xref: /dragonfly/sys/net/tap/if_tap.c (revision 1bf4b486)
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 1.20 2005/06/14 18:37:26 joerg 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/thread2.h>
56 #include <sys/ttycom.h>
57 #include <sys/uio.h>
58 #include <sys/vnode.h>
59 
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/if.h>
63 #include <net/ifq_var.h>
64 #include <net/if_arp.h>
65 #include <net/route.h>
66 
67 #include <netinet/in.h>
68 
69 #include "if_tapvar.h"
70 #include "if_tap.h"
71 
72 
73 #define CDEV_NAME	"tap"
74 #define CDEV_MAJOR	149
75 #define TAPDEBUG	if (tapdebug) if_printf
76 
77 #define TAP		"tap"
78 #define VMNET		"vmnet"
79 #define VMNET_DEV_MASK	0x00010000
80 
81 /* module */
82 static int 		tapmodevent	(module_t, int, void *);
83 
84 /* device */
85 static void		tapcreate	(dev_t);
86 
87 /* network interface */
88 static void		tapifstart	(struct ifnet *);
89 static int		tapifioctl	(struct ifnet *, u_long, caddr_t,
90 					 struct ucred *);
91 static void		tapifinit	(void *);
92 
93 /* character device */
94 static d_open_t		tapopen;
95 static d_close_t	tapclose;
96 static d_read_t		tapread;
97 static d_write_t	tapwrite;
98 static d_ioctl_t	tapioctl;
99 static d_poll_t		tappoll;
100 
101 static struct cdevsw	tap_cdevsw = {
102 	/* dev name */	CDEV_NAME,
103 	/* dev major */	CDEV_MAJOR,
104 	/* flags */	0,
105 	/* port */	NULL,
106 	/* clone */	NULL,
107 
108 	/* open */	tapopen,
109 	/* close */	tapclose,
110 	/* read */	tapread,
111 	/* write */	tapwrite,
112 	/* ioctl */	tapioctl,
113 	/* poll */	tappoll,
114 	/* mmap */	nommap,
115 	/* startegy */	nostrategy,
116 	/* dump */	nodump,
117 	/* psize */	nopsize
118 };
119 
120 static int		taprefcnt = 0;		/* module ref. counter   */
121 static int		taplastunit = -1;	/* max. open unit number */
122 static int		tapdebug = 0;		/* debug flag            */
123 
124 MALLOC_DECLARE(M_TAP);
125 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
126 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
127 DEV_MODULE(if_tap, tapmodevent, NULL);
128 
129 /*
130  * tapmodevent
131  *
132  * module event handler
133  */
134 static int
135 tapmodevent(mod, type, data)
136 	module_t	 mod;
137 	int		 type;
138 	void		*data;
139 {
140 	static int		 attached = 0;
141 	struct ifnet		*ifp = NULL;
142 	int			 unit;
143 
144 	switch (type) {
145 	case MOD_LOAD:
146 		if (attached)
147 			return (EEXIST);
148 
149 		cdevsw_add(&tap_cdevsw, 0, 0);
150 		attached = 1;
151 	break;
152 
153 	case MOD_UNLOAD:
154 		if (taprefcnt > 0)
155 			return (EBUSY);
156 
157 		cdevsw_remove(&tap_cdevsw, 0, 0);
158 
159 		/* XXX: maintain tap ifs in a local list */
160 		unit = 0;
161 		while (unit <= taplastunit) {
162 			crit_enter();
163 			TAILQ_FOREACH(ifp, &ifnet, if_link)
164 				if ((strcmp(ifp->if_dname, TAP) == 0) ||
165 				    (strcmp(ifp->if_dname, VMNET) == 0))
166 					if (ifp->if_dunit == unit)
167 						break;
168 			crit_exit();
169 
170 			if (ifp != NULL) {
171 				struct tap_softc	*tp = ifp->if_softc;
172 
173 				TAPDEBUG(ifp, "detached. minor = %#x, " \
174 					"taplastunit = %d\n",
175 					minor(tp->tap_dev), taplastunit);
176 
