if_tun.c revision 1.83 1 /* $NetBSD: if_tun.c,v 1.83 2006/03/29 19:29:00 rpaulo Exp $ */
2
3 /*
4 * Copyright (c) 1988, Julian Onions <jpo (at) cs.nott.ac.uk>
5 * Nottingham University 1987.
6 *
7 * This source may be freely distributed, however I would be interested
8 * in any changes that are made.
9 *
10 * This driver takes packets off the IP i/f and hands them up to a
11 * user process to have its wicked way with. This driver has its
12 * roots in a similar driver written by Phil Cockcroft (formerly) at
13 * UCL. This driver is based much more on read/write/poll mode of
14 * operation though.
15 */
16
17 #include <sys/cdefs.h>
18 __KERNEL_RCSID(0, "$NetBSD: if_tun.c,v 1.83 2006/03/29 19:29:00 rpaulo Exp $");
19
20 #include "opt_inet.h"
21 #include "opt_ns.h"
22
23 #include <sys/param.h>
24 #include <sys/proc.h>
25 #include <sys/systm.h>
26 #include <sys/mbuf.h>
27 #include <sys/buf.h>
28 #include <sys/protosw.h>
29 #include <sys/socket.h>
30 #include <sys/ioctl.h>
31 #include <sys/errno.h>
32 #include <sys/syslog.h>
33 #include <sys/select.h>
34 #include <sys/poll.h>
35 #include <sys/file.h>
36 #include <sys/signalvar.h>
37 #include <sys/conf.h>
38
39 #include <machine/cpu.h>
40
41 #include <net/if.h>
42 #include <net/if_types.h>
43 #include <net/netisr.h>
44 #include <net/route.h>
45
46
47 #ifdef INET
48 #include <netinet/in.h>
49 #include <netinet/in_systm.h>
50 #include <netinet/in_var.h>
51 #include <netinet/ip.h>
52 #include <netinet/if_inarp.h>
53 #endif
54
55 #ifdef NS
56 #include <netns/ns.h>
57 #include <netns/ns_if.h>
58 #endif
59
60 #include "bpfilter.h"
61 #if NBPFILTER > 0
62 #include <sys/time.h>
63 #include <net/bpf.h>
64 #endif
65
66 #include <net/if_tun.h>
67
68 #define TUNDEBUG if (tundebug) printf
69 int tundebug = 0;
70
71 extern int ifqmaxlen;
72 void tunattach(int);
73
74 static LIST_HEAD(, tun_softc) tun_softc_list;
75 static LIST_HEAD(, tun_softc) tunz_softc_list;
76 static struct simplelock tun_softc_lock;
77
78 static int tun_ioctl(struct ifnet *, u_long, caddr_t);
79 static int tun_output(struct ifnet *, struct mbuf *, struct sockaddr *,
80 struct rtentry *rt);
81 static int tun_clone_create(struct if_clone *, int);
82 static int tun_clone_destroy(struct ifnet *);
83
84 static struct if_clone tun_cloner =
85 IF_CLONE_INITIALIZER("tun", tun_clone_create, tun_clone_destroy);
86
87 static void tunattach0(struct tun_softc *);
88 static void tuninit(struct tun_softc *);
89 #ifdef ALTQ
90 static void tunstart(struct ifnet *);
91 #endif
92 static struct tun_softc *tun_find_unit(dev_t);
93 static struct tun_softc *tun_find_zunit(int);
94
95 static dev_type_open(tunopen);
96 static dev_type_close(tunclose);
97 static dev_type_read(tunread);
98 static dev_type_write(tunwrite);
99 static dev_type_ioctl(tunioctl);
100 static dev_type_poll(tunpoll);
101 static dev_type_kqfilter(tunkqfilter);
102
103 const struct cdevsw tun_cdevsw = {
104 tunopen, tunclose, tunread, tunwrite, tunioctl,
105 nostop, notty, tunpoll, nommap, tunkqfilter,
106 };
107
108 void
109 tunattach(int unused)
110 {
111
112 simple_lock_init(&tun_softc_lock);
113 LIST_INIT(&tun_softc_list);
114 LIST_INIT(&tunz_softc_list);
115 if_clone_attach(&tun_cloner);
116 }
117
118 /*
119 * Find driver instance from dev_t.
120 * Call at splnet().
121 * Returns with tp locked (if found).
