FreeBSD/Linux Kernel Cross Reference
sys/net/if_tun.c
1 /* $NetBSD: if_tun.c,v 1.14 1994/06/29 06:36:25 cgd Exp $ */
2
3 /*-
4 * Copyright (c) 1988, Julian Onions <jpo@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 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
14 * operation though.
15 *
16 * $FreeBSD$
17 */
18
19 #include "opt_atalk.h"
20 #include "opt_inet.h"
21 #include "opt_inet6.h"
22 #include "opt_ipx.h"
23
24 #include <sys/param.h>
25 #include <sys/priv.h>
26 #include <sys/proc.h>
27 #include <sys/systm.h>
28 #include <sys/jail.h>
29 #include <sys/mbuf.h>
30 #include <sys/module.h>
31 #include <sys/socket.h>
32 #include <sys/fcntl.h>
33 #include <sys/filio.h>
34 #include <sys/sockio.h>
35 #include <sys/ttycom.h>
36 #include <sys/poll.h>
37 #include <sys/selinfo.h>
38 #include <sys/signalvar.h>
39 #include <sys/filedesc.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
42 #include <sys/conf.h>
43 #include <sys/uio.h>
44 #include <sys/malloc.h>
45 #include <sys/random.h>
46
47 #include <net/if.h>
48 #include <net/if_clone.h>
49 #include <net/if_types.h>
50 #include <net/netisr.h>
51 #include <net/route.h>
52 #include <net/vnet.h>
53 #ifdef INET
54 #include <netinet/in.h>
55 #endif
56 #include <net/bpf.h>
57 #include <net/if_tun.h>
58
59 #include <sys/queue.h>
60 #include <sys/condvar.h>
61
62 #include <security/mac/mac_framework.h>
63
64 /*
65 * tun_list is protected by global tunmtx. Other mutable fields are
66 * protected by tun->tun_mtx, or by their owning subsystem. tun_dev is
67 * static for the duration of a tunnel interface.
68 */
69 struct tun_softc {
70 TAILQ_ENTRY(tun_softc) tun_list;
71 struct cdev *tun_dev;
72 u_short tun_flags; /* misc flags */
73 #define TUN_OPEN 0x0001
74 #define TUN_INITED 0x0002
75 #define TUN_RCOLL 0x0004
76 #define TUN_IASET 0x0008
77 #define TUN_DSTADDR 0x0010
78 #define TUN_LMODE 0x0020
79 #define TUN_RWAIT 0x0040
80 #define TUN_ASYNC 0x0080
81 #define TUN_IFHEAD 0x0100
82
83 #define TUN_READY (TUN_OPEN | TUN_INITED)
84
85 /*
86 * XXXRW: tun_pid is used to exclusively lock /dev/tun. Is this
87 * actually needed? Can we just return EBUSY if already open?
88 * Problem is that this involved inherent races when a tun device
89 * is handed off from one process to another, as opposed to just
90 * being slightly stale informationally.
91 */
92 pid_t tun_pid; /* owning pid */
93 struct ifnet *tun_ifp; /* the interface */
94 struct sigio *tun_sigio; /* information for async I/O */
95 struct selinfo tun_rsel; /* read select */
96 struct mtx tun_mtx; /* protect mutable softc fields */
97 struct cv tun_cv; /* protect against ref'd dev destroy */
98 };
99 #define TUN2IFP(sc) ((sc)->tun_ifp)
100
101 #define TUNDEBUG if (tundebug) if_printf
102 #define TUNNAME "tun"
103
104 /*
105 * All mutable global variables in if_tun are locked using tunmtx, with
106 * the exception of tundebug, which is used unlocked, and tunclones,
107 * which is static after setup.
