xref: /freebsd/sys/net/if_vlan.c (revision 07b16b1e2aeab0b30f68a013de31a4c322a61246)
1 /*-
2  * Copyright 1998 Massachusetts Institute of Technology
3  * Copyright 2012 ADARA Networks, Inc.
4  * Copyright 2017 Dell EMC Isilon
5  *
6  * Portions of this software were developed by Robert N. M. Watson under
7  * contract to ADARA Networks, Inc.
8  *
9  * Permission to use, copy, modify, and distribute this software and
10  * its documentation for any purpose and without fee is hereby
11  * granted, provided that both the above copyright notice and this
12  * permission notice appear in all copies, that both the above
13  * copyright notice and this permission notice appear in all
14  * supporting documentation, and that the name of M.I.T. not be used
15  * in advertising or publicity pertaining to distribution of the
16  * software without specific, written prior permission.  M.I.T. makes
17  * no representations about the suitability of this software for any
18  * purpose.  It is provided "as is" without express or implied
19  * warranty.
20  *
21  * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
22  * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
23  * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
24  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
25  * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
28  * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
29  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * if_vlan.c - pseudo-device driver for IEEE 802.1Q virtual LANs.
37  * This is sort of sneaky in the implementation, since
38  * we need to pretend to be enough of an Ethernet implementation
39  * to make arp work.  The way we do this is by telling everyone
40  * that we are an Ethernet, and then catch the packets that
41  * ether_output() sends to us via if_transmit(), rewrite them for
42  * use by the real outgoing interface, and ask it to send them.
43  */
44 
45 #include "opt_inet.h"
46 #include "opt_inet6.h"
47 #include "opt_ipsec.h"
48 #include "opt_kern_tls.h"
49 #include "opt_vlan.h"
50 #include "opt_ratelimit.h"
51 
52 #include <sys/param.h>
53 #include <sys/eventhandler.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/malloc.h>
57 #include <sys/mbuf.h>
58 #include <sys/module.h>
59 #include <sys/rmlock.h>
60 #include <sys/priv.h>
61 #include <sys/queue.h>
62 #include <sys/socket.h>
63 #include <sys/sockio.h>
64 #include <sys/sysctl.h>
65 #include <sys/systm.h>
66 #include <sys/sx.h>
67 #include <sys/taskqueue.h>
68 
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/if_clone.h>
75 #include <net/if_dl.h>
76 #include <net/if_types.h>
77 #include <net/if_vlan_var.h>
78 #include <net/route.h>
79 #include <net/vnet.h>
80 
81 #ifdef INET
82 #include <netinet/in.h>
83 #include <netinet/if_ether.h>
84 #endif
85 
86 #include <netlink/netlink.h>
87 #include <netlink/netlink_ctl.h>
88 #include <netlink/netlink_route.h>
89 #include <netlink/route/route_var.h>
90 
91 #define	VLAN_DEF_HWIDTH	4
92 #define	VLAN_IFFLAGS	(IFF_BROADCAST | IFF_MULTICAST)
93 
94 #define	UP_AND_RUNNING(ifp) \
95     ((ifp)->if_flags & IFF_UP && (ifp)->if_drv_flags & IFF_DRV_RUNNING)
96 
97 CK_SLIST_HEAD(ifvlanhead, ifvlan);
98 
99 struct ifvlantrunk {
100 	struct	ifnet   *parent;	/* parent interface of this trunk */
101 	struct	mtx	lock;
102 #ifdef VLAN_ARRAY
103 #define	VLAN_ARRAY_SIZE	(EVL_VLID_MASK + 1)
104 	struct	ifvlan	*vlans[VLAN_ARRAY_SIZE]; /* static table */
105 #else
106 	struct	ifvlanhead *hash;	/* dynamic hash-list table */
107 	uint16_t	hmask;
108 	uint16_t	hwidth;
109 #endif
110 	int		refcnt;
111 };
112 
113 #if defined(KERN_TLS) || defined(RATELIMIT)
114 struct vlan_snd_tag {
115 	struct m_snd_tag com;
116 	struct m_snd_tag *tag;
117 };
118 
119 static inline struct vlan_snd_tag *
mst_to_vst(struct m_snd_tag * mst)120 mst_to_vst(struct m_snd_tag *mst)
121 {
122 
123 	return (__containerof(mst, struct vlan_snd_tag, com));
124 }
125 #endif
126 
127 /*
128  * This macro provides a facility to iterate over every vlan on a trunk with
129  * the assumption that none will be added/removed during iteration.
130  */
131 #ifdef VLAN_ARRAY
132 #define VLAN_FOREACH(_ifv, _trunk) \
133 	size_t _i; \
134 	for (_i = 0; _i < VLAN_ARRAY_SIZE; _i++) \
135 		if (((_ifv) = (_trunk)->vlans[_i]) != NULL)
136 #else /* VLAN_ARRAY */
137 #define VLAN_FOREACH(_ifv, _trunk) \
138 	struct ifvlan *_next; \
139 	size_t _i; \
140 	for (_i = 0; _i < (1 << (_trunk)->hwidth); _i++) \
141 		CK_SLIST_FOREACH_SAFE((_ifv), &(_trunk)->hash[_i], ifv_list, _next)
142 #endif /* VLAN_ARRAY */
143 
144 /*
145  * This macro provides a facility to iterate over every vlan on a trunk while
146  * also modifying the number of vlans on the trunk. The iteration continues
147  * until some condition is met or there are no more vlans on the trunk.
148  */
149 #ifdef VLAN_ARRAY
150 /* The VLAN_ARRAY case is simple -- just a for loop using the condition. */
151 #define VLAN_FOREACH_UNTIL_SAFE(_ifv, _trunk, _cond) \
152 	size_t _i; \
153 	for (_i = 0; !(_cond) && _i < VLAN_ARRAY_SIZE; _i++) \
154 		if (((_ifv) = (_trunk)->vlans[_i]))
155 #else /* VLAN_ARRAY */
156 /*
157  * The hash table case is more complicated. We allow for the hash table to be
158  * modified (i.e. vlans removed) while we are iterating over it. To allow for
159  * this we must restart the iteration every time we "touch" something during
160  * the iteration, since removal will resize the hash table and invalidate our
161  * current position. If acting on the touched element causes the trunk to be
162  * emptied, then iteration also stops.
163  */
164 #define VLAN_FOREACH_UNTIL_SAFE(_ifv, _trunk, _cond) \
165 	size_t _i; \
166 	bool _touch = false; \
167 	for (_i = 0; \
168 	    !(_cond) && _i < (1 << (_trunk)->hwidth); \
169 	    _i = (_touch && ((_trunk) != NULL) ? 0 : _i + 1), _touch = false) \
170 		if (((_ifv) = CK_SLIST_FIRST(&(_trunk)->hash[_i])) != NULL && \
171 		    (_touch = true))
172 #endif /* VLAN_ARRAY */
173 
174 struct vlan_mc_entry {
175 	struct sockaddr_dl		mc_addr;
176 	CK_SLIST_ENTRY(vlan_mc_entry)	mc_entries;
177 	struct epoch_context		mc_epoch_ctx;
178 };
179 
180 struct ifvlan {
181 	struct	ifvlantrunk *ifv_trunk;
182 	struct	ifnet *ifv_ifp;
183 #define	TRUNK(ifv)	((ifv)->ifv_trunk)
184 #define	PARENT(ifv)	(TRUNK(ifv)->parent)
185 	void	*ifv_cookie;
186 	int	ifv_pflags;	/* special flags we have set on parent */
187 	int	ifv_capenable;
188   	int	ifv_capenable2;
189 	int	ifv_encaplen;	/* encapsulation length */
190 	int	ifv_mtufudge;	/* MTU fudged by this much */
191 	int	ifv_mintu;	/* min transmission unit */
192 	struct  ether_8021q_tag ifv_qtag;
193 #define ifv_proto	ifv_qtag.proto
194 #define ifv_vid		ifv_qtag.vid
195 #define ifv_pcp		ifv_qtag.pcp
196 	struct task lladdr_task;
197 	CK_SLIST_HEAD(, vlan_mc_entry) vlan_mc_listhead;
198 #ifndef VLAN_ARRAY
199 	CK_SLIST_ENTRY(ifvlan) ifv_list;
200 #endif
201 };
202 
203 /* Special flags we should propagate to parent. */
204 static struct {
205 	int flag;
206 	int (*func)(struct ifnet *, int);
207 } vlan_pflags[] = {
208 	{IFF_PROMISC, ifpromisc},
209 	{IFF_ALLMULTI, if_allmulti},
210 	{0, NULL}
211 };
212 
213 VNET_DECLARE(int, vlan_mtag_pcp);
214 #define	V_vlan_mtag_pcp	VNET(vlan_mtag_pcp)
215 
216 static const char vlanname[] = "vlan";
217 static MALLOC_DEFINE(M_VLAN, vlanname, "802.1Q Virtual LAN Interface");
218 
219 static eventhandler_tag ifdetach_tag;
220 static eventhandler_tag iflladdr_tag;
221 static eventhandler_tag ifevent_tag;
222 
223 /*
224  * if_vlan uses two module-level synchronizations primitives to allow concurrent
225  * modification of vlan interfaces and (mostly) allow for vlans to be destroyed
226  * while they are being used for tx/rx. To accomplish this in a way that has
227  * acceptable performance and cooperation with other parts of the network stack
228  * there is a non-sleepable epoch(9) and an sx(9).
229  *
230  * The performance-sensitive paths that warrant using the epoch(9) are
231  * vlan_transmit and vlan_input. Both have to check for the vlan interface's
232  * existence using if_vlantrunk, and being in the network tx/rx paths the use
233  * of an epoch(9) gives a measureable improvement in performance.
234  *
235  * The reason for having an sx(9) is mostly because there are still areas that
236  * must be sleepable and also have safe concurrent access to a vlan interface.
237  * Since the sx(9) exists, it is used by default in most paths unless sleeping
238  * is not permitted, or if it is not clear whether sleeping is permitted.
239  *
240  */
241 #define _VLAN_SX_ID ifv_sx
242 
243 static struct sx _VLAN_SX_ID;
244 
245 #define VLAN_LOCKING_INIT() \
246 	sx_init_flags(&_VLAN_SX_ID, "vlan_sx", SX_RECURSE)
247 
248 #define VLAN_LOCKING_DESTROY() \
249 	sx_destroy(&_VLAN_SX_ID)
250 
251 #define	VLAN_SLOCK()			sx_slock(&_VLAN_SX_ID)
252 #define	VLAN_SUNLOCK()			sx_sunlock(&_VLAN_SX_ID)
253 #define	VLAN_XLOCK()			sx_xlock(&_VLAN_SX_ID)
254 #define	VLAN_XUNLOCK()			sx_xunlock(&_VLAN_SX_ID)
255 #define	VLAN_SLOCK_ASSERT()		sx_assert(&_VLAN_SX_ID, SA_SLOCKED)
256 #define	VLAN_XLOCK_ASSERT()		sx_assert(&_VLAN_SX_ID, SA_XLOCKED)
257 #define	VLAN_SXLOCK_ASSERT()		sx_assert(&_VLAN_SX_ID, SA_LOCKED)
258 
259 /*
260  * We also have a per-trunk mutex that should be acquired when changing
261  * its state.
262  */
263 #define	TRUNK_LOCK_INIT(trunk)		mtx_init(&(trunk)->lock, vlanname, NULL, MTX_DEF)
264 #define	TRUNK_LOCK_DESTROY(trunk)	mtx_destroy(&(trunk)->lock)
265 #define	TRUNK_WLOCK(trunk)		mtx_lock(&(trunk)->lock)
266 #define	TRUNK_WUNLOCK(trunk)		mtx_unlock(&(trunk)->lock)
267 #define	TRUNK_WLOCK_ASSERT(trunk)	mtx_assert(&(trunk)->lock, MA_OWNED);
268 
269 /*
270  * The VLAN_ARRAY substitutes the dynamic hash with a static array
271  * with 4096 entries. In theory this can give a boost in processing,
272  * however in practice it does not. Probably this is because the array
273  * is too big to fit into CPU cache.