122 */
123 static struct tun_softc *
124 tun_find_unit(dev_t dev)
125 {
126 struct tun_softc *tp;
127 int unit = minor(dev);
128
129 simple_lock(&tun_softc_lock);
130 LIST_FOREACH(tp, &tun_softc_list, tun_list)
131 if (unit == tp->tun_unit)
132 break;
133 if (tp)
134 simple_lock(&tp->tun_lock);
135 simple_unlock(&tun_softc_lock);
136
137 return (tp);
138 }
139
140 /*
141 * Find zombie driver instance by unit number.
142 * Call at splnet().
143 * Remove tp from list and return it unlocked (if found).
144 */
145 static struct tun_softc *
146 tun_find_zunit(int unit)
147 {
148 struct tun_softc *tp;
149
150 simple_lock(&tun_softc_lock);
151 LIST_FOREACH(tp, &tunz_softc_list, tun_list)
152 if (unit == tp->tun_unit)
153 break;
154 if (tp)
155 LIST_REMOVE(tp, tun_list);
156 simple_unlock(&tun_softc_lock);
157 #ifdef DIAGNOSTIC
158 if (tp != NULL && (tp->tun_flags & (TUN_INITED|TUN_OPEN)) != TUN_OPEN)
159 printf("tun%d: inconsistent flags: %x\n", unit, tp->tun_flags);
160 #endif
161
162 return (tp);
163 }
164
165 static int
166 tun_clone_create(struct if_clone *ifc, int unit)
167 {
168 struct tun_softc *tp;
169
170 if ((tp = tun_find_zunit(unit)) == NULL) {
171 /* Allocate a new instance */
172 tp = malloc(sizeof(struct tun_softc), M_DEVBUF, M_WAITOK);
173 (void)memset(tp, 0, sizeof(struct tun_softc));
174
175 tp->tun_unit = unit;
176 simple_lock_init(&tp->tun_lock);
177 } else {
178 /* Revive tunnel instance; clear ifp part */
179 (void)memset(&tp->tun_if, 0, sizeof(struct ifnet));
180 }
181
182 (void)snprintf(tp->tun_if.if_xname, sizeof(tp->tun_if.if_xname),
183 "%s%d", ifc->ifc_name, unit);
184 tunattach0(tp);
185 tp->tun_flags |= TUN_INITED;
186
187 simple_lock(&tun_softc_lock);
188 LIST_INSERT_HEAD(&tun_softc_list, tp, tun_list);
189 simple_unlock(&tun_softc_lock);
190
191 return (0);
192 }
193
194 static void
195 tunattach0(struct tun_softc *tp)
196 {
197 struct ifnet *ifp;
198
199 ifp = &tp->tun_if;
200 ifp->if_softc = tp;
201 ifp->if_mtu = TUNMTU;
202 ifp->if_ioctl = tun_ioctl;
203 ifp->if_output = tun_output;
204 #ifdef ALTQ
205 ifp->if_start = tunstart;
206 #endif
207 ifp->if_flags = IFF_POINTOPOINT;
208 ifp->if_type = IFT_TUNNEL;
209 ifp->if_snd.ifq_maxlen = ifqmaxlen;
210 ifp->if_collisions = 0;
211 ifp->if_ierrors = 0;
212 ifp->if_oerrors = 0;
213 ifp->if_ipackets = 0;
214 ifp->if_opackets = 0;
215 ifp->if_ibytes = 0;
216 ifp->if_obytes = 0;
217 ifp->if_dlt = DLT_NULL;
218 IFQ_SET_READY(&ifp->if_snd);
219 if_attach(ifp);
220 if_alloc_sadl(ifp);
221 #if NBPFILTER > 0
222 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
223 #endif
224 }
225
226 static int
227 tun_clone_destroy(struct ifnet *ifp)
228 {
229 struct tun_softc *tp = (void *)ifp;
230 int s, zombie = 0;
231
232 s = splnet();
233 simple_lock(&tun_softc_lock);
234 simple_lock(&tp->tun_lock);
235 LIST_REMOVE(tp, tun_list);
236 if (tp->tun_flags & TUN_OPEN) {