108 */
109 static struct mtx tunmtx;
110 static MALLOC_DEFINE(M_TUN, TUNNAME, "Tunnel Interface");
111 static int tundebug = 0;
112 static int tundclone = 1;
113 static struct clonedevs *tunclones;
114 static TAILQ_HEAD(,tun_softc) tunhead = TAILQ_HEAD_INITIALIZER(tunhead);
115 SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, "");
116
117 SYSCTL_DECL(_net_link);
118 SYSCTL_NODE(_net_link, OID_AUTO, tun, CTLFLAG_RW, 0,
119 "IP tunnel software network interface.");
120 SYSCTL_INT(_net_link_tun, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tundclone, 0,
121 "Enable legacy devfs interface creation.");
122
123 TUNABLE_INT("net.link.tun.devfs_cloning", &tundclone);
124
125 static void tunclone(void *arg, struct ucred *cred, char *name,
126 int namelen, struct cdev **dev);
127 static void tuncreate(const char *name, struct cdev *dev);
128 static int tunifioctl(struct ifnet *, u_long, caddr_t);
129 static void tuninit(struct ifnet *);
130 static int tunmodevent(module_t, int, void *);
131 static int tunoutput(struct ifnet *, struct mbuf *, struct sockaddr *,
132 struct route *ro);
133 static void tunstart(struct ifnet *);
134
135 static int tun_clone_create(struct if_clone *, int, caddr_t);
136 static void tun_clone_destroy(struct ifnet *);
137
138 IFC_SIMPLE_DECLARE(tun, 0);
139
140 static d_open_t tunopen;
141 static d_close_t tunclose;
142 static d_read_t tunread;
143 static d_write_t tunwrite;
144 static d_ioctl_t tunioctl;
145 static d_poll_t tunpoll;
146 static d_kqfilter_t tunkqfilter;
147
148 static int tunkqread(struct knote *, long);
149 static int tunkqwrite(struct knote *, long);
150 static void tunkqdetach(struct knote *);
151
152 static struct filterops tun_read_filterops = {
153 .f_isfd = 1,
154 .f_attach = NULL,
155 .f_detach = tunkqdetach,
156 .f_event = tunkqread,
157 };
158
159 static struct filterops tun_write_filterops = {
160 .f_isfd = 1,
161 .f_attach = NULL,
162 .f_detach = tunkqdetach,
163 .f_event = tunkqwrite,
164 };
165
166 static struct cdevsw tun_cdevsw = {
167 .d_version = D_VERSION,
168 .d_flags = D_PSEUDO | D_NEEDMINOR,
169 .d_open = tunopen,
170 .d_close = tunclose,
171 .d_read = tunread,
172 .d_write = tunwrite,
173 .d_ioctl = tunioctl,
174 .d_poll = tunpoll,
175 .d_kqfilter = tunkqfilter,
176 .d_name = TUNNAME,
177 };
178
179 static int
180 tun_clone_create(struct if_clone *ifc, int unit, caddr_t params)
181 {
182 struct cdev *dev;
183 int i;
184
185 /* find any existing device, or allocate new unit number */
186 i = clone_create(&tunclones, &tun_cdevsw, &unit, &dev, 0);
187 if (i) {
188 /* No preexisting struct cdev *, create one */
189 dev = make_dev(&tun_cdevsw, unit,
190 UID_UUCP, GID_DIALER, 0600, "%s%d", ifc->ifc_name, unit);
191 }
192 tuncreate(ifc->ifc_name, dev);
193
194 return (0);
195 }
196
197 static void
198 tunclone(void *arg, struct ucred *cred, char *name, int namelen,
199 struct cdev **dev)
200 {
201 char devname[SPECNAMELEN + 1];
202 int u, i, append_unit;
203
204 if (*dev != NULL)
205 return;
206
207 /*
208 * If tun cloning is enabled, only the superuser can create an
209 * interface.
210 */
211 if (!tundclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0)
212 return;
213
214 if (strcmp(name, TUNNAME) == 0) {
215 u = -1;
216 } else if (dev_stdclone(name, NULL, TUNNAME, &u) != 1)
217 return; /* Don't recognise the name */
218 if (u != -1 && u > IF_MAXUNIT)
219 return; /* Unit number too high */
220
221 if (u == -1)
222 append_unit = 1;
223 else
224 append_unit = 0;
225
226 CURVNET_SET(CRED_TO_VNET(cred));
227 /* find any existing device, or allocate new unit number */
228 i = clone_create(&tunclones, &tun_cdevsw, &u, dev, 0);
229 if (i) {
230 if (append_unit) {
231 namelen = snprintf(devname, sizeof(devname), "%s%d",
232 name, u);
233 name = devname;
234 }
235 /* No preexisting struct cdev *, create one */
236 *dev = make_dev_credf(MAKEDEV_REF, &tun_cdevsw, u, cred,
237 UID_UUCP, GID_DIALER, 0600, "%s", name);
238 }
239
240 if_clone_create(name, namelen, NULL);
241 CURVNET_RESTORE();
242 }
243
244 static void
245 tun_destroy(struct tun_softc *tp)
246 {
247 struct cdev *dev;
248
249 /* Unlocked read. */
250 mtx_lock(&tp->tun_mtx);
251 if ((tp->tun_flags & TUN_OPEN) != 0)
