xref: /dragonfly/sys/net/tap/if_tap.c (revision 8a7bdfea) 
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 1.38 2007/12/31 04:58:53 sephe 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/device.h>
44 #include <sys/filedesc.h>
45 #include <sys/filio.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/poll.h>
50 #include <sys/proc.h>
51 #include <sys/signalvar.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56 #include <sys/thread2.h>
57 #include <sys/ttycom.h>
58 #include <sys/uio.h>
59 #include <sys/vnode.h>
60 #include <sys/serialize.h>
61 
62 #include <net/bpf.h>
63 #include <net/ethernet.h>
64 #include <net/if.h>
65 #include <net/ifq_var.h>
66 #include <net/if_arp.h>
67 #include <net/route.h>
68 
69 #include <netinet/in.h>
70 
71 #include "if_tapvar.h"
72 #include "if_tap.h"
73 
74 
75 #define CDEV_NAME	"tap"
76 #define CDEV_MAJOR	149
77 #define TAPDEBUG	if (tapdebug) if_printf
78 
79 #define TAP		"tap"
80 #define VMNET		"vmnet"
81 #define VMNET_DEV_MASK	0x00010000
82 
83 /* module */
84 static int 		tapmodevent	(module_t, int, void *);
85 
86 /* device */
87 static void		tapcreate	(cdev_t);
88 
89 /* network interface */
90 static void		tapifstart	(struct ifnet *);
91 static int		tapifioctl	(struct ifnet *, u_long, caddr_t,
92 					 struct ucred *);
93 static void		tapifinit	(void *);
94 
95 /* character device */
96 static d_open_t		tapopen;
97 static d_close_t	tapclose;
98 static d_read_t		tapread;
99 static d_write_t	tapwrite;
100 static d_ioctl_t	tapioctl;
101 static d_poll_t		tappoll;
102 static d_kqfilter_t	tapkqfilter;
103 
104 static struct dev_ops	tap_ops = {
105 	{ CDEV_NAME, CDEV_MAJOR, 0 },
106 	.d_open =	tapopen,
107 	.d_close =	tapclose,
108 	.d_read =	tapread,
109 	.d_write =	tapwrite,
110 	.d_ioctl =	tapioctl,
111 	.d_poll =	tappoll,
112 	.d_kqfilter =	tapkqfilter
113 };
114 
115 static int		taprefcnt = 0;		/* module ref. counter   */
116 static int		taplastunit = -1;	/* max. open unit number */
117 static int		tapdebug = 0;		/* debug flag            */
118 
119 MALLOC_DECLARE(M_TAP);
120 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
121 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
122 DEV_MODULE(if_tap, tapmodevent, NULL);
123 
124 /*
125  * tapmodevent
126  *
127  * module event handler
128  */
129 static int
130 tapmodevent(module_t mod, int type, void *data)
131 {
132 	static int		 attached = 0;
133 	struct ifnet		*ifp = NULL;
134 	int			 unit;
135 
136 	switch (type) {
137 	case MOD_LOAD:
138 		if (attached)
139 			return (EEXIST);
140 
141 		dev_ops_add(&tap_ops, 0, 0);
142 		attached = 1;
143 	break;
144 
145 	case MOD_UNLOAD:
146 		if (taprefcnt > 0)
147 			return (EBUSY);
148 
149 		dev_ops_remove(&tap_ops, 0, 0);
150 
151 		/* XXX: maintain tap ifs in a local list */
152 		unit = 0;
153 		while (unit <= taplastunit) {
154 			TAILQ_FOREACH(ifp, &ifnet, if_link) {
155 				if ((strcmp(ifp->if_dname, TAP) == 0) ||
156 				    (strcmp(ifp->if_dname, VMNET) == 0)) {
157 					if (ifp->if_dunit == unit)
158 						break;
159 				}
160 			}
161 
162 			if (ifp != NULL) {
163 				struct tap_softc	*tp = ifp->if_softc;
164 
165 				TAPDEBUG(ifp, "detached. minor = %#x, " \
166 					"taplastunit = %d\n",
167 					minor(tp->tap_dev), taplastunit);
168 
169 				ether_ifdetach(ifp);
170 				destroy_dev(tp->tap_dev);
171 				kfree(tp, M_TAP);
172 			}
173 			else
174 				unit ++;
175 		}
176 
177 		attached = 0;
178 	break;
179 
180 	default:
181 		return (EOPNOTSUPP);
182 	}
183 
184 	return (0);