274  */
275 #ifndef VLAN_ARRAY
276 static	void vlan_inithash(struct ifvlantrunk *trunk);
277 static	void vlan_freehash(struct ifvlantrunk *trunk);
278 static	int vlan_inshash(struct ifvlantrunk *trunk, struct ifvlan *ifv);
279 static	int vlan_remhash(struct ifvlantrunk *trunk, struct ifvlan *ifv);
280 static	void vlan_growhash(struct ifvlantrunk *trunk, int howmuch);
281 static __inline struct ifvlan * vlan_gethash(struct ifvlantrunk *trunk,
282 	uint16_t vid);
283 #endif
284 static	void trunk_destroy(struct ifvlantrunk *trunk);
285 
286 static	void vlan_init(void *foo);
287 static	void vlan_input(struct ifnet *ifp, struct mbuf *m);
288 static	int vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr);
289 #if defined(KERN_TLS) || defined(RATELIMIT)
290 static	int vlan_snd_tag_alloc(struct ifnet *,
291     union if_snd_tag_alloc_params *, struct m_snd_tag **);
292 static	int vlan_snd_tag_modify(struct m_snd_tag *,
293     union if_snd_tag_modify_params *);
294 static	int vlan_snd_tag_query(struct m_snd_tag *,
295     union if_snd_tag_query_params *);
296 static	void vlan_snd_tag_free(struct m_snd_tag *);
297 static struct m_snd_tag *vlan_next_snd_tag(struct m_snd_tag *);
298 static void vlan_ratelimit_query(struct ifnet *,
299     struct if_ratelimit_query_results *);
300 #endif
301 static	void vlan_qflush(struct ifnet *ifp);
302 static	int vlan_setflag(struct ifnet *ifp, int flag, int status,
303     int (*func)(struct ifnet *, int));
304 static	int vlan_setflags(struct ifnet *ifp, int status);
305 static	int vlan_setmulti(struct ifnet *ifp);
306 static	int vlan_transmit(struct ifnet *ifp, struct mbuf *m);
307 #ifdef ALTQ
308 static void vlan_altq_start(struct ifnet *ifp);
309 static	int vlan_altq_transmit(struct ifnet *ifp, struct mbuf *m);
310 #endif
311 static	int vlan_output(struct ifnet *ifp, struct mbuf *m,
312     const struct sockaddr *dst, struct route *ro);
313 static	void vlan_unconfig(struct ifnet *ifp);
314 static	void vlan_unconfig_locked(struct ifnet *ifp, int departing);
315 static	int vlan_config(struct ifvlan *ifv, struct ifnet *p, uint16_t tag,
316 	uint16_t proto);
317 static	void vlan_link_state(struct ifnet *ifp);
318 static	void vlan_capabilities(struct ifvlan *ifv);
319 static	void vlan_trunk_capabilities(struct ifnet *ifp);
320 
321 static	struct ifnet *vlan_clone_match_ethervid(const char *, int *);
322 static	int vlan_clone_match(struct if_clone *, const char *);
323 static	int vlan_clone_create(struct if_clone *, char *, size_t,
324     struct ifc_data *, struct ifnet **);
325 static	int vlan_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
326 
327 static int vlan_clone_create_nl(struct if_clone *ifc, char *name, size_t len,
328     struct ifc_data_nl *ifd);
329 static int vlan_clone_modify_nl(struct ifnet *ifp, struct ifc_data_nl *ifd);
330 static void vlan_clone_dump_nl(struct ifnet *ifp, struct nl_writer *nw);
331 
332 static	void vlan_ifdetach(void *arg, struct ifnet *ifp);
333 static  void vlan_iflladdr(void *arg, struct ifnet *ifp);
334 static  void vlan_ifevent(void *arg, struct ifnet *ifp, int event);
335 
336 static  void vlan_lladdr_fn(void *arg, int pending);
337 
338 static struct if_clone *vlan_cloner;
339 
340 #ifdef VIMAGE
341 VNET_DEFINE_STATIC(struct if_clone *, vlan_cloner);
342 #define	V_vlan_cloner	VNET(vlan_cloner)
343 #endif
344 
345 #ifdef RATELIMIT
346 static const struct if_snd_tag_sw vlan_snd_tag_ul_sw = {
347 	.snd_tag_modify = vlan_snd_tag_modify,
348 	.snd_tag_query = vlan_snd_tag_query,
349 	.snd_tag_free = vlan_snd_tag_free,
350 	.next_snd_tag = vlan_next_snd_tag,
351 	.type = IF_SND_TAG_TYPE_UNLIMITED
352 };
353 
354 static const struct if_snd_tag_sw vlan_snd_tag_rl_sw = {
355 	.snd_tag_modify = vlan_snd_tag_modify,
356 	.snd_tag_query = vlan_snd_tag_query,
357 	.snd_tag_free = vlan_snd_tag_free,
358 	.next_snd_tag = vlan_next_snd_tag,
359 	.type = IF_SND_TAG_TYPE_RATE_LIMIT
360 };
361 #endif
362 
363 #ifdef KERN_TLS
364 static const struct if_snd_tag_sw vlan_snd_tag_tls_sw = {
365 	.snd_tag_modify = vlan_snd_tag_modify,
366 	.snd_tag_query = vlan_snd_tag_query,
367 	.snd_tag_free = vlan_snd_tag_free,
368 	.next_snd_tag = vlan_next_snd_tag,
369 	.type = IF_SND_TAG_TYPE_TLS
370 };
371 
372 #ifdef RATELIMIT
373 static const struct if_snd_tag_sw vlan_snd_tag_tls_rl_sw = {
374 	.snd_tag_modify = vlan_snd_tag_modify,
375 	.snd_tag_query = vlan_snd_tag_query,
376 	.snd_tag_free = vlan_snd_tag_free,
377 	.next_snd_tag = vlan_next_snd_tag,
378 	.type = IF_SND_TAG_TYPE_TLS_RATE_LIMIT
379 };
380 #endif
381 #endif
382 
383 static void
vlan_mc_free(struct epoch_context * ctx)384 vlan_mc_free(struct epoch_context *ctx)
385 {
386 	struct vlan_mc_entry *mc = __containerof(ctx, struct vlan_mc_entry, mc_epoch_ctx);
387 	free(mc, M_VLAN);
388 }
389 
390 #ifndef VLAN_ARRAY
391 #define HASH(n, m)	((((n) >> 8) ^ ((n) >> 4) ^ (n)) & (m))
392 
393 static void
vlan_inithash(struct ifvlantrunk * trunk)394 vlan_inithash(struct ifvlantrunk *trunk)
395 {
396 	int i, n;
397 
398 	/*
399 	 * The trunk must not be locked here since we call malloc(M_WAITOK).
400 	 * It is OK in case this function is called before the trunk struct
401 	 * gets hooked up and becomes visible from other threads.
402 	 */
403 
404 	KASSERT(trunk->hwidth == 0 && trunk->hash == NULL,
405 	    ("%s: hash already initialized", __func__));
406 
407 	trunk->hwidth = VLAN_DEF_HWIDTH;
408 	n = 1 << trunk->hwidth;
409 	trunk->hmask = n - 1;
410 	trunk->hash = malloc(sizeof(struct ifvlanhead) * n, M_VLAN, M_WAITOK);
411 	for (i = 0; i < n; i++)
412 		CK_SLIST_INIT(&trunk->hash[i]);
413 }
414 
415 static void
vlan_freehash(struct ifvlantrunk * trunk)416 vlan_freehash(struct ifvlantrunk *trunk)
417 {
418 #ifdef INVARIANTS
419 	int i;
420 
421 	KASSERT(trunk->hwidth > 0, ("%s: hwidth not positive", __func__));
422 	for (i = 0; i < (1 << trunk->hwidth); i++)
423 		KASSERT(CK_SLIST_EMPTY(&trunk->hash[i]),
424 		    ("%s: hash table not empty", __func__));
425 #endif
426 	free(trunk->hash, M_VLAN);
427 	trunk->hash = NULL;
428 	trunk->hwidth = trunk->hmask = 0;
429 }
430 
431 static int
vlan_inshash(struct ifvlantrunk * trunk,struct ifvlan * ifv)432 vlan_inshash(struct ifvlantrunk *trunk, struct ifvlan *ifv)
433 {
434 	int i, b;
435 	struct ifvlan *ifv2;
436 
437 	VLAN_XLOCK_ASSERT();
438 	KASSERT(trunk->hwidth > 0, ("%s: hwidth not positive", __func__));
439 
440 	b = 1 << trunk->hwidth;
441 	i = HASH(ifv->ifv_vid, trunk->hmask);
442 	CK_SLIST_FOREACH(ifv2, &trunk->hash[i], ifv_list)
443 		if (ifv->ifv_vid == ifv2->ifv_vid)
444 			return (EEXIST);
445 
446 	/*
447 	 * Grow the hash when the number of vlans exceeds half of the number of
448 	 * hash buckets squared. This will make the average linked-list length
449 	 * buckets/2.
450 	 */
451 	if (trunk->refcnt > (b * b) / 2) {
452 		vlan_growhash(trunk, 1);
453 		i = HASH(ifv->ifv_vid, trunk->hmask);
454 	}
455 	CK_SLIST_INSERT_HEAD(&trunk->hash[i], ifv, ifv_list);
456 	trunk->refcnt++;
457 
458 	return (0);
459 }
460 
461 static int
vlan_remhash(struct ifvlantrunk * trunk,struct ifvlan * ifv)462 vlan_remhash(struct ifvlantrunk *trunk, struct ifvlan *ifv)
463 {
464 	int i, b;
465 	struct ifvlan *ifv2;
466 
467 	VLAN_XLOCK_ASSERT();
468 	KASSERT(trunk->hwidth > 0, ("%s: hwidth not positive", __func__));
469 
470 	b = 1 << (trunk->hwidth - 1);
471 	i = HASH(ifv->ifv_vid, trunk->hmask);
472 	CK_SLIST_FOREACH(ifv2, &trunk->hash[i], ifv_list)
473 		if (ifv2 == ifv) {
474 			trunk->refcnt--;
475 			CK_SLIST_REMOVE(&trunk->hash[i], ifv2, ifvlan, ifv_list);
476 			if (trunk->refcnt < (b * b) / 2)
477 				vlan_growhash(trunk, -1);
478 			return (0);
479 		}
480 
481 	panic("%s: vlan not found\n", __func__);
482 	return (ENOENT); /*NOTREACHED*/
483 }
484 
485 /*
486  * Grow the hash larger or smaller if memory permits.
487  */
488 static void
vlan_growhash(struct ifvlantrunk * trunk,int howmuch)489 vlan_growhash(struct ifvlantrunk *trunk, int howmuch)
490 {
491 	struct ifvlan *ifv;
492 	struct ifvlanhead *hash2;
493 	int hwidth2, i, j, n, n2;
494 
495 	VLAN_XLOCK_ASSERT();
496 	KASSERT(trunk->hwidth > 0, ("%s: hwidth not positive", __func__));
497 
498 	if (howmuch == 0) {
499 		/* Harmless yet obvious coding error */
500 		printf("%s: howmuch is 0\n", __func__);
501 		return;
502 	}
503 
504 	hwidth2 = trunk->hwidth + howmuch;
505 	n = 1 << trunk->hwidth;
506 	n2 = 1 << hwidth2;
507 	/* Do not shrink the table below the default */
508 	if (hwidth2 < VLAN_DEF_HWIDTH)
509 		return;
510 
511 	hash2 = malloc(sizeof(struct ifvlanhead) * n2, M_VLAN, M_WAITOK);
512 	for (j = 0; j < n2; j++)
513 		CK_SLIST_INIT(&hash2[j]);
514 	for (i = 0; i < n; i++)
515 		while ((ifv = CK_SLIST_FIRST(&trunk->hash[i])) != NULL) {
516 			CK_SLIST_REMOVE(&trunk->hash[i], ifv, ifvlan, ifv_list);
517 			j = HASH(ifv->ifv_vid, n2 - 1);
518 			CK_SLIST_INSERT_HEAD(&hash2[j], ifv, ifv_list);
519 		}
520 	NET_EPOCH_WAIT();
521 	free(trunk->hash, M_VLAN);
522 	trunk->hash = hash2;
523 	trunk->hwidth = hwidth2;
524 	trunk->hmask = n2 - 1;
525 
526 	if (bootverbose)
527 		if_printf(trunk->parent,
528 		    "VLAN hash table resized from %d to %d buckets\n", n, n2);
529 }
530 
531 static __inline struct ifvlan *
vlan_gethash(struct ifvlantrunk * trunk,uint16_t vid)532 vlan_gethash(struct ifvlantrunk *trunk, uint16_t vid)
533 {
534 	struct ifvlan *ifv;
535 
536 	NET_EPOCH_ASSERT();
537 
538 	CK_SLIST_FOREACH(ifv, &trunk->hash[HASH(vid, trunk->hmask)], ifv_list)
539 		if (ifv->ifv_vid == vid)
540 			return (ifv);
541 	return (NULL);
542 }
543 
544 #if 0
545 /* Debugging code to view the hashtables. */
546 static void
547 vlan_dumphash(struct ifvlantrunk *trunk)
548 {
549 	int i;
550 	struct ifvlan *ifv;
551 
552 	for (i = 0; i < (1 << trunk->hwidth); i++) {
553 		printf("%d: ", i);
554 		CK_SLIST_FOREACH(ifv, &trunk->hash[i], ifv_list)
555 			printf("%s ", ifv->ifv_ifp->if_xname);
556 		printf("\n");
557 	}
558 }
559 #endif /* 0 */
560 #else
561 
562 static __inline struct ifvlan *
vlan_gethash(struct ifvlantrunk * trunk,uint16_t vid)563 vlan_gethash(struct ifvlantrunk *trunk, uint16_t vid)
564 {
565 
566 	return trunk->vlans[vid];
567 }
568 
569 static __inline int
vlan_inshash(struct ifvlantrunk * trunk,struct ifvlan * ifv)570 vlan_inshash(struct ifvlantrunk *trunk, struct ifvlan *ifv)
571 {
572 
573 	if (trunk->vlans[ifv->ifv_vid] != NULL)
574 		return EEXIST;
575 	trunk->vlans[ifv->ifv_vid] = ifv;
576 	trunk->refcnt++;
577 
578 	return (0);
579 }
580 
581 static __inline int
vlan_remhash(struct ifvlantrunk * trunk,struct ifvlan * ifv)582 vlan_remhash(struct ifvlantrunk *trunk, struct ifvlan *ifv)
583 {
584 
585 	trunk->vlans[ifv->ifv_vid] = NULL;
586 	trunk->refcnt--;
587 
588 	return (0);
589 }
590 
591 static __inline void
vlan_freehash(struct ifvlantrunk * trunk)592 vlan_freehash(struct ifvlantrunk *trunk)
593 {
594 }
595 
596 static __inline void
vlan_inithash(struct ifvlantrunk * trunk)597 vlan_inithash(struct ifvlantrunk *trunk)
598 {
599 }
600 
601 #endif /* !VLAN_ARRAY */
602 
603 static void
trunk_destroy(struct ifvlantrunk * trunk)604 trunk_destroy(struct ifvlantrunk *trunk)
605 {
606 	VLAN_XLOCK_ASSERT();
607 
608 	vlan_freehash(trunk);
609 	trunk->parent->if_vlantrunk = NULL;
610 	TRUNK_LOCK_DESTROY(trunk);
611 	if_rele(trunk->parent);
612 	free(trunk, M_VLAN);
613 }
614 
615 /*
616  * Program our multicast filter. What we're actually doing is
617  * programming the multicast filter of the parent. This has the
618  * side effect of causing the parent interface to receive multicast
619  * traffic that it doesn't really want, which ends up being discarded
620  * later by the upper protocol layers. Unfortunately, there's no way
621  * to avoid this: there really is only one physical interface.