252 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx);
253 else
254 mtx_unlock(&tp->tun_mtx);
255
256 CURVNET_SET(TUN2IFP(tp)->if_vnet);
257 dev = tp->tun_dev;
258 bpfdetach(TUN2IFP(tp));
259 if_detach(TUN2IFP(tp));
260 if_free(TUN2IFP(tp));
261 destroy_dev(dev);
262 seldrain(&tp->tun_rsel);
263 knlist_destroy(&tp->tun_rsel.si_note);
264 mtx_destroy(&tp->tun_mtx);
265 cv_destroy(&tp->tun_cv);
266 free(tp, M_TUN);
267 CURVNET_RESTORE();
268 }
269
270 static void
271 tun_clone_destroy(struct ifnet *ifp)
272 {
273 struct tun_softc *tp = ifp->if_softc;
274
275 mtx_lock(&tunmtx);
276 TAILQ_REMOVE(&tunhead, tp, tun_list);
277 mtx_unlock(&tunmtx);
278 tun_destroy(tp);
279 }
280
281 static int
282 tunmodevent(module_t mod, int type, void *data)
283 {
284 static eventhandler_tag tag;
285 struct tun_softc *tp;
286
287 switch (type) {
288 case MOD_LOAD:
289 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF);
290 clone_setup(&tunclones);
291 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000);
292 if (tag == NULL)
293 return (ENOMEM);
294 if_clone_attach(&tun_cloner);
295 break;
296 case MOD_UNLOAD:
297 if_clone_detach(&tun_cloner);
298 EVENTHANDLER_DEREGISTER(dev_clone, tag);
299 drain_dev_clone_events();
300
301 mtx_lock(&tunmtx);
302 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) {
303 TAILQ_REMOVE(&tunhead, tp, tun_list);
304 mtx_unlock(&tunmtx);
305 tun_destroy(tp);
306 mtx_lock(&tunmtx);
307 }
308 mtx_unlock(&tunmtx);
309 clone_cleanup(&tunclones);
310 mtx_destroy(&tunmtx);
311 break;
312 default:
313 return EOPNOTSUPP;
314 }
315 return 0;
316 }
317
318 static moduledata_t tun_mod = {
319 "if_tun",
320 tunmodevent,
321 0
322 };
323
324 DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
325 MODULE_VERSION(if_tun, 1);
326
327 static void
328 tunstart(struct ifnet *ifp)
329 {
330 struct tun_softc *tp = ifp->if_softc;
331 struct mbuf *m;
332
333 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname);
334 if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
335 IFQ_LOCK(&ifp->if_snd);
336 IFQ_POLL_NOLOCK(&ifp->if_snd, m);
337 if (m == NULL) {
338 IFQ_UNLOCK(&ifp->if_snd);
339 return;
340 }
341 IFQ_UNLOCK(&ifp->if_snd);
342 }
343
344 mtx_lock(&tp->tun_mtx);
345 if (tp->tun_flags & TUN_RWAIT) {
346 tp->tun_flags &= ~TUN_RWAIT;
347 wakeup(tp);
348 }
349 selwakeuppri(&tp->tun_rsel, PZERO + 1);
350 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0);
351 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) {
352 mtx_unlock(&tp->tun_mtx);
353 pgsigio(&tp->tun_sigio, SIGIO, 0);
354 } else
355 mtx_unlock(&tp->tun_mtx);
356 }
357
358 /* XXX: should return an error code so it can fail. */
359 static void
360 tuncreate(const char *name, struct cdev *dev)
361 {
362 struct tun_softc *sc;
363 struct ifnet *ifp;
364
365 dev->si_flags &= ~SI_CHEAPCLONE;
366
367 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO);
368 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF);
369 cv_init(&sc->tun_cv, "tun_condvar");
370 sc->tun_flags = TUN_INITED;
371 sc->tun_dev = dev;
372 mtx_lock(&tunmtx);
373 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list);
374 mtx_unlock(&tunmtx);
375
376 ifp = sc->tun_ifp = if_alloc(IFT_PPP);
377 if (ifp == NULL)
378 panic("%s%d: failed to if_alloc() interface.\n",
379 name, dev2unit(dev));
380 if_initname(ifp, name, dev2unit(dev));
381 ifp->if_mtu = TUNMTU;
382 ifp->if_ioctl = tunifioctl;
383 ifp->if_output = tunoutput;
384 ifp->if_start = tunstart;
385 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
386 ifp->if_softc = sc;
387 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
388 ifp->if_snd.ifq_drv_maxlen = 0;
389 IFQ_SET_READY(&ifp->if_snd);
390 knlist_init_mtx(&sc->tun_rsel.si_note, &sc->tun_mtx);
391 ifp->if_capabilities |= IFCAP_LINKSTATE;
392 ifp->if_capenable |= IFCAP_LINKSTATE;
393
394 if_attach(ifp);
395 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
396 dev->si_drv1 = sc;
397 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n",
398 ifp->if_xname, dev2unit(dev));
399 }
400
401 static int
402 tunopen(struct cdev *dev, int flag, int mode, struct thread *td)
403 {
404 struct ifnet *ifp;
405 struct tun_softc *tp;
406
407 /*
408 * XXXRW: Non-atomic test and set of dev->si_drv1 requires
409 * synchronization.