237 /* Hang on to storage until last close */
238 zombie = 1;
239 tp->tun_flags &= ~TUN_INITED;
240 LIST_INSERT_HEAD(&tunz_softc_list, tp, tun_list);
241 }
242 simple_unlock(&tun_softc_lock);
243
244 IF_PURGE(&ifp->if_snd);
245 ifp->if_flags &= ~IFF_RUNNING;
246
247 if (tp->tun_flags & TUN_RWAIT) {
248 tp->tun_flags &= ~TUN_RWAIT;
249 wakeup((caddr_t)tp);
250 }
251 if (tp->tun_flags & TUN_ASYNC && tp->tun_pgid)
252 fownsignal(tp->tun_pgid, SIGIO, POLL_HUP, 0, NULL);
253
254 selwakeup(&tp->tun_rsel);
255
256 simple_unlock(&tp->tun_lock);
257 splx(s);
258
259 #if NBPFILTER > 0
260 bpfdetach(ifp);
261 #endif
262 if_detach(ifp);
263
264 if (!zombie)
265 free(tp, M_DEVBUF);
266
267 return (0);
268 }
269
270 /*
271 * tunnel open - must be superuser & the device must be
272 * configured in
273 */
274 static int
275 tunopen(dev_t dev, int flag, int mode, struct lwp *l)
276 {
277 struct proc *p = l->l_proc;
278 struct ifnet *ifp;
279 struct tun_softc *tp;
280 int s, error;
281
282 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
283 return (error);
284
285 s = splnet();
286 tp = tun_find_unit(dev);
287
288 if (tp == NULL) {
289 (void)tun_clone_create(&tun_cloner, minor(dev));
290 tp = tun_find_unit(dev);
291 if (tp == NULL) {
292 error = ENXIO;
293 goto out_nolock;
294 }
295 }
296
297 if (tp->tun_flags & TUN_OPEN) {
298 error = EBUSY;
299 goto out;
300 }
301
302 ifp = &tp->tun_if;
303 tp->tun_flags |= TUN_OPEN;
304 TUNDEBUG("%s: open\n", ifp->if_xname);
305 out:
306 simple_unlock(&tp->tun_lock);
307 out_nolock:
308 splx(s);
309 return (error);
310 }
311
312 /*
313 * tunclose - close the device - mark i/f down & delete
314 * routing info
315 */
316 int
317 tunclose(dev_t dev, int flag, int mode, struct lwp *l)
318 {
319 int s;
320 struct tun_softc *tp;
321 struct ifnet *ifp;
322
323 s = splnet();
324 if ((tp = tun_find_zunit(minor(dev))) != NULL) {
325 /* interface was "destroyed" before the close */
326 free(tp, M_DEVBUF);
327 goto out_nolock;
328 }
329
330 if ((tp = tun_find_unit(dev)) == NULL)
331 goto out_nolock;
332
333 ifp = &tp->tun_if;
334
335 tp->tun_flags &= ~TUN_OPEN;
336
337 /*
338 * junk all pending output
339 */
340 IFQ_PURGE(&ifp->if_snd);
341
342 if (ifp->if_flags & IFF_UP) {
343 if_down(ifp);
344 if (ifp->if_flags & IFF_RUNNING) {
345 /* find internet addresses and delete routes */
346 struct ifaddr *ifa;
347 IFADDR_FOREACH(ifa, ifp) {
348 #if defined(INET) || defined(INET6)
349 if (ifa->ifa_addr->sa_family == AF_INET ||
350 ifa->ifa_addr->sa_family == AF_INET6) {
351 rtinit(ifa, (int)RTM_DELETE,
352 tp->tun_flags & TUN_DSTADDR
353 ? RTF_HOST
354 : 0);
355 }
356 #endif
357 }
358 }
359 }
360 tp->tun_pgid = 0;
361 selnotify(&tp->tun_rsel, 0);
362
363 TUNDEBUG ("%s: closed\n", ifp->if_xname);
364 simple_unlock(&tp->tun_lock);
365 out_nolock:
366 splx(s);
367 return (0);
368 }
369
370 /*
371 * Call at splnet() with tp locked.
