xref: /freebsd/sys/net/if_vlan.c (revision a19b353d354d4ef808965c53253103cb6e7e6708)
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 /* For if_link_state_change() eyes only... */
892 extern	void (*vlan_link_state_p)(struct ifnet *);
893 
894 static struct if_clone_addreq_v2 vlan_addreq = {
895 	.version = 2,
896 	.match_f = vlan_clone_match,
897 	.create_f = vlan_clone_create,
898 	.destroy_f = vlan_clone_destroy,
899 	.create_nl_f = vlan_clone_create_nl,
900 	.modify_nl_f = vlan_clone_modify_nl,
901 	.dump_nl_f = vlan_clone_dump_nl,
902 };
903 
904 static int
vlan_modevent(module_t mod,int type,void * data)905 vlan_modevent(module_t mod, int type, void *data)
906 {
907 
908 	switch (type) {
909 	case MOD_LOAD:
910 		ifdetach_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
911 		    vlan_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
912 		if (ifdetach_tag == NULL)
913 			return (ENOMEM);
914 		iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event,
915 		    vlan_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
916 		if (iflladdr_tag == NULL)
917 			return (ENOMEM);
918 		ifevent_tag = EVENTHANDLER_REGISTER(ifnet_event,
919 		    vlan_ifevent, NULL, EVENTHANDLER_PRI_ANY);
920 		if (ifevent_tag == NULL)
921 			return (ENOMEM);
922 		VLAN_LOCKING_INIT();
923 		vlan_input_p = vlan_input;
924 		vlan_link_state_p = vlan_link_state;
925 		vlan_trunk_cap_p = vlan_trunk_capabilities;
926 		vlan_trunkdev_p = vlan_trunkdev;
927 		vlan_cookie_p = vlan_cookie;
928 		vlan_setcookie_p = vlan_setcookie;
929 		vlan_tag_p = vlan_tag;
930 		vlan_pcp_p = vlan_pcp;
931 		vlan_devat_p = vlan_devat;
932 #ifndef VIMAGE
933 		vlan_cloner = ifc_attach_cloner(vlanname, (struct if_clone_addreq *)&vlan_addreq);
934 #endif
935 		if (bootverbose)
936 			printf("vlan: initialized, using "
937 #ifdef VLAN_ARRAY
938 			       "full-size arrays"
939 #else
940 			       "hash tables with chaining"
941 #endif
942 
943 			       "\n");
944 		break;
945 	case MOD_UNLOAD:
946 #ifndef VIMAGE
947 		ifc_detach_cloner(vlan_cloner);
948 #endif
949 		EVENTHANDLER_DEREGISTER(ifnet_departure_event, ifdetach_tag);
950 		EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_tag);
951 		EVENTHANDLER_DEREGISTER(ifnet_event, ifevent_tag);
952 		vlan_input_p = NULL;
953 		vlan_link_state_p = NULL;
954 		vlan_trunk_cap_p = NULL;
955 		vlan_trunkdev_p = NULL;
956 		vlan_tag_p = NULL;
957 		vlan_cookie_p = NULL;
958 		vlan_setcookie_p = NULL;
959 		vlan_devat_p = NULL;
960 		VLAN_LOCKING_DESTROY();
961 		if (bootverbose)
962 			printf("vlan: unloaded\n");
963 		break;
964 	default:
965 		return (EOPNOTSUPP);
966 	}
967 	return (0);
968 }
969 
970 static moduledata_t vlan_mod = {
971 	"if_vlan",
972 	vlan_modevent,
973 	0
974 };
975 
976 DECLARE_MODULE(if_vlan, vlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
977 MODULE_VERSION(if_vlan, 3);
978 
979 #ifdef VIMAGE
980 static void
vnet_vlan_init(const void * unused __unused)981 vnet_vlan_init(const void *unused __unused)
982 {
983 	vlan_cloner = ifc_attach_cloner(vlanname, (struct if_clone_addreq *)&vlan_addreq);
984 	V_vlan_cloner = vlan_cloner;
985 }
986 VNET_SYSINIT(vnet_vlan_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
987     vnet_vlan_init, NULL);
988 
989 static void
vnet_vlan_uninit(const void * unused __unused)990 vnet_vlan_uninit(const void *unused __unused)
991 {
992 
993 	ifc_detach_cloner(V_vlan_cloner);
994 }
995 VNET_SYSUNINIT(vnet_vlan_uninit, SI_SUB_INIT_IF, SI_ORDER_ANY,
996     vnet_vlan_uninit, NULL);
997 #endif
998 
999 /*
1000  * Check for <etherif>.<vlan>[.<vlan> ...] style interface names.
1001  */
1002 static struct ifnet *
vlan_clone_match_ethervid(const char * name,int * vidp)1003 vlan_clone_match_ethervid(const char *name, int *vidp)
1004 {
1005 	char ifname[IFNAMSIZ];
1006 	char *cp;
1007 	struct ifnet *ifp;
1008 	int vid;
1009 
1010 	strlcpy(ifname, name, IFNAMSIZ);
1011 	if ((cp = strrchr(ifname, '.')) == NULL)
1012 		return (NULL);
1013 	*cp = '\0';
1014 	if ((ifp = ifunit_ref(ifname)) == NULL)
1015 		return (NULL);
1016 	/* Parse VID. */
1017 	if (*++cp == '\0') {
1018 		if_rele(ifp);
1019 		return (NULL);
1020 	}
1021 	vid = 0;
1022 	for(; *cp >= '0' && *cp <= '9'; cp++)
1023 		vid = (vid * 10) + (*cp - '0');
1024 	if (*cp != '\0') {
1025 		if_rele(ifp);
1026 		return (NULL);
1027 	}
1028 	if (vidp != NULL)
1029 		*vidp = vid;
1030 
1031 	return (ifp);
1032 }
1033 
1034 static int
vlan_clone_match(struct if_clone * ifc,const char * name)1035 vlan_clone_match(struct if_clone *ifc, const char *name)
1036 {
1037 	struct ifnet *ifp;
1038 	const char *cp;
1039 
1040 	ifp = vlan_clone_match_ethervid(name, NULL);
1041 	if (ifp != NULL) {
1042 		if_rele(ifp);
1043 		return (1);
1044 	}
1045 
1046 	if (strncmp(vlanname, name, strlen(vlanname)) != 0)
1047 		return (0);
1048 	for (cp = name + 4; *cp != '\0'; cp++) {
1049 		if (*cp < '0' || *cp > '9')
1050 			return (0);
1051 	}
1052 
1053 	return (1);
1054 }
1055 
1056 static int
vlan_clone_create(struct if_clone * ifc,char * name,size_t len,struct ifc_data * ifd,struct ifnet ** ifpp)1057 vlan_clone_create(struct if_clone *ifc, char *name, size_t len,
1058     struct ifc_data *ifd, struct ifnet **ifpp)
1059 {
1060 	char *dp;
1061 	bool wildcard = false;
1062 	bool subinterface = false;
1063 	int unit;
1064 	int error;
1065 	int vid = 0;
1066 	uint16_t proto = ETHERTYPE_VLAN;
1067 	struct ifvlan *ifv;
1068 	struct ifnet *ifp;
1069 	struct ifnet *p = NULL;
1070 	struct ifaddr *ifa;
1071 	struct sockaddr_dl *sdl;
1072 	struct vlanreq vlr;
1073 	static const u_char eaddr[ETHER_ADDR_LEN];	/* 00:00:00:00:00:00 */
1074 
1075 
1076 	/*
1077 	 * There are three ways to specify the cloned device:
1078 	 * o pass a parameter block with the clone request.
1079 	 * o specify parameters in the text of the clone device name
1080 	 * o specify no parameters and get an unattached device that
1081 	 *   must be configured separately.
1082 	 * The first technique is preferred; the latter two are supported
1083 	 * for backwards compatibility.
1084 	 *
1085 	 * XXXRW: Note historic use of the word "tag" here.  New ioctls may be
1086 	 * called for.