185 } /* tapmodevent */
186 
187 
188 /*
189  * tapcreate
190  *
191  * to create interface
192  */
193 static void
194 tapcreate(cdev_t dev)
195 {
196 	struct ifnet		*ifp = NULL;
197 	struct tap_softc	*tp = NULL;
198 	uint8_t			ether_addr[ETHER_ADDR_LEN];
199 	int			 unit;
200 	char			*name = NULL;
201 
202 	/* allocate driver storage and create device */
203 	MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
204 
205 	/* select device: tap or vmnet */
206 	if (minor(dev) & VMNET_DEV_MASK) {
207 		name = VMNET;
208 		unit = lminor(dev) & 0xff;
209 		tp->tap_flags |= TAP_VMNET;
210 	}
211 	else {
212 		name = TAP;
213 		unit = lminor(dev);
214 	}
215 
216 	tp->tap_dev = make_dev(&tap_ops, minor(dev), UID_ROOT, GID_WHEEL,
217 						0600, "%s%d", name, unit);
218 	tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
219 	reference_dev(tp->tap_dev);	/* so we can destroy it later */
220 
221 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
222 	ether_addr[0] = 0x00;
223 	ether_addr[1] = 0xbd;
224 	bcopy(&ticks, &ether_addr[2], 3);
225 	ether_addr[5] = (u_char)unit;
226 
227 	/* fill the rest and attach interface */
228 	ifp = &tp->tap_if;
229 	ifp->if_softc = tp;
230 
231 	if_initname(ifp, name, unit);
232 	if (unit > taplastunit)
233 		taplastunit = unit;
234 
235 	ifp->if_init = tapifinit;
236 	ifp->if_start = tapifstart;
237 	ifp->if_ioctl = tapifioctl;
238 	ifp->if_mtu = ETHERMTU;
239 	ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
240 	ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
241 	ifq_set_ready(&ifp->if_snd);
242 
243 	ether_ifattach(ifp, ether_addr, NULL);
244 
245 	tp->tap_flags |= TAP_INITED;
246 
247 	TAPDEBUG(ifp, "created. minor = %#x\n", minor(tp->tap_dev));
248 } /* tapcreate */
249 
250 
251 /*
252  * tapopen
253  *
254  * to open tunnel. must be superuser
255  */
256 static int
257 tapopen(struct dev_open_args *ap)
258 {
259 	cdev_t dev = ap->a_head.a_dev;
260 	struct tap_softc *tp = NULL;
261 	struct ifnet *ifp = NULL;
262 	int error;
263 
264 	if ((error = suser_cred(ap->a_cred, 0)) != 0)
265 		return (error);
266 
267 	get_mplock();
268 	tp = dev->si_drv1;
269 	if (tp == NULL) {
270 		tapcreate(dev);
271 		tp = dev->si_drv1;
272 		ifp = &tp->arpcom.ac_if;
273 	} else {
274 		ifp = &tp->arpcom.ac_if;
275 
276                 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
277 
278 		/* Announce the return of the interface. */
279 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
280 	}
281 
282 	if (tp->tap_flags & TAP_OPEN) {
283 		rel_mplock();
284 		return (EBUSY);
285 	}
286 
287 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
288 
289 	tp->tap_td = curthread;
290 	tp->tap_flags |= TAP_OPEN;
291 	taprefcnt ++;
292 
293 	TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
294 		 minor(tp->tap_dev), taprefcnt, taplastunit);
295 
296 	rel_mplock();
297 	return (0);
298 }
299 
300 
301 /*
302  * tapclose
303  *
304  * close the device - mark i/f down & delete routing info
305  */
306 static int
307 tapclose(struct dev_close_args *ap)
308 {
309 	cdev_t dev = ap->a_head.a_dev;
310 	struct tap_softc	*tp = dev->si_drv1;
311 	struct ifnet		*ifp = &tp->tap_if;
312 
313 	/* junk all pending output */
314 
315 	get_mplock();
316 	lwkt_serialize_enter(ifp->if_serializer);
317 	ifq_purge(&ifp->if_snd);
318 	lwkt_serialize_exit(ifp->if_serializer);
319 
320 	/*
321 	 * do not bring the interface down, and do not anything with
322 	 * interface, if we are in VMnet mode. just close the device.
