xref: /dragonfly/sys/net/tap/if_tap.c (revision 2d8a3be7) 
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: src/sys/net/if_tap.c,v 1.3.2.3 2002/04/14 21:41:48 luigi Exp $
35  * $DragonFly: src/sys/net/tap/if_tap.c,v 1.8 2003/08/26 20:49:49 rob Exp $
36  * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
37  */
38 
39 #include "opt_inet.h"
40 
41 #include <sys/param.h>
42 #include <sys/conf.h>
43 #include <sys/filedesc.h>
44 #include <sys/filio.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/poll.h>
49 #include <sys/proc.h>
50 #include <sys/signalvar.h>
51 #include <sys/socket.h>
52 #include <sys/sockio.h>
53 #include <sys/sysctl.h>
54 #include <sys/systm.h>
55 #include <sys/ttycom.h>
56 #include <sys/uio.h>
57 #include <sys/vnode.h>
58 
59 #include <net/bpf.h>
60 #include <net/ethernet.h>
61 #include <net/if.h>
62 #include <net/if_arp.h>
63 #include <net/route.h>
64 
65 #include <netinet/in.h>
66 
67 #include "if_tapvar.h"
68 #include "if_tap.h"
69 
70 
71 #define CDEV_NAME	"tap"
72 #define CDEV_MAJOR	149
73 #define TAPDEBUG	if (tapdebug) printf
74 
75 #define TAP		"tap"
76 #define VMNET		"vmnet"
77 #define VMNET_DEV_MASK	0x00010000
78 
79 /* module */
80 static int 		tapmodevent	(module_t, int, void *);
81 
82 /* device */
83 static void		tapcreate	(dev_t);
84 
85 /* network interface */
86 static void		tapifstart	(struct ifnet *);
87 static int		tapifioctl	(struct ifnet *, u_long, caddr_t);
88 static void		tapifinit	(void *);
89 
90 /* character device */
91 static d_open_t		tapopen;
92 static d_close_t	tapclose;
93 static d_read_t		tapread;
94 static d_write_t	tapwrite;
95 static d_ioctl_t	tapioctl;
96 static d_poll_t		tappoll;
97 
98 static struct cdevsw	tap_cdevsw = {
99 	/* dev name */	CDEV_NAME,
100 	/* dev major */	CDEV_MAJOR,
101 	/* flags */	0,
102 	/* port */	NULL,
103 	/* autoq */	0,
104 
105 	/* open */	tapopen,
106 	/* close */	tapclose,
107 	/* read */	tapread,
108 	/* write */	tapwrite,
109 	/* ioctl */	tapioctl,
110 	/* poll */	tappoll,
111 	/* mmap */	nommap,
112 	/* startegy */	nostrategy,
113 	/* dump */	nodump,
114 	/* psize */	nopsize
115 };
116 
117 static int		taprefcnt = 0;		/* module ref. counter   */
118 static int		taplastunit = -1;	/* max. open unit number */
119 static int		tapdebug = 0;		/* debug flag            */
120 
121 MALLOC_DECLARE(M_TAP);
122 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
123 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
124 DEV_MODULE(if_tap, tapmodevent, NULL);
125 
126 /*
127  * tapmodevent
128  *
129  * module event handler
130  */
131 static int
132 tapmodevent(mod, type, data)
133 	module_t	 mod;
134 	int		 type;
135 	void		*data;
136 {
137 	static int		 attached = 0;
138 	struct ifnet		*ifp = NULL;
139 	int			 unit, s;
140 
141 	switch (type) {
142 	case MOD_LOAD:
143 		if (attached)
144 			return (EEXIST);
145 
146 		cdevsw_add(&tap_cdevsw);
147 		attached = 1;
148 	break;
149 
150 	case MOD_UNLOAD:
151 		if (taprefcnt > 0)
152 			return (EBUSY);
153 
154 		cdevsw_remove(&tap_cdevsw);
155 
156 		unit = 0;
157 		while (unit <= taplastunit) {
158 			s = splimp();
159 			TAILQ_FOREACH(ifp, &ifnet, if_link)
160 				if ((strcmp(ifp->if_name, TAP) == 0) ||
161 				    (strcmp(ifp->if_name, VMNET) == 0))
162 					if (ifp->if_unit == unit)
163 						break;
164 			splx(s);
165 
166 			if (ifp != NULL) {
167 				struct tap_softc	*tp = ifp->if_softc;
168 
169 				TAPDEBUG("detaching %s%d. minor = %#x, " \
170 					"taplastunit = %d\n",
171 					ifp->if_name, unit, minor(tp->tap_dev),
172 					taplastunit);
173 
174 				s = splimp();
175 				ether_ifdetach(ifp, 1);