1087 	 */
1088 
1089 	if (ifd->params != NULL) {
1090 		error = ifc_copyin(ifd, &vlr, sizeof(vlr));
1091 		if (error)
1092 			return error;
1093 		vid = vlr.vlr_tag;
1094 		proto = vlr.vlr_proto;
1095 		if (proto == 0)
1096 			proto = ETHERTYPE_VLAN;
1097 		p = ifunit_ref(vlr.vlr_parent);
1098 		if (p == NULL)
1099 			return (ENXIO);
1100 	}
1101 
1102 	if ((error = ifc_name2unit(name, &unit)) == 0) {
1103 
1104 		/*
1105 		 * vlanX interface. Set wildcard to true if the unit number
1106 		 * is not fixed (-1)
1107 		 */
1108 		wildcard = (unit < 0);
1109 	} else {
1110 		struct ifnet *p_tmp = vlan_clone_match_ethervid(name, &vid);
1111 		if (p_tmp != NULL) {
1112 			error = 0;
1113 			subinterface = true;
1114 			unit = IF_DUNIT_NONE;
1115 			wildcard = false;
1116 			if (p != NULL) {
1117 				if_rele(p_tmp);
1118 				if (p != p_tmp)
1119 					error = EINVAL;
1120 			} else
1121 				p = p_tmp;
1122 		} else
1123 			error = ENXIO;
1124 	}
1125 
1126 	if (error != 0) {
1127 		if (p != NULL)
1128 			if_rele(p);
1129 		return (error);
1130 	}
1131 
1132 	if (!subinterface) {
1133 		/* vlanX interface, mark X as busy or allocate new unit # */
1134 		error = ifc_alloc_unit(ifc, &unit);
1135 		if (error != 0) {
1136 			if (p != NULL)
1137 				if_rele(p);
1138 			return (error);
1139 		}
1140 	}
1141 
1142 	/* In the wildcard case, we need to update the name. */
1143 	if (wildcard) {
1144 		for (dp = name; *dp != '\0'; dp++);
1145 		if (snprintf(dp, len - (dp-name), "%d", unit) >
1146 		    len - (dp-name) - 1) {
1147 			panic("%s: interface name too long", __func__);
1148 		}
1149 	}
1150 
1151 	ifv = malloc(sizeof(struct ifvlan), M_VLAN, M_WAITOK | M_ZERO);
1152 	ifp = ifv->ifv_ifp = if_alloc(IFT_ETHER);
1153 	CK_SLIST_INIT(&ifv->vlan_mc_listhead);
1154 	ifp->if_softc = ifv;
1155 	/*
1156 	 * Set the name manually rather than using if_initname because
1157 	 * we don't conform to the default naming convention for interfaces.
1158 	 */
1159 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
1160 	ifp->if_dname = vlanname;
1161 	ifp->if_dunit = unit;
1162 
1163 	ifp->if_init = vlan_init;
1164 #ifdef ALTQ
1165 	ifp->if_start = vlan_altq_start;
1166 	ifp->if_transmit = vlan_altq_transmit;
1167 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
1168 	ifp->if_snd.ifq_drv_maxlen = 0;
1169 	IFQ_SET_READY(&ifp->if_snd);
1170 #else
1171 	ifp->if_transmit = vlan_transmit;
1172 #endif
1173 	ifp->if_qflush = vlan_qflush;
1174 	ifp->if_ioctl = vlan_ioctl;
1175 #if defined(KERN_TLS) || defined(RATELIMIT)
1176 	ifp->if_snd_tag_alloc = vlan_snd_tag_alloc;
1177 	ifp->if_ratelimit_query = vlan_ratelimit_query;
1178 #endif
1179 	ifp->if_flags = VLAN_IFFLAGS;
1180 	ifp->if_type = IFT_L2VLAN;
1181 	ether_ifattach(ifp, eaddr);
1182 	/* Now undo some of the damage... */
1183 	ifp->if_baudrate = 0;
1184 	ifp->if_hdrlen = ETHER_VLAN_ENCAP_LEN;
1185 	ifa = ifp->if_addr;
1186 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1187 	sdl->sdl_type = IFT_L2VLAN;
1188 
1189 	if (p != NULL) {
1190 		error = vlan_config(ifv, p, vid, proto);
1191 		if_rele(p);
1192 		if (error != 0) {
1193 			/*
1194 			 * Since we've partially failed, we need to back
1195 			 * out all the way, otherwise userland could get
1196 			 * confused.  Thus, we destroy the interface.
1197 			 */
1198 			ether_ifdetach(ifp);
1199 			vlan_unconfig(ifp);
1200 			if_free(ifp);
1201 			if (!subinterface)
1202 				ifc_free_unit(ifc, unit);
1203 			free(ifv, M_VLAN);
1204 
1205 			return (error);
1206 		}
1207 	}
1208 	*ifpp = ifp;
1209 
1210 	return (0);
1211 }
1212 
1213 /*
1214  *
1215  * Parsers of IFLA_INFO_DATA inside IFLA_LINKINFO of RTM_NEWLINK
1216  *    {{nla_len=8, nla_type=IFLA_LINK}, 2},
1217  *    {{nla_len=12, nla_type=IFLA_IFNAME}, "xvlan22"},
1218  *    {{nla_len=24, nla_type=IFLA_LINKINFO},
1219  *     [
1220  *      {{nla_len=8, nla_type=IFLA_INFO_KIND}, "vlan"...},
1221  *      {{nla_len=12, nla_type=IFLA_INFO_DATA}, "\x06\x00\x01\x00\x16\x00\x00\x00"}]}
1222  */
1223 
1224 struct nl_parsed_vlan {
1225 	uint16_t vlan_id;
1226 	uint16_t vlan_proto;
1227 	struct ifla_vlan_flags vlan_flags;
1228 };
1229 
1230 #define	_OUT(_field)	offsetof(struct nl_parsed_vlan, _field)
1231 static const struct nlattr_parser nla_p_vlan[] = {
1232 	{ .type = IFLA_VLAN_ID, .off = _OUT(vlan_id), .cb = nlattr_get_uint16 },
1233 	{ .type = IFLA_VLAN_FLAGS, .off = _OUT(vlan_flags), .cb = nlattr_get_nla },
1234 	{ .type = IFLA_VLAN_PROTOCOL, .off = _OUT(vlan_proto), .cb = nlattr_get_uint16 },
1235 };
1236 #undef _OUT
1237 NL_DECLARE_ATTR_PARSER(vlan_parser, nla_p_vlan);
1238 
1239 static int
vlan_clone_create_nl(struct if_clone * ifc,char * name,size_t len,struct ifc_data_nl * ifd)1240 vlan_clone_create_nl(struct if_clone *ifc, char *name, size_t len,
1241     struct ifc_data_nl *ifd)
1242 {
1243 	struct epoch_tracker et;
1244         struct ifnet *ifp_parent;
1245 	struct nl_pstate *npt = ifd->npt;
1246 	struct nl_parsed_link *lattrs = ifd->lattrs;
1247 	int error;
1248 
1249 	/*
1250 	 * lattrs.ifla_ifname is the new interface name
1251 	 * lattrs.ifi_index contains parent interface index
1252 	 * lattrs.ifla_idata contains un-parsed vlan data
1253 	 */
1254 	struct nl_parsed_vlan attrs = {
1255 		.vlan_id = 0xFEFE,
1256 		.vlan_proto = ETHERTYPE_VLAN
1257 	};
1258 
1259 	if (lattrs->ifla_idata == NULL) {
1260 		nlmsg_report_err_msg(npt, "vlan id is required, guessing not supported");
1261 		return (ENOTSUP);
1262 	}
1263 
1264 	error = nl_parse_nested(lattrs->ifla_idata, &vlan_parser, npt, &attrs);
1265 	if (error != 0)
1266 		return (error);
1267 	if (attrs.vlan_id > DOT1Q_VID_MAX) {
1268 		nlmsg_report_err_msg(npt, "Invalid VID: %d", attrs.vlan_id);
1269 		return (EINVAL);
1270 	}
1271 	if (attrs.vlan_proto != ETHERTYPE_VLAN && attrs.vlan_proto != ETHERTYPE_QINQ) {
1272 		nlmsg_report_err_msg(npt, "Unsupported ethertype: 0x%04X", attrs.vlan_proto);
1273 		return (ENOTSUP);
1274 	}
1275 
1276 	struct vlanreq params = {
1277 		.vlr_tag = attrs.vlan_id,
1278 		.vlr_proto = attrs.vlan_proto,
1279 	};
1280 	struct ifc_data ifd_new = { .flags = IFC_F_SYSSPACE, .unit = ifd->unit, .params = &params };
1281 
1282 	NET_EPOCH_ENTER(et);
1283 	ifp_parent = ifnet_byindex(lattrs->ifi_index);
1284 	if (ifp_parent != NULL)
1285 		strlcpy(params.vlr_parent, if_name(ifp_parent), sizeof(params.vlr_parent));
1286 	NET_EPOCH_EXIT(et);
1287 
1288 	if (ifp_parent == NULL) {
1289 		nlmsg_report_err_msg(npt, "unable to find parent interface %u", lattrs->ifi_index);
1290 		return (ENOENT);
1291 	}
1292 
1293 	error = vlan_clone_create(ifc, name, len, &ifd_new, &ifd->ifp);
1294 
1295 	return (error);
1296 }
1297 
1298 static int
vlan_clone_modify_nl(struct ifnet * ifp,struct ifc_data_nl * ifd)1299 vlan_clone_modify_nl(struct ifnet *ifp, struct ifc_data_nl *ifd)
1300 {
