xref: /dragonfly/sys/net/tap/if_tap.c (revision a32bc35d)
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  * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
36  */
37 
38 #include "opt_inet.h"
39 #include "use_tap.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/proc.h>
50 #include <sys/priv.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/ttycom.h>
57 #include <sys/uio.h>
58 #include <sys/vnode.h>
59 #include <sys/serialize.h>
60 
61 #include <sys/thread2.h>
62 #include <sys/mplock2.h>
63 
64 #include <net/bpf.h>
65 #include <net/ethernet.h>
66 #include <net/if.h>
67 #include <net/ifq_var.h>
68 #include <net/if_arp.h>
69 #include <net/if_clone.h>
70 #include <net/if_media.h>
71 #include <net/route.h>
72 #include <sys/devfs.h>
73 
74 #include <netinet/in.h>
75 
76 #include "if_tapvar.h"
77 #include "if_tap.h"
78 
79 #define TAP_IFFLAGS	(IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST)
80 
81 #if NTAP <= 1
82 #define TAP_PREALLOCATED_UNITS	4
83 #else
84 #define TAP_PREALLOCATED_UNITS	NTAP
85 #endif
86 
87 #define CDEV_NAME	"tap"
88 #define TAPDEBUG	if (tapdebug) if_printf
89 
90 #define TAP		"tap"
91 #define VMNET		"vmnet"
92 #define VMNET_DEV_MASK	0x00010000
93 
94 DEVFS_DECLARE_CLONE_BITMAP(tap);
95 
96 /* module */
97 static int 		tapmodevent	(module_t, int, void *);
98 
99 /* device */
100 static struct tap_softc *tapcreate(int, cdev_t);
101 static void		tapdestroy(struct tap_softc *);
102 
103 /* clone */
104 static int		tap_clone_create(struct if_clone *, int, caddr_t);
105 static int		tap_clone_destroy(struct ifnet *);
106 
107 
108 /* network interface */
109 static void		tapifstart	(struct ifnet *);
110 static int		tapifioctl	(struct ifnet *, u_long, caddr_t,
111 					 struct ucred *);
112 static void		tapifinit	(void *);
113 static void		tapifstop(struct tap_softc *, int);
114 static void		tapifflags(struct tap_softc *);
115 
116 
117 /* character device */
118 static d_open_t		tapopen;
119 static d_clone_t	tapclone;
120 static d_close_t	tapclose;
121 static d_read_t		tapread;
122 static d_write_t	tapwrite;
123 static d_ioctl_t	tapioctl;
124 static d_kqfilter_t	tapkqfilter;
125 
126 static struct dev_ops	tap_ops = {
127 	{ CDEV_NAME, 0, 0 },
128 	.d_open =	tapopen,
129 	.d_close =	tapclose,
130 	.d_read =	tapread,
131 	.d_write =	tapwrite,
132 	.d_ioctl =	tapioctl,
133 	.d_kqfilter =	tapkqfilter
134 };
135 
136 static int		taprefcnt = 0;		/* module ref. counter   */
137 static int		taplastunit = -1;	/* max. open unit number */
138 static int		tapdebug = 0;		/* debug flag            */
139 static int		tapuopen = 0;		/* all user open()       */
140 static int		tapuponopen = 0;	/* IFF_UP       */
141 
142 MALLOC_DECLARE(M_TAP);
143 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
144 struct if_clone tap_cloner = IF_CLONE_INITIALIZER("tap",
145 			     tap_clone_create, tap_clone_destroy,
146 			     0, IF_MAXUNIT);
147 static SLIST_HEAD(,tap_softc) tap_listhead =
148 	SLIST_HEAD_INITIALIZER(&tap_listhead);
149 
150 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
151 SYSCTL_DECL(_net_link);
152 SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
153 	    "Ethernet tunnel software network interface");
154 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
