xref: /dragonfly/sys/net/tap/if_tap.c (revision fe76c4fb)
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 1.25 2006/06/13 08:12:03 dillon 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 #include <sys/serialize.h>
60 
61 #include <net/bpf.h>
62 #include <net/ethernet.h>
63 #include <net/if.h>
64 #include <net/ifq_var.h>
65 #include <net/if_arp.h>
66 #include <net/route.h>
67 
68 #include <netinet/in.h>
69 
70 #include "if_tapvar.h"
71 #include "if_tap.h"
72 
73 
74 #define CDEV_NAME	"tap"
75 #define CDEV_MAJOR	149
76 #define TAPDEBUG	if (tapdebug) if_printf
77 
78 #define TAP		"tap"
79 #define VMNET		"vmnet"
80 #define VMNET_DEV_MASK	0x00010000
81 
82 /* module */
83 static int 		tapmodevent	(module_t, int, void *);
84 
85 /* device */
86 static void		tapcreate	(dev_t);
87 
88 /* network interface */
89 static void		tapifstart	(struct ifnet *);
90 static int		tapifioctl	(struct ifnet *, u_long, caddr_t,
91 					 struct ucred *);
92 static void		tapifinit	(void *);
93 
94 /* character device */
95 static d_open_t		tapopen;
96 static d_close_t	tapclose;
97 static d_read_t		tapread;
98 static d_write_t	tapwrite;
99 static d_ioctl_t	tapioctl;
100 static d_poll_t		tappoll;
101 
102 static struct cdevsw	tap_cdevsw = {
103 	/* dev name */	CDEV_NAME,
104 	/* dev major */	CDEV_MAJOR,
105 	/* flags */	0,
106 	/* port */	NULL,
107 	/* clone */	NULL,
108 
109 	/* open */	tapopen,
110 	/* close */	tapclose,
111 	/* read */	tapread,
112 	/* write */	tapwrite,
113 	/* ioctl */	tapioctl,
114 	/* poll */	tappoll,
115 	/* mmap */	nommap,
116 	/* startegy */	nostrategy,
117 	/* dump */	nodump,
118 	/* psize */	nopsize
119 };
120 
121 static int		taprefcnt = 0;		/* module ref. counter   */
122 static int		taplastunit = -1;	/* max. open unit number */
123 static int		tapdebug = 0;		/* debug flag            */
124 
125 MALLOC_DECLARE(M_TAP);
126 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
127 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
128 DEV_MODULE(if_tap, tapmodevent, NULL);
129 
130 /*
131  * tapmodevent
132  *
133  * module event handler
134  */
135 static int
136 tapmodevent(module_t mod, int type, void *data)
137 {
138 	static int		 attached = 0;
139 	struct ifnet		*ifp = NULL;
140 	int			 unit;
141 
142 	switch (type) {
143 	case MOD_LOAD:
144 		if (attached)
145 			return (EEXIST);
146 
147 		cdevsw_add(&tap_cdevsw, 0, 0);
148 		attached = 1;
149 	break;
150 
151 	case MOD_UNLOAD:
152 		if (taprefcnt > 0)
153 			return (EBUSY);
154 
155 		cdevsw_remove(&tap_cdevsw, 0, 0);
156 
157 		/* XXX: maintain tap ifs in a local list */
158 		unit = 0;
159 		while (unit <= taplastunit) {
160 			TAILQ_FOREACH(ifp, &ifnet, if_link) {
161 				if ((strcmp(ifp->if_dname, TAP) == 0) ||
162 				    (strcmp(ifp->if_dname, VMNET) == 0)) {
163 					if (ifp->if_dunit == unit)
164 						break;
165 				}
166 			}
167 
168 			if (ifp != NULL) {
169 				struct tap_softc	*tp = ifp->if_softc;
170 
171 				TAPDEBUG(ifp, "detached. minor = %#x, " \
172 					"taplastunit = %d\n",
173 					minor(tp->tap_dev), taplastunit);
174 
175 				ether_ifdetach(ifp);
176 				destroy_dev(tp->tap_dev);
177 				free(tp, M_TAP);
178 			}
179 			else
180 				unit ++;
181 		}
182 
183 		attached = 0;