622  */
623 static int
vlan_setmulti(struct ifnet * ifp)624 vlan_setmulti(struct ifnet *ifp)
625 {
626 	struct ifnet		*ifp_p;
627 	struct ifmultiaddr	*ifma;
628 	struct ifvlan		*sc;
629 	struct vlan_mc_entry	*mc;
630 	int			error;
631 
632 	VLAN_XLOCK_ASSERT();
633 
634 	/* Find the parent. */
635 	sc = ifp->if_softc;
636 	ifp_p = PARENT(sc);
637 
638 	CURVNET_SET_QUIET(ifp_p->if_vnet);
639 
640 	/* First, remove any existing filter entries. */
641 	while ((mc = CK_SLIST_FIRST(&sc->vlan_mc_listhead)) != NULL) {
642 		CK_SLIST_REMOVE_HEAD(&sc->vlan_mc_listhead, mc_entries);
643 		(void)if_delmulti(ifp_p, (struct sockaddr *)&mc->mc_addr);
644 		NET_EPOCH_CALL(vlan_mc_free, &mc->mc_epoch_ctx);
645 	}
646 
647 	/* Now program new ones. */
648 	IF_ADDR_WLOCK(ifp);
649 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
650 		if (ifma->ifma_addr->sa_family != AF_LINK)
651 			continue;
652 		mc = malloc(sizeof(struct vlan_mc_entry), M_VLAN, M_NOWAIT);
653 		if (mc == NULL) {
654 			IF_ADDR_WUNLOCK(ifp);
655 			CURVNET_RESTORE();
656 			return (ENOMEM);
657 		}
658 		bcopy(ifma->ifma_addr, &mc->mc_addr, ifma->ifma_addr->sa_len);
659 		mc->mc_addr.sdl_index = ifp_p->if_index;
660 		CK_SLIST_INSERT_HEAD(&sc->vlan_mc_listhead, mc, mc_entries);
661 	}
662 	IF_ADDR_WUNLOCK(ifp);
663 	CK_SLIST_FOREACH (mc, &sc->vlan_mc_listhead, mc_entries) {
664 		error = if_addmulti(ifp_p, (struct sockaddr *)&mc->mc_addr,
665 		    NULL);
666 		if (error) {
667 			CURVNET_RESTORE();
668 			return (error);
669 		}
670 	}
671 
672 	CURVNET_RESTORE();
673 	return (0);
674 }
675 
676 /*
677  * A handler for interface ifnet events.
678  */
679 static void
vlan_ifevent(void * arg __unused,struct ifnet * ifp,int event)680 vlan_ifevent(void *arg __unused, struct ifnet *ifp, int event)
681 {
682 	struct epoch_tracker et;
683 	struct ifvlan *ifv;
684 	struct ifvlantrunk *trunk;
685 
686 	if (event != IFNET_EVENT_UPDATE_BAUDRATE)
687 		return;
688 
689 	NET_EPOCH_ENTER(et);
690 	trunk = ifp->if_vlantrunk;
691 	if (trunk == NULL) {
692 		NET_EPOCH_EXIT(et);
693 		return;
694 	}
695 
696 	TRUNK_WLOCK(trunk);
697 	VLAN_FOREACH(ifv, trunk) {
698 		ifv->ifv_ifp->if_baudrate = ifp->if_baudrate;
699 	}
700 	TRUNK_WUNLOCK(trunk);
701 	NET_EPOCH_EXIT(et);
702 }
703 
704 /*
705  * A handler for parent interface link layer address changes.
706  * If the parent interface link layer address is changed we
707  * should also change it on all children vlans.
708  */
709 static void
vlan_iflladdr(void * arg __unused,struct ifnet * ifp)710 vlan_iflladdr(void *arg __unused, struct ifnet *ifp)
711 {
712 	struct epoch_tracker et;
713 	struct ifvlan *ifv;
714 	struct ifnet *ifv_ifp;
715 	struct ifvlantrunk *trunk;
716 	struct sockaddr_dl *sdl;
717 
718 	/* Need the epoch since this is run on taskqueue_swi. */
719 	NET_EPOCH_ENTER(et);
720 	trunk = ifp->if_vlantrunk;
721 	if (trunk == NULL) {
722 		NET_EPOCH_EXIT(et);
723 		return;
724 	}
725 
726 	/*
727 	 * OK, it's a trunk.  Loop over and change all vlan's lladdrs on it.
728 	 * We need an exclusive lock here to prevent concurrent SIOCSIFLLADDR
729 	 * ioctl calls on the parent garbling the lladdr of the child vlan.
730 	 */
731 	TRUNK_WLOCK(trunk);
732 	VLAN_FOREACH(ifv, trunk) {
733 		/*
734 		 * Copy new new lladdr into the ifv_ifp, enqueue a task
735 		 * to actually call if_setlladdr. if_setlladdr needs to
736 		 * be deferred to a taskqueue because it will call into
737 		 * the if_vlan ioctl path and try to acquire the global
738 		 * lock.
739 		 */
740 		ifv_ifp = ifv->ifv_ifp;
741 		bcopy(IF_LLADDR(ifp), IF_LLADDR(ifv_ifp),
742 		    ifp->if_addrlen);
743 		sdl = (struct sockaddr_dl *)ifv_ifp->if_addr->ifa_addr;
744 		sdl->sdl_alen = ifp->if_addrlen;
745 		taskqueue_enqueue(taskqueue_thread, &ifv->lladdr_task);
746 	}
747 	TRUNK_WUNLOCK(trunk);
748 	NET_EPOCH_EXIT(et);
749 }
750 
751 /*
752  * A handler for network interface departure events.
753  * Track departure of trunks here so that we don't access invalid
754  * pointers or whatever if a trunk is ripped from under us, e.g.,
755  * by ejecting its hot-plug card.  However, if an ifnet is simply
756  * being renamed, then there's no need to tear down the state.
757  */
758 static void
vlan_ifdetach(void * arg __unused,struct ifnet * ifp)759 vlan_ifdetach(void *arg __unused, struct ifnet *ifp)
760 {
761 	struct ifvlan *ifv;
762 	struct ifvlantrunk *trunk;
763 
764 	/* If the ifnet is just being renamed, don't do anything. */
765 	if (ifp->if_flags & IFF_RENAMING)
766 		return;
767 	VLAN_XLOCK();
768 	trunk = ifp->if_vlantrunk;
769 	if (trunk == NULL) {
770 		VLAN_XUNLOCK();
771 		return;
772 	}
773 
774 	/*
775 	 * OK, it's a trunk.  Loop over and detach all vlan's on it.
776 	 * Check trunk pointer after each vlan_unconfig() as it will
777 	 * free it and set to NULL after the last vlan was detached.
778 	 */
779 	VLAN_FOREACH_UNTIL_SAFE(ifv, ifp->if_vlantrunk,
780 	    ifp->if_vlantrunk == NULL)
781 		vlan_unconfig_locked(ifv->ifv_ifp, 1);
782 
783 	/* Trunk should have been destroyed in vlan_unconfig(). */
784 	KASSERT(ifp->if_vlantrunk == NULL, ("%s: purge failed", __func__));
785 	VLAN_XUNLOCK();
786 }
787 
788 /*
789  * Return the trunk device for a virtual interface.
790  */
791 static struct ifnet  *
vlan_trunkdev(struct ifnet * ifp)792 vlan_trunkdev(struct ifnet *ifp)
793 {
794 	struct ifvlan *ifv;
795 
796 	NET_EPOCH_ASSERT();
797 
798 	if (ifp->if_type != IFT_L2VLAN)
799 		return (NULL);
800 
801 	ifv = ifp->if_softc;
802 	ifp = NULL;
803 	if (ifv->ifv_trunk)
804 		ifp = PARENT(ifv);
805 	return (ifp);
806 }
807 
808 /*
809  * Return the 12-bit VLAN VID for this interface, for use by external
810  * components such as Infiniband.
811  *
812  * XXXRW: Note that the function name here is historical; it should be named
813  * vlan_vid().
814  */
815 static int
vlan_tag(struct ifnet * ifp,uint16_t * vidp)816 vlan_tag(struct ifnet *ifp, uint16_t *vidp)
817 {
818 	struct ifvlan *ifv;
819 
820 	if (ifp->if_type != IFT_L2VLAN)
821 		return (EINVAL);
822 	ifv = ifp->if_softc;
823 	*vidp = ifv->ifv_vid;
824 	return (0);
825 }
826 
827 static int
vlan_pcp(struct ifnet * ifp,uint16_t * pcpp)828 vlan_pcp(struct ifnet *ifp, uint16_t *pcpp)
829 {
830 	struct ifvlan *ifv;
831 
832 	if (ifp->if_type != IFT_L2VLAN)
833 		return (EINVAL);
834 	ifv = ifp->if_softc;
835 	*pcpp = ifv->ifv_pcp;
836 	return (0);
837 }
838 
839 /*
840  * Return a driver specific cookie for this interface.  Synchronization
841  * with setcookie must be provided by the driver.
842  */
843 static void *
vlan_cookie(struct ifnet * ifp)844 vlan_cookie(struct ifnet *ifp)
845 {
846 	struct ifvlan *ifv;
847 
848 	if (ifp->if_type != IFT_L2VLAN)
849 		return (NULL);
850 	ifv = ifp->if_softc;
851 	return (ifv->ifv_cookie);
852 }
853 
854 /*
855  * Store a cookie in our softc that drivers can use to store driver
856  * private per-instance data in.
857  */
858 static int
vlan_setcookie(struct ifnet * ifp,void * cookie)859 vlan_setcookie(struct ifnet *ifp, void *cookie)
860 {
861 	struct ifvlan *ifv;
862 
863 	if (ifp->if_type != IFT_L2VLAN)
864 		return (EINVAL);
865 	ifv = ifp->if_softc;
866 	ifv->ifv_cookie = cookie;
867 	return (0);
868 }
869 
870 /*
871  * Return the vlan device present at the specific VID.
872  */
873 static struct ifnet *
vlan_devat(struct ifnet * ifp,uint16_t vid)874 vlan_devat(struct ifnet *ifp, uint16_t vid)
875 {
876 	struct ifvlantrunk *trunk;
877 	struct ifvlan *ifv;
878 
879 	NET_EPOCH_ASSERT();
880 
881 	trunk = ifp->if_vlantrunk;
882 	if (trunk == NULL)
883 		return (NULL);
884 	ifp = NULL;
885 	ifv = vlan_gethash(trunk, vid);
886 	if (ifv)
887 		ifp = ifv->ifv_ifp;
888 	return (ifp);
889 }
890 
891 /*
892  * VLAN support can be loaded as a module.  The only place in the
893  * system that's intimately aware of this is ether_input.  We hook
894  * into this code through vlan_input_p which is defined there and
895  * set here.  No one else in the system should be aware of this so
896  * we use an explicit reference here.