155 	   "Allow user to open /dev/tap (based on node permissions)");
156 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
157 	   "Bring interface up when /dev/tap is opened");
158 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
159 
160 DEV_MODULE(if_tap, tapmodevent, NULL);
161 
162 /*
163  * tapmodevent
164  *
165  * module event handler
166  */
167 static int
168 tapmodevent(module_t mod, int type, void *data)
169 {
170 	static int attached = 0;
171 	struct tap_softc *tp, *ntp;
172 	int i;
173 
174 	switch (type) {
175 	case MOD_LOAD:
176 		if (attached)
177 			return (EEXIST);
178 
179 		make_autoclone_dev(&tap_ops, &DEVFS_CLONE_BITMAP(tap), tapclone,
180 				   UID_ROOT, GID_WHEEL, 0600, "tap");
181 		SLIST_INIT(&tap_listhead);
182 		if_clone_attach(&tap_cloner);
183 
184 		for (i = 0; i < TAP_PREALLOCATED_UNITS; ++i) {
185 			make_dev(&tap_ops, i, UID_ROOT, GID_WHEEL,
186 				 0600, "tap%d", i);
187 			devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(tap), i);
188 		}
189 
190 		attached = 1;
191 		break;
192 
193 	case MOD_UNLOAD:
194 		if (taprefcnt > 0)
195 			return (EBUSY);
196 
197 		if_clone_detach(&tap_cloner);
198 
199 		/* Maintain tap ifs in a local list */
200 		SLIST_FOREACH_MUTABLE(tp, &tap_listhead, tap_link, ntp)
201 			tapdestroy(tp);
202 
203 		attached = 0;
204 
205 		devfs_clone_handler_del("tap");
206 		dev_ops_remove_all(&tap_ops);
207 		devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(tap));
208 		break;
209 
210 	default:
211 		return (EOPNOTSUPP);
212 	}
213 
214 	return (0);
215 } /* tapmodevent */
216 
217 
218 /*
219  * tapcreate - create or clone an interface
220  */
221 static struct tap_softc *
222 tapcreate(int unit, cdev_t dev)
223 {
224 	const char	*name = TAP;
225 	struct ifnet	*ifp;
226 	struct tap_softc *tp;
227 	uint8_t		ether_addr[ETHER_ADDR_LEN];
228 
229 	tp = kmalloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
230 	dev->si_drv1 = tp;
231 	tp->tap_dev = dev;
232 	tp->tap_unit = unit;
233 
234 	reference_dev(dev);	/* tp association */
235 
236 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
237 	ether_addr[0] = 0x00;
238 	ether_addr[1] = 0xbd;
239 	bcopy(&ticks, &ether_addr[2], 3);
240 	ether_addr[5] = (u_char)unit;
241 
242 	/* fill the rest and attach interface */
243 	ifp = &tp->tap_if;
244 	ifp->if_softc = tp;
245 
246 	if_initname(ifp, name, unit);
247 	if (unit > taplastunit)
248 		taplastunit = unit;
249 
250 	ifp->if_init = tapifinit;
251 	ifp->if_start = tapifstart;
252 	ifp->if_ioctl = tapifioctl;
253 	ifp->if_mtu = ETHERMTU;
254 	ifp->if_flags = TAP_IFFLAGS;
255 	ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
256 	ifq_set_ready(&ifp->if_snd);
257 
258 	ether_ifattach(ifp, ether_addr, NULL);
259 
260 	tp->tap_flags |= TAP_INITED;
261 	tp->tap_devq.ifq_maxlen = ifqmaxlen;
262 
263 	SLIST_INSERT_HEAD(&tap_listhead, tp, tap_link);
264 
265 	TAPDEBUG(ifp, "created. minor = %#x\n", minor(dev));
266 	return (tp);
267 }
268 
269 static
270 struct tap_softc *
271 tapfind(int unit)
272 {
273 	struct tap_softc *tp;
274 
275 	SLIST_FOREACH(tp, &tap_listhead, tap_link) {
276 		if (tp->tap_unit == unit)
277 			return(tp);
278 	}
279 	return (NULL);
280 }
281 
282 /*
283  * tap_clone_create:
284  *
285  * Create a new tap instance via ifconfig.
