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: releng/7.4/sys/net/if_tap.c 213727 2010-10-12 16:09:08Z jhb $
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,
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 }
196
197 tapcreate(dev);
198 return (0);
199 }
200
201 /* vmnet devices are tap devices in disguise */
202 static int
203 vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
204 {
205 return tap_clone_create(ifc, unit, params);
206 }
207
208 static void
209 tap_destroy(struct tap_softc *tp)
210 {
211 struct ifnet *ifp = tp->tap_ifp;
212
213 /* Unlocked read. */
214 KASSERT(!(tp->tap_flags & TAP_OPEN),
215 ("%s flags is out of sync", ifp->if_xname));
216
217 knlist_destroy(&tp->tap_rsel.si_note);
218 destroy_dev(tp->tap_dev);
219 ether_ifdetach(ifp);
220 if_free_type(ifp, IFT_ETHER);
221
222 mtx_destroy(&tp->tap_mtx);
223 free(tp, M_TAP);
224 }
225
226 static void
227 tap_clone_destroy(struct ifnet *ifp)
228 {
229 struct tap_softc *tp = ifp->if_softc;
230
231 mtx_lock(&tapmtx);
232 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
233 mtx_unlock(&tapmtx);
234 tap_destroy(tp);
235 }
236
237 /* vmnet devices are tap devices in disguise */
238 static void
239 vmnet_clone_destroy(struct ifnet *ifp)
240 {
241 tap_clone_destroy(ifp);
242 }
243
244 /*
245 * tapmodevent
246 *
247 * module event handler
248 */
249 static int
250 tapmodevent(module_t mod, int type, void *data)
251 {
252 static eventhandler_tag eh_tag = NULL;
253 struct tap_softc *tp = NULL;
254 struct ifnet *ifp = NULL;
255
256 switch (type) {
257 case MOD_LOAD:
258
259 /* intitialize device */
260
261 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
262 SLIST_INIT(&taphead);
263
264 clone_setup(&tapclones);
265 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
266 if (eh_tag == NULL) {
267 clone_cleanup(&tapclones);
268 mtx_destroy(&tapmtx);
269 return (ENOMEM);
270 }
271 if_clone_attach(&tap_cloner);
272 if_clone_attach(&vmnet_cloner);
273 return (0);
274
275 case MOD_UNLOAD:
276 /*
277 * The EBUSY algorithm here can't quite atomically
278 * guarantee that this is race-free since we have to
279 * release the tap mtx to deregister the clone handler.
280 */
281 mtx_lock(&tapmtx);
282 SLIST_FOREACH(tp, &taphead, tap_next) {
283 mtx_lock(&tp->tap_mtx);
284 if (tp->tap_flags & TAP_OPEN) {
285 mtx_unlock(&tp->tap_mtx);
286 mtx_unlock(&tapmtx);
287 return (EBUSY);
288 }
289 mtx_unlock(&tp->tap_mtx);
290 }
291 mtx_unlock(&tapmtx);
292
293 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
294 if_clone_detach(&tap_cloner);
295 if_clone_detach(&vmnet_cloner);
296 drain_dev_clone_events();
297
298 mtx_lock(&tapmtx);
299 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
300 SLIST_REMOVE_HEAD(&taphead, tap_next);
301 mtx_unlock(&tapmtx);
302
303 ifp = tp->tap_ifp;
304
305 TAPDEBUG("detaching %s\n", ifp->if_xname);
306
307 tap_destroy(tp);
308 mtx_lock(&tapmtx);
309 }
310 mtx_unlock(&tapmtx);
311 clone_cleanup(&tapclones);
312
313 mtx_destroy(&tapmtx);
314
315 break;
316
317 default:
318 return (EOPNOTSUPP);
319 }
320
321 return (0);
322 } /* tapmodevent */
323
324
325 /*
326 * DEVFS handler
327 *
328 * We need to support two kind of devices - tap and vmnet
329 */
330 static void
331 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
332 {
333 char devname[SPECNAMELEN + 1];
334 int i, unit, append_unit;
335 int extra;
336
337 if (*dev != NULL)
338 return;
339
340 if (!tapdclone ||
341 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
342 return;
343
344 unit = 0;
345 append_unit = 0;
346 extra = 0;
347
348 /* We're interested in only tap/vmnet devices. */
349 if (strcmp(name, TAP) == 0) {
350 unit = -1;
351 } else if (strcmp(name, VMNET) == 0) {
352 unit = -1;
353 extra = VMNET_DEV_MASK;
354 } else if (dev_stdclone(name, NULL, TAP, &unit) != 1) {
355 if (dev_stdclone(name, NULL, VMNET, &unit) != 1) {
356 return;
357 } else {
358 extra = VMNET_DEV_MASK;
359 }
360 }
361
362 if (unit == -1)
363 append_unit = 1;
364
365 /* find any existing device, or allocate new unit number */
366 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
367 if (i) {
368 if (append_unit) {
369 /*
370 * We were passed 'tun' or 'tap', with no unit specified
371 * so we'll need to append it now.