897  */
898 extern	void (*vlan_input_p)(struct ifnet *, struct mbuf *);
899 
900 /* For if_link_state_change() eyes only... */
901 extern	void (*vlan_link_state_p)(struct ifnet *);
902 
903 static struct if_clone_addreq_v2 vlan_addreq = {
904 	.version = 2,
905 	.match_f = vlan_clone_match,
906 	.create_f = vlan_clone_create,
907 	.destroy_f = vlan_clone_destroy,
908 	.create_nl_f = vlan_clone_create_nl,
909 	.modify_nl_f = vlan_clone_modify_nl,
910 	.dump_nl_f = vlan_clone_dump_nl,
911 };
912 
913 static int
vlan_modevent(module_t mod,int type,void * data)914 vlan_modevent(module_t mod, int type, void *data)
915 {
916 
917 	switch (type) {
918 	case MOD_LOAD:
919 		ifdetach_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
920 		    vlan_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
921 		if (ifdetach_tag == NULL)
922 			return (ENOMEM);
923 		iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event,
924 		    vlan_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
925 		if (iflladdr_tag == NULL)
926 			return (ENOMEM);
927 		ifevent_tag = EVENTHANDLER_REGISTER(ifnet_event,
928 		    vlan_ifevent, NULL, EVENTHANDLER_PRI_ANY);
929 		if (ifevent_tag == NULL)
930 			return (ENOMEM);
931 		VLAN_LOCKING_INIT();
932 		vlan_input_p = vlan_input;
933 		vlan_link_state_p = vlan_link_state;
934 		vlan_trunk_cap_p = vlan_trunk_capabilities;
935 		vlan_trunkdev_p = vlan_trunkdev;
936 		vlan_cookie_p = vlan_cookie;
937 		vlan_setcookie_p = vlan_setcookie;
938 		vlan_tag_p = vlan_tag;
939 		vlan_pcp_p = vlan_pcp;
940 		vlan_devat_p = vlan_devat;
941 #ifndef VIMAGE
942 		vlan_cloner = ifc_attach_cloner(vlanname, (struct if_clone_addreq *)&vlan_addreq);
943 #endif
944 		if (bootverbose)
945 			printf("vlan: initialized, using "
946 #ifdef VLAN_ARRAY
947 			       "full-size arrays"
948 #else
949 			       "hash tables with chaining"
950 #endif
951 
952 			       "\n");
953 		break;
954 	case MOD_UNLOAD:
955 #ifndef VIMAGE
956 		ifc_detach_cloner(vlan_cloner);
957 #endif
958 		EVENTHANDLER_DEREGISTER(ifnet_departure_event, ifdetach_tag);
959 		EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_tag);
960 		EVENTHANDLER_DEREGISTER(ifnet_event, ifevent_tag);
961 		vlan_input_p = NULL;
962 		vlan_link_state_p = NULL;
963 		vlan_trunk_cap_p = NULL;
964 		vlan_trunkdev_p = NULL;
965 		vlan_tag_p = NULL;
966 		vlan_cookie_p = NULL;
967 		vlan_setcookie_p = NULL;
968 		vlan_devat_p = NULL;
969 		VLAN_LOCKING_DESTROY();
970 		if (bootverbose)
971 			printf("vlan: unloaded\n");
972 		break;
973 	default:
974 		return (EOPNOTSUPP);
975 	}
976 	return (0);
977 }
978 
979 static moduledata_t vlan_mod = {
980 	"if_vlan",
981 	vlan_modevent,
982 	0
983 };
984 
985 DECLARE_MODULE(if_vlan, vlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
986 MODULE_VERSION(if_vlan, 3);
987 
988 #ifdef VIMAGE
989 static void
vnet_vlan_init(const void * unused __unused)990 vnet_vlan_init(const void *unused __unused)
991 {
992 	vlan_cloner = ifc_attach_cloner(vlanname, (struct if_clone_addreq *)&vlan_addreq);
993 	V_vlan_cloner = vlan_cloner;
994 }
995 VNET_SYSINIT(vnet_vlan_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
996     vnet_vlan_init, NULL);
997 
998 static void
vnet_vlan_uninit(const void * unused __unused)999 vnet_vlan_uninit(const void *unused __unused)
1000 {
1001 
1002 	ifc_detach_cloner(V_vlan_cloner);
1003 }
1004 VNET_SYSUNINIT(vnet_vlan_uninit, SI_SUB_INIT_IF, SI_ORDER_ANY,
1005     vnet_vlan_uninit, NULL);
1006 #endif
1007 
1008 /*
1009  * Check for <etherif>.<vlan>[.<vlan> ...] style interface names.
1010  */
1011 static struct ifnet *
vlan_clone_match_ethervid(const char * name,int * vidp)1012 vlan_clone_match_ethervid(const char *name, int *vidp)
1013 {
1014 	char ifname[IFNAMSIZ];
1015 	char *cp;
1016 	struct ifnet *ifp;
1017 	int vid;
1018 
1019 	strlcpy(ifname, name, IFNAMSIZ);
1020 	if ((cp = strrchr(ifname, '.')) == NULL)
1021 		return (NULL);
1022 	*cp = '\0';
1023 	if ((ifp = ifunit_ref(ifname)) == NULL)
1024 		return (NULL);
1025 	/* Parse VID. */
1026 	if (*++cp == '\0') {
1027 		if_rele(ifp);
1028 		return (NULL);
1029 	}
1030 	vid = 0;
1031 	for(; *cp >= '0' && *cp <= '9'; cp++)
1032 		vid = (vid * 10) + (*cp - '0');
1033 	if (*cp != '\0') {
1034 		if_rele(ifp);
1035 		return (NULL);
1036 	}
1037 	if (vidp != NULL)
1038 		*vidp = vid;
1039 
1040 	return (ifp);
1041 }
1042 
1043 static int
vlan_clone_match(struct if_clone * ifc,const char * name)1044 vlan_clone_match(struct if_clone *ifc, const char *name)
1045 {
1046 	struct ifnet *ifp;
1047 	const char *cp;
1048 
1049 	ifp = vlan_clone_match_ethervid(name, NULL);
1050 	if (ifp != NULL) {
1051 		if_rele(ifp);
1052 		return (1);
1053 	}
1054 
1055 	if (strncmp(vlanname, name, strlen(vlanname)) != 0)
1056 		return (0);
1057 	for (cp = name + 4; *cp != '\0'; cp++) {
1058 		if (*cp < '0' || *cp > '9')
1059 			return (0);
1060 	}
1061 
1062 	return (1);
1063 }
1064 
1065 static int
vlan_clone_create(struct if_clone * ifc,char * name,size_t len,struct ifc_data * ifd,struct ifnet ** ifpp)1066 vlan_clone_create(struct if_clone *ifc, char *name, size_t len,
1067     struct ifc_data *ifd, struct ifnet **ifpp)
1068 {
1069 	char *dp;
1070 	bool wildcard = false;
1071 	bool subinterface = false;
1072 	int unit;
1073 	int error;
1074 	int vid = 0;
1075 	uint16_t proto = ETHERTYPE_VLAN;
1076 	struct ifvlan *ifv;
1077 	struct ifnet *ifp;
1078 	struct ifnet *p = NULL;
1079 	struct ifaddr *ifa;
1080 	struct sockaddr_dl *sdl;
1081 	struct vlanreq vlr;
1082 	static const u_char eaddr[ETHER_ADDR_LEN];	/* 00:00:00:00:00:00 */
1083 
1084 
1085 	/*
1086 	 * There are three ways to specify the cloned device:
1087 	 * o pass a parameter block with the clone request.
1088 	 * o specify parameters in the text of the clone device name
1089 	 * o specify no parameters and get an unattached device that
1090 	 *   must be configured separately.
1091 	 * The first technique is preferred; the latter two are supported
1092 	 * for backwards compatibility.
1093 	 *
1094 	 * XXXRW: Note historic use of the word "tag" here.  New ioctls may be
1095 	 * called for.
1096 	 */
1097 
1098 	if (ifd->params != NULL) {
1099 		error = ifc_copyin(ifd, &vlr, sizeof(vlr));
1100 		if (error)
1101 			return error;
1102 		vid = vlr.vlr_tag;
1103 		proto = vlr.vlr_proto;
1104 		if (proto == 0)
1105 			proto = ETHERTYPE_VLAN;
1106 		p = ifunit_ref(vlr.vlr_parent);
1107 		if (p == NULL)
1108 			return (ENXIO);
1109 	}
1110 
1111 	if ((error = ifc_name2unit(name, &unit)) == 0) {
1112 
1113 		/*
1114 		 * vlanX interface. Set wildcard to true if the unit number
1115 		 * is not fixed (-1)
1116 		 */
1117 		wildcard = (unit < 0);
1118 	} else {
1119 		struct ifnet *p_tmp = vlan_clone_match_ethervid(name, &vid);
1120 		if (p_tmp != NULL) {
1121 			error = 0;
1122 			subinterface = true;
1123 			unit = IF_DUNIT_NONE;
1124 			wildcard = false;
1125 			if (p != NULL) {
1126 				if_rele(p_tmp);
1127 				if (p != p_tmp)
1128 					error = EINVAL;
1129 			} else
1130 				p = p_tmp;
1131 		} else
1132 			error = ENXIO;
1133 	}
1134 
1135 	if (error != 0) {
1136 		if (p != NULL)
1137 			if_rele(p);
1138 		return (error);
1139 	}
1140 
1141 	if (!subinterface) {
1142 		/* vlanX interface, mark X as busy or allocate new unit # */
1143 		error = ifc_alloc_unit(ifc, &unit);
1144 		if (error != 0) {
1145 			if (p != NULL)
1146 				if_rele(p);
1147 			return (error);
1148 		}
1149 	}
1150 
1151 	/* In the wildcard case, we need to update the name. */
1152 	if (wildcard) {
1153 		for (dp = name; *dp != '\0'; dp++);
1154 		if (snprintf(dp, len - (dp-name), "%d", unit) >
1155 		    len - (dp-name) - 1) {
1156 			panic("%s: interface name too long", __func__);
1157 		}
1158 	}
1159 
1160 	ifv = malloc(sizeof(struct ifvlan), M_VLAN, M_WAITOK | M_ZERO);
1161 	ifp = ifv->ifv_ifp = if_alloc(IFT_ETHER);
1162 	CK_SLIST_INIT(&ifv->vlan_mc_listhead);
1163 	ifp->if_softc = ifv;
1164 	/*
1165 	 * Set the name manually rather than using if_initname because
1166 	 * we don't conform to the default naming convention for interfaces.
1167 	 */
1168 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
1169 	ifp->if_dname = vlanname;
1170 	ifp->if_dunit = unit;
1171 
1172 	ifp->if_init = vlan_init;
1173 #ifdef ALTQ
1174 	ifp->if_start = vlan_altq_start;
1175 	ifp->if_transmit = vlan_altq_transmit;
1176 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
1177 	ifp->if_snd.ifq_drv_maxlen = 0;
1178 	IFQ_SET_READY(&ifp->if_snd);
1179 #else
1180 	ifp->if_transmit = vlan_transmit;
1181 #endif
1182 	ifp->if_qflush = vlan_qflush;
1183 	ifp->if_ioctl = vlan_ioctl;
1184 #if defined(KERN_TLS) || defined(RATELIMIT)
1185 	ifp->if_snd_tag_alloc = vlan_snd_tag_alloc;
1186 	ifp->if_ratelimit_query = vlan_ratelimit_query;
1187 #endif
1188 	ifp->if_flags = VLAN_IFFLAGS;
1189 	ether_ifattach(ifp, eaddr);
1190 	/* Now undo some of the damage... */
1191 	ifp->if_baudrate = 0;
1192 	ifp->if_type = IFT_L2VLAN;
1193 	ifp->if_hdrlen = ETHER_VLAN_ENCAP_LEN;
1194 	ifa = ifp->if_addr;
1195 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1196 	sdl->sdl_type = IFT_L2VLAN;
1197 
1198 	if (p != NULL) {
1199 		error = vlan_config(ifv, p, vid, proto);
1200 		if_rele(p);
1201 		if (error != 0) {
1202 			/*
1203 			 * Since we've partially failed, we need to back
1204 			 * out all the way, otherwise userland could get
1205 			 * confused.  Thus, we destroy the interface.
