2745 		const char *groupname;
2746 
2747 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2748 		if (error)
2749 			return (error);
2750 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2751 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2752 			return (EINVAL);
2753 		error = if_addgroup(ifp, groupname);
2754 		if (error != 0)
2755 			return (error);
2756 		break;
2757 	}
2758 	case SIOCGIFGROUP:
2759 	{
2760 		struct epoch_tracker et;
2761 
2762 		NET_EPOCH_ENTER(et);
2763 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2764 		NET_EPOCH_EXIT(et);
2765 		break;
2766 	}
2767 
2768 	case SIOCDIFGROUP:
2769 	{
2770 		const char *groupname;
2771 
2772 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2773 		if (error)
2774 			return (error);
2775 		groupname = ((struct ifgroupreq *)data)->ifgr_group;
2776 		if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
2777 			return (EINVAL);
2778 		error = if_delgroup(ifp, groupname);
2779 		if (error != 0)
2780 			return (error);
2781 		break;
2782 	}
2783 	default:
2784 		error = ENOIOCTL;
2785 		break;
2786 	}
2787 	return (error);
2788 }
2789 
2790 /*
2791  * Interface ioctls.
2792  */
2793 int
ifioctl(struct socket * so,u_long cmd,caddr_t data,struct thread * td)2794 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2795 {
2796 #ifdef COMPAT_FREEBSD32
2797 	union {
2798 		struct ifconf ifc;
2799 		struct ifdrv ifd;
2800 		struct ifgroupreq ifgr;
2801 		struct ifmediareq ifmr;
2802 	} thunk;
2803 	u_long saved_cmd;
2804 	struct ifconf32 *ifc32;
2805 	struct ifdrv32 *ifd32;
2806 	struct ifgroupreq32 *ifgr32;
2807 	struct ifmediareq32 *ifmr32;
2808 #endif
2809 	struct ifnet *ifp;
2810 	struct ifreq *ifr;
2811 	int error;
2812 	int oif_flags;
2813 #ifdef VIMAGE
2814 	bool shutdown;
2815 #endif
2816 
2817 	CURVNET_SET(so->so_vnet);
2818 #ifdef VIMAGE
2819 	/* Make sure the VNET is stable. */
2820 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2821 	if (shutdown) {
2822 		CURVNET_RESTORE();
2823 		return (EBUSY);
2824 	}
2825 #endif
2826 
2827 #ifdef COMPAT_FREEBSD32
2828 	saved_cmd = cmd;
2829 	switch (cmd) {
2830 	case SIOCGIFCONF32:
2831 		ifc32 = (struct ifconf32 *)data;
2832 		thunk.ifc.ifc_len = ifc32->ifc_len;
2833 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2834 		data = (caddr_t)&thunk.ifc;
2835 		cmd = SIOCGIFCONF;
2836 		break;
2837 	case SIOCGDRVSPEC32:
2838 	case SIOCSDRVSPEC32:
2839 		ifd32 = (struct ifdrv32 *)data;
2840 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2841 		    sizeof(thunk.ifd.ifd_name));
2842 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2843 		thunk.ifd.ifd_len = ifd32->ifd_len;
2844 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2845 		data = (caddr_t)&thunk.ifd;
2846 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2847 		break;
2848 	case SIOCAIFGROUP32:
2849 	case SIOCGIFGROUP32:
2850 	case SIOCDIFGROUP32:
2851 	case SIOCGIFGMEMB32:
2852 		ifgr32 = (struct ifgroupreq32 *)data;
2853 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2854 		    sizeof(thunk.ifgr.ifgr_name));
2855 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2856 		switch (cmd) {
2857 		case SIOCAIFGROUP32:
2858 		case SIOCDIFGROUP32:
2859 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2860 			    sizeof(thunk.ifgr.ifgr_group));
2861 			break;
2862 		case SIOCGIFGROUP32:
2863 		case SIOCGIFGMEMB32:
2864 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2865 			break;
2866 		}
2867 		data = (caddr_t)&thunk.ifgr;
2868 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2869 		break;
2870 	case SIOCGIFMEDIA32:
2871 	case SIOCGIFXMEDIA32:
2872 		ifmr32 = (struct ifmediareq32 *)data;
2873 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2874 		    sizeof(thunk.ifmr.ifm_name));
2875 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2876 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2877 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2878 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2879 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2880 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2881 		data = (caddr_t)&thunk.ifmr;
2882 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2883 		break;
2884 	}
2885 #endif
2886 
2887 	switch (cmd) {
2888 	case SIOCGIFCONF:
2889 		error = ifconf(cmd, data);
2890 		goto out_noref;
2891 	}
2892 
2893 	ifr = (struct ifreq *)data;
2894 	switch (cmd) {
2895 #ifdef VIMAGE
2896 	case SIOCSIFVNET:
2897 		error = priv_check(td, PRIV_NET_SETIFVNET);
2898 		if (error == 0)
2899 			error = if_vmove_loan(td, ifr->ifr_name, ifr->ifr_jid);
2900 		goto out_noref;
2901 
2902 	case SIOCSIFRVNET:
2903 		error = priv_check(td, PRIV_NET_SETIFVNET);
2904 		if (error == 0)
2905 			error = if_vmove_reclaim(td, ifr->ifr_name,
2906 			    ifr->ifr_jid);
2907 		goto out_noref;
2908 #endif
2909 	case SIOCIFCREATE:
2910 	case SIOCIFCREATE2:
2911 		error = priv_check(td, PRIV_NET_IFCREATE);
2912 		if (error == 0)
2913 			error = if_clone_create(ifr->ifr_name,
2914 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
2915 			    ifr_data_get_ptr(ifr) : NULL);
2916 		goto out_noref;
2917 	case SIOCIFDESTROY:
2918 		error = priv_check(td, PRIV_NET_IFDESTROY);
2919 
2920 		if (error == 0)
2921 			error = if_clone_destroy(ifr->ifr_name);
2922 		goto out_noref;
2923 
2924 	case SIOCIFGCLONERS:
2925 		error = if_clone_list((struct if_clonereq *)data);
2926 		goto out_noref;
2927 
2928 	case SIOCGIFGMEMB:
2929 	{
2930 		struct ifgroupreq *req;
2931 
2932 		req = (struct ifgroupreq *)data;
2933 		if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) {
2934 			error = EINVAL;
2935 			goto out_noref;
2936 		}
2937 		error = if_getgroupmembers(req);
2938 		goto out_noref;
2939 	}
2940 	}
2941 
2942 	ifp = ifunit_ref(ifr->ifr_name);
2943 	if (ifp == NULL) {
2944 		error = ENXIO;
2945 		goto out_noref;
2946 	}
2947 
2948 	error = ifhwioctl(cmd, ifp, data, td);
2949 	if (error != ENOIOCTL)
2950 		goto out_ref;
2951 
2952 	oif_flags = ifp->if_flags;
2953 	if (so->so_proto == NULL) {
2954 		error = EOPNOTSUPP;
2955 		goto out_ref;
2956 	}
2957 
2958 	/*
2959 	 * Pass the request on to the socket control method, and if the
2960 	 * latter returns EOPNOTSUPP, directly to the interface.
2961 	 *
2962 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
2963 	 * trust SIOCSIFADDR et al to come from an already privileged
2964 	 * layer, and do not perform any credentials checks or input
2965 	 * validation.
2966 	 */
2967 	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
2968 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
2969 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
2970 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
2971 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2972 
2973 	if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP))
2974 		if_up(ifp);
2975 out_ref:
2976 	if_rele(ifp);
2977 out_noref:
2978 	CURVNET_RESTORE();
2979 #ifdef COMPAT_FREEBSD32
2980 	if (error != 0)
2981 		return (error);
2982 	switch (saved_cmd) {
2983 	case SIOCGIFCONF32:
2984 		ifc32->ifc_len = thunk.ifc.ifc_len;
2985 		break;
2986 	case SIOCGDRVSPEC32:
2987 		/*
2988 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
2989 		 * the struct so just assert that ifd_len (the only
2990 		 * field it might make sense to update) hasn't
2991 		 * changed.
2992 		 */
2993 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
2994 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
2995 			thunk.ifd.ifd_len));
2996 		break;
2997 	case SIOCGIFGROUP32:
2998 	case SIOCGIFGMEMB32:
2999 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3000 		break;
3001 	case SIOCGIFMEDIA32:
3002 	case SIOCGIFXMEDIA32:
3003 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3004 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3005 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3006 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3007 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3008 		break;
3009 	}
3010 #endif
3011 	return (error);
3012 }
3013 
3014 int
if_rename(struct ifnet * ifp,char * new_name)3015 if_rename(struct ifnet *ifp, char *new_name)
3016 {
3017 	struct ifaddr *ifa;
3018 	struct sockaddr_dl *sdl;
3019 	size_t namelen, onamelen;
3020 	char old_name[IFNAMSIZ];
3021 	char strbuf[IFNAMSIZ + 8];
3022 
3023 	if (new_name[0] == '\0')
3024 		return (EINVAL);
3025 	if (strcmp(new_name, ifp->if_xname) == 0)
3026 		return (0);
3027 	if (ifunit(new_name) != NULL)
3028 		return (EEXIST);
3029 
3030 	if_printf(ifp, "changing name to '%s'\n", new_name);
3031 
3032 	IF_ADDR_WLOCK(ifp);
3033 	strlcpy(old_name, ifp->if_xname, sizeof(old_name));
3034 	strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
3035 	ifa = ifp->if_addr;
3036 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3037 	namelen = strlen(new_name);
3038 	onamelen = sdl->sdl_nlen;
3039 	/*
3040 	 * Move the address if needed.  This is safe because we
3041 	 * allocate space for a name of length IFNAMSIZ when we
3042 	 * create this in if_attach().
