xref: /dragonfly/sys/net/tap/if_tap.c (revision 88cfb1f7)
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 
40 #include <sys/param.h>
41 #include <sys/conf.h>
42 #include <sys/device.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/proc.h>
49 #include <sys/priv.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 #include <sys/serialize.h>
59 
60 #include <sys/thread2.h>
61 #include <sys/mplock2.h>
62 
63 #include <net/bpf.h>
64 #include <net/ethernet.h>
65 #include <net/if.h>
66 #include <net/ifq_var.h>
67 #include <net/if_arp.h>
68 #include <net/if_clone.h>
69 #include <net/route.h>
70 #include <sys/devfs.h>
71 
72 #include <netinet/in.h>
73 
74 #include "if_tapvar.h"
75 #include "if_tap.h"
76 
77 #define TAP_IFFLAGS	(IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST)
78 
79 #define TAP_PREALLOCATED_UNITS	4
80 
81 #define CDEV_NAME	"tap"
82 #define TAPDEBUG	if (tapdebug) if_printf
83 
84 #define TAP		"tap"
85 #define VMNET		"vmnet"
86 #define VMNET_DEV_MASK	0x00010000
87 
88 DEVFS_DECLARE_CLONE_BITMAP(tap);
89 
90 /* module */
91 static int 		tapmodevent	(module_t, int, void *);
92 
93 /* device */
94 static struct tap_softc *tapcreate(int, cdev_t);
95 static void		tapdestroy(struct tap_softc *);
96 
97 /* clone */
98 static int		tap_clone_create(struct if_clone *, int, caddr_t);
99 static void		tap_clone_destroy(struct ifnet *);
100 
101 
102 /* network interface */
103 static void		tapifstart	(struct ifnet *);
104 static int		tapifioctl	(struct ifnet *, u_long, caddr_t,
105 					 struct ucred *);
106 static void		tapifinit	(void *);
107 static void		tapifstop(struct tap_softc *, int);
108 static void		tapifflags(struct tap_softc *);
109 
110 /* character device */
111 static d_open_t		tapopen;
112 static d_clone_t	tapclone;
113 static d_close_t	tapclose;
114 static d_read_t		tapread;
115 static d_write_t	tapwrite;
116 static d_ioctl_t	tapioctl;
117 static d_kqfilter_t	tapkqfilter;
118 
119 static struct dev_ops	tap_ops = {
120 	{ CDEV_NAME, 0, 0 },
121 	.d_open =	tapopen,
122 	.d_close =	tapclose,
123 	.d_read =	tapread,
124 	.d_write =	tapwrite,
125 	.d_ioctl =	tapioctl,
126 	.d_kqfilter =	tapkqfilter
127 };
128 
129 static int		taprefcnt = 0;		/* module ref. counter   */
130 static int		taplastunit = -1;	/* max. open unit number */
131 static int		tapdebug = 0;		/* debug flag            */
132 static int		tapuopen = 0;		/* all user open()       */
133 static int		tapuponopen = 0;	/* IFF_UP       */
134 
135 MALLOC_DECLARE(M_TAP);
136 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
137 struct if_clone tap_cloner = IF_CLONE_INITIALIZER("tap",
138 			     tap_clone_create, tap_clone_destroy,
139 			     0, IF_MAXUNIT);
140 static SLIST_HEAD(,tap_softc) tap_listhead =
141 	SLIST_HEAD_INITIALIZER(&tap_listhead);
142 
143 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
144 SYSCTL_DECL(_net_link);
145 SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
146 	    "Ethernet tunnel software network interface");
147 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
148 	   "Allow user to open /dev/tap (based on node permissions)");
149 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
