FreeBSD/Linux Kernel Cross Reference
sys/net/if_tap.c
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 * $FreeBSD$
35 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
36 */
37
38 #include "opt_compat.h"
39 #include "opt_inet.h"
40
41 #include <sys/param.h>
42 #include <sys/conf.h>
43 #include <sys/fcntl.h>
44 #include <sys/filio.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/module.h>
49 #include <sys/poll.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/selinfo.h>
53 #include <sys/signalvar.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/systm.h>
58 #include <sys/ttycom.h>
59 #include <sys/uio.h>
60 #include <sys/queue.h>
61
62 #include <net/bpf.h>
63 #include <net/ethernet.h>
64 #include <net/if.h>
65 #include <net/if_clone.h>
66 #include <net/if_dl.h>
67 #include <net/route.h>
68 #include <net/if_types.h>
69
70 #include <netinet/in.h>
71
72 #include <net/if_tapvar.h>
73 #include <net/if_tap.h>
74
75
76 #define CDEV_NAME "tap"
77 #define TAPDEBUG if (tapdebug) printf
78
79 #define TAP "tap"
80 #define VMNET "vmnet"
81 #define TAPMAXUNIT 0x7fff
82 #define VMNET_DEV_MASK CLONE_FLAG0
83
84 /* module */
85 static int tapmodevent(module_t, int, void *);
86
87 /* device */
88 static void tapclone(void *, struct ucred *, char *, int,
89 struct cdev **);
90 static void tapcreate(struct cdev *);
91
92 /* network interface */
93 static void tapifstart(struct ifnet *);
94 static int tapifioctl(struct ifnet *, u_long, caddr_t);
95 static void tapifinit(void *);
96
97 static int tap_clone_create(struct if_clone *, int, caddr_t);
98 static void tap_clone_destroy(struct ifnet *);
99 static int vmnet_clone_create(struct if_clone *, int, caddr_t);
100 static void vmnet_clone_destroy(struct ifnet *);
101
102 IFC_SIMPLE_DECLARE(tap, 0);
103 IFC_SIMPLE_DECLARE(vmnet, 0);
104
105 /* character device */
106 static d_open_t tapopen;
107 static d_close_t tapclose;
108 static d_read_t tapread;
109 static d_write_t tapwrite;
110 static d_ioctl_t tapioctl;
111 static d_poll_t tappoll;
112 static d_kqfilter_t tapkqfilter;
113
114 /* kqueue(2) */
115 static int tapkqread(struct knote *, long);
116 static int tapkqwrite(struct knote *, long);
117 static void tapkqdetach(struct knote *);
118
119 static struct filterops tap_read_filterops = {
120 .f_isfd = 1,
121 .f_attach = NULL,
122 .f_detach = tapkqdetach,
123 .f_event = tapkqread,
124 };
125
126 static struct filterops tap_write_filterops = {
127 .f_isfd = 1,
128 .f_attach = NULL,
129 .f_detach = tapkqdetach,
130 .f_event = tapkqwrite,
131 };
132
133 static struct cdevsw tap_cdevsw = {
134 .d_version = D_VERSION,
135 .d_flags = D_PSEUDO | D_NEEDGIANT,
136 .d_open = tapopen,
137 .d_close = tapclose,
138 .d_read = tapread,
139 .d_write = tapwrite,
140 .d_ioctl = tapioctl,
141 .d_poll = tappoll,
142 .d_name = CDEV_NAME,
143 .d_kqfilter = tapkqfilter,
144 };
145
146 /*
147 * All global variables in if_tap.c are locked with tapmtx, with the
148 * exception of tapdebug, which is accessed unlocked; tapclones is
149 * static at runtime.
