if_tap.c revision 166497 
1/*-
2 * Copyright (C) 1999-2000 by Maksim Yevmenkin <m_evmenkin@yahoo.com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * BASED ON:
27 * -------------------------------------------------------------------------
28 *
29 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk>
30 * Nottingham University 1987.
31 */
32
33/*
34 * $FreeBSD: head/sys/net/if_tap.c 166497 2007-02-04 16:32:46Z bms $
35 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
36 */
37
38#include "opt_compat.h"
39#include "opt_inet.h"
40
41#include <sys/param.h>
42#include <sys/conf.h>
43#include <sys/fcntl.h>
44#include <sys/filio.h>
45#include <sys/kernel.h>
46#include <sys/malloc.h>
47#include <sys/mbuf.h>
48#include <sys/module.h>
49#include <sys/poll.h>
50#include <sys/priv.h>
51#include <sys/proc.h>
52#include <sys/selinfo.h>
53#include <sys/signalvar.h>
54#include <sys/socket.h>
55#include <sys/sockio.h>
56#include <sys/sysctl.h>
57#include <sys/systm.h>
58#include <sys/ttycom.h>
59#include <sys/uio.h>
60#include <sys/queue.h>
61
62#include <net/bpf.h>
63#include <net/ethernet.h>
64#include <net/if.h>
65#include <net/if_clone.h>
66#include <net/if_dl.h>
67#include <net/route.h>
68#include <net/if_types.h>
69
70#include <netinet/in.h>
71
72#include <net/if_tapvar.h>
73#include <net/if_tap.h>
74
75
76#define CDEV_NAME	"tap"
77#define TAPDEBUG	if (tapdebug) printf
78
79#define TAP		"tap"
80#define VMNET		"vmnet"
81#define TAPMAXUNIT	0x7fff
82#define VMNET_DEV_MASK	CLONE_FLAG0
83
84/* module */
85static int		tapmodevent(module_t, int, void *);
86
87/* device */
88static void		tapclone(void *, struct ucred *, char *, int,
89			    struct cdev **);
90static void		tapcreate(struct cdev *);
91
92/* network interface */
93static void		tapifstart(struct ifnet *);
94static int		tapifioctl(struct ifnet *, u_long, caddr_t);
95static void		tapifinit(void *);
96
97static int		tap_clone_create(struct if_clone *, int, caddr_t);
98static void		tap_clone_destroy(struct ifnet *);
99static int		vmnet_clone_create(struct if_clone *, int, caddr_t);
100static void		vmnet_clone_destroy(struct ifnet *);
101
102IFC_SIMPLE_DECLARE(tap, 0);
103IFC_SIMPLE_DECLARE(vmnet, 0);
104
105/* character device */
106static d_open_t		tapopen;
107static d_close_t	tapclose;
108static d_read_t		tapread;
109static d_write_t	tapwrite;
110static d_ioctl_t	tapioctl;
111static d_poll_t		tappoll;
112static d_kqfilter_t	tapkqfilter;
113
114/* kqueue(2) */
115static int		tapkqread(struct knote *, long);
116static int		tapkqwrite(struct knote *, long);
117static void		tapkqdetach(struct knote *);
118
119static struct filterops	tap_read_filterops = {
120	.f_isfd =	1,
121	.f_attach =	NULL,
122	.f_detach =	tapkqdetach,
123	.f_event =	tapkqread,
124};
125
126static struct filterops	tap_write_filterops = {
127	.f_isfd =	1,
128	.f_attach =	NULL,
129	.f_detach =	tapkqdetach,
130	.f_event =	tapkqwrite,
131};
132
133static struct cdevsw	tap_cdevsw = {
134	.d_version =	D_VERSION,
135	.d_flags =	D_PSEUDO | D_NEEDGIANT,
136	.d_open =	tapopen,
137	.d_close =	tapclose,
138	.d_read =	tapread,
139	.d_write =	tapwrite,
140	.d_ioctl =	tapioctl,
141	.d_poll =	tappoll,
142	.d_name =	CDEV_NAME,
143	.d_kqfilter =	tapkqfilter,
144};
145
146/*
147 * All global variables in if_tap.c are locked with tapmtx, with the
148 * exception of tapdebug, which is accessed unlocked; tapclones is
149 * static at runtime.