286  */
287 static int
288 tap_clone_create(struct if_clone *ifc __unused, int unit,
289     caddr_t param __unused)
290 {
291 	struct tap_softc *tp;
292 	cdev_t dev;
293 
294 	tp = tapfind(unit);
295 	if (tp == NULL) {
296 		if (!devfs_clone_bitmap_chk(&DEVFS_CLONE_BITMAP(tap), unit)) {
297 			devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(tap), unit);
298 			dev = make_dev(&tap_ops, unit, UID_ROOT, GID_WHEEL,
299 					   0600, "%s%d", TAP, unit);
300 		} else {
301 			dev = devfs_find_device_by_name("%s%d", TAP, unit);
302 		}
303 
304 		KKASSERT(dev != NULL);
305 		tp = tapcreate(unit, dev);
306 	}
307 	tp->tap_flags |= TAP_CLONE;
308 	TAPDEBUG(&tp->tap_if, "clone created. minor = %#x tap_flags = 0x%x\n",
309 		 minor(tp->tap_dev), tp->tap_flags);
310 
311 	return (0);
312 }
313 
314 /*
315  * tapopen
316  *
317  * to open tunnel. must be superuser
318  */
319 static int
320 tapopen(struct dev_open_args *ap)
321 {
322 	cdev_t dev = NULL;
323 	struct tap_softc *tp = NULL;
324 	struct ifnet *ifp = NULL;
325 	int error;
326 
327 	if (tapuopen == 0 &&
328 	    (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)) != 0)
329 		return (error);
330 
331 	get_mplock();
332 	dev = ap->a_head.a_dev;
333 	tp = dev->si_drv1;
334 	if (tp == NULL)
335 		tp = tapcreate(minor(dev), dev);
336 	if (tp->tap_flags & TAP_OPEN) {
337 		rel_mplock();
338 		return (EBUSY);
339 	}
340 	ifp = &tp->arpcom.ac_if;
341 
342 	if ((tp->tap_flags & TAP_CLONE) == 0) {
343 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
344 
345 		/* Announce the return of the interface. */
346 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
347 	}
348 
349 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
350 
351 	if (curthread->td_proc)
352 		fsetown(curthread->td_proc->p_pid, &tp->tap_sigtd);
353 	tp->tap_flags |= TAP_OPEN;
354 	taprefcnt ++;
355 
356 	if (tapuponopen && (ifp->if_flags & IFF_UP) == 0) {
357 		crit_enter();
358 		if_up(ifp);
359 		crit_exit();
360 
361 		ifnet_serialize_all(ifp);
362 		tapifflags(tp);
363 		ifnet_deserialize_all(ifp);
364 
365 		tp->tap_flags |= TAP_CLOSEDOWN;
366 	}
367 
368 	TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
369 		 minor(tp->tap_dev), taprefcnt, taplastunit);
370 
371 	rel_mplock();
372 	return (0);
373 }
374 
375 static int
376 tapclone(struct dev_clone_args *ap)
377 {
378 	int unit;
379 
380 	unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(tap), 0);
381 	ap->a_dev = make_only_dev(&tap_ops, unit, UID_ROOT, GID_WHEEL,
382 				  0600, "%s%d", TAP, unit);
383 	tapcreate(unit, ap->a_dev);
384 	return (0);
385 }
386 
387 /*
388  * tapclose
389  *
390  * close the device - mark i/f down & delete routing info
391  */
392 static int
393 tapclose(struct dev_close_args *ap)
394 {
395 	cdev_t dev = ap->a_head.a_dev;
396 	struct tap_softc *tp = dev->si_drv1;
397 	struct ifnet *ifp = &tp->tap_if;
398 	int clear_flags = 0;
399 
400 	get_mplock();
401 
402 	/* Junk all pending output */
403 	ifq_purge(&ifp->if_snd);
404 
405 	/*
406 	 * Do not bring the interface down, and do not anything with
407 	 * interface, if we are in VMnet mode. just close the device.
408 	 *
409 	 * If the interface is not cloned, we always bring it down.
410 	 *
411 	 * If the interface is cloned, then we bring it down during
412 	 * closing only if it was brought up during opening.