176 				splx(s);
177 				destroy_dev(tp->tap_dev);
178 				free(tp, M_TAP);
179 			}
180 			else
181 				unit ++;
182 		}
183 
184 		attached = 0;
185 	break;
186 
187 	default:
188 		return (EOPNOTSUPP);
189 	}
190 
191 	return (0);
192 } /* tapmodevent */
193 
194 
195 /*
196  * tapcreate
197  *
198  * to create interface
199  */
200 static void
201 tapcreate(dev)
202 	dev_t	dev;
203 {
204 	struct ifnet		*ifp = NULL;
205 	struct tap_softc	*tp = NULL;
206 	unsigned short		 macaddr_hi;
207 	int			 unit, s;
208 	char			*name = NULL;
209 
210 	/* allocate driver storage and create device */
211 	MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK);
212 	bzero(tp, sizeof(*tp));
213 
214 	/* select device: tap or vmnet */
215 	if (minor(dev) & VMNET_DEV_MASK) {
216 		name = VMNET;
217 		unit = lminor(dev) & 0xff;
218 		tp->tap_flags |= TAP_VMNET;
219 	}
220 	else {
221 		name = TAP;
222 		unit = lminor(dev);
223 	}
224 
225 	tp->tap_dev = make_dev(&tap_cdevsw, minor(dev), UID_ROOT, GID_WHEEL,
226 						0600, "%s%d", name, unit);
227 	tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
228 
229 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
230 	macaddr_hi = htons(0x00bd);
231 	bcopy(&macaddr_hi, &tp->arpcom.ac_enaddr[0], sizeof(short));
232 	bcopy(&ticks, &tp->arpcom.ac_enaddr[2], sizeof(long));
233 	tp->arpcom.ac_enaddr[5] = (u_char)unit;
234 
235 	/* fill the rest and attach interface */
236 	ifp = &tp->tap_if;
237 	ifp->if_softc = tp;
238 
239 	ifp->if_unit = unit;
240 	if (unit > taplastunit)
241 		taplastunit = unit;
242 
243 	ifp->if_name = name;
244 	ifp->if_init = tapifinit;
245 	ifp->if_output = ether_output;
246 	ifp->if_start = tapifstart;
247 	ifp->if_ioctl = tapifioctl;
248 	ifp->if_mtu = ETHERMTU;
249 	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
250 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
251 
252 	s = splimp();
253 	ether_ifattach(ifp, 1);
254 	splx(s);
255 
256 	tp->tap_flags |= TAP_INITED;
257 
258 	TAPDEBUG("interface %s%d created. minor = %#x\n",
259 			ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
260 } /* tapcreate */
261 
262 
263 /*
264  * tapopen
265  *
266  * to open tunnel. must be superuser
267  */
268 static int
269 tapopen(dev_t dev, int flag, int mode, d_thread_t *td)
270 {
271 	struct tap_softc	*tp = NULL;
272 	int			 error;
273 
274 	KKASSERT(p != NULL);
275 
276 	if ((error = suser(td)) != 0)
277 		return (error);
278 
279 	tp = dev->si_drv1;
280 	if (tp == NULL) {
281 		tapcreate(dev);
282 		tp = dev->si_drv1;
283 	}
284 
285 	if (tp->tap_flags & TAP_OPEN)
286 		return (EBUSY);
287 
288 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
289 
290 	tp->tap_td = td;
291 	tp->tap_flags |= TAP_OPEN;
292 	taprefcnt ++;
293 
294 	TAPDEBUG("%s%d is open. minor = %#x, refcnt = %d, taplastunit = %d\n",
295 		tp->tap_if.if_name, tp->tap_if.if_unit,
296 		minor(tp->tap_dev), taprefcnt, taplastunit);
297 
298 	return (0);
299 } /* tapopen */
300 
301 
302 /*
303  * tapclose
304  *
305  * close the device - mark i/f down & delete routing info
306  */
307 static int
308 tapclose(dev_t dev, int foo, int bar, d_thread_t *td)
309 {
310 	int			 s;
311 	struct tap_softc	*tp = dev->si_drv1;
312 	struct ifnet		*ifp = &tp->tap_if;
313 	struct mbuf		*m = NULL;
314 
315 	/* junk all pending output */
316 
317 	s = splimp();
318 	do {
319 		IF_DEQUEUE(&ifp->if_snd, m);
320 		if (m != NULL)
321 			m_freem(m);
322 	} while (m != NULL);
323 	splx(s);
324 
325 	/*
326 	 * do not bring the interface down, and do not anything with
327 	 * interface, if we are in VMnet mode. just close the device.