323 	 */
324 
325 	if ((tp->tap_flags & TAP_VMNET) == 0) {
326 		if (ifp->if_flags & IFF_UP) {
327 			lwkt_serialize_enter(ifp->if_serializer);
328 			if_down(ifp);
329 			lwkt_serialize_exit(ifp->if_serializer);
330 		}
331 		ifp->if_flags &= ~IFF_RUNNING;
332 	}
333 	if_purgeaddrs_nolink(ifp);
334 
335 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
336 
337 	/* Announce the departure of the interface. */
338 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
339 
340 	funsetown(tp->tap_sigio);
341 	selwakeup(&tp->tap_rsel);
342 
343 	tp->tap_flags &= ~TAP_OPEN;
344 	tp->tap_td = NULL;
345 
346 	taprefcnt --;
347 	if (taprefcnt < 0) {
348 		taprefcnt = 0;
349 		if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
350 			"set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
351 	}
352 
353 	TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
354 		 minor(tp->tap_dev), taprefcnt, taplastunit);
355 
356 	rel_mplock();
357 	return (0);
358 }
359 
360 
361 /*
362  * tapifinit
363  *
364  * Network interface initialization function (called with if serializer held)
365  *
366  * MPSAFE
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 }
382 
383 
384 /*
385  * tapifioctl
386  *
387  * Process an ioctl request on network interface (called with if serializer
388  * held).
389  *
390  * MPSAFE
391  */
392 static int
393 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
394 {
395 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
396 	struct ifstat		*ifs = NULL;
397 	int			 dummy;
398 
399 	switch (cmd) {
400 		case SIOCSIFADDR:
401 		case SIOCGIFADDR:
402 		case SIOCSIFMTU:
403 			dummy = ether_ioctl(ifp, cmd, data);
404 			return (dummy);
405 
406 		case SIOCSIFFLAGS:
407 			if ((tp->tap_flags & TAP_VMNET) == 0) {
408 				/*
409 				 * Only for non-vmnet tap(4)
410 				 */
411 				if (ifp->if_flags & IFF_UP) {
412 					if ((ifp->if_flags & IFF_RUNNING) == 0)
413 						tapifinit(tp);
414 				} else {
415 					/* We don't have a tapifstop() */
416 					ifp->if_flags &= ~IFF_RUNNING;
417 				}
418 			}
419 			break;
420 		case SIOCADDMULTI: /* XXX -- just like vmnet does */
421 		case SIOCDELMULTI:
422 			break;
423 
424 		case SIOCGIFSTATUS:
425 			ifs = (struct ifstat *)data;
426 			dummy = strlen(ifs->ascii);
427 			if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
428 				if (tp->tap_td->td_proc) {
429 				    ksnprintf(ifs->ascii + dummy,
430 					sizeof(ifs->ascii) - dummy,
431 					"\tOpened by pid %d\n",
432 					(int)tp->tap_td->td_proc->p_pid);
433 				} else {
434 				    ksnprintf(ifs->ascii + dummy,
435 					sizeof(ifs->ascii) - dummy,
436 					"\tOpened by td %p\n", tp->tap_td);
437 				}
438 			}
439 			break;
440 
441 		default:
442 			return (EINVAL);
443 	}
444 
445 	return (0);
446 }
447 
448 
449 /*
450  * tapifstart
451  *
452  * Queue packets from higher level ready to put out (called with if serializer
453  * held)
454  *
455  * MPSAFE
456  */
457 static void
458 tapifstart(struct ifnet *ifp)
459 {
460 	struct tap_softc	*tp = ifp->if_softc;
461 
462 	TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
463 
464 	/*
465 	 * do not junk pending output if we are in VMnet mode.
466 	 * XXX: can this do any harm because of queue overflow?