150 */
151static struct mtx		tapmtx;
152static int			tapdebug = 0;        /* debug flag   */
153static int			tapuopen = 0;        /* allow user open() */
154static int			tapdclone = 1;	/* enable devfs cloning */
155static SLIST_HEAD(, tap_softc)	taphead;             /* first device */
156static struct clonedevs 	*tapclones;
157
158MALLOC_DECLARE(M_TAP);
159MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
160SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
161
162SYSCTL_DECL(_net_link);
163SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
164    "Ethernet tunnel software network interface");
165SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
166	"Allow user to open /dev/tap (based on node permissions)");
167SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0,
168	"Enably legacy devfs interface creation");
169SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
170
171TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone);
172
173DEV_MODULE(if_tap, tapmodevent, NULL);
174
175static int
176tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
177{
178	struct cdev *dev;
179	int i;
180	int extra;
181
182	if (strcmp(ifc->ifc_name, VMNET) == 0)
183		extra = VMNET_DEV_MASK;
184	else
185		extra = 0;
186
187	/* find any existing device, or allocate new unit number */
188	i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, extra);
189	if (i) {
190		dev = make_dev(&tap_cdevsw, unit2minor(unit | extra),
191		     UID_ROOT, GID_WHEEL, 0600, "%s%d", ifc->ifc_name, unit);
192		if (dev != NULL) {
193			dev_ref(dev);
194			dev->si_flags |= SI_CHEAPCLONE;
195		}
196	}
197
198	tapcreate(dev);
199	return (0);
200}
201
202/* vmnet devices are tap devices in disguise */
203static int
204vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
205{
206	return tap_clone_create(ifc, unit, params);
207}
208
209static void
210tap_destroy(struct tap_softc *tp)
211{
212	struct ifnet *ifp = tp->tap_ifp;
213	int s;
214
215	/* Unlocked read. */
216	KASSERT(!(tp->tap_flags & TAP_OPEN),
217		("%s flags is out of sync", ifp->if_xname));
218
219	knlist_destroy(&tp->tap_rsel.si_note);
220	destroy_dev(tp->tap_dev);
221	s = splimp();
222	ether_ifdetach(ifp);
223	if_free_type(ifp, IFT_ETHER);
224	splx(s);
225
226	mtx_destroy(&tp->tap_mtx);
227	free(tp, M_TAP);
228}
229
230static void
231tap_clone_destroy(struct ifnet *ifp)
232{
233	struct tap_softc *tp = ifp->if_softc;
234
235	mtx_lock(&tapmtx);
236	SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
237	mtx_unlock(&tapmtx);
238	tap_destroy(tp);
239}
240
241/* vmnet devices are tap devices in disguise */
242static void
243vmnet_clone_destroy(struct ifnet *ifp)
244{
245	tap_clone_destroy(ifp);
246}
247
248/*
249 * tapmodevent
250 *
251 * module event handler
252 */
253static int
254tapmodevent(module_t mod, int type, void *data)
255{
256	static eventhandler_tag	 eh_tag = NULL;
257	struct tap_softc	*tp = NULL;
258	struct ifnet		*ifp = NULL;
259
260	switch (type) {
261	case MOD_LOAD:
262
263		/* intitialize device */
264
265		mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
266		SLIST_INIT(&taphead);
267
268		clone_setup(&tapclones);
269		eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
270		if (eh_tag == NULL) {
271			clone_cleanup(&tapclones);
272			mtx_destroy(&tapmtx);
273			return (ENOMEM);
274		}
275		if_clone_attach(&tap_cloner);
276		if_clone_attach(&vmnet_cloner);
277		return (0);
278
279	case MOD_UNLOAD:
280		/*
281		 * The EBUSY algorithm here can't quite atomically
282		 * guarantee that this is race-free since we have to
283		 * release the tap mtx to deregister the clone handler.