1301 	struct nl_parsed_link *lattrs = ifd->lattrs;
1302 
1303 	if ((lattrs->ifla_idata != NULL) && ((ifd->flags & IFC_F_CREATE) == 0)) {
1304 		struct epoch_tracker et;
1305 		struct nl_parsed_vlan attrs = {
1306 			.vlan_proto = ETHERTYPE_VLAN,
1307 		};
1308 		int error;
1309 
1310 		error = nl_parse_nested(lattrs->ifla_idata, &vlan_parser, ifd->npt, &attrs);
1311 		if (error != 0)
1312 			return (error);
1313 
1314 		NET_EPOCH_ENTER(et);
1315 		struct ifnet *ifp_parent = ifnet_byindex_ref(lattrs->ifla_link);
1316 		NET_EPOCH_EXIT(et);
1317 
1318 		if (ifp_parent == NULL) {
1319 			nlmsg_report_err_msg(ifd->npt, "unable to find parent interface %u",
1320 			    lattrs->ifla_link);
1321 			return (ENOENT);
1322 		}
1323 
1324 		struct ifvlan *ifv = ifp->if_softc;
1325 		error = vlan_config(ifv, ifp_parent, attrs.vlan_id, attrs.vlan_proto);
1326 
1327 		if_rele(ifp_parent);
1328 		if (error != 0)
1329 			return (error);
1330 	}
1331 
1332 	return (nl_modify_ifp_generic(ifp, ifd->lattrs, ifd->bm, ifd->npt));
1333 }
1334 
1335 /*
1336  *    {{nla_len=24, nla_type=IFLA_LINKINFO},
1337  *     [
1338  *      {{nla_len=8, nla_type=IFLA_INFO_KIND}, "vlan"...},
1339  *      {{nla_len=12, nla_type=IFLA_INFO_DATA}, "\x06\x00\x01\x00\x16\x00\x00\x00"}]}
1340  */
1341 static void
vlan_clone_dump_nl(struct ifnet * ifp,struct nl_writer * nw)1342 vlan_clone_dump_nl(struct ifnet *ifp, struct nl_writer *nw)
1343 {
1344 	uint32_t parent_index = 0;
1345 	uint16_t vlan_id = 0;
1346 	uint16_t vlan_proto = 0;
1347 
1348 	VLAN_SLOCK();
1349 	struct ifvlan *ifv = ifp->if_softc;
1350 	if (TRUNK(ifv) != NULL)
1351 		parent_index = PARENT(ifv)->if_index;
1352 	vlan_id = ifv->ifv_vid;
1353 	vlan_proto = ifv->ifv_proto;
1354 	VLAN_SUNLOCK();
1355 
1356 	if (parent_index != 0)
1357 		nlattr_add_u32(nw, IFLA_LINK, parent_index);
1358 
1359 	int off = nlattr_add_nested(nw, IFLA_LINKINFO);
1360 	if (off != 0) {
1361 		nlattr_add_string(nw, IFLA_INFO_KIND, "vlan");
1362 		int off2 = nlattr_add_nested(nw, IFLA_INFO_DATA);
1363 		if (off2 != 0) {
1364 			nlattr_add_u16(nw, IFLA_VLAN_ID, vlan_id);
1365 			nlattr_add_u16(nw, IFLA_VLAN_PROTOCOL, vlan_proto);
1366 			nlattr_set_len(nw, off2);
1367 		}
1368 		nlattr_set_len(nw, off);
1369 	}
1370 }
1371 
1372 static int
vlan_clone_destroy(struct if_clone * ifc,struct ifnet * ifp,uint32_t flags)1373 vlan_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
1374 {
1375 	struct ifvlan *ifv = ifp->if_softc;
1376 	int unit = ifp->if_dunit;
1377 
1378 	if (ifp->if_vlantrunk)
1379 		return (EBUSY);
1380 
1381 #ifdef ALTQ
1382 	IFQ_PURGE(&ifp->if_snd);
1383 #endif
1384 	ether_ifdetach(ifp);	/* first, remove it from system-wide lists */
1385 	vlan_unconfig(ifp);	/* now it can be unconfigured and freed */
1386 	/*
1387 	 * We should have the only reference to the ifv now, so we can now
1388 	 * drain any remaining lladdr task before freeing the ifnet and the
1389 	 * ifvlan.
1390 	 */
1391 	taskqueue_drain(taskqueue_thread, &ifv->lladdr_task);
1392 	NET_EPOCH_WAIT();
1393 	if_free(ifp);
1394 	free(ifv, M_VLAN);
1395 	if (unit != IF_DUNIT_NONE)
1396 		ifc_free_unit(ifc, unit);
1397 
1398 	return (0);
1399 }
1400 
1401 /*
1402  * The ifp->if_init entry point for vlan(4) is a no-op.
1403  */
1404 static void
vlan_init(void * foo __unused)1405 vlan_init(void *foo __unused)
1406 {
1407 }
1408 
1409 /*
1410  * The if_transmit method for vlan(4) interface.
1411  */
1412 static int
vlan_transmit(struct ifnet * ifp,struct mbuf * m)1413 vlan_transmit(struct ifnet *ifp, struct mbuf *m)
1414 {
1415 	struct ifvlan *ifv;
1416 	struct ifnet *p;
1417 	int error, len, mcast;
1418 
1419 	NET_EPOCH_ASSERT();
1420 
1421 	ifv = ifp->if_softc;
1422 	if (TRUNK(ifv) == NULL) {
1423 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1424 		m_freem(m);
1425 		return (ENETDOWN);
1426 	}
1427 	p = PARENT(ifv);
1428 	len = m->m_pkthdr.len;
1429 	mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
1430 
1431 	BPF_MTAP(ifp, m);
1432 
1433 #if defined(KERN_TLS) || defined(RATELIMIT)
1434 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) {
1435 		struct vlan_snd_tag *vst;
1436 		struct m_snd_tag *mst;
1437 
1438 		MPASS(m->m_pkthdr.snd_tag->ifp == ifp);
1439 		mst = m->m_pkthdr.snd_tag;
1440 		vst = mst_to_vst(mst);
1441 		if (vst->tag->ifp != p) {
1442 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1443 			m_freem(m);
1444 			return (EAGAIN);
1445 		}
1446 
1447 		m->m_pkthdr.snd_tag = m_snd_tag_ref(vst->tag);
1448 		m_snd_tag_rele(mst);
1449 	}
1450 #endif
1451 
1452 	/*
1453 	 * Do not run parent's if_transmit() if the parent is not up,
1454 	 * or parent's driver will cause a system crash.
1455 	 */
1456 	if (!UP_AND_RUNNING(p)) {
1457 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1458 		m_freem(m);
1459 		return (ENETDOWN);
1460 	}
1461 
1462 	if (!ether_8021q_frame(&m, ifp, p, &ifv->ifv_qtag)) {
1463 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1464 		return (0);
1465 	}
1466 
1467 	/*
1468 	 * Send it, precisely as ether_output() would have.
1469 	 */
1470 	error = (p->if_transmit)(p, m);
1471 	if (error == 0) {
1472 		if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
1473 		if_inc_counter(ifp, IFCOUNTER_OBYTES, len);
1474 		if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast);
1475 	} else
1476 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
1477 	return (error);
1478 }
1479 
1480 static int
vlan_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)1481 vlan_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
1482     struct route *ro)
1483 {
1484 	struct ifvlan *ifv;
1485 	struct ifnet *p;
1486 
1487 	NET_EPOCH_ASSERT();
1488 
1489 	/*
1490 	 * Find the first non-VLAN parent interface.
1491 	 */
1492 	ifv = ifp->if_softc;
1493 	do {
1494 		if (TRUNK(ifv) == NULL) {
1495 			m_freem(m);
1496 			return (ENETDOWN);
1497 		}
1498 		p = PARENT(ifv);
1499 		ifv = p->if_softc;
1500 	} while (p->if_type == IFT_L2VLAN);
1501 
1502 	return p->if_output(ifp, m, dst, ro);
1503 }
1504 
1505 #ifdef ALTQ
1506 static void
vlan_altq_start(if_t ifp)1507 vlan_altq_start(if_t ifp)
1508 {
1509 	struct ifaltq *ifq = &ifp->if_snd;
1510 	struct mbuf *m;
1511 
1512 	IFQ_LOCK(ifq);
1513 	IFQ_DEQUEUE_NOLOCK(ifq, m);
1514 	while (m != NULL) {
1515 		vlan_transmit(ifp, m);
1516 		IFQ_DEQUEUE_NOLOCK(ifq, m);
1517 	}
1518 	IFQ_UNLOCK(ifq);
1519 }
1520 
1521 static int
vlan_altq_transmit(if_t ifp,struct mbuf * m)1522 vlan_altq_transmit(if_t ifp, struct mbuf *m)
1523 {
1524 	int err;
1525 
1526 	if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
1527 		IFQ_ENQUEUE(&ifp->if_snd, m, err);
1528 		if (err == 0)
1529 			vlan_altq_start(ifp);
1530 	} else
1531 		err = vlan_transmit(ifp, m);
1532 
1533 	return (err);
1534 }
1535 #endif	/* ALTQ */
1536 
1537 /*
1538  * The ifp->if_qflush entry point for vlan(4) is a no-op.
