if_tap.c revision 226500
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 226500 2011-10-18 08:09:44Z ed $
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_NEEDMINOR,
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			tapuponopen = 0;    /* IFF_UP on open() */
155static int			tapdclone = 1;	/* enable devfs cloning */
156static SLIST_HEAD(, tap_softc)	taphead;             /* first device */
157static struct clonedevs 	*tapclones;
158
159MALLOC_DECLARE(M_TAP);
160MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
161SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
162
163SYSCTL_DECL(_net_link);
164SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
165    "Ethernet tunnel software network interface");
166SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
167	"Allow user to open /dev/tap (based on node permissions)");
168SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
169	"Bring interface up when /dev/tap is opened");
170SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0,
171	"Enably legacy devfs interface creation");
172SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
173
174TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone);
175
176DEV_MODULE(if_tap, tapmodevent, NULL);
177
178static int
179tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
180{
181	struct cdev *dev;
182	int i;
183	int extra;
184
185	if (strcmp(ifc->ifc_name, VMNET) == 0)
186		extra = VMNET_DEV_MASK;
187	else
188		extra = 0;
189
190	/* find any existing device, or allocate new unit number */
191	i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, extra);
192	if (i) {
193		dev = make_dev(&tap_cdevsw, unit | extra,
194		     UID_ROOT, GID_WHEEL, 0600, "%s%d", ifc->ifc_name, unit);
195	}
196
197	tapcreate(dev);
198	return (0);
199}
200
201/* vmnet devices are tap devices in disguise */
202static int
203vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
204{
205	return tap_clone_create(ifc, unit, params);
206}
207
208static void
209tap_destroy(struct tap_softc *tp)
210{
211	struct ifnet *ifp = tp->tap_ifp;
212
213	/* Unlocked read. */
214	KASSERT(!(tp->tap_flags & TAP_OPEN),
215		("%s flags is out of sync", ifp->if_xname));
216
217	seldrain(&tp->tap_rsel);
218	knlist_destroy(&tp->tap_rsel.si_note);
219	destroy_dev(tp->tap_dev);
220	ether_ifdetach(ifp);
221	if_free_type(ifp, IFT_ETHER);
222
223	mtx_destroy(&tp->tap_mtx);
224	free(tp, M_TAP);
225}
226
227static void
228tap_clone_destroy(struct ifnet *ifp)
229{
230	struct tap_softc *tp = ifp->if_softc;
231
232	mtx_lock(&tapmtx);
233	SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
234	mtx_unlock(&tapmtx);
235	tap_destroy(tp);
236}
237
238/* vmnet devices are tap devices in disguise */
239static void
240vmnet_clone_destroy(struct ifnet *ifp)
241{
242	tap_clone_destroy(ifp);
243}
244
245/*
246 * tapmodevent
247 *
248 * module event handler
249 */
250static int
251tapmodevent(module_t mod, int type, void *data)
252{
253	static eventhandler_tag	 eh_tag = NULL;
254	struct tap_softc	*tp = NULL;
255	struct ifnet		*ifp = NULL;
256
257	switch (type) {
258	case MOD_LOAD:
259
260		/* intitialize device */
261
262		mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
263		SLIST_INIT(&taphead);
264
265		clone_setup(&tapclones);
266		eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
267		if (eh_tag == NULL) {
268			clone_cleanup(&tapclones);
269			mtx_destroy(&tapmtx);
270			return (ENOMEM);
271		}
272		if_clone_attach(&tap_cloner);
273		if_clone_attach(&vmnet_cloner);
274		return (0);
275
276	case MOD_UNLOAD:
277		/*
278		 * The EBUSY algorithm here can't quite atomically
279		 * guarantee that this is race-free since we have to
280		 * release the tap mtx to deregister the clone handler.