284		 */
285		mtx_lock(&tapmtx);
286		SLIST_FOREACH(tp, &taphead, tap_next) {
287			mtx_lock(&tp->tap_mtx);
288			if (tp->tap_flags & TAP_OPEN) {
289				mtx_unlock(&tp->tap_mtx);
290				mtx_unlock(&tapmtx);
291				return (EBUSY);
292			}
293			mtx_unlock(&tp->tap_mtx);
294		}
295		mtx_unlock(&tapmtx);
296
297		EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
298		if_clone_detach(&tap_cloner);
299		if_clone_detach(&vmnet_cloner);
300
301		mtx_lock(&tapmtx);
302		while ((tp = SLIST_FIRST(&taphead)) != NULL) {
303			SLIST_REMOVE_HEAD(&taphead, tap_next);
304			mtx_unlock(&tapmtx);
305
306			ifp = tp->tap_ifp;
307
308			TAPDEBUG("detaching %s\n", ifp->if_xname);
309
310			tap_destroy(tp);
311			mtx_lock(&tapmtx);
312		}
313		mtx_unlock(&tapmtx);
314		clone_cleanup(&tapclones);
315
316		mtx_destroy(&tapmtx);
317
318		break;
319
320	default:
321		return (EOPNOTSUPP);
322	}
323
324	return (0);
325} /* tapmodevent */
326
327
328/*
329 * DEVFS handler
330 *
331 * We need to support two kind of devices - tap and vmnet
332 */
333static void
334tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
335{
336	char		devname[SPECNAMELEN + 1];
337	int		i, unit, append_unit;
338	int		extra;
339
340	if (*dev != NULL)
341		return;
342
343	/*
344	 * If tap cloning is enabled, only the superuser can create
345	 * an interface.
346	 */
347	if (!tapdclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0)
348		return;
349
350	unit = 0;
351	append_unit = 0;
352	extra = 0;
353
354	/* We're interested in only tap/vmnet devices. */
355	if (strcmp(name, TAP) == 0) {
356		unit = -1;
357	} else if (strcmp(name, VMNET) == 0) {
358		unit = -1;
359		extra = VMNET_DEV_MASK;
360	} else if (dev_stdclone(name, NULL, TAP, &unit) != 1) {
361		if (dev_stdclone(name, NULL, VMNET, &unit) != 1) {
362			return;
363		} else {
364			extra = VMNET_DEV_MASK;
365		}
366	}
367
368	if (unit == -1)
369		append_unit = 1;
370
371	/* find any existing device, or allocate new unit number */
372	i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
373	if (i) {
374		if (append_unit) {
375			/*
376			 * We were passed 'tun' or 'tap', with no unit specified
377			 * so we'll need to append it now.