177 				crit_enter();
178 				ether_ifdetach(ifp);
179 				crit_exit();
180 				destroy_dev(tp->tap_dev);
181 				free(tp, M_TAP);
182 			}
183 			else
184 				unit ++;
185 		}
186 
187 		attached = 0;
188 	break;
189 
190 	default:
191 		return (EOPNOTSUPP);
192 	}
193 
194 	return (0);
195 } /* tapmodevent */
196 
197 
198 /*
199  * tapcreate
200  *
201  * to create interface
202  */
203 static void
204 tapcreate(dev)
205 	dev_t	dev;
206 {
207 	struct ifnet		*ifp = NULL;
208 	struct tap_softc	*tp = NULL;
209 	uint8_t			ether_addr[ETHER_ADDR_LEN];
210 	int			 unit;
211 	char			*name = NULL;
212 
213 	/* allocate driver storage and create device */
214 	MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK);
215 	bzero(tp, sizeof(*tp));
216 
217 	/* select device: tap or vmnet */
218 	if (minor(dev) & VMNET_DEV_MASK) {
219 		name = VMNET;
220 		unit = lminor(dev) & 0xff;
221 		tp->tap_flags |= TAP_VMNET;
222 	}
223 	else {
224 		name = TAP;
225 		unit = lminor(dev);
226 	}
227 
228 	tp->tap_dev = make_dev(&tap_cdevsw, minor(dev), UID_ROOT, GID_WHEEL,
229 						0600, "%s%d", name, unit);
230 	tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
231 	reference_dev(tp->tap_dev);	/* so we can destroy it later */
232 
233 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
234 	ether_addr[0] = 0x00;
235 	ether_addr[1] = 0xbd;
236 	bcopy(&ticks, ether_addr, 4);
237 	ether_addr[5] = (u_char)unit;
238 
239 	/* fill the rest and attach interface */
240 	ifp = &tp->tap_if;
241 	ifp->if_softc = tp;
242 
243 	if_initname(ifp, name, unit);
244 	if (unit > taplastunit)
245 		taplastunit = unit;
246 
247 	ifp->if_init = tapifinit;
248 	ifp->if_start = tapifstart;
249 	ifp->if_ioctl = tapifioctl;
250 	ifp->if_mtu = ETHERMTU;
251 	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
252 	ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
253 	ifq_set_ready(&ifp->if_snd);
254 
255 	ether_ifattach(ifp, ether_addr);
256 
257 	tp->tap_flags |= TAP_INITED;
258 
259 	TAPDEBUG(ifp, "created. minor = %#x\n", 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 	if ((error = suser(td)) != 0)
275 		return (error);
276 
277 	tp = dev->si_drv1;
278 	if (tp == NULL) {
279 		tapcreate(dev);
280 		tp = dev->si_drv1;
281 	}
282 
283 	if (tp->tap_flags & TAP_OPEN)
284 		return (EBUSY);
285 
286 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
287 
288 	tp->tap_td = td;
289 	tp->tap_flags |= TAP_OPEN;
290 	taprefcnt ++;
291 
292 	TAPDEBUG(&tp->arpcom.ac_if,
293 		 "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
294 		 minor(tp->tap_dev), taprefcnt, taplastunit);
295 
296 	return (0);
297 } /* tapopen */
298 
299 
300 /*
301  * tapclose
302  *
303  * close the device - mark i/f down & delete routing info
304  */
305 static int
306 tapclose(dev_t dev, int foo, int bar, d_thread_t *td)
307 {
308 	struct tap_softc	*tp = dev->si_drv1;
309 	struct ifnet		*ifp = &tp->tap_if;
310 
311 	/* junk all pending output */
312 
313 	crit_enter();
314 	ifq_purge(&ifp->if_snd);
315 	crit_exit();
316 
317 	/*
318 	 * do not bring the interface down, and do not anything with
319 	 * interface, if we are in VMnet mode. just close the device.