184 	break;
185 
186 	default:
187 		return (EOPNOTSUPP);
188 	}
189 
190 	return (0);
191 } /* tapmodevent */
192 
193 
194 /*
195  * tapcreate
196  *
197  * to create interface
198  */
199 static void
200 tapcreate(dev_t dev)
201 {
202 	struct ifnet		*ifp = NULL;
203 	struct tap_softc	*tp = NULL;
204 	uint8_t			ether_addr[ETHER_ADDR_LEN];
205 	int			 unit;
206 	char			*name = NULL;
207 
208 	/* allocate driver storage and create device */
209 	MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK);
210 	bzero(tp, sizeof(*tp));
211 
212 	/* select device: tap or vmnet */
213 	if (minor(dev) & VMNET_DEV_MASK) {
214 		name = VMNET;
215 		unit = lminor(dev) & 0xff;
216 		tp->tap_flags |= TAP_VMNET;
217 	}
218 	else {
219 		name = TAP;
220 		unit = lminor(dev);
221 	}
222 
223 	tp->tap_dev = make_dev(&tap_cdevsw, minor(dev), UID_ROOT, GID_WHEEL,
224 						0600, "%s%d", name, unit);
225 	tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
226 	reference_dev(tp->tap_dev);	/* so we can destroy it later */
227 
228 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
229 	ether_addr[0] = 0x00;
230 	ether_addr[1] = 0xbd;
231 	bcopy(&ticks, ether_addr, 4);
232 	ether_addr[5] = (u_char)unit;
233 
234 	/* fill the rest and attach interface */
235 	ifp = &tp->tap_if;
236 	ifp->if_softc = tp;
237 
238 	if_initname(ifp, name, unit);
239 	if (unit > taplastunit)
240 		taplastunit = unit;
241 
242 	ifp->if_init = tapifinit;
243 	ifp->if_start = tapifstart;
244 	ifp->if_ioctl = tapifioctl;
245 	ifp->if_mtu = ETHERMTU;
246 	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
247 	ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
248 	ifq_set_ready(&ifp->if_snd);
249 
250 	ether_ifattach(ifp, ether_addr, NULL);
251 
252 	tp->tap_flags |= TAP_INITED;
253 
254 	TAPDEBUG(ifp, "created. minor = %#x\n", minor(tp->tap_dev));
255 } /* tapcreate */
256 
257 
258 /*
259  * tapopen
260  *
261  * to open tunnel. must be superuser
262  */
263 static int
264 tapopen(dev_t dev, int flag, int mode, d_thread_t *td)
265 {
266 	struct tap_softc	*tp = NULL;
267 	int			 error;
268 
269 	if ((error = suser(td)) != 0)
270 		return (error);
271 
272 	tp = dev->si_drv1;
273 	if (tp == NULL) {
274 		tapcreate(dev);
275 		tp = dev->si_drv1;
276 	}
277 
278 	if (tp->tap_flags & TAP_OPEN)
279 		return (EBUSY);
280 
281 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
282 
283 	tp->tap_td = td;
284 	tp->tap_flags |= TAP_OPEN;
285 	taprefcnt ++;
286 
287 	TAPDEBUG(&tp->arpcom.ac_if,
288 		 "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
289 		 minor(tp->tap_dev), taprefcnt, taplastunit);
290 
291 	return (0);
292 } /* tapopen */
293 
294 
295 /*
296  * tapclose
297  *
298  * close the device - mark i/f down & delete routing info
299  */
300 static int
301 tapclose(dev_t dev, int foo, int bar, d_thread_t *td)
302 {
303 	struct tap_softc	*tp = dev->si_drv1;
304 	struct ifnet		*ifp = &tp->tap_if;
305 
306 	/* junk all pending output */
307 
308 	lwkt_serialize_enter(ifp->if_serializer);
309 	ifq_purge(&ifp->if_snd);
310 	lwkt_serialize_exit(ifp->if_serializer);
311 
312 	/*
313 	 * do not bring the interface down, and do not anything with
314 	 * interface, if we are in VMnet mode. just close the device.