413 	 */
414 	if ((tp->tap_flags & TAP_VMNET) == 0 &&
415 	    ((tp->tap_flags & TAP_CLONE) == 0 ||
416 	     (tp->tap_flags & TAP_CLOSEDOWN))) {
417 		if (ifp->if_flags & IFF_UP)
418 			if_down(ifp);
419 		clear_flags = 1;
420 	}
421 	ifnet_serialize_all(ifp);
422 	tapifstop(tp, clear_flags);
423 	ifnet_deserialize_all(ifp);
424 
425 	if ((tp->tap_flags & TAP_CLONE) == 0) {
426 		if_purgeaddrs_nolink(ifp);
427 
428 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
429 
430 		/* Announce the departure of the interface. */
431 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
432 	}
433 
434 	funsetown(&tp->tap_sigio);
435 	tp->tap_sigio = NULL;
436 	KNOTE(&tp->tap_rkq.ki_note, 0);
437 
438 	tp->tap_flags &= ~TAP_OPEN;
439 	funsetown(&tp->tap_sigtd);
440 	tp->tap_sigtd = NULL;
441 
442 	taprefcnt --;
443 	if (taprefcnt < 0) {
444 		taprefcnt = 0;
445 		if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
446 			"set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
447 	}
448 
449 	TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
450 		 minor(tp->tap_dev), taprefcnt, taplastunit);
451 
452 	if (tp->tap_unit >= TAP_PREALLOCATED_UNITS)
453 		tapdestroy(tp);
454 
455 	rel_mplock();
456 	return (0);
457 }
458 
459 /*
460  * tapdestroy:
461  *
462  *	Destroy a tap instance.
463  */
464 static void
465 tapdestroy(struct tap_softc *tp)
466 {
467 	struct ifnet *ifp = &tp->arpcom.ac_if;
468 	cdev_t dev;
469 
470 	TAPDEBUG(ifp, "destroyed. minor = %#x, refcnt = %d, taplastunit = %d\n",
471 		 minor(tp->tap_dev), taprefcnt, taplastunit);
472 
473 	ifnet_serialize_all(ifp);
474 	tapifstop(tp, 1);
475 	ifnet_deserialize_all(ifp);
476 
477 	ether_ifdetach(ifp);
478 	SLIST_REMOVE(&tap_listhead, tp, tap_softc, tap_link);
479 
480 	dev = tp->tap_dev;
481 	tp->tap_dev = NULL;
482 	dev->si_drv1 = NULL;
483 
484 	release_dev(dev);	/* tp association */
485 
486 	/*
487 	 * Also destroy the cloned device
488 	 */
489 	if (tp->tap_unit >= TAP_PREALLOCATED_UNITS) {
490 		destroy_dev(dev);
491 		devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(tap), tp->tap_unit);
492 	}
493 
494 	kfree(tp, M_TAP);
495 
496 	taplastunit--;
497 }
498 
499 /*
500  * tap_clone_destroy:
501  *
502  *	Destroy a tap instance.
503  */
504 static int
505 tap_clone_destroy(struct ifnet *ifp)
506 {
507 	struct tap_softc *tp = ifp->if_softc;
508 
509 	if ((tp->tap_flags & TAP_CLONE) == 0)
510 		return ENXIO;
511 
512 	TAPDEBUG(&tp->tap_if, "clone destroyed. minor = %#x tap_flags = 0x%x\n",
513 		 minor(tp->tap_dev), tp->tap_flags);
514 	tapdestroy(tp);
515 
516 	return 0;
517 }
518 
519 /*
520  * tapifinit
521  *
522  * Network interface initialization function (called with if serializer held)
523  *
524  * MPSAFE
525  */
526 static void
527 tapifinit(void *xtp)
528 {
529 	struct tap_softc *tp = xtp;
530 	struct ifnet *ifp = &tp->tap_if;
531 
532 	TAPDEBUG(ifp, "initializing, minor = %#x tap_flags = 0x%x\n",
533 		 minor(tp->tap_dev), tp->tap_flags);
534 
535 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
536 
537 	tapifstop(tp, 1);
538 
539 	ifp->if_flags |= IFF_RUNNING;
540 	ifp->if_flags &= ~IFF_OACTIVE;
541 
542 	/* attempt to start output */
543 	tapifstart(ifp);
544 }
545 
546 
547 /*
548  * tapifioctl
549  *
550  * Process an ioctl request on network interface (called with if serializer
551  * held).