1206 			 */
1207 			ether_ifdetach(ifp);
1208 			vlan_unconfig(ifp);
1209 			if_free(ifp);
1210 			if (!subinterface)
1211 				ifc_free_unit(ifc, unit);
1212 			free(ifv, M_VLAN);
1213 
1214 			return (error);
1215 		}
1216 	}
1217 	*ifpp = ifp;
1218 
1219 	return (0);
1220 }
1221 
1222 /*
1223  *
1224  * Parsers of IFLA_INFO_DATA inside IFLA_LINKINFO of RTM_NEWLINK
1225  *    {{nla_len=8, nla_type=IFLA_LINK}, 2},
1226  *    {{nla_len=12, nla_type=IFLA_IFNAME}, "xvlan22"},
1227  *    {{nla_len=24, nla_type=IFLA_LINKINFO},
1228  *     [
1229  *      {{nla_len=8, nla_type=IFLA_INFO_KIND}, "vlan"...},
1230  *      {{nla_len=12, nla_type=IFLA_INFO_DATA}, "\x06\x00\x01\x00\x16\x00\x00\x00"}]}
1231  */
1232 
1233 struct nl_parsed_vlan {
1234 	uint16_t vlan_id;
1235 	uint16_t vlan_proto;
1236 	struct ifla_vlan_flags vlan_flags;
1237 };
1238 
1239 #define	_OUT(_field)	offsetof(struct nl_parsed_vlan, _field)
1240 static const struct nlattr_parser nla_p_vlan[] = {
1241 	{ .type = IFLA_VLAN_ID, .off = _OUT(vlan_id), .cb = nlattr_get_uint16 },
1242 	{ .type = IFLA_VLAN_FLAGS, .off = _OUT(vlan_flags), .cb = nlattr_get_nla },
1243 	{ .type = IFLA_VLAN_PROTOCOL, .off = _OUT(vlan_proto), .cb = nlattr_get_uint16 },
1244 };
1245 #undef _OUT
1246 NL_DECLARE_ATTR_PARSER(vlan_parser, nla_p_vlan);
1247 
1248 static int
vlan_clone_create_nl(struct if_clone * ifc,char * name,size_t len,struct ifc_data_nl * ifd)1249 vlan_clone_create_nl(struct if_clone *ifc, char *name, size_t len,
1250     struct ifc_data_nl *ifd)
1251 {
1252 	struct epoch_tracker et;
1253         struct ifnet *ifp_parent;
1254 	struct nl_pstate *npt = ifd->npt;
1255 	struct nl_parsed_link *lattrs = ifd->lattrs;
1256 	int error;
1257 
1258 	/*
1259 	 * lattrs.ifla_ifname is the new interface name
1260 	 * lattrs.ifi_index contains parent interface index
1261 	 * lattrs.ifla_idata contains un-parsed vlan data
1262 	 */
1263 	struct nl_parsed_vlan attrs = {
1264 		.vlan_id = 0xFEFE,
1265 		.vlan_proto = ETHERTYPE_VLAN
1266 	};
1267 
1268 	if (lattrs->ifla_idata == NULL) {
1269 		nlmsg_report_err_msg(npt, "vlan id is required, guessing not supported");
1270 		return (ENOTSUP);
1271 	}
1272 
1273 	error = nl_parse_nested(lattrs->ifla_idata, &vlan_parser, npt, &attrs);
1274 	if (error != 0)
1275 		return (error);
1276 	if (attrs.vlan_id > 4095) {
1277 		nlmsg_report_err_msg(npt, "Invalid VID: %d", attrs.vlan_id);
1278 		return (EINVAL);
1279 	}
1280 	if (attrs.vlan_proto != ETHERTYPE_VLAN && attrs.vlan_proto != ETHERTYPE_QINQ) {
1281 		nlmsg_report_err_msg(npt, "Unsupported ethertype: 0x%04X", attrs.vlan_proto);
1282 		return (ENOTSUP);
1283 	}
1284 
1285 	struct vlanreq params = {
1286 		.vlr_tag = attrs.vlan_id,
1287 		.vlr_proto = attrs.vlan_proto,
1288 	};
1289 	struct ifc_data ifd_new = { .flags = IFC_F_SYSSPACE, .unit = ifd->unit, .params = &params };
1290 
1291 	NET_EPOCH_ENTER(et);
1292 	ifp_parent = ifnet_byindex(lattrs->ifi_index);
1293 	if (ifp_parent != NULL)
1294 		strlcpy(params.vlr_parent, if_name(ifp_parent), sizeof(params.vlr_parent));
1295 	NET_EPOCH_EXIT(et);
1296 
1297 	if (ifp_parent == NULL) {
1298 		nlmsg_report_err_msg(npt, "unable to find parent interface %u", lattrs->ifi_index);
1299 		return (ENOENT);
1300 	}
1301 
1302 	error = vlan_clone_create(ifc, name, len, &ifd_new, &ifd->ifp);
1303 
1304 	return (error);
1305 }
1306 
1307 static int
vlan_clone_modify_nl(struct ifnet * ifp,struct ifc_data_nl * ifd)1308 vlan_clone_modify_nl(struct ifnet *ifp, struct ifc_data_nl *ifd)
1309 {
1310 	struct nl_parsed_link *lattrs = ifd->lattrs;
1311 
1312 	if ((lattrs->ifla_idata != NULL) && ((ifd->flags & IFC_F_CREATE) == 0)) {
1313 		struct epoch_tracker et;
1314 		struct nl_parsed_vlan attrs = {
1315 			.vlan_proto = ETHERTYPE_VLAN,
1316 		};
1317 		int error;
1318 
1319 		error = nl_parse_nested(lattrs->ifla_idata, &vlan_parser, ifd->npt, &attrs);
1320 		if (error != 0)
1321 			return (error);
1322 
1323 		NET_EPOCH_ENTER(et);
1324 		struct ifnet *ifp_parent = ifnet_byindex_ref(lattrs->ifla_link);
1325 		NET_EPOCH_EXIT(et);
1326 
1327 		if (ifp_parent == NULL) {
1328 			nlmsg_report_err_msg(ifd->npt, "unable to find parent interface %u",
1329 			    lattrs->ifla_link);
1330 			return (ENOENT);
1331 		}
1332 
1333 		struct ifvlan *ifv = ifp->if_softc;
1334 		error = vlan_config(ifv, ifp_parent, attrs.vlan_id, attrs.vlan_proto);
1335 
1336 		if_rele(ifp_parent);
1337 		if (error != 0)
1338 			return (error);
1339 	}
1340 
1341 	return (nl_modify_ifp_generic(ifp, ifd->lattrs, ifd->bm, ifd->npt));
1342 }
1343 
1344 /*
1345  *    {{nla_len=24, nla_type=IFLA_LINKINFO},
1346  *     [
1347  *      {{nla_len=8, nla_type=IFLA_INFO_KIND}, "vlan"...},
1348  *      {{nla_len=12, nla_type=IFLA_INFO_DATA}, "\x06\x00\x01\x00\x16\x00\x00\x00"}]}
1349  */
1350 static void
vlan_clone_dump_nl(struct ifnet * ifp,struct nl_writer * nw)1351 vlan_clone_dump_nl(struct ifnet *ifp, struct nl_writer *nw)
1352 {
1353 	uint32_t parent_index = 0;
1354 	uint16_t vlan_id = 0;
1355 	uint16_t vlan_proto = 0;
1356 
1357 	VLAN_SLOCK();
1358 	struct ifvlan *ifv = ifp->if_softc;
1359 	if (TRUNK(ifv) != NULL)
1360 		parent_index = PARENT(ifv)->if_index;
1361 	vlan_id = ifv->ifv_vid;
1362 	vlan_proto = ifv->ifv_proto;
1363 	VLAN_SUNLOCK();
1364 
1365 	if (parent_index != 0)
1366 		nlattr_add_u32(nw, IFLA_LINK, parent_index);
1367 
1368 	int off = nlattr_add_nested(nw, IFLA_LINKINFO);
1369 	if (off != 0) {
1370 		nlattr_add_string(nw, IFLA_INFO_KIND, "vlan");
1371 		int off2 = nlattr_add_nested(nw, IFLA_INFO_DATA);
1372 		if (off2 != 0) {
1373 			nlattr_add_u16(nw, IFLA_VLAN_ID, vlan_id);
1374 			nlattr_add_u16(nw, IFLA_VLAN_PROTOCOL, vlan_proto);
1375 			nlattr_set_len(nw, off2);
1376 		}
1377 		nlattr_set_len(nw, off);
1378 	}
1379 }
1380 
1381 static int
vlan_clone_destroy(struct if_clone * ifc,struct ifnet * ifp,uint32_t flags)1382 vlan_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
1383 {
1384 	struct ifvlan *ifv = ifp->if_softc;
1385 	int unit = ifp->if_dunit;
1386 
1387 	if (ifp->if_vlantrunk)
1388 		return (EBUSY);
1389 
1390 #ifdef ALTQ
1391 	IFQ_PURGE(&ifp->if_snd);
1392 #endif
1393 	ether_ifdetach(ifp);	/* first, remove it from system-wide lists */
1394 	vlan_unconfig(ifp);	/* now it can be unconfigured and freed */
1395 	/*
1396 	 * We should have the only reference to the ifv now, so we can now
1397 	 * drain any remaining lladdr task before freeing the ifnet and the
1398 	 * ifvlan.
1399 	 */
1400 	taskqueue_drain(taskqueue_thread, &ifv->lladdr_task);
1401 	NET_EPOCH_WAIT();
1402 	if_free(ifp);
1403 	free(ifv, M_VLAN);
1404 	if (unit != IF_DUNIT_NONE)
1405 		ifc_free_unit(ifc, unit);
1406 
1407 	return (0);
1408 }
1409 
1410 /*
1411  * The ifp->if_init entry point for vlan(4) is a no-op.
1412  */
1413 static void
vlan_init(void * foo __unused)1414 vlan_init(void *foo __unused)
1415 {
1416 }
1417 
1418 /*
1419  * The if_transmit method for vlan(4) interface.
1420  */
1421 static int
vlan_transmit(struct ifnet * ifp,struct mbuf * m)1422 vlan_transmit(struct ifnet *ifp, struct mbuf *m)
1423 {
1424 	struct ifvlan *ifv;
1425 	struct ifnet *p;
1426 	int error, len, mcast;
1427 
1428 	NET_EPOCH_ASSERT();
1429 
1430 	ifv = ifp->if_softc;
1431 	if (TRUNK(ifv) == NULL) {
1432 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1433 		m_freem(m);
1434 		return (ENETDOWN);
1435 	}
1436 	p = PARENT(ifv);
1437 	len = m->m_pkthdr.len;
1438 	mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
1439 
1440 	BPF_MTAP(ifp, m);
1441 
1442 #if defined(KERN_TLS) || defined(RATELIMIT)
1443 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) {
1444 		struct vlan_snd_tag *vst;
1445 		struct m_snd_tag *mst;
1446 
1447 		MPASS(m->m_pkthdr.snd_tag->ifp == ifp);
1448 		mst = m->m_pkthdr.snd_tag;
1449 		vst = mst_to_vst(mst);
1450 		if (vst->tag->ifp != p) {
1451 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1452 			m_freem(m);
1453 			return (EAGAIN);
1454 		}
1455 
1456 		m->m_pkthdr.snd_tag = m_snd_tag_ref(vst->tag);
1457 		m_snd_tag_rele(mst);
1458 	}
1459 #endif
1460 
1461 	/*
1462 	 * Do not run parent's if_transmit() if the parent is not up,
1463 	 * or parent's driver will cause a system crash.
1464 	 */
1465 	if (!UP_AND_RUNNING(p)) {
1466 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1467 		m_freem(m);
1468 		return (ENETDOWN);
1469 	}
1470 
1471 	if (!ether_8021q_frame(&m, ifp, p, &ifv->ifv_qtag)) {
1472 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1473 		return (0);
1474 	}
1475 
1476 	/*
1477 	 * Send it, precisely as ether_output() would have.
1478 	 */
1479 	error = (p->if_transmit)(p, m);
1480 	if (error == 0) {
1481 		if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
1482 		if_inc_counter(ifp, IFCOUNTER_OBYTES, len);
1483 		if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast);
1484 	} else
1485 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1486 	return (error);
1487 }
1488 
1489 static int
vlan_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)1490 vlan_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
1491     struct route *ro)
1492 {
1493 	struct ifvlan *ifv;
1494 	struct ifnet *p;
1495 
1496 	NET_EPOCH_ASSERT();
1497 
1498 	/*
1499 	 * Find the first non-VLAN parent interface.
1500 	 */
1501 	ifv = ifp->if_softc;
1502 	do {
1503 		if (TRUNK(ifv) == NULL) {
1504 			m_freem(m);
1505 			return (ENETDOWN);
1506 		}
1507 		p = PARENT(ifv);
1508 		ifv = p->if_softc;
1509 	} while (p->if_type == IFT_L2VLAN);
1510 
1511 	return p->if_output(ifp, m, dst, ro);
1512 }
1513 
1514 #ifdef ALTQ
1515 static void
vlan_altq_start(if_t ifp)1516 vlan_altq_start(if_t ifp)
1517 {
1518 	struct ifaltq *ifq = &ifp->if_snd;
1519 	struct mbuf *m;
1520 
1521 	IFQ_LOCK(ifq);
1522 	IFQ_DEQUEUE_NOLOCK(ifq, m);
1523 	while (m != NULL) {
1524 		vlan_transmit(ifp, m);
1525 		IFQ_DEQUEUE_NOLOCK(ifq, m);
1526 	}
1527 	IFQ_UNLOCK(ifq);
1528 }
1529 
1530 static int
vlan_altq_transmit(if_t ifp,struct mbuf * m)1531 vlan_altq_transmit(if_t ifp, struct mbuf *m)
1532 {
1533 	int err;
1534 
1535 	if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
1536 		IFQ_ENQUEUE(&ifp->if_snd, m, err);
1537 		if (err == 0)
1538 			vlan_altq_start(ifp);
1539 	} else
1540 		err = vlan_transmit(ifp, m);
1541 
1542 	return (err);
1543 }
1544 #endif	/* ALTQ */
1545 
1546 /*
1547  * The ifp->if_qflush entry point for vlan(4) is a no-op.
1548  */
1549 static void
vlan_qflush(struct ifnet * ifp __unused)1550 vlan_qflush(struct ifnet *ifp __unused)
1551 {
1552 }
1553 
1554 static void
vlan_input(struct ifnet * ifp,struct mbuf * m)1555 vlan_input(struct ifnet *ifp, struct mbuf *m)
1556 {
1557 	struct ifvlantrunk *trunk;
1558 	struct ifvlan *ifv;
1559 	struct m_tag *mtag;
1560 	uint16_t vid, tag;
1561 
1562 	NET_EPOCH_ASSERT();
1563 
1564 	trunk = ifp->if_vlantrunk;
1565 	if (trunk == NULL) {
1566 		m_freem(m);
1567 		return;
1568 	}
1569 
1570 	if (m->m_flags & M_VLANTAG) {
1571 		/*
1572 		 * Packet is tagged, but m contains a normal
1573 		 * Ethernet frame; the tag is stored out-of-band.
1574 		 */
1575 		tag = m->m_pkthdr.ether_vtag;
1576 		m->m_flags &= ~M_VLANTAG;
1577 	} else {
1578 		struct ether_vlan_header *evl;
1579 
1580 		/*
1581 		 * Packet is tagged in-band as specified by 802.1q.
1582 		 */
1583 		switch (ifp->if_type) {
1584 		case IFT_ETHER:
1585 			if (m->m_len < sizeof(*evl) &&
1586 			    (m = m_pullup(m, sizeof(*evl))) == NULL) {
1587 				if_printf(ifp, "cannot pullup VLAN header\n");
1588 				return;
1589 			}
1590 			evl = mtod(m, struct ether_vlan_header *);
1591 			tag = ntohs(evl->evl_tag);
1592 
1593 			/*
1594 			 * Remove the 802.1q header by copying the Ethernet
1595 			 * addresses over it and adjusting the beginning of
1596 			 * the data in the mbuf.  The encapsulated Ethernet
1597 			 * type field is already in place.