410 */
411 tp = dev->si_drv1;
412 if (!tp) {
413 tuncreate(TUNNAME, dev);
414 tp = dev->si_drv1;
415 }
416
417 /*
418 * XXXRW: This use of tun_pid is subject to error due to the
419 * fact that a reference to the tunnel can live beyond the
420 * death of the process that created it. Can we replace this
421 * with a simple busy flag?
422 */
423 mtx_lock(&tp->tun_mtx);
424 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) {
425 mtx_unlock(&tp->tun_mtx);
426 return (EBUSY);
427 }
428 tp->tun_pid = td->td_proc->p_pid;
429
430 tp->tun_flags |= TUN_OPEN;
431 ifp = TUN2IFP(tp);
432 if_link_state_change(ifp, LINK_STATE_UP);
433 TUNDEBUG(ifp, "open\n");
434 mtx_unlock(&tp->tun_mtx);
435
436 return (0);
437 }
438
439 /*
440 * tunclose - close the device - mark i/f down & delete
441 * routing info
442 */
443 static int
444 tunclose(struct cdev *dev, int foo, int bar, struct thread *td)
445 {
446 struct tun_softc *tp;
447 struct ifnet *ifp;
448
449 tp = dev->si_drv1;
450 ifp = TUN2IFP(tp);
451
452 mtx_lock(&tp->tun_mtx);
453 tp->tun_flags &= ~TUN_OPEN;
454 tp->tun_pid = 0;
455
456 /*
457 * junk all pending output
458 */
459 CURVNET_SET(ifp->if_vnet);
460 IFQ_PURGE(&ifp->if_snd);
461
462 if (ifp->if_flags & IFF_UP) {
463 mtx_unlock(&tp->tun_mtx);
464 if_down(ifp);
465 mtx_lock(&tp->tun_mtx);
466 }
467
468 /* Delete all addresses and routes which reference this interface. */
469 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
470 struct ifaddr *ifa;
471
472 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
473 mtx_unlock(&tp->tun_mtx);
474 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
475 /* deal w/IPv4 PtP destination; unlocked read */
476 if (ifa->ifa_addr->sa_family == AF_INET) {
477 rtinit(ifa, (int)RTM_DELETE,
478 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0);
479 } else {
480 rtinit(ifa, (int)RTM_DELETE, 0);
481 }
482 }
483 if_purgeaddrs(ifp);
484 mtx_lock(&tp->tun_mtx);
485 }
486 if_link_state_change(ifp, LINK_STATE_DOWN);
487 CURVNET_RESTORE();
488
489 funsetown(&tp->tun_sigio);
490 selwakeuppri(&tp->tun_rsel, PZERO + 1);
491 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0);
492 TUNDEBUG (ifp, "closed\n");
493
494 cv_broadcast(&tp->tun_cv);
495 mtx_unlock(&tp->tun_mtx);
496 return (0);
497 }
498
499 static void
500 tuninit(struct ifnet *ifp)
501 {
502 struct tun_softc *tp = ifp->if_softc;
503 #ifdef INET
504 struct ifaddr *ifa;
505 #endif
506
507 TUNDEBUG(ifp, "tuninit\n");
508
509 mtx_lock(&tp->tun_mtx);
510 ifp->if_flags |= IFF_UP;
511 ifp->if_drv_flags |= IFF_DRV_RUNNING;
512 getmicrotime(&ifp->if_lastchange);
513
514 #ifdef INET
515 if_addr_rlock(ifp);
516 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
517 if (ifa->ifa_addr->sa_family == AF_INET) {
518 struct sockaddr_in *si;
519
520 si = (struct sockaddr_in *)ifa->ifa_addr;
521 if (si->sin_addr.s_addr)
522 tp->tun_flags |= TUN_IASET;
523
524 si = (struct sockaddr_in *)ifa->ifa_dstaddr;
525 if (si && si->sin_addr.s_addr)
526 tp->tun_flags |= TUN_DSTADDR;
527 }
528 }
529 if_addr_runlock(ifp);
530 #endif
531 mtx_unlock(&tp->tun_mtx);
532 }
533
534 /*
535 * Process an ioctl request.