150 	   "Bring interface up when /dev/tap is opened");
151 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
152 
153 DEV_MODULE(if_tap, tapmodevent, NULL);
154 
155 /*
156  * tapmodevent
157  *
158  * module event handler
159  */
160 static int
161 tapmodevent(module_t mod, int type, void *data)
162 {
163 	static int attached = 0;
164 	struct tap_softc *tp, *ntp;
165 	int i;
166 
167 	switch (type) {
168 	case MOD_LOAD:
169 		if (attached)
170 			return (EEXIST);
171 
172 		make_autoclone_dev(&tap_ops, &DEVFS_CLONE_BITMAP(tap), tapclone,
173 				   UID_ROOT, GID_WHEEL, 0600, "tap");
174 		SLIST_INIT(&tap_listhead);
175 		if_clone_attach(&tap_cloner);
176 
177 		for (i = 0; i < TAP_PREALLOCATED_UNITS; ++i) {
178 			make_dev(&tap_ops, i, UID_ROOT, GID_WHEEL,
179 				 0600, "tap%d", i);
180 			devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(tap), i);
181 		}
182 
183 		attached = 1;
184 		break;
185 
186 	case MOD_UNLOAD:
187 		if (taprefcnt > 0)
188 			return (EBUSY);
189 
190 		if_clone_detach(&tap_cloner);
191 
192 		/* Maintain tap ifs in a local list */
193 		SLIST_FOREACH_MUTABLE(tp, &tap_listhead, tap_link, ntp)
194 			tapdestroy(tp);
195 
196 		attached = 0;
197 
198 		devfs_clone_handler_del("tap");
199 		dev_ops_remove_all(&tap_ops);
200 		devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(tap));
201 		break;
202 
203 	default:
204 		return (EOPNOTSUPP);
205 	}
206 
207 	return (0);
208 } /* tapmodevent */
209 
210 
211 /*
212  * tapcreate - create or clone an interface
213  */
214 static struct tap_softc *
215 tapcreate(int unit, cdev_t dev)
216 {
217 	const char	*name = TAP;
218 	struct ifnet	*ifp;
219 	struct tap_softc *tp;
220 	uint8_t		ether_addr[ETHER_ADDR_LEN];
221 
222 	tp = kmalloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
223 	dev->si_drv1 = tp;
224 	tp->tap_dev = dev;
225 	tp->tap_unit = unit;
226 
227 	reference_dev(dev);	/* tp association */
228 
229 	/* generate fake MAC address: 00 bd xx xx xx unit_no */
230 	ether_addr[0] = 0x00;
231 	ether_addr[1] = 0xbd;
232 	bcopy(&ticks, &ether_addr[2], 3);
233 	ether_addr[5] = (u_char)unit;
234 
235 	/* fill the rest and attach interface */
236 	ifp = &tp->tap_if;
237 	ifp->if_softc = tp;
238 
239 	if_initname(ifp, name, unit);
240 	if (unit > taplastunit)
241 		taplastunit = unit;
242 
243 	ifp->if_init = tapifinit;
244 	ifp->if_start = tapifstart;
245 	ifp->if_ioctl = tapifioctl;
246 	ifp->if_mtu = ETHERMTU;
247 	ifp->if_flags = TAP_IFFLAGS;
248 	ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
249 	ifq_set_ready(&ifp->if_snd);
250 
251 	ether_ifattach(ifp, ether_addr, NULL);
252 
253 	tp->tap_flags |= TAP_INITED;
254 	tp->tap_devq.ifq_maxlen = ifqmaxlen;
255 
256 	SLIST_INSERT_HEAD(&tap_listhead, tp, tap_link);
257 
258 	TAPDEBUG(ifp, "created. minor = %#x\n", minor(dev));
259 	return (tp);
260 }
261 
262 static
263 struct tap_softc *
264 tapfind(int unit)
265 {
266 	struct tap_softc *tp;
267 
268 	SLIST_FOREACH(tp, &tap_listhead, tap_link) {
269 		if (tp->tap_unit == unit)
270 			return(tp);
271 	}
272 	return (NULL);
273 }
274 
275 /*
276  * tap_clone_create:
277  *
278  * Create a new tap instance via ifconfig.