552  *
553  * MPSAFE
554  */
555 static int
556 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
557 {
558 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
559 	struct ifstat		*ifs = NULL;
560 	struct ifmediareq	*ifmr = NULL;
561 	int			error = 0;
562 	int			dummy;
563 
564 	switch (cmd) {
565 		case SIOCSIFADDR:
566 		case SIOCGIFADDR:
567 		case SIOCSIFMTU:
568 			error = ether_ioctl(ifp, cmd, data);
569 			break;
570 
571 		case SIOCSIFFLAGS:
572 			tapifflags(tp);
573 			break;
574 
575 		case SIOCADDMULTI: /* XXX -- just like vmnet does */
576 		case SIOCDELMULTI:
577 			break;
578 
579 		case SIOCGIFMEDIA:
580 			/*
581 			 * The bridge code needs this when running the
582 			 * spanning tree protocol.
583 			 */
584 			ifmr = (struct ifmediareq *)data;
585 			dummy = ifmr->ifm_count;
586 			ifmr->ifm_count = 1;
587 			ifmr->ifm_status = IFM_AVALID;
588 			ifmr->ifm_active = IFM_ETHER;
589 			if (tp->tap_flags & TAP_OPEN)
590 				ifmr->ifm_status |= IFM_ACTIVE;
591 			ifmr->ifm_current = ifmr->ifm_active;
592 			if (dummy >= 1) {
593 				int media = IFM_ETHER;
594 				error = copyout(&media,
595 						ifmr->ifm_ulist,
596 						sizeof(int));
597 			}
598 			break;
599 
600 		case SIOCGIFSTATUS:
601 			ifs = (struct ifstat *)data;
602 			dummy = strlen(ifs->ascii);
603 			if ((tp->tap_flags & TAP_OPEN) &&
604 			    dummy < sizeof(ifs->ascii)) {
605 				if (tp->tap_sigtd && tp->tap_sigtd->sio_proc) {
606 				    ksnprintf(ifs->ascii + dummy,
607 					sizeof(ifs->ascii) - dummy,
608 					"\tOpened by pid %d\n",
609 					(int)tp->tap_sigtd->sio_proc->p_pid);
610 				} else {
611 				    ksnprintf(ifs->ascii + dummy,
612 					sizeof(ifs->ascii) - dummy,
613 					"\tOpened by <unknown>\n");
614 				}
615 			}
616 			break;
617 
618 		default:
619 			error = EINVAL;
620 			break;
621 	}
622 
623 	return (error);
624 }
625 
626 
627 /*
628  * tapifstart
629  *
630  * Queue packets from higher level ready to put out (called with if serializer
631  * held)
632  *
633  * MPSAFE
634  */
635 static void
636 tapifstart(struct ifnet *ifp)
637 {
638 	struct tap_softc *tp = ifp->if_softc;
639 	struct ifqueue *ifq;
640 	struct mbuf *m;
641 	int has_data = 0;
642 
643 	TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
644 
645 	/*
646 	 * do not junk pending output if we are in VMnet mode.
647 	 * XXX: can this do any harm because of queue overflow?
648 	 */
649 
650 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
651 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
652 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
653 			 minor(tp->tap_dev), tp->tap_flags);
654 		ifq_purge(&ifp->if_snd);
655 		return;
656 	}
657 
658 	ifp->if_flags |= IFF_OACTIVE;
659 
660 	ifq = &tp->tap_devq;
661 	while ((m = ifq_dequeue(&ifp->if_snd, NULL)) != NULL) {
662 		if (IF_QFULL(ifq)) {
663 			IF_DROP(ifq);
664 			ifp->if_oerrors++;
665 			m_freem(m);
666 		} else {
667 			IF_ENQUEUE(ifq, m);
668 			ifp->if_opackets++;
669 			has_data = 1;
670 		}
671 	}
672 
673 	if (has_data) {
674 		if (tp->tap_flags & TAP_RWAIT) {
675 			tp->tap_flags &= ~TAP_RWAIT;
676 			wakeup((caddr_t)tp);
677 		}
678 
679 		KNOTE(&tp->tap_rkq.ki_note, 0);
680 
681 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
682 			get_mplock();
683 			pgsigio(tp->tap_sigio, SIGIO, 0);
684 			rel_mplock();
685 		}
686 	}
687 
688 	ifp->if_flags &= ~IFF_OACTIVE;
689 }
690 
691 
692 /*
693  * tapioctl
694  *
695  * The ops interface is now pretty minimal.  Called via fileops with nothing
696  * held.