3043 	 */
3044 	if (namelen != onamelen) {
3045 		bcopy(sdl->sdl_data + onamelen,
3046 		    sdl->sdl_data + namelen, sdl->sdl_alen);
3047 	}
3048 	bcopy(new_name, sdl->sdl_data, namelen);
3049 	sdl->sdl_nlen = namelen;
3050 	sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
3051 	bzero(sdl->sdl_data, onamelen);
3052 	while (namelen != 0)
3053 		sdl->sdl_data[--namelen] = 0xff;
3054 	IF_ADDR_WUNLOCK(ifp);
3055 
3056 	EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name);
3057 
3058 	snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
3059 	devctl_notify("IFNET", old_name, "RENAME", strbuf);
3060 
3061 	return (0);
3062 }
3063 
3064 /*
3065  * The code common to handling reference counted flags,
3066  * e.g., in ifpromisc() and if_allmulti().
3067  * The "pflag" argument can specify a permanent mode flag to check,
3068  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3069  *
3070  * Only to be used on stack-owned flags, not driver-owned flags.
3071  */
3072 static int
if_setflag(struct ifnet * ifp,int flag,int pflag,int * refcount,int onswitch)3073 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3074 {
3075 	struct ifreq ifr;
3076 	int error;
3077 	int oldflags, oldcount;
3078 
3079 	/* Sanity checks to catch programming errors */
3080 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3081 	    ("%s: setting driver-owned flag %d", __func__, flag));
3082 
3083 	if (onswitch)
3084 		KASSERT(*refcount >= 0,
3085 		    ("%s: increment negative refcount %d for flag %d",
3086 		    __func__, *refcount, flag));
3087 	else
3088 		KASSERT(*refcount > 0,
3089 		    ("%s: decrement non-positive refcount %d for flag %d",
3090 		    __func__, *refcount, flag));
3091 
3092 	/* In case this mode is permanent, just touch refcount */
3093 	if (ifp->if_flags & pflag) {
3094 		*refcount += onswitch ? 1 : -1;
3095 		return (0);
3096 	}
3097 
3098 	/* Save ifnet parameters for if_ioctl() may fail */
3099 	oldcount = *refcount;
3100 	oldflags = ifp->if_flags;
3101 
3102 	/*
3103 	 * See if we aren't the only and touching refcount is enough.
3104 	 * Actually toggle interface flag if we are the first or last.
3105 	 */
3106 	if (onswitch) {
3107 		if ((*refcount)++)
3108 			return (0);
3109 		ifp->if_flags |= flag;
3110 	} else {
3111 		if (--(*refcount))
3112 			return (0);
3113 		ifp->if_flags &= ~flag;
3114 	}
3115 
3116 	/* Call down the driver since we've changed interface flags */
3117 	if (ifp->if_ioctl == NULL) {
3118 		error = EOPNOTSUPP;
3119 		goto recover;
3120 	}
3121 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3122 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3123 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3124 	if (error)
3125 		goto recover;
3126 	/* Notify userland that interface flags have changed */
3127 	rt_ifmsg(ifp, flag);
3128 	return (0);
3129 
3130 recover:
3131 	/* Recover after driver error */
3132 	*refcount = oldcount;
3133 	ifp->if_flags = oldflags;
3134 	return (error);
3135 }
3136 
3137 /*
3138  * Set/clear promiscuous mode on interface ifp based on the truth value
3139  * of pswitch.  The calls are reference counted so that only the first
3140  * "on" request actually has an effect, as does the final "off" request.
3141  * Results are undefined if the "off" and "on" requests are not matched.
3142  */
3143 int
ifpromisc(struct ifnet * ifp,int pswitch)3144 ifpromisc(struct ifnet *ifp, int pswitch)
3145 {
3146 	int error;
3147 	int oldflags = ifp->if_flags;
3148 
3149 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3150 			   &ifp->if_pcount, pswitch);
3151 	/* If promiscuous mode status has changed, log a message */
3152 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3153             log_promisc_mode_change)
3154 		if_printf(ifp, "promiscuous mode %s\n",
3155 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3156 	return (error);
3157 }
3158 
3159 /*
3160  * Return interface configuration
3161  * of system.  List may be used
3162  * in later ioctl's (above) to get
3163  * other information.
3164  */
3165 /*ARGSUSED*/
3166 static int
ifconf(u_long cmd,caddr_t data)3167 ifconf(u_long cmd, caddr_t data)
3168 {
3169 	struct ifconf *ifc = (struct ifconf *)data;
3170 	struct ifnet *ifp;
3171 	struct ifaddr *ifa;
3172 	struct ifreq ifr;
3173 	struct sbuf *sb;
3174 	int error, full = 0, valid_len, max_len;
3175 
3176 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3177 	max_len = maxphys - 1;
3178 
3179 	/* Prevent hostile input from being able to crash the system */
3180 	if (ifc->ifc_len <= 0)
3181 		return (EINVAL);
3182 
3183 again:
3184 	if (ifc->ifc_len <= max_len) {
3185 		max_len = ifc->ifc_len;
3186 		full = 1;
3187 	}
3188 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3189 	max_len = 0;
3190 	valid_len = 0;
3191 
3192 	IFNET_RLOCK();
3193 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3194 		struct epoch_tracker et;
3195 		int addrs;
3196 
3197 		/*
3198 		 * Zero the ifr to make sure we don't disclose the contents
3199 		 * of the stack.
3200 		 */
3201 		memset(&ifr, 0, sizeof(ifr));
3202 
3203 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3204 		    >= sizeof(ifr.ifr_name)) {
3205 			sbuf_delete(sb);
3206 			IFNET_RUNLOCK();
3207 			return (ENAMETOOLONG);
3208 		}
3209 
3210 		addrs = 0;
3211 		NET_EPOCH_ENTER(et);
3212 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3213 			struct sockaddr *sa = ifa->ifa_addr;
3214 
3215 			if (prison_if(curthread->td_ucred, sa) != 0)
3216 				continue;
3217 			addrs++;
3218 			if (sa->sa_len <= sizeof(*sa)) {
3219 				if (sa->sa_len < sizeof(*sa)) {
3220 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3221 					    sizeof(ifr.ifr_ifru.ifru_addr));
3222 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3223 					    sa->sa_len);
3224 				} else
3225 					ifr.ifr_ifru.ifru_addr = *sa;
3226 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3227 				max_len += sizeof(ifr);
3228 			} else {
3229 				sbuf_bcat(sb, &ifr,
3230 				    offsetof(struct ifreq, ifr_addr));
3231 				max_len += offsetof(struct ifreq, ifr_addr);
3232 				sbuf_bcat(sb, sa, sa->sa_len);
3233 				max_len += sa->sa_len;
3234 			}
3235 
3236 			if (sbuf_error(sb) == 0)
3237 				valid_len = sbuf_len(sb);
3238 		}
3239 		NET_EPOCH_EXIT(et);
3240 		if (addrs == 0) {
3241 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3242 			max_len += sizeof(ifr);
3243 
3244 			if (sbuf_error(sb) == 0)
3245 				valid_len = sbuf_len(sb);
3246 		}
3247 	}
3248 	IFNET_RUNLOCK();
3249 
3250 	/*
3251 	 * If we didn't allocate enough space (uncommon), try again.  If
3252 	 * we have already allocated as much space as we are allowed,
3253 	 * return what we've got.
3254 	 */
3255 	if (valid_len != max_len && !full) {
3256 		sbuf_delete(sb);
3257 		goto again;
3258 	}
3259 
3260 	ifc->ifc_len = valid_len;
3261 	sbuf_finish(sb);
3262 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3263 	sbuf_delete(sb);
3264 	return (error);
3265 }
3266 
3267 /*
3268  * Just like ifpromisc(), but for all-multicast-reception mode.
3269  */
3270 int
if_allmulti(struct ifnet * ifp,int onswitch)3271 if_allmulti(struct ifnet *ifp, int onswitch)
3272 {
3273 
3274 	return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount,
3275 	    onswitch));
3276 }
3277 
3278 struct ifmultiaddr *
if_findmulti(struct ifnet * ifp,const struct sockaddr * sa)3279 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3280 {
3281 	struct ifmultiaddr *ifma;
3282 
3283 	IF_ADDR_LOCK_ASSERT(ifp);
3284 
3285 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3286 		if (sa->sa_family == AF_LINK) {
3287 			if (sa_dl_equal(ifma->ifma_addr, sa))
3288 				break;
3289 		} else {
3290 			if (sa_equal(ifma->ifma_addr, sa))
3291 				break;
3292 		}
3293 	}
3294 
3295 	return ifma;
3296 }
3297 
3298 /*
3299  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3300  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3301  * the ifnet multicast address list here, so the caller must do that and
3302  * other setup work (such as notifying the device driver).  The reference
3303  * count is initialized to 1.
3304  */
3305 static struct ifmultiaddr *
if_allocmulti(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr * llsa,int mflags)3306 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3307     int mflags)
3308 {
3309 	struct ifmultiaddr *ifma;
3310 	struct sockaddr *dupsa;
3311 
3312 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3313 	    M_ZERO);
3314 	if (ifma == NULL)
3315 		return (NULL);
3316 
3317 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3318 	if (dupsa == NULL) {
3319 		free(ifma, M_IFMADDR);
3320 		return (NULL);
3321 	}
3322 	bcopy(sa, dupsa, sa->sa_len);
3323 	ifma->ifma_addr = dupsa;
3324 
3325 	ifma->ifma_ifp = ifp;
3326 	ifma->ifma_refcount = 1;
3327 	ifma->ifma_protospec = NULL;
3328 
3329 	if (llsa == NULL) {
3330 		ifma->ifma_lladdr = NULL;
3331 		return (ifma);
3332 	}
3333 
3334 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3335 	if (dupsa == NULL) {
3336 		free(ifma->ifma_addr, M_IFMADDR);
3337 		free(ifma, M_IFMADDR);
3338 		return (NULL);
3339 	}
3340 	bcopy(llsa, dupsa, llsa->sa_len);
3341 	ifma->ifma_lladdr = dupsa;
3342 
3343 	return (ifma);
3344 }
3345 
3346 /*
3347  * if_freemulti: free ifmultiaddr structure and possibly attached related
3348  * addresses.  The caller is responsible for implementing reference
3349  * counting, notifying the driver, handling routing messages, and releasing
3350  * any dependent link layer state.