281		 */
282		mtx_lock(&tapmtx);
283		SLIST_FOREACH(tp, &taphead, tap_next) {
284			mtx_lock(&tp->tap_mtx);
285			if (tp->tap_flags & TAP_OPEN) {
286				mtx_unlock(&tp->tap_mtx);
287				mtx_unlock(&tapmtx);
288				return (EBUSY);
289			}
290			mtx_unlock(&tp->tap_mtx);
291		}
292		mtx_unlock(&tapmtx);
293
294		EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
295		if_clone_detach(&tap_cloner);
296		if_clone_detach(&vmnet_cloner);
297		drain_dev_clone_events();
298
299		mtx_lock(&tapmtx);
300		while ((tp = SLIST_FIRST(&taphead)) != NULL) {
301			SLIST_REMOVE_HEAD(&taphead, tap_next);
302			mtx_unlock(&tapmtx);
303
304			ifp = tp->tap_ifp;
305
306			TAPDEBUG("detaching %s\n", ifp->if_xname);
307
308			tap_destroy(tp);
309			mtx_lock(&tapmtx);
310		}
311		mtx_unlock(&tapmtx);
312		clone_cleanup(&tapclones);
313
314		mtx_destroy(&tapmtx);
315
316		break;
317
318	default:
319		return (EOPNOTSUPP);
320	}
321
322	return (0);
323} /* tapmodevent */
324
325
326/*
327 * DEVFS handler
328 *
329 * We need to support two kind of devices - tap and vmnet
330 */
331static void
332tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
333{
334	char		devname[SPECNAMELEN + 1];
335	int		i, unit, append_unit;
336	int		extra;
337
338	if (*dev != NULL)
339		return;
340
341	if (!tapdclone ||
342	    (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
343		return;
344
345	unit = 0;
346	append_unit = 0;
347	extra = 0;
348
349	/* We're interested in only tap/vmnet devices. */
350	if (strcmp(name, TAP) == 0) {
351		unit = -1;
352	} else if (strcmp(name, VMNET) == 0) {
353		unit = -1;
354		extra = VMNET_DEV_MASK;
355	} else if (dev_stdclone(name, NULL, TAP, &unit) != 1) {
356		if (dev_stdclone(name, NULL, VMNET, &unit) != 1) {
357			return;
358		} else {
359			extra = VMNET_DEV_MASK;
360		}
361	}
362
363	if (unit == -1)
364		append_unit = 1;
365
366	/* find any existing device, or allocate new unit number */
367	i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
368	if (i) {
369		if (append_unit) {
370			/*
371			 * We were passed 'tun' or 'tap', with no unit specified
372			 * so we'll need to append it now.
373			 */
374			namelen = snprintf(devname, sizeof(devname), "%s%d", name,
375			    unit);
376			name = devname;
377		}
378
379		*dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra,
380		     cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
381	}
382
383	if_clone_create(name, namelen, NULL);
384} /* tapclone */
385
386
387/*
388 * tapcreate
389 *
390 * to create interface
391 */
392static void
393tapcreate(struct cdev *dev)
394{
395	struct ifnet		*ifp = NULL;
396	struct tap_softc	*tp = NULL;
397	unsigned short		 macaddr_hi;
398	uint32_t		 macaddr_mid;
399	int			 unit;
400	char			*name = NULL;
401	u_char			eaddr[6];
402
403	dev->si_flags &= ~SI_CHEAPCLONE;
404
405	/* allocate driver storage and create device */
406	tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
407	mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
408	mtx_lock(&tapmtx);
409	SLIST_INSERT_HEAD(&taphead, tp, tap_next);
410	mtx_unlock(&tapmtx);
411
412	unit = dev2unit(dev);
413
414	/* select device: tap or vmnet */
415	if (unit & VMNET_DEV_MASK) {
416		name = VMNET;
417		tp->tap_flags |= TAP_VMNET;
418	} else
419		name = TAP;
420
421	unit &= TAPMAXUNIT;
422
423	TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev));
424
425	/* generate fake MAC address: 00 bd xx xx xx unit_no */
426	macaddr_hi = htons(0x00bd);
427	macaddr_mid = (uint32_t) ticks;
428	bcopy(&macaddr_hi, eaddr, sizeof(short));
429	bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
430	eaddr[5] = (u_char)unit;
431
432	/* fill the rest and attach interface */
433	ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
434	if (ifp == NULL)
435		panic("%s%d: can not if_alloc()", name, unit);
436	ifp->if_softc = tp;
437	if_initname(ifp, name, unit);
438	ifp->if_init = tapifinit;
439	ifp->if_start = tapifstart;
440	ifp->if_ioctl = tapifioctl;
441	ifp->if_mtu = ETHERMTU;
442	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
443	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
444	ifp->if_capabilities |= IFCAP_LINKSTATE;
445	ifp->if_capenable |= IFCAP_LINKSTATE;
446
447	dev->si_drv1 = tp;
448	tp->tap_dev = dev;
449
450	ether_ifattach(ifp, eaddr);
451
452	mtx_lock(&tp->tap_mtx);
453	tp->tap_flags |= TAP_INITED;
454	mtx_unlock(&tp->tap_mtx);
455
456	knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
457