315 	 */
316 
317 	if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
318 		if_down(ifp);
319 		lwkt_serialize_enter(ifp->if_serializer);
320 		if (ifp->if_flags & IFF_RUNNING) {
321 			/* find internet addresses and delete routes */
322 			struct ifaddr	*ifa = NULL;
323 
324 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
325 				if (ifa->ifa_addr->sa_family == AF_INET) {
326 					rtinit(ifa, (int)RTM_DELETE, 0);
327 
328 					/* remove address from interface */
329 					bzero(ifa->ifa_addr,
330 						   sizeof(*(ifa->ifa_addr)));
331 					bzero(ifa->ifa_dstaddr,
332 						   sizeof(*(ifa->ifa_dstaddr)));
333 					bzero(ifa->ifa_netmask,
334 						   sizeof(*(ifa->ifa_netmask)));
335 				}
336 			}
337 
338 			ifp->if_flags &= ~IFF_RUNNING;
339 		}
340 		lwkt_serialize_exit(ifp->if_serializer);
341 	}
342 
343 	funsetown(tp->tap_sigio);
344 	selwakeup(&tp->tap_rsel);
345 
346 	tp->tap_flags &= ~TAP_OPEN;
347 	tp->tap_td = NULL;
348 
349 	taprefcnt --;
350 	if (taprefcnt < 0) {
351 		taprefcnt = 0;
352 		if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
353 			"set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
354 	}
355 
356 	TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
357 		 minor(tp->tap_dev), taprefcnt, taplastunit);
358 
359 	return (0);
360 } /* tapclose */
361 
362 
363 /*
364  * tapifinit
365  *
366  * network interface initialization function
367  */
368 static void
369 tapifinit(void *xtp)
370 {
371 	struct tap_softc	*tp = (struct tap_softc *)xtp;
372 	struct ifnet		*ifp = &tp->tap_if;
373 
374 	TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev));
375 
376 	ifp->if_flags |= IFF_RUNNING;
377 	ifp->if_flags &= ~IFF_OACTIVE;
378 
379 	/* attempt to start output */
380 	tapifstart(ifp);
381 } /* tapifinit */
382 
383 
384 /*
385  * tapifioctl
386  *
387  * Process an ioctl request on network interface
388  *
389  * MPSAFE
390  */
391 int
392 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
393 {
394 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
395 	struct ifstat		*ifs = NULL;
396 	int			 dummy;
397 
398 	switch (cmd) {
399 		case SIOCSIFADDR:
400 		case SIOCGIFADDR:
401 		case SIOCSIFMTU:
402 			dummy = ether_ioctl(ifp, cmd, data);
403 			return (dummy);
404 
405 		case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
406 		case SIOCADDMULTI:
407 		case SIOCDELMULTI:
408 			break;
409 
410 		case SIOCGIFSTATUS:
411 			ifs = (struct ifstat *)data;
412 			dummy = strlen(ifs->ascii);
413 			if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
414 				if (tp->tap_td->td_proc) {
415 				    snprintf(ifs->ascii + dummy,
416 					sizeof(ifs->ascii) - dummy,
417 					"\tOpened by pid %d\n",
418 					(int)tp->tap_td->td_proc->p_pid);
419 				} else {
420 				    snprintf(ifs->ascii + dummy,
421 					sizeof(ifs->ascii) - dummy,
422 					"\tOpened by td %p\n", tp->tap_td);
423 				}
424 			}
425 			break;
426 
427 		default:
428 			return (EINVAL);
429 	}
430 
431 	return (0);
432 } /* tapifioctl */
433 
434 
435 /*
436  * tapifstart
437  *
438  * queue packets from higher level ready to put out
439  */
440 static void
441 tapifstart(struct ifnet *ifp)
442 {
443 	struct tap_softc	*tp = ifp->if_softc;
444 
445 	TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
446 
447 	/*
448 	 * do not junk pending output if we are in VMnet mode.
449 	 * XXX: can this do any harm because of queue overflow?
450 	 */
451 
452 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
453 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
454 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
455 			 minor(tp->tap_dev), tp->tap_flags);
456 
457 		ifq_purge(&ifp->if_snd);
458 		return;
459 	}
460 
461 	ifp->if_flags |= IFF_OACTIVE;
462 
463 	if (!ifq_is_empty(&ifp->if_snd)) {
464 		if (tp->tap_flags & TAP_RWAIT) {
465 			tp->tap_flags &= ~TAP_RWAIT;
466 			wakeup((caddr_t)tp);
467 		}
468 
469 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL))
470 			pgsigio(tp->tap_sigio, SIGIO, 0);
471 
472 		/*
473 		 * selwakeup is not MPSAFE.  tapifstart is.