372 */
373 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
374 unit);
375 name = devname;
376 }
377
378 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw,
379 unit2minor(unit | extra),
380 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
381 }
382
383 if_clone_create(name, namelen, NULL);
384 } /* tapclone */
385
386
387 /*
388 * tapcreate
389 *
390 * to create interface
391 */
392 static void
393 tapcreate(struct cdev *dev)
394 {
395 struct ifnet *ifp = NULL;
396 struct tap_softc *tp = NULL;
397 unsigned short macaddr_hi;
398 uint32_t macaddr_mid;
399 int unit;
400 char *name = NULL;
401 u_char eaddr[6];
402
403 dev->si_flags &= ~SI_CHEAPCLONE;
404
405 /* allocate driver storage and create device */
406 MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
407 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
408 mtx_lock(&tapmtx);
409 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
410 mtx_unlock(&tapmtx);
411
412 unit = dev2unit(dev);
413
414 /* select device: tap or vmnet */
415 if (unit & VMNET_DEV_MASK) {
416 name = VMNET;
417 tp->tap_flags |= TAP_VMNET;
418 } else
419 name = TAP;
420
421 unit &= TAPMAXUNIT;
422
423 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, minor(dev));
424
425 /* generate fake MAC address: 00 bd xx xx xx unit_no */
426 macaddr_hi = htons(0x00bd);
427 macaddr_mid = (uint32_t) ticks;
428 bcopy(&macaddr_hi, eaddr, sizeof(short));
429 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
430 eaddr[5] = (u_char)unit;
431
432 /* fill the rest and attach interface */
433 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
434 if (ifp == NULL)
435 panic("%s%d: can not if_alloc()", name, unit);
436 ifp->if_softc = tp;
437 if_initname(ifp, name, unit);
438 ifp->if_init = tapifinit;
439 ifp->if_start = tapifstart;
440 ifp->if_ioctl = tapifioctl;
441 ifp->if_mtu = ETHERMTU;
442 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
443 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
444
445 dev->si_drv1 = tp;
446 tp->tap_dev = dev;
447
448 ether_ifattach(ifp, eaddr);
449
450 mtx_lock(&tp->tap_mtx);
451 tp->tap_flags |= TAP_INITED;
452 mtx_unlock(&tp->tap_mtx);
453
454 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
455
456 TAPDEBUG("interface %s is created. minor = %#x\n",
457 ifp->if_xname, minor(dev));
458 } /* tapcreate */
459
460
461 /*
462 * tapopen
463 *
464 * to open tunnel. must be superuser
465 */
466 static int
467 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
468 {
469 struct tap_softc *tp = NULL;
470 struct ifnet *ifp = NULL;
471 int error;
472
473 if (tapuopen == 0) {
474 error = priv_check(td, PRIV_NET_TAP);
475 if (error)
476 return (error);
477 }
478
479 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
480 return (ENXIO);
481
482 tp = dev->si_drv1;
483
484 mtx_lock(&tp->tap_mtx);
485 if (tp->tap_flags & TAP_OPEN) {
486 mtx_unlock(&tp->tap_mtx);
487 return (EBUSY);
488 }
489
490 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
491 tp->tap_pid = td->td_proc->p_pid;
492 tp->tap_flags |= TAP_OPEN;
493 ifp = tp->tap_ifp;
494
495 ifp->if_drv_flags |= IFF_DRV_RUNNING;
496 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
497 if (tapuponopen)
498 ifp->if_flags |= IFF_UP;
499 mtx_unlock(&tp->tap_mtx);
500
501 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, minor(dev));
502
503 return (0);
504 } /* tapopen */
505
506
507 /*
508 * tapclose
509 *
510 * close the device - mark i/f down & delete routing info
511 */
512 static int
513 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
514 {
515 struct ifaddr *ifa;
516 struct tap_softc *tp = dev->si_drv1;
517 struct ifnet *ifp = tp->tap_ifp;
518
519 /* junk all pending output */
520 mtx_lock(&tp->tap_mtx);
521 IF_DRAIN(&ifp->if_snd);
522
523 /*
524 * do not bring the interface down, and do not anything with
525 * interface, if we are in VMnet mode. just close the device.