458	TAPDEBUG("interface %s is created. minor = %#x\n",
459		ifp->if_xname, dev2unit(dev));
460} /* tapcreate */
461
462
463/*
464 * tapopen
465 *
466 * to open tunnel. must be superuser
467 */
468static int
469tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
470{
471	struct tap_softc	*tp = NULL;
472	struct ifnet		*ifp = NULL;
473	int			 error;
474
475	if (tapuopen == 0) {
476		error = priv_check(td, PRIV_NET_TAP);
477		if (error)
478			return (error);
479	}
480
481	if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
482		return (ENXIO);
483
484	tp = dev->si_drv1;
485
486	mtx_lock(&tp->tap_mtx);
487	if (tp->tap_flags & TAP_OPEN) {
488		mtx_unlock(&tp->tap_mtx);
489		return (EBUSY);
490	}
491
492	bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
493	tp->tap_pid = td->td_proc->p_pid;
494	tp->tap_flags |= TAP_OPEN;
495	ifp = tp->tap_ifp;
496
497	ifp->if_drv_flags |= IFF_DRV_RUNNING;
498	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
499	if (tapuponopen)
500		ifp->if_flags |= IFF_UP;
501	if_link_state_change(ifp, LINK_STATE_UP);
502	mtx_unlock(&tp->tap_mtx);
503
504	TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev));
505
506	return (0);
507} /* tapopen */
508
509
510/*
511 * tapclose
512 *
513 * close the device - mark i/f down & delete routing info
514 */
515static int
516tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
517{
518	struct ifaddr		*ifa;
519	struct tap_softc	*tp = dev->si_drv1;
520	struct ifnet		*ifp = tp->tap_ifp;
521
522	/* junk all pending output */
523	mtx_lock(&tp->tap_mtx);
524	IF_DRAIN(&ifp->if_snd);
525
526	/*
527	 * do not bring the interface down, and do not anything with
528	 * interface, if we are in VMnet mode. just close the device.
529	 */
530
531	if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
532		mtx_unlock(&tp->tap_mtx);
533		if_down(ifp);
534		mtx_lock(&tp->tap_mtx);
535		if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
536			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
537			mtx_unlock(&tp->tap_mtx);
538			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
539				rtinit(ifa, (int)RTM_DELETE, 0);
540			}
541			if_purgeaddrs(ifp);
542			mtx_lock(&tp->tap_mtx);
543		}
544	}
545
546	if_link_state_change(ifp, LINK_STATE_DOWN);
547	funsetown(&tp->tap_sigio);
548	selwakeuppri(&tp->tap_rsel, PZERO+1);
549	KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
550
551	tp->tap_flags &= ~TAP_OPEN;
552	tp->tap_pid = 0;
553	mtx_unlock(&tp->tap_mtx);
554
555	TAPDEBUG("%s is closed. minor = %#x\n",
556		ifp->if_xname, dev2unit(dev));
557
558	return (0);
559} /* tapclose */
560
561
562/*
563 * tapifinit
564 *
565 * network interface initialization function
566 */
567static void
568tapifinit(void *xtp)
569{
570	struct tap_softc	*tp = (struct tap_softc *)xtp;
571	struct ifnet		*ifp = tp->tap_ifp;
572
573	TAPDEBUG("initializing %s\n", ifp->if_xname);
574
575	mtx_lock(&tp->tap_mtx);
576	ifp->if_drv_flags |= IFF_DRV_RUNNING;
577	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
578	mtx_unlock(&tp->tap_mtx);
579
580	/* attempt to start output */
581	tapifstart(ifp);
582} /* tapifinit */
583
584
585/*
586 * tapifioctl
587 *
588 * Process an ioctl request on network interface
589 */
590static int
591tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
592{
593	struct tap_softc	*tp = ifp->if_softc;
594	struct ifreq		*ifr = (struct ifreq *)data;
595	struct ifstat		*ifs = NULL;
596	int			 dummy;
597
598	switch (cmd) {
599		case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
600		case SIOCADDMULTI:
601		case SIOCDELMULTI:
602			break;
603
604		case SIOCSIFMTU:
605			ifp->if_mtu = ifr->ifr_mtu;
606			break;
607
608		case SIOCGIFSTATUS:
609			ifs = (struct ifstat *)data;
610			dummy = strlen(ifs->ascii);
611			mtx_lock(&tp->tap_mtx);
612			if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
613				snprintf(ifs->ascii + dummy,
614					sizeof(ifs->ascii) - dummy,
615					"\tOpened by PID %d\n", tp->tap_pid);
616			mtx_unlock(&tp->tap_mtx);
617			break;
618
619		default:
620			return (ether_ioctl(ifp, cmd, data));
621			/* NOT REACHED */
622	}
623
624	return (0);
625} /* tapifioctl */
626
627
628/*
629 * tapifstart
630 *
631 * queue packets from higher level ready to put out
632 */
633static void
634tapifstart(struct ifnet *ifp)
635{
636	struct tap_softc	*tp = ifp->if_softc;
637
638	TAPDEBUG("%s starting\n", ifp->if_xname);
639
640	/*
641	 * do not junk pending output if we are in VMnet mode.