1539  */
1540 static void
vlan_qflush(struct ifnet * ifp __unused)1541 vlan_qflush(struct ifnet *ifp __unused)
1542 {
1543 }
1544 
1545 static void
vlan_input(struct ifnet * ifp,struct mbuf * m)1546 vlan_input(struct ifnet *ifp, struct mbuf *m)
1547 {
1548 	struct ifvlantrunk *trunk;
1549 	struct ifvlan *ifv;
1550 	struct m_tag *mtag;
1551 	uint16_t vid, tag;
1552 
1553 	NET_EPOCH_ASSERT();
1554 
1555 	trunk = ifp->if_vlantrunk;
1556 	if (trunk == NULL) {
1557 		m_freem(m);
1558 		return;
1559 	}
1560 
1561 	if (m->m_flags & M_VLANTAG) {
1562 		/*
1563 		 * Packet is tagged, but m contains a normal
1564 		 * Ethernet frame; the tag is stored out-of-band.
1565 		 */
1566 		tag = m->m_pkthdr.ether_vtag;
1567 		m->m_flags &= ~M_VLANTAG;
1568 	} else {
1569 		struct ether_vlan_header *evl;
1570 
1571 		/*
1572 		 * Packet is tagged in-band as specified by 802.1q.
1573 		 */
1574 		switch (ifp->if_type) {
1575 		case IFT_ETHER:
1576 			if (m->m_len < sizeof(*evl) &&
1577 			    (m = m_pullup(m, sizeof(*evl))) == NULL) {
1578 				if_printf(ifp, "cannot pullup VLAN header\n");
1579 				return;
1580 			}
1581 			evl = mtod(m, struct ether_vlan_header *);
1582 			tag = ntohs(evl->evl_tag);
1583 
1584 			/*
1585 			 * Remove the 802.1q header by copying the Ethernet
1586 			 * addresses over it and adjusting the beginning of
1587 			 * the data in the mbuf.  The encapsulated Ethernet
1588 			 * type field is already in place.
1589 			 */
1590 			bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
1591 			      ETHER_HDR_LEN - ETHER_TYPE_LEN);
1592 			m_adj(m, ETHER_VLAN_ENCAP_LEN);
1593 			break;
1594 
1595 		default:
1596 #ifdef INVARIANTS
1597 			panic("%s: %s has unsupported if_type %u",
1598 			      __func__, ifp->if_xname, ifp->if_type);
1599 #endif
1600 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
1601 			m_freem(m);
1602 			return;
1603 		}
1604 	}
1605 
1606 	vid = EVL_VLANOFTAG(tag);
1607 
1608 	ifv = vlan_gethash(trunk, vid);
1609 	if (ifv == NULL || !UP_AND_RUNNING(ifv->ifv_ifp)) {
1610 		if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
1611 		m_freem(m);
1612 		return;
1613 	}
1614 
1615 	if (V_vlan_mtag_pcp) {
1616 		/*
1617 		 * While uncommon, it is possible that we will find a 802.1q
1618 		 * packet encapsulated inside another packet that also had an
1619 		 * 802.1q header.  For example, ethernet tunneled over IPSEC
1620 		 * arriving over ethernet.  In that case, we replace the
1621 		 * existing 802.1q PCP m_tag value.
1622 		 */
1623 		mtag = m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_IN, NULL);
1624 		if (mtag == NULL) {
1625 			mtag = m_tag_alloc(MTAG_8021Q, MTAG_8021Q_PCP_IN,
1626 			    sizeof(uint8_t), M_NOWAIT);
1627 			if (mtag == NULL) {
1628 				if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
1629 				m_freem(m);
1630 				return;
1631 			}
1632 			m_tag_prepend(m, mtag);
1633 		}
1634 		*(uint8_t *)(mtag + 1) = EVL_PRIOFTAG(tag);
1635 	}
1636 
1637 	m->m_pkthdr.rcvif = ifv->ifv_ifp;
1638 	if_inc_counter(ifv->ifv_ifp, IFCOUNTER_IPACKETS, 1);
1639 
1640 	/* Pass it back through the parent's input routine. */
1641 	(*ifv->ifv_ifp->if_input)(ifv->ifv_ifp, m);
1642 }
1643 
1644 static void
vlan_lladdr_fn(void * arg,int pending __unused)1645 vlan_lladdr_fn(void *arg, int pending __unused)
1646 {
1647 	struct ifvlan *ifv;
1648 	struct ifnet *ifp;
1649 
1650 	ifv = (struct ifvlan *)arg;
1651 	ifp = ifv->ifv_ifp;
1652 
1653 	CURVNET_SET(ifp->if_vnet);
1654 
1655 	/* The ifv_ifp already has the lladdr copied in. */
1656 	if_setlladdr(ifp, IF_LLADDR(ifp), ifp->if_addrlen);
1657 
1658 	CURVNET_RESTORE();
1659 }
1660 
1661 static int
vlan_config(struct ifvlan * ifv,struct ifnet * p,uint16_t vid,uint16_t proto)1662 vlan_config(struct ifvlan *ifv, struct ifnet *p, uint16_t vid,
1663 	uint16_t proto)
1664 {
1665 	struct epoch_tracker et;
1666 	struct ifvlantrunk *trunk;
1667 	struct ifnet *ifp;
1668 	int error = 0;
1669 
1670 	/*
1671 	 * We can handle non-ethernet hardware types as long as
1672 	 * they handle the tagging and headers themselves.
1673 	 */
1674 	if (p->if_type != IFT_ETHER &&
1675 	    p->if_type != IFT_L2VLAN &&
1676 	    (p->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
1677 		return (EPROTONOSUPPORT);
1678 	if ((p->if_flags & VLAN_IFFLAGS) != VLAN_IFFLAGS)
1679 		return (EPROTONOSUPPORT);
1680 	/*
1681 	 * Don't let the caller set up a VLAN VID with
1682 	 * anything except VLID bits.
1683 	 * VID numbers 0x0 and 0xFFF are reserved.
1684 	 */
1685 	if (vid == 0 || vid == 0xFFF || (vid & ~EVL_VLID_MASK))
1686 		return (EINVAL);
1687 	if (ifv->ifv_trunk) {
1688 		trunk = ifv->ifv_trunk;
1689 		if (trunk->parent != p)
1690 			return (EBUSY);
1691 
1692 		VLAN_XLOCK();
1693 
1694 		ifv->ifv_proto = proto;
1695 
1696 		if (ifv->ifv_vid != vid) {
1697 			int oldvid = ifv->ifv_vid;
1698 
1699 			/* Re-hash */
1700 			vlan_remhash(trunk, ifv);
1701 			ifv->ifv_vid = vid;
1702 			error = vlan_inshash(trunk, ifv);
1703 			if (error) {
1704 				int ret __diagused;
1705 
1706 				ifv->ifv_vid = oldvid;
1707 				/* Re-insert back where we found it. */
1708 				ret = vlan_inshash(trunk, ifv);
1709 				MPASS(ret == 0);
1710 			}
1711 		}
1712 		/* Will unlock */
1713 		goto done;
1714 	}
1715 
1716 	VLAN_XLOCK();
1717 	if (p->if_vlantrunk == NULL) {
1718 		trunk = malloc(sizeof(struct ifvlantrunk),
1719 		    M_VLAN, M_WAITOK | M_ZERO);
1720 		vlan_inithash(trunk);
1721 		TRUNK_LOCK_INIT(trunk);
1722 		TRUNK_WLOCK(trunk);
1723 		p->if_vlantrunk = trunk;
1724 		trunk->parent = p;
1725 		if_ref(trunk->parent);
1726 		TRUNK_WUNLOCK(trunk);
1727 	} else {
1728 		trunk = p->if_vlantrunk;
1729 	}
1730 
1731 	ifv->ifv_vid = vid;	/* must set this before vlan_inshash() */
1732 	ifv->ifv_pcp = 0;       /* Default: best effort delivery. */
1733 	error = vlan_inshash(trunk, ifv);
1734 	if (error)
1735 		goto done;
1736 	ifv->ifv_proto = proto;
1737 	ifv->ifv_encaplen = ETHER_VLAN_ENCAP_LEN;
1738 	ifv->ifv_mintu = ETHERMIN;
1739 	ifv->ifv_pflags = 0;
1740 	ifv->ifv_capenable = -1;
1741 	ifv->ifv_capenable2 = -1;
1742 
1743 	/*
1744 	 * If the parent supports the VLAN_MTU capability,
1745 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames,
1746 	 * use it.