150 */
151 static struct mtx tapmtx;
152 static int tapdebug = 0; /* debug flag */
153 static int tapuopen = 0; /* allow user open() */
154 static int tapuponopen = 0; /* IFF_UP on open() */
155 static int tapdclone = 1; /* enable devfs cloning */
156 static SLIST_HEAD(, tap_softc) taphead; /* first device */
157 static struct clonedevs *tapclones;
158
159 MALLOC_DECLARE(M_TAP);
160 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
161 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
162
163 SYSCTL_DECL(_net_link);
164 SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
165 "Ethernet tunnel software network interface");
166 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
167 "Allow user to open /dev/tap (based on node permissions)");
168 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
169 "Bring interface up when /dev/tap is opened");
170 SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0,
171 "Enably legacy devfs interface creation");
172 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
173
174 TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone);
175
176 DEV_MODULE(if_tap, tapmodevent, NULL);
177
178 static int
179 tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
180 {
181 struct cdev *dev;
182 int i;
183 int extra;
184
185 if (strcmp(ifc->ifc_name, VMNET) == 0)
186 extra = VMNET_DEV_MASK;
187 else
188 extra = 0;
189
190 /* find any existing device, or allocate new unit number */
191 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, extra);
192 if (i) {
193 dev = make_dev(&tap_cdevsw, unit2minor(unit | extra),
194 UID_ROOT, GID_WHEEL, 0600, "%s%d", ifc->ifc_name, unit);
195 if (dev != NULL) {
196 dev_ref(dev);
197 dev->si_flags |= SI_CHEAPCLONE;
198 }
199 }
200
201 tapcreate(dev);
202 return (0);
203 }
204
205 /* vmnet devices are tap devices in disguise */
206 static int
207 vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
208 {
209 return tap_clone_create(ifc, unit, params);
210 }
211
212 static void
213 tap_destroy(struct tap_softc *tp)
214 {
215 struct ifnet *ifp = tp->tap_ifp;
216 int s;
217
218 /* Unlocked read. */
219 KASSERT(!(tp->tap_flags & TAP_OPEN),
220 ("%s flags is out of sync", ifp->if_xname));
221
222 knlist_destroy(&tp->tap_rsel.si_note);
223 destroy_dev(tp->tap_dev);
224 s = splimp();
225 ether_ifdetach(ifp);
226 if_free_type(ifp, IFT_ETHER);
227 splx(s);
228
229 mtx_destroy(&tp->tap_mtx);
230 free(tp, M_TAP);
231 }
232
233 static void
234 tap_clone_destroy(struct ifnet *ifp)
235 {
236 struct tap_softc *tp = ifp->if_softc;
237
238 mtx_lock(&tapmtx);
239 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
240 mtx_unlock(&tapmtx);
241 tap_destroy(tp);
242 }
243
244 /* vmnet devices are tap devices in disguise */
245 static void
246 vmnet_clone_destroy(struct ifnet *ifp)
247 {
248 tap_clone_destroy(ifp);
249 }
250
251 /*
252 * tapmodevent
253 *
254 * module event handler
255 */
256 static int
257 tapmodevent(module_t mod, int type, void *data)
258 {
259 static eventhandler_tag eh_tag = NULL;
260 struct tap_softc *tp = NULL;
261 struct ifnet *ifp = NULL;
262
263 switch (type) {
264 case MOD_LOAD:
265
266 /* intitialize device */
267
268 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
269 SLIST_INIT(&taphead);
270
271 clone_setup(&tapclones);
272 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
273 if (eh_tag == NULL) {
274 clone_cleanup(&tapclones);
275 mtx_destroy(&tapmtx);
276 return (ENOMEM);
277 }
278 if_clone_attach(&tap_cloner);
279 if_clone_attach(&vmnet_cloner);
280 return (0);
281
282 case MOD_UNLOAD:
283 /*
284 * The EBUSY algorithm here can't quite atomically
285 * guarantee that this is race-free since we have to
286 * release the tap mtx to deregister the clone handler.