474 		 */
475 		get_mplock();
476 		selwakeup(&tp->tap_rsel);
477 		rel_mplock();
478 		ifp->if_opackets ++; /* obytes are counted in ether_output */
479 	}
480 
481 	ifp->if_flags &= ~IFF_OACTIVE;
482 } /* tapifstart */
483 
484 
485 /*
486  * tapioctl
487  *
488  * the cdevsw interface is now pretty minimal
489  */
490 static int
491 tapioctl(dev_t dev, u_long cmd, caddr_t data, int flag, d_thread_t *td)
492 {
493 	struct tap_softc	*tp = dev->si_drv1;
494 	struct ifnet		*ifp = &tp->tap_if;
495  	struct tapinfo		*tapp = NULL;
496 	struct mbuf *mb;
497 	short f;
498 	int error;
499 
500 	lwkt_serialize_enter(ifp->if_serializer);
501 	error = 0;
502 
503 	switch (cmd) {
504 	case TAPSIFINFO:
505 		tapp = (struct tapinfo *)data;
506 		ifp->if_mtu = tapp->mtu;
507 		ifp->if_type = tapp->type;
508 		ifp->if_baudrate = tapp->baudrate;
509 		break;
510 
511 	case TAPGIFINFO:
512 		tapp = (struct tapinfo *)data;
513 		tapp->mtu = ifp->if_mtu;
514 		tapp->type = ifp->if_type;
515 		tapp->baudrate = ifp->if_baudrate;
516 		break;
517 
518 	case TAPSDEBUG:
519 		tapdebug = *(int *)data;
520 		break;
521 
522 	case TAPGDEBUG:
523 		*(int *)data = tapdebug;
524 		break;
525 
526 	case FIOASYNC:
527 		if (*(int *)data)
528 			tp->tap_flags |= TAP_ASYNC;
529 		else
530 			tp->tap_flags &= ~TAP_ASYNC;
531 		break;
532 
533 	case FIONREAD:
534 		*(int *)data = 0;
535 		if ((mb = ifq_poll(&ifp->if_snd)) != NULL) {
536 			for(; mb != NULL; mb = mb->m_next)
537 				*(int *)data += mb->m_len;
538 		}
539 		break;
540 
541 	case FIOSETOWN:
542 		error = fsetown(*(int *)data, &tp->tap_sigio);
543 		break;
544 
545 	case FIOGETOWN:
546 		*(int *)data = fgetown(tp->tap_sigio);
547 		break;
548 
549 	/* this is deprecated, FIOSETOWN should be used instead */
550 	case TIOCSPGRP:
551 		error = fsetown(-(*(int *)data), &tp->tap_sigio);
552 		break;
553 
554 	/* this is deprecated, FIOGETOWN should be used instead */
555 	case TIOCGPGRP:
556 		*(int *)data = -fgetown(tp->tap_sigio);
557 		break;
558 
559 	/* VMware/VMnet port ioctl's */
560 
561 	case SIOCGIFFLAGS:	/* get ifnet flags */
562 		bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
563 		break;
564 
565 	case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
566 		f = *(short *)data;
567 		f &= 0x0fff;
568 		f &= ~IFF_CANTCHANGE;
569 		f |= IFF_UP;
570 		ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
571 		break;
572 
573 	case OSIOCGIFADDR:	/* get MAC address of the remote side */
574 	case SIOCGIFADDR:
575 		bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
576 		break;
577 
578 	case SIOCSIFADDR:	/* set MAC address of the remote side */
579 		bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
580 		break;
581 
582 	default:
583 		error = ENOTTY;
584 		break;
585 	}
586 	lwkt_serialize_exit(ifp->if_serializer);
587 	return (error);
588 } /* tapioctl */
589 
590 
591 /*
592  * tapread
593  *
594  * the cdevsw read interface - reads a packet at a time, or at
595  * least as much of a packet as can be read
596  */
597 static int
598 tapread(dev_t dev, struct uio *uio, int flag)
599 {
600 	struct tap_softc	*tp = dev->si_drv1;
601 	struct ifnet		*ifp = &tp->tap_if;
602 	struct mbuf		*m0 = NULL;
603 	int			 error = 0, len;
604 
605 	TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
606 
607 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
608 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
609 			 minor(tp->tap_dev), tp->tap_flags);
610 
611 		return (EHOSTDOWN);
612 	}
613 
614 	tp->tap_flags &= ~TAP_RWAIT;
615 
616 	/* sleep until we get a packet */
617 	do {
618 		lwkt_serialize_enter(ifp->if_serializer);
619 		m0 = ifq_dequeue(&ifp->if_snd, NULL);
620 		if (m0 == NULL) {
621 			tp->tap_flags |= TAP_RWAIT;
622 			tsleep_interlock(tp);
623 			lwkt_serialize_exit(ifp->if_serializer);
624 			if (flag & IO_NDELAY)
625 				return (EWOULDBLOCK);
626 			error = tsleep(tp, PCATCH, "taprd", 0);
627 			if (error)
628 				return (error);
629 		} else {
630 			lwkt_serialize_exit(ifp->if_serializer);
631 		}
632 	} while (m0 == NULL);
633 
634 	BPF_MTAP(ifp, m0);
635 
636 	/* xfer packet to user space */
637 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
638 		len = min(uio->uio_resid, m0->m_len);
639 		if (len == 0)
640 			break;
641 
642 		error = uiomove(mtod(m0, caddr_t), len, uio);
643 		m0 = m_free(m0);
644 	}
645 
646 	if (m0 != NULL) {
647 		TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
648 			 minor(tp->tap_dev));
649 		m_freem(m0);
650 	}
651 
652 	return (error);
653 } /* tapread */
654 
655 
656 /*
657  * tapwrite
658  *
659  * the cdevsw write interface - an atomic write is a packet - or else!