1747 	 */
1748 	if (p->if_capenable & IFCAP_VLAN_MTU) {
1749 		/*
1750 		 * No need to fudge the MTU since the parent can
1751 		 * handle extended frames.
1752 		 */
1753 		ifv->ifv_mtufudge = 0;
1754 	} else {
1755 		/*
1756 		 * Fudge the MTU by the encapsulation size.  This
1757 		 * makes us incompatible with strictly compliant
1758 		 * 802.1Q implementations, but allows us to use
1759 		 * the feature with other NetBSD implementations,
1760 		 * which might still be useful.
1761 		 */
1762 		ifv->ifv_mtufudge = ifv->ifv_encaplen;
1763 	}
1764 
1765 	ifv->ifv_trunk = trunk;
1766 	ifp = ifv->ifv_ifp;
1767 	/*
1768 	 * Initialize fields from our parent.  This duplicates some
1769 	 * work with ether_ifattach() but allows for non-ethernet
1770 	 * interfaces to also work.
1771 	 */
1772 	ifp->if_mtu = p->if_mtu - ifv->ifv_mtufudge;
1773 	ifp->if_baudrate = p->if_baudrate;
1774 	ifp->if_input = p->if_input;
1775 	ifp->if_resolvemulti = p->if_resolvemulti;
1776 	ifp->if_addrlen = p->if_addrlen;
1777 	ifp->if_broadcastaddr = p->if_broadcastaddr;
1778 	ifp->if_pcp = ifv->ifv_pcp;
1779 
1780 	/*
1781 	 * We wrap the parent's if_output using vlan_output to ensure that it
1782 	 * can't become stale.
1783 	 */
1784 	ifp->if_output = vlan_output;
1785 
1786 	/*
1787 	 * Copy only a selected subset of flags from the parent.
1788 	 * Other flags are none of our business.
1789 	 */
1790 #define VLAN_COPY_FLAGS (IFF_SIMPLEX)
1791 	ifp->if_flags &= ~VLAN_COPY_FLAGS;
1792 	ifp->if_flags |= p->if_flags & VLAN_COPY_FLAGS;
1793 #undef VLAN_COPY_FLAGS
1794 
1795 	ifp->if_link_state = p->if_link_state;
1796 
1797 	NET_EPOCH_ENTER(et);
1798 	vlan_capabilities(ifv);
1799 	NET_EPOCH_EXIT(et);
1800 
1801 	/*
1802 	 * Set up our interface address to reflect the underlying
1803 	 * physical interface's.
1804 	 */
1805 	TASK_INIT(&ifv->lladdr_task, 0, vlan_lladdr_fn, ifv);
1806 	((struct sockaddr_dl *)ifp->if_addr->ifa_addr)->sdl_alen =
1807 	    p->if_addrlen;
1808 
1809 	/*
1810 	 * Do not schedule link address update if it was the same
1811 	 * as previous parent's. This helps avoid updating for each
1812 	 * associated llentry.
1813 	 */
1814 	if (memcmp(IF_LLADDR(p), IF_LLADDR(ifp), p->if_addrlen) != 0) {
1815 		bcopy(IF_LLADDR(p), IF_LLADDR(ifp), p->if_addrlen);
1816 		taskqueue_enqueue(taskqueue_thread, &ifv->lladdr_task);
1817 	}
1818 
1819 	/* We are ready for operation now. */
1820 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
1821 
1822 	/* Update flags on the parent, if necessary. */
1823 	vlan_setflags(ifp, 1);
1824 
1825 	/*
1826 	 * Configure multicast addresses that may already be
1827 	 * joined on the vlan device.
1828 	 */
1829 	(void)vlan_setmulti(ifp);
1830 
1831 done:
1832 	if (error == 0)
1833 		EVENTHANDLER_INVOKE(vlan_config, p, ifv->ifv_vid);
1834 	VLAN_XUNLOCK();
1835 
1836 	return (error);
1837 }
1838 
1839 static void
vlan_unconfig(struct ifnet * ifp)1840 vlan_unconfig(struct ifnet *ifp)
1841 {
1842 
1843 	VLAN_XLOCK();
1844 	vlan_unconfig_locked(ifp, 0);
1845 	VLAN_XUNLOCK();
1846 }
1847 
1848 static void
vlan_unconfig_locked(struct ifnet * ifp,int departing)1849 vlan_unconfig_locked(struct ifnet *ifp, int departing)
1850 {
1851 	struct ifvlantrunk *trunk;
1852 	struct vlan_mc_entry *mc;
1853 	struct ifvlan *ifv;
1854 	struct ifnet  *parent;
1855 	int error;
1856 
1857 	VLAN_XLOCK_ASSERT();
1858 
1859 	ifv = ifp->if_softc;
1860 	trunk = ifv->ifv_trunk;
1861 	parent = NULL;
1862 
1863 	if (trunk != NULL) {
1864 		parent = trunk->parent;
1865 
1866 		/*
1867 		 * Since the interface is being unconfigured, we need to
1868 		 * empty the list of multicast groups that we may have joined
1869 		 * while we were alive from the parent's list.
1870 		 */
1871 		while ((mc = CK_SLIST_FIRST(&ifv->vlan_mc_listhead)) != NULL) {
1872 			/*
1873 			 * If the parent interface is being detached,
1874 			 * all its multicast addresses have already
1875 			 * been removed.  Warn about errors if
1876 			 * if_delmulti() does fail, but don't abort as
1877 			 * all callers expect vlan destruction to
1878 			 * succeed.
1879 			 */
1880 			if (!departing) {
1881 				error = if_delmulti(parent,
1882 				    (struct sockaddr *)&mc->mc_addr);
1883 				if (error)
1884 					if_printf(ifp,
1885 		    "Failed to delete multicast address from parent: %d\n",
1886 					    error);
1887 			}
1888 			CK_SLIST_REMOVE_HEAD(&ifv->vlan_mc_listhead, mc_entries);
1889 			NET_EPOCH_CALL(vlan_mc_free, &mc->mc_epoch_ctx);
1890 		}
1891 
1892 		vlan_setflags(ifp, 0); /* clear special flags on parent */
1893 
1894 		vlan_remhash(trunk, ifv);
1895 		ifv->ifv_trunk = NULL;
1896 
1897 		/*
1898 		 * Check if we were the last.
1899 		 */
1900 		if (trunk->refcnt == 0) {
1901 			parent->if_vlantrunk = NULL;
1902 			NET_EPOCH_WAIT();
1903 			trunk_destroy(trunk);
1904 		}
1905 	}
1906 
1907 	/* Disconnect from parent. */
1908 	if (ifv->ifv_pflags)
1909 		if_printf(ifp, "%s: ifv_pflags unclean\n", __func__);
1910 	ifp->if_mtu = ETHERMTU;
1911 	ifp->if_link_state = LINK_STATE_UNKNOWN;
1912 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1913 
1914 	/*
1915 	 * Only dispatch an event if vlan was
1916 	 * attached, otherwise there is nothing
1917 	 * to cleanup anyway.
1918 	 */
1919 	if (parent != NULL)
1920 		EVENTHANDLER_INVOKE(vlan_unconfig, parent, ifv->ifv_vid);
1921 }
1922 
1923 /* Handle a reference counted flag that should be set on the parent as well */
1924 static int
vlan_setflag(struct ifnet * ifp,int flag,int status,int (* func)(struct ifnet *,int))1925 vlan_setflag(struct ifnet *ifp, int flag, int status,
1926 	     int (*func)(struct ifnet *, int))
1927 {
1928 	struct ifvlan *ifv;
1929 	int error;
1930 
1931 	VLAN_SXLOCK_ASSERT();
1932 
1933 	ifv = ifp->if_softc;
1934 	status = status ? (ifp->if_flags & flag) : 0;
1935 	/* Now "status" contains the flag value or 0 */
1936 
1937 	/*
1938 	 * See if recorded parent's status is different from what
1939 	 * we want it to be.  If it is, flip it.  We record parent's
1940 	 * status in ifv_pflags so that we won't clear parent's flag
1941 	 * we haven't set.  In fact, we don't clear or set parent's
1942 	 * flags directly, but get or release references to them.
1943 	 * That's why we can be sure that recorded flags still are
1944 	 * in accord with actual parent's flags.
1945 	 */
1946 	if (status != (ifv->ifv_pflags & flag)) {
1947 		error = (*func)(PARENT(ifv), status);
1948 		if (error)
1949 			return (error);
1950 		ifv->ifv_pflags &= ~flag;
1951 		ifv->ifv_pflags |= status;
1952 	}
1953 	return (0);
1954 }
1955 
1956 /*
1957  * Handle IFF_* flags that require certain changes on the parent:
1958  * if "status" is true, update parent's flags respective to our if_flags;
1959  * if "status" is false, forcedly clear the flags set on parent.