526 */
527
528 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
529 mtx_unlock(&tp->tap_mtx);
530 if_down(ifp);
531 mtx_lock(&tp->tap_mtx);
532 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
533 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
534 mtx_unlock(&tp->tap_mtx);
535 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
536 rtinit(ifa, (int)RTM_DELETE, 0);
537 }
538 if_purgeaddrs(ifp);
539 mtx_lock(&tp->tap_mtx);
540 }
541 }
542
543 funsetown(&tp->tap_sigio);
544 selwakeuppri(&tp->tap_rsel, PZERO+1);
545 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
546
547 tp->tap_flags &= ~TAP_OPEN;
548 tp->tap_pid = 0;
549 mtx_unlock(&tp->tap_mtx);
550
551 TAPDEBUG("%s is closed. minor = %#x\n",
552 ifp->if_xname, minor(dev));
553
554 return (0);
555 } /* tapclose */
556
557
558 /*
559 * tapifinit
560 *
561 * network interface initialization function
562 */
563 static void
564 tapifinit(void *xtp)
565 {
566 struct tap_softc *tp = (struct tap_softc *)xtp;
567 struct ifnet *ifp = tp->tap_ifp;
568
569 TAPDEBUG("initializing %s\n", ifp->if_xname);
570
571 mtx_lock(&tp->tap_mtx);
572 ifp->if_drv_flags |= IFF_DRV_RUNNING;
573 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
574 mtx_unlock(&tp->tap_mtx);
575
576 /* attempt to start output */
577 tapifstart(ifp);
578 } /* tapifinit */
579
580
581 /*
582 * tapifioctl
583 *
584 * Process an ioctl request on network interface
585 */
586 static int
587 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
588 {
589 struct tap_softc *tp = ifp->if_softc;
590 struct ifreq *ifr = (struct ifreq *)data;
591 struct ifstat *ifs = NULL;
592 int dummy;
593
594 switch (cmd) {
595 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
596 case SIOCADDMULTI:
597 case SIOCDELMULTI:
598 break;
599
600 case SIOCSIFMTU:
601 ifp->if_mtu = ifr->ifr_mtu;
602 break;
603
604 case SIOCGIFSTATUS:
605 ifs = (struct ifstat *)data;
606 dummy = strlen(ifs->ascii);
607 mtx_lock(&tp->tap_mtx);
608 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
609 snprintf(ifs->ascii + dummy,
610 sizeof(ifs->ascii) - dummy,
611 "\tOpened by PID %d\n", tp->tap_pid);
612 mtx_unlock(&tp->tap_mtx);
613 break;
614
615 default:
616 return (ether_ioctl(ifp, cmd, data));
617 /* NOT REACHED */
618 }
619
620 return (0);
621 } /* tapifioctl */
622
623
624 /*
625 * tapifstart
626 *
627 * queue packets from higher level ready to put out
628 */
629 static void
630 tapifstart(struct ifnet *ifp)
631 {
632 struct tap_softc *tp = ifp->if_softc;
633
634 TAPDEBUG("%s starting\n", ifp->if_xname);
635
636 /*
637 * do not junk pending output if we are in VMnet mode.
638 * XXX: can this do any harm because of queue overflow?