287 */
288 mtx_lock(&tapmtx);
289 SLIST_FOREACH(tp, &taphead, tap_next) {
290 mtx_lock(&tp->tap_mtx);
291 if (tp->tap_flags & TAP_OPEN) {
292 mtx_unlock(&tp->tap_mtx);
293 mtx_unlock(&tapmtx);
294 return (EBUSY);
295 }
296 mtx_unlock(&tp->tap_mtx);
297 }
298 mtx_unlock(&tapmtx);
299
300 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
301 if_clone_detach(&tap_cloner);
302 if_clone_detach(&vmnet_cloner);
303
304 mtx_lock(&tapmtx);
305 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
306 SLIST_REMOVE_HEAD(&taphead, tap_next);
307 mtx_unlock(&tapmtx);
308
309 ifp = tp->tap_ifp;
310
311 TAPDEBUG("detaching %s\n", ifp->if_xname);
312
313 tap_destroy(tp);
314 mtx_lock(&tapmtx);
315 }
316 mtx_unlock(&tapmtx);
317 clone_cleanup(&tapclones);
318
319 mtx_destroy(&tapmtx);
320
321 break;
322
323 default:
324 return (EOPNOTSUPP);
325 }
326
327 return (0);
328 } /* tapmodevent */
329
330
331 /*
332 * DEVFS handler
333 *
334 * We need to support two kind of devices - tap and vmnet
335 */
336 static void
337 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
338 {
339 char devname[SPECNAMELEN + 1];
340 int i, unit, append_unit;
341 int extra;
342
343 if (*dev != NULL)
344 return;
345
346 if (!tapdclone ||
347 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
348 return;
349
350 unit = 0;
351 append_unit = 0;
352 extra = 0;
353
354 /* We're interested in only tap/vmnet devices. */
355 if (strcmp(name, TAP) == 0) {
356 unit = -1;
357 } else if (strcmp(name, VMNET) == 0) {
358 unit = -1;
359 extra = VMNET_DEV_MASK;
360 } else if (dev_stdclone(name, NULL, TAP, &unit) != 1) {
361 if (dev_stdclone(name, NULL, VMNET, &unit) != 1) {
362 return;
363 } else {
364 extra = VMNET_DEV_MASK;
365 }
366 }
367
368 if (unit == -1)
369 append_unit = 1;
370
371 /* find any existing device, or allocate new unit number */
372 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
373 if (i) {
374 if (append_unit) {
375 /*
376 * We were passed 'tun' or 'tap', with no unit specified
377 * so we'll need to append it now.
378 */
379 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
380 unit);
381 name = devname;
382 }
383
384 *dev = make_dev(&tap_cdevsw, unit2minor(unit | extra),
385 UID_ROOT, GID_WHEEL, 0600, "%s", name);
386 if (*dev != NULL) {
387 dev_ref(*dev);
388 (*dev)->si_flags |= SI_CHEAPCLONE;
389 }
390 }
391
392 if_clone_create(name, namelen, NULL);
393 } /* tapclone */
394
395
396 /*
397 * tapcreate
398 *
399 * to create interface
400 */
401 static void
402 tapcreate(struct cdev *dev)
403 {
404 struct ifnet *ifp = NULL;
405 struct tap_softc *tp = NULL;
406 unsigned short macaddr_hi;
407 uint32_t macaddr_mid;
408 int unit, s;
409 char *name = NULL;
410 u_char eaddr[6];
411
412 dev->si_flags &= ~SI_CHEAPCLONE;
413
414 /* allocate driver storage and create device */
415 MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
416 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
417 mtx_lock(&tapmtx);
418 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
419 mtx_unlock(&tapmtx);
420
421 unit = dev2unit(dev);
422
423 /* select device: tap or vmnet */
424 if (unit & VMNET_DEV_MASK) {
425 name = VMNET;
426 tp->tap_flags |= TAP_VMNET;
427 } else
428 name = TAP;
429
430 unit &= TAPMAXUNIT;
431
432 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, minor(dev));
433
434 /* generate fake MAC address: 00 bd xx xx xx unit_no */
435 macaddr_hi = htons(0x00bd);
436 macaddr_mid = (uint32_t) ticks;
437 bcopy(&macaddr_hi, eaddr, sizeof(short));
438 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
439 eaddr[5] = (u_char)unit;
440
441 /* fill the rest and attach interface */
442 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
443 if (ifp == NULL)
444 panic("%s%d: can not if_alloc()", name, unit);
445 ifp->if_softc = tp;
446 if_initname(ifp, name, unit);
447 ifp->if_init = tapifinit;
448 ifp->if_start = tapifstart;
449 ifp->if_ioctl = tapifioctl;
450 ifp->if_mtu = ETHERMTU;
451 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
452 ifp->if_snd.ifq_maxlen = ifqmaxlen;
453
454 dev->si_drv1 = tp;
455 tp->tap_dev = dev;
456
457 s = splimp();
458 ether_ifattach(ifp, eaddr);
459 splx(s);
460
461 mtx_lock(&tp->tap_mtx);
462 tp->tap_flags |= TAP_INITED;
463 mtx_unlock(&tp->tap_mtx);
464
465 knlist_init(&tp->tap_rsel.si_note, NULL, NULL, NULL, NULL);
