1 /* $NetBSD: if_tun.c,v 1.14 1994/06/29 06:36:25 cgd Exp $ */
4 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk>
5 * Nottingham University 1987.
7 * This source may be freely distributed, however I would be interested
8 * in any changes that are made.
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 it's
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
16 * $FreeBSD: src/sys/net/if_tun.c,v 1.74.2.8 2002/02/13 00:43:11 dillon Exp $
17 * $DragonFly: src/sys/net/tun/if_tun.c,v 1.37 2008/06/05 18:06:32 swildner Exp $
20 #include "opt_atalk.h"
22 #include "opt_inet6.h"
25 #include <sys/param.h>
28 #include <sys/systm.h>
30 #include <sys/socket.h>
32 #include <sys/device.h>
33 #include <sys/filio.h>
34 #include <sys/sockio.h>
35 #include <sys/thread2.h>
36 #include <sys/ttycom.h>
38 #include <sys/signalvar.h>
39 #include <sys/filedesc.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
43 #include <sys/vnode.h>
44 #include <sys/malloc.h>
47 #include <net/if_types.h>
48 #include <net/ifq_var.h>
49 #include <net/netisr.h>
50 #include <net/route.h>
53 #include <netinet/in.h>
58 #include "if_tunvar.h"
61 static MALLOC_DEFINE(M_TUN, "tun", "Tunnel Interface");
63 static void tunattach (void *);
64 PSEUDO_SET(tunattach, if_tun);
66 static void tuncreate (cdev_t dev);
68 #define TUNDEBUG if (tundebug) if_printf
69 static int tundebug = 0;
70 SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, "");
72 static int tunoutput (struct ifnet *, struct mbuf *, struct sockaddr *,
74 static int tunifioctl (struct ifnet *, u_long, caddr_t, struct ucred *);
75 static int tuninit (struct ifnet *);
76 static void tunstart(struct ifnet *);
78 static d_open_t tunopen;
79 static d_close_t tunclose;
80 static d_read_t tunread;
81 static d_write_t tunwrite;
82 static d_ioctl_t tunioctl;
83 static d_poll_t tunpoll;
86 static struct dev_ops tun_ops = {
87 { "tun", CDEV_MAJOR, 0 },
97 tunattach(void *dummy)
99 dev_ops_add(&tun_ops, 0, 0);
103 tuncreate(cdev_t dev)
105 struct tun_softc *sc;
108 dev = make_dev(&tun_ops, minor(dev),
109 UID_UUCP, GID_DIALER, 0600, "tun%d", lminor(dev));
111 MALLOC(sc, struct tun_softc *, sizeof(*sc), M_TUN, M_WAITOK | M_ZERO);
112 sc->tun_flags = TUN_INITED;
115 if_initname(ifp, "tun", lminor(dev));
116 ifp->if_mtu = TUNMTU;
117 ifp->if_ioctl = tunifioctl;
118 ifp->if_output = tunoutput;
119 ifp->if_start = tunstart;
120 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
121 ifp->if_type = IFT_PPP;
122 ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
123 ifq_set_ready(&ifp->if_snd);
125 if_attach(ifp, NULL);
126 bpfattach(ifp, DLT_NULL, sizeof(u_int));
131 * tunnel open - must be superuser & the device must be
135 tunopen(struct dev_open_args *ap)
137 cdev_t dev = ap->a_head.a_dev;
139 struct tun_softc *tp;
142 if ((error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)) != 0)
150 if (tp->tun_flags & TUN_OPEN)
152 tp->tun_pid = curproc->p_pid;
154 tp->tun_flags |= TUN_OPEN;
155 TUNDEBUG(ifp, "open\n");
160 * tunclose - close the device - mark i/f down & delete
164 tunclose(struct dev_close_args *ap)
166 cdev_t dev = ap->a_head.a_dev;
167 struct tun_softc *tp;
173 tp->tun_flags &= ~TUN_OPEN;
176 /* Junk all pending output. */
177 lwkt_serialize_enter(ifp->if_serializer);
178 ifq_purge(&ifp->if_snd);
179 lwkt_serialize_exit(ifp->if_serializer);
181 if (ifp->if_flags & IFF_UP) {
182 lwkt_serialize_enter(ifp->if_serializer);
184 lwkt_serialize_exit(ifp->if_serializer);
186 ifp->if_flags &= ~IFF_RUNNING;
187 if_purgeaddrs_nolink(ifp);
189 funsetown(tp->tun_sigio);
190 selwakeup(&tp->tun_rsel);
192 TUNDEBUG(ifp, "closed\n");
197 tuninit(struct ifnet *ifp)
199 struct tun_softc *tp = ifp->if_softc;
200 struct ifaddr_container *ifac;
203 TUNDEBUG(ifp, "tuninit\n");
205 ifp->if_flags |= IFF_UP | IFF_RUNNING;
206 getmicrotime(&ifp->if_lastchange);
208 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
209 struct ifaddr *ifa = ifac->ifa;
211 if (ifa->ifa_addr == NULL) {
213 /* XXX: Should maybe return straight off? */
216 if (ifa->ifa_addr->sa_family == AF_INET) {
217 struct sockaddr_in *si;
219 si = (struct sockaddr_in *)ifa->ifa_addr;
220 if (si->sin_addr.s_addr)
221 tp->tun_flags |= TUN_IASET;
230 * Process an ioctl request.