467 	 */
468 
469 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
470 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
471 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
472 			 minor(tp->tap_dev), tp->tap_flags);
473 
474 		ifq_purge(&ifp->if_snd);
475 		return;
476 	}
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 		KNOTE(&tp->tap_rsel.si_note, 0);
486 
487 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
488 			get_mplock();
489 			pgsigio(tp->tap_sigio, SIGIO, 0);
490 			rel_mplock();
491 		}
492 
493 		/*
494 		 * selwakeup is not MPSAFE.  tapifstart is.
495 		 */
496 		get_mplock();
497 		selwakeup(&tp->tap_rsel);
498 		rel_mplock();
499 		ifp->if_opackets ++; /* obytes are counted in ether_output */
500 	}
501 
502 	ifp->if_flags &= ~IFF_OACTIVE;
503 }
504 
505 
506 /*
507  * tapioctl
508  *
509  * The ops interface is now pretty minimal.  Called via fileops with nothing
510  * held.
511  *
512  * MPSAFE
513  */
514 static int
515 tapioctl(struct dev_ioctl_args *ap)
516 {
517 	cdev_t dev = ap->a_head.a_dev;
518 	caddr_t data = ap->a_data;
519 	struct tap_softc	*tp = dev->si_drv1;
520 	struct ifnet		*ifp = &tp->tap_if;
521  	struct tapinfo		*tapp = NULL;
522 	struct mbuf *mb;
523 	short f;
524 	int error;
525 
526 	lwkt_serialize_enter(ifp->if_serializer);
527 	error = 0;
528 
529 	switch (ap->a_cmd) {
530 	case TAPSIFINFO:
531 		tapp = (struct tapinfo *)data;
532 		ifp->if_mtu = tapp->mtu;
533 		ifp->if_type = tapp->type;
534 		ifp->if_baudrate = tapp->baudrate;
535 		break;
536 
537 	case TAPGIFINFO:
538 		tapp = (struct tapinfo *)data;
539 		tapp->mtu = ifp->if_mtu;
540 		tapp->type = ifp->if_type;
541 		tapp->baudrate = ifp->if_baudrate;
542 		break;
543 
544 	case TAPSDEBUG:
545 		tapdebug = *(int *)data;
546 		break;
547 
548 	case TAPGDEBUG:
549 		*(int *)data = tapdebug;
550 		break;
551 
552 	case FIOASYNC:
553 		if (*(int *)data)
554 			tp->tap_flags |= TAP_ASYNC;
555 		else
556 			tp->tap_flags &= ~TAP_ASYNC;
557 		break;
558 
559 	case FIONREAD:
560 		*(int *)data = 0;
561 		if ((mb = ifq_poll(&ifp->if_snd)) != NULL) {
562 			for(; mb != NULL; mb = mb->m_next)
563 				*(int *)data += mb->m_len;
564 		}
565 		break;
566 
567 	case FIOSETOWN:
568 		error = fsetown(*(int *)data, &tp->tap_sigio);
569 		break;
570 
571 	case FIOGETOWN:
572 		*(int *)data = fgetown(tp->tap_sigio);
573 		break;
574 
575 	/* this is deprecated, FIOSETOWN should be used instead */
576 	case TIOCSPGRP:
577 		error = fsetown(-(*(int *)data), &tp->tap_sigio);
578 		break;
579 
580 	/* this is deprecated, FIOGETOWN should be used instead */
581 	case TIOCGPGRP:
582 		*(int *)data = -fgetown(tp->tap_sigio);
583 		break;
584 
585 	/* VMware/VMnet port ioctl's */
586 
587 	case SIOCGIFFLAGS:	/* get ifnet flags */
588 		bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
589 		break;
590 
591 	case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
592 		f = *(short *)data;
593 		f &= 0x0fff;
594 		f &= ~IFF_CANTCHANGE;
595 		f |= IFF_UP;
596 		ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
597 		break;
598 
599 	case OSIOCGIFADDR:	/* get MAC address of the remote side */
600 	case SIOCGIFADDR:
601 		bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
602 		break;
603 
604 	case SIOCSIFADDR:	/* set MAC address of the remote side */
605 		bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
606 		break;
607 
608 	default:
609 		error = ENOTTY;
610 		break;
611 	}
612 	lwkt_serialize_exit(ifp->if_serializer);
613 	return (error);
614 }
615 
616 
617 /*
618  * tapread
619  *
620  * The ops read interface - reads a packet at a time, or at
621  * least as much of a packet as can be read.