466
467 TAPDEBUG("interface %s is created. minor = %#x\n",
468 ifp->if_xname, minor(dev));
469 } /* tapcreate */
470
471
472 /*
473 * tapopen
474 *
475 * to open tunnel. must be superuser
476 */
477 static int
478 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
479 {
480 struct tap_softc *tp = NULL;
481 struct ifnet *ifp = NULL;
482 int error, s;
483
484 if (tapuopen == 0) {
485 error = priv_check(td, PRIV_NET_TAP);
486 if (error)
487 return (error);
488 }
489
490 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
491 return (ENXIO);
492
493 tp = dev->si_drv1;
494
495 mtx_lock(&tp->tap_mtx);
496 if (tp->tap_flags & TAP_OPEN) {
497 mtx_unlock(&tp->tap_mtx);
498 return (EBUSY);
499 }
500
501 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
502 tp->tap_pid = td->td_proc->p_pid;
503 tp->tap_flags |= TAP_OPEN;
504 ifp = tp->tap_ifp;
505 mtx_unlock(&tp->tap_mtx);
506
507 s = splimp();
508 ifp->if_drv_flags |= IFF_DRV_RUNNING;
509 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
510 if (tapuponopen)
511 ifp->if_flags |= IFF_UP;
512 splx(s);
513
514 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, minor(dev));
515
516 return (0);
517 } /* tapopen */
518
519
520 /*
521 * tapclose
522 *
523 * close the device - mark i/f down & delete routing info
524 */
525 static int
526 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
527 {
528 struct ifaddr *ifa;
529 struct tap_softc *tp = dev->si_drv1;
530 struct ifnet *ifp = tp->tap_ifp;
531 int s;
532
533 /* junk all pending output */
534 IF_DRAIN(&ifp->if_snd);
535
536 /*
537 * do not bring the interface down, and do not anything with
538 * interface, if we are in VMnet mode. just close the device.
539 */
540
541 mtx_lock(&tp->tap_mtx);
542 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
543 mtx_unlock(&tp->tap_mtx);
544 s = splimp();
545 if_down(ifp);
546 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
547 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
548 rtinit(ifa, (int)RTM_DELETE, 0);
549 }
550 if_purgeaddrs(ifp);
551 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
552 }
553 splx(s);
554 } else
555 mtx_unlock(&tp->tap_mtx);
556
557 funsetown(&tp->tap_sigio);
558 selwakeuppri(&tp->tap_rsel, PZERO+1);
559 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0);
560
561 mtx_lock(&tp->tap_mtx);
562 tp->tap_flags &= ~TAP_OPEN;
563 tp->tap_pid = 0;
564 mtx_unlock(&tp->tap_mtx);
565
566 TAPDEBUG("%s is closed. minor = %#x\n",
567 ifp->if_xname, minor(dev));
568
569 return (0);
570 } /* tapclose */
571
572
573 /*
574 * tapifinit
575 *
576 * network interface initialization function
577 */
578 static void
579 tapifinit(void *xtp)
580 {
581 struct tap_softc *tp = (struct tap_softc *)xtp;
582 struct ifnet *ifp = tp->tap_ifp;
583
584 TAPDEBUG("initializing %s\n", ifp->if_xname);
585
586 ifp->if_drv_flags |= IFF_DRV_RUNNING;
587 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
588
589 /* attempt to start output */
590 tapifstart(ifp);
591 } /* tapifinit */
592
593
594 /*
595 * tapifioctl
596 *
597 * Process an ioctl request on network interface
598 */
599 static int
600 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
601 {
602 struct tap_softc *tp = ifp->if_softc;
603 struct ifstat *ifs = NULL;
604 int s, dummy;
605
606 switch (cmd) {
607 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
608 case SIOCADDMULTI:
609 case SIOCDELMULTI:
610 break;
611
612 case SIOCGIFSTATUS:
613 s = splimp();
614 ifs = (struct ifstat *)data;
615 dummy = strlen(ifs->ascii);
616 mtx_lock(&tp->tap_mtx);
617 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
618 snprintf(ifs->ascii + dummy,
619 sizeof(ifs->ascii) - dummy,
620 "\tOpened by PID %d\n", tp->tap_pid);
621 mtx_unlock(&tp->tap_mtx);
622 splx(s);
623 break;
624
625 default:
626 s = splimp();
627 dummy = ether_ioctl(ifp, cmd, data);
628 splx(s);
629 return (dummy);
630 /* NOT REACHED */
631 }
632
633 return (0);
634 } /* tapifioctl */
635
636
637 /*
638 * tapifstart
639 *
640 * queue packets from higher level ready to put out
641 */
642 static void
643 tapifstart(struct ifnet *ifp)
644 {
645 struct tap_softc *tp = ifp->if_softc;
646 int s;
647
648 TAPDEBUG("%s starting\n", ifp->if_xname);
649
650 /*
651 * do not junk pending output if we are in VMnet mode.