3351  */
3352 #ifdef MCAST_VERBOSE
3353 extern void kdb_backtrace(void);
3354 #endif
3355 static void
if_freemulti_internal(struct ifmultiaddr * ifma)3356 if_freemulti_internal(struct ifmultiaddr *ifma)
3357 {
3358 
3359 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3360 	    ifma->ifma_refcount));
3361 
3362 	if (ifma->ifma_lladdr != NULL)
3363 		free(ifma->ifma_lladdr, M_IFMADDR);
3364 #ifdef MCAST_VERBOSE
3365 	kdb_backtrace();
3366 	printf("%s freeing ifma: %p\n", __func__, ifma);
3367 #endif
3368 	free(ifma->ifma_addr, M_IFMADDR);
3369 	free(ifma, M_IFMADDR);
3370 }
3371 
3372 static void
if_destroymulti(epoch_context_t ctx)3373 if_destroymulti(epoch_context_t ctx)
3374 {
3375 	struct ifmultiaddr *ifma;
3376 
3377 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3378 	if_freemulti_internal(ifma);
3379 }
3380 
3381 void
if_freemulti(struct ifmultiaddr * ifma)3382 if_freemulti(struct ifmultiaddr *ifma)
3383 {
3384 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3385 	    ifma->ifma_refcount));
3386 
3387 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3388 }
3389 
3390 /*
3391  * Register an additional multicast address with a network interface.
3392  *
3393  * - If the address is already present, bump the reference count on the
3394  *   address and return.
3395  * - If the address is not link-layer, look up a link layer address.
3396  * - Allocate address structures for one or both addresses, and attach to the
3397  *   multicast address list on the interface.  If automatically adding a link
3398  *   layer address, the protocol address will own a reference to the link
3399  *   layer address, to be freed when it is freed.
3400  * - Notify the network device driver of an addition to the multicast address
3401  *   list.
3402  *
3403  * 'sa' points to caller-owned memory with the desired multicast address.
3404  *
3405  * 'retifma' will be used to return a pointer to the resulting multicast
3406  * address reference, if desired.
3407  */
3408 int
if_addmulti(struct ifnet * ifp,struct sockaddr * sa,struct ifmultiaddr ** retifma)3409 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3410     struct ifmultiaddr **retifma)
3411 {
3412 	struct ifmultiaddr *ifma, *ll_ifma;
3413 	struct sockaddr *llsa;
3414 	struct sockaddr_dl sdl;
3415 	int error;
3416 
3417 #ifdef INET
3418 	IN_MULTI_LIST_UNLOCK_ASSERT();
3419 #endif
3420 #ifdef INET6
3421 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3422 #endif
3423 	/*
3424 	 * If the address is already present, return a new reference to it;
3425 	 * otherwise, allocate storage and set up a new address.
3426 	 */
3427 	IF_ADDR_WLOCK(ifp);
3428 	ifma = if_findmulti(ifp, sa);
3429 	if (ifma != NULL) {
3430 		ifma->ifma_refcount++;
3431 		if (retifma != NULL)
3432 			*retifma = ifma;
3433 		IF_ADDR_WUNLOCK(ifp);
3434 		return (0);
3435 	}
3436 
3437 	/*
3438 	 * The address isn't already present; resolve the protocol address
3439 	 * into a link layer address, and then look that up, bump its
3440 	 * refcount or allocate an ifma for that also.
3441 	 * Most link layer resolving functions returns address data which
3442 	 * fits inside default sockaddr_dl structure. However callback
3443 	 * can allocate another sockaddr structure, in that case we need to
3444 	 * free it later.
3445 	 */
3446 	llsa = NULL;
3447 	ll_ifma = NULL;
3448 	if (ifp->if_resolvemulti != NULL) {
3449 		/* Provide called function with buffer size information */
3450 		sdl.sdl_len = sizeof(sdl);
3451 		llsa = (struct sockaddr *)&sdl;
3452 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3453 		if (error)
3454 			goto unlock_out;
3455 	}
3456 
3457 	/*
3458 	 * Allocate the new address.  Don't hook it up yet, as we may also
3459 	 * need to allocate a link layer multicast address.
3460 	 */
3461 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3462 	if (ifma == NULL) {
3463 		error = ENOMEM;
3464 		goto free_llsa_out;
3465 	}
3466 
3467 	/*
3468 	 * If a link layer address is found, we'll need to see if it's
3469 	 * already present in the address list, or allocate is as well.
3470 	 * When this block finishes, the link layer address will be on the
3471 	 * list.
3472 	 */
3473 	if (llsa != NULL) {
3474 		ll_ifma = if_findmulti(ifp, llsa);
3475 		if (ll_ifma == NULL) {
3476 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3477 			if (ll_ifma == NULL) {
3478 				--ifma->ifma_refcount;
3479 				if_freemulti(ifma);
3480 				error = ENOMEM;
3481 				goto free_llsa_out;
3482 			}
3483 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3484 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3485 			    ifma_link);
3486 		} else
3487 			ll_ifma->ifma_refcount++;
3488 		ifma->ifma_llifma = ll_ifma;
3489 	}
3490 
3491 	/*
3492 	 * We now have a new multicast address, ifma, and possibly a new or
3493 	 * referenced link layer address.  Add the primary address to the
3494 	 * ifnet address list.
3495 	 */
3496 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3497 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3498 
3499 	if (retifma != NULL)
3500 		*retifma = ifma;
3501 
3502 	/*
3503 	 * Must generate the message while holding the lock so that 'ifma'
3504 	 * pointer is still valid.
3505 	 */
3506 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3507 	IF_ADDR_WUNLOCK(ifp);
3508 
3509 	/*
3510 	 * We are certain we have added something, so call down to the
3511 	 * interface to let them know about it.
3512 	 */
3513 	if (ifp->if_ioctl != NULL) {
3514 		if (THREAD_CAN_SLEEP())
3515 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3516 		else
3517 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3518 	}
3519 
3520 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3521 		link_free_sdl(llsa);
3522 
3523 	return (0);
3524 
3525 free_llsa_out:
3526 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3527 		link_free_sdl(llsa);
3528 
3529 unlock_out:
3530 	IF_ADDR_WUNLOCK(ifp);
3531 	return (error);
3532 }
3533 
3534 static void
if_siocaddmulti(void * arg,int pending)3535 if_siocaddmulti(void *arg, int pending)
3536 {
3537 	struct ifnet *ifp;
3538 
3539 	ifp = arg;
3540 #ifdef DIAGNOSTIC
3541 	if (pending > 1)
3542 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3543 #endif
3544 	CURVNET_SET(ifp->if_vnet);
3545 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3546 	CURVNET_RESTORE();
3547 }
3548 
3549 /*
3550  * Delete a multicast group membership by network-layer group address.
3551  *
3552  * Returns ENOENT if the entry could not be found. If ifp no longer
3553  * exists, results are undefined. This entry point should only be used
3554  * from subsystems which do appropriate locking to hold ifp for the
3555  * duration of the call.
3556  * Network-layer protocol domains must use if_delmulti_ifma().
3557  */
3558 int
if_delmulti(struct ifnet * ifp,struct sockaddr * sa)3559 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3560 {
3561 	struct ifmultiaddr *ifma;
3562 	int lastref;
3563 
3564 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3565 
3566 	IF_ADDR_WLOCK(ifp);
3567 	lastref = 0;
3568 	ifma = if_findmulti(ifp, sa);
3569 	if (ifma != NULL)
3570 		lastref = if_delmulti_locked(ifp, ifma, 0);
3571 	IF_ADDR_WUNLOCK(ifp);
3572 
3573 	if (ifma == NULL)
3574 		return (ENOENT);
3575 
3576 	if (lastref && ifp->if_ioctl != NULL) {
3577 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3578 	}
3579 
3580 	return (0);
3581 }
3582 
3583 /*
3584  * Delete all multicast group membership for an interface.
3585  * Should be used to quickly flush all multicast filters.
3586  */
3587 void
if_delallmulti(struct ifnet * ifp)3588 if_delallmulti(struct ifnet *ifp)
3589 {
3590 	struct ifmultiaddr *ifma;
3591 	struct ifmultiaddr *next;
3592 
3593 	IF_ADDR_WLOCK(ifp);
3594 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3595 		if_delmulti_locked(ifp, ifma, 0);
3596 	IF_ADDR_WUNLOCK(ifp);
3597 }
3598 
3599 void
if_delmulti_ifma(struct ifmultiaddr * ifma)3600 if_delmulti_ifma(struct ifmultiaddr *ifma)
3601 {
3602 	if_delmulti_ifma_flags(ifma, 0);
3603 }
3604 
3605 /*
3606  * Delete a multicast group membership by group membership pointer.
3607  * Network-layer protocol domains must use this routine.
3608  *
3609  * It is safe to call this routine if the ifp disappeared.
3610  */
3611 void
if_delmulti_ifma_flags(struct ifmultiaddr * ifma,int flags)3612 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3613 {
3614 	struct ifnet *ifp;
3615 	int lastref;
3616 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3617 #ifdef INET
3618 	IN_MULTI_LIST_UNLOCK_ASSERT();
3619 #endif
3620 	ifp = ifma->ifma_ifp;
3621 #ifdef DIAGNOSTIC
3622 	if (ifp == NULL) {
3623 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3624 	} else {
3625 		struct epoch_tracker et;
3626 		struct ifnet *oifp;
3627 
3628 		NET_EPOCH_ENTER(et);
3629 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3630 			if (ifp == oifp)
3631 				break;
3632 		NET_EPOCH_EXIT(et);
3633 		if (ifp != oifp)
3634 			ifp = NULL;
3635 	}
3636 #endif
3637 	/*
3638 	 * If and only if the ifnet instance exists: Acquire the address lock.
