if_l2tp.c revision 1.10 1 /* $NetBSD: if_l2tp.c,v 1.10 2017/04/20 09:11:58 knakahara Exp $ */
2
3 /*
4 * Copyright (c) 2017 Internet Initiative Japan Inc.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26 * POSSIBILITY OF SUCH DAMAGE.
27 */
28
29 /*
30 * L2TPv3 kernel interface
31 */
32
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: if_l2tp.c,v 1.10 2017/04/20 09:11:58 knakahara Exp $");
35
36 #ifdef _KERNEL_OPT
37 #include "opt_inet.h"
38 #endif
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/mbuf.h>
44 #include <sys/socket.h>
45 #include <sys/sockio.h>
46 #include <sys/errno.h>
47 #include <sys/ioctl.h>
48 #include <sys/time.h>
49 #include <sys/syslog.h>
50 #include <sys/proc.h>
51 #include <sys/conf.h>
52 #include <sys/kauth.h>
53 #include <sys/cpu.h>
54 #include <sys/cprng.h>
55 #include <sys/intr.h>
56 #include <sys/kmem.h>
57 #include <sys/mutex.h>
58 #include <sys/atomic.h>
59 #include <sys/pserialize.h>
60 #include <sys/device.h>
61 #include <sys/module.h>
62
63 #include <net/if.h>
64 #include <net/if_dl.h>
65 #include <net/if_ether.h>
66 #include <net/if_types.h>
67 #include <net/netisr.h>
68 #include <net/route.h>
69 #include <net/bpf.h>
70 #include <net/if_vlanvar.h>
71
72 #include <netinet/in.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/ip_encap.h>
76 #ifdef INET
77 #include <netinet/in_var.h>
78 #include <netinet/in_l2tp.h>
79 #endif /* INET */
80 #ifdef INET6
81 #include <netinet6/in6_l2tp.h>
82 #endif
83
84 #include <net/if_l2tp.h>
85
86 #include <net/if_vlanvar.h>
87
88 /* TODO: IP_TCPMSS support */
89 #undef IP_TCPMSS
90 #ifdef IP_TCPMSS
91 #include <netinet/ip_tcpmss.h>
92 #endif
93
94 #include <net/bpf.h>
95 #include <net/net_osdep.h>
96
97 /*
98 * l2tp global variable definitions
99 */
100 LIST_HEAD(l2tp_sclist, l2tp_softc);
101 static struct {
102 struct l2tp_sclist list;
103 kmutex_t lock;
104 } l2tp_softcs __cacheline_aligned;
105
106
107 #if !defined(L2TP_ID_HASH_SIZE)
108 #define L2TP_ID_HASH_SIZE 64
109 #endif
110 static struct {
111 kmutex_t lock;
112 struct pslist_head *lists;
113 u_long mask;
114 } l2tp_hash __cacheline_aligned = {
115 .lists = NULL,
116 };
117
118 pserialize_t l2tp_psz __read_mostly;
119 struct psref_class *lv_psref_class __read_mostly;
120
121 static void l2tp_ro_init_pc(void *, void *, struct cpu_info *);
122 static void l2tp_ro_fini_pc(void *, void *, struct cpu_info *);
123
124 static int l2tp_clone_create(struct if_clone *, int);
125 static int l2tp_clone_destroy(struct ifnet *);
126
127 struct if_clone l2tp_cloner =
128 IF_CLONE_INITIALIZER("l2tp", l2tp_clone_create, l2tp_clone_destroy);
129
130 static int l2tp_output(struct ifnet *, struct mbuf *,
131 const struct sockaddr *, const struct rtentry *);
132 static void l2tpintr(struct l2tp_variant *);
133
134 static void l2tp_hash_init(void);
135 static int l2tp_hash_fini(void);
136
137 static void l2tp_start(struct ifnet *);
138 static int l2tp_transmit(struct ifnet *, struct mbuf *);
139
140 static int l2tp_set_tunnel(struct ifnet *, struct sockaddr *,
141 struct sockaddr *);
142 static void l2tp_delete_tunnel(struct ifnet *);
143
144 static int id_hash_func(uint32_t, u_long);
145
146 static void l2tp_variant_update(struct l2tp_softc *, struct l2tp_variant *);
147 static int l2tp_set_session(struct l2tp_softc *, uint32_t, uint32_t);
148 static int l2tp_clear_session(struct l2tp_softc *);
149 static int l2tp_set_cookie(struct l2tp_softc *, uint64_t, u_int, uint64_t, u_int);
150 static void l2tp_clear_cookie(struct l2tp_softc *);
151 static void l2tp_set_state(struct l2tp_softc *, int);
152 static int l2tp_encap_attach(struct l2tp_variant *);
153 static int l2tp_encap_detach(struct l2tp_variant *);
154
155 #ifndef MAX_L2TP_NEST
156 /*
157 * This macro controls the upper limitation on nesting of l2tp tunnels.
158 * Since, setting a large value to this macro with a careless configuration
159 * may introduce system crash, we don't allow any nestings by default.
160 * If you need to configure nested l2tp tunnels, you can define this macro
161 * in your kernel configuration file. However, if you do so, please be
162 * careful to configure the tunnels so that it won't make a loop.
163 */
164 /*
165 * XXX
166 * Currently, if in_l2tp_output recursively calls, it causes locking against
167 * myself of struct l2tp_ro->lr_lock. So, nested l2tp tunnels is prohibited.
