if_vlan.c revision 80296
1/*
2 * Copyright 1998 Massachusetts Institute of Technology
3 *
4 * Permission to use, copy, modify, and distribute this software and
5 * its documentation for any purpose and without fee is hereby
6 * granted, provided that both the above copyright notice and this
7 * permission notice appear in all copies, that both the above
8 * copyright notice and this permission notice appear in all
9 * supporting documentation, and that the name of M.I.T. not be used
10 * in advertising or publicity pertaining to distribution of the
11 * software without specific, written prior permission.  M.I.T. makes
12 * no representations about the suitability of this software for any
13 * purpose.  It is provided "as is" without express or implied
14 * warranty.
15 *
16 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
17 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
18 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
20 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
23 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
24 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
25 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
26 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $FreeBSD: head/sys/net/if_vlan.c 80296 2001-07-24 17:14:37Z fenner $
30 */
31
32/*
33 * if_vlan.c - pseudo-device driver for IEEE 802.1Q virtual LANs.
34 * Might be extended some day to also handle IEEE 802.1p priority
35 * tagging.  This is sort of sneaky in the implementation, since
36 * we need to pretend to be enough of an Ethernet implementation
37 * to make arp work.  The way we do this is by telling everyone
38 * that we are an Ethernet, and then catch the packets that
39 * ether_output() left on our output queue when it calls
40 * if_start(), rewrite them for use by the real outgoing interface,
41 * and ask it to send them.
42 *
43 *
44 * XXX It's incorrect to assume that we must always kludge up
45 * headers on the physical device's behalf: some devices support
46 * VLAN tag insertion and extraction in firmware. For these cases,
47 * one can change the behavior of the vlan interface by setting
48 * the LINK0 flag on it (that is setting the vlan interface's LINK0
49 * flag, _not_ the parent's LINK0 flag; we try to leave the parent
50 * alone). If the interface has the LINK0 flag set, then it will
51 * not modify the ethernet header on output, because the parent
52 * can do that for itself. On input, the parent can call vlan_input_tag()
53 * directly in order to supply us with an incoming mbuf and the vlan
54 * tag value that goes with it.
55 */
56
57#include "vlan.h"
58#include "opt_inet.h"
59
60#include <sys/param.h>
61#include <sys/kernel.h>
62#include <sys/malloc.h>
63#include <sys/mbuf.h>
64#include <sys/module.h>
65#include <sys/queue.h>
66#include <sys/socket.h>
67#include <sys/sockio.h>
68#include <sys/sysctl.h>
69#include <sys/systm.h>
70
71#include <net/bpf.h>
72#include <net/ethernet.h>
73#include <net/if.h>
74#include <net/if_arp.h>
75#include <net/if_dl.h>
76#include <net/if_types.h>
77#include <net/if_vlan_var.h>
78
79#ifdef INET
80#include <netinet/in.h>
81#include <netinet/if_ether.h>
82#endif
83
84SYSCTL_DECL(_net_link);
85SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0, "IEEE 802.1Q VLAN");
86SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0, "for consistency");
87
88u_int	vlan_proto = ETHERTYPE_VLAN;
89SYSCTL_INT(_net_link_vlan_link, VLANCTL_PROTO, proto, CTLFLAG_RW, &vlan_proto,
90	   0, "Ethernet protocol used for VLAN encapsulation");
91
92static	struct ifvlan ifv_softc[NVLAN];
93
94static	void vlan_start(struct ifnet *ifp);
95static	void vlan_ifinit(void *foo);
96static	int vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr);
97static	int vlan_setmulti(struct ifnet *ifp);
98static	int vlan_unconfig(struct ifnet *ifp);
99static	int vlan_config(struct ifvlan *ifv, struct ifnet *p);
100
101/*
102 * Program our multicast filter. What we're actually doing is
103 * programming the multicast filter of the parent. This has the
104 * side effect of causing the parent interface to receive multicast
105 * traffic that it doesn't really want, which ends up being discarded
106 * later by the upper protocol layers. Unfortunately, there's no way
107 * to avoid this: there really is only one physical interface.