235 tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
237 struct ifreq *ifr = (struct ifreq *)data;
238 struct tun_softc *tp = ifp->if_softc;
244 ifs = (struct ifstat *)data;
246 ksprintf(ifs->ascii + strlen(ifs->ascii),
247 "\tOpened by PID %d\n", tp->tun_pid);
250 error = tuninit(ifp);
251 TUNDEBUG(ifp, "address set, error=%d\n", error);
254 error = tuninit(ifp);
255 TUNDEBUG(ifp, "destination address set, error=%d\n", error);
258 ifp->if_mtu = ifr->ifr_mtu;
259 TUNDEBUG(ifp, "mtu set\n");
272 * tunoutput - queue packets from higher level ready to put out.
277 tunoutput_serialized(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst,
280 struct tun_softc *tp = ifp->if_softc;
282 struct altq_pktattr pktattr;
284 TUNDEBUG(ifp, "tunoutput\n");
286 if ((tp->tun_flags & TUN_READY) != TUN_READY) {
287 TUNDEBUG(ifp, "not ready 0%o\n", tp->tun_flags);
293 * if the queueing discipline needs packet classification,
294 * do it before prepending link headers.
296 ifq_classify(&ifp->if_snd, m0, dst->sa_family, &pktattr);
298 /* BPF write needs to be handled specially */
299 if (dst->sa_family == AF_UNSPEC) {
300 dst->sa_family = *(mtod(m0, int *));
301 m0->m_len -= sizeof(int);
302 m0->m_pkthdr.len -= sizeof(int);
303 m0->m_data += sizeof(int);
308 * We need to prepend the address family as
311 uint32_t af = dst->sa_family;
313 bpf_ptap(ifp->if_bpf, m0, &af, sizeof(af));
316 /* prepend sockaddr? this may abort if the mbuf allocation fails */
317 if (tp->tun_flags & TUN_LMODE) {
318 /* allocate space for sockaddr */
319 M_PREPEND(m0, dst->sa_len, MB_DONTWAIT);
321 /* if allocation failed drop packet */
323 IF_DROP(&ifp->if_snd);
327 bcopy(dst, m0->m_data, dst->sa_len);
331 if (tp->tun_flags & TUN_IFHEAD) {
332 /* Prepend the address family */
333 M_PREPEND(m0, 4, MB_DONTWAIT);
335 /* if allocation failed drop packet */
337 IF_DROP(&ifp->if_snd);
341 *(u_int32_t *)m0->m_data = htonl(dst->sa_family);
344 if (dst->sa_family != AF_INET)
352 error = ifq_handoff(ifp, m0, &pktattr);
354 ifp->if_collisions++;
357 if (tp->tun_flags & TUN_RWAIT) {
358 tp->tun_flags &= ~TUN_RWAIT;
362 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio)
363 pgsigio(tp->tun_sigio, SIGIO, 0);
364 selwakeup(&tp->tun_rsel);
371 tunoutput(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst,
376 lwkt_serialize_enter(ifp->if_serializer);
377 error = tunoutput_serialized(ifp, m0, dst, rt);
378 lwkt_serialize_exit(ifp->if_serializer);
384 * the ops interface is now pretty minimal.