1598 			 */
1599 			bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
1600 			      ETHER_HDR_LEN - ETHER_TYPE_LEN);
1601 			m_adj(m, ETHER_VLAN_ENCAP_LEN);
1602 			break;
1603 
1604 		default:
1605 #ifdef INVARIANTS
1606 			panic("%s: %s has unsupported if_type %u",
1607 			      __func__, ifp->if_xname, ifp->if_type);
1608 #endif
1609 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
1610 			m_freem(m);
1611 			return;
1612 		}
1613 	}
1614 
1615 	vid = EVL_VLANOFTAG(tag);
1616 
1617 	ifv = vlan_gethash(trunk, vid);
1618 	if (ifv == NULL || !UP_AND_RUNNING(ifv->ifv_ifp)) {
1619 		if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
1620 		m_freem(m);
1621 		return;
1622 	}
1623 
1624 	if (V_vlan_mtag_pcp) {
1625 		/*
1626 		 * While uncommon, it is possible that we will find a 802.1q
1627 		 * packet encapsulated inside another packet that also had an
1628 		 * 802.1q header.  For example, ethernet tunneled over IPSEC
1629 		 * arriving over ethernet.  In that case, we replace the
1630 		 * existing 802.1q PCP m_tag value.
1631 		 */
1632 		mtag = m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_IN, NULL);
1633 		if (mtag == NULL) {
1634 			mtag = m_tag_alloc(MTAG_8021Q, MTAG_8021Q_PCP_IN,
1635 			    sizeof(uint8_t), M_NOWAIT);
1636 			if (mtag == NULL) {
1637 				if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
1638 				m_freem(m);
1639 				return;
1640 			}
1641 			m_tag_prepend(m, mtag);
1642 		}
1643 		*(uint8_t *)(mtag + 1) = EVL_PRIOFTAG(tag);
1644 	}
1645 
1646 	m->m_pkthdr.rcvif = ifv->ifv_ifp;
1647 	if_inc_counter(ifv->ifv_ifp, IFCOUNTER_IPACKETS, 1);
1648 
1649 	/* Pass it back through the parent's input routine. */
1650 	(*ifv->ifv_ifp->if_input)(ifv->ifv_ifp, m);
1651 }
1652 
1653 static void
vlan_lladdr_fn(void * arg,int pending __unused)1654 vlan_lladdr_fn(void *arg, int pending __unused)
1655 {
1656 	struct ifvlan *ifv;
1657 	struct ifnet *ifp;
1658 
1659 	ifv = (struct ifvlan *)arg;
1660 	ifp = ifv->ifv_ifp;
1661 
1662 	CURVNET_SET(ifp->if_vnet);
1663 
1664 	/* The ifv_ifp already has the lladdr copied in. */
1665 	if_setlladdr(ifp, IF_LLADDR(ifp), ifp->if_addrlen);
1666 
1667 	CURVNET_RESTORE();
1668 }
1669 
1670 static int
vlan_config(struct ifvlan * ifv,struct ifnet * p,uint16_t vid,uint16_t proto)1671 vlan_config(struct ifvlan *ifv, struct ifnet *p, uint16_t vid,
1672 	uint16_t proto)
1673 {
1674 	struct epoch_tracker et;
1675 	struct ifvlantrunk *trunk;
1676 	struct ifnet *ifp;
1677 	int error = 0;
1678 
1679 	/*
1680 	 * We can handle non-ethernet hardware types as long as
1681 	 * they handle the tagging and headers themselves.
1682 	 */
1683 	if (p->if_type != IFT_ETHER &&
1684 	    p->if_type != IFT_L2VLAN &&
1685 	    (p->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
1686 		return (EPROTONOSUPPORT);
1687 	if ((p->if_flags & VLAN_IFFLAGS) != VLAN_IFFLAGS)
1688 		return (EPROTONOSUPPORT);
1689 	/*
1690 	 * Don't let the caller set up a VLAN VID with
1691 	 * anything except VLID bits.
1692 	 * VID numbers 0x0 and 0xFFF are reserved.
1693 	 */
1694 	if (vid == 0 || vid == 0xFFF || (vid & ~EVL_VLID_MASK))
1695 		return (EINVAL);
1696 	if (ifv->ifv_trunk) {
1697 		trunk = ifv->ifv_trunk;
1698 		if (trunk->parent != p)
1699 			return (EBUSY);
1700 
1701 		VLAN_XLOCK();
1702 
1703 		ifv->ifv_proto = proto;
1704 
1705 		if (ifv->ifv_vid != vid) {
1706 			int oldvid = ifv->ifv_vid;
1707 
1708 			/* Re-hash */
1709 			vlan_remhash(trunk, ifv);
1710 			ifv->ifv_vid = vid;
1711 			error = vlan_inshash(trunk, ifv);
1712 			if (error) {
1713 				int ret __diagused;
1714 
1715 				ifv->ifv_vid = oldvid;
1716 				/* Re-insert back where we found it. */
1717 				ret = vlan_inshash(trunk, ifv);
1718 				MPASS(ret == 0);
1719 			}
1720 		}
1721 		/* Will unlock */
1722 		goto done;
1723 	}
1724 
1725 	VLAN_XLOCK();
1726 	if (p->if_vlantrunk == NULL) {
1727 		trunk = malloc(sizeof(struct ifvlantrunk),
1728 		    M_VLAN, M_WAITOK | M_ZERO);
1729 		vlan_inithash(trunk);
1730 		TRUNK_LOCK_INIT(trunk);
1731 		TRUNK_WLOCK(trunk);
1732 		p->if_vlantrunk = trunk;
1733 		trunk->parent = p;
1734 		if_ref(trunk->parent);
1735 		TRUNK_WUNLOCK(trunk);
1736 	} else {
1737 		trunk = p->if_vlantrunk;
1738 	}
1739 
1740 	ifv->ifv_vid = vid;	/* must set this before vlan_inshash() */
1741 	ifv->ifv_pcp = 0;       /* Default: best effort delivery. */
1742 	error = vlan_inshash(trunk, ifv);
1743 	if (error)
1744 		goto done;
1745 	ifv->ifv_proto = proto;
1746 	ifv->ifv_encaplen = ETHER_VLAN_ENCAP_LEN;
1747 	ifv->ifv_mintu = ETHERMIN;
1748 	ifv->ifv_pflags = 0;
1749 	ifv->ifv_capenable = -1;
1750 	ifv->ifv_capenable2 = -1;
1751 
1752 	/*
1753 	 * If the parent supports the VLAN_MTU capability,
1754 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames,
1755 	 * use it.
1756 	 */
1757 	if (p->if_capenable & IFCAP_VLAN_MTU) {
1758 		/*
1759 		 * No need to fudge the MTU since the parent can
1760 		 * handle extended frames.
1761 		 */
1762 		ifv->ifv_mtufudge = 0;
1763 	} else {
1764 		/*
1765 		 * Fudge the MTU by the encapsulation size.  This
1766 		 * makes us incompatible with strictly compliant
1767 		 * 802.1Q implementations, but allows us to use
1768 		 * the feature with other NetBSD implementations,
1769 		 * which might still be useful.
1770 		 */
1771 		ifv->ifv_mtufudge = ifv->ifv_encaplen;
1772 	}
1773 
1774 	ifv->ifv_trunk = trunk;
1775 	ifp = ifv->ifv_ifp;
1776 	/*
1777 	 * Initialize fields from our parent.  This duplicates some
1778 	 * work with ether_ifattach() but allows for non-ethernet
1779 	 * interfaces to also work.
1780 	 */
1781 	ifp->if_mtu = p->if_mtu - ifv->ifv_mtufudge;
1782 	ifp->if_baudrate = p->if_baudrate;
1783 	ifp->if_input = p->if_input;
1784 	ifp->if_resolvemulti = p->if_resolvemulti;
1785 	ifp->if_addrlen = p->if_addrlen;
1786 	ifp->if_broadcastaddr = p->if_broadcastaddr;
1787 	ifp->if_pcp = ifv->ifv_pcp;
1788 
1789 	/*
1790 	 * We wrap the parent's if_output using vlan_output to ensure that it
1791 	 * can't become stale.
1792 	 */
1793 	ifp->if_output = vlan_output;
1794 
1795 	/*
1796 	 * Copy only a selected subset of flags from the parent.
1797 	 * Other flags are none of our business.
1798 	 */
1799 #define VLAN_COPY_FLAGS (IFF_SIMPLEX)
1800 	ifp->if_flags &= ~VLAN_COPY_FLAGS;
1801 	ifp->if_flags |= p->if_flags & VLAN_COPY_FLAGS;
1802 #undef VLAN_COPY_FLAGS
1803 
1804 	ifp->if_link_state = p->if_link_state;
1805 
1806 	NET_EPOCH_ENTER(et);
1807 	vlan_capabilities(ifv);
1808 	NET_EPOCH_EXIT(et);
1809 
1810 	/*
1811 	 * Set up our interface address to reflect the underlying
1812 	 * physical interface's.
1813 	 */
1814 	TASK_INIT(&ifv->lladdr_task, 0, vlan_lladdr_fn, ifv);
1815 	((struct sockaddr_dl *)ifp->if_addr->ifa_addr)->sdl_alen =
1816 	    p->if_addrlen;
1817 
1818 	/*
1819 	 * Do not schedule link address update if it was the same
1820 	 * as previous parent's. This helps avoid updating for each
1821 	 * associated llentry.
1822 	 */
1823 	if (memcmp(IF_LLADDR(p), IF_LLADDR(ifp), p->if_addrlen) != 0) {
1824 		bcopy(IF_LLADDR(p), IF_LLADDR(ifp), p->if_addrlen);
1825 		taskqueue_enqueue(taskqueue_thread, &ifv->lladdr_task);
1826 	}
1827 
1828 	/* We are ready for operation now. */
1829 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
1830 
1831 	/* Update flags on the parent, if necessary. */
1832 	vlan_setflags(ifp, 1);
1833 
1834 	/*
1835 	 * Configure multicast addresses that may already be
1836 	 * joined on the vlan device.
1837 	 */
1838 	(void)vlan_setmulti(ifp);
1839 
1840 done:
1841 	if (error == 0)
1842 		EVENTHANDLER_INVOKE(vlan_config, p, ifv->ifv_vid);
1843 	VLAN_XUNLOCK();
1844 
1845 	return (error);
1846 }
1847 
1848 static void
vlan_unconfig(struct ifnet * ifp)1849 vlan_unconfig(struct ifnet *ifp)
1850 {
1851 
1852 	VLAN_XLOCK();
1853 	vlan_unconfig_locked(ifp, 0);
1854 	VLAN_XUNLOCK();
1855 }
1856 
1857 static void
vlan_unconfig_locked(struct ifnet * ifp,int departing)1858 vlan_unconfig_locked(struct ifnet *ifp, int departing)
1859 {
1860 	struct ifvlantrunk *trunk;
1861 	struct vlan_mc_entry *mc;
1862 	struct ifvlan *ifv;
1863 	struct ifnet  *parent;
1864 	int error;
1865 
1866 	VLAN_XLOCK_ASSERT();
1867 
1868 	ifv = ifp->if_softc;
1869 	trunk = ifv->ifv_trunk;
1870 	parent = NULL;
1871 
1872 	if (trunk != NULL) {
1873 		parent = trunk->parent;
1874 
1875 		/*
1876 		 * Since the interface is being unconfigured, we need to
1877 		 * empty the list of multicast groups that we may have joined
1878 		 * while we were alive from the parent's list.
1879 		 */
1880 		while ((mc = CK_SLIST_FIRST(&ifv->vlan_mc_listhead)) != NULL) {
1881 			/*
1882 			 * If the parent interface is being detached,
1883 			 * all its multicast addresses have already
1884 			 * been removed.  Warn about errors if
1885 			 * if_delmulti() does fail, but don't abort as
1886 			 * all callers expect vlan destruction to
1887 			 * succeed.
1888 			 */
1889 			if (!departing) {
1890 				error = if_delmulti(parent,
1891 				    (struct sockaddr *)&mc->mc_addr);
1892 				if (error)
1893 					if_printf(ifp,
1894 		    "Failed to delete multicast address from parent: %d\n",
1895 					    error);
1896 			}
1897 			CK_SLIST_REMOVE_HEAD(&ifv->vlan_mc_listhead, mc_entries);
1898 			NET_EPOCH_CALL(vlan_mc_free, &mc->mc_epoch_ctx);
1899 		}
1900 
1901 		vlan_setflags(ifp, 0); /* clear special flags on parent */
1902 
1903 		vlan_remhash(trunk, ifv);
1904 		ifv->ifv_trunk = NULL;
1905 
1906 		/*
1907 		 * Check if we were the last.
1908 		 */
1909 		if (trunk->refcnt == 0) {
1910 			parent->if_vlantrunk = NULL;
1911 			NET_EPOCH_WAIT();
1912 			trunk_destroy(trunk);
1913 		}
1914 	}
1915 
1916 	/* Disconnect from parent. */
1917 	if (ifv->ifv_pflags)
1918 		if_printf(ifp, "%s: ifv_pflags unclean\n", __func__);
1919 	ifp->if_mtu = ETHERMTU;
1920 	ifp->if_link_state = LINK_STATE_UNKNOWN;
1921 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1922 
1923 	/*
1924 	 * Only dispatch an event if vlan was
1925 	 * attached, otherwise there is nothing
1926 	 * to cleanup anyway.
1927 	 */
1928 	if (parent != NULL)
1929 		EVENTHANDLER_INVOKE(vlan_unconfig, parent, ifv->ifv_vid);
1930 }
1931 
1932 /* Handle a reference counted flag that should be set on the parent as well */
1933 static int
vlan_setflag(struct ifnet * ifp,int flag,int status,int (* func)(struct ifnet *,int))1934 vlan_setflag(struct ifnet *ifp, int flag, int status,
1935 	     int (*func)(struct ifnet *, int))
1936 {
1937 	struct ifvlan *ifv;
1938 	int error;
1939 
1940 	VLAN_SXLOCK_ASSERT();
1941 
1942 	ifv = ifp->if_softc;
1943 	status = status ? (ifp->if_flags & flag) : 0;
1944 	/* Now "status" contains the flag value or 0 */
1945 
1946 	/*
1947 	 * See if recorded parent's status is different from what
1948 	 * we want it to be.  If it is, flip it.  We record parent's
1949 	 * status in ifv_pflags so that we won't clear parent's flag
1950 	 * we haven't set.  In fact, we don't clear or set parent's
1951 	 * flags directly, but get or release references to them.