378			 */
379			namelen = snprintf(devname, sizeof(devname), "%s%d", name,
380			    unit);
381			name = devname;
382		}
383
384		*dev = make_dev(&tap_cdevsw, unit2minor(unit | extra),
385		     UID_ROOT, GID_WHEEL, 0600, "%s", name);
386		if (*dev != NULL) {
387			dev_ref(*dev);
388			(*dev)->si_flags |= SI_CHEAPCLONE;
389		}
390	}
391
392	if_clone_create(name, namelen, NULL);
393} /* tapclone */
394
395
396/*
397 * tapcreate
398 *
399 * to create interface
400 */
401static void
402tapcreate(struct cdev *dev)
403{
404	struct ifnet		*ifp = NULL;
405	struct tap_softc	*tp = NULL;
406	unsigned short		 macaddr_hi;
407	int			 unit, s;
408	char			*name = NULL;
409	u_char			eaddr[6];
410
411	dev->si_flags &= ~SI_CHEAPCLONE;
412
413	/* allocate driver storage and create device */
414	MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
415	mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
416	mtx_lock(&tapmtx);
417	SLIST_INSERT_HEAD(&taphead, tp, tap_next);
418	mtx_unlock(&tapmtx);
419
420	unit = dev2unit(dev);
421
422	/* select device: tap or vmnet */
423	if (unit & VMNET_DEV_MASK) {
424		name = VMNET;
425		tp->tap_flags |= TAP_VMNET;
426	} else
427		name = TAP;
428
429	unit &= TAPMAXUNIT;
430
431	TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, minor(dev));
432
433	/* generate fake MAC address: 00 bd xx xx xx unit_no */
434	macaddr_hi = htons(0x00bd);
435	bcopy(&macaddr_hi, eaddr, sizeof(short));
436	bcopy(&ticks, &eaddr[2], sizeof(long));
437	eaddr[5] = (u_char)unit;
438
439	/* fill the rest and attach interface */
440	ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
441	if (ifp == NULL)
442		panic("%s%d: can not if_alloc()", name, unit);
443	ifp->if_softc = tp;
444	if_initname(ifp, name, unit);
445	ifp->if_init = tapifinit;
446	ifp->if_start = tapifstart;
447	ifp->if_ioctl = tapifioctl;
448	ifp->if_mtu = ETHERMTU;
449	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
450	ifp->if_snd.ifq_maxlen = ifqmaxlen;
451
452	dev->si_drv1 = tp;
453	tp->tap_dev = dev;
454
455	s = splimp();
456	ether_ifattach(ifp, eaddr);
457	splx(s);
458
459	mtx_lock(&tp->tap_mtx);
460	tp->tap_flags |= TAP_INITED;
461	mtx_unlock(&tp->tap_mtx);
462
463	knlist_init(&tp->tap_rsel.si_note, NULL, NULL, NULL, NULL);
464
465	TAPDEBUG("interface %s is created. minor = %#x\n",
466		ifp->if_xname, minor(dev));
467} /* tapcreate */
468
469
470/*
471 * tapopen
472 *
473 * to open tunnel. must be superuser
474 */
475static int
476tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
477{
478	struct tap_softc	*tp = NULL;
479	struct ifnet		*ifp = NULL;
480	int			 error, s;
481
482	if (tapuopen == 0) {
483		error = priv_check(td, PRIV_NET_TAP);
484		if (error)
485			return (error);
486	}
487
488	if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
489		return (ENXIO);
490
491	tp = dev->si_drv1;
492
493	mtx_lock(&tp->tap_mtx);
494	if (tp->tap_flags & TAP_OPEN) {
495		mtx_unlock(&tp->tap_mtx);
496		return (EBUSY);
497	}
498
499	bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
500	tp->tap_pid = td->td_proc->p_pid;
501	tp->tap_flags |= TAP_OPEN;
502	ifp = tp->tap_ifp;
503	mtx_unlock(&tp->tap_mtx);
504
505	s = splimp();
506	ifp->if_drv_flags |= IFF_DRV_RUNNING;
507	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
508	splx(s);
509
510	TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, minor(dev));
511
512	return (0);
513} /* tapopen */
514
515
516/*
517 * tapclose
518 *
519 * close the device - mark i/f down & delete routing info
520 */
521static int
522tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
523{
524	struct ifaddr		*ifa;
525	struct tap_softc	*tp = dev->si_drv1;
526	struct ifnet		*ifp = tp->tap_ifp;
527	int			s;
528
529	/* junk all pending output */
530	IF_DRAIN(&ifp->if_snd);
531
532	/*
533	 * do not bring the interface down, and do not anything with
534	 * interface, if we are in VMnet mode. just close the device.