387 tunioctl(struct dev_ioctl_args *ap)
389 cdev_t dev = ap->a_head.a_dev;
390 struct tun_softc *tp = dev->si_drv1;
391 struct tuninfo *tunp;
395 tunp = (struct tuninfo *)ap->a_data;
396 if (tunp->mtu < IF_MINMTU)
398 tp->tun_if.if_mtu = tunp->mtu;
399 tp->tun_if.if_type = tunp->type;
400 tp->tun_if.if_baudrate = tunp->baudrate;
403 tunp = (struct tuninfo *)ap->a_data;
404 tunp->mtu = tp->tun_if.if_mtu;
405 tunp->type = tp->tun_if.if_type;
406 tunp->baudrate = tp->tun_if.if_baudrate;
409 tundebug = *(int *)ap->a_data;
412 *(int *)ap->a_data = tundebug;
415 if (*(int *)ap->a_data) {
416 tp->tun_flags |= TUN_LMODE;
417 tp->tun_flags &= ~TUN_IFHEAD;
419 tp->tun_flags &= ~TUN_LMODE;
422 if (*(int *)ap->a_data) {
423 tp->tun_flags |= TUN_IFHEAD;
424 tp->tun_flags &= ~TUN_LMODE;
426 tp->tun_flags &= ~TUN_IFHEAD;
429 *(int *)ap->a_data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0;
432 /* deny this if UP */
433 if (tp->tun_if.if_flags & IFF_UP)
436 switch (*(int *)ap->a_data & ~IFF_MULTICAST) {
437 case IFF_POINTOPOINT:
439 tp->tun_if.if_flags &= ~(IFF_BROADCAST|IFF_POINTOPOINT);
440 tp->tun_if.if_flags |= *(int *)ap->a_data;
447 tp->tun_pid = curproc->p_pid;
450 if (*(int *)ap->a_data)
451 tp->tun_flags |= TUN_ASYNC;
453 tp->tun_flags &= ~TUN_ASYNC;
456 lwkt_serialize_enter(tp->tun_if.if_serializer);
457 if (!ifq_is_empty(&tp->tun_if.if_snd)) {
460 mb = ifq_poll(&tp->tun_if.if_snd);
461 for( *(int *)ap->a_data = 0; mb != 0; mb = mb->m_next)
462 *(int *)ap->a_data += mb->m_len;
464 *(int *)ap->a_data = 0;
466 lwkt_serialize_exit(tp->tun_if.if_serializer);
469 return (fsetown(*(int *)ap->a_data, &tp->tun_sigio));
472 *(int *)ap->a_data = fgetown(tp->tun_sigio);
475 /* This is deprecated, FIOSETOWN should be used instead. */
477 return (fsetown(-(*(int *)ap->a_data), &tp->tun_sigio));
479 /* This is deprecated, FIOGETOWN should be used instead. */
481 *(int *)ap->a_data = -fgetown(tp->tun_sigio);
491 * The ops read interface - reads a packet at a time, or at
492 * least as much of a packet as can be read.
495 tunread(struct dev_read_args *ap)
497 cdev_t dev = ap->a_head.a_dev;
498 struct uio *uio = ap->a_uio;
499 struct tun_softc *tp = dev->si_drv1;
500 struct ifnet *ifp = &tp->tun_if;
504 TUNDEBUG(ifp, "read\n");
505 if ((tp->tun_flags & TUN_READY) != TUN_READY) {
506 TUNDEBUG(ifp, "not ready 0%o\n", tp->tun_flags);
510 tp->tun_flags &= ~TUN_RWAIT;
512 lwkt_serialize_enter(ifp->if_serializer);
514 while ((m0 = ifq_dequeue(&ifp->if_snd, NULL)) == NULL) {
515 if (ap->a_ioflag & IO_NDELAY) {
516 lwkt_serialize_exit(ifp->if_serializer);
519 tp->tun_flags |= TUN_RWAIT;
520 lwkt_serialize_exit(ifp->if_serializer);
521 if ((error = tsleep(tp, PCATCH, "tunread", 0)) != 0)
523 lwkt_serialize_enter(ifp->if_serializer);
526 lwkt_serialize_exit(ifp->if_serializer);
528 while (m0 && uio->uio_resid > 0 && error == 0) {
529 len = min(uio->uio_resid, m0->m_len);
531 error = uiomove(mtod(m0, caddr_t), len, uio);
536 TUNDEBUG(ifp, "Dropping mbuf\n");
543 * the ops write interface - an atomic write is a packet - or else!