279  */
280 static int
281 tap_clone_create(struct if_clone *ifc __unused, int unit,
282     caddr_t param __unused)
283 {
284 	struct tap_softc *tp;
285 	cdev_t dev;
286 
287 	tp = tapfind(unit);
288 	if (tp == NULL) {
289 		if (!devfs_clone_bitmap_chk(&DEVFS_CLONE_BITMAP(tap), unit)) {
290 			devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(tap), unit);
291 			dev = make_dev(&tap_ops, unit, UID_ROOT, GID_WHEEL,
292 					   0600, "%s%d", TAP, unit);
293 		} else {
294 			dev = devfs_find_device_by_name("%s%d", TAP, unit);
295 		}
296 
297 		KKASSERT(dev != NULL);
298 		tp = tapcreate(unit, dev);
299 	}
300 	tp->tap_flags |= TAP_CLONE;
301 	TAPDEBUG(&tp->tap_if, "clone created. minor = %#x tap_flags = 0x%x\n",
302 		 minor(tp->tap_dev), tp->tap_flags);
303 
304 	return (0);
305 }
306 
307 /*
308  * tapopen
309  *
310  * to open tunnel. must be superuser
311  */
312 static int
313 tapopen(struct dev_open_args *ap)
314 {
315 	cdev_t dev = NULL;
316 	struct tap_softc *tp = NULL;
317 	struct ifnet *ifp = NULL;
318 	int error;
319 
320 	if (tapuopen == 0 &&
321 	    (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)) != 0)
322 		return (error);
323 
324 	get_mplock();
325 	dev = ap->a_head.a_dev;
326 	tp = dev->si_drv1;
327 	if (tp == NULL)
328 		tp = tapcreate(minor(dev), dev);
329 	if (tp->tap_flags & TAP_OPEN) {
330 		rel_mplock();
331 		return (EBUSY);
332 	}
333 	ifp = &tp->arpcom.ac_if;
334 
335 	if ((tp->tap_flags & TAP_CLONE) == 0) {
336 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
337 
338 		/* Announce the return of the interface. */
339 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
340 	}
341 
342 	bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
343 
344 	if (curthread->td_proc)
345 		fsetown(curthread->td_proc->p_pid, &tp->tap_sigtd);
346 	tp->tap_flags |= TAP_OPEN;
347 	taprefcnt ++;
348 
349 	if (tapuponopen && (ifp->if_flags & IFF_UP) == 0) {
350 		crit_enter();
351 		if_up(ifp);
352 		crit_exit();
353 
354 		ifnet_serialize_all(ifp);
355 		tapifflags(tp);
356 		ifnet_deserialize_all(ifp);
357 
358 		tp->tap_flags |= TAP_CLOSEDOWN;
359 	}
360 
361 	TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
362 		 minor(tp->tap_dev), taprefcnt, taplastunit);
363 
364 	rel_mplock();
365 	return (0);
366 }
367 
368 static int
369 tapclone(struct dev_clone_args *ap)
370 {
371 	int unit;
372 
373 	unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(tap), 0);
374 	ap->a_dev = make_only_dev(&tap_ops, unit, UID_ROOT, GID_WHEEL,
375 				  0600, "%s%d", TAP, unit);
376 	tapcreate(unit, ap->a_dev);
377 	return (0);
378 }
379 
380 /*
381  * tapclose
382  *
383  * close the device - mark i/f down & delete routing info
384  */
385 static int
386 tapclose(struct dev_close_args *ap)
387 {
388 	cdev_t dev = ap->a_head.a_dev;
389 	struct tap_softc *tp = dev->si_drv1;
390 	struct ifnet *ifp = &tp->tap_if;
391 	int clear_flags = 0;
392 
393 	get_mplock();
394 
395 	/* Junk all pending output */
396 	ifq_purge(&ifp->if_snd);
397 
398 	/*
399 	 * Do not bring the interface down, and do not anything with
400 	 * interface, if we are in VMnet mode. just close the device.
401 	 *
402 	 * If the interface is not cloned, we always bring it down.
403 	 *
404 	 * If the interface is cloned, then we bring it down during
405 	 * closing only if it was brought up during opening.