108 */
109static int
110vlan_setmulti(struct ifnet *ifp)
111{
112	struct ifnet		*ifp_p;
113	struct ifmultiaddr	*ifma, *rifma = NULL;
114	struct ifvlan		*sc;
115	struct vlan_mc_entry	*mc = NULL;
116	struct sockaddr_dl	sdl;
117	int			error;
118
119	/* Find the parent. */
120	sc = ifp->if_softc;
121	ifp_p = sc->ifv_p;
122
123	/*
124	 * If we don't have a parent, just remember the membership for
125	 * when we do.
126	 */
127	if (ifp_p == NULL)
128		return(0);
129
130	bzero((char *)&sdl, sizeof sdl);
131	sdl.sdl_len = sizeof sdl;
132	sdl.sdl_family = AF_LINK;
133	sdl.sdl_index = ifp_p->if_index;
134	sdl.sdl_type = IFT_ETHER;
135	sdl.sdl_alen = ETHER_ADDR_LEN;
136
137	/* First, remove any existing filter entries. */
138	while(SLIST_FIRST(&sc->vlan_mc_listhead) != NULL) {
139		mc = SLIST_FIRST(&sc->vlan_mc_listhead);
140		bcopy((char *)&mc->mc_addr, LLADDR(&sdl), ETHER_ADDR_LEN);
141		error = if_delmulti(ifp_p, (struct sockaddr *)&sdl);
142		if (error)
143			return(error);
144		SLIST_REMOVE_HEAD(&sc->vlan_mc_listhead, mc_entries);
145		free(mc, M_DEVBUF);
146	}
147
148	/* Now program new ones. */
149	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
150		if (ifma->ifma_addr->sa_family != AF_LINK)
151			continue;
152		mc = malloc(sizeof(struct vlan_mc_entry), M_DEVBUF, M_WAITOK);
153		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
154		    (char *)&mc->mc_addr, ETHER_ADDR_LEN);
155		SLIST_INSERT_HEAD(&sc->vlan_mc_listhead, mc, mc_entries);
156		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
157		    LLADDR(&sdl), ETHER_ADDR_LEN);
158		error = if_addmulti(ifp_p, (struct sockaddr *)&sdl, &rifma);
159		if (error)
160			return(error);
161	}
162
163	return(0);
164}
165
166static void
167vlaninit(void)
168{
169	int i;
170
171	for (i = 0; i < NVLAN; i++) {
172		struct ifnet *ifp = &ifv_softc[i].ifv_if;
173
174		ifp->if_softc = &ifv_softc[i];
175		ifp->if_name = "vlan";
176		ifp->if_unit = i;
177		/* NB: flags are not set here */
178		ifp->if_linkmib = &ifv_softc[i].ifv_mib;
179		ifp->if_linkmiblen = sizeof ifv_softc[i].ifv_mib;
180		/* NB: mtu is not set here */
181
182		ifp->if_init = vlan_ifinit;
183		ifp->if_start = vlan_start;
184		ifp->if_ioctl = vlan_ioctl;
185		ifp->if_output = ether_output;
186		ifp->if_snd.ifq_maxlen = ifqmaxlen;
187		ether_ifattach(ifp, ETHER_BPF_SUPPORTED);
188		/* Now undo some of the damage... */
189		ifp->if_data.ifi_type = IFT_L2VLAN;
190		ifp->if_data.ifi_hdrlen = EVL_ENCAPLEN;
191	}
192}
193
194static int
195vlan_modevent(module_t mod, int type, void *data)
196{
197	switch (type) {
198	case MOD_LOAD:
199		vlaninit();
200		break;
201	case MOD_UNLOAD:
202		printf("if_vlan module unload - not possible for this module type\n");
203		return EINVAL;
204	}
205	return 0;
206}
207
208static moduledata_t vlan_mod = {
209	"if_vlan",
210	vlan_modevent,
211	0
212};
213
214DECLARE_MODULE(if_vlan, vlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
215
216static void
217vlan_ifinit(void *foo)
218{
219	return;
220}
221
222static void
223vlan_start(struct ifnet *ifp)
224{
225	struct ifvlan *ifv;
226	struct ifnet *p;
227	struct ether_vlan_header *evl;
228	struct mbuf *m;
229
230	ifv = ifp->if_softc;
231	p = ifv->ifv_p;
232
233	ifp->if_flags |= IFF_OACTIVE;
234	for (;;) {
235		IF_DEQUEUE(&ifp->if_snd, m);
236		if (m == 0)
237			break;
238		if (ifp->if_bpf)
239			bpf_mtap(ifp, m);
240
241		/*
242		 * Do not run parent's if_start() if the parent is not up,
243		 * or parent's driver will cause a system crash.