535	 */
536
537	mtx_lock(&tp->tap_mtx);
538	if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
539		mtx_unlock(&tp->tap_mtx);
540		s = splimp();
541		if_down(ifp);
542		if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
543			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
544				rtinit(ifa, (int)RTM_DELETE, 0);
545			}
546			if_purgeaddrs(ifp);
547			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
548		}
549		splx(s);
550	} else
551		mtx_unlock(&tp->tap_mtx);
552
553	funsetown(&tp->tap_sigio);
554	selwakeuppri(&tp->tap_rsel, PZERO+1);
555	KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0);
556
557	mtx_lock(&tp->tap_mtx);
558	tp->tap_flags &= ~TAP_OPEN;
559	tp->tap_pid = 0;
560	mtx_unlock(&tp->tap_mtx);
561
562	TAPDEBUG("%s is closed. minor = %#x\n",
563		ifp->if_xname, minor(dev));
564
565	return (0);
566} /* tapclose */
567
568
569/*
570 * tapifinit
571 *
572 * network interface initialization function
573 */
574static void
575tapifinit(void *xtp)
576{
577	struct tap_softc	*tp = (struct tap_softc *)xtp;
578	struct ifnet		*ifp = tp->tap_ifp;
579
580	TAPDEBUG("initializing %s\n", ifp->if_xname);
581
582	ifp->if_drv_flags |= IFF_DRV_RUNNING;
583	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
584
585	/* attempt to start output */
586	tapifstart(ifp);
587} /* tapifinit */
588
589
590/*
591 * tapifioctl
592 *
593 * Process an ioctl request on network interface
594 */
595static int
596tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
597{
598	struct tap_softc	*tp = ifp->if_softc;
599	struct ifstat		*ifs = NULL;
600	int			 s, dummy;
601
602	switch (cmd) {
603		case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
604		case SIOCADDMULTI:
605		case SIOCDELMULTI:
606			break;
607
608		case SIOCGIFSTATUS:
609			s = splimp();
610			ifs = (struct ifstat *)data;
611			dummy = strlen(ifs->ascii);
612			mtx_lock(&tp->tap_mtx);
613			if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
614				snprintf(ifs->ascii + dummy,
615					sizeof(ifs->ascii) - dummy,
616					"\tOpened by PID %d\n", tp->tap_pid);
617			mtx_unlock(&tp->tap_mtx);
618			splx(s);
619			break;
620
621		default:
622			s = splimp();
623			dummy = ether_ioctl(ifp, cmd, data);
624			splx(s);
625			return (dummy);
626			/* NOT REACHED */
627	}
628
629	return (0);
630} /* tapifioctl */
631
632
633/*
634 * tapifstart
635 *
636 * queue packets from higher level ready to put out
637 */
638static void
639tapifstart(struct ifnet *ifp)
640{
641	struct tap_softc	*tp = ifp->if_softc;
642	int			 s;
643
644	TAPDEBUG("%s starting\n", ifp->if_xname);
645
646	/*
647	 * do not junk pending output if we are in VMnet mode.
648	 * XXX: can this do any harm because of queue overflow?
649	 */
650
651	mtx_lock(&tp->tap_mtx);
652	if (((tp->tap_flags & TAP_VMNET) == 0) &&
653	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
654		struct mbuf	*m = NULL;
655
656		mtx_unlock(&tp->tap_mtx);
657
658		/* Unlocked read. */
659		TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
660		    tp->tap_flags);
661
662		s = splimp();
663		do {
664			IF_DEQUEUE(&ifp->if_snd, m);
665			if (m != NULL)
666				m_freem(m);
667			ifp->if_oerrors ++;
668		} while (m != NULL);
669		splx(s);
670
671		return;
672	}
673	mtx_unlock(&tp->tap_mtx);
674
675	s = splimp();