639 */
640
641 mtx_lock(&tp->tap_mtx);
642 if (((tp->tap_flags & TAP_VMNET) == 0) &&
643 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
644 struct mbuf *m;
645
646 /* Unlocked read. */
647 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
648 tp->tap_flags);
649
650 for (;;) {
651 IF_DEQUEUE(&ifp->if_snd, m);
652 if (m != NULL) {
653 m_freem(m);
654 ifp->if_oerrors++;
655 } else
656 break;
657 }
658 mtx_unlock(&tp->tap_mtx);
659
660 return;
661 }
662
663 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
664
665 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
666 if (tp->tap_flags & TAP_RWAIT) {
667 tp->tap_flags &= ~TAP_RWAIT;
668 wakeup(tp);
669 }
670
671 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
672 mtx_unlock(&tp->tap_mtx);
673 pgsigio(&tp->tap_sigio, SIGIO, 0);
674 mtx_lock(&tp->tap_mtx);
675 }
676
677 selwakeuppri(&tp->tap_rsel, PZERO+1);
678 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
679 ifp->if_opackets ++; /* obytes are counted in ether_output */
680 }
681
682 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
683 mtx_unlock(&tp->tap_mtx);
684 } /* tapifstart */
685
686
687 /*
688 * tapioctl
689 *
690 * the cdevsw interface is now pretty minimal
691 */
692 static int
693 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
694 {
695 struct tap_softc *tp = dev->si_drv1;
696 struct ifnet *ifp = tp->tap_ifp;
697 struct tapinfo *tapp = NULL;
698 int f;
699 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
700 defined(COMPAT_FREEBSD4)
701 int ival;
702 #endif
703
704 switch (cmd) {
705 case TAPSIFINFO:
706 tapp = (struct tapinfo *)data;
707 mtx_lock(&tp->tap_mtx);
708 ifp->if_mtu = tapp->mtu;
709 ifp->if_type = tapp->type;
710 ifp->if_baudrate = tapp->baudrate;
711 mtx_unlock(&tp->tap_mtx);
712 break;
713
714 case TAPGIFINFO:
715 tapp = (struct tapinfo *)data;
716 mtx_lock(&tp->tap_mtx);
717 tapp->mtu = ifp->if_mtu;
718 tapp->type = ifp->if_type;
719 tapp->baudrate = ifp->if_baudrate;
720 mtx_unlock(&tp->tap_mtx);
721 break;
722
723 case TAPSDEBUG:
724 tapdebug = *(int *)data;
725 break;
726
727 case TAPGDEBUG:
728 *(int *)data = tapdebug;
729 break;
730
731 case TAPGIFNAME: {
732 struct ifreq *ifr = (struct ifreq *) data;
733
734 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
735 } break;
736
737 case FIONBIO:
738 break;
739
740 case FIOASYNC:
741 mtx_lock(&tp->tap_mtx);
742 if (*(int *)data)
743 tp->tap_flags |= TAP_ASYNC;
744 else
745 tp->tap_flags &= ~TAP_ASYNC;
746 mtx_unlock(&tp->tap_mtx);
747 break;
748
749 case FIONREAD:
750 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
751 struct mbuf *mb;
752
753 IFQ_LOCK(&ifp->if_snd);
754 IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
755 for (*(int *)data = 0; mb != NULL;
756 mb = mb->m_next)
757 *(int *)data += mb->m_len;
758 IFQ_UNLOCK(&ifp->if_snd);
759 } else
760 *(int *)data = 0;
761 break;
762
763 case FIOSETOWN:
764 return (fsetown(*(int *)data, &tp->tap_sigio));
765
766 case FIOGETOWN:
767 *(int *)data = fgetown(&tp->tap_sigio);
768 return (0);
769
770 /* this is deprecated, FIOSETOWN should be used instead */
771 case TIOCSPGRP:
772 return (fsetown(-(*(int *)data), &tp->tap_sigio));
773
774 /* this is deprecated, FIOGETOWN should be used instead */
775 case TIOCGPGRP:
776 *(int *)data = -fgetown(&tp->tap_sigio);
777 return (0);
778
779 /* VMware/VMnet port ioctl's */
780
781 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
782 defined(COMPAT_FREEBSD4)
783 case _IO('V', 0):