244		 */
245		if ((p->if_flags & (IFF_UP | IFF_RUNNING)) !=
246					(IFF_UP | IFF_RUNNING)) {
247			m_freem(m);
248			ifp->if_data.ifi_collisions++;
249			continue;
250		}
251
252		/*
253		 * If the LINK0 flag is set, it means the underlying interface
254		 * can do VLAN tag insertion itself and doesn't require us to
255	 	 * create a special header for it. In this case, we just pass
256		 * the packet along. However, we need some way to tell the
257		 * interface where the packet came from so that it knows how
258		 * to find the VLAN tag to use, so we set the rcvif in the
259		 * mbuf header to our ifnet.
260		 *
261		 * Note: we also set the M_PROTO1 flag in the mbuf to let
262		 * the parent driver know that the rcvif pointer is really
263		 * valid. We need to do this because sometimes mbufs will
264		 * be allocated by other parts of the system that contain
265		 * garbage in the rcvif pointer. Using the M_PROTO1 flag
266		 * lets the driver perform a proper sanity check and avoid
267		 * following potentially bogus rcvif pointers off into
268		 * never-never land.
269		 */
270		if (ifp->if_flags & IFF_LINK0) {
271			m->m_pkthdr.rcvif = ifp;
272			m->m_flags |= M_PROTO1;
273		} else {
274			M_PREPEND(m, EVL_ENCAPLEN, M_DONTWAIT);
275			if (m == NULL) {
276				printf("vlan%d: M_PREPEND failed", ifp->if_unit);
277				ifp->if_ierrors++;
278				continue;
279			}
280			/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
281
282			m = m_pullup(m, ETHER_HDR_LEN + EVL_ENCAPLEN);
283			if (m == NULL) {
284				printf("vlan%d: m_pullup failed", ifp->if_unit);
285				ifp->if_ierrors++;
286				continue;
287			}
288
289			/*
290			 * Transform the Ethernet header into an Ethernet header
291			 * with 802.1Q encapsulation.
292			 */
293			bcopy(mtod(m, char *) + EVL_ENCAPLEN, mtod(m, char *),
294			      sizeof(struct ether_header));
295			evl = mtod(m, struct ether_vlan_header *);
296			evl->evl_proto = evl->evl_encap_proto;
297			evl->evl_encap_proto = htons(vlan_proto);
298			evl->evl_tag = htons(ifv->ifv_tag);
299#ifdef DEBUG
300			printf("vlan_start: %*D\n", sizeof *evl,
301			    (unsigned char *)evl, ":");
302#endif
303		}
304
305		/*
306		 * Send it, precisely as ether_output() would have.
307		 * We are already running at splimp.
308		 */
309		if (IF_HANDOFF(&p->if_snd, m, p))
310			ifp->if_opackets++;
311		else
312			ifp->if_oerrors++;
313	}
314	ifp->if_flags &= ~IFF_OACTIVE;
315
316	return;
317}
318
319int
320vlan_input_tag(struct ether_header *eh, struct mbuf *m, u_int16_t t)
321{
322	int i;
323	struct ifvlan *ifv;
324
325	for (i = 0; i < NVLAN; i++) {
326		ifv = &ifv_softc[i];
327		if (ifv->ifv_tag == t)
328			break;
329	}
330
331	if (i >= NVLAN || (ifv->ifv_if.if_flags & IFF_UP) == 0) {
332		m_free(m);
333		return -1;	/* So the parent can take note */
334	}
335
336	/*
337	 * Having found a valid vlan interface corresponding to
338	 * the given source interface and vlan tag, run the
339	 * the real packet through ethert_input().