406 	 */
407 	if ((tp->tap_flags & TAP_VMNET) == 0 &&
408 	    ((tp->tap_flags & TAP_CLONE) == 0 ||
409 	     (tp->tap_flags & TAP_CLOSEDOWN))) {
410 		if (ifp->if_flags & IFF_UP)
411 			if_down(ifp);
412 		clear_flags = 1;
413 	}
414 	ifnet_serialize_all(ifp);
415 	tapifstop(tp, clear_flags);
416 	ifnet_deserialize_all(ifp);
417 
418 	if ((tp->tap_flags & TAP_CLONE) == 0) {
419 		if_purgeaddrs_nolink(ifp);
420 
421 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
422 
423 		/* Announce the departure of the interface. */
424 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
425 	}
426 
427 	funsetown(tp->tap_sigio);
428 	tp->tap_sigio = NULL;
429 	KNOTE(&tp->tap_rkq.ki_note, 0);
430 
431 	tp->tap_flags &= ~TAP_OPEN;
432 	funsetown(tp->tap_sigtd);
433 	tp->tap_sigtd = NULL;
434 
435 	taprefcnt --;
436 	if (taprefcnt < 0) {
437 		taprefcnt = 0;
438 		if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
439 			"set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
440 	}
441 
442 	TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
443 		 minor(tp->tap_dev), taprefcnt, taplastunit);
444 
445 	if (tp->tap_unit >= TAP_PREALLOCATED_UNITS)
446 		tapdestroy(tp);
447 
448 	rel_mplock();
449 	return (0);
450 }
451 
452 /*
453  * tapdestroy:
454  *
455  *	Destroy a tap instance.
456  */
457 static void
458 tapdestroy(struct tap_softc *tp)
459 {
460 	struct ifnet *ifp = &tp->arpcom.ac_if;
461 	cdev_t dev;
462 
463 	TAPDEBUG(ifp, "destroyed. minor = %#x, refcnt = %d, taplastunit = %d\n",
464 		 minor(tp->tap_dev), taprefcnt, taplastunit);
465 
466 	ifnet_serialize_all(ifp);
467 	tapifstop(tp, 1);
468 	ifnet_deserialize_all(ifp);
469 
470 	ether_ifdetach(ifp);
471 	SLIST_REMOVE(&tap_listhead, tp, tap_softc, tap_link);
472 
473 	dev = tp->tap_dev;
474 	tp->tap_dev = NULL;
475 	dev->si_drv1 = NULL;
476 
477 	release_dev(dev);	/* tp association */
478 
479 	/*
480 	 * Also destroy the cloned device
481 	 */
482 	if (tp->tap_unit >= TAP_PREALLOCATED_UNITS) {
483 		destroy_dev(dev);
484 		devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(tap), tp->tap_unit);
485 	}
486 
487 	kfree(tp, M_TAP);
488 
489 	taplastunit--;
490 }
491 
492 /*
493  * tap_clone_destroy:
494  *
495  *	Destroy a tap instance.
496  */
497 static void
498 tap_clone_destroy(struct ifnet *ifp)
499 {
500 	struct tap_softc *tp = ifp->if_softc;
501 
502 	TAPDEBUG(&tp->tap_if, "clone destroyed. minor = %#x tap_flags = 0x%x\n",
503 		 minor(tp->tap_dev), tp->tap_flags);
504 	if (tp->tap_flags & TAP_CLONE)
505 		tapdestroy(tp);
506 }
507 
508 /*
509  * tapifinit
510  *
511  * Network interface initialization function (called with if serializer held)
512  *
513  * MPSAFE
514  */
515 static void
516 tapifinit(void *xtp)
517 {
518 	struct tap_softc *tp = xtp;
519 	struct ifnet *ifp = &tp->tap_if;
520 
521 	TAPDEBUG(ifp, "initializing, minor = %#x tap_flags = 0x%x\n",
522 		 minor(tp->tap_dev), tp->tap_flags);
523 
524 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
525 
526 	tapifstop(tp, 1);
527 
528 	ifp->if_flags |= IFF_RUNNING;
529 	ifp->if_flags &= ~IFF_OACTIVE;
530 
531 	/* attempt to start output */
532 	tapifstart(ifp);
533 }
534 
535 
536 /*
537  * tapifioctl
538  *
539  * Process an ioctl request on network interface (called with if serializer
540  * held).