536 */
537 static int
538 tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
539 {
540 struct ifreq *ifr = (struct ifreq *)data;
541 struct tun_softc *tp = ifp->if_softc;
542 struct ifstat *ifs;
543 int error = 0;
544
545 switch(cmd) {
546 case SIOCGIFSTATUS:
547 ifs = (struct ifstat *)data;
548 mtx_lock(&tp->tun_mtx);
549 if (tp->tun_pid)
550 sprintf(ifs->ascii + strlen(ifs->ascii),
551 "\tOpened by PID %d\n", tp->tun_pid);
552 mtx_unlock(&tp->tun_mtx);
553 break;
554 case SIOCSIFADDR:
555 tuninit(ifp);
556 TUNDEBUG(ifp, "address set\n");
557 break;
558 case SIOCSIFDSTADDR:
559 tuninit(ifp);
560 TUNDEBUG(ifp, "destination address set\n");
561 break;
562 case SIOCSIFMTU:
563 ifp->if_mtu = ifr->ifr_mtu;
564 TUNDEBUG(ifp, "mtu set\n");
565 break;
566 case SIOCSIFFLAGS:
567 case SIOCADDMULTI:
568 case SIOCDELMULTI:
569 break;
570 default:
571 error = EINVAL;
572 }
573 return (error);
574 }
575
576 /*
577 * tunoutput - queue packets from higher level ready to put out.
578 */
579 static int
580 tunoutput(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst,
581 struct route *ro)
582 {
583 struct tun_softc *tp = ifp->if_softc;
584 u_short cached_tun_flags;
585 int error;
586 u_int32_t af;
587
588 TUNDEBUG (ifp, "tunoutput\n");
589
590 #ifdef MAC
591 error = mac_ifnet_check_transmit(ifp, m0);
592 if (error) {
593 m_freem(m0);
594 return (error);
595 }
596 #endif
597
598 /* Could be unlocked read? */
599 mtx_lock(&tp->tun_mtx);
600 cached_tun_flags = tp->tun_flags;
601 mtx_unlock(&tp->tun_mtx);
602 if ((cached_tun_flags & TUN_READY) != TUN_READY) {
603 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags);
604 m_freem (m0);
605 return (EHOSTDOWN);
606 }
607
608 if ((ifp->if_flags & IFF_UP) != IFF_UP) {
609 m_freem (m0);
610 return (EHOSTDOWN);
611 }
612
613 /* BPF writes need to be handled specially. */
614 if (dst->sa_family == AF_UNSPEC) {
615 bcopy(dst->sa_data, &af, sizeof(af));
616 dst->sa_family = af;
617 }
618
619 if (bpf_peers_present(ifp->if_bpf)) {
620 af = dst->sa_family;
621 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0);
622 }
623
624 /* prepend sockaddr? this may abort if the mbuf allocation fails */
625 if (cached_tun_flags & TUN_LMODE) {
626 /* allocate space for sockaddr */
627 M_PREPEND(m0, dst->sa_len, M_DONTWAIT);
628
629 /* if allocation failed drop packet */
630 if (m0 == NULL) {
631 ifp->if_iqdrops++;
632 ifp->if_oerrors++;
633 return (ENOBUFS);
634 } else {
635 bcopy(dst, m0->m_data, dst->sa_len);
636 }
637 }
638
639 if (cached_tun_flags & TUN_IFHEAD) {
640 /* Prepend the address family */
641 M_PREPEND(m0, 4, M_DONTWAIT);
642
643 /* if allocation failed drop packet */
644 if (m0 == NULL) {
645 ifp->if_iqdrops++;
646 ifp->if_oerrors++;
647 return (ENOBUFS);
648 } else
649 *(u_int32_t *)m0->m_data = htonl(dst->sa_family);
650 } else {
651 #ifdef INET
652 if (dst->sa_family != AF_INET)
653 #endif
654 {
655 m_freem(m0);
656 return (EAFNOSUPPORT);
657 }
658 }
659
660 error = (ifp->if_transmit)(ifp, m0);
661 if (error)
662 return (ENOBUFS);
663 ifp->if_opackets++;
664 return (0);
665 }
666
667 /*
668 * the cdevsw interface is now pretty minimal.