784 ival = IOCPARM_IVAL(data);
785 data = (caddr_t)&ival;
786 /* FALLTHROUGH */
787 #endif
788 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
789 f = *(int *)data;
790 f &= 0x0fff;
791 f &= ~IFF_CANTCHANGE;
792 f |= IFF_UP;
793
794 mtx_lock(&tp->tap_mtx);
795 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
796 mtx_unlock(&tp->tap_mtx);
797 break;
798
799 case OSIOCGIFADDR: /* get MAC address of the remote side */
800 case SIOCGIFADDR:
801 mtx_lock(&tp->tap_mtx);
802 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
803 mtx_unlock(&tp->tap_mtx);
804 break;
805
806 case SIOCSIFADDR: /* set MAC address of the remote side */
807 mtx_lock(&tp->tap_mtx);
808 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
809 mtx_unlock(&tp->tap_mtx);
810 break;
811
812 default:
813 return (ENOTTY);
814 }
815 return (0);
816 } /* tapioctl */
817
818
819 /*
820 * tapread
821 *
822 * the cdevsw read interface - reads a packet at a time, or at
823 * least as much of a packet as can be read
824 */
825 static int
826 tapread(struct cdev *dev, struct uio *uio, int flag)
827 {
828 struct tap_softc *tp = dev->si_drv1;
829 struct ifnet *ifp = tp->tap_ifp;
830 struct mbuf *m = NULL;
831 int error = 0, len;
832
833 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, minor(dev));
834
835 mtx_lock(&tp->tap_mtx);
836 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
837 mtx_unlock(&tp->tap_mtx);
838
839 /* Unlocked read. */
840 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
841 ifp->if_xname, minor(dev), tp->tap_flags);
842
843 return (EHOSTDOWN);
844 }
845
846 tp->tap_flags &= ~TAP_RWAIT;
847
848 /* sleep until we get a packet */
849 do {
850 IF_DEQUEUE(&ifp->if_snd, m);
851
852 if (m == NULL) {
853 if (flag & O_NONBLOCK) {
854 mtx_unlock(&tp->tap_mtx);
855 return (EWOULDBLOCK);
856 }
857
858 tp->tap_flags |= TAP_RWAIT;
859 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
860 "taprd", 0);
861 if (error) {
862 mtx_unlock(&tp->tap_mtx);
863 return (error);
864 }
865 }
866 } while (m == NULL);
867 mtx_unlock(&tp->tap_mtx);
868
869 /* feed packet to bpf */
870 BPF_MTAP(ifp, m);
871
872 /* xfer packet to user space */
873 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
874 len = min(uio->uio_resid, m->m_len);
875 if (len == 0)
876 break;
877
878 error = uiomove(mtod(m, void *), len, uio);
879 m = m_free(m);
880 }
881
882 if (m != NULL) {
883 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
884 minor(dev));
885 m_freem(m);
886 }
887
888 return (error);
889 } /* tapread */
890
891
892 /*
893 * tapwrite
894 *
895 * the cdevsw write interface - an atomic write is a packet - or else!
896 */
897 static int
898 tapwrite(struct cdev *dev, struct uio *uio, int flag)
899 {
900 struct ether_header *eh;
901 struct tap_softc *tp = dev->si_drv1;
902 struct ifnet *ifp = tp->tap_ifp;
903 struct mbuf *m;
904
905 TAPDEBUG("%s writting, minor = %#x\n",
906 ifp->if_xname, minor(dev));
907
908 if (uio->uio_resid == 0)
909 return (0);
910
911 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
912 TAPDEBUG("%s invalid packet len = %d, minor = %#x\n",
913 ifp->if_xname, uio->uio_resid, minor(dev));
914
915 return (EIO);
916 }
917
918 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN,
919 M_PKTHDR)) == NULL) {
920 ifp->if_ierrors ++;
921 return (ENOBUFS);
922 }
923
924 m->m_pkthdr.rcvif = ifp;
925
926 /*
927 * Only pass a unicast frame to ether_input(), if it would actually
928 * have been received by non-virtual hardware.