642	 * XXX: can this do any harm because of queue overflow?
643	 */
644
645	mtx_lock(&tp->tap_mtx);
646	if (((tp->tap_flags & TAP_VMNET) == 0) &&
647	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
648		struct mbuf *m;
649
650		/* Unlocked read. */
651		TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
652		    tp->tap_flags);
653
654		for (;;) {
655			IF_DEQUEUE(&ifp->if_snd, m);
656			if (m != NULL) {
657				m_freem(m);
658				ifp->if_oerrors++;
659			} else
660				break;
661		}
662		mtx_unlock(&tp->tap_mtx);
663
664		return;
665	}
666
667	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
668
669	if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
670		if (tp->tap_flags & TAP_RWAIT) {
671			tp->tap_flags &= ~TAP_RWAIT;
672			wakeup(tp);
673		}
674
675		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
676			mtx_unlock(&tp->tap_mtx);
677			pgsigio(&tp->tap_sigio, SIGIO, 0);
678			mtx_lock(&tp->tap_mtx);
679		}
680
681		selwakeuppri(&tp->tap_rsel, PZERO+1);
682		KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
683		ifp->if_opackets ++; /* obytes are counted in ether_output */
684	}
685
686	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
687	mtx_unlock(&tp->tap_mtx);
688} /* tapifstart */
689
690
691/*
692 * tapioctl
693 *
694 * the cdevsw interface is now pretty minimal
695 */
696static int
697tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
698{
699	struct tap_softc	*tp = dev->si_drv1;
700	struct ifnet		*ifp = tp->tap_ifp;
701	struct tapinfo		*tapp = NULL;
702	int			 f;
703#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
704    defined(COMPAT_FREEBSD4)
705	int			 ival;
706#endif
707
708	switch (cmd) {
709		case TAPSIFINFO:
710			tapp = (struct tapinfo *)data;
711			mtx_lock(&tp->tap_mtx);
712			ifp->if_mtu = tapp->mtu;
713			ifp->if_type = tapp->type;
714			ifp->if_baudrate = tapp->baudrate;
715			mtx_unlock(&tp->tap_mtx);
716			break;
717
718		case TAPGIFINFO:
719			tapp = (struct tapinfo *)data;
720			mtx_lock(&tp->tap_mtx);
721			tapp->mtu = ifp->if_mtu;
722			tapp->type = ifp->if_type;
723			tapp->baudrate = ifp->if_baudrate;
724			mtx_unlock(&tp->tap_mtx);
725			break;
726
727		case TAPSDEBUG:
728			tapdebug = *(int *)data;
729			break;
730
731		case TAPGDEBUG:
732			*(int *)data = tapdebug;
733			break;
734
735		case TAPGIFNAME: {
736			struct ifreq	*ifr = (struct ifreq *) data;
737
738			strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
739			} break;
740
741		case FIONBIO:
742			break;
743
744		case FIOASYNC:
745			mtx_lock(&tp->tap_mtx);
746			if (*(int *)data)
747				tp->tap_flags |= TAP_ASYNC;
748			else
749				tp->tap_flags &= ~TAP_ASYNC;
750			mtx_unlock(&tp->tap_mtx);
751			break;
752
753		case FIONREAD:
754			if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
755				struct mbuf *mb;
756
757				IFQ_LOCK(&ifp->if_snd);
758				IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
759				for (*(int *)data = 0; mb != NULL;
760				     mb = mb->m_next)
761					*(int *)data += mb->m_len;
762				IFQ_UNLOCK(&ifp->if_snd);
763			} else
764				*(int *)data = 0;
765			break;
766
767		case FIOSETOWN:
768			return (fsetown(*(int *)data, &tp->tap_sigio));
769
770		case FIOGETOWN:
771			*(int *)data = fgetown(&tp->tap_sigio);
772			return (0);
773
774		/* this is deprecated, FIOSETOWN should be used instead */
775		case TIOCSPGRP:
776			return (fsetown(-(*(int *)data), &tp->tap_sigio));
777
778		/* this is deprecated, FIOGETOWN should be used instead */
779		case TIOCGPGRP:
780			*(int *)data = -fgetown(&tp->tap_sigio);
781			return (0);
782
783		/* VMware/VMnet port ioctl's */
784
785#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