652 * XXX: can this do any harm because of queue overflow?
653 */
654
655 mtx_lock(&tp->tap_mtx);
656 if (((tp->tap_flags & TAP_VMNET) == 0) &&
657 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
658 struct mbuf *m = NULL;
659
660 mtx_unlock(&tp->tap_mtx);
661
662 /* Unlocked read. */
663 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
664 tp->tap_flags);
665
666 s = splimp();
667 do {
668 IF_DEQUEUE(&ifp->if_snd, m);
669 if (m != NULL)
670 m_freem(m);
671 ifp->if_oerrors ++;
672 } while (m != NULL);
673 splx(s);
674
675 return;
676 }
677 mtx_unlock(&tp->tap_mtx);
678
679 s = splimp();
680 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
681
682 if (ifp->if_snd.ifq_len != 0) {
683 mtx_lock(&tp->tap_mtx);
684 if (tp->tap_flags & TAP_RWAIT) {
685 tp->tap_flags &= ~TAP_RWAIT;
686 wakeup(tp);
687 }
688
689 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
690 mtx_unlock(&tp->tap_mtx);
691 pgsigio(&tp->tap_sigio, SIGIO, 0);
692 } else
693 mtx_unlock(&tp->tap_mtx);
694
695 selwakeuppri(&tp->tap_rsel, PZERO+1);
696 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0);
697 ifp->if_opackets ++; /* obytes are counted in ether_output */
698 }
699
700 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
701 splx(s);
702 } /* tapifstart */
703
704
705 /*
706 * tapioctl
707 *
708 * the cdevsw interface is now pretty minimal
709 */
710 static int
711 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
712 {
713 struct tap_softc *tp = dev->si_drv1;
714 struct ifnet *ifp = tp->tap_ifp;
715 struct tapinfo *tapp = NULL;
716 int s;
717 int f;
718 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
719 defined(COMPAT_FREEBSD4)
720 int ival;
721 #endif
722
723 switch (cmd) {
724 case TAPSIFINFO:
725 s = splimp();
726 tapp = (struct tapinfo *)data;
727 ifp->if_mtu = tapp->mtu;
728 ifp->if_type = tapp->type;
729 ifp->if_baudrate = tapp->baudrate;
730 splx(s);
731 break;
732
733 case TAPGIFINFO:
734 tapp = (struct tapinfo *)data;
735 tapp->mtu = ifp->if_mtu;
736 tapp->type = ifp->if_type;
737 tapp->baudrate = ifp->if_baudrate;
738 break;
739
740 case TAPSDEBUG:
741 tapdebug = *(int *)data;
742 break;
743
744 case TAPGDEBUG:
745 *(int *)data = tapdebug;
746 break;
747
748 case FIONBIO:
749 break;
750
751 case FIOASYNC:
752 s = splimp();
753 mtx_lock(&tp->tap_mtx);
754 if (*(int *)data)
755 tp->tap_flags |= TAP_ASYNC;
756 else
757 tp->tap_flags &= ~TAP_ASYNC;
758 mtx_unlock(&tp->tap_mtx);
759 splx(s);
760 break;
761
762 case FIONREAD:
763 s = splimp();
764 if (ifp->if_snd.ifq_head) {
765 struct mbuf *mb = ifp->if_snd.ifq_head;
766
767 for(*(int *)data = 0;mb != NULL;mb = mb->m_next)
768 *(int *)data += mb->m_len;
769 } else
770 *(int *)data = 0;
771 splx(s);
772 break;
773
774 case FIOSETOWN:
775 return (fsetown(*(int *)data, &tp->tap_sigio));
776
777 case FIOGETOWN:
778 *(int *)data = fgetown(&tp->tap_sigio);
779 return (0);
780
781 /* this is deprecated, FIOSETOWN should be used instead */
782 case TIOCSPGRP:
783 return (fsetown(-(*(int *)data), &tp->tap_sigio));
784
785 /* this is deprecated, FIOGETOWN should be used instead */
786 case TIOCGPGRP:
787 *(int *)data = -fgetown(&tp->tap_sigio);
788 return (0);
789
790 /* VMware/VMnet port ioctl's */
791
792 case SIOCGIFFLAGS: /* get ifnet flags */
793 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
794 break;
795
796 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
797 defined(COMPAT_FREEBSD4)
798 case _IO('V', 0):