622  *
623  * Called from the fileops interface with nothing held.
624  *
625  * MPSAFE
626  */
627 static int
628 tapread(struct dev_read_args *ap)
629 {
630 	cdev_t dev = ap->a_head.a_dev;
631 	struct uio *uio = ap->a_uio;
632 	struct tap_softc	*tp = dev->si_drv1;
633 	struct ifnet		*ifp = &tp->tap_if;
634 	struct mbuf		*m0 = NULL;
635 	int			 error = 0, len;
636 
637 	TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
638 
639 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
640 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
641 			 minor(tp->tap_dev), tp->tap_flags);
642 
643 		return (EHOSTDOWN);
644 	}
645 
646 	tp->tap_flags &= ~TAP_RWAIT;
647 
648 	/* sleep until we get a packet */
649 	do {
650 		lwkt_serialize_enter(ifp->if_serializer);
651 		m0 = ifq_dequeue(&ifp->if_snd, NULL);
652 		if (m0 == NULL) {
653 			if (ap->a_ioflag & IO_NDELAY) {
654 				lwkt_serialize_exit(ifp->if_serializer);
655 				return (EWOULDBLOCK);
656 			}
657 			tp->tap_flags |= TAP_RWAIT;
658 			crit_enter();
659 			tsleep_interlock(tp);
660 			lwkt_serialize_exit(ifp->if_serializer);
661 			error = tsleep(tp, PCATCH, "taprd", 0);
662 			crit_exit();
663 			if (error)
664 				return (error);
665 		} else {
666 			lwkt_serialize_exit(ifp->if_serializer);
667 		}
668 	} while (m0 == NULL);
669 
670 	BPF_MTAP(ifp, m0);
671 
672 	/* xfer packet to user space */
673 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
674 		len = min(uio->uio_resid, m0->m_len);
675 		if (len == 0)
676 			break;
677 
678 		error = uiomove(mtod(m0, caddr_t), len, uio);
679 		m0 = m_free(m0);
680 	}
681 
682 	if (m0 != NULL) {
683 		TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
684 			 minor(tp->tap_dev));
685 		m_freem(m0);
686 	}
687 
688 	return (error);
689 }
690 
691 /*
692  * tapwrite
693  *
694  * The ops write interface - an atomic write is a packet - or else!
695  *
696  * Called from the fileops interface with nothing held.
697  *
698  * MPSAFE
699  */
700 static int
701 tapwrite(struct dev_write_args *ap)
702 {
703 	cdev_t dev = ap->a_head.a_dev;
704 	struct uio *uio = ap->a_uio;
705 	struct tap_softc	*tp = dev->si_drv1;
706 	struct ifnet		*ifp = &tp->tap_if;
707 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
708 	int		 	 error = 0, tlen, mlen;
709 
710 	TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
711 
712 	if (uio->uio_resid == 0)
713 		return (0);
714 
715 	if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
716 		TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
717 			 uio->uio_resid, minor(tp->tap_dev));
718 
719 		return (EIO);
720 	}
721 	tlen = uio->uio_resid;
722 
723 	/* get a header mbuf */
724 	MGETHDR(m, MB_DONTWAIT, MT_DATA);
725 	if (m == NULL)
726 		return (ENOBUFS);
727 	mlen = MHLEN;
728 
729 	top = 0;
730 	mp = &top;
731 	while ((error == 0) && (uio->uio_resid > 0)) {
732 		m->m_len = min(mlen, uio->uio_resid);
733 		error = uiomove(mtod(m, caddr_t), m->m_len, uio);
734 		*mp = m;
735 		mp = &m->m_next;
736 		if (uio->uio_resid > 0) {
737 			MGET(m, MB_DONTWAIT, MT_DATA);
738 			if (m == NULL) {
739 				error = ENOBUFS;
740 				break;
741 			}
742 			mlen = MLEN;
743 		}
744 	}
745 	if (error) {
746 		ifp->if_ierrors ++;
747 		if (top)
748 			m_freem(top);
749 		return (error);
750 	}
751 
752 	top->m_pkthdr.len = tlen;
753 	top->m_pkthdr.rcvif = ifp;
754 
755 	/*
756 	 * Ethernet bridge and bpf are handled in ether_input
757 	 *
758 	 * adjust mbuf and give packet to the ether_input
759 	 */
760 	lwkt_serialize_enter(ifp->if_serializer);
761 	ifp->if_input(ifp, top);
762 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
763 	lwkt_serialize_exit(ifp->if_serializer);
764 
765 	return (0);
766 }
767 
768 /*
769  * tappoll
770  *
771  * The poll interface, this is only useful on reads really. The write
772  * detect always returns true, write never blocks anyway, it either
773  * accepts the packet or drops it
774  *
775  * Called from the fileops interface with nothing held.