340	 */
341	m->m_pkthdr.rcvif = &ifv->ifv_if;
342
343	ifv->ifv_if.if_ipackets++;
344	ether_input(&ifv->ifv_if, eh, m);
345	return 0;
346}
347
348int
349vlan_input(struct ether_header *eh, struct mbuf *m)
350{
351	int i;
352	struct ifvlan *ifv;
353
354	for (i = 0; i < NVLAN; i++) {
355		ifv = &ifv_softc[i];
356		if (m->m_pkthdr.rcvif == ifv->ifv_p
357		    && (EVL_VLANOFTAG(ntohs(*mtod(m, u_int16_t *)))
358			== ifv->ifv_tag))
359			break;
360	}
361
362	if (i >= NVLAN || (ifv->ifv_if.if_flags & IFF_UP) == 0) {
363		m_freem(m);
364		return -1;	/* so ether_input can take note */
365	}
366
367	/*
368	 * Having found a valid vlan interface corresponding to
369	 * the given source interface and vlan tag, remove the
370	 * encapsulation, and run the real packet through
371	 * ether_input() a second time (it had better be
372	 * reentrant!).
373	 */
374	m->m_pkthdr.rcvif = &ifv->ifv_if;
375	eh->ether_type = mtod(m, u_int16_t *)[1];
376	m->m_data += EVL_ENCAPLEN;
377	m->m_len -= EVL_ENCAPLEN;
378	m->m_pkthdr.len -= EVL_ENCAPLEN;
379
380	ifv->ifv_if.if_ipackets++;
381	ether_input(&ifv->ifv_if, eh, m);
382	return 0;
383}
384
385static int
386vlan_config(struct ifvlan *ifv, struct ifnet *p)
387{
388	struct ifaddr *ifa1, *ifa2;
389	struct sockaddr_dl *sdl1, *sdl2;
390
391	if (p->if_data.ifi_type != IFT_ETHER)
392		return EPROTONOSUPPORT;
393	if (ifv->ifv_p)
394		return EBUSY;
395	ifv->ifv_p = p;
396	if (p->if_data.ifi_hdrlen == sizeof(struct ether_vlan_header))
397		ifv->ifv_if.if_mtu = p->if_mtu;
398	else
399		ifv->ifv_if.if_mtu = p->if_data.ifi_mtu - EVL_ENCAPLEN;
400
401	/*
402	 * Copy only a selected subset of flags from the parent.
403	 * Other flags are none of our business.
404	 */
405	ifv->ifv_if.if_flags = (p->if_flags &
406	    (IFF_BROADCAST | IFF_MULTICAST | IFF_SIMPLEX | IFF_POINTOPOINT));
407
408	/*
409	 * Set up our ``Ethernet address'' to reflect the underlying
410	 * physical interface's.
411	 */
412	ifa1 = ifnet_addrs[ifv->ifv_if.if_index - 1];
413	ifa2 = ifnet_addrs[p->if_index - 1];
414	sdl1 = (struct sockaddr_dl *)ifa1->ifa_addr;
415	sdl2 = (struct sockaddr_dl *)ifa2->ifa_addr;
416	sdl1->sdl_type = IFT_ETHER;
417	sdl1->sdl_alen = ETHER_ADDR_LEN;
418	bcopy(LLADDR(sdl2), LLADDR(sdl1), ETHER_ADDR_LEN);
419	bcopy(LLADDR(sdl2), ifv->ifv_ac.ac_enaddr, ETHER_ADDR_LEN);
420
421	/*
422	 * Configure multicast addresses that may already be
423	 * joined on the vlan device.
424	 */
425	(void)vlan_setmulti(&ifv->ifv_if);
426
427	return 0;
428}
429
430static int
431vlan_unconfig(struct ifnet *ifp)
432{
433	struct ifaddr *ifa;
434	struct sockaddr_dl *sdl;
435	struct vlan_mc_entry *mc;
436	struct ifvlan *ifv;
437	struct ifnet *p;
438	int error;
439
440	ifv = ifp->if_softc;
441	p = ifv->ifv_p;
442
443	if (p) {
444		struct sockaddr_dl sdl;
445
446		/*
447		 * Since the interface is being unconfigured, we need to
448		 * empty the list of multicast groups that we may have joined
449		 * while we were alive from the parent's list.