3639 	 */
3640 	if (ifp != NULL)
3641 		IF_ADDR_WLOCK(ifp);
3642 
3643 	lastref = if_delmulti_locked(ifp, ifma, flags);
3644 
3645 	if (ifp != NULL) {
3646 		/*
3647 		 * If and only if the ifnet instance exists:
3648 		 *  Release the address lock.
3649 		 *  If the group was left: update the hardware hash filter.
3650 		 */
3651 		IF_ADDR_WUNLOCK(ifp);
3652 		if (lastref && ifp->if_ioctl != NULL) {
3653 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3654 		}
3655 	}
3656 }
3657 
3658 /*
3659  * Perform deletion of network-layer and/or link-layer multicast address.
3660  *
3661  * Return 0 if the reference count was decremented.
3662  * Return 1 if the final reference was released, indicating that the
3663  * hardware hash filter should be reprogrammed.
3664  */
3665 static int
if_delmulti_locked(struct ifnet * ifp,struct ifmultiaddr * ifma,int detaching)3666 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3667 {
3668 	struct ifmultiaddr *ll_ifma;
3669 
3670 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3671 		KASSERT(ifma->ifma_ifp == ifp,
3672 		    ("%s: inconsistent ifp %p", __func__, ifp));
3673 		IF_ADDR_WLOCK_ASSERT(ifp);
3674 	}
3675 
3676 	ifp = ifma->ifma_ifp;
3677 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3678 
3679 	/*
3680 	 * If the ifnet is detaching, null out references to ifnet,
3681 	 * so that upper protocol layers will notice, and not attempt
3682 	 * to obtain locks for an ifnet which no longer exists. The
3683 	 * routing socket announcement must happen before the ifnet
3684 	 * instance is detached from the system.
3685 	 */
3686 	if (detaching) {
3687 #ifdef DIAGNOSTIC
3688 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3689 #endif
3690 		/*
3691 		 * ifp may already be nulled out if we are being reentered
3692 		 * to delete the ll_ifma.
3693 		 */
3694 		if (ifp != NULL) {
3695 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3696 			ifma->ifma_ifp = NULL;
3697 		}
3698 	}
3699 
3700 	if (--ifma->ifma_refcount > 0)
3701 		return 0;
3702 
3703 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3704 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3705 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3706 	}
3707 	/*
3708 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3709 	 * have been associated with it. Release it first if so.
3710 	 */
3711 	ll_ifma = ifma->ifma_llifma;
3712 	if (ll_ifma != NULL) {
3713 		KASSERT(ifma->ifma_lladdr != NULL,
3714 		    ("%s: llifma w/o lladdr", __func__));
3715 		if (detaching)
3716 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3717 		if (--ll_ifma->ifma_refcount == 0) {
3718 			if (ifp != NULL) {
3719 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3720 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3721 						ifma_link);
3722 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3723 				}
3724 			}
3725 			if_freemulti(ll_ifma);
3726 		}
3727 	}
3728 #ifdef INVARIANTS
3729 	if (ifp) {
3730 		struct ifmultiaddr *ifmatmp;
3731 
3732 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3733 			MPASS(ifma != ifmatmp);
3734 	}
3735 #endif
3736 	if_freemulti(ifma);
3737 	/*
3738 	 * The last reference to this instance of struct ifmultiaddr
3739 	 * was released; the hardware should be notified of this change.
3740 	 */
3741 	return 1;
3742 }
3743 
3744 /*
3745  * Set the link layer address on an interface.
3746  *
3747  * At this time we only support certain types of interfaces,
3748  * and we don't allow the length of the address to change.
3749  *
3750  * Set noinline to be dtrace-friendly
3751  */
3752 __noinline int
if_setlladdr(struct ifnet * ifp,const u_char * lladdr,int len)3753 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3754 {
3755 	struct sockaddr_dl *sdl;
3756 	struct ifaddr *ifa;
3757 	struct ifreq ifr;
3758 
3759 	ifa = ifp->if_addr;
3760 	if (ifa == NULL)
3761 		return (EINVAL);
3762 
3763 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3764 	if (sdl == NULL)
3765 		return (EINVAL);
3766 
3767 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3768 		return (EINVAL);
3769 
3770 	switch (ifp->if_type) {
3771 	case IFT_ETHER:
3772 	case IFT_XETHER:
3773 	case IFT_L2VLAN:
3774 	case IFT_BRIDGE:
3775 	case IFT_IEEE8023ADLAG:
3776 		bcopy(lladdr, LLADDR(sdl), len);
3777 		break;
3778 	default:
3779 		return (ENODEV);
3780 	}
3781 
3782 	/*
3783 	 * If the interface is already up, we need
3784 	 * to re-init it in order to reprogram its
3785 	 * address filter.
3786 	 */
3787 	if ((ifp->if_flags & IFF_UP) != 0) {
3788 		if (ifp->if_ioctl) {
3789 			ifp->if_flags &= ~IFF_UP;
3790 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3791 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3792 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3793 			ifp->if_flags |= IFF_UP;
3794 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3795 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3796 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3797 		}
3798 	}
3799 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3800 
3801 	return (0);
3802 }
3803 
3804 /*
3805  * Compat function for handling basic encapsulation requests.
3806  * Not converted stacks (FDDI, IB, ..) supports traditional
3807  * output model: ARP (and other similar L2 protocols) are handled
3808  * inside output routine, arpresolve/nd6_resolve() returns MAC
3809  * address instead of full prepend.
3810  *
3811  * This function creates calculated header==MAC for IPv4/IPv6 and
3812  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3813  * address families.
3814  */
3815 static int
if_requestencap_default(struct ifnet * ifp,struct if_encap_req * req)3816 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3817 {
3818 	if (req->rtype != IFENCAP_LL)
3819 		return (EOPNOTSUPP);
3820 
3821 	if (req->bufsize < req->lladdr_len)
3822 		return (ENOMEM);
3823 
3824 	switch (req->family) {
3825 	case AF_INET:
3826 	case AF_INET6:
3827 		break;
3828 	default:
3829 		return (EAFNOSUPPORT);
3830 	}
3831 
3832 	/* Copy lladdr to storage as is */
3833 	memmove(req->buf, req->lladdr, req->lladdr_len);
3834 	req->bufsize = req->lladdr_len;
3835 	req->lladdr_off = 0;
3836 
3837 	return (0);
3838 }
3839 
3840 /*
3841  * Tunnel interfaces can nest, also they may cause infinite recursion
3842  * calls when misconfigured. We'll prevent this by detecting loops.
3843  * High nesting level may cause stack exhaustion. We'll prevent this
3844  * by introducing upper limit.
3845  *
3846  * Return 0, if tunnel nesting count is equal or less than limit.
3847  */
3848 int
if_tunnel_check_nesting(struct ifnet * ifp,struct mbuf * m,uint32_t cookie,int limit)3849 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3850     int limit)
3851 {
3852 	struct m_tag *mtag;
3853 	int count;
3854 
3855 	count = 1;
3856 	mtag = NULL;
3857 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3858 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3859 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3860 			return (EIO);
3861 		}
3862 		count++;
3863 	}
3864 	if (count > limit) {
3865 		log(LOG_NOTICE,
3866 		    "%s: if_output recursively called too many times(%d)\n",
3867 		    if_name(ifp), count);
3868 		return (EIO);
3869 	}
3870 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3871 	if (mtag == NULL)
3872 		return (ENOMEM);
3873 	*(struct ifnet **)(mtag + 1) = ifp;
3874 	m_tag_prepend(m, mtag);
3875 	return (0);
3876 }
3877 
3878 /*
3879  * Get the link layer address that was read from the hardware at attach.
3880  *
3881  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3882  * their component interfaces as IFT_IEEE8023ADLAG.
3883  */
3884 int
if_gethwaddr(struct ifnet * ifp,struct ifreq * ifr)3885 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3886 {
3887 	if (ifp->if_hw_addr == NULL)
3888 		return (ENODEV);
3889 
3890 	switch (ifp->if_type) {
3891 	case IFT_ETHER:
3892 	case IFT_IEEE8023ADLAG:
3893 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3894 		return (0);
3895 	default:
3896 		return (ENODEV);
3897 	}
3898 }
3899 
3900 /*
3901  * The name argument must be a pointer to storage which will last as
3902  * long as the interface does.  For physical devices, the result of
3903  * device_get_name(dev) is a good choice and for pseudo-devices a
3904  * static string works well.
3905  */
3906 void
if_initname(struct ifnet * ifp,const char * name,int unit)3907 if_initname(struct ifnet *ifp, const char *name, int unit)
3908 {
3909 	ifp->if_dname = name;
3910 	ifp->if_dunit = unit;
3911 	if (unit != IF_DUNIT_NONE)
3912 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3913 	else
3914 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3915 }
3916 
3917 static int
if_vlog(struct ifnet * ifp,int pri,const char * fmt,va_list ap)3918 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
3919 {
3920 	char if_fmt[256];
3921 
3922 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
3923 	vlog(pri, if_fmt, ap);
3924 	return (0);
3925 }
3926 
3927 
3928 int
if_printf(struct ifnet * ifp,const char * fmt,...)3929 if_printf(struct ifnet *ifp, const char *fmt, ...)