372 */
373 static void
374 tuninit(struct tun_softc *tp)
375 {
376 struct ifnet *ifp = &tp->tun_if;
377 struct ifaddr *ifa;
378
379 TUNDEBUG("%s: tuninit\n", ifp->if_xname);
380
381 ifp->if_flags |= IFF_UP | IFF_RUNNING;
382
383 tp->tun_flags &= ~(TUN_IASET|TUN_DSTADDR);
384 IFADDR_FOREACH(ifa, ifp) {
385 #ifdef INET
386 if (ifa->ifa_addr->sa_family == AF_INET) {
387 struct sockaddr_in *sin;
388
389 sin = satosin(ifa->ifa_addr);
390 if (sin && sin->sin_addr.s_addr)
391 tp->tun_flags |= TUN_IASET;
392
393 if (ifp->if_flags & IFF_POINTOPOINT) {
394 sin = satosin(ifa->ifa_dstaddr);
395 if (sin && sin->sin_addr.s_addr)
396 tp->tun_flags |= TUN_DSTADDR;
397 }
398 }
399 #endif
400 #ifdef INET6
401 if (ifa->ifa_addr->sa_family == AF_INET6) {
402 struct sockaddr_in6 *sin;
403
404 sin = (struct sockaddr_in6 *)ifa->ifa_addr;
405 if (!IN6_IS_ADDR_UNSPECIFIED(&sin->sin6_addr))
406 tp->tun_flags |= TUN_IASET;
407
408 if (ifp->if_flags & IFF_POINTOPOINT) {
409 sin = (struct sockaddr_in6 *)ifa->ifa_dstaddr;
410 if (sin &&
411 !IN6_IS_ADDR_UNSPECIFIED(&sin->sin6_addr))
412 tp->tun_flags |= TUN_DSTADDR;
413 } else
414 tp->tun_flags &= ~TUN_DSTADDR;
415 }
416 #endif /* INET6 */
417 }
418
419 return;
420 }
421
422 /*
423 * Process an ioctl request.
424 */
425 static int
426 tun_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
427 {
428 int error = 0, s;
429 struct tun_softc *tp = (struct tun_softc *)(ifp->if_softc);
430
431 s = splnet();
432 simple_lock(&tp->tun_lock);
433
434 switch (cmd) {
435 case SIOCSIFADDR:
436 tuninit(tp);
437 TUNDEBUG("%s: address set\n", ifp->if_xname);
438 break;
439 case SIOCSIFDSTADDR:
440 tuninit(tp);
441 TUNDEBUG("%s: destination address set\n", ifp->if_xname);
442 break;
443 case SIOCSIFBRDADDR:
444 TUNDEBUG("%s: broadcast address set\n", ifp->if_xname);
445 break;
446 case SIOCSIFMTU: {
447 struct ifreq *ifr = (struct ifreq *) data;
448 if (ifr->ifr_mtu > TUNMTU || ifr->ifr_mtu < 576) {
449 error = EINVAL;
450 break;
451 }
452 TUNDEBUG("%s: interface mtu set\n", ifp->if_xname);
453 ifp->if_mtu = ifr->ifr_mtu;
454 break;
455 }
456 case SIOCADDMULTI:
457 case SIOCDELMULTI: {
458 struct ifreq *ifr = (struct ifreq *) data;
459 if (ifr == 0) {
460 error = EAFNOSUPPORT; /* XXX */
461 break;
462 }
463 switch (ifr->ifr_addr.sa_family) {
464 #ifdef INET
465 case AF_INET:
466 break;
467 #endif
468 #ifdef INET6
469 case AF_INET6:
470 break;
471 #endif
472 default:
473 error = EAFNOSUPPORT;
474 break;
475 }
476 break;
477 }
478 case SIOCSIFFLAGS:
479 break;
480 default:
481 error = EINVAL;
482 }
483
484 simple_unlock(&tp->tun_lock);
485 splx(s);
486 return (error);
487 }
488
489 /*
490 * tun_output - queue packets from higher level ready to put out.
491 */
492 static int
493 tun_output(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst,
494 struct rtentry *rt)
495 {
496 struct tun_softc *tp = ifp->if_softc;
497 int s;
498 int error;
499 #if defined(INET) || defined(INET6)
500 int mlen;
501 uint32_t *af;
502 #endif
503 ALTQ_DECL(struct altq_pktattr pktattr;)
504
505 s = splnet();
506 simple_lock(&tp->tun_lock);
507 TUNDEBUG ("%s: tun_output\n", ifp->if_xname);
508
509 if ((tp->tun_flags & TUN_READY) != TUN_READY) {
510 TUNDEBUG ("%s: not ready 0%o\n", ifp->if_xname,
511 tp->tun_flags);
512 m_freem (m0);
513 error = EHOSTDOWN;
514 goto out;
515 }
516
517 /*
518 * if the queueing discipline needs packet classification,
519 * do it before prepending link headers.