320 	 */
321 
322 	if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
323 		crit_enter();
324 		if_down(ifp);
325 		if (ifp->if_flags & IFF_RUNNING) {
326 			/* find internet addresses and delete routes */
327 			struct ifaddr	*ifa = NULL;
328 
329 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
330 				if (ifa->ifa_addr->sa_family == AF_INET) {
331 					rtinit(ifa, (int)RTM_DELETE, 0);
332 
333 					/* remove address from interface */
334 					bzero(ifa->ifa_addr,
335 						   sizeof(*(ifa->ifa_addr)));
336 					bzero(ifa->ifa_dstaddr,
337 						   sizeof(*(ifa->ifa_dstaddr)));
338 					bzero(ifa->ifa_netmask,
339 						   sizeof(*(ifa->ifa_netmask)));
340 				}
341 			}
342 
343 			ifp->if_flags &= ~IFF_RUNNING;
344 		}
345 		crit_exit();
346 	}
347 
348 	funsetown(tp->tap_sigio);
349 	selwakeup(&tp->tap_rsel);
350 
351 	tp->tap_flags &= ~TAP_OPEN;
352 	tp->tap_td = NULL;
353 
354 	taprefcnt --;
355 	if (taprefcnt < 0) {
356 		taprefcnt = 0;
357 		if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
358 			"set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
359 	}
360 
361 	TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
362 		 minor(tp->tap_dev), taprefcnt, taplastunit);
363 
364 	return (0);
365 } /* tapclose */
366 
367 
368 /*
369  * tapifinit
370  *
371  * network interface initialization function
372  */
373 static void
374 tapifinit(xtp)
375 	void	*xtp;
376 {
377 	struct tap_softc	*tp = (struct tap_softc *)xtp;
378 	struct ifnet		*ifp = &tp->tap_if;
379 
380 	TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev));
381 
382 	ifp->if_flags |= IFF_RUNNING;
383 	ifp->if_flags &= ~IFF_OACTIVE;
384 
385 	/* attempt to start output */
386 	tapifstart(ifp);
387 } /* tapifinit */
388 
389 
390 /*
391  * tapifioctl
392  *
393  * Process an ioctl request on network interface
394  */
395 int
396 tapifioctl(ifp, cmd, data, cr)
397 	struct ifnet	*ifp;
398 	u_long		 cmd;
399 	caddr_t		 data;
400 	struct ucred	*cr;
401 {
402 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
403 	struct ifstat		*ifs = NULL;
404 	int			 dummy;
405 
406 	switch (cmd) {
407 		case SIOCSIFADDR:
408 		case SIOCGIFADDR:
409 		case SIOCSIFMTU:
410 			crit_enter();
411 			dummy = ether_ioctl(ifp, cmd, data);
412 			crit_exit();
413 			return (dummy);
414 
415 		case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
416 		case SIOCADDMULTI:
417 		case SIOCDELMULTI:
418 			break;
419 
420 		case SIOCGIFSTATUS:
421 			crit_enter();
422 			ifs = (struct ifstat *)data;
423 			dummy = strlen(ifs->ascii);
424 			if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
425 				if (tp->tap_td->td_proc) {
426 				    snprintf(ifs->ascii + dummy,
427 					sizeof(ifs->ascii) - dummy,
428 					"\tOpened by pid %d\n",
429 					(int)tp->tap_td->td_proc->p_pid);
430 				} else {
431 				    snprintf(ifs->ascii + dummy,
432 					sizeof(ifs->ascii) - dummy,
433 					"\tOpened by td %p\n", tp->tap_td);
434 				}
435 			}
436 			crit_exit();
437 			break;
438 
439 		default:
440 			return (EINVAL);
441 	}
442 
443 	return (0);
444 } /* tapifioctl */
445 
446 
447 /*
448  * tapifstart
449  *
450  * queue packets from higher level ready to put out
451  */
452 static void
453 tapifstart(ifp)
454 	struct ifnet	*ifp;
455 {
456 	struct tap_softc	*tp = ifp->if_softc;
457 
458 	TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
459 
460 	/*
461 	 * do not junk pending output if we are in VMnet mode.
462 	 * XXX: can this do any harm because of queue overflow?