786    defined(COMPAT_FREEBSD4)
787		case _IO('V', 0):
788			ival = IOCPARM_IVAL(data);
789			data = (caddr_t)&ival;
790			/* FALLTHROUGH */
791#endif
792		case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
793			f = *(int *)data;
794			f &= 0x0fff;
795			f &= ~IFF_CANTCHANGE;
796			f |= IFF_UP;
797
798			mtx_lock(&tp->tap_mtx);
799			ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
800			mtx_unlock(&tp->tap_mtx);
801			break;
802
803		case OSIOCGIFADDR:	/* get MAC address of the remote side */
804		case SIOCGIFADDR:
805			mtx_lock(&tp->tap_mtx);
806			bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
807			mtx_unlock(&tp->tap_mtx);
808			break;
809
810		case SIOCSIFADDR:	/* set MAC address of the remote side */
811			mtx_lock(&tp->tap_mtx);
812			bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
813			mtx_unlock(&tp->tap_mtx);
814			break;
815
816		default:
817			return (ENOTTY);
818	}
819	return (0);
820} /* tapioctl */
821
822
823/*
824 * tapread
825 *
826 * the cdevsw read interface - reads a packet at a time, or at
827 * least as much of a packet as can be read
828 */
829static int
830tapread(struct cdev *dev, struct uio *uio, int flag)
831{
832	struct tap_softc	*tp = dev->si_drv1;
833	struct ifnet		*ifp = tp->tap_ifp;
834	struct mbuf		*m = NULL;
835	int			 error = 0, len;
836
837	TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev));
838
839	mtx_lock(&tp->tap_mtx);
840	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
841		mtx_unlock(&tp->tap_mtx);
842
843		/* Unlocked read. */
844		TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
845			ifp->if_xname, dev2unit(dev), tp->tap_flags);
846
847		return (EHOSTDOWN);
848	}
849
850	tp->tap_flags &= ~TAP_RWAIT;
851
852	/* sleep until we get a packet */
853	do {
854		IF_DEQUEUE(&ifp->if_snd, m);
855
856		if (m == NULL) {
857			if (flag & O_NONBLOCK) {
858				mtx_unlock(&tp->tap_mtx);
859				return (EWOULDBLOCK);
860			}
861
862			tp->tap_flags |= TAP_RWAIT;
863			error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
864			    "taprd", 0);
865			if (error) {
866				mtx_unlock(&tp->tap_mtx);
867				return (error);
868			}
869		}
870	} while (m == NULL);
871	mtx_unlock(&tp->tap_mtx);
872
873	/* feed packet to bpf */
874	BPF_MTAP(ifp, m);
875
876	/* xfer packet to user space */
877	while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
878		len = min(uio->uio_resid, m->m_len);
879		if (len == 0)
880			break;
881
882		error = uiomove(mtod(m, void *), len, uio);
883		m = m_free(m);
884	}
885
886	if (m != NULL) {
887		TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
888			dev2unit(dev));
889		m_freem(m);
890	}
891
892	return (error);
893} /* tapread */
894
895
896/*
897 * tapwrite
898 *
899 * the cdevsw write interface - an atomic write is a packet - or else!
900 */
901static int
902tapwrite(struct cdev *dev, struct uio *uio, int flag)
903{
904	struct ether_header	*eh;
905	struct tap_softc	*tp = dev->si_drv1;
906	struct ifnet		*ifp = tp->tap_ifp;
907	struct mbuf		*m;
908
909	TAPDEBUG("%s writting, minor = %#x\n",
910		ifp->if_xname, dev2unit(dev));
911
912	if (uio->uio_resid == 0)
913		return (0);
914
915	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
916		TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n",
917			ifp->if_xname, uio->uio_resid, dev2unit(dev));
918
919		return (EIO);
920	}
921
922	if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN,
923	    M_PKTHDR)) == NULL) {
924		ifp->if_ierrors ++;
925		return (ENOBUFS);
926	}
927
928	m->m_pkthdr.rcvif = ifp;
929
930	/*
931	 * Only pass a unicast frame to ether_input(), if it would actually
932	 * have been received by non-virtual hardware.