450		 */
451		bzero((char *)&sdl, sizeof sdl);
452		sdl.sdl_len = sizeof sdl;
453		sdl.sdl_family = AF_LINK;
454		sdl.sdl_index = p->if_index;
455		sdl.sdl_type = IFT_ETHER;
456		sdl.sdl_alen = ETHER_ADDR_LEN;
457
458		while(SLIST_FIRST(&ifv->vlan_mc_listhead) != NULL) {
459			mc = SLIST_FIRST(&ifv->vlan_mc_listhead);
460			bcopy((char *)&mc->mc_addr, LLADDR(&sdl), ETHER_ADDR_LEN);
461			error = if_delmulti(p, (struct sockaddr *)&sdl);
462			if (error)
463				return(error);
464			SLIST_REMOVE_HEAD(&ifv->vlan_mc_listhead, mc_entries);
465			free(mc, M_DEVBUF);
466		}
467	}
468
469	/* Disconnect from parent. */
470	ifv->ifv_p = NULL;
471	ifv->ifv_if.if_mtu = ETHERMTU;
472
473	/* Clear our MAC address. */
474	ifa = ifnet_addrs[ifv->ifv_if.if_index - 1];
475	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
476	sdl->sdl_type = IFT_ETHER;
477	sdl->sdl_alen = ETHER_ADDR_LEN;
478	bzero(LLADDR(sdl), ETHER_ADDR_LEN);
479	bzero(ifv->ifv_ac.ac_enaddr, ETHER_ADDR_LEN);
480
481	return 0;
482}
483
484static int
485vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
486{
487	struct ifaddr *ifa;
488	struct ifnet *p;
489	struct ifreq *ifr;
490	struct ifvlan *ifv;
491	struct vlanreq vlr;
492	int error = 0;
493
494	ifr = (struct ifreq *)data;
495	ifa = (struct ifaddr *)data;
496	ifv = ifp->if_softc;
497
498	switch (cmd) {
499	case SIOCSIFADDR:
500		ifp->if_flags |= IFF_UP;
501
502		switch (ifa->ifa_addr->sa_family) {
503#ifdef INET
504		case AF_INET:
505			arp_ifinit(&ifv->ifv_ac, ifa);
506			break;
507#endif
508		default:
509			break;
510		}
511		break;
512
513	case SIOCGIFADDR:
514		{
515			struct sockaddr *sa;
516
517			sa = (struct sockaddr *) &ifr->ifr_data;
518			bcopy(((struct arpcom *)ifp->if_softc)->ac_enaddr,
519			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
520		}
521		break;
522
523	case SIOCSIFMTU:
524		/*
525		 * Set the interface MTU.
526		 * This is bogus. The underlying interface might support
527	 	 * jumbo frames.
528		 */
529		if (ifr->ifr_mtu > ETHERMTU) {
530			error = EINVAL;
531		} else {
532			ifp->if_mtu = ifr->ifr_mtu;
533		}
534		break;
535
536	case SIOCSETVLAN:
537		error = copyin(ifr->ifr_data, &vlr, sizeof vlr);
538		if (error)
539			break;
540		if (vlr.vlr_parent[0] == '\0') {
541			vlan_unconfig(ifp);
542			if (ifp->if_flags & IFF_UP) {
543				int s = splimp();
544				if_down(ifp);
545				splx(s);
546			}
547			ifp->if_flags &= ~IFF_RUNNING;
548			break;
549		}
550		p = ifunit(vlr.vlr_parent);
551		if (p == 0) {
552			error = ENOENT;
553			break;
554		}
555		error = vlan_config(ifv, p);
556		if (error)
557			break;
558		ifv->ifv_tag = vlr.vlr_tag;
559		ifp->if_flags |= IFF_RUNNING;
560		break;
561
562	case SIOCGETVLAN:
563		bzero(&vlr, sizeof vlr);
564		if (ifv->ifv_p) {
565			snprintf(vlr.vlr_parent, sizeof(vlr.vlr_parent),
566			    "%s%d", ifv->ifv_p->if_name, ifv->ifv_p->if_unit);
567			vlr.vlr_tag = ifv->ifv_tag;
568		}
569		error = copyout(&vlr, ifr->ifr_data, sizeof vlr);
570		break;
571
572	case SIOCSIFFLAGS:
573		/*
574		 * We don't support promiscuous mode
575		 * right now because it would require help from the
576		 * underlying drivers, which hasn't been implemented.
577		 */
578		if (ifr->ifr_flags & (IFF_PROMISC)) {
579			ifp->if_flags &= ~(IFF_PROMISC);
580			error = EINVAL;
581		}
582		break;
583	case SIOCADDMULTI:
584	case SIOCDELMULTI:
585		error = vlan_setmulti(ifp);
586		break;
587	default:
588		error = EINVAL;
589	}
590	return error;
591}
592