520 */
521 IFQ_CLASSIFY(&ifp->if_snd, m0, dst->sa_family, &pktattr);
522
523 #if NBPFILTER > 0
524 if (ifp->if_bpf)
525 bpf_mtap_af(ifp->if_bpf, dst->sa_family, m0);
526 #endif
527
528 switch(dst->sa_family) {
529 #ifdef INET6
530 case AF_INET6:
531 #endif
532 #ifdef INET
533 case AF_INET:
534 #endif
535 #if defined(INET) || defined(INET6)
536 if (tp->tun_flags & TUN_PREPADDR) {
537 /* Simple link-layer header */
538 M_PREPEND(m0, dst->sa_len, M_DONTWAIT);
539 if (m0 == NULL) {
540 IF_DROP(&ifp->if_snd);
541 error = ENOBUFS;
542 goto out;
543 }
544 bcopy(dst, mtod(m0, char *), dst->sa_len);
545 } else {
546 /* Prepend the address family */
547 M_PREPEND(m0, sizeof(*af), M_DONTWAIT);
548 if (m0 == NULL) {
549 IF_DROP(&ifp->if_snd);
550 error = ENOBUFS;
551 goto out;
552 }
553 af = mtod(m0,uint32_t *);
554 *af = htonl(dst->sa_family);
555 }
556 /* FALLTHROUGH */
557 case AF_UNSPEC:
558 IFQ_ENQUEUE(&ifp->if_snd, m0, &pktattr, error);
559 if (error) {
560 ifp->if_collisions++;
561 error = EAFNOSUPPORT;
562 goto out;
563 }
564 mlen = m0->m_pkthdr.len;
565 ifp->if_opackets++;
566 ifp->if_obytes += mlen;
567 break;
568 #endif
569 default:
570 m_freem(m0);
571 error = EAFNOSUPPORT;
572 goto out;
573 }
574
575 if (tp->tun_flags & TUN_RWAIT) {
576 tp->tun_flags &= ~TUN_RWAIT;
577 wakeup((caddr_t)tp);
578 }
579 if (tp->tun_flags & TUN_ASYNC && tp->tun_pgid)
580 fownsignal(tp->tun_pgid, SIGIO, POLL_IN, POLLIN|POLLRDNORM,
581 NULL);
582
583 selnotify(&tp->tun_rsel, 0);
584 out:
585 simple_unlock(&tp->tun_lock);
586 splx(s);
587 return (0);
588 }
589
590 /*
591 * the cdevsw interface is now pretty minimal.
592 */
593 int
594 tunioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct lwp *l)
595 {
596 struct tun_softc *tp;
597 int s, error = 0;
598
599 s = splnet();
600 tp = tun_find_unit(dev);
601
602 /* interface was "destroyed" already */
603 if (tp == NULL) {
604 error = ENXIO;
605 goto out_nolock;
606 }
607
608 switch (cmd) {
609 case TUNSDEBUG:
610 tundebug = *(int *)data;
611 break;
612
613 case TUNGDEBUG:
614 *(int *)data = tundebug;
615 break;
616
617 case TUNSIFMODE:
618 switch (*(int *)data & (IFF_POINTOPOINT|IFF_BROADCAST)) {
619 case IFF_POINTOPOINT:
620 case IFF_BROADCAST:
621 if (tp->tun_if.if_flags & IFF_UP) {
622 error = EBUSY;
623 goto out;
624 }
625 tp->tun_if.if_flags &=
626 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST);
627 tp->tun_if.if_flags |= *(int *)data;
628 break;
629 default:
630 error = EINVAL;
631 goto out;
632 }
633 break;
634
635 case TUNSLMODE:
636 if (*(int *)data)
637 tp->tun_flags |= TUN_PREPADDR;
638 else
639 tp->tun_flags &= ~TUN_PREPADDR;
640 break;
641
642 case FIONBIO:
643 if (*(int *)data)
644 tp->tun_flags |= TUN_NBIO;
645 else
646 tp->tun_flags &= ~TUN_NBIO;
647 break;
648
649 case FIOASYNC:
650 if (*(int *)data)
651 tp->tun_flags |= TUN_ASYNC;
652 else
653 tp->tun_flags &= ~TUN_ASYNC;
654 break;
655
656 case FIONREAD:
657 if (tp->tun_if.if_snd.ifq_head)
658 *(int *)data = tp->tun_if.if_snd.ifq_head->m_pkthdr.len;
659 else
660 *(int *)data = 0;
661 break;
662
663 case TIOCSPGRP:
664 case FIOSETOWN:
665 error = fsetown(l->l_proc, &tp->tun_pgid, cmd, data);
666 break;
667
668 case TIOCGPGRP:
669 case FIOGETOWN:
670 error = fgetown(l->l_proc, tp->tun_pgid, cmd, data);
671 break;
672
673 default:
674 error = ENOTTY;
675 }
676
677 out:
678 simple_unlock(&tp->tun_lock);
679 out_nolock:
680 splx(s);
681 return (error);
682 }
683
684 /*
685 * The cdevsw read interface - reads a packet at a time, or at
686 * least as much of a packet as can be read.