697  *
698  * MPSAFE
699  */
700 static int
701 tapioctl(struct dev_ioctl_args *ap)
702 {
703 	cdev_t dev = ap->a_head.a_dev;
704 	caddr_t data = ap->a_data;
705 	struct tap_softc	*tp = dev->si_drv1;
706 	struct ifnet		*ifp = &tp->tap_if;
707  	struct tapinfo		*tapp = NULL;
708 	struct mbuf *mb;
709 	short f;
710 	int error;
711 
712 	ifnet_serialize_all(ifp);
713 	error = 0;
714 
715 	switch (ap->a_cmd) {
716 	case TAPSIFINFO:
717 		tapp = (struct tapinfo *)data;
718 		ifp->if_mtu = tapp->mtu;
719 		ifp->if_type = tapp->type;
720 		ifp->if_baudrate = tapp->baudrate;
721 		break;
722 
723 	case TAPGIFINFO:
724 		tapp = (struct tapinfo *)data;
725 		tapp->mtu = ifp->if_mtu;
726 		tapp->type = ifp->if_type;
727 		tapp->baudrate = ifp->if_baudrate;
728 		break;
729 
730 	case TAPSDEBUG:
731 		tapdebug = *(int *)data;
732 		break;
733 
734 	case TAPGDEBUG:
735 		*(int *)data = tapdebug;
736 		break;
737 
738 	case FIOASYNC:
739 		if (*(int *)data)
740 			tp->tap_flags |= TAP_ASYNC;
741 		else
742 			tp->tap_flags &= ~TAP_ASYNC;
743 		break;
744 
745 	case FIONREAD:
746 		*(int *)data = 0;
747 
748 		/* Take a look at devq first */
749 		IF_POLL(&tp->tap_devq, mb);
750 		if (mb == NULL)
751 			mb = ifq_poll(&ifp->if_snd);
752 
753 		if (mb != NULL) {
754 			for(; mb != NULL; mb = mb->m_next)
755 				*(int *)data += mb->m_len;
756 		}
757 		break;
758 
759 	case FIOSETOWN:
760 		error = fsetown(*(int *)data, &tp->tap_sigio);
761 		break;
762 
763 	case FIOGETOWN:
764 		*(int *)data = fgetown(&tp->tap_sigio);
765 		break;
766 
767 	/* this is deprecated, FIOSETOWN should be used instead */
768 	case TIOCSPGRP:
769 		error = fsetown(-(*(int *)data), &tp->tap_sigio);
770 		break;
771 
772 	/* this is deprecated, FIOGETOWN should be used instead */
773 	case TIOCGPGRP:
774 		*(int *)data = -fgetown(&tp->tap_sigio);
775 		break;
776 
777 	/* VMware/VMnet port ioctl's */
778 
779 	case SIOCGIFFLAGS:	/* get ifnet flags */
780 		bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
781 		break;
782 
783 	case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
784 		f = *(short *)data;
785 		f &= 0x0fff;
786 		f &= ~IFF_CANTCHANGE;
787 		f |= IFF_UP;
788 		ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
789 		break;
790 
791 	case OSIOCGIFADDR:	/* get MAC address of the remote side */
792 	case SIOCGIFADDR:
793 		bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
794 		break;
795 
796 	case SIOCSIFADDR:	/* set MAC address of the remote side */
797 		bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
798 		break;
799 
800 	default:
801 		error = ENOTTY;
802 		break;
803 	}
804 	ifnet_deserialize_all(ifp);
805 	return (error);
806 }
807 
808 
809 /*
810  * tapread
811  *
812  * The ops read interface - reads a packet at a time, or at
813  * least as much of a packet as can be read.