1960  */
1961 static int
vlan_setflags(struct ifnet * ifp,int status)1962 vlan_setflags(struct ifnet *ifp, int status)
1963 {
1964 	int error, i;
1965 
1966 	for (i = 0; vlan_pflags[i].flag; i++) {
1967 		error = vlan_setflag(ifp, vlan_pflags[i].flag,
1968 				     status, vlan_pflags[i].func);
1969 		if (error)
1970 			return (error);
1971 	}
1972 	return (0);
1973 }
1974 
1975 /* Inform all vlans that their parent has changed link state */
1976 static void
vlan_link_state(struct ifnet * ifp)1977 vlan_link_state(struct ifnet *ifp)
1978 {
1979 	struct epoch_tracker et;
1980 	struct ifvlantrunk *trunk;
1981 	struct ifvlan *ifv;
1982 
1983 	NET_EPOCH_ENTER(et);
1984 	trunk = ifp->if_vlantrunk;
1985 	if (trunk == NULL) {
1986 		NET_EPOCH_EXIT(et);
1987 		return;
1988 	}
1989 
1990 	TRUNK_WLOCK(trunk);
1991 	VLAN_FOREACH(ifv, trunk) {
1992 		ifv->ifv_ifp->if_baudrate = trunk->parent->if_baudrate;
1993 		if_link_state_change(ifv->ifv_ifp,
1994 		    trunk->parent->if_link_state);
1995 	}
1996 	TRUNK_WUNLOCK(trunk);
1997 	NET_EPOCH_EXIT(et);
1998 }
1999 
2000 #ifdef IPSEC_OFFLOAD
2001 #define	VLAN_IPSEC_METHOD(exp)				\
2002 	if_t p;						\
2003 	struct ifvlan *ifv;				\
2004 	int error;					\
2005 							\
2006 	ifv = ifp->if_softc;				\
2007 	VLAN_SLOCK();					\
2008 	if (TRUNK(ifv) != NULL) {			\
2009 		p = PARENT(ifv);			\
2010 		if_ref(p);				\
2011 		error = p->if_ipsec_accel_m->exp;	\
2012 		if_rele(p);				\
2013 	} else {					\
2014 		error = ENXIO;				\
2015 	}						\
2016 	VLAN_SUNLOCK();					\
2017 	return (error);
2018 
2019 
2020 static int
vlan_if_spdadd(if_t ifp,void * sp,void * inp,void ** priv)2021 vlan_if_spdadd(if_t ifp, void *sp, void *inp, void **priv)
2022 {
2023 	VLAN_IPSEC_METHOD(if_spdadd(ifp, sp, inp, priv));
2024 }
2025 
2026 static int
vlan_if_spddel(if_t ifp,void * sp,void * priv)2027 vlan_if_spddel(if_t ifp, void *sp, void *priv)
2028 {
2029 	VLAN_IPSEC_METHOD(if_spddel(ifp, sp, priv));
2030 }
2031 
2032 static int
vlan_if_sa_newkey(if_t ifp,void * sav,u_int drv_spi,void ** privp)2033 vlan_if_sa_newkey(if_t ifp, void *sav, u_int drv_spi, void **privp)
2034 {
2035 	VLAN_IPSEC_METHOD(if_sa_newkey(ifp, sav, drv_spi, privp));
2036 }
2037 
2038 static int
vlan_if_sa_deinstall(if_t ifp,u_int drv_spi,void * priv)2039 vlan_if_sa_deinstall(if_t ifp, u_int drv_spi, void *priv)
2040 {
2041 	VLAN_IPSEC_METHOD(if_sa_deinstall(ifp, drv_spi, priv));
2042 }
2043 
2044 static int
vlan_if_sa_cnt(if_t ifp,void * sa,uint32_t drv_spi,void * priv,struct seclifetime * lt)2045 vlan_if_sa_cnt(if_t ifp, void *sa, uint32_t drv_spi, void *priv,
2046     struct seclifetime *lt)
2047 {
2048 	VLAN_IPSEC_METHOD(if_sa_cnt(ifp, sa, drv_spi, priv, lt));
2049 }
2050 
2051 static int
vlan_if_ipsec_hwassist(if_t ifp,void * sav,u_int drv_spi,void * priv)2052 vlan_if_ipsec_hwassist(if_t ifp, void *sav, u_int drv_spi,void *priv)
2053 {
2054 	if_t trunk;
2055 
2056 	NET_EPOCH_ASSERT();
2057 	trunk = vlan_trunkdev(ifp);
2058 	if (trunk == NULL)
2059 		return (0);
2060 	return (trunk->if_ipsec_accel_m->if_hwassist(trunk, sav,
2061 	    drv_spi, priv));
2062 }
2063 
2064 static const struct if_ipsec_accel_methods vlan_if_ipsec_accel_methods = {
2065 	.if_spdadd = vlan_if_spdadd,
2066 	.if_spddel = vlan_if_spddel,
2067 	.if_sa_newkey = vlan_if_sa_newkey,
2068 	.if_sa_deinstall = vlan_if_sa_deinstall,
2069 	.if_sa_cnt = vlan_if_sa_cnt,
2070 	.if_hwassist = vlan_if_ipsec_hwassist,
2071 };
2072 
2073 #undef VLAN_IPSEC_METHOD
2074 #endif	/* IPSEC_OFFLOAD */
2075 
2076 static void
vlan_capabilities(struct ifvlan * ifv)2077 vlan_capabilities(struct ifvlan *ifv)
2078 {
2079 	struct ifnet *p;
2080 	struct ifnet *ifp;
2081 	struct ifnet_hw_tsomax hw_tsomax;
2082 	int cap = 0, ena = 0, mena, cap2 = 0, ena2 = 0;
2083 	int mena2 __unused;
2084 	u_long hwa = 0;
2085 
2086 	NET_EPOCH_ASSERT();
2087 	VLAN_SXLOCK_ASSERT();
2088 
2089 	p = PARENT(ifv);
2090 	ifp = ifv->ifv_ifp;
2091 
2092 	/* Mask parent interface enabled capabilities disabled by user. */
2093 	mena = p->if_capenable & ifv->ifv_capenable;
2094 	mena2 = p->if_capenable2 & ifv->ifv_capenable2;
2095 
2096 	/*
2097 	 * If the parent interface can do checksum offloading
2098 	 * on VLANs, then propagate its hardware-assisted
2099 	 * checksumming flags. Also assert that checksum
2100 	 * offloading requires hardware VLAN tagging.
2101 	 */
2102 	if (p->if_capabilities & IFCAP_VLAN_HWCSUM)
2103 		cap |= p->if_capabilities & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
2104 	if (p->if_capenable & IFCAP_VLAN_HWCSUM &&
2105 	    p->if_capenable & IFCAP_VLAN_HWTAGGING) {
2106 		ena |= mena & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
2107 		if (ena & IFCAP_TXCSUM)
2108 			hwa |= p->if_hwassist & (CSUM_IP | CSUM_TCP |
2109 			    CSUM_UDP | CSUM_SCTP);
2110 		if (ena & IFCAP_TXCSUM_IPV6)
2111 			hwa |= p->if_hwassist & (CSUM_TCP_IPV6 |
2112 			    CSUM_UDP_IPV6 | CSUM_SCTP_IPV6);
2113 	}
2114 
2115 	/*
2116 	 * If the parent interface can do TSO on VLANs then
2117 	 * propagate the hardware-assisted flag. TSO on VLANs
2118 	 * does not necessarily require hardware VLAN tagging.
2119 	 */
2120 	memset(&hw_tsomax, 0, sizeof(hw_tsomax));
2121 	if_hw_tsomax_common(p, &hw_tsomax);
2122 	if_hw_tsomax_update(ifp, &hw_tsomax);
2123 	if (p->if_capabilities & IFCAP_VLAN_HWTSO)
2124 		cap |= p->if_capabilities & IFCAP_TSO;
2125 	if (p->if_capenable & IFCAP_VLAN_HWTSO) {
2126 		ena |= mena & IFCAP_TSO;
2127 		if (ena & IFCAP_TSO)
2128 			hwa |= p->if_hwassist & CSUM_TSO;
2129 	}
2130 
2131 	/*
2132 	 * If the parent interface can do LRO and checksum offloading on
2133 	 * VLANs, then guess it may do LRO on VLANs.  False positive here
2134 	 * cost nothing, while false negative may lead to some confusions.
2135 	 */
2136 	if (p->if_capabilities & IFCAP_VLAN_HWCSUM)
2137 		cap |= p->if_capabilities & IFCAP_LRO;
2138 	if (p->if_capenable & IFCAP_VLAN_HWCSUM)
2139 		ena |= mena & IFCAP_LRO;
2140 
2141 	/*
2142 	 * If the parent interface can offload TCP connections over VLANs then
2143 	 * propagate its TOE capability to the VLAN interface.