669 */
670 static int
671 tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag,
672 struct thread *td)
673 {
674 int error;
675 struct tun_softc *tp = dev->si_drv1;
676 struct tuninfo *tunp;
677
678 switch (cmd) {
679 case TUNSIFINFO:
680 tunp = (struct tuninfo *)data;
681 if (tunp->mtu < IF_MINMTU)
682 return (EINVAL);
683 if (TUN2IFP(tp)->if_mtu != tunp->mtu) {
684 error = priv_check(td, PRIV_NET_SETIFMTU);
685 if (error)
686 return (error);
687 }
688 if (TUN2IFP(tp)->if_type != tunp->type)
689 return (EPROTOTYPE);
690 mtx_lock(&tp->tun_mtx);
691 TUN2IFP(tp)->if_mtu = tunp->mtu;
692 TUN2IFP(tp)->if_baudrate = tunp->baudrate;
693 mtx_unlock(&tp->tun_mtx);
694 break;
695 case TUNGIFINFO:
696 tunp = (struct tuninfo *)data;
697 mtx_lock(&tp->tun_mtx);
698 tunp->mtu = TUN2IFP(tp)->if_mtu;
699 tunp->type = TUN2IFP(tp)->if_type;
700 tunp->baudrate = TUN2IFP(tp)->if_baudrate;
701 mtx_unlock(&tp->tun_mtx);
702 break;
703 case TUNSDEBUG:
704 tundebug = *(int *)data;
705 break;
706 case TUNGDEBUG:
707 *(int *)data = tundebug;
708 break;
709 case TUNSLMODE:
710 mtx_lock(&tp->tun_mtx);
711 if (*(int *)data) {
712 tp->tun_flags |= TUN_LMODE;
713 tp->tun_flags &= ~TUN_IFHEAD;
714 } else
715 tp->tun_flags &= ~TUN_LMODE;
716 mtx_unlock(&tp->tun_mtx);
717 break;
718 case TUNSIFHEAD:
719 mtx_lock(&tp->tun_mtx);
720 if (*(int *)data) {
721 tp->tun_flags |= TUN_IFHEAD;
722 tp->tun_flags &= ~TUN_LMODE;
723 } else
724 tp->tun_flags &= ~TUN_IFHEAD;
725 mtx_unlock(&tp->tun_mtx);
726 break;
727 case TUNGIFHEAD:
728 mtx_lock(&tp->tun_mtx);
729 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0;
730 mtx_unlock(&tp->tun_mtx);
731 break;
732 case TUNSIFMODE:
733 /* deny this if UP */
734 if (TUN2IFP(tp)->if_flags & IFF_UP)
735 return(EBUSY);
736
737 switch (*(int *)data & ~IFF_MULTICAST) {
738 case IFF_POINTOPOINT:
739 case IFF_BROADCAST:
740 mtx_lock(&tp->tun_mtx);
741 TUN2IFP(tp)->if_flags &=
742 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST);
743 TUN2IFP(tp)->if_flags |= *(int *)data;
744 mtx_unlock(&tp->tun_mtx);
745 break;
746 default:
747 return(EINVAL);
748 }
749 break;
750 case TUNSIFPID:
751 mtx_lock(&tp->tun_mtx);
752 tp->tun_pid = curthread->td_proc->p_pid;
753 mtx_unlock(&tp->tun_mtx);
754 break;
755 case FIONBIO:
756 break;
757 case FIOASYNC:
758 mtx_lock(&tp->tun_mtx);
759 if (*(int *)data)
760 tp->tun_flags |= TUN_ASYNC;
761 else
762 tp->tun_flags &= ~TUN_ASYNC;
763 mtx_unlock(&tp->tun_mtx);
764 break;
765 case FIONREAD:
766 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) {
767 struct mbuf *mb;
768 IFQ_LOCK(&TUN2IFP(tp)->if_snd);
769 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb);
770 for (*(int *)data = 0; mb != NULL; mb = mb->m_next)
771 *(int *)data += mb->m_len;
772 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd);
773 } else
774 *(int *)data = 0;
775 break;
776 case FIOSETOWN:
777 return (fsetown(*(int *)data, &tp->tun_sigio));
778
779 case FIOGETOWN:
780 *(int *)data = fgetown(&tp->tun_sigio);
781 return (0);
782
783 /* This is deprecated, FIOSETOWN should be used instead. */
784 case TIOCSPGRP:
785 return (fsetown(-(*(int *)data), &tp->tun_sigio));
786
787 /* This is deprecated, FIOGETOWN should be used instead. */
788 case TIOCGPGRP:
789 *(int *)data = -fgetown(&tp->tun_sigio);
790 return (0);
791
792 default:
793 return (ENOTTY);
794 }
795 return (0);
796 }
797
798 /*
799 * The cdevsw read interface - reads a packet at a time, or at
800 * least as much of a packet as can be read.