676	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
677
678	if (ifp->if_snd.ifq_len != 0) {
679		mtx_lock(&tp->tap_mtx);
680		if (tp->tap_flags & TAP_RWAIT) {
681			tp->tap_flags &= ~TAP_RWAIT;
682			wakeup(tp);
683		}
684
685		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
686			mtx_unlock(&tp->tap_mtx);
687			pgsigio(&tp->tap_sigio, SIGIO, 0);
688		} else
689			mtx_unlock(&tp->tap_mtx);
690
691		selwakeuppri(&tp->tap_rsel, PZERO+1);
692		KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0);
693		ifp->if_opackets ++; /* obytes are counted in ether_output */
694	}
695
696	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
697	splx(s);
698} /* tapifstart */
699
700
701/*
702 * tapioctl
703 *
704 * the cdevsw interface is now pretty minimal
705 */
706static int
707tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
708{
709	struct tap_softc	*tp = dev->si_drv1;
710	struct ifnet		*ifp = tp->tap_ifp;
711	struct tapinfo		*tapp = NULL;
712	int			 s;
713	int			 f;
714#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
715    defined(COMPAT_FREEBSD4)
716	int			 ival;
717#endif
718
719	switch (cmd) {
720		case TAPSIFINFO:
721			s = splimp();
722			tapp = (struct tapinfo *)data;
723			ifp->if_mtu = tapp->mtu;
724			ifp->if_type = tapp->type;
725			ifp->if_baudrate = tapp->baudrate;
726			splx(s);
727			break;
728
729		case TAPGIFINFO:
730			tapp = (struct tapinfo *)data;
731			tapp->mtu = ifp->if_mtu;
732			tapp->type = ifp->if_type;
733			tapp->baudrate = ifp->if_baudrate;
734			break;
735
736		case TAPSDEBUG:
737			tapdebug = *(int *)data;
738			break;
739
740		case TAPGDEBUG:
741			*(int *)data = tapdebug;
742			break;
743
744		case FIONBIO:
745			break;
746
747		case FIOASYNC:
748			s = splimp();
749			mtx_lock(&tp->tap_mtx);
750			if (*(int *)data)
751				tp->tap_flags |= TAP_ASYNC;
752			else
753				tp->tap_flags &= ~TAP_ASYNC;
754			mtx_unlock(&tp->tap_mtx);
755			splx(s);
756			break;
757
758		case FIONREAD:
759			s = splimp();
760			if (ifp->if_snd.ifq_head) {
761				struct mbuf	*mb = ifp->if_snd.ifq_head;
762
763				for(*(int *)data = 0;mb != NULL;mb = mb->m_next)
764					*(int *)data += mb->m_len;
765			} else
766				*(int *)data = 0;
767			splx(s);
768			break;
769
770		case FIOSETOWN:
771			return (fsetown(*(int *)data, &tp->tap_sigio));
772
773		case FIOGETOWN:
774			*(int *)data = fgetown(&tp->tap_sigio);
775			return (0);
776
777		/* this is deprecated, FIOSETOWN should be used instead */
778		case TIOCSPGRP:
779			return (fsetown(-(*(int *)data), &tp->tap_sigio));
780
781		/* this is deprecated, FIOGETOWN should be used instead */
782		case TIOCGPGRP:
783			*(int *)data = -fgetown(&tp->tap_sigio);
784			return (0);
785
786		/* VMware/VMnet port ioctl's */
787
788		case SIOCGIFFLAGS:	/* get ifnet flags */
789			bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
790			break;
791
792#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
793    defined(COMPAT_FREEBSD4)
794		case _IO('V', 0):
795			ival = IOCPARM_IVAL(data);
796			data = (caddr_t)&ival;
797			/* FALLTHROUGH */
798#endif
799		case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
800			f = *(int *)data;
801			f &= 0x0fff;
802			f &= ~IFF_CANTCHANGE;
803			f |= IFF_UP;
804
805			s = splimp();