541  *
542  * MPSAFE
543  */
544 static int
545 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
546 {
547 	struct tap_softc 	*tp = (struct tap_softc *)(ifp->if_softc);
548 	struct ifstat		*ifs = NULL;
549 	int			 dummy;
550 
551 	switch (cmd) {
552 		case SIOCSIFADDR:
553 		case SIOCGIFADDR:
554 		case SIOCSIFMTU:
555 			dummy = ether_ioctl(ifp, cmd, data);
556 			return (dummy);
557 
558 		case SIOCSIFFLAGS:
559 			tapifflags(tp);
560 			break;
561 
562 		case SIOCADDMULTI: /* XXX -- just like vmnet does */
563 		case SIOCDELMULTI:
564 			break;
565 
566 		case SIOCGIFSTATUS:
567 			ifs = (struct ifstat *)data;
568 			dummy = strlen(ifs->ascii);
569 			if ((tp->tap_flags & TAP_OPEN) &&
570 			    dummy < sizeof(ifs->ascii)) {
571 				if (tp->tap_sigtd && tp->tap_sigtd->sio_proc) {
572 				    ksnprintf(ifs->ascii + dummy,
573 					sizeof(ifs->ascii) - dummy,
574 					"\tOpened by pid %d\n",
575 					(int)tp->tap_sigtd->sio_proc->p_pid);
576 				} else {
577 				    ksnprintf(ifs->ascii + dummy,
578 					sizeof(ifs->ascii) - dummy,
579 					"\tOpened by <unknown>\n");
580 				}
581 			}
582 			break;
583 
584 		default:
585 			return (EINVAL);
586 	}
587 
588 	return (0);
589 }
590 
591 
592 /*
593  * tapifstart
594  *
595  * Queue packets from higher level ready to put out (called with if serializer
596  * held)
597  *
598  * MPSAFE
599  */
600 static void
601 tapifstart(struct ifnet *ifp)
602 {
603 	struct tap_softc *tp = ifp->if_softc;
604 	struct ifqueue *ifq;
605 	struct mbuf *m;
606 	int has_data = 0;
607 
608 	TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
609 
610 	/*
611 	 * do not junk pending output if we are in VMnet mode.
612 	 * XXX: can this do any harm because of queue overflow?
613 	 */
614 
615 	if (((tp->tap_flags & TAP_VMNET) == 0) &&
616 	    ((tp->tap_flags & TAP_READY) != TAP_READY)) {
617 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
618 			 minor(tp->tap_dev), tp->tap_flags);
619 		ifq_purge(&ifp->if_snd);
620 		return;
621 	}
622 
623 	ifp->if_flags |= IFF_OACTIVE;
624 
625 	ifq = &tp->tap_devq;
626 	while ((m = ifq_dequeue(&ifp->if_snd, NULL)) != NULL) {
627 		if (IF_QFULL(ifq)) {
628 			IF_DROP(ifq);
629 			ifp->if_oerrors++;
630 			m_freem(m);
631 		} else {
632 			IF_ENQUEUE(ifq, m);
633 			ifp->if_opackets++;
634 			has_data = 1;
635 		}
636 	}
637 
638 	if (has_data) {
639 		if (tp->tap_flags & TAP_RWAIT) {
640 			tp->tap_flags &= ~TAP_RWAIT;
641 			wakeup((caddr_t)tp);
642 		}
643 
644 		KNOTE(&tp->tap_rkq.ki_note, 0);
645 
646 		if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
647 			get_mplock();
648 			pgsigio(tp->tap_sigio, SIGIO, 0);
649 			rel_mplock();
650 		}
651 	}
652 
653 	ifp->if_flags &= ~IFF_OACTIVE;
654 }
655 
656 
657 /*
658  * tapioctl
659  *
660  * The ops interface is now pretty minimal.  Called via fileops with nothing
661  * held.