546 tunwrite(struct dev_write_args *ap)
548 cdev_t dev = ap->a_head.a_dev;
549 struct uio *uio = ap->a_uio;
550 struct tun_softc *tp = dev->si_drv1;
551 struct ifnet *ifp = &tp->tun_if;
552 struct mbuf *top, **mp, *m;
553 int error=0, tlen, mlen;
557 TUNDEBUG(ifp, "tunwrite\n");
559 if (uio->uio_resid == 0)
562 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) {
563 TUNDEBUG(ifp, "len=%d!\n", uio->uio_resid);
566 tlen = uio->uio_resid;
568 /* get a header mbuf */
569 MGETHDR(m, MB_DONTWAIT, MT_DATA);
576 while (error == 0 && uio->uio_resid > 0) {
577 m->m_len = min(mlen, uio->uio_resid);
578 error = uiomove(mtod (m, caddr_t), m->m_len, uio);
581 if (uio->uio_resid > 0) {
582 MGET (m, MB_DONTWAIT, MT_DATA);
597 top->m_pkthdr.len = tlen;
598 top->m_pkthdr.rcvif = ifp;
601 if (tp->tun_flags & TUN_IFHEAD) {
603 * Conveniently, we already have a 4-byte address
604 * family prepended to our packet !
605 * Inconveniently, it's in the wrong byte order !
607 if ((top = m_pullup(top, sizeof(family))) == NULL)
609 *mtod(top, u_int32_t *) =
610 ntohl(*mtod(top, u_int32_t *));
611 bpf_mtap(ifp->if_bpf, top);
612 *mtod(top, u_int32_t *) =
613 htonl(*mtod(top, u_int32_t *));
616 * We need to prepend the address family as
619 static const uint32_t af = AF_INET;
621 bpf_ptap(ifp->if_bpf, top, &af, sizeof(af));
625 if (tp->tun_flags & TUN_IFHEAD) {
626 if (top->m_len < sizeof(family) &&
627 (top = m_pullup(top, sizeof(family))) == NULL)
629 family = ntohl(*mtod(top, u_int32_t *));
630 m_adj(top, sizeof(family));
634 ifp->if_ibytes += top->m_pkthdr.len;
660 return (EAFNOSUPPORT);
663 netisr_dispatch(isr, top);
668 * tunpoll - the poll interface, this is only useful on reads
669 * really. The write detect always returns true, write never blocks
670 * anyway, it either accepts the packet or drops it.
673 tunpoll(struct dev_poll_args *ap)
675 cdev_t dev = ap->a_head.a_dev;
676 struct tun_softc *tp = dev->si_drv1;
677 struct ifnet *ifp = &tp->tun_if;
680 TUNDEBUG(ifp, "tunpoll\n");
682 lwkt_serialize_enter(ifp->if_serializer);
684 if (ap->a_events & (POLLIN | POLLRDNORM)) {
685 if (!ifq_is_empty(&ifp->if_snd)) {
686 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len);
687 revents |= ap->a_events & (POLLIN | POLLRDNORM);
689 TUNDEBUG(ifp, "tunpoll waiting\n");
690 selrecord(curthread, &tp->tun_rsel);
693 if (ap->a_events & (POLLOUT | POLLWRNORM))
694 revents |= ap->a_events & (POLLOUT | POLLWRNORM);
696 lwkt_serialize_exit(ifp->if_serializer);
697 ap->a_events = revents;
702 * Start packet transmission on the interface.
703 * when the interface queue is rate-limited by ALTQ,
704 * if_start is needed to drain packets from the queue in order
705 * to notify readers when outgoing packets become ready.
708 tunstart(struct ifnet *ifp)
710 struct tun_softc *tp = ifp->if_softc;
713 if (!ifq_is_enabled(&ifp->if_snd))
716 m = ifq_poll(&ifp->if_snd);
718 if (tp->tun_flags & TUN_RWAIT) {
719 tp->tun_flags &= ~TUN_RWAIT;
722 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio)
723 pgsigio(tp->tun_sigio, SIGIO, 0);
724 selwakeup(&tp->tun_rsel);