3930 {
3931 	va_list ap;
3932 
3933 	va_start(ap, fmt);
3934 	if_vlog(ifp, LOG_INFO, fmt, ap);
3935 	va_end(ap);
3936 	return (0);
3937 }
3938 
3939 int
if_log(struct ifnet * ifp,int pri,const char * fmt,...)3940 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
3941 {
3942 	va_list ap;
3943 
3944 	va_start(ap, fmt);
3945 	if_vlog(ifp, pri, fmt, ap);
3946 	va_end(ap);
3947 	return (0);
3948 }
3949 
3950 void
if_start(struct ifnet * ifp)3951 if_start(struct ifnet *ifp)
3952 {
3953 
3954 	(*(ifp)->if_start)(ifp);
3955 }
3956 
3957 /*
3958  * Backwards compatibility interface for drivers
3959  * that have not implemented it
3960  */
3961 static int
if_transmit_default(struct ifnet * ifp,struct mbuf * m)3962 if_transmit_default(struct ifnet *ifp, struct mbuf *m)
3963 {
3964 	int error;
3965 
3966 	IFQ_HANDOFF(ifp, m, error);
3967 	return (error);
3968 }
3969 
3970 static void
if_input_default(struct ifnet * ifp __unused,struct mbuf * m)3971 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3972 {
3973 	m_freem(m);
3974 }
3975 
3976 int
if_handoff(struct ifqueue * ifq,struct mbuf * m,struct ifnet * ifp,int adjust)3977 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3978 {
3979 	int active = 0;
3980 
3981 	IF_LOCK(ifq);
3982 	if (_IF_QFULL(ifq)) {
3983 		IF_UNLOCK(ifq);
3984 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3985 		m_freem(m);
3986 		return (0);
3987 	}
3988 	if (ifp != NULL) {
3989 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3990 		if (m->m_flags & (M_BCAST|M_MCAST))
3991 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3992 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3993 	}
3994 	_IF_ENQUEUE(ifq, m);
3995 	IF_UNLOCK(ifq);
3996 	if (ifp != NULL && !active)
3997 		(*(ifp)->if_start)(ifp);
3998 	return (1);
3999 }
4000 
4001 void
if_register_com_alloc(u_char type,if_com_alloc_t * a,if_com_free_t * f)4002 if_register_com_alloc(u_char type,
4003     if_com_alloc_t *a, if_com_free_t *f)
4004 {
4005 
4006 	KASSERT(if_com_alloc[type] == NULL,
4007 	    ("if_register_com_alloc: %d already registered", type));
4008 	KASSERT(if_com_free[type] == NULL,
4009 	    ("if_register_com_alloc: %d free already registered", type));
4010 
4011 	if_com_alloc[type] = a;
4012 	if_com_free[type] = f;
4013 }
4014 
4015 void
if_deregister_com_alloc(u_char type)4016 if_deregister_com_alloc(u_char type)
4017 {
4018 
4019 	KASSERT(if_com_alloc[type] != NULL,
4020 	    ("if_deregister_com_alloc: %d not registered", type));
4021 	KASSERT(if_com_free[type] != NULL,
4022 	    ("if_deregister_com_alloc: %d free not registered", type));
4023 
4024 	/*
4025 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4026 	 * fixes issues about late invocation of if_destroy(), which leads
4027 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4028 	 */
4029 	NET_EPOCH_DRAIN_CALLBACKS();
4030 
4031 	if_com_alloc[type] = NULL;
4032 	if_com_free[type] = NULL;
4033 }
4034 
4035 /* API for driver access to network stack owned ifnet.*/
4036 uint64_t
if_setbaudrate(struct ifnet * ifp,uint64_t baudrate)4037 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4038 {
4039 	uint64_t oldbrate;
4040 
4041 	oldbrate = ifp->if_baudrate;
4042 	ifp->if_baudrate = baudrate;
4043 	return (oldbrate);
4044 }
4045 
4046 uint64_t
if_getbaudrate(const if_t ifp)4047 if_getbaudrate(const if_t ifp)
4048 {
4049 	return (ifp->if_baudrate);
4050 }
4051 
4052 int
if_setcapabilities(if_t ifp,int capabilities)4053 if_setcapabilities(if_t ifp, int capabilities)
4054 {
4055 	ifp->if_capabilities = capabilities;
4056 	return (0);
4057 }
4058 
4059 int
if_setcapabilitiesbit(if_t ifp,int setbit,int clearbit)4060 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4061 {
4062 	ifp->if_capabilities &= ~clearbit;
4063 	ifp->if_capabilities |= setbit;
4064 	return (0);
4065 }
4066 
4067 int
if_getcapabilities(const if_t ifp)4068 if_getcapabilities(const if_t ifp)
4069 {
4070 	return (ifp->if_capabilities);
4071 }
4072 
4073 int
if_setcapenable(if_t ifp,int capabilities)4074 if_setcapenable(if_t ifp, int capabilities)
4075 {
4076 	ifp->if_capenable = capabilities;
4077 	return (0);
4078 }
4079 
4080 int
if_setcapenablebit(if_t ifp,int setcap,int clearcap)4081 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4082 {
4083 	ifp->if_capenable &= ~clearcap;
4084 	ifp->if_capenable |= setcap;
4085 	return (0);
4086 }
4087 
4088 int
if_setcapabilities2(if_t ifp,int capabilities)4089 if_setcapabilities2(if_t ifp, int capabilities)
4090 {
4091 	ifp->if_capabilities2 = capabilities;
4092 	return (0);
4093 }
4094 
4095 int
if_setcapabilities2bit(if_t ifp,int setbit,int clearbit)4096 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit)
4097 {
4098 	ifp->if_capabilities2 &= ~clearbit;
4099 	ifp->if_capabilities2 |= setbit;
4100 	return (0);
4101 }
4102 
4103 int
if_getcapabilities2(const if_t ifp)4104 if_getcapabilities2(const if_t ifp)
4105 {
4106 	return (ifp->if_capabilities2);
4107 }
4108 
4109 int
if_setcapenable2(if_t ifp,int capabilities2)4110 if_setcapenable2(if_t ifp, int capabilities2)
4111 {
4112 	ifp->if_capenable2 = capabilities2;
4113 	return (0);
4114 }
4115 
4116 int
if_setcapenable2bit(if_t ifp,int setcap,int clearcap)4117 if_setcapenable2bit(if_t ifp, int setcap, int clearcap)
4118 {
4119 	ifp->if_capenable2 &= ~clearcap;
4120 	ifp->if_capenable2 |= setcap;
4121 	return (0);
4122 }
4123 
4124 const char *
if_getdname(const if_t ifp)4125 if_getdname(const if_t ifp)
4126 {
4127 	return (ifp->if_dname);
4128 }
4129 
4130 void
if_setdname(if_t ifp,const char * dname)4131 if_setdname(if_t ifp, const char *dname)
4132 {
4133 	ifp->if_dname = dname;
4134 }
4135 
4136 const char *
if_name(if_t ifp)4137 if_name(if_t ifp)
4138 {
4139 	return (ifp->if_xname);
4140 }
4141 
4142 int
if_setname(if_t ifp,const char * name)4143 if_setname(if_t ifp, const char *name)
4144 {
4145 	if (strlen(name) > sizeof(ifp->if_xname) - 1)
4146 		return (ENAMETOOLONG);
4147 	strcpy(ifp->if_xname, name);
4148 
4149 	return (0);
4150 }
4151 
4152 int
if_togglecapenable(if_t ifp,int togglecap)4153 if_togglecapenable(if_t ifp, int togglecap)
4154 {
4155 	ifp->if_capenable ^= togglecap;
4156 	return (0);
4157 }
4158 
4159 int
if_getcapenable(const if_t ifp)4160 if_getcapenable(const if_t ifp)
4161 {
4162 	return (ifp->if_capenable);
4163 }
4164 
4165 int
if_togglecapenable2(if_t ifp,int togglecap)4166 if_togglecapenable2(if_t ifp, int togglecap)
4167 {
4168 	ifp->if_capenable2 ^= togglecap;
4169 	return (0);
4170 }
4171 
4172 int
if_getcapenable2(const if_t ifp)4173 if_getcapenable2(const if_t ifp)
4174 {
4175 	return (ifp->if_capenable2);
4176 }
4177 
4178 int
if_getdunit(const if_t ifp)4179 if_getdunit(const if_t ifp)
4180 {
4181 	return (ifp->if_dunit);
4182 }
4183 
4184 int
if_getindex(const if_t ifp)4185 if_getindex(const if_t ifp)
4186 {
4187 	return (ifp->if_index);
4188 }
4189 
4190 int
if_getidxgen(const if_t ifp)4191 if_getidxgen(const if_t ifp)
4192 {
4193 	return (ifp->if_idxgen);
4194 }
4195 
4196 const char *
if_getdescr(if_t ifp)4197 if_getdescr(if_t ifp)
4198 {
4199 	return (ifp->if_description);
4200 }
4201 
4202 void
if_setdescr(if_t ifp,char * descrbuf)4203 if_setdescr(if_t ifp, char *descrbuf)
4204 {
4205 	sx_xlock(&ifdescr_sx);
4206 	char *odescrbuf = ifp->if_description;
4207 	ifp->if_description = descrbuf;
4208 	sx_xunlock(&ifdescr_sx);
4209 
4210 	if_freedescr(odescrbuf);
4211 }
4212 
4213 char *
if_allocdescr(size_t sz,int malloc_flag)4214 if_allocdescr(size_t sz, int malloc_flag)
4215 {
4216 	malloc_flag &= (M_WAITOK | M_NOWAIT);
4217 	return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag));
4218 }
4219 
4220 void
if_freedescr(char * descrbuf)4221 if_freedescr(char *descrbuf)
4222 {
4223 	free(descrbuf, M_IFDESCR);
4224 }
4225 
4226 int
if_getalloctype(const if_t ifp)4227 if_getalloctype(const if_t ifp)
4228 {
4229 	return (ifp->if_alloctype);
4230 }
4231 
4232 void
if_setlastchange(if_t ifp)4233 if_setlastchange(if_t ifp)
4234 {
4235 	getmicrotime(&ifp->if_lastchange);
4236 }
4237 
4238 /*
4239  * This is largely undesirable because it ties ifnet to a device, but does
4240  * provide flexiblity for an embedded product vendor. Should be used with
4241  * the understanding that it violates the interface boundaries, and should be
4242  * a last resort only.