662  *
663  * MPSAFE
664  */
665 static int
666 tapioctl(struct dev_ioctl_args *ap)
667 {
668 	cdev_t dev = ap->a_head.a_dev;
669 	caddr_t data = ap->a_data;
670 	struct tap_softc	*tp = dev->si_drv1;
671 	struct ifnet		*ifp = &tp->tap_if;
672  	struct tapinfo		*tapp = NULL;
673 	struct mbuf *mb;
674 	short f;
675 	int error;
676 
677 	ifnet_serialize_all(ifp);
678 	error = 0;
679 
680 	switch (ap->a_cmd) {
681 	case TAPSIFINFO:
682 		tapp = (struct tapinfo *)data;
683 		ifp->if_mtu = tapp->mtu;
684 		ifp->if_type = tapp->type;
685 		ifp->if_baudrate = tapp->baudrate;
686 		break;
687 
688 	case TAPGIFINFO:
689 		tapp = (struct tapinfo *)data;
690 		tapp->mtu = ifp->if_mtu;
691 		tapp->type = ifp->if_type;
692 		tapp->baudrate = ifp->if_baudrate;
693 		break;
694 
695 	case TAPSDEBUG:
696 		tapdebug = *(int *)data;
697 		break;
698 
699 	case TAPGDEBUG:
700 		*(int *)data = tapdebug;
701 		break;
702 
703 	case FIOASYNC:
704 		if (*(int *)data)
705 			tp->tap_flags |= TAP_ASYNC;
706 		else
707 			tp->tap_flags &= ~TAP_ASYNC;
708 		break;
709 
710 	case FIONREAD:
711 		*(int *)data = 0;
712 
713 		/* Take a look at devq first */
714 		IF_POLL(&tp->tap_devq, mb);
715 		if (mb == NULL)
716 			mb = ifq_poll(&ifp->if_snd);
717 
718 		if (mb != NULL) {
719 			for(; mb != NULL; mb = mb->m_next)
720 				*(int *)data += mb->m_len;
721 		}
722 		break;
723 
724 	case FIOSETOWN:
725 		error = fsetown(*(int *)data, &tp->tap_sigio);
726 		break;
727 
728 	case FIOGETOWN:
729 		*(int *)data = fgetown(tp->tap_sigio);
730 		break;
731 
732 	/* this is deprecated, FIOSETOWN should be used instead */
733 	case TIOCSPGRP:
734 		error = fsetown(-(*(int *)data), &tp->tap_sigio);
735 		break;
736 
737 	/* this is deprecated, FIOGETOWN should be used instead */
738 	case TIOCGPGRP:
739 		*(int *)data = -fgetown(tp->tap_sigio);
740 		break;
741 
742 	/* VMware/VMnet port ioctl's */
743 
744 	case SIOCGIFFLAGS:	/* get ifnet flags */
745 		bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
746 		break;
747 
748 	case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
749 		f = *(short *)data;
750 		f &= 0x0fff;
751 		f &= ~IFF_CANTCHANGE;
752 		f |= IFF_UP;
753 		ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
754 		break;
755 
756 	case OSIOCGIFADDR:	/* get MAC address of the remote side */
757 	case SIOCGIFADDR:
758 		bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
759 		break;
760 
761 	case SIOCSIFADDR:	/* set MAC address of the remote side */
762 		bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
763 		break;
764 
765 	default:
766 		error = ENOTTY;
767 		break;
768 	}
769 	ifnet_deserialize_all(ifp);
770 	return (error);
771 }
772 
773 
774 /*
775  * tapread
776  *
777  * The ops read interface - reads a packet at a time, or at
778  * least as much of a packet as can be read.
779  *
780  * Called from the fileops interface with nothing held.