687 */
688 int
689 tunread(dev_t dev, struct uio *uio, int ioflag)
690 {
691 struct tun_softc *tp;
692 struct ifnet *ifp;
693 struct mbuf *m, *m0;
694 int error = 0, len, s, index;
695
696 s = splnet();
697 tp = tun_find_unit(dev);
698
699 /* interface was "destroyed" already */
700 if (tp == NULL) {
701 error = ENXIO;
702 goto out_nolock;
703 }
704
705 index = tp->tun_if.if_index;
706 ifp = &tp->tun_if;
707
708 TUNDEBUG ("%s: read\n", ifp->if_xname);
709 if ((tp->tun_flags & TUN_READY) != TUN_READY) {
710 TUNDEBUG ("%s: not ready 0%o\n", ifp->if_xname, tp->tun_flags);
711 error = EHOSTDOWN;
712 goto out;
713 }
714
715 tp->tun_flags &= ~TUN_RWAIT;
716
717 do {
718 IFQ_DEQUEUE(&ifp->if_snd, m0);
719 if (m0 == 0) {
720 if (tp->tun_flags & TUN_NBIO) {
721 error = EWOULDBLOCK;
722 goto out;
723 }
724 tp->tun_flags |= TUN_RWAIT;
725 if (ltsleep((caddr_t)tp, PZERO|PCATCH|PNORELOCK,
726 "tunread", 0, &tp->tun_lock) != 0) {
727 error = EINTR;
728 goto out_nolock;
729 } else {
730 /*
731 * Maybe the interface was destroyed while
732 * we were sleeping, so let's ensure that
733 * we're looking at the same (valid) tun
734 * interface before looping.
735 */
736 tp = tun_find_unit(dev);
737 if (tp == NULL) {
738 error = ENXIO;
739 goto out_nolock;
740 }
741 if (tp->tun_if.if_index != index) {
742 error = ENXIO;
743 goto out;
744 }
745 }
746 }
747 } while (m0 == 0);
748
749 simple_unlock(&tp->tun_lock);
750 splx(s);
751
752 /* Copy the mbuf chain */
753 while (m0 && uio->uio_resid > 0 && error == 0) {
754 len = min(uio->uio_resid, m0->m_len);
755 if (len != 0)
756 error = uiomove(mtod(m0, caddr_t), len, uio);
757 MFREE(m0, m);
758 m0 = m;
759 }
760
761 if (m0) {
762 TUNDEBUG("Dropping mbuf\n");
763 m_freem(m0);
764 }
765 if (error)
766 ifp->if_ierrors++;
767
768 return (error);
769
770 out:
771 simple_unlock(&tp->tun_lock);
772 out_nolock:
773 splx(s);
774 return (error);
775 }
776
777 /*
778 * the cdevsw write interface - an atomic write is a packet - or else!