1952 	 * That's why we can be sure that recorded flags still are
1953 	 * in accord with actual parent's flags.
1954 	 */
1955 	if (status != (ifv->ifv_pflags & flag)) {
1956 		error = (*func)(PARENT(ifv), status);
1957 		if (error)
1958 			return (error);
1959 		ifv->ifv_pflags &= ~flag;
1960 		ifv->ifv_pflags |= status;
1961 	}
1962 	return (0);
1963 }
1964 
1965 /*
1966  * Handle IFF_* flags that require certain changes on the parent:
1967  * if "status" is true, update parent's flags respective to our if_flags;
1968  * if "status" is false, forcedly clear the flags set on parent.
1969  */
1970 static int
vlan_setflags(struct ifnet * ifp,int status)1971 vlan_setflags(struct ifnet *ifp, int status)
1972 {
1973 	int error, i;
1974 
1975 	for (i = 0; vlan_pflags[i].flag; i++) {
1976 		error = vlan_setflag(ifp, vlan_pflags[i].flag,
1977 				     status, vlan_pflags[i].func);
1978 		if (error)
1979 			return (error);
1980 	}
1981 	return (0);
1982 }
1983 
1984 /* Inform all vlans that their parent has changed link state */
1985 static void
vlan_link_state(struct ifnet * ifp)1986 vlan_link_state(struct ifnet *ifp)
1987 {
1988 	struct epoch_tracker et;
1989 	struct ifvlantrunk *trunk;
1990 	struct ifvlan *ifv;
1991 
1992 	NET_EPOCH_ENTER(et);
1993 	trunk = ifp->if_vlantrunk;
1994 	if (trunk == NULL) {
1995 		NET_EPOCH_EXIT(et);
1996 		return;
1997 	}
1998 
1999 	TRUNK_WLOCK(trunk);
2000 	VLAN_FOREACH(ifv, trunk) {
2001 		ifv->ifv_ifp->if_baudrate = trunk->parent->if_baudrate;
2002 		if_link_state_change(ifv->ifv_ifp,
2003 		    trunk->parent->if_link_state);
2004 	}
2005 	TRUNK_WUNLOCK(trunk);
2006 	NET_EPOCH_EXIT(et);
2007 }
2008 
2009 #ifdef IPSEC_OFFLOAD
2010 #define	VLAN_IPSEC_METHOD(exp)				\
2011 	if_t p;						\
2012 	struct ifvlan *ifv;				\
2013 	int error;					\
2014 							\
2015 	ifv = ifp->if_softc;				\
2016 	VLAN_SLOCK();					\
2017 	if (TRUNK(ifv) != NULL) {			\
2018 		p = PARENT(ifv);			\
2019 		if_ref(p);				\
2020 		error = p->if_ipsec_accel_m->exp;	\
2021 		if_rele(p);				\
2022 	} else {					\
2023 		error = ENXIO;				\
2024 	}						\
2025 	VLAN_SUNLOCK();					\
2026 	return (error);
2027 
2028 
2029 static int
vlan_if_spdadd(if_t ifp,void * sp,void * inp,void ** priv)2030 vlan_if_spdadd(if_t ifp, void *sp, void *inp, void **priv)
2031 {
2032 	VLAN_IPSEC_METHOD(if_spdadd(ifp, sp, inp, priv));
2033 }
2034 
2035 static int
vlan_if_spddel(if_t ifp,void * sp,void * priv)2036 vlan_if_spddel(if_t ifp, void *sp, void *priv)
2037 {
2038 	VLAN_IPSEC_METHOD(if_spddel(ifp, sp, priv));
2039 }
2040 
2041 static int
vlan_if_sa_newkey(if_t ifp,void * sav,u_int drv_spi,void ** privp)2042 vlan_if_sa_newkey(if_t ifp, void *sav, u_int drv_spi, void **privp)
2043 {
2044 	VLAN_IPSEC_METHOD(if_sa_newkey(ifp, sav, drv_spi, privp));
2045 }
2046 
2047 static int
vlan_if_sa_deinstall(if_t ifp,u_int drv_spi,void * priv)2048 vlan_if_sa_deinstall(if_t ifp, u_int drv_spi, void *priv)
2049 {
2050 	VLAN_IPSEC_METHOD(if_sa_deinstall(ifp, drv_spi, priv));
2051 }
2052 
2053 static int
vlan_if_sa_cnt(if_t ifp,void * sa,uint32_t drv_spi,void * priv,struct seclifetime * lt)2054 vlan_if_sa_cnt(if_t ifp, void *sa, uint32_t drv_spi, void *priv,
2055     struct seclifetime *lt)
2056 {
2057 	VLAN_IPSEC_METHOD(if_sa_cnt(ifp, sa, drv_spi, priv, lt));
2058 }
2059 
2060 static int
vlan_if_ipsec_hwassist(if_t ifp,void * sav,u_int drv_spi,void * priv)2061 vlan_if_ipsec_hwassist(if_t ifp, void *sav, u_int drv_spi,void *priv)
2062 {
2063 	if_t trunk;
2064 
2065 	NET_EPOCH_ASSERT();
2066 	trunk = vlan_trunkdev(ifp);
2067 	if (trunk == NULL)
2068 		return (0);
2069 	return (trunk->if_ipsec_accel_m->if_hwassist(trunk, sav,
2070 	    drv_spi, priv));
2071 }
2072 
2073 static const struct if_ipsec_accel_methods vlan_if_ipsec_accel_methods = {
2074 	.if_spdadd = vlan_if_spdadd,
2075 	.if_spddel = vlan_if_spddel,
2076 	.if_sa_newkey = vlan_if_sa_newkey,
2077 	.if_sa_deinstall = vlan_if_sa_deinstall,
2078 	.if_sa_cnt = vlan_if_sa_cnt,
2079 	.if_hwassist = vlan_if_ipsec_hwassist,
2080 };
2081 
2082 #undef VLAN_IPSEC_METHOD
2083 #endif	/* IPSEC_OFFLOAD */
2084 
2085 static void
vlan_capabilities(struct ifvlan * ifv)2086 vlan_capabilities(struct ifvlan *ifv)
2087 {
2088 	struct ifnet *p;
2089 	struct ifnet *ifp;
2090 	struct ifnet_hw_tsomax hw_tsomax;
2091 	int cap = 0, ena = 0, mena, cap2 = 0, ena2 = 0;
2092 	int mena2 __unused;
2093 	u_long hwa = 0;
2094 
2095 	NET_EPOCH_ASSERT();
2096 	VLAN_SXLOCK_ASSERT();
2097 
2098 	p = PARENT(ifv);
2099 	ifp = ifv->ifv_ifp;
2100 
2101 	/* Mask parent interface enabled capabilities disabled by user. */
2102 	mena = p->if_capenable & ifv->ifv_capenable;
2103 	mena2 = p->if_capenable2 & ifv->ifv_capenable2;
2104 
2105 	/*
2106 	 * If the parent interface can do checksum offloading
2107 	 * on VLANs, then propagate its hardware-assisted
2108 	 * checksumming flags. Also assert that checksum
2109 	 * offloading requires hardware VLAN tagging.
2110 	 */
2111 	if (p->if_capabilities & IFCAP_VLAN_HWCSUM)
2112 		cap |= p->if_capabilities & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
2113 	if (p->if_capenable & IFCAP_VLAN_HWCSUM &&
2114 	    p->if_capenable & IFCAP_VLAN_HWTAGGING) {
2115 		ena |= mena & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
2116 		if (ena & IFCAP_TXCSUM)
2117 			hwa |= p->if_hwassist & (CSUM_IP | CSUM_TCP |
2118 			    CSUM_UDP | CSUM_SCTP);
2119 		if (ena & IFCAP_TXCSUM_IPV6)
2120 			hwa |= p->if_hwassist & (CSUM_TCP_IPV6 |
2121 			    CSUM_UDP_IPV6 | CSUM_SCTP_IPV6);
2122 	}
2123 
2124 	/*
2125 	 * If the parent interface can do TSO on VLANs then
2126 	 * propagate the hardware-assisted flag. TSO on VLANs
2127 	 * does not necessarily require hardware VLAN tagging.
2128 	 */
2129 	memset(&hw_tsomax, 0, sizeof(hw_tsomax));
2130 	if_hw_tsomax_common(p, &hw_tsomax);
2131 	if_hw_tsomax_update(ifp, &hw_tsomax);
2132 	if (p->if_capabilities & IFCAP_VLAN_HWTSO)
2133 		cap |= p->if_capabilities & IFCAP_TSO;
2134 	if (p->if_capenable & IFCAP_VLAN_HWTSO) {
2135 		ena |= mena & IFCAP_TSO;
2136 		if (ena & IFCAP_TSO)
2137 			hwa |= p->if_hwassist & CSUM_TSO;
2138 	}
2139 
2140 	/*
2141 	 * If the parent interface can do LRO and checksum offloading on
2142 	 * VLANs, then guess it may do LRO on VLANs.  False positive here
2143 	 * cost nothing, while false negative may lead to some confusions.
2144 	 */
2145 	if (p->if_capabilities & IFCAP_VLAN_HWCSUM)
2146 		cap |= p->if_capabilities & IFCAP_LRO;
2147 	if (p->if_capenable & IFCAP_VLAN_HWCSUM)
2148 		ena |= mena & IFCAP_LRO;
2149 
2150 	/*
2151 	 * If the parent interface can offload TCP connections over VLANs then
2152 	 * propagate its TOE capability to the VLAN interface.
2153 	 *
2154 	 * All TOE drivers in the tree today can deal with VLANs.  If this
2155 	 * changes then IFCAP_VLAN_TOE should be promoted to a full capability
2156 	 * with its own bit.
2157 	 */
2158 #define	IFCAP_VLAN_TOE IFCAP_TOE
2159 	if (p->if_capabilities & IFCAP_VLAN_TOE)
2160 		cap |= p->if_capabilities & IFCAP_TOE;
2161 	if (p->if_capenable & IFCAP_VLAN_TOE) {
2162 		SETTOEDEV(ifp, TOEDEV(p));
2163 		ena |= mena & IFCAP_TOE;
2164 	}
2165 
2166 	/*
2167 	 * If the parent interface supports dynamic link state, so does the
2168 	 * VLAN interface.
2169 	 */
2170 	cap |= (p->if_capabilities & IFCAP_LINKSTATE);
2171 	ena |= (mena & IFCAP_LINKSTATE);
2172 
2173 #ifdef RATELIMIT
2174 	/*
2175 	 * If the parent interface supports ratelimiting, so does the
2176 	 * VLAN interface.
2177 	 */
2178 	cap |= (p->if_capabilities & IFCAP_TXRTLMT);
2179 	ena |= (mena & IFCAP_TXRTLMT);
2180 #endif
2181 
2182 	/*
2183 	 * If the parent interface supports unmapped mbufs, so does
2184 	 * the VLAN interface.  Note that this should be fine even for
2185 	 * interfaces that don't support hardware tagging as headers
2186 	 * are prepended in normal mbufs to unmapped mbufs holding
2187 	 * payload data.
2188 	 */
2189 	cap |= (p->if_capabilities & IFCAP_MEXTPG);
2190 	ena |= (mena & IFCAP_MEXTPG);
2191 
2192 	/*
2193 	 * If the parent interface can offload encryption and segmentation
2194 	 * of TLS records over TCP, propagate it's capability to the VLAN
2195 	 * interface.
2196 	 *
2197 	 * All TLS drivers in the tree today can deal with VLANs.  If
2198 	 * this ever changes, then a new IFCAP_VLAN_TXTLS can be
2199 	 * defined.