4243  */
4244 int
if_setdev(if_t ifp,void * dev)4245 if_setdev(if_t ifp, void *dev)
4246 {
4247 	return (0);
4248 }
4249 
4250 int
if_setdrvflagbits(if_t ifp,int set_flags,int clear_flags)4251 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4252 {
4253 	ifp->if_drv_flags &= ~clear_flags;
4254 	ifp->if_drv_flags |= set_flags;
4255 
4256 	return (0);
4257 }
4258 
4259 int
if_getdrvflags(const if_t ifp)4260 if_getdrvflags(const if_t ifp)
4261 {
4262 	return (ifp->if_drv_flags);
4263 }
4264 
4265 int
if_setdrvflags(if_t ifp,int flags)4266 if_setdrvflags(if_t ifp, int flags)
4267 {
4268 	ifp->if_drv_flags = flags;
4269 	return (0);
4270 }
4271 
4272 int
if_setflags(if_t ifp,int flags)4273 if_setflags(if_t ifp, int flags)
4274 {
4275 	ifp->if_flags = flags;
4276 	return (0);
4277 }
4278 
4279 int
if_setflagbits(if_t ifp,int set,int clear)4280 if_setflagbits(if_t ifp, int set, int clear)
4281 {
4282 	ifp->if_flags &= ~clear;
4283 	ifp->if_flags |= set;
4284 	return (0);
4285 }
4286 
4287 int
if_getflags(const if_t ifp)4288 if_getflags(const if_t ifp)
4289 {
4290 	return (ifp->if_flags);
4291 }
4292 
4293 int
if_clearhwassist(if_t ifp)4294 if_clearhwassist(if_t ifp)
4295 {
4296 	ifp->if_hwassist = 0;
4297 	return (0);
4298 }
4299 
4300 int
if_sethwassistbits(if_t ifp,int toset,int toclear)4301 if_sethwassistbits(if_t ifp, int toset, int toclear)
4302 {
4303 	ifp->if_hwassist &= ~toclear;
4304 	ifp->if_hwassist |= toset;
4305 
4306 	return (0);
4307 }
4308 
4309 int
if_sethwassist(if_t ifp,int hwassist_bit)4310 if_sethwassist(if_t ifp, int hwassist_bit)
4311 {
4312 	ifp->if_hwassist = hwassist_bit;
4313 	return (0);
4314 }
4315 
4316 int
if_gethwassist(const if_t ifp)4317 if_gethwassist(const if_t ifp)
4318 {
4319 	return (ifp->if_hwassist);
4320 }
4321 
4322 int
if_togglehwassist(if_t ifp,int toggle_bits)4323 if_togglehwassist(if_t ifp, int toggle_bits)
4324 {
4325 	ifp->if_hwassist ^= toggle_bits;
4326 	return (0);
4327 }
4328 
4329 int
if_setmtu(if_t ifp,int mtu)4330 if_setmtu(if_t ifp, int mtu)
4331 {
4332 	ifp->if_mtu = mtu;
4333 	return (0);
4334 }
4335 
4336 void
if_notifymtu(if_t ifp)4337 if_notifymtu(if_t ifp)
4338 {
4339 #ifdef INET6
4340 	nd6_setmtu(ifp);
4341 #endif
4342 	rt_updatemtu(ifp);
4343 }
4344 
4345 int
if_getmtu(const if_t ifp)4346 if_getmtu(const if_t ifp)
4347 {
4348 	return (ifp->if_mtu);
4349 }
4350 
4351 void
if_setppromisc(if_t ifp,bool ppromisc)4352 if_setppromisc(if_t ifp, bool ppromisc)
4353 {
4354 	int new_flags;
4355 
4356 	if (ppromisc)
4357 		new_flags = ifp->if_flags | IFF_PPROMISC;
4358 	else
4359 		new_flags = ifp->if_flags & ~IFF_PPROMISC;
4360 	if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
4361 		if (new_flags & IFF_PPROMISC)
4362 			new_flags |= IFF_PROMISC;
4363 		/*
4364 		 * Only unset IFF_PROMISC if there are no more consumers of
4365 		 * promiscuity, i.e. the ifp->if_pcount refcount is 0.
4366 		 */
4367 		else if (ifp->if_pcount == 0)
4368 			new_flags &= ~IFF_PROMISC;
4369 		if (log_promisc_mode_change)
4370                         if_printf(ifp, "permanently promiscuous mode %s\n",
4371                             ((new_flags & IFF_PPROMISC) ?
4372                              "enabled" : "disabled"));
4373 	}
4374 	ifp->if_flags = new_flags;
4375 }
4376 
4377 /*
4378  * Methods for drivers to access interface unicast and multicast
4379  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4380  * 'struct ifmultiaddr'.
4381  */
4382 u_int
if_lladdr_count(if_t ifp)4383 if_lladdr_count(if_t ifp)
4384 {
4385 	struct epoch_tracker et;
4386 	struct ifaddr *ifa;
4387 	u_int count;
4388 
4389 	count = 0;
4390 	NET_EPOCH_ENTER(et);
4391 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4392 		if (ifa->ifa_addr->sa_family == AF_LINK)
4393 			count++;
4394 	NET_EPOCH_EXIT(et);
4395 
4396 	return (count);
4397 }
4398 
4399 int
if_foreach(if_foreach_cb_t cb,void * cb_arg)4400 if_foreach(if_foreach_cb_t cb, void *cb_arg)
4401 {
4402 	if_t ifp;
4403 	int error;
4404 
4405 	NET_EPOCH_ASSERT();
4406 	MPASS(cb);
4407 
4408 	error = 0;
4409 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4410 		error = cb(ifp, cb_arg);
4411 		if (error != 0)
4412 			break;
4413 	}
4414 
4415 	return (error);
4416 }
4417 
4418 /*
4419  * Iterates over the list of interfaces, permitting callback function @cb to sleep.
4420  * Stops iteration if @cb returns non-zero error code.
4421  * Returns the last error code from @cb.
4422  * @match_cb: optional match callback limiting the iteration to only matched interfaces
4423  * @match_arg: argument to pass to @match_cb
4424  * @cb: iteration callback
4425  * @cb_arg: argument to pass to @cb
4426  */
4427 int
if_foreach_sleep(if_foreach_match_t match_cb,void * match_arg,if_foreach_cb_t cb,void * cb_arg)4428 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb,
4429     void *cb_arg)
4430 {
4431 	int match_count = 0, array_size = 16; /* 128 bytes for malloc */
4432 	struct ifnet **match_array = NULL;
4433 	int error = 0;
4434 
4435 	MPASS(cb);
4436 
4437 	while (true) {
4438 		struct ifnet **new_array;
4439 		int new_size = array_size;
4440 		struct epoch_tracker et;
4441 		struct ifnet *ifp;
4442 
4443 		while (new_size < match_count)
4444 			new_size *= 2;
4445 		new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK);
4446 		if (match_array != NULL)
4447 			memcpy(new_array, match_array, array_size * sizeof(void *));
4448 		free(match_array, M_TEMP);
4449 		match_array = new_array;
4450 		array_size = new_size;
4451 
4452 		match_count = 0;
4453 		NET_EPOCH_ENTER(et);
4454 		CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4455 			if (match_cb != NULL && !match_cb(ifp, match_arg))
4456 				continue;
4457 			if (match_count < array_size) {
4458 				if (if_try_ref(ifp))
4459 					match_array[match_count++] = ifp;
4460 			} else
4461 				match_count++;
4462 		}
4463 		NET_EPOCH_EXIT(et);
4464 
4465 		if (match_count > array_size) {
4466 			for (int i = 0; i < array_size; i++)
4467 				if_rele(match_array[i]);
4468 			continue;
4469 		} else {
4470 			for (int i = 0; i < match_count; i++) {
4471 				if (error == 0)
4472 					error = cb(match_array[i], cb_arg);
4473 				if_rele(match_array[i]);
4474 			}
4475 			free(match_array, M_TEMP);
4476 			break;
4477 		}
4478 	}
4479 
4480 	return (error);
4481 }
4482 
4483 
4484 /*
4485  * Uses just 1 pointer of the 4 available in the public struct.