660  */
661 static int
662 tapwrite(dev_t dev, struct uio *uio, int flag)
663 {
664 	struct tap_softc	*tp = dev->si_drv1;
665 	struct ifnet		*ifp = &tp->tap_if;
666 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
667 	int		 	 error = 0, tlen, mlen;
668 
669 	TAPDEBUG(ifp, "writting, minor = %#x\n", minor(tp->tap_dev));
670 
671 	if (uio->uio_resid == 0)
672 		return (0);
673 
674 	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
675 		TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
676 			 uio->uio_resid, minor(tp->tap_dev));
677 
678 		return (EIO);
679 	}
680 	tlen = uio->uio_resid;
681 
682 	/* get a header mbuf */
683 	MGETHDR(m, MB_DONTWAIT, MT_DATA);
684 	if (m == NULL)
685 		return (ENOBUFS);
686 	mlen = MHLEN;
687 
688 	top = 0;
689 	mp = &top;
690 	while ((error == 0) && (uio->uio_resid > 0)) {
691 		m->m_len = min(mlen, uio->uio_resid);
692 		error = uiomove(mtod(m, caddr_t), m->m_len, uio);
693 		*mp = m;
694 		mp = &m->m_next;
695 		if (uio->uio_resid > 0) {
696 			MGET(m, MB_DONTWAIT, MT_DATA);
697 			if (m == NULL) {
698 				error = ENOBUFS;
699 				break;
700 			}
701 			mlen = MLEN;
702 		}
703 	}
704 	if (error) {
705 		ifp->if_ierrors ++;
706 		if (top)
707 			m_freem(top);
708 		return (error);
709 	}
710 
711 	top->m_pkthdr.len = tlen;
712 	top->m_pkthdr.rcvif = ifp;
713 
714 	/*
715 	 * Ethernet bridge and bpf are handled in ether_input
716 	 *
717 	 * adjust mbuf and give packet to the ether_input
718 	 */
719 	lwkt_serialize_enter(ifp->if_serializer);
720 	ifp->if_input(ifp, top);
721 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
722 	lwkt_serialize_exit(ifp->if_serializer);
723 
724 	return (0);
725 } /* tapwrite */
726 
727 
728 /*
729  * tappoll
730  *
731  * the poll interface, this is only useful on reads
732  * really. the write detect always returns true, write never blocks
733  * anyway, it either accepts the packet or drops it
734  */
735 static int
736 tappoll(dev_t dev, int events, d_thread_t *td)
737 {
738 	struct tap_softc	*tp = dev->si_drv1;
739 	struct ifnet		*ifp = &tp->tap_if;
740 	int		 	 revents = 0;
741 
742 	TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
743 
744 	lwkt_serialize_enter(ifp->if_serializer);
745 	if (events & (POLLIN | POLLRDNORM)) {
746 		if (!ifq_is_empty(&ifp->if_snd)) {
747 			TAPDEBUG(ifp,
748 				 "has data in queue. minor = %#x\n",
749 				 minor(tp->tap_dev));
750 
751 			revents |= (events & (POLLIN | POLLRDNORM));
752 		}
753 		else {
754 			TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
755 				 minor(tp->tap_dev));
756 
757 			selrecord(td, &tp->tap_rsel);
758 		}
759 	}
760 	lwkt_serialize_exit(ifp->if_serializer);
761 
762 	if (events & (POLLOUT | POLLWRNORM))
763 		revents |= (events & (POLLOUT | POLLWRNORM));
764 	return (revents);
765 } /* tappoll */
766