2200 	 */
2201 	if (p->if_capabilities & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT))
2202 		cap |= p->if_capabilities & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT);
2203 	if (p->if_capenable & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT))
2204 		ena |= mena & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT);
2205 
2206 	ifp->if_capabilities = cap;
2207 	ifp->if_capenable = ena;
2208 	ifp->if_hwassist = hwa;
2209 
2210 #ifdef IPSEC_OFFLOAD
2211 	cap2 |= p->if_capabilities2 & IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD);
2212 	ena2 |= mena2 & IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD);
2213 	ifp->if_ipsec_accel_m = &vlan_if_ipsec_accel_methods;
2214 #endif
2215 
2216 	ifp->if_capabilities2 = cap2;
2217 	ifp->if_capenable2 = ena2;
2218 }
2219 
2220 static void
vlan_trunk_capabilities(struct ifnet * ifp)2221 vlan_trunk_capabilities(struct ifnet *ifp)
2222 {
2223 	struct epoch_tracker et;
2224 	struct ifvlantrunk *trunk;
2225 	struct ifvlan *ifv;
2226 
2227 	VLAN_SLOCK();
2228 	trunk = ifp->if_vlantrunk;
2229 	if (trunk == NULL) {
2230 		VLAN_SUNLOCK();
2231 		return;
2232 	}
2233 	NET_EPOCH_ENTER(et);
2234 	VLAN_FOREACH(ifv, trunk)
2235 		vlan_capabilities(ifv);
2236 	NET_EPOCH_EXIT(et);
2237 	VLAN_SUNLOCK();
2238 }
2239 
2240 static int
vlan_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)2241 vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
2242 {
2243 	struct ifnet *p;
2244 	struct ifreq *ifr;
2245 #ifdef INET
2246 	struct ifaddr *ifa;
2247 #endif
2248 	struct ifvlan *ifv;
2249 	struct ifvlantrunk *trunk;
2250 	struct vlanreq vlr;
2251 	int error = 0, oldmtu;
2252 
2253 	ifr = (struct ifreq *)data;
2254 #ifdef INET
2255 	ifa = (struct ifaddr *) data;
2256 #endif
2257 	ifv = ifp->if_softc;
2258 
2259 	switch (cmd) {
2260 	case SIOCSIFADDR:
2261 		ifp->if_flags |= IFF_UP;
2262 #ifdef INET
2263 		if (ifa->ifa_addr->sa_family == AF_INET)
2264 			arp_ifinit(ifp, ifa);
2265 #endif
2266 		break;
2267 	case SIOCGIFADDR:
2268 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
2269 		    ifp->if_addrlen);
2270 		break;
2271 	case SIOCGIFMEDIA:
2272 		VLAN_SLOCK();
2273 		if (TRUNK(ifv) != NULL) {
2274 			p = PARENT(ifv);
2275 			if_ref(p);
2276 			error = (*p->if_ioctl)(p, SIOCGIFMEDIA, data);
2277 			if_rele(p);
2278 			/* Limit the result to the parent's current config. */
2279 			if (error == 0) {
2280 				struct ifmediareq *ifmr;
2281 
2282 				ifmr = (struct ifmediareq *)data;
2283 				if (ifmr->ifm_count >= 1 && ifmr->ifm_ulist) {
2284 					ifmr->ifm_count = 1;
2285 					error = copyout(&ifmr->ifm_current,
2286 						ifmr->ifm_ulist,
2287 						sizeof(int));
2288 				}
2289 			}
2290 		} else {
2291 			error = EINVAL;
2292 		}
2293 		VLAN_SUNLOCK();
2294 		break;
2295 
2296 	case SIOCSIFMEDIA:
2297 		error = EINVAL;
2298 		break;
2299 
2300 	case SIOCSIFMTU:
2301 		/*
2302 		 * Set the interface MTU.
2303 		 */
2304 		VLAN_SLOCK();
2305 		trunk = TRUNK(ifv);
2306 		if (trunk != NULL) {
2307 			TRUNK_WLOCK(trunk);
2308 			if (ifr->ifr_mtu >
2309 			     (PARENT(ifv)->if_mtu - ifv->ifv_mtufudge) ||
2310 			    ifr->ifr_mtu <
2311 			     (ifv->ifv_mintu - ifv->ifv_mtufudge))
2312 				error = EINVAL;
2313 			else
2314 				ifp->if_mtu = ifr->ifr_mtu;
2315 			TRUNK_WUNLOCK(trunk);
2316 		} else
2317 			error = EINVAL;
2318 		VLAN_SUNLOCK();
2319 		break;
2320 
2321 	case SIOCSETVLAN:
2322 #ifdef VIMAGE
2323 		/*
2324 		 * XXXRW/XXXBZ: The goal in these checks is to allow a VLAN
2325 		 * interface to be delegated to a jail without allowing the
2326 		 * jail to change what underlying interface/VID it is
2327 		 * associated with.  We are not entirely convinced that this
2328 		 * is the right way to accomplish that policy goal.
2329 		 */
2330 		if (ifp->if_vnet != ifp->if_home_vnet) {
2331 			error = EPERM;
2332 			break;
2333 		}
2334 #endif
2335 		error = copyin(ifr_data_get_ptr(ifr), &vlr, sizeof(vlr));
2336 		if (error)
2337 			break;
2338 		if (vlr.vlr_parent[0] == '\0') {
2339 			vlan_unconfig(ifp);
2340 			break;
2341 		}
2342 		p = ifunit_ref(vlr.vlr_parent);
2343 		if (p == NULL) {
2344 			error = ENOENT;
2345 			break;
2346 		}
2347 		if (vlr.vlr_proto == 0)
2348 			vlr.vlr_proto = ETHERTYPE_VLAN;
2349 		oldmtu = ifp->if_mtu;
2350 		error = vlan_config(ifv, p, vlr.vlr_tag, vlr.vlr_proto);
2351 		if_rele(p);
2352 
2353 		/*
2354 		 * VLAN MTU may change during addition of the vlandev.
2355 		 * If it did, do network layer specific procedure.
2356 		 */
2357 		if (ifp->if_mtu != oldmtu)
2358 			if_notifymtu(ifp);
2359 		break;
2360 
2361 	case SIOCGETVLAN:
2362 #ifdef VIMAGE
2363 		if (ifp->if_vnet != ifp->if_home_vnet) {
2364 			error = EPERM;
2365 			break;
2366 		}
2367 #endif
2368 		bzero(&vlr, sizeof(vlr));
2369 		VLAN_SLOCK();
2370 		if (TRUNK(ifv) != NULL) {
2371 			strlcpy(vlr.vlr_parent, PARENT(ifv)->if_xname,
2372 			    sizeof(vlr.vlr_parent));
2373 			vlr.vlr_tag = ifv->ifv_vid;
2374 			vlr.vlr_proto = ifv->ifv_proto;
2375 		}
2376 		VLAN_SUNLOCK();
2377 		error = copyout(&vlr, ifr_data_get_ptr(ifr), sizeof(vlr));
2378 		break;
2379 
2380 	case SIOCSIFFLAGS:
2381 		/*
2382 		 * We should propagate selected flags to the parent,
2383 		 * e.g., promiscuous mode.
2384 		 */
2385 		VLAN_SLOCK();
2386 		if (TRUNK(ifv) != NULL)
2387 			error = vlan_setflags(ifp, 1);
2388 		VLAN_SUNLOCK();
2389 		break;
2390 
2391 	case SIOCADDMULTI:
2392 	case SIOCDELMULTI:
2393 		/*
2394 		 * If we don't have a parent, just remember the membership for
2395 		 * when we do.
2396 		 *
2397 		 * XXX We need the rmlock here to avoid sleeping while
2398 		 * holding in6_multi_mtx.
2399 		 */
2400 		VLAN_XLOCK();
2401 		trunk = TRUNK(ifv);
2402 		if (trunk != NULL)
2403 			error = vlan_setmulti(ifp);
2404 		VLAN_XUNLOCK();
2405 
2406 		break;
2407 	case SIOCGVLANPCP:
2408 #ifdef VIMAGE
2409 		if (ifp->if_vnet != ifp->if_home_vnet) {
2410 			error = EPERM;
2411 			break;
2412 		}
2413 #endif
2414 		ifr->ifr_vlan_pcp = ifv->ifv_pcp;
2415 		break;
2416 
2417 	case SIOCSVLANPCP:
2418 #ifdef VIMAGE
2419 		if (ifp->if_vnet != ifp->if_home_vnet) {
2420 			error = EPERM;
2421 			break;
2422 		}
2423 #endif
2424 		error = priv_check(curthread, PRIV_NET_SETVLANPCP);
2425 		if (error)
2426 			break;
2427 		if (ifr->ifr_vlan_pcp > VLAN_PCP_MAX) {
2428 			error = EINVAL;
2429 			break;
2430 		}
2431 		ifv->ifv_pcp = ifr->ifr_vlan_pcp;
2432 		ifp->if_pcp = ifv->ifv_pcp;
2433 		/* broadcast event about PCP change */
2434 		EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
2435 		break;
2436 
2437 	case SIOCSIFCAP:
2438 		VLAN_SLOCK();
2439 		ifv->ifv_capenable = ifr->ifr_reqcap;
2440 		trunk = TRUNK(ifv);
2441 		if (trunk != NULL) {
2442 			struct epoch_tracker et;
2443 
2444 			NET_EPOCH_ENTER(et);
2445 			vlan_capabilities(ifv);
2446 			NET_EPOCH_EXIT(et);
2447 		}
2448 		VLAN_SUNLOCK();
2449 		break;
2450 
2451 	default:
2452 		error = EINVAL;
2453 		break;
2454 	}
2455 
2456 	return (error);
2457 }
2458 
2459 #if defined(KERN_TLS) || defined(RATELIMIT)
2460 static int
vlan_snd_tag_alloc(struct ifnet * ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** ppmt)2461 vlan_snd_tag_alloc(struct ifnet *ifp,
2462     union if_snd_tag_alloc_params *params,
2463     struct m_snd_tag **ppmt)
2464 {
2465 	struct epoch_tracker et;
2466 	const struct if_snd_tag_sw *sw;
2467 	struct vlan_snd_tag *vst;
2468 	struct ifvlan *ifv;
2469 	struct ifnet *parent;
2470 	struct m_snd_tag *mst;
2471 	int error;
2472 
2473 	NET_EPOCH_ENTER(et);
2474 	ifv = ifp->if_softc;
2475 
2476 	switch (params->hdr.type) {
2477 #ifdef RATELIMIT
2478 	case IF_SND_TAG_TYPE_UNLIMITED:
2479 		sw = &vlan_snd_tag_ul_sw;
2480 		break;
2481 	case IF_SND_TAG_TYPE_RATE_LIMIT:
2482 		sw = &vlan_snd_tag_rl_sw;
2483 		break;
2484 #endif
2485 #ifdef KERN_TLS
2486 	case IF_SND_TAG_TYPE_TLS:
2487 		sw = &vlan_snd_tag_tls_sw;
2488 		break;
2489 	case IF_SND_TAG_TYPE_TLS_RX:
2490 		sw = NULL;
2491 		if (params->tls_rx.vlan_id != 0)
2492 			goto failure;
2493 		params->tls_rx.vlan_id = ifv->ifv_vid;
2494 		break;
2495 #ifdef RATELIMIT
2496 	case IF_SND_TAG_TYPE_TLS_RATE_LIMIT:
2497 		sw = &vlan_snd_tag_tls_rl_sw;
2498 		break;
2499 #endif
2500 #endif
2501 	default:
2502 		goto failure;
2503 	}
2504 
2505 	if (ifv->ifv_trunk != NULL)
2506 		parent = PARENT(ifv);
2507 	else
2508 		parent = NULL;
2509 	if (parent == NULL)
2510 		goto failure;
2511 	if_ref(parent);
2512 	NET_EPOCH_EXIT(et);
2513 
2514 	if (sw != NULL) {
2515 		vst = malloc(sizeof(*vst), M_VLAN, M_NOWAIT);
2516 		if (vst == NULL) {
2517 			if_rele(parent);
2518 			return (ENOMEM);
2519 		}
2520 	} else
2521 		vst = NULL;
2522 
2523 	error = m_snd_tag_alloc(parent, params, &mst);
2524 	if_rele(parent);
2525 	if (error) {
2526 		free(vst, M_VLAN);
2527 		return (error);
2528 	}
2529 
2530 	if (sw != NULL) {
2531 		m_snd_tag_init(&vst->com, ifp, sw);
2532 		vst->tag = mst;
2533 
2534 		*ppmt = &vst->com;
2535 	} else
2536 		*ppmt = mst;
2537 
2538 	return (0);
2539 failure:
2540 	NET_EPOCH_EXIT(et);
2541 	return (EOPNOTSUPP);
2542 }
2543 
2544 static struct m_snd_tag *
vlan_next_snd_tag(struct m_snd_tag * mst)2545 vlan_next_snd_tag(struct m_snd_tag *mst)
2546 {
2547 	struct vlan_snd_tag *vst;
2548 
2549 	vst = mst_to_vst(mst);
2550 	return (vst->tag);
2551 }
2552 
2553 static int
vlan_snd_tag_modify(struct m_snd_tag * mst,union if_snd_tag_modify_params * params)2554 vlan_snd_tag_modify(struct m_snd_tag *mst,
2555     union if_snd_tag_modify_params *params)
2556 {
2557 	struct vlan_snd_tag *vst;
2558 
2559 	vst = mst_to_vst(mst);
2560 	return (vst->tag->sw->snd_tag_modify(vst->tag, params));
2561 }
2562 
2563 static int
vlan_snd_tag_query(struct m_snd_tag * mst,union if_snd_tag_query_params * params)2564 vlan_snd_tag_query(struct m_snd_tag *mst,
2565     union if_snd_tag_query_params *params)
2566 {
2567 	struct vlan_snd_tag *vst;
2568 
2569 	vst = mst_to_vst(mst);
2570 	return (vst->tag->sw->snd_tag_query(vst->tag, params));
2571 }
2572 
2573 static void
vlan_snd_tag_free(struct m_snd_tag * mst)2574 vlan_snd_tag_free(struct m_snd_tag *mst)
2575 {
2576 	struct vlan_snd_tag *vst;
2577 
2578 	vst = mst_to_vst(mst);
2579 	m_snd_tag_rele(vst->tag);
2580 	free(vst, M_VLAN);
2581 }
2582 
2583 static void
vlan_ratelimit_query(struct ifnet * ifp __unused,struct if_ratelimit_query_results * q)2584 vlan_ratelimit_query(struct ifnet *ifp __unused, struct if_ratelimit_query_results *q)
2585 {
2586 	/*
2587 	 * For vlan, we have an indirect
2588 	 * interface. The caller needs to
2589 	 * get a ratelimit tag on the actual
2590 	 * interface the flow will go on.
2591 	 */
2592 	q->rate_table = NULL;
2593 	q->flags = RT_IS_INDIRECT;
2594 	q->max_flows = 0;
2595 	q->number_of_rates = 0;
2596 }
2597 
2598 #endif
2599