799 ival = IOCPARM_IVAL(data);
800 data = (caddr_t)&ival;
801 /* FALLTHROUGH */
802 #endif
803 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
804 f = *(int *)data;
805 f &= 0x0fff;
806 f &= ~IFF_CANTCHANGE;
807 f |= IFF_UP;
808
809 s = splimp();
810 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
811 splx(s);
812 break;
813
814 case OSIOCGIFADDR: /* get MAC address of the remote side */
815 case SIOCGIFADDR:
816 mtx_lock(&tp->tap_mtx);
817 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
818 mtx_unlock(&tp->tap_mtx);
819 break;
820
821 case SIOCSIFADDR: /* set MAC address of the remote side */
822 mtx_lock(&tp->tap_mtx);
823 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
824 mtx_unlock(&tp->tap_mtx);
825 break;
826
827 default:
828 return (ENOTTY);
829 }
830 return (0);
831 } /* tapioctl */
832
833
834 /*
835 * tapread
836 *
837 * the cdevsw read interface - reads a packet at a time, or at
838 * least as much of a packet as can be read
839 */
840 static int
841 tapread(struct cdev *dev, struct uio *uio, int flag)
842 {
843 struct tap_softc *tp = dev->si_drv1;
844 struct ifnet *ifp = tp->tap_ifp;
845 struct mbuf *m = NULL;
846 int error = 0, len, s;
847
848 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, minor(dev));
849
850 mtx_lock(&tp->tap_mtx);
851 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
852 mtx_unlock(&tp->tap_mtx);
853
854 /* Unlocked read. */
855 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
856 ifp->if_xname, minor(dev), tp->tap_flags);
857
858 return (EHOSTDOWN);
859 }
860
861 tp->tap_flags &= ~TAP_RWAIT;
862 mtx_unlock(&tp->tap_mtx);
863
864 /* sleep until we get a packet */
865 do {
866 s = splimp();
867 IF_DEQUEUE(&ifp->if_snd, m);
868 splx(s);
869
870 if (m == NULL) {
871 if (flag & O_NONBLOCK)
872 return (EWOULDBLOCK);
873
874 mtx_lock(&tp->tap_mtx);
875 tp->tap_flags |= TAP_RWAIT;
876 mtx_unlock(&tp->tap_mtx);
877 error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0);
878 if (error)
879 return (error);
880 }
881 } while (m == NULL);
882
883 /* feed packet to bpf */
884 BPF_MTAP(ifp, m);
885
886 /* xfer packet to user space */
887 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
888 len = min(uio->uio_resid, m->m_len);
889 if (len == 0)
890 break;
891
892 error = uiomove(mtod(m, void *), len, uio);
893 m = m_free(m);
894 }
895
896 if (m != NULL) {
897 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
898 minor(dev));
899 m_freem(m);
900 }
901
902 return (error);
903 } /* tapread */
904
905
906 /*
907 * tapwrite
908 *
909 * the cdevsw write interface - an atomic write is a packet - or else!
910 */
911 static int
912 tapwrite(struct cdev *dev, struct uio *uio, int flag)
913 {
914 struct ether_header *eh;
915 struct tap_softc *tp = dev->si_drv1;
916 struct ifnet *ifp = tp->tap_ifp;
917 struct mbuf *m;
918
919 TAPDEBUG("%s writting, minor = %#x\n",
920 ifp->if_xname, minor(dev));
921
922 if (uio->uio_resid == 0)
923 return (0);
924
925 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
926 TAPDEBUG("%s invalid packet len = %d, minor = %#x\n",
927 ifp->if_xname, uio->uio_resid, minor(dev));
928
929 return (EIO);
930 }
931
932 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN,
933 M_PKTHDR)) == NULL) {
934 ifp->if_ierrors ++;
935 return (ENOBUFS);
936 }
937
938 m->m_pkthdr.rcvif = ifp;
939
940 /*
941 * Only pass a unicast frame to ether_input(), if it would actually
942 * have been received by non-virtual hardware.