463 	 */
464 
465 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
466 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
467 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
468 			 minor(tp->tap_dev), tp->tap_flags);
469 
470 		crit_enter();
471 		ifq_purge(&ifp->if_snd);
472 		crit_exit();
473 		return;
474 	}
475 
476 	crit_enter();
477 
478 	ifp->if_flags |= IFF_OACTIVE;
479 
480 	if (!ifq_is_empty(&ifp->if_snd)) {
481 		if (tp->tap_flags & TAP_RWAIT) {
482 			tp->tap_flags &= ~TAP_RWAIT;
483 			wakeup((caddr_t)tp);
484 		}
485 
486 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL))
487 			pgsigio(tp->tap_sigio, SIGIO, 0);
488 
489 		selwakeup(&tp->tap_rsel);
490 		ifp->if_opackets ++; /* obytes are counted in ether_output */
491 	}
492 
493 	ifp->if_flags &= ~IFF_OACTIVE;
494 
495 	crit_exit();
496 } /* tapifstart */
497 
498 
499 /*
500  * tapioctl
501  *
502  * the cdevsw interface is now pretty minimal
503  */
504 static int
505 tapioctl(dev_t dev, u_long cmd, caddr_t data, int flag, d_thread_t *td)
506 {
507 	struct tap_softc	*tp = dev->si_drv1;
508 	struct ifnet		*ifp = &tp->tap_if;
509  	struct tapinfo		*tapp = NULL;
510 	struct mbuf *mb;
511 
512 	switch (cmd) {
513  		case TAPSIFINFO:
514 			crit_enter();
515  		        tapp = (struct tapinfo *)data;
516  			ifp->if_mtu = tapp->mtu;
517  			ifp->if_type = tapp->type;
518  			ifp->if_baudrate = tapp->baudrate;
519 			crit_exit();
520  		break;
521 
522 	 	case TAPGIFINFO:
523  			tapp = (struct tapinfo *)data;
524  			tapp->mtu = ifp->if_mtu;
525  			tapp->type = ifp->if_type;
526  			tapp->baudrate = ifp->if_baudrate;
527  		break;
528 
529 		case TAPSDEBUG:
530 			tapdebug = *(int *)data;
531 		break;
532 
533 		case TAPGDEBUG:
534 			*(int *)data = tapdebug;
535 		break;
536 
537 		case FIONBIO:
538 		break;
539 
540 		case FIOASYNC:
541 			crit_enter();
542 			if (*(int *)data)
543 				tp->tap_flags |= TAP_ASYNC;
544 			else
545 				tp->tap_flags &= ~TAP_ASYNC;
546 			crit_exit();
547 		break;
548 
549 		case FIONREAD:
550 			crit_enter();
551 			*(int *)data = 0;
552 			if ((mb = ifq_poll(&ifp->if_snd)) != NULL) {
553 				for(; mb != NULL; mb = mb->m_next)
554 					*(int *)data += mb->m_len;
555 			}
556 			crit_exit();
557 		break;
558 
559 		case FIOSETOWN:
560 			return (fsetown(*(int *)data, &tp->tap_sigio));
561 
562 		case FIOGETOWN:
563 			*(int *)data = fgetown(tp->tap_sigio);
564 			return (0);
565 
566 		/* this is deprecated, FIOSETOWN should be used instead */
567 		case TIOCSPGRP:
568 			return (fsetown(-(*(int *)data), &tp->tap_sigio));
569 
570 		/* this is deprecated, FIOGETOWN should be used instead */
571 		case TIOCGPGRP:
572 			*(int *)data = -fgetown(tp->tap_sigio);
573 			return (0);
574 
575 		/* VMware/VMnet port ioctl's */
576 
577 		case SIOCGIFFLAGS:	/* get ifnet flags */
578 			bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
579 		break;
580 
581 		case VMIO_SIOCSIFFLAGS: { /* VMware/VMnet SIOCSIFFLAGS */
582 			short	f = *(short *)data;
583 
584 			f &= 0x0fff;
585 			f &= ~IFF_CANTCHANGE;
586 			f |= IFF_UP;
587 
588 			crit_enter();
589 			ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
590 			crit_exit();
591 		} break;
592 
593 		case OSIOCGIFADDR:	/* get MAC address of the remote side */
594 		case SIOCGIFADDR:
595 			bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
596 		break;
597 
598 		case SIOCSIFADDR:	/* set MAC address of the remote side */
599 			bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
600 		break;
601 
602 		default:
603 			return (ENOTTY);
604 	}
605 	return (0);
606 } /* tapioctl */
607 
608 
609 /*
610  * tapread
611  *
612  * the cdevsw read interface - reads a packet at a time, or at
613  * least as much of a packet as can be read
614  */
615 static int
616 tapread(dev, uio, flag)
617 	dev_t		 dev;
618 	struct uio	*uio;
619 	int		 flag;
620 {
621 	struct tap_softc	*tp = dev->si_drv1;
622 	struct ifnet		*ifp = &tp->tap_if;
623 	struct mbuf		*m0 = NULL;
624 	int			 error = 0, len;
625 
626 	TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
627 
628 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
629 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
630 			 minor(tp->tap_dev), tp->tap_flags);
631 
632 		return (EHOSTDOWN);
633 	}
634 
635 	tp->tap_flags &= ~TAP_RWAIT;
636 
637 	/* sleep until we get a packet */
638 	do {
639 		crit_enter();
640 		m0 = ifq_dequeue(&ifp->if_snd);
641 		crit_exit();
642 
643 		if (m0 == NULL) {
644 			if (flag & IO_NDELAY)
645 				return (EWOULDBLOCK);
646 
647 			tp->tap_flags |= TAP_RWAIT;
648 			error = tsleep((caddr_t)tp, PCATCH, "taprd", 0);
649 			if (error)
650 				return (error);
651 		}
652 	} while (m0 == NULL);
653 
654 	BPF_MTAP(ifp, m0);
655 
656 	/* xfer packet to user space */
657 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
658 		len = min(uio->uio_resid, m0->m_len);
659 		if (len == 0)
660 			break;
661 
662 		error = uiomove(mtod(m0, caddr_t), len, uio);
663 		m0 = m_free(m0);
664 	}
665 
666 	if (m0 != NULL) {
667 		TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
668 			 minor(tp->tap_dev));
669 		m_freem(m0);
670 	}
671 
672 	return (error);
673 } /* tapread */
674 
675 
676 /*
677  * tapwrite
678  *
679  * the cdevsw write interface - an atomic write is a packet - or else!