801 */
802 static int
803 tunread(struct cdev *dev, struct uio *uio, int flag)
804 {
805 struct tun_softc *tp = dev->si_drv1;
806 struct ifnet *ifp = TUN2IFP(tp);
807 struct mbuf *m;
808 int error=0, len;
809
810 TUNDEBUG (ifp, "read\n");
811 mtx_lock(&tp->tun_mtx);
812 if ((tp->tun_flags & TUN_READY) != TUN_READY) {
813 mtx_unlock(&tp->tun_mtx);
814 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags);
815 return (EHOSTDOWN);
816 }
817
818 tp->tun_flags &= ~TUN_RWAIT;
819
820 do {
821 IFQ_DEQUEUE(&ifp->if_snd, m);
822 if (m == NULL) {
823 if (flag & O_NONBLOCK) {
824 mtx_unlock(&tp->tun_mtx);
825 return (EWOULDBLOCK);
826 }
827 tp->tun_flags |= TUN_RWAIT;
828 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1),
829 "tunread", 0);
830 if (error != 0) {
831 mtx_unlock(&tp->tun_mtx);
832 return (error);
833 }
834 }
835 } while (m == NULL);
836 mtx_unlock(&tp->tun_mtx);
837
838 while (m && uio->uio_resid > 0 && error == 0) {
839 len = min(uio->uio_resid, m->m_len);
840 if (len != 0)
841 error = uiomove(mtod(m, void *), len, uio);
842 m = m_free(m);
843 }
844
845 if (m) {
846 TUNDEBUG(ifp, "Dropping mbuf\n");
847 m_freem(m);
848 }
849 return (error);
850 }
851
852 /*
853 * the cdevsw write interface - an atomic write is a packet - or else!
854 */
855 static int
856 tunwrite(struct cdev *dev, struct uio *uio, int flag)
857 {
858 struct tun_softc *tp = dev->si_drv1;
859 struct ifnet *ifp = TUN2IFP(tp);
860 struct mbuf *m;
861 uint32_t family;
862 int isr;
863
864 TUNDEBUG(ifp, "tunwrite\n");
865
866 if ((ifp->if_flags & IFF_UP) != IFF_UP)
867 /* ignore silently */
868 return (0);
869
870 if (uio->uio_resid == 0)
871 return (0);
872
873 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) {
874 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid);
875 return (EIO);
876 }
877
878 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, 0, M_PKTHDR)) == NULL) {
879 ifp->if_ierrors++;
880 return (ENOBUFS);
881 }
882
883 m->m_pkthdr.rcvif = ifp;
884 #ifdef MAC
885 mac_ifnet_create_mbuf(ifp, m);
886 #endif
887
888 /* Could be unlocked read? */
889 mtx_lock(&tp->tun_mtx);
890 if (tp->tun_flags & TUN_IFHEAD) {
891 mtx_unlock(&tp->tun_mtx);
892 if (m->m_len < sizeof(family) &&
893 (m = m_pullup(m, sizeof(family))) == NULL)
894 return (ENOBUFS);
895 family = ntohl(*mtod(m, u_int32_t *));
896 m_adj(m, sizeof(family));
897 } else {
898 mtx_unlock(&tp->tun_mtx);
899 family = AF_INET;
900 }
901
902 BPF_MTAP2(ifp, &family, sizeof(family), m);
903
904 switch (family) {
905 #ifdef INET
906 case AF_INET:
907 isr = NETISR_IP;
908 break;
909 #endif
910 #ifdef INET6
911 case AF_INET6:
912 isr = NETISR_IPV6;
913 break;
914 #endif
915 #ifdef IPX
916 case AF_IPX:
917 isr = NETISR_IPX;
918 break;
919 #endif
920 #ifdef NETATALK
921 case AF_APPLETALK:
922 isr = NETISR_ATALK2;
923 break;
924 #endif
925 default:
926 m_freem(m);
927 return (EAFNOSUPPORT);
928 }
929 /* First chunk of an mbuf contains good junk */
930 if (harvest.point_to_point)
931 random_harvest(m, 16, 3, 0, RANDOM_NET);
932 ifp->if_ibytes += m->m_pkthdr.len;
933 ifp->if_ipackets++;
934 CURVNET_SET(ifp->if_vnet);
935 M_SETFIB(m, ifp->if_fib);
936 netisr_dispatch(isr, m);
937 CURVNET_RESTORE();
938 return (0);
939 }
940
941 /*
942 * tunpoll - the poll interface, this is only useful on reads
943 * really. The write detect always returns true, write never blocks
944 * anyway, it either accepts the packet or drops it.