4486  */
4487 if_t
if_iter_start(struct if_iter * iter)4488 if_iter_start(struct if_iter *iter)
4489 {
4490 	if_t ifp;
4491 
4492 	NET_EPOCH_ASSERT();
4493 
4494 	bzero(iter, sizeof(*iter));
4495 	ifp = CK_STAILQ_FIRST(&V_ifnet);
4496 	if (ifp != NULL)
4497 		iter->context[0] = CK_STAILQ_NEXT(ifp, if_link);
4498 	else
4499 		iter->context[0] = NULL;
4500 	return (ifp);
4501 }
4502 
4503 if_t
if_iter_next(struct if_iter * iter)4504 if_iter_next(struct if_iter *iter)
4505 {
4506 	if_t cur_ifp = iter->context[0];
4507 
4508 	if (cur_ifp != NULL)
4509 		iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link);
4510 	return (cur_ifp);
4511 }
4512 
4513 void
if_iter_finish(struct if_iter * iter)4514 if_iter_finish(struct if_iter *iter)
4515 {
4516 	/* Nothing to do here for now. */
4517 }
4518 
4519 u_int
if_foreach_lladdr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4520 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4521 {
4522 	struct epoch_tracker et;
4523 	struct ifaddr *ifa;
4524 	u_int count;
4525 
4526 	MPASS(cb);
4527 
4528 	count = 0;
4529 	NET_EPOCH_ENTER(et);
4530 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4531 		if (ifa->ifa_addr->sa_family != AF_LINK)
4532 			continue;
4533 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4534 		    count);
4535 	}
4536 	NET_EPOCH_EXIT(et);
4537 
4538 	return (count);
4539 }
4540 
4541 u_int
if_llmaddr_count(if_t ifp)4542 if_llmaddr_count(if_t ifp)
4543 {
4544 	struct epoch_tracker et;
4545 	struct ifmultiaddr *ifma;
4546 	int count;
4547 
4548 	count = 0;
4549 	NET_EPOCH_ENTER(et);
4550 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4551 		if (ifma->ifma_addr->sa_family == AF_LINK)
4552 			count++;
4553 	NET_EPOCH_EXIT(et);
4554 
4555 	return (count);
4556 }
4557 
4558 bool
if_maddr_empty(if_t ifp)4559 if_maddr_empty(if_t ifp)
4560 {
4561 
4562 	return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs));
4563 }
4564 
4565 u_int
if_foreach_llmaddr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4566 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4567 {
4568 	struct epoch_tracker et;
4569 	struct ifmultiaddr *ifma;
4570 	u_int count;
4571 
4572 	MPASS(cb);
4573 
4574 	count = 0;
4575 	NET_EPOCH_ENTER(et);
4576 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4577 		if (ifma->ifma_addr->sa_family != AF_LINK)
4578 			continue;
4579 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4580 		    count);
4581 	}
4582 	NET_EPOCH_EXIT(et);
4583 
4584 	return (count);
4585 }
4586 
4587 u_int
if_foreach_addr_type(if_t ifp,int type,if_addr_cb_t cb,void * cb_arg)4588 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg)
4589 {
4590 	struct epoch_tracker et;
4591 	struct ifaddr *ifa;
4592 	u_int count;
4593 
4594 	MPASS(cb);
4595 
4596 	count = 0;
4597 	NET_EPOCH_ENTER(et);
4598 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4599 		if (ifa->ifa_addr->sa_family != type)
4600 			continue;
4601 		count += (*cb)(cb_arg, ifa, count);
4602 	}
4603 	NET_EPOCH_EXIT(et);
4604 
4605 	return (count);
4606 }
4607 
4608 struct ifaddr *
ifa_iter_start(if_t ifp,struct ifa_iter * iter)4609 ifa_iter_start(if_t ifp, struct ifa_iter *iter)
4610 {
4611 	struct ifaddr *ifa;
4612 
4613 	NET_EPOCH_ASSERT();
4614 
4615 	bzero(iter, sizeof(*iter));
4616 	ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
4617 	if (ifa != NULL)
4618 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4619 	else
4620 		iter->context[0] = NULL;
4621 	return (ifa);
4622 }
4623 
4624 struct ifaddr *
ifa_iter_next(struct ifa_iter * iter)4625 ifa_iter_next(struct ifa_iter *iter)
4626 {
4627 	struct ifaddr *ifa = iter->context[0];
4628 
4629 	if (ifa != NULL)
4630 		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4631 	return (ifa);
4632 }
4633 
4634 void
ifa_iter_finish(struct ifa_iter * iter)4635 ifa_iter_finish(struct ifa_iter *iter)
4636 {
4637 	/* Nothing to do here for now. */
4638 }
4639 
4640 int
if_setsoftc(if_t ifp,void * softc)4641 if_setsoftc(if_t ifp, void *softc)
4642 {
4643 	ifp->if_softc = softc;
4644 	return (0);
4645 }
4646 
4647 void *
if_getsoftc(const if_t ifp)4648 if_getsoftc(const if_t ifp)
4649 {
4650 	return (ifp->if_softc);
4651 }
4652 
4653 void
if_setrcvif(struct mbuf * m,if_t ifp)4654 if_setrcvif(struct mbuf *m, if_t ifp)
4655 {
4656 
4657 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4658 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4659 }
4660 
4661 void
if_setvtag(struct mbuf * m,uint16_t tag)4662 if_setvtag(struct mbuf *m, uint16_t tag)
4663 {
4664 	m->m_pkthdr.ether_vtag = tag;
4665 }
4666 
4667 uint16_t
if_getvtag(struct mbuf * m)4668 if_getvtag(struct mbuf *m)
4669 {
4670 	return (m->m_pkthdr.ether_vtag);
4671 }
4672 
4673 int
if_sendq_empty(if_t ifp)4674 if_sendq_empty(if_t ifp)
4675 {
4676 	return (IFQ_DRV_IS_EMPTY(&ifp->if_snd));
4677 }
4678 
4679 struct ifaddr *
if_getifaddr(const if_t ifp)4680 if_getifaddr(const if_t ifp)
4681 {
4682 	return (ifp->if_addr);
4683 }
4684 
4685 int
if_setsendqready(if_t ifp)4686 if_setsendqready(if_t ifp)
4687 {
4688 	IFQ_SET_READY(&ifp->if_snd);
4689 	return (0);
4690 }
4691 
4692 int
if_setsendqlen(if_t ifp,int tx_desc_count)4693 if_setsendqlen(if_t ifp, int tx_desc_count)
4694 {
4695 	IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count);
4696 	ifp->if_snd.ifq_drv_maxlen = tx_desc_count;
4697 	return (0);
4698 }
4699 
4700 void
if_setnetmapadapter(if_t ifp,struct netmap_adapter * na)4701 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na)
4702 {
4703 	ifp->if_netmap = na;
4704 }
4705 
4706 struct netmap_adapter *
if_getnetmapadapter(if_t ifp)4707 if_getnetmapadapter(if_t ifp)
4708 {
4709 	return (ifp->if_netmap);
4710 }
4711 
4712 int
if_vlantrunkinuse(if_t ifp)4713 if_vlantrunkinuse(if_t ifp)
4714 {
4715 	return (ifp->if_vlantrunk != NULL);
4716 }
4717 
4718 void
if_init(if_t ifp,void * ctx)4719 if_init(if_t ifp, void *ctx)
4720 {
4721 	(*ifp->if_init)(ctx);
4722 }
4723 
4724 void
if_input(if_t ifp,struct mbuf * sendmp)4725 if_input(if_t ifp, struct mbuf* sendmp)
4726 {
4727 	(*ifp->if_input)(ifp, sendmp);
4728 }
4729 
4730 int
if_transmit(if_t ifp,struct mbuf * m)4731 if_transmit(if_t ifp, struct mbuf *m)
4732 {
4733 	return ((*ifp->if_transmit)(ifp, m));
4734 }
4735 
4736 int
if_resolvemulti(if_t ifp,struct sockaddr ** srcs,struct sockaddr * dst)4737 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst)
4738 {
4739 	if (ifp->if_resolvemulti == NULL)
4740 		return (EOPNOTSUPP);
4741 
4742 	return (ifp->if_resolvemulti(ifp, srcs, dst));
4743 }
4744 
4745 int
if_ioctl(if_t ifp,u_long cmd,void * data)4746 if_ioctl(if_t ifp, u_long cmd, void *data)
4747 {
4748 	if (ifp->if_ioctl == NULL)
4749 		return (EOPNOTSUPP);
4750 
4751 	return (ifp->if_ioctl(ifp, cmd, data));
4752 }
4753 
4754 struct mbuf *
if_dequeue(if_t ifp)4755 if_dequeue(if_t ifp)
4756 {
4757 	struct mbuf *m;
4758 
4759 	IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
4760 	return (m);
4761 }
4762 
4763 int
if_sendq_prepend(if_t ifp,struct mbuf * m)4764 if_sendq_prepend(if_t ifp, struct mbuf *m)
4765 {
4766 	IFQ_DRV_PREPEND(&ifp->if_snd, m);
4767 	return (0);
4768 }
4769 
4770 int
if_setifheaderlen(if_t ifp,int len)4771 if_setifheaderlen(if_t ifp, int len)
4772 {
4773 	ifp->if_hdrlen = len;
4774 	return (0);
4775 }
4776 
4777 char *
if_getlladdr(const if_t ifp)4778 if_getlladdr(const if_t ifp)
4779 {
4780 	return (IF_LLADDR(ifp));
4781 }
4782 
4783 void *
if_gethandle(u_char type)4784 if_gethandle(u_char type)
4785 {
4786 	return (if_alloc(type));
4787 }
4788 
4789 void
if_vlancap(if_t ifp)4790 if_vlancap(if_t ifp)
4791 {
4792 	VLAN_CAPABILITIES(ifp);
4793 }
4794 
4795 int
if_sethwtsomax(if_t ifp,u_int if_hw_tsomax)4796 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4797 {
4798 	ifp->if_hw_tsomax = if_hw_tsomax;
4799         return (0);
4800 }
4801 
4802 int
if_sethwtsomaxsegcount(if_t ifp,u_int if_hw_tsomaxsegcount)4803 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4804 {
4805 	ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4806         return (0);
4807 }
4808 
4809 int
if_sethwtsomaxsegsize(if_t ifp,u_int if_hw_tsomaxsegsize)4810 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4811 {
4812 	ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4813         return (0);
4814 }
4815 
4816 u_int
if_gethwtsomax(const if_t ifp)4817 if_gethwtsomax(const if_t ifp)
4818 {
4819 	return (ifp->if_hw_tsomax);
4820 }