2144 	 *
2145 	 * All TOE drivers in the tree today can deal with VLANs.  If this
2146 	 * changes then IFCAP_VLAN_TOE should be promoted to a full capability
2147 	 * with its own bit.
2148 	 */
2149 #define	IFCAP_VLAN_TOE IFCAP_TOE
2150 	if (p->if_capabilities & IFCAP_VLAN_TOE)
2151 		cap |= p->if_capabilities & IFCAP_TOE;
2152 	if (p->if_capenable & IFCAP_VLAN_TOE) {
2153 		SETTOEDEV(ifp, TOEDEV(p));
2154 		ena |= mena & IFCAP_TOE;
2155 	}
2156 
2157 	/*
2158 	 * If the parent interface supports dynamic link state, so does the
2159 	 * VLAN interface.
2160 	 */
2161 	cap |= (p->if_capabilities & IFCAP_LINKSTATE);
2162 	ena |= (mena & IFCAP_LINKSTATE);
2163 
2164 #ifdef RATELIMIT
2165 	/*
2166 	 * If the parent interface supports ratelimiting, so does the
2167 	 * VLAN interface.
2168 	 */
2169 	cap |= (p->if_capabilities & IFCAP_TXRTLMT);
2170 	ena |= (mena & IFCAP_TXRTLMT);
2171 #endif
2172 
2173 	/*
2174 	 * If the parent interface supports unmapped mbufs, so does
2175 	 * the VLAN interface.  Note that this should be fine even for
2176 	 * interfaces that don't support hardware tagging as headers
2177 	 * are prepended in normal mbufs to unmapped mbufs holding
2178 	 * payload data.
2179 	 */
2180 	cap |= (p->if_capabilities & IFCAP_MEXTPG);
2181 	ena |= (mena & IFCAP_MEXTPG);
2182 
2183 	/*
2184 	 * If the parent interface can offload encryption and segmentation
2185 	 * of TLS records over TCP, propagate it's capability to the VLAN
2186 	 * interface.
2187 	 *
2188 	 * All TLS drivers in the tree today can deal with VLANs.  If
2189 	 * this ever changes, then a new IFCAP_VLAN_TXTLS can be
2190 	 * defined.
2191 	 */
2192 	if (p->if_capabilities & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT))
2193 		cap |= p->if_capabilities & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT);
2194 	if (p->if_capenable & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT))
2195 		ena |= mena & (IFCAP_TXTLS | IFCAP_TXTLS_RTLMT);
2196 
2197 	ifp->if_capabilities = cap;
2198 	ifp->if_capenable = ena;
2199 	ifp->if_hwassist = hwa;
2200 
2201 #ifdef IPSEC_OFFLOAD
2202 	cap2 |= p->if_capabilities2 & IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD);
2203 	ena2 |= mena2 & IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD);
2204 	ifp->if_ipsec_accel_m = &vlan_if_ipsec_accel_methods;
2205 #endif
2206 
2207 	ifp->if_capabilities2 = cap2;
2208 	ifp->if_capenable2 = ena2;
2209 }
2210 
2211 static void
vlan_trunk_capabilities(struct ifnet * ifp)2212 vlan_trunk_capabilities(struct ifnet *ifp)
2213 {
2214 	struct epoch_tracker et;
2215 	struct ifvlantrunk *trunk;
2216 	struct ifvlan *ifv;
2217 
2218 	VLAN_SLOCK();
2219 	trunk = ifp->if_vlantrunk;
2220 	if (trunk == NULL) {
2221 		VLAN_SUNLOCK();
2222 		return;
2223 	}
2224 	NET_EPOCH_ENTER(et);
2225 	VLAN_FOREACH(ifv, trunk)
2226 		vlan_capabilities(ifv);
2227 	NET_EPOCH_EXIT(et);
2228 	VLAN_SUNLOCK();
2229 }
2230 
2231 static int
vlan_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)2232 vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
2233 {
2234 	struct ifnet *p;
2235 	struct ifreq *ifr;
2236 #ifdef INET
2237 	struct ifaddr *ifa;
2238 #endif
2239 	struct ifvlan *ifv;
2240 	struct ifvlantrunk *trunk;
2241 	struct vlanreq vlr;
2242 	int error = 0, oldmtu;
2243 
2244 	ifr = (struct ifreq *)data;
2245 #ifdef INET
2246 	ifa = (struct ifaddr *) data;
2247 #endif
2248 	ifv = ifp->if_softc;
2249 
2250 	switch (cmd) {
2251 	case SIOCSIFADDR:
2252 		ifp->if_flags |= IFF_UP;
2253 #ifdef INET
2254 		if (ifa->ifa_addr->sa_family == AF_INET)
2255 			arp_ifinit(ifp, ifa);
2256 #endif
2257 		break;
2258 	case SIOCGIFADDR:
2259 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
2260 		    ifp->if_addrlen);
2261 		break;
2262 	case SIOCGIFMEDIA:
2263 		VLAN_SLOCK();
2264 		if (TRUNK(ifv) != NULL) {
2265 			p = PARENT(ifv);
2266 			if_ref(p);
2267 			error = (*p->if_ioctl)(p, SIOCGIFMEDIA, data);
2268 			if_rele(p);
2269 			/* Limit the result to the parent's current config. */
2270 			if (error == 0) {
2271 				struct ifmediareq *ifmr;
2272 
2273 				ifmr = (struct ifmediareq *)data;
2274 				if (ifmr->ifm_count >= 1 && ifmr->ifm_ulist) {
2275 					ifmr->ifm_count = 1;
2276 					error = copyout(&ifmr->ifm_current,
2277 						ifmr->ifm_ulist,
2278 						sizeof(int));
2279 				}
2280 			}
2281 		} else {
2282 			error = EINVAL;
2283 		}
2284 		VLAN_SUNLOCK();
2285 		break;
2286 
2287 	case SIOCSIFMEDIA:
2288 		error = EINVAL;
2289 		break;
2290 
2291 	case SIOCSIFMTU:
2292 		/*
2293 		 * Set the interface MTU.
2294 		 */
2295 		VLAN_SLOCK();
2296 		trunk = TRUNK(ifv);
2297 		if (trunk != NULL) {
2298 			TRUNK_WLOCK(trunk);
2299 			if (ifr->ifr_mtu >
2300 			     (PARENT(ifv)->if_mtu - ifv->ifv_mtufudge) ||
2301 			    ifr->ifr_mtu <
2302 			     (ifv->ifv_mintu - ifv->ifv_mtufudge))
2303 				error = EINVAL;
2304 			else
2305 				ifp->if_mtu = ifr->ifr_mtu;
2306 			TRUNK_WUNLOCK(trunk);
2307 		} else
2308 			error = EINVAL;
2309 		VLAN_SUNLOCK();
2310 		break;
2311 
2312 	case SIOCSETVLAN:
2313 #ifdef VIMAGE
2314 		/*
2315 		 * XXXRW/XXXBZ: The goal in these checks is to allow a VLAN
2316 		 * interface to be delegated to a jail without allowing the
2317 		 * jail to change what underlying interface/VID it is
2318 		 * associated with.  We are not entirely convinced that this
2319 		 * is the right way to accomplish that policy goal.
2320 		 */
2321 		if (ifp->if_vnet != ifp->if_home_vnet) {
2322 			error = EPERM;
2323 			break;
2324 		}
2325 #endif
2326 		error = copyin(ifr_data_get_ptr(ifr), &vlr, sizeof(vlr));
2327 		if (error)
2328 			break;
2329 		if (vlr.vlr_parent[0] == '\0') {
2330 			vlan_unconfig(ifp);
2331 			break;
2332 		}
2333 		p = ifunit_ref(vlr.vlr_parent);
2334 		if (p == NULL) {
2335 			error = ENOENT;
2336 			break;
2337 		}
2338 		if (vlr.vlr_proto == 0)
2339 			vlr.vlr_proto = ETHERTYPE_VLAN;
2340 		oldmtu = ifp->if_mtu;
2341 		error = vlan_config(ifv, p, vlr.vlr_tag, vlr.vlr_proto);
2342 		if_rele(p);
2343 
2344 		/*
2345 		 * VLAN MTU may change during addition of the vlandev.
2346 		 * If it did, do network layer specific procedure.