814  *
815  * Called from the fileops interface with nothing held.
816  *
817  * MPSAFE
818  */
819 static int
820 tapread(struct dev_read_args *ap)
821 {
822 	cdev_t dev = ap->a_head.a_dev;
823 	struct uio *uio = ap->a_uio;
824 	struct tap_softc	*tp = dev->si_drv1;
825 	struct ifnet		*ifp = &tp->tap_if;
826 	struct mbuf		*m0 = NULL;
827 	int			 error = 0, len;
828 
829 	TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
830 
831 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
832 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
833 			 minor(tp->tap_dev), tp->tap_flags);
834 
835 		return (EHOSTDOWN);
836 	}
837 
838 	tp->tap_flags &= ~TAP_RWAIT;
839 
840 	/* sleep until we get a packet */
841 	do {
842 		ifnet_serialize_all(ifp);
843 		IF_DEQUEUE(&tp->tap_devq, m0);
844 		if (m0 == NULL) {
845 			if (ap->a_ioflag & IO_NDELAY) {
846 				ifnet_deserialize_all(ifp);
847 				return (EWOULDBLOCK);
848 			}
849 			tp->tap_flags |= TAP_RWAIT;
850 			tsleep_interlock(tp, PCATCH);
851 			ifnet_deserialize_all(ifp);
852 			error = tsleep(tp, PCATCH | PINTERLOCKED, "taprd", 0);
853 			if (error)
854 				return (error);
855 		} else {
856 			ifnet_deserialize_all(ifp);
857 		}
858 	} while (m0 == NULL);
859 
860 	BPF_MTAP(ifp, m0);
861 
862 	/* xfer packet to user space */
863 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
864 		len = (int)szmin(uio->uio_resid, m0->m_len);
865 		if (len == 0)
866 			break;
867 
868 		error = uiomove(mtod(m0, caddr_t), (size_t)len, uio);
869 		m0 = m_free(m0);
870 	}
871 
872 	if (m0 != NULL) {
873 		TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
874 			 minor(tp->tap_dev));
875 		m_freem(m0);
876 	}
877 
878 	return (error);
879 }
880 
881 /*
882  * tapwrite
883  *
884  * The ops write interface - an atomic write is a packet - or else!
885  *
886  * Called from the fileops interface with nothing held.
887  *
888  * MPSAFE
889  */
890 static int
891 tapwrite(struct dev_write_args *ap)
892 {
893 	cdev_t dev = ap->a_head.a_dev;
894 	struct uio *uio = ap->a_uio;
895 	struct tap_softc	*tp = dev->si_drv1;
896 	struct ifnet		*ifp = &tp->tap_if;
897 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
898 	int			error;
899 	size_t			tlen, mlen;
900 
901 	TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
902 
903 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
904 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
905 			 minor(tp->tap_dev), tp->tap_flags);
906 		return (EHOSTDOWN);
907 	}
908 
909 	if (uio->uio_resid == 0)
910 		return (0);
911 
912 	if (uio->uio_resid > TAPMRU) {
913 		TAPDEBUG(ifp, "invalid packet len = %zu, minor = %#x\n",
914 			 uio->uio_resid, minor(tp->tap_dev));
915 
916 		return (EIO);
917 	}
918 	tlen = uio->uio_resid;
919 
920 	/* get a header mbuf */
921 	MGETHDR(m, MB_WAIT, MT_DATA);
922 	if (m == NULL)
923 		return (ENOBUFS);
924 	mlen = MHLEN;
925 
926 	top = NULL;
927 	mp = &top;
928 	error = 0;
929 
930 	while (error == 0 && uio->uio_resid > 0) {
931 		m->m_len = (int)szmin(mlen, uio->uio_resid);
932 		error = uiomove(mtod(m, caddr_t), (size_t)m->m_len, uio);
933 		*mp = m;
934 		mp = &m->m_next;
935 		if (uio->uio_resid > 0) {
936 			MGET(m, MB_WAIT, MT_DATA);