781  *
782  * MPSAFE
783  */
784 static int
785 tapread(struct dev_read_args *ap)
786 {
787 	cdev_t dev = ap->a_head.a_dev;
788 	struct uio *uio = ap->a_uio;
789 	struct tap_softc	*tp = dev->si_drv1;
790 	struct ifnet		*ifp = &tp->tap_if;
791 	struct mbuf		*m0 = NULL;
792 	int			 error = 0, len;
793 
794 	TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
795 
796 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
797 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
798 			 minor(tp->tap_dev), tp->tap_flags);
799 
800 		return (EHOSTDOWN);
801 	}
802 
803 	tp->tap_flags &= ~TAP_RWAIT;
804 
805 	/* sleep until we get a packet */
806 	do {
807 		ifnet_serialize_all(ifp);
808 		IF_DEQUEUE(&tp->tap_devq, m0);
809 		if (m0 == NULL) {
810 			if (ap->a_ioflag & IO_NDELAY) {
811 				ifnet_deserialize_all(ifp);
812 				return (EWOULDBLOCK);
813 			}
814 			tp->tap_flags |= TAP_RWAIT;
815 			tsleep_interlock(tp, PCATCH);
816 			ifnet_deserialize_all(ifp);
817 			error = tsleep(tp, PCATCH | PINTERLOCKED, "taprd", 0);
818 			if (error)
819 				return (error);
820 		} else {
821 			ifnet_deserialize_all(ifp);
822 		}
823 	} while (m0 == NULL);
824 
825 	BPF_MTAP(ifp, m0);
826 
827 	/* xfer packet to user space */
828 	while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
829 		len = (int)szmin(uio->uio_resid, m0->m_len);
830 		if (len == 0)
831 			break;
832 
833 		error = uiomove(mtod(m0, caddr_t), (size_t)len, uio);
834 		m0 = m_free(m0);
835 	}
836 
837 	if (m0 != NULL) {
838 		TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
839 			 minor(tp->tap_dev));
840 		m_freem(m0);
841 	}
842 
843 	return (error);
844 }
845 
846 /*
847  * tapwrite
848  *
849  * The ops write interface - an atomic write is a packet - or else!
850  *
851  * Called from the fileops interface with nothing held.
852  *
853  * MPSAFE
854  */
855 static int
856 tapwrite(struct dev_write_args *ap)
857 {
858 	cdev_t dev = ap->a_head.a_dev;
859 	struct uio *uio = ap->a_uio;
860 	struct tap_softc	*tp = dev->si_drv1;
861 	struct ifnet		*ifp = &tp->tap_if;
862 	struct mbuf		*top = NULL, **mp = NULL, *m = NULL;
863 	int			error = 0;
864 	size_t			tlen, mlen;
865 
866 	TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
867 
868 	if ((tp->tap_flags & TAP_READY) != TAP_READY) {
869 		TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
870 			 minor(tp->tap_dev), tp->tap_flags);
871 		return (EHOSTDOWN);
872 	}
873 
874 	if (uio->uio_resid == 0)
875 		return (0);
876 
877 	if (uio->uio_resid > TAPMRU) {
878 		TAPDEBUG(ifp, "invalid packet len = %zu, minor = %#x\n",
879 			 uio->uio_resid, minor(tp->tap_dev));
880 
881 		return (EIO);
882 	}
883 	tlen = uio->uio_resid;
884 
885 	/* get a header mbuf */
886 	MGETHDR(m, MB_DONTWAIT, MT_DATA);
887 	if (m == NULL)
888 		return (ENOBUFS);
889 	mlen = MHLEN;
890 
891 	top = 0;
892 	mp = &top;
893 	while ((error == 0) && (uio->uio_resid > 0)) {
894 		m->m_len = (int)szmin(mlen, uio->uio_resid);
895 		error = uiomove(mtod(m, caddr_t), (size_t)m->m_len, uio);
896 		*mp = m;
897 		mp = &m->m_next;
898 		if (uio->uio_resid > 0) {
899 			MGET(m, MB_DONTWAIT, MT_DATA);
900 			if (m == NULL) {
901 				error = ENOBUFS;