2347 		 */
2348 		if (ifp->if_mtu != oldmtu)
2349 			if_notifymtu(ifp);
2350 		break;
2351 
2352 	case SIOCGETVLAN:
2353 #ifdef VIMAGE
2354 		if (ifp->if_vnet != ifp->if_home_vnet) {
2355 			error = EPERM;
2356 			break;
2357 		}
2358 #endif
2359 		bzero(&vlr, sizeof(vlr));
2360 		VLAN_SLOCK();
2361 		if (TRUNK(ifv) != NULL) {
2362 			strlcpy(vlr.vlr_parent, PARENT(ifv)->if_xname,
2363 			    sizeof(vlr.vlr_parent));
2364 			vlr.vlr_tag = ifv->ifv_vid;
2365 			vlr.vlr_proto = ifv->ifv_proto;
2366 		}
2367 		VLAN_SUNLOCK();
2368 		error = copyout(&vlr, ifr_data_get_ptr(ifr), sizeof(vlr));
2369 		break;
2370 
2371 	case SIOCSIFFLAGS:
2372 		/*
2373 		 * We should propagate selected flags to the parent,
2374 		 * e.g., promiscuous mode.
2375 		 */
2376 		VLAN_SLOCK();
2377 		if (TRUNK(ifv) != NULL)
2378 			error = vlan_setflags(ifp, 1);
2379 		VLAN_SUNLOCK();
2380 		break;
2381 
2382 	case SIOCADDMULTI:
2383 	case SIOCDELMULTI:
2384 		/*
2385 		 * If we don't have a parent, just remember the membership for
2386 		 * when we do.
2387 		 *
2388 		 * XXX We need the rmlock here to avoid sleeping while
2389 		 * holding in6_multi_mtx.
2390 		 */
2391 		VLAN_XLOCK();
2392 		trunk = TRUNK(ifv);
2393 		if (trunk != NULL)
2394 			error = vlan_setmulti(ifp);
2395 		VLAN_XUNLOCK();
2396 
2397 		break;
2398 	case SIOCGVLANPCP:
2399 #ifdef VIMAGE
2400 		if (ifp->if_vnet != ifp->if_home_vnet) {
2401 			error = EPERM;
2402 			break;
2403 		}
2404 #endif
2405 		ifr->ifr_vlan_pcp = ifv->ifv_pcp;
2406 		break;
2407 
2408 	case SIOCSVLANPCP:
2409 #ifdef VIMAGE
2410 		if (ifp->if_vnet != ifp->if_home_vnet) {
2411 			error = EPERM;
2412 			break;
2413 		}
2414 #endif
2415 		error = priv_check(curthread, PRIV_NET_SETVLANPCP);
2416 		if (error)
2417 			break;
2418 		if (ifr->ifr_vlan_pcp > VLAN_PCP_MAX) {
2419 			error = EINVAL;
2420 			break;
2421 		}
2422 		ifv->ifv_pcp = ifr->ifr_vlan_pcp;
2423 		ifp->if_pcp = ifv->ifv_pcp;
2424 		/* broadcast event about PCP change */
2425 		EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
2426 		break;
2427 
2428 	case SIOCSIFCAP:
2429 		VLAN_SLOCK();
2430 		ifv->ifv_capenable = ifr->ifr_reqcap;
2431 		trunk = TRUNK(ifv);
2432 		if (trunk != NULL) {
2433 			struct epoch_tracker et;
2434 
2435 			NET_EPOCH_ENTER(et);
2436 			vlan_capabilities(ifv);
2437 			NET_EPOCH_EXIT(et);
2438 		}
2439 		VLAN_SUNLOCK();
2440 		break;
2441 
2442 	default:
2443 		error = EINVAL;
2444 		break;
2445 	}
2446 
2447 	return (error);
2448 }
2449 
2450 #if defined(KERN_TLS) || defined(RATELIMIT)
2451 static int
vlan_snd_tag_alloc(struct ifnet * ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** ppmt)2452 vlan_snd_tag_alloc(struct ifnet *ifp,
2453     union if_snd_tag_alloc_params *params,
2454     struct m_snd_tag **ppmt)
2455 {
2456 	struct epoch_tracker et;
2457 	const struct if_snd_tag_sw *sw;
2458 	struct vlan_snd_tag *vst;
2459 	struct ifvlan *ifv;
2460 	struct ifnet *parent;
2461 	struct m_snd_tag *mst;
2462 	int error;
2463 
2464 	NET_EPOCH_ENTER(et);
2465 	ifv = ifp->if_softc;
2466 
2467 	switch (params->hdr.type) {
2468 #ifdef RATELIMIT
2469 	case IF_SND_TAG_TYPE_UNLIMITED:
2470 		sw = &vlan_snd_tag_ul_sw;
2471 		break;
2472 	case IF_SND_TAG_TYPE_RATE_LIMIT:
2473 		sw = &vlan_snd_tag_rl_sw;
2474 		break;
2475 #endif
2476 #ifdef KERN_TLS
2477 	case IF_SND_TAG_TYPE_TLS:
2478 		sw = &vlan_snd_tag_tls_sw;
2479 		break;
2480 	case IF_SND_TAG_TYPE_TLS_RX:
2481 		sw = NULL;
2482 		if (params->tls_rx.vlan_id != 0)
2483 			goto failure;
2484 		params->tls_rx.vlan_id = ifv->ifv_vid;
2485 		break;
2486 #ifdef RATELIMIT
2487 	case IF_SND_TAG_TYPE_TLS_RATE_LIMIT:
2488 		sw = &vlan_snd_tag_tls_rl_sw;
2489 		break;
2490 #endif
2491 #endif
2492 	default:
2493 		goto failure;
2494 	}
2495 
2496 	if (ifv->ifv_trunk != NULL)
2497 		parent = PARENT(ifv);
2498 	else
2499 		parent = NULL;
2500 	if (parent == NULL)
2501 		goto failure;
2502 	if_ref(parent);
2503 	NET_EPOCH_EXIT(et);
2504 
2505 	if (sw != NULL) {
2506 		vst = malloc(sizeof(*vst), M_VLAN, M_NOWAIT);
2507 		if (vst == NULL) {
2508 			if_rele(parent);
2509 			return (ENOMEM);
2510 		}
2511 	} else
2512 		vst = NULL;
2513 
2514 	error = m_snd_tag_alloc(parent, params, &mst);
2515 	if_rele(parent);
2516 	if (error) {
2517 		free(vst, M_VLAN);
2518 		return (error);
2519 	}
2520 
2521 	if (sw != NULL) {
2522 		m_snd_tag_init(&vst->com, ifp, sw);
2523 		vst->tag = mst;
2524 
2525 		*ppmt = &vst->com;
2526 	} else
2527 		*ppmt = mst;
2528 
2529 	return (0);
2530 failure:
2531 	NET_EPOCH_EXIT(et);
2532 	return (EOPNOTSUPP);
2533 }
2534 
2535 static struct m_snd_tag *
vlan_next_snd_tag(struct m_snd_tag * mst)2536 vlan_next_snd_tag(struct m_snd_tag *mst)
2537 {
2538 	struct vlan_snd_tag *vst;
2539 
2540 	vst = mst_to_vst(mst);
2541 	return (vst->tag);
2542 }
2543 
2544 static int
vlan_snd_tag_modify(struct m_snd_tag * mst,union if_snd_tag_modify_params * params)2545 vlan_snd_tag_modify(struct m_snd_tag *mst,
2546     union if_snd_tag_modify_params *params)
2547 {
2548 	struct vlan_snd_tag *vst;
2549 
2550 	vst = mst_to_vst(mst);
2551 	return (vst->tag->sw->snd_tag_modify(vst->tag, params));
2552 }
2553 
2554 static int
vlan_snd_tag_query(struct m_snd_tag * mst,union if_snd_tag_query_params * params)2555 vlan_snd_tag_query(struct m_snd_tag *mst,
2556     union if_snd_tag_query_params *params)
2557 {
2558 	struct vlan_snd_tag *vst;
2559 
2560 	vst = mst_to_vst(mst);
2561 	return (vst->tag->sw->snd_tag_query(vst->tag, params));
2562 }
2563 
2564 static void
vlan_snd_tag_free(struct m_snd_tag * mst)2565 vlan_snd_tag_free(struct m_snd_tag *mst)
2566 {
2567 	struct vlan_snd_tag *vst;
2568 
2569 	vst = mst_to_vst(mst);
2570 	m_snd_tag_rele(vst->tag);
2571 	free(vst, M_VLAN);
2572 }
2573 
2574 static void
vlan_ratelimit_query(struct ifnet * ifp __unused,struct if_ratelimit_query_results * q)2575 vlan_ratelimit_query(struct ifnet *ifp __unused, struct if_ratelimit_query_results *q)
2576 {
2577 	/*
2578 	 * For vlan, we have an indirect
2579 	 * interface. The caller needs to
2580 	 * get a ratelimit tag on the actual
2581 	 * interface the flow will go on.
2582 	 */
2583 	q->rate_table = NULL;
2584 	q->flags = RT_IS_INDIRECT;
2585 	q->max_flows = 0;
2586 	q->number_of_rates = 0;
2587 }
2588 
2589 #endif
2590