4821 
4822 u_int
if_gethwtsomaxsegcount(const if_t ifp)4823 if_gethwtsomaxsegcount(const if_t ifp)
4824 {
4825 	return (ifp->if_hw_tsomaxsegcount);
4826 }
4827 
4828 u_int
if_gethwtsomaxsegsize(const if_t ifp)4829 if_gethwtsomaxsegsize(const if_t ifp)
4830 {
4831 	return (ifp->if_hw_tsomaxsegsize);
4832 }
4833 
4834 void
if_setinitfn(if_t ifp,if_init_fn_t init_fn)4835 if_setinitfn(if_t ifp, if_init_fn_t init_fn)
4836 {
4837 	ifp->if_init = init_fn;
4838 }
4839 
4840 void
if_setinputfn(if_t ifp,if_input_fn_t input_fn)4841 if_setinputfn(if_t ifp, if_input_fn_t input_fn)
4842 {
4843 	ifp->if_input = input_fn;
4844 }
4845 
4846 if_input_fn_t
if_getinputfn(if_t ifp)4847 if_getinputfn(if_t ifp)
4848 {
4849 	return (ifp->if_input);
4850 }
4851 
4852 void
if_setioctlfn(if_t ifp,if_ioctl_fn_t ioctl_fn)4853 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn)
4854 {
4855 	ifp->if_ioctl = ioctl_fn;
4856 }
4857 
4858 void
if_setoutputfn(if_t ifp,if_output_fn_t output_fn)4859 if_setoutputfn(if_t ifp, if_output_fn_t output_fn)
4860 {
4861 	ifp->if_output = output_fn;
4862 }
4863 
4864 void
if_setstartfn(if_t ifp,if_start_fn_t start_fn)4865 if_setstartfn(if_t ifp, if_start_fn_t start_fn)
4866 {
4867 	ifp->if_start = start_fn;
4868 }
4869 
4870 if_start_fn_t
if_getstartfn(if_t ifp)4871 if_getstartfn(if_t ifp)
4872 {
4873 	return (ifp->if_start);
4874 }
4875 
4876 void
if_settransmitfn(if_t ifp,if_transmit_fn_t start_fn)4877 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4878 {
4879 	ifp->if_transmit = start_fn;
4880 }
4881 
4882 if_transmit_fn_t
if_gettransmitfn(if_t ifp)4883 if_gettransmitfn(if_t ifp)
4884 {
4885 	return (ifp->if_transmit);
4886 }
4887 
4888 void
if_setqflushfn(if_t ifp,if_qflush_fn_t flush_fn)4889 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4890 {
4891 	ifp->if_qflush = flush_fn;
4892 }
4893 
4894 void
if_setsndtagallocfn(if_t ifp,if_snd_tag_alloc_t alloc_fn)4895 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn)
4896 {
4897 	ifp->if_snd_tag_alloc = alloc_fn;
4898 }
4899 
4900 int
if_snd_tag_alloc(if_t ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** mstp)4901 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
4902     struct m_snd_tag **mstp)
4903 {
4904 	if (ifp->if_snd_tag_alloc == NULL)
4905 		return (EOPNOTSUPP);
4906 	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
4907 }
4908 
4909 void
if_setgetcounterfn(if_t ifp,if_get_counter_t fn)4910 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4911 {
4912 	ifp->if_get_counter = fn;
4913 }
4914 
4915 void
if_setreassignfn(if_t ifp,if_reassign_fn_t fn)4916 if_setreassignfn(if_t ifp, if_reassign_fn_t fn)
4917 {
4918 	ifp->if_reassign = fn;
4919 }
4920 
4921 void
if_setratelimitqueryfn(if_t ifp,if_ratelimit_query_t fn)4922 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn)
4923 {
4924 	ifp->if_ratelimit_query = fn;
4925 }
4926 
4927 void
if_setdebugnet_methods(if_t ifp,struct debugnet_methods * m)4928 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m)
4929 {
4930 	ifp->if_debugnet_methods = m;
4931 }
4932 
4933 struct label *
if_getmaclabel(if_t ifp)4934 if_getmaclabel(if_t ifp)
4935 {
4936 	return (ifp->if_label);
4937 }
4938 
4939 void
if_setmaclabel(if_t ifp,struct label * label)4940 if_setmaclabel(if_t ifp, struct label *label)
4941 {
4942 	ifp->if_label = label;
4943 }
4944 
4945 int
if_gettype(if_t ifp)4946 if_gettype(if_t ifp)
4947 {
4948 	return (ifp->if_type);
4949 }
4950 
4951 void *
if_getllsoftc(if_t ifp)4952 if_getllsoftc(if_t ifp)
4953 {
4954 	return (ifp->if_llsoftc);
4955 }
4956 
4957 void
if_setllsoftc(if_t ifp,void * llsoftc)4958 if_setllsoftc(if_t ifp, void *llsoftc)
4959 {
4960 	ifp->if_llsoftc = llsoftc;
4961 };
4962 
4963 int
if_getlinkstate(if_t ifp)4964 if_getlinkstate(if_t ifp)
4965 {
4966 	return (ifp->if_link_state);
4967 }
4968 
4969 const uint8_t *
if_getbroadcastaddr(if_t ifp)4970 if_getbroadcastaddr(if_t ifp)
4971 {
4972 	return (ifp->if_broadcastaddr);
4973 }
4974 
4975 void
if_setbroadcastaddr(if_t ifp,const uint8_t * addr)4976 if_setbroadcastaddr(if_t ifp, const uint8_t *addr)
4977 {
4978 	ifp->if_broadcastaddr = addr;
4979 }
4980 
4981 int
if_getnumadomain(if_t ifp)4982 if_getnumadomain(if_t ifp)
4983 {
4984 	return (ifp->if_numa_domain);
4985 }
4986 
4987 uint64_t
if_getcounter(if_t ifp,ift_counter counter)4988 if_getcounter(if_t ifp, ift_counter counter)
4989 {
4990 	return (ifp->if_get_counter(ifp, counter));
4991 }
4992 
4993 bool
if_altq_is_enabled(if_t ifp)4994 if_altq_is_enabled(if_t ifp)
4995 {
4996 	return (ALTQ_IS_ENABLED(&ifp->if_snd));
4997 }
4998 
4999 struct vnet *
if_getvnet(if_t ifp)5000 if_getvnet(if_t ifp)
5001 {
5002 	return (ifp->if_vnet);
5003 }
5004 
5005 struct in_ifinfo *
if_getinet(if_t ifp)5006 if_getinet(if_t ifp)
5007 {
5008 	return (ifp->if_inet);
5009 }
5010 
5011 struct in6_ifextra *
if_getinet6(if_t ifp)5012 if_getinet6(if_t ifp)
5013 {
5014 	return (ifp->if_inet6);
5015 }
5016 
5017 u_int
if_getfib(if_t ifp)5018 if_getfib(if_t ifp)
5019 {
5020 	return (ifp->if_fib);
5021 }
5022 
5023 uint8_t
if_getaddrlen(if_t ifp)5024 if_getaddrlen(if_t ifp)
5025 {
5026 	return (ifp->if_addrlen);
5027 }
5028 
5029 struct bpf_if *
if_getbpf(if_t ifp)5030 if_getbpf(if_t ifp)
5031 {
5032 	return (ifp->if_bpf);
5033 }
5034 
5035 struct ifvlantrunk *
if_getvlantrunk(if_t ifp)5036 if_getvlantrunk(if_t ifp)
5037 {
5038 	return (ifp->if_vlantrunk);
5039 }
5040 
5041 uint8_t
if_getpcp(if_t ifp)5042 if_getpcp(if_t ifp)
5043 {
5044 	return (ifp->if_pcp);
5045 }
5046 
5047 void *
if_getl2com(if_t ifp)5048 if_getl2com(if_t ifp)
5049 {
5050 	return (ifp->if_l2com);
5051 }
5052 
5053 void
if_setipsec_accel_methods(if_t ifp,const struct if_ipsec_accel_methods * m)5054 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m)
5055 {
5056 	ifp->if_ipsec_accel_m = m;
5057 }
5058 
5059 #ifdef DDB
5060 static void
if_show_ifnet(struct ifnet * ifp)5061 if_show_ifnet(struct ifnet *ifp)
5062 {
5063 	if (ifp == NULL)
5064 		return;
5065 	db_printf("%s:\n", ifp->if_xname);
5066 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
5067 	IF_DB_PRINTF("%s", if_dname);
5068 	IF_DB_PRINTF("%d", if_dunit);
5069 	IF_DB_PRINTF("%s", if_description);
5070 	IF_DB_PRINTF("%u", if_index);
5071 	IF_DB_PRINTF("%d", if_idxgen);
5072 	IF_DB_PRINTF("%u", if_refcount);
5073 	IF_DB_PRINTF("%p", if_softc);
5074 	IF_DB_PRINTF("%p", if_l2com);
5075 	IF_DB_PRINTF("%p", if_llsoftc);
5076 	IF_DB_PRINTF("%d", if_amcount);
5077 	IF_DB_PRINTF("%p", if_addr);
5078 	IF_DB_PRINTF("%p", if_broadcastaddr);
5079 	IF_DB_PRINTF("%u", if_fib);
5080 	IF_DB_PRINTF("%p", if_vnet);
5081 	IF_DB_PRINTF("%p", if_home_vnet);
5082 	IF_DB_PRINTF("%p", if_vlantrunk);
5083 	IF_DB_PRINTF("%p", if_bpf);
5084 	IF_DB_PRINTF("%u", if_pcount);
5085 	IF_DB_PRINTF("%p", if_bridge);
5086 	IF_DB_PRINTF("%p", if_lagg);
5087 	IF_DB_PRINTF("%p", if_pf_kif);
5088 	IF_DB_PRINTF("%p", if_carp);
5089 	IF_DB_PRINTF("%p", if_label);
5090 	IF_DB_PRINTF("%p", if_netmap);
5091 	IF_DB_PRINTF("0x%08x", if_flags);
5092 	IF_DB_PRINTF("0x%08x", if_drv_flags);
5093 	IF_DB_PRINTF("0x%08x", if_capabilities);
5094 	IF_DB_PRINTF("0x%08x", if_capenable);
5095 	IF_DB_PRINTF("%p", if_snd.ifq_head);
5096 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
5097 	IF_DB_PRINTF("%d", if_snd.ifq_len);
5098 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
5099 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
5100 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
5101 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
5102 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
5103 	IF_DB_PRINTF("%d", if_snd.altq_type);
5104 	IF_DB_PRINTF("%x", if_snd.altq_flags);
5105 #undef IF_DB_PRINTF
5106 }
5107 
DB_SHOW_COMMAND(ifnet,db_show_ifnet)5108 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
5109 {
5110 	if (!have_addr) {
5111 		db_printf("usage: show ifnet <struct ifnet *>\n");
5112 		return;
5113 	}
5114 
5115 	if_show_ifnet((struct ifnet *)addr);
5116 }
5117 
DB_SHOW_ALL_COMMAND(ifnets,db_show_all_ifnets)5118 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
5119 {
5120 	struct ifnet *ifp;
5121 	u_short idx;
5122 
5123 	for (idx = 1; idx <= if_index; idx++) {
5124 		ifp = ifindex_table[idx].ife_ifnet;
5125 		if (ifp == NULL)
5126 			continue;
5127 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
5128 		if (db_pager_quit)
5129 			break;
5130 	}
5131 }
5132 #endif	/* DDB */
5133