943 */
944 if (m->m_len < sizeof(struct ether_header)) {
945 m_freem(m);
946 return (0);
947 }
948 eh = mtod(m, struct ether_header *);
949
950 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
951 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
952 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
953 m_freem(m);
954 return (0);
955 }
956
957 /* Pass packet up to parent. */
958 (*ifp->if_input)(ifp, m);
959 ifp->if_ipackets ++; /* ibytes are counted in parent */
960
961 return (0);
962 } /* tapwrite */
963
964
965 /*
966 * tappoll
967 *
968 * the poll interface, this is only useful on reads
969 * really. the write detect always returns true, write never blocks
970 * anyway, it either accepts the packet or drops it
971 */
972 static int
973 tappoll(struct cdev *dev, int events, struct thread *td)
974 {
975 struct tap_softc *tp = dev->si_drv1;
976 struct ifnet *ifp = tp->tap_ifp;
977 int s, revents = 0;
978
979 TAPDEBUG("%s polling, minor = %#x\n",
980 ifp->if_xname, minor(dev));
981
982 s = splimp();
983 if (events & (POLLIN | POLLRDNORM)) {
984 if (ifp->if_snd.ifq_len > 0) {
985 TAPDEBUG("%s have data in queue. len = %d, " \
986 "minor = %#x\n", ifp->if_xname,
987 ifp->if_snd.ifq_len, minor(dev));
988
989 revents |= (events & (POLLIN | POLLRDNORM));
990 } else {
991 TAPDEBUG("%s waiting for data, minor = %#x\n",
992 ifp->if_xname, minor(dev));
993
994 selrecord(td, &tp->tap_rsel);
995 }
996 }
997
998 if (events & (POLLOUT | POLLWRNORM))
999 revents |= (events & (POLLOUT | POLLWRNORM));
1000
1001 splx(s);
1002 return (revents);
1003 } /* tappoll */
1004
1005
1006 /*
1007 * tap_kqfilter
1008 *
1009 * support for kevent() system call
1010 */
1011 static int
1012 tapkqfilter(struct cdev *dev, struct knote *kn)
1013 {
1014 int s;
1015 struct tap_softc *tp = dev->si_drv1;
1016 struct ifnet *ifp = tp->tap_ifp;
1017
1018 s = splimp();
1019 switch (kn->kn_filter) {
1020 case EVFILT_READ:
1021 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1022 ifp->if_xname, minor(dev));
1023 kn->kn_fop = &tap_read_filterops;
1024 break;
1025
1026 case EVFILT_WRITE:
1027 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1028 ifp->if_xname, minor(dev));
1029 kn->kn_fop = &tap_write_filterops;
1030 break;
1031
1032 default:
1033 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1034 ifp->if_xname, minor(dev));
1035 splx(s);
1036 return (EINVAL);
1037 /* NOT REACHED */
1038 }
1039 splx(s);
1040
1041 kn->kn_hook = (caddr_t) dev;
1042 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1043
1044 return (0);
1045 } /* tapkqfilter */
1046
1047
1048 /*
1049 * tap_kqread
1050 *
1051 * Return true if there is data in the interface queue
1052 */
1053 static int
1054 tapkqread(struct knote *kn, long hint)
1055 {
1056 int ret, s;
1057 struct cdev *dev = (struct cdev *)(kn->kn_hook);
1058 struct tap_softc *tp = dev->si_drv1;
1059 struct ifnet *ifp = tp->tap_ifp;
1060
1061 s = splimp();
1062 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1063 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1064 ifp->if_xname, ifp->if_snd.ifq_len, minor(dev));
1065 ret = 1;
1066 } else {
1067 TAPDEBUG("%s waiting for data, minor = %#x\n",
1068 ifp->if_xname, minor(dev));
1069 ret = 0;
1070 }
1071 splx(s);
1072
1073 return (ret);
1074 } /* tapkqread */
1075
1076
1077 /*
1078 * tap_kqwrite
1079 *
1080 * Always can write. Return the MTU in kn->data
1081 */
1082 static int
1083 tapkqwrite(struct knote *kn, long hint)
1084 {
1085 int s;
1086 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1;
1087 struct ifnet *ifp = tp->tap_ifp;
1088
1089 s = splimp();
1090 kn->kn_data = ifp->if_mtu;
1091 splx(s);
1092
1093 return (1);
1094 } /* tapkqwrite */
1095
1096
1097 static void
1098 tapkqdetach(struct knote *kn)
1099 {
1100 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1;
1101
1102 knlist_remove(&tp->tap_rsel.si_note, kn, 0);
1103 } /* tapkqdetach */
1104
Cache object: 0957058ba322ce6509ad5bf83bea1d89
|