929 */
930 if (m->m_len < sizeof(struct ether_header)) {
931 m_freem(m);
932 return (0);
933 }
934 eh = mtod(m, struct ether_header *);
935
936 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
937 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
938 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
939 m_freem(m);
940 return (0);
941 }
942
943 /* Pass packet up to parent. */
944 (*ifp->if_input)(ifp, m);
945 ifp->if_ipackets ++; /* ibytes are counted in parent */
946
947 return (0);
948 } /* tapwrite */
949
950
951 /*
952 * tappoll
953 *
954 * the poll interface, this is only useful on reads
955 * really. the write detect always returns true, write never blocks
956 * anyway, it either accepts the packet or drops it
957 */
958 static int
959 tappoll(struct cdev *dev, int events, struct thread *td)
960 {
961 struct tap_softc *tp = dev->si_drv1;
962 struct ifnet *ifp = tp->tap_ifp;
963 int revents = 0;
964
965 TAPDEBUG("%s polling, minor = %#x\n",
966 ifp->if_xname, minor(dev));
967
968 if (events & (POLLIN | POLLRDNORM)) {
969 IFQ_LOCK(&ifp->if_snd);
970 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
971 TAPDEBUG("%s have data in queue. len = %d, " \
972 "minor = %#x\n", ifp->if_xname,
973 ifp->if_snd.ifq_len, minor(dev));
974
975 revents |= (events & (POLLIN | POLLRDNORM));
976 } else {
977 TAPDEBUG("%s waiting for data, minor = %#x\n",
978 ifp->if_xname, minor(dev));
979
980 selrecord(td, &tp->tap_rsel);
981 }
982 IFQ_UNLOCK(&ifp->if_snd);
983 }
984
985 if (events & (POLLOUT | POLLWRNORM))
986 revents |= (events & (POLLOUT | POLLWRNORM));
987
988 return (revents);
989 } /* tappoll */
990
991
992 /*
993 * tap_kqfilter
994 *
995 * support for kevent() system call
996 */
997 static int
998 tapkqfilter(struct cdev *dev, struct knote *kn)
999 {
1000 struct tap_softc *tp = dev->si_drv1;
1001 struct ifnet *ifp = tp->tap_ifp;
1002
1003 switch (kn->kn_filter) {
1004 case EVFILT_READ:
1005 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1006 ifp->if_xname, minor(dev));
1007 kn->kn_fop = &tap_read_filterops;
1008 break;
1009
1010 case EVFILT_WRITE:
1011 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1012 ifp->if_xname, minor(dev));
1013 kn->kn_fop = &tap_write_filterops;
1014 break;
1015
1016 default:
1017 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1018 ifp->if_xname, minor(dev));
1019 return (EINVAL);
1020 /* NOT REACHED */
1021 }
1022
1023 kn->kn_hook = tp;
1024 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1025
1026 return (0);
1027 } /* tapkqfilter */
1028
1029
1030 /*
1031 * tap_kqread
1032 *
1033 * Return true if there is data in the interface queue
1034 */
1035 static int
1036 tapkqread(struct knote *kn, long hint)
1037 {
1038 int ret;
1039 struct tap_softc *tp = kn->kn_hook;
1040 struct cdev *dev = tp->tap_dev;
1041 struct ifnet *ifp = tp->tap_ifp;
1042
1043 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1044 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1045 ifp->if_xname, ifp->if_snd.ifq_len, minor(dev));
1046 ret = 1;
1047 } else {
1048 TAPDEBUG("%s waiting for data, minor = %#x\n",
1049 ifp->if_xname, minor(dev));
1050 ret = 0;
1051 }
1052
1053 return (ret);
1054 } /* tapkqread */
1055
1056
1057 /*
1058 * tap_kqwrite
1059 *
1060 * Always can write. Return the MTU in kn->data
1061 */
1062 static int
1063 tapkqwrite(struct knote *kn, long hint)
1064 {
1065 struct tap_softc *tp = kn->kn_hook;
1066 struct ifnet *ifp = tp->tap_ifp;
1067
1068 kn->kn_data = ifp->if_mtu;
1069
1070 return (1);
1071 } /* tapkqwrite */
1072
1073
1074 static void
1075 tapkqdetach(struct knote *kn)
1076 {
1077 struct tap_softc *tp = kn->kn_hook;
1078
1079 knlist_remove(&tp->tap_rsel.si_note, kn, 0);
1080 } /* tapkqdetach */
1081
Cache object: a092ea6e60df6e48754a0b63120ecf8c
|