779 */
780 int
781 tunwrite(dev_t dev, struct uio *uio, int ioflag)
782 {
783 struct tun_softc *tp;
784 struct ifnet *ifp;
785 struct mbuf *top, **mp, *m;
786 struct ifqueue *ifq;
787 struct sockaddr dst;
788 int isr, error = 0, s, tlen, mlen;
789 uint32_t family;
790
791 s = splnet();
792 tp = tun_find_unit(dev);
793
794 /* interface was "destroyed" already */
795 if (tp == NULL) {
796 error = ENXIO;
797 goto out_nolock;
798 }
799
800 /* Unlock until we've got the data */
801 simple_unlock(&tp->tun_lock);
802 splx(s);
803
804 ifp = &tp->tun_if;
805
806 TUNDEBUG("%s: tunwrite\n", ifp->if_xname);
807
808 if (tp->tun_flags & TUN_PREPADDR) {
809 if (uio->uio_resid < sizeof(dst)) {
810 error = EIO;
811 goto out0;
812 }
813 error = uiomove((caddr_t)&dst, sizeof(dst), uio);
814 if (dst.sa_len > sizeof(dst)) {
815 /* Duh.. */
816 char discard;
817 int n = dst.sa_len - sizeof(dst);
818 while (n--)
819 if ((error = uiomove(&discard, 1, uio)) != 0) {
820 goto out0;
821 }
822 }
823 } else {
824 if (uio->uio_resid < sizeof(family)){
825 error = EIO;
826 goto out0;
827 }
828 error = uiomove((caddr_t)&family, sizeof(family), uio);
829 dst.sa_family = ntohl(family);
830 }
831
832 if (uio->uio_resid > TUNMTU) {
833 TUNDEBUG("%s: len=%lu!\n", ifp->if_xname,
834 (unsigned long)uio->uio_resid);
835 error = EIO;
836 goto out0;
837 }
838
839 switch (dst.sa_family) {
840 #ifdef INET
841 case AF_INET:
842 ifq = &ipintrq;
843 isr = NETISR_IP;
844 break;
845 #endif
846 #ifdef INET6
847 case AF_INET6:
848 ifq = &ip6intrq;
849 isr = NETISR_IPV6;
850 break;
851 #endif
852 default:
853 error = EAFNOSUPPORT;
854 goto out0;
855 }
856
857 tlen = uio->uio_resid;
858
859 /* get a header mbuf */
860 MGETHDR(m, M_DONTWAIT, MT_DATA);
861 if (m == NULL) {
862 error = ENOBUFS;
863 goto out0;
864 }
865 mlen = MHLEN;
866
867 top = NULL;
868 mp = ⊤
869 while (error == 0 && uio->uio_resid > 0) {
870 m->m_len = min(mlen, uio->uio_resid);
871 error = uiomove(mtod(m, caddr_t), m->m_len, uio);
872 *mp = m;
873 mp = &m->m_next;
874 if (error == 0 && uio->uio_resid > 0) {
875 MGET(m, M_DONTWAIT, MT_DATA);
876 if (m == NULL) {
877 error = ENOBUFS;
878 break;
879 }
880 mlen = MLEN;
881 }
882 }
883 if (error) {
884 if (top != NULL)
885 m_freem (top);
886 ifp->if_ierrors++;
887 goto out0;
888 }
889
890 top->m_pkthdr.len = tlen;
891 top->m_pkthdr.rcvif = ifp;
892
893 #if NBPFILTER > 0
894 if (ifp->if_bpf)
895 bpf_mtap_af(ifp->if_bpf, dst.sa_family, top);
896 #endif
897
898 s = splnet();
899 simple_lock(&tp->tun_lock);
900 if ((tp->tun_flags & TUN_INITED) == 0) {
901 /* Interface was destroyed */
902 error = ENXIO;
903 goto out;
904 }
905 if (IF_QFULL(ifq)) {
906 IF_DROP(ifq);
907 ifp->if_collisions++;
908 m_freem(top);
909 error = ENOBUFS;
910 goto out;
911 }
912
913 IF_ENQUEUE(ifq, top);
914 ifp->if_ipackets++;
915 ifp->if_ibytes += tlen;
916 schednetisr(isr);
917 out:
918 simple_unlock(&tp->tun_lock);
919 out_nolock:
920 splx(s);
921 out0:
922 return (error);
923 }
924
925 #ifdef ALTQ
926 /*
927 * Start packet transmission on the interface.
928 * when the interface queue is rate-limited by ALTQ or TBR,
929 * if_start is needed to drain packets from the queue in order
930 * to notify readers when outgoing packets become ready.
931 *
932 * Should be called at splnet.