937 			if (m == NULL) {
938 				error = ENOBUFS;
939 				break;
940 			}
941 			mlen = MLEN;
942 		}
943 	}
944 	if (error) {
945 		ifp->if_ierrors ++;
946 		if (top)
947 			m_freem(top);
948 		return (error);
949 	}
950 
951 	top->m_pkthdr.len = (int)tlen;
952 	top->m_pkthdr.rcvif = ifp;
953 
954 	/*
955 	 * Ethernet bridge and bpf are handled in ether_input
956 	 *
957 	 * adjust mbuf and give packet to the ether_input
958 	 */
959 	ifnet_serialize_all(ifp);
960 	ifp->if_input(ifp, top);
961 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
962 	ifnet_deserialize_all(ifp);
963 
964 	return (0);
965 }
966 
967 /*
968  * tapkqfilter - called from the fileops interface with nothing held
969  *
970  * MPSAFE
971  */
972 static int filt_tapread(struct knote *kn, long hint);
973 static int filt_tapwrite(struct knote *kn, long hint);
974 static void filt_tapdetach(struct knote *kn);
975 static struct filterops tapread_filtops =
976 	{ FILTEROP_ISFD, NULL, filt_tapdetach, filt_tapread };
977 static struct filterops tapwrite_filtops =
978 	{ FILTEROP_ISFD, NULL, filt_tapdetach, filt_tapwrite };
979 
980 static int
981 tapkqfilter(struct dev_kqfilter_args *ap)
982 {
983 	cdev_t dev = ap->a_head.a_dev;
984 	struct knote *kn = ap->a_kn;
985 	struct tap_softc *tp;
986 	struct klist *list;
987 	struct ifnet *ifp;
988 
989 	tp = dev->si_drv1;
990 	list = &tp->tap_rkq.ki_note;
991 	ifp = &tp->tap_if;
992 	ap->a_result =0;
993 
994 	switch(kn->kn_filter) {
995 	case EVFILT_READ:
996 		kn->kn_fop = &tapread_filtops;
997 		kn->kn_hook = (void *)tp;
998 		break;
999 	case EVFILT_WRITE:
1000 		kn->kn_fop = &tapwrite_filtops;
1001 		kn->kn_hook = (void *)tp;
1002 		break;
1003 	default:
1004 		ap->a_result = EOPNOTSUPP;
1005 		return(0);
1006 	}
1007 
1008 	knote_insert(list, kn);
1009 	return(0);
1010 }
1011 
1012 static int
1013 filt_tapread(struct knote *kn, long hint)
1014 {
1015 	struct tap_softc *tp = (void *)kn->kn_hook;
1016 
1017 	if (IF_QEMPTY(&tp->tap_devq) == 0)	/* XXX serializer */
1018 		return(1);
1019 	else
1020 		return(0);
1021 }
1022 
1023 static int
1024 filt_tapwrite(struct knote *kn, long hint)
1025 {
1026 	/* Always ready for a write */
1027 	return (1);
1028 }
1029 
1030 static void
1031 filt_tapdetach(struct knote *kn)
1032 {
1033 	struct tap_softc *tp = (void *)kn->kn_hook;
1034 
1035 	knote_remove(&tp->tap_rkq.ki_note, kn);
1036 }
1037 
1038 static void
1039 tapifstop(struct tap_softc *tp, int clear_flags)
1040 {
1041 	struct ifnet *ifp = &tp->tap_if;
1042 
1043 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
1044 	IF_DRAIN(&tp->tap_devq);
1045 	tp->tap_flags &= ~TAP_CLOSEDOWN;
1046 	if (clear_flags)
1047 		ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
1048 }
1049 
1050 static void
1051 tapifflags(struct tap_softc *tp)
1052 {
1053 	struct ifnet *ifp = &tp->arpcom.ac_if;
1054 
1055 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
1056 	if ((tp->tap_flags & TAP_VMNET) == 0) {
1057 		/*
1058 		 * Only for non-vmnet tap(4)
1059 		 */
1060 		if (ifp->if_flags & IFF_UP) {
1061 			if ((ifp->if_flags & IFF_RUNNING) == 0)
1062 				tapifinit(tp);
1063 		} else {
1064 			tapifstop(tp, 1);
1065 		}
1066 	} else {
1067 		/* XXX */
1068 	}
1069 }
1070