902 				break;
903 			}
904 			mlen = MLEN;
905 		}
906 	}
907 	if (error) {
908 		ifp->if_ierrors ++;
909 		if (top)
910 			m_freem(top);
911 		return (error);
912 	}
913 
914 	top->m_pkthdr.len = (int)tlen;
915 	top->m_pkthdr.rcvif = ifp;
916 
917 	/*
918 	 * Ethernet bridge and bpf are handled in ether_input
919 	 *
920 	 * adjust mbuf and give packet to the ether_input
921 	 */
922 	ifnet_serialize_all(ifp);
923 	ifp->if_input(ifp, top);
924 	ifp->if_ipackets ++; /* ibytes are counted in ether_input */
925 	ifnet_deserialize_all(ifp);
926 
927 	return (0);
928 }
929 
930 /*
931  * tapkqfilter - called from the fileops interface with nothing held
932  *
933  * MPSAFE
934  */
935 static int filt_tapread(struct knote *kn, long hint);
936 static int filt_tapwrite(struct knote *kn, long hint);
937 static void filt_tapdetach(struct knote *kn);
938 static struct filterops tapread_filtops =
939 	{ FILTEROP_ISFD, NULL, filt_tapdetach, filt_tapread };
940 static struct filterops tapwrite_filtops =
941 	{ FILTEROP_ISFD, NULL, filt_tapdetach, filt_tapwrite };
942 
943 static int
944 tapkqfilter(struct dev_kqfilter_args *ap)
945 {
946 	cdev_t dev = ap->a_head.a_dev;
947 	struct knote *kn = ap->a_kn;
948 	struct tap_softc *tp;
949 	struct klist *list;
950 	struct ifnet *ifp;
951 
952 	tp = dev->si_drv1;
953 	list = &tp->tap_rkq.ki_note;
954 	ifp = &tp->tap_if;
955 	ap->a_result =0;
956 
957 	switch(kn->kn_filter) {
958 	case EVFILT_READ:
959 		kn->kn_fop = &tapread_filtops;
960 		kn->kn_hook = (void *)tp;
961 		break;
962 	case EVFILT_WRITE:
963 		kn->kn_fop = &tapwrite_filtops;
964 		kn->kn_hook = (void *)tp;
965 		break;
966 	default:
967 		ap->a_result = EOPNOTSUPP;
968 		rel_mplock();
969 		return(0);
970 	}
971 
972 	knote_insert(list, kn);
973 	return(0);
974 }
975 
976 static int
977 filt_tapread(struct knote *kn, long hint)
978 {
979 	struct tap_softc *tp = (void *)kn->kn_hook;
980 
981 	if (IF_QEMPTY(&tp->tap_devq) == 0)	/* XXX serializer */
982 		return(1);
983 	else
984 		return(0);
985 }
986 
987 static int
988 filt_tapwrite(struct knote *kn, long hint)
989 {
990 	/* Always ready for a write */
991 	return (1);
992 }
993 
994 static void
995 filt_tapdetach(struct knote *kn)
996 {
997 	struct tap_softc *tp = (void *)kn->kn_hook;
998 
999 	knote_remove(&tp->tap_rkq.ki_note, kn);
1000 }
1001 
1002 static void
1003 tapifstop(struct tap_softc *tp, int clear_flags)
1004 {
1005 	struct ifnet *ifp = &tp->tap_if;
1006 
1007 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
1008 	IF_DRAIN(&tp->tap_devq);
1009 	tp->tap_flags &= ~TAP_CLOSEDOWN;
1010 	if (clear_flags)
1011 		ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
1012 }
1013 
1014 static void
1015 tapifflags(struct tap_softc *tp)
1016 {
1017 	struct ifnet *ifp = &tp->arpcom.ac_if;
1018 
1019 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
1020 	if ((tp->tap_flags & TAP_VMNET) == 0) {
1021 		/*
1022 		 * Only for non-vmnet tap(4)
1023 		 */
1024 		if (ifp->if_flags & IFF_UP) {
1025 			if ((ifp->if_flags & IFF_RUNNING) == 0)
1026 				tapifinit(tp);
1027 		} else {
1028 			tapifstop(tp, 1);
1029 		}
1030 	} else {
1031 		/* XXX */
1032 	}
1033 }
1034