933 */
934 static void
935 tunstart(struct ifnet *ifp)
936 {
937 struct tun_softc *tp = ifp->if_softc;
938
939 if (!ALTQ_IS_ENABLED(&ifp->if_snd) && !TBR_IS_ENABLED(&ifp->if_snd))
940 return;
941
942 simple_lock(&tp->tun_lock);
943 if (!IF_IS_EMPTY(&ifp->if_snd)) {
944 if (tp->tun_flags & TUN_RWAIT) {
945 tp->tun_flags &= ~TUN_RWAIT;
946 wakeup((caddr_t)tp);
947 }
948 if (tp->tun_flags & TUN_ASYNC && tp->tun_pgid)
949 fownsignal(tp->tun_pgid, SIGIO, POLL_OUT,
950 POLLOUT|POLLWRNORM, NULL);
951
952 selwakeup(&tp->tun_rsel);
953 }
954 simple_unlock(&tp->tun_lock);
955 }
956 #endif /* ALTQ */
957 /*
958 * tunpoll - the poll interface, this is only useful on reads
959 * really. The write detect always returns true, write never blocks
960 * anyway, it either accepts the packet or drops it.
961 */
962 int
963 tunpoll(dev_t dev, int events, struct lwp *l)
964 {
965 struct tun_softc *tp;
966 struct ifnet *ifp;
967 int s, revents = 0;
968
969 s = splnet();
970 tp = tun_find_unit(dev);
971
972 /* interface was "destroyed" already */
973 if (tp == NULL)
974 goto out_nolock;
975
976 ifp = &tp->tun_if;
977
978 TUNDEBUG("%s: tunpoll\n", ifp->if_xname);
979
980 if (events & (POLLIN | POLLRDNORM)) {
981 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
982 TUNDEBUG("%s: tunpoll q=%d\n", ifp->if_xname,
983 ifp->if_snd.ifq_len);
984 revents |= events & (POLLIN | POLLRDNORM);
985 } else {
986 TUNDEBUG("%s: tunpoll waiting\n", ifp->if_xname);
987 selrecord(l, &tp->tun_rsel);
988 }
989 }
990
991 if (events & (POLLOUT | POLLWRNORM))
992 revents |= events & (POLLOUT | POLLWRNORM);
993
994 simple_unlock(&tp->tun_lock);
995 out_nolock:
996 splx(s);
997 return (revents);
998 }
999
1000 static void
1001 filt_tunrdetach(struct knote *kn)
1002 {
1003 struct tun_softc *tp = kn->kn_hook;
1004 int s;
1005
1006 s = splnet();
1007 SLIST_REMOVE(&tp->tun_rsel.sel_klist, kn, knote, kn_selnext);
1008 splx(s);
1009 }
1010
1011 static int
1012 filt_tunread(struct knote *kn, long hint)
1013 {
1014 struct tun_softc *tp = kn->kn_hook;
1015 struct ifnet *ifp = &tp->tun_if;
1016 struct mbuf *m;
1017 int s;
1018
1019 s = splnet();
1020 IF_POLL(&ifp->if_snd, m);
1021 if (m == NULL) {
1022 splx(s);
1023 return (0);
1024 }
1025
1026 for (kn->kn_data = 0; m != NULL; m = m->m_next)
1027 kn->kn_data += m->m_len;
1028
1029 splx(s);
1030 return (1);
1031 }
1032
1033 static const struct filterops tunread_filtops =
1034 { 1, NULL, filt_tunrdetach, filt_tunread };
1035
1036 static const struct filterops tun_seltrue_filtops =
1037 { 1, NULL, filt_tunrdetach, filt_seltrue };
1038
1039 int
1040 tunkqfilter(dev_t dev, struct knote *kn)
1041 {
1042 struct tun_softc *tp;
1043 struct klist *klist;
1044 int rv = 0, s;
1045
1046 s = splnet();
1047 tp = tun_find_unit(dev);
1048 if (tp == NULL)
1049 goto out_nolock;
1050
1051 switch (kn->kn_filter) {
1052 case EVFILT_READ:
1053 klist = &tp->tun_rsel.sel_klist;
1054 kn->kn_fop = &tunread_filtops;
1055 break;
1056
1057 case EVFILT_WRITE:
1058 klist = &tp->tun_rsel.sel_klist;
1059 kn->kn_fop = &tun_seltrue_filtops;
1060 break;
1061
1062 default:
1063 rv = 1;
1064 goto out;
1065 }
1066
1067 kn->kn_hook = tp;
1068
1069 SLIST_INSERT_HEAD(klist, kn, kn_selnext);
1070
1071 out:
1072 simple_unlock(&tp->tun_lock);
1073 out_nolock:
1074 splx(s);
1075 return (rv);
1076 }
1077