328 	 */
329 
330 	if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
331 		s = splimp();
332 		if_down(ifp);
333 		if (ifp->if_flags & IFF_RUNNING) {
334 			/* find internet addresses and delete routes */
335 			struct ifaddr	*ifa = NULL;
336 
337 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
338 				if (ifa->ifa_addr->sa_family == AF_INET) {
339 					rtinit(ifa, (int)RTM_DELETE, 0);
340 
341 					/* remove address from interface */
342 					bzero(ifa->ifa_addr,
343 						   sizeof(*(ifa->ifa_addr)));
344 					bzero(ifa->ifa_dstaddr,
345 						   sizeof(*(ifa->ifa_dstaddr)));
346 					bzero(ifa->ifa_netmask,
347 						   sizeof(*(ifa->ifa_netmask)));
348 				}
349 			}
350 
351 			ifp->if_flags &= ~IFF_RUNNING;
352 		}
353 		splx(s);
354 	}
355 
356 	funsetown(tp->tap_sigio);
357 	selwakeup(&tp->tap_rsel);
358 
359 	tp->tap_flags &= ~TAP_OPEN;
360 	tp->tap_td = NULL;
361 
362 	taprefcnt --;
363 	if (taprefcnt < 0) {
364 		taprefcnt = 0;
365 		printf("%s%d minor = %#x, refcnt = %d is out of sync. " \
366 			"set refcnt to 0\n", ifp->if_name, ifp->if_unit,
367 			minor(tp->tap_dev), taprefcnt);
368 	}
369 
370 	TAPDEBUG("%s%d is closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
371 			ifp->if_name, ifp->if_unit, minor(tp->tap_dev),
372 			taprefcnt, taplastunit);
373 
374 	return (0);
375 } /* tapclose */
376 
377 
378 /*
379  * tapifinit
380  *
381  * network interface initialization function
382  */
383 static void
384 tapifinit(xtp)
385 	void	*xtp;
386 {
387 	struct tap_softc	*tp = (struct tap_softc *)xtp;
388 	struct ifnet		*ifp = &tp->tap_if;
389 
390 	TAPDEBUG("initializing %s%d, minor = %#x\n",
391 			ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
392 
393 	ifp->if_flags |= IFF_RUNNING;
394 	ifp->if_flags &= ~IFF_OACTIVE;
395 
396 	/* attempt to start output */
397 	tapifstart(ifp);
398 } /* tapifinit */
399 
400 
401 /*
402  * tapifioctl
403  *
404  * Process an ioctl request on network interface
405  */
406 int
407 tapifioctl(ifp, cmd, data)
408 	struct ifnet	*ifp;
409 	u_long		 cmd;
410 	caddr_t		 data;
411 {
412 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
413 	struct ifstat		*ifs = NULL;
414 	int			 s, dummy;
415 
416 	switch (cmd) {
417 		case SIOCSIFADDR:
418 		case SIOCGIFADDR:
419 		case SIOCSIFMTU:
420 			s = splimp();
421 			dummy = ether_ioctl(ifp, cmd, data);
422 			splx(s);
423 			return (dummy);
424 
425 		case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
426 		case SIOCADDMULTI:
427 		case SIOCDELMULTI:
428 		break;
429 
430 		case SIOCGIFSTATUS:
431 			s = splimp();
432 			ifs = (struct ifstat *)data;
433 			dummy = strlen(ifs->ascii);
434 			if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
435 				if (tp->tap_td->td_proc) {
436 				    snprintf(ifs->ascii + dummy,
437 					sizeof(ifs->ascii) - dummy,
438 					"\tOpened by pid %d\n",
439 					(int)tp->tap_td->td_proc->p_pid);
440 				} else {
441 				    snprintf(ifs->ascii + dummy,
442 					sizeof(ifs->ascii) - dummy,
443 					"\tOpened by td %p\n", tp->tap_td);
444 				}
445 			}
446 			splx(s);
447 		break;
448 
449 		default:
450 			return (EINVAL);
451 	}
452 
453 	return (0);
454 } /* tapifioctl */
455 
456 
457 /*
458  * tapifstart
459  *
460  * queue packets from higher level ready to put out
461  */
462 static void
463 tapifstart(ifp)
464 	struct ifnet	*ifp;
465 {
466 	struct tap_softc	*tp = ifp->if_softc;
467 	int			 s;
468 
469 	TAPDEBUG("%s%d starting, minor = %#x\n",
470 			ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
471 
472 	/*
473 	 * do not junk pending output if we are in VMnet mode.
474 	 * XXX: can this do any harm because of queue overflow?
475 	 */
476 
477 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
478 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
479 		struct mbuf	*m = NULL;
480 
481 		TAPDEBUG("%s%d not ready. minor = %#x, tap_flags = 0x%x\n",
482 			ifp->if_name, ifp->if_unit,
483 			minor(tp->tap_dev), tp->tap_flags);
484 
485 		s = splimp();
486 		do {
487 			IF_DEQUEUE(&ifp->if_snd, m);
488 			if (m != NULL)
489 				m_freem(m);
490 			ifp->if_oerrors ++;
491 		} while (m != NULL);
492 		splx(s);
493 
494 		return;
495 	}
496 
497 	s = splimp();
498 	ifp->if_flags |= IFF_OACTIVE;
499 
500 	if (ifp->if_snd.ifq_len != 0) {
501 		if (tp->tap_flags & TAP_RWAIT) {
502 			tp->tap_flags &= ~TAP_RWAIT;
503 			wakeup((caddr_t)tp);
504 		}
505 
506 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL))
507 			pgsigio(tp->tap_sigio, SIGIO, 0);
508 
509 		selwakeup(&tp->tap_rsel);
510 		ifp->if_opackets ++; /* obytes are counted in ether_output */
511 	}
512 
513 	ifp->if_flags &= ~IFF_OACTIVE;
514 	splx(s);
515 } /* tapifstart */
516 
517 
518 /*
519  * tapioctl
520  *
521  * the cdevsw interface is now pretty minimal
522  */
523 static int
524 tapioctl(dev_t dev, u_long cmd, caddr_t data, int flag, d_thread_t *td)
525 {
526 	struct tap_softc	*tp = dev->si_drv1;
527 	struct ifnet		*ifp = &tp->tap_if;
528  	struct tapinfo		*tapp = NULL;
529 	int			 s;
530 
531 	switch (cmd) {
532  		case TAPSIFINFO:
533 			s = splimp();
534  		        tapp = (struct tapinfo *)data;
535  			ifp->if_mtu = tapp->mtu;
536  			ifp->if_type = tapp->type;
537  			ifp->if_baudrate = tapp->baudrate;
538 			splx(s);
539  		break;
540 
541 	 	case TAPGIFINFO:
542  			tapp = (struct tapinfo *)data;
543  			tapp->mtu = ifp->if_mtu;
544  			tapp->type = ifp->if_type;
545  			tapp->baudrate = ifp->if_baudrate;
546  		break;
547 
548 		case TAPSDEBUG:
549 			tapdebug = *(int *)data;
550 		break;
551 
552 		case TAPGDEBUG:
553 			*(int *)data = tapdebug;
554 		break;
555 
556 		case FIONBIO:
557 		break;
558 
559 		case FIOASYNC:
560 			s = splimp();
561 			if (*(int *)data)
562 				tp->tap_flags |= TAP_ASYNC;
563 			else
564 				tp->tap_flags &= ~TAP_ASYNC;
565 			splx(s);
566 		break;
567 
568 		case FIONREAD:
569 			s = splimp();
570 			if (ifp->if_snd.ifq_head) {
571 				struct mbuf	*mb = ifp->if_snd.ifq_head;
572 
573 				for(*(int *)data = 0;mb != NULL;mb = mb->m_next)
574 					*(int *)data += mb->m_len;
575 			}
576 			else
577 				*(int *)data = 0;
578 			splx(s);
579 		break;
580 
581 		case FIOSETOWN:
582 			return (fsetown(*(int *)data, &tp->tap_sigio));
583 
584 		case FIOGETOWN:
585 			*(int *)data = fgetown(tp->tap_sigio);
586 			return (0);
587 
588 		/* this is deprecated, FIOSETOWN should be used instead */
589 		case TIOCSPGRP:
590 			return (fsetown(-(*(int *)data), &tp->tap_sigio));
591 
592 		/* this is deprecated, FIOGETOWN should be used instead */
593 		case TIOCGPGRP:
594 			*(int *)data = -fgetown(tp->tap_sigio);
595 			return (0);
596 
597 		/* VMware/VMnet port ioctl's */
598 
599 		case SIOCGIFFLAGS:	/* get ifnet flags */
600 			bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
601 		break;
602 
603 		case VMIO_SIOCSIFFLAGS: { /* VMware/VMnet SIOCSIFFLAGS */
604 			short	f = *(short *)data;
605 
606 			f &= 0x0fff;
607 			f &= ~IFF_CANTCHANGE;
608 			f |= IFF_UP;
609 
610 			s = splimp();
611 			ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
612 			splx(s);
613 		} break;
614 
615 		case OSIOCGIFADDR:	/* get MAC address of the remote side */
616 		case SIOCGIFADDR:
617 			bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
618 		break;
619 
620 		case SIOCSIFADDR:	/* set MAC address of the remote side */
621 			bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
622 		break;
623 
624 		default:
625 			return (ENOTTY);
626 	}
627 	return (0);
628 } /* tapioctl */
629 
630 
631 /*
632  * tapread
633  *
634  * the cdevsw read interface - reads a packet at a time, or at
635  * least as much of a packet as can be read
636  */
637 static int
638 tapread(dev, uio, flag)
639 	dev_t		 dev;
640 	struct uio	*uio;
641 	int		 flag;
642 {
643 	struct tap_softc	*tp = dev->si_drv1;
644 	struct ifnet		*ifp = &tp->tap_if;
645 	struct mbuf		*m0 = NULL;
646 	int			 error = 0, len, s;
647 
648 	TAPDEBUG("%s%d reading, minor = %#x\n",
649 			ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
650 
651 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
652 		TAPDEBUG("%s%d not ready. minor = %#x, tap_flags = 0x%x\n",
653 				ifp->if_name, ifp->if_unit,
654 				minor(tp->tap_dev), tp->tap_flags);
655 
656 		return (EHOSTDOWN);
657 	}
658 
659 	tp->tap_flags &= ~TAP_RWAIT;
660 
661 	/* sleep until we get a packet */
662 	do {
663 		s = splimp();
664 		IF_DEQUEUE(&ifp->if_snd, m0);
665 		splx(s);
666 
667 		if (m0 == NULL) {
668 			if (flag & IO_NDELAY)
669 				return (EWOULDBLOCK);
670 
671 			tp->tap_flags |= TAP_RWAIT;
672 			error = tsleep((caddr_t)tp, PCATCH, "taprd", 0);
673 			if (error)
674 				return (error);
675 		}
676 	} while (m0 == NULL);
677 
678 	/* feed packet to bpf */
679 	if (ifp->if_bpf != NULL)
680 		bpf_mtap(ifp, m0);
681 
682 	/* xfer packet to user space */
683 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
684 		len = min(uio->uio_resid, m0->m_len);
685 		if (len == 0)
686 			break;
687 
688 		error = uiomove(mtod(m0, caddr_t), len, uio);
689 		m0 = m_free(m0);
690 	}
691 
692 	if (m0 != NULL) {
693 		TAPDEBUG("%s%d dropping mbuf, minor = %#x\n",
694 				ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
695 		m_freem(m0);
696 	}
697 
698 	return (error);
699 } /* tapread */
700 
701 
702 /*
703  * tapwrite
704  *
705  * the cdevsw write interface - an atomic write is a packet - or else!