933	 */
934	if (m->m_len < sizeof(struct ether_header)) {
935		m_freem(m);
936		return (0);
937	}
938	eh = mtod(m, struct ether_header *);
939
940	if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
941	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
942	    bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
943		m_freem(m);
944		return (0);
945	}
946
947	/* Pass packet up to parent. */
948	(*ifp->if_input)(ifp, m);
949	ifp->if_ipackets ++; /* ibytes are counted in parent */
950
951	return (0);
952} /* tapwrite */
953
954
955/*
956 * tappoll
957 *
958 * the poll interface, this is only useful on reads
959 * really. the write detect always returns true, write never blocks
960 * anyway, it either accepts the packet or drops it
961 */
962static int
963tappoll(struct cdev *dev, int events, struct thread *td)
964{
965	struct tap_softc	*tp = dev->si_drv1;
966	struct ifnet		*ifp = tp->tap_ifp;
967	int			 revents = 0;
968
969	TAPDEBUG("%s polling, minor = %#x\n",
970		ifp->if_xname, dev2unit(dev));
971
972	if (events & (POLLIN | POLLRDNORM)) {
973		IFQ_LOCK(&ifp->if_snd);
974		if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
975			TAPDEBUG("%s have data in queue. len = %d, " \
976				"minor = %#x\n", ifp->if_xname,
977				ifp->if_snd.ifq_len, dev2unit(dev));
978
979			revents |= (events & (POLLIN | POLLRDNORM));
980		} else {
981			TAPDEBUG("%s waiting for data, minor = %#x\n",
982				ifp->if_xname, dev2unit(dev));
983
984			selrecord(td, &tp->tap_rsel);
985		}
986		IFQ_UNLOCK(&ifp->if_snd);
987	}
988
989	if (events & (POLLOUT | POLLWRNORM))
990		revents |= (events & (POLLOUT | POLLWRNORM));
991
992	return (revents);
993} /* tappoll */
994
995
996/*
997 * tap_kqfilter
998 *
999 * support for kevent() system call
1000 */
1001static int
1002tapkqfilter(struct cdev *dev, struct knote *kn)
1003{
1004	struct tap_softc	*tp = dev->si_drv1;
1005	struct ifnet		*ifp = tp->tap_ifp;
1006
1007	switch (kn->kn_filter) {
1008	case EVFILT_READ:
1009		TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1010			ifp->if_xname, dev2unit(dev));
1011		kn->kn_fop = &tap_read_filterops;
1012		break;
1013
1014	case EVFILT_WRITE:
1015		TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1016			ifp->if_xname, dev2unit(dev));
1017		kn->kn_fop = &tap_write_filterops;
1018		break;
1019
1020	default:
1021		TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1022			ifp->if_xname, dev2unit(dev));
1023		return (EINVAL);
1024		/* NOT REACHED */
1025	}
1026
1027	kn->kn_hook = tp;
1028	knlist_add(&tp->tap_rsel.si_note, kn, 0);
1029
1030	return (0);
1031} /* tapkqfilter */
1032
1033
1034/*
1035 * tap_kqread
1036 *
1037 * Return true if there is data in the interface queue
1038 */
1039static int
1040tapkqread(struct knote *kn, long hint)
1041{
1042	int			 ret;
1043	struct tap_softc	*tp = kn->kn_hook;
1044	struct cdev		*dev = tp->tap_dev;
1045	struct ifnet		*ifp = tp->tap_ifp;
1046
1047	if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1048		TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1049			ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1050		ret = 1;
1051	} else {
1052		TAPDEBUG("%s waiting for data, minor = %#x\n",
1053			ifp->if_xname, dev2unit(dev));
1054		ret = 0;
1055	}
1056
1057	return (ret);
1058} /* tapkqread */
1059
1060
1061/*
1062 * tap_kqwrite
1063 *
1064 * Always can write. Return the MTU in kn->data
1065 */
1066static int
1067tapkqwrite(struct knote *kn, long hint)
1068{
1069	struct tap_softc	*tp = kn->kn_hook;
1070	struct ifnet		*ifp = tp->tap_ifp;
1071
1072	kn->kn_data = ifp->if_mtu;
1073
1074	return (1);
1075} /* tapkqwrite */
1076
1077
1078static void
1079tapkqdetach(struct knote *kn)
1080{
1081	struct tap_softc	*tp = kn->kn_hook;
1082
1083	knlist_remove(&tp->tap_rsel.si_note, kn, 0);
1084} /* tapkqdetach */
1085
1086