806			ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
807			splx(s);
808			break;
809
810		case OSIOCGIFADDR:	/* get MAC address of the remote side */
811		case SIOCGIFADDR:
812			mtx_lock(&tp->tap_mtx);
813			bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
814			mtx_unlock(&tp->tap_mtx);
815			break;
816
817		case SIOCSIFADDR:	/* set MAC address of the remote side */
818			mtx_lock(&tp->tap_mtx);
819			bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
820			mtx_unlock(&tp->tap_mtx);
821			break;
822
823		default:
824			return (ENOTTY);
825	}
826	return (0);
827} /* tapioctl */
828
829
830/*
831 * tapread
832 *
833 * the cdevsw read interface - reads a packet at a time, or at
834 * least as much of a packet as can be read
835 */
836static int
837tapread(struct cdev *dev, struct uio *uio, int flag)
838{
839	struct tap_softc	*tp = dev->si_drv1;
840	struct ifnet		*ifp = tp->tap_ifp;
841	struct mbuf		*m = NULL;
842	int			 error = 0, len, s;
843
844	TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, minor(dev));
845
846	mtx_lock(&tp->tap_mtx);
847	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
848		mtx_unlock(&tp->tap_mtx);
849
850		/* Unlocked read. */
851		TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
852			ifp->if_xname, minor(dev), tp->tap_flags);
853
854		return (EHOSTDOWN);
855	}
856
857	tp->tap_flags &= ~TAP_RWAIT;
858	mtx_unlock(&tp->tap_mtx);
859
860	/* sleep until we get a packet */
861	do {
862		s = splimp();
863		IF_DEQUEUE(&ifp->if_snd, m);
864		splx(s);
865
866		if (m == NULL) {
867			if (flag & O_NONBLOCK)
868				return (EWOULDBLOCK);
869
870			mtx_lock(&tp->tap_mtx);
871			tp->tap_flags |= TAP_RWAIT;
872			mtx_unlock(&tp->tap_mtx);
873			error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0);
874			if (error)
875				return (error);
876		}
877	} while (m == NULL);
878
879	/* feed packet to bpf */
880	BPF_MTAP(ifp, m);
881
882	/* xfer packet to user space */
883	while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
884		len = min(uio->uio_resid, m->m_len);
885		if (len == 0)
886			break;
887
888		error = uiomove(mtod(m, void *), len, uio);
889		m = m_free(m);
890	}
891
892	if (m != NULL) {
893		TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
894			minor(dev));
895		m_freem(m);
896	}
897
898	return (error);
899} /* tapread */
900
901
902/*
903 * tapwrite
904 *
905 * the cdevsw write interface - an atomic write is a packet - or else!
906 */
907static int
908tapwrite(struct cdev *dev, struct uio *uio, int flag)
909{
910	struct ether_header	*eh;
911	struct tap_softc	*tp = dev->si_drv1;
912	struct ifnet		*ifp = tp->tap_ifp;
913	struct mbuf		*m;
914
915	TAPDEBUG("%s writting, minor = %#x\n",
916		ifp->if_xname, minor(dev));
917
918	if (uio->uio_resid == 0)
919		return (0);
920
921	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
922		TAPDEBUG("%s invalid packet len = %d, minor = %#x\n",
923			ifp->if_xname, uio->uio_resid, minor(dev));
924
925		return (EIO);
926	}
927
928	if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN,
929	    M_PKTHDR)) == NULL) {
930		ifp->if_ierrors ++;
931		return (ENOBUFS);
932	}
933
934	m->m_pkthdr.rcvif = ifp;
935
936	/*
937	 * Only pass a unicast frame to ether_input(), if it would actually
938	 * have been received by non-virtual hardware.