706  */
707 static int
708 tapwrite(dev, uio, flag)
709 	dev_t		 dev;
710 	struct uio	*uio;
711 	int		 flag;
712 {
713 	struct tap_softc	*tp = dev->si_drv1;
714 	struct ifnet		*ifp = &tp->tap_if;
715 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
716 	struct ether_header	*eh = NULL;
717 	int		 	 error = 0, tlen, mlen;
718 
719 	TAPDEBUG("%s%d writting, minor = %#x\n",
720 				ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
721 
722 	if (uio->uio_resid == 0)
723 		return (0);
724 
725 	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
726 		TAPDEBUG("%s%d invalid packet len = %d, minor = %#x\n",
727 				ifp->if_name, ifp->if_unit,
728 				uio->uio_resid, minor(tp->tap_dev));
729 
730 		return (EIO);
731 	}
732 	tlen = uio->uio_resid;
733 
734 	/* get a header mbuf */
735 	MGETHDR(m, M_DONTWAIT, MT_DATA);
736 	if (m == NULL)
737 		return (ENOBUFS);
738 	mlen = MHLEN;
739 
740 	top = 0;
741 	mp = &top;
742 	while ((error == 0) && (uio->uio_resid > 0)) {
743 		m->m_len = min(mlen, uio->uio_resid);
744 		error = uiomove(mtod(m, caddr_t), m->m_len, uio);
745 		*mp = m;
746 		mp = &m->m_next;
747 		if (uio->uio_resid > 0) {
748 			MGET(m, M_DONTWAIT, MT_DATA);
749 			if (m == NULL) {
750 				error = ENOBUFS;
751 				break;
752 			}
753 			mlen = MLEN;
754 		}
755 	}
756 	if (error) {
757 		ifp->if_ierrors ++;
758 		if (top)
759 			m_freem(top);
760 		return (error);
761 	}
762 
763 	top->m_pkthdr.len = tlen;
764 	top->m_pkthdr.rcvif = ifp;
765 
766 	/*
767 	 * Ethernet bridge and bpf are handled in ether_input
768 	 *
769 	 * adjust mbuf and give packet to the ether_input
770 	 */
771 
772 	eh = mtod(top, struct ether_header *);
773 	m_adj(top, sizeof(struct ether_header));
774 	ether_input(ifp, eh, top);
775 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
776 
777 	return (0);
778 } /* tapwrite */
779 
780 
781 /*
782  * tappoll
783  *
784  * the poll interface, this is only useful on reads
785  * really. the write detect always returns true, write never blocks
786  * anyway, it either accepts the packet or drops it
787  */
788 static int
789 tappoll(dev_t dev, int events, d_thread_t *td)
790 {
791 	struct tap_softc	*tp = dev->si_drv1;
792 	struct ifnet		*ifp = &tp->tap_if;
793 	int		 	 s, revents = 0;
794 
795 	TAPDEBUG("%s%d polling, minor = %#x\n",
796 				ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
797 
798 	s = splimp();
799 	if (events & (POLLIN | POLLRDNORM)) {
800 		if (ifp->if_snd.ifq_len > 0) {
801 			TAPDEBUG("%s%d have data in queue. len = %d, " \
802 				"minor = %#x\n", ifp->if_name, ifp->if_unit,
803 				ifp->if_snd.ifq_len, minor(tp->tap_dev));
804 
805 			revents |= (events & (POLLIN | POLLRDNORM));
806 		}
807 		else {
808 			TAPDEBUG("%s%d waiting for data, minor = %#x\n",
809 				ifp->if_name, ifp->if_unit, minor(tp->tap_dev));
810 
811 			selrecord(td, &tp->tap_rsel);
812 		}
813 	}
814 
815 	if (events & (POLLOUT | POLLWRNORM))
816 		revents |= (events & (POLLOUT | POLLWRNORM));
817 
818 	splx(s);
819 	return (revents);
820 } /* tappoll */
821