776  *
777  * MPSAFE
778  */
779 static int
780 tappoll(struct dev_poll_args *ap)
781 {
782 	cdev_t dev = ap->a_head.a_dev;
783 	struct tap_softc	*tp = dev->si_drv1;
784 	struct ifnet		*ifp = &tp->tap_if;
785 	int		 	 revents = 0;
786 
787 	TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
788 
789 	lwkt_serialize_enter(ifp->if_serializer);
790 	if (ap->a_events & (POLLIN | POLLRDNORM)) {
791 		if (!ifq_is_empty(&ifp->if_snd)) {
792 			TAPDEBUG(ifp,
793 				 "has data in queue. minor = %#x\n",
794 				 minor(tp->tap_dev));
795 
796 			revents |= (ap->a_events & (POLLIN | POLLRDNORM));
797 		} else {
798 			TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
799 				 minor(tp->tap_dev));
800 
801 			get_mplock();
802 			selrecord(curthread, &tp->tap_rsel);
803 			rel_mplock();
804 		}
805 	}
806 	lwkt_serialize_exit(ifp->if_serializer);
807 
808 	if (ap->a_events & (POLLOUT | POLLWRNORM))
809 		revents |= (ap->a_events & (POLLOUT | POLLWRNORM));
810 	ap->a_events = revents;
811 	return(0);
812 }
813 
814 /*
815  * tapkqfilter - called from the fileops interface with nothing held
816  *
817  * MPSAFE
818  */
819 static int filt_tapread(struct knote *kn, long hint);
820 static void filt_tapdetach(struct knote *kn);
821 static struct filterops tapread_filtops =
822 	{ 1, NULL, filt_tapdetach, filt_tapread };
823 
824 static int
825 tapkqfilter(struct dev_kqfilter_args *ap)
826 {
827 	cdev_t dev = ap->a_head.a_dev;
828 	struct knote *kn = ap->a_kn;
829 	struct tap_softc *tp;
830 	struct klist *list;
831 	struct ifnet *ifp;
832 
833 	get_mplock();
834 	tp = dev->si_drv1;
835 	ifp = &tp->tap_if;
836 	ap->a_result =0;
837 
838 	switch(kn->kn_filter) {
839 	case EVFILT_READ:
840 		list = &tp->tap_rsel.si_note;
841 		kn->kn_fop = &tapread_filtops;
842 		kn->kn_hook = (void *)tp;
843 		break;
844 	case EVFILT_WRITE:
845 		/* fall through */
846 	default:
847 		ap->a_result = 1;
848 		rel_mplock();
849 		return(0);
850 	}
851 	crit_enter();
852 	SLIST_INSERT_HEAD(list, kn, kn_selnext);
853 	crit_exit();
854 	rel_mplock();
855 	return(0);
856 }
857 
858 static int
859 filt_tapread(struct knote *kn, long hint)
860 {
861 	struct tap_softc *tp = (void *)kn->kn_hook;
862 	struct ifnet *ifp = &tp->tap_if;
863 
864 	if (ifq_is_empty(&ifp->if_snd) == 0) {
865 		return(1);
866 	} else {
867 		return(0);
868 	}
869 }
870 
871 static void
872 filt_tapdetach(struct knote *kn)
873 {
874 	struct tap_softc *tp = (void *)kn->kn_hook;
875 
876 	SLIST_REMOVE(&tp->tap_rsel.si_note, kn, knote, kn_selnext);
877 }
878