680  */
681 static int
682 tapwrite(dev, uio, flag)
683 	dev_t		 dev;
684 	struct uio	*uio;
685 	int		 flag;
686 {
687 	struct tap_softc	*tp = dev->si_drv1;
688 	struct ifnet		*ifp = &tp->tap_if;
689 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
690 	int		 	 error = 0, tlen, mlen;
691 
692 	TAPDEBUG(ifp, "writting, minor = %#x\n", minor(tp->tap_dev));
693 
694 	if (uio->uio_resid == 0)
695 		return (0);
696 
697 	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
698 		TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
699 			 uio->uio_resid, minor(tp->tap_dev));
700 
701 		return (EIO);
702 	}
703 	tlen = uio->uio_resid;
704 
705 	/* get a header mbuf */
706 	MGETHDR(m, MB_DONTWAIT, MT_DATA);
707 	if (m == NULL)
708 		return (ENOBUFS);
709 	mlen = MHLEN;
710 
711 	top = 0;
712 	mp = &top;
713 	while ((error == 0) && (uio->uio_resid > 0)) {
714 		m->m_len = min(mlen, uio->uio_resid);
715 		error = uiomove(mtod(m, caddr_t), m->m_len, uio);
716 		*mp = m;
717 		mp = &m->m_next;
718 		if (uio->uio_resid > 0) {
719 			MGET(m, MB_DONTWAIT, MT_DATA);
720 			if (m == NULL) {
721 				error = ENOBUFS;
722 				break;
723 			}
724 			mlen = MLEN;
725 		}
726 	}
727 	if (error) {
728 		ifp->if_ierrors ++;
729 		if (top)
730 			m_freem(top);
731 		return (error);
732 	}
733 
734 	top->m_pkthdr.len = tlen;
735 	top->m_pkthdr.rcvif = ifp;
736 
737 	/*
738 	 * Ethernet bridge and bpf are handled in ether_input
739 	 *
740 	 * adjust mbuf and give packet to the ether_input
741 	 */
742 
743 	(*ifp->if_input)(ifp, top);
744 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
745 
746 	return (0);
747 } /* tapwrite */
748 
749 
750 /*
751  * tappoll
752  *
753  * the poll interface, this is only useful on reads
754  * really. the write detect always returns true, write never blocks
755  * anyway, it either accepts the packet or drops it
756  */
757 static int
758 tappoll(dev_t dev, int events, d_thread_t *td)
759 {
760 	struct tap_softc	*tp = dev->si_drv1;
761 	struct ifnet		*ifp = &tp->tap_if;
762 	int		 	 revents = 0;
763 
764 	TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
765 
766 	crit_enter();
767 
768 	if (events & (POLLIN | POLLRDNORM)) {
769 		if (!ifq_is_empty(&ifp->if_snd)) {
770 			TAPDEBUG(ifp,
771 				 "has data in queue. minor = %#x\n",
772 				 minor(tp->tap_dev));
773 
774 			revents |= (events & (POLLIN | POLLRDNORM));
775 		}
776 		else {
777 			TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
778 				 minor(tp->tap_dev));
779 
780 			selrecord(td, &tp->tap_rsel);
781 		}
782 	}
783 
784 	if (events & (POLLOUT | POLLWRNORM))
785 		revents |= (events & (POLLOUT | POLLWRNORM));
786 
787 	crit_exit();
788 	return (revents);
789 } /* tappoll */
790