939	 */
940	if (m->m_len < sizeof(struct ether_header)) {
941		m_freem(m);
942		return (0);
943	}
944	eh = mtod(m, struct ether_header *);
945
946	if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
947	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
948	    bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
949		m_freem(m);
950		return (0);
951	}
952
953	/* Pass packet up to parent. */
954	(*ifp->if_input)(ifp, m);
955	ifp->if_ipackets ++; /* ibytes are counted in parent */
956
957	return (0);
958} /* tapwrite */
959
960
961/*
962 * tappoll
963 *
964 * the poll interface, this is only useful on reads
965 * really. the write detect always returns true, write never blocks
966 * anyway, it either accepts the packet or drops it
967 */
968static int
969tappoll(struct cdev *dev, int events, struct thread *td)
970{
971	struct tap_softc	*tp = dev->si_drv1;
972	struct ifnet		*ifp = tp->tap_ifp;
973	int			 s, revents = 0;
974
975	TAPDEBUG("%s polling, minor = %#x\n",
976		ifp->if_xname, minor(dev));
977
978	s = splimp();
979	if (events & (POLLIN | POLLRDNORM)) {
980		if (ifp->if_snd.ifq_len > 0) {
981			TAPDEBUG("%s have data in queue. len = %d, " \
982				"minor = %#x\n", ifp->if_xname,
983				ifp->if_snd.ifq_len, minor(dev));
984
985			revents |= (events & (POLLIN | POLLRDNORM));
986		} else {
987			TAPDEBUG("%s waiting for data, minor = %#x\n",
988				ifp->if_xname, minor(dev));
989
990			selrecord(td, &tp->tap_rsel);
991		}
992	}
993
994	if (events & (POLLOUT | POLLWRNORM))
995		revents |= (events & (POLLOUT | POLLWRNORM));
996
997	splx(s);
998	return (revents);
999} /* tappoll */
1000
1001
1002/*
1003 * tap_kqfilter
1004 *
1005 * support for kevent() system call
1006 */
1007static int
1008tapkqfilter(struct cdev *dev, struct knote *kn)
1009{
1010    	int			 s;
1011	struct tap_softc	*tp = dev->si_drv1;
1012	struct ifnet		*ifp = tp->tap_ifp;
1013
1014	s = splimp();
1015	switch (kn->kn_filter) {
1016	case EVFILT_READ:
1017		TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1018			ifp->if_xname, minor(dev));
1019		kn->kn_fop = &tap_read_filterops;
1020		break;
1021
1022	case EVFILT_WRITE:
1023		TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1024			ifp->if_xname, minor(dev));
1025		kn->kn_fop = &tap_write_filterops;
1026		break;
1027
1028	default:
1029		TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1030			ifp->if_xname, minor(dev));
1031		splx(s);
1032		return (EINVAL);
1033		/* NOT REACHED */
1034	}
1035	splx(s);
1036
1037	kn->kn_hook = (caddr_t) dev;
1038	knlist_add(&tp->tap_rsel.si_note, kn, 0);
1039
1040	return (0);
1041} /* tapkqfilter */
1042
1043
1044/*
1045 * tap_kqread
1046 *
1047 * Return true if there is data in the interface queue
1048 */
1049static int
1050tapkqread(struct knote *kn, long hint)
1051{
1052	int			 ret, s;
1053	struct cdev		*dev = (struct cdev *)(kn->kn_hook);
1054	struct tap_softc	*tp = dev->si_drv1;
1055	struct ifnet		*ifp = tp->tap_ifp;
1056
1057	s = splimp();
1058	if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1059		TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1060			ifp->if_xname, ifp->if_snd.ifq_len, minor(dev));
1061		ret = 1;
1062	} else {
1063		TAPDEBUG("%s waiting for data, minor = %#x\n",
1064			ifp->if_xname, minor(dev));
1065		ret = 0;
1066	}
1067	splx(s);
1068
1069	return (ret);
1070} /* tapkqread */
1071
1072
1073/*
1074 * tap_kqwrite
1075 *
1076 * Always can write. Return the MTU in kn->data
1077 */
1078static int
1079tapkqwrite(struct knote *kn, long hint)
1080{
1081	int			 s;
1082	struct tap_softc	*tp = ((struct cdev *) kn->kn_hook)->si_drv1;
1083	struct ifnet		*ifp = tp->tap_ifp;
1084
1085	s = splimp();
1086	kn->kn_data = ifp->if_mtu;
1087	splx(s);
1088
1089	return (1);
1090} /* tapkqwrite */
1091
1092
1093static void
1094tapkqdetach(struct knote *kn)
1095{
1096	struct tap_softc	*tp = ((struct cdev *) kn->kn_hook)->si_drv1;
1097
1098	knlist_remove(&tp->tap_rsel.si_note, kn, 0);
1099} /* tapkqdetach */
1100
1101