945 */
946 static int
947 tunpoll(struct cdev *dev, int events, struct thread *td)
948 {
949 struct tun_softc *tp = dev->si_drv1;
950 struct ifnet *ifp = TUN2IFP(tp);
951 int revents = 0;
952 struct mbuf *m;
953
954 TUNDEBUG(ifp, "tunpoll\n");
955
956 if (events & (POLLIN | POLLRDNORM)) {
957 IFQ_LOCK(&ifp->if_snd);
958 IFQ_POLL_NOLOCK(&ifp->if_snd, m);
959 if (m != NULL) {
960 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len);
961 revents |= events & (POLLIN | POLLRDNORM);
962 } else {
963 TUNDEBUG(ifp, "tunpoll waiting\n");
964 selrecord(td, &tp->tun_rsel);
965 }
966 IFQ_UNLOCK(&ifp->if_snd);
967 }
968 if (events & (POLLOUT | POLLWRNORM))
969 revents |= events & (POLLOUT | POLLWRNORM);
970
971 return (revents);
972 }
973
974 /*
975 * tunkqfilter - support for the kevent() system call.
976 */
977 static int
978 tunkqfilter(struct cdev *dev, struct knote *kn)
979 {
980 struct tun_softc *tp = dev->si_drv1;
981 struct ifnet *ifp = TUN2IFP(tp);
982
983 switch(kn->kn_filter) {
984 case EVFILT_READ:
985 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n",
986 ifp->if_xname, dev2unit(dev));
987 kn->kn_fop = &tun_read_filterops;
988 break;
989
990 case EVFILT_WRITE:
991 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n",
992 ifp->if_xname, dev2unit(dev));
993 kn->kn_fop = &tun_write_filterops;
994 break;
995
996 default:
997 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n",
998 ifp->if_xname, dev2unit(dev));
999 return(EINVAL);
1000 }
1001
1002 kn->kn_hook = tp;
1003 knlist_add(&tp->tun_rsel.si_note, kn, 0);
1004
1005 return (0);
1006 }
1007
1008 /*
1009 * Return true of there is data in the interface queue.
1010 */
1011 static int
1012 tunkqread(struct knote *kn, long hint)
1013 {
1014 int ret;
1015 struct tun_softc *tp = kn->kn_hook;
1016 struct cdev *dev = tp->tun_dev;
1017 struct ifnet *ifp = TUN2IFP(tp);
1018
1019 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1020 TUNDEBUG(ifp,
1021 "%s have data in the queue. Len = %d, minor = %#x\n",
1022 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1023 ret = 1;
1024 } else {
1025 TUNDEBUG(ifp,
1026 "%s waiting for data, minor = %#x\n", ifp->if_xname,
1027 dev2unit(dev));
1028 ret = 0;
1029 }
1030
1031 return (ret);
1032 }
1033
1034 /*
1035 * Always can write, always return MTU in kn->data.
1036 */
1037 static int
1038 tunkqwrite(struct knote *kn, long hint)
1039 {
1040 struct tun_softc *tp = kn->kn_hook;
1041 struct ifnet *ifp = TUN2IFP(tp);
1042
1043 kn->kn_data = ifp->if_mtu;
1044
1045 return (1);
1046 }
1047
1048 static void
1049 tunkqdetach(struct knote *kn)
1050 {
1051 struct tun_softc *tp = kn->kn_hook;
1052
1053 knlist_remove(&tp->tun_rsel.si_note, kn, 0);
1054 }
Cache object: 1fa559e5992ea1a10b6d4cd6b4e58711
|