168 */
169 #define MAX_L2TP_NEST 0
170 #endif
171
172 static int max_l2tp_nesting = MAX_L2TP_NEST;
173
174 /* ARGSUSED */
175 void
176 l2tpattach(int count)
177 {
178 /*
179 * Nothing to do here, initialization is handled by the
180 * module initialization code in l2tpinit() below).
181 */
182 }
183
184 static void
185 l2tpinit(void)
186 {
187
188 mutex_init(&l2tp_softcs.lock, MUTEX_DEFAULT, IPL_NONE);
189 LIST_INIT(&l2tp_softcs.list);
190
191 mutex_init(&l2tp_hash.lock, MUTEX_DEFAULT, IPL_NONE);
192 l2tp_psz = pserialize_create();
193 lv_psref_class = psref_class_create("l2tpvar", IPL_SOFTNET);
194 if_clone_attach(&l2tp_cloner);
195
196 l2tp_hash_init();
197 }
198
199 static int
200 l2tpdetach(void)
201 {
202 int error;
203
204 mutex_enter(&l2tp_softcs.lock);
205 if (!LIST_EMPTY(&l2tp_softcs.list)) {
206 mutex_exit(&l2tp_softcs.lock);
207 return EBUSY;
208 }
209 mutex_exit(&l2tp_softcs.lock);
210
211 error = l2tp_hash_fini();
212 if (error)
213 return error;
214
215 if_clone_detach(&l2tp_cloner);
216 psref_class_destroy(lv_psref_class);
217 pserialize_destroy(l2tp_psz);
218 mutex_destroy(&l2tp_hash.lock);
219
220 mutex_destroy(&l2tp_softcs.lock);
221
222 return error;
223 }
224
225 static int
226 l2tp_clone_create(struct if_clone *ifc, int unit)
227 {
228 struct l2tp_softc *sc;
229 struct l2tp_variant *var;
230
231 sc = kmem_zalloc(sizeof(struct l2tp_softc), KM_SLEEP);
232 var = kmem_zalloc(sizeof(struct l2tp_variant), KM_SLEEP);
233
234 var->lv_softc = sc;
235 var->lv_state = L2TP_STATE_DOWN;
236 var->lv_use_cookie = L2TP_COOKIE_OFF;
237 psref_target_init(&var->lv_psref, lv_psref_class);
238
239 sc->l2tp_var = var;
240 mutex_init(&sc->l2tp_lock, MUTEX_DEFAULT, IPL_NONE);
241 PSLIST_ENTRY_INIT(sc, l2tp_hash);
242
243 if_initname(&sc->l2tp_ec.ec_if, ifc->ifc_name, unit);
244
245 l2tpattach0(sc);
246
247 sc->l2tp_ro_percpu = percpu_alloc(sizeof(struct l2tp_ro));
248 KASSERTMSG(sc->l2tp_ro_percpu != NULL,
249 "failed to allocate sc->l2tp_ro_percpu");
250 percpu_foreach(sc->l2tp_ro_percpu, l2tp_ro_init_pc, NULL);
251
252 mutex_enter(&l2tp_softcs.lock);
253 LIST_INSERT_HEAD(&l2tp_softcs.list, sc, l2tp_list);
254 mutex_exit(&l2tp_softcs.lock);
255
256 return (0);
257 }
258
259 void
260 l2tpattach0(struct l2tp_softc *sc)
261 {
262
263 sc->l2tp_ec.ec_if.if_addrlen = 0;
264 sc->l2tp_ec.ec_if.if_mtu = L2TP_MTU;
265 sc->l2tp_ec.ec_if.if_flags = IFF_POINTOPOINT|IFF_MULTICAST|IFF_SIMPLEX;
266 sc->l2tp_ec.ec_if.if_extflags = IFEF_OUTPUT_MPSAFE|IFEF_START_MPSAFE;
267 sc->l2tp_ec.ec_if.if_ioctl = l2tp_ioctl;
268 sc->l2tp_ec.ec_if.if_output = l2tp_output;
269 sc->l2tp_ec.ec_if.if_type = IFT_L2TP;
270 sc->l2tp_ec.ec_if.if_dlt = DLT_NULL;
271 sc->l2tp_ec.ec_if.if_start = l2tp_start;
272 sc->l2tp_ec.ec_if.if_transmit = l2tp_transmit;
273 sc->l2tp_ec.ec_if._if_input = ether_input;
274 IFQ_SET_READY(&sc->l2tp_ec.ec_if.if_snd);
275 if_attach(&sc->l2tp_ec.ec_if);
276 if_alloc_sadl(&sc->l2tp_ec.ec_if);
277 bpf_attach(&sc->l2tp_ec.ec_if, DLT_EN10MB, sizeof(struct ether_header));
278 }
279
280 void
281 l2tp_ro_init_pc(void *p, void *arg __unused, struct cpu_info *ci __unused)
282 {
283 struct l2tp_ro *lro = p;
284
285 mutex_init(&lro->lr_lock, MUTEX_DEFAULT, IPL_NONE);
286 }
287
288 void
289 l2tp_ro_fini_pc(void *p, void *arg __unused, struct cpu_info *ci __unused)
290 {
291 struct l2tp_ro *lro = p;
292
293 rtcache_free(&lro->lr_ro);
294
295 mutex_destroy(&lro->lr_lock);
296 }
297
298 static int
299 l2tp_clone_destroy(struct ifnet *ifp)
300 {
301 struct l2tp_variant *var;
302 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
303 l2tp_ec.ec_if);
304
305 l2tp_clear_session(sc);
306 l2tp_delete_tunnel(&sc->l2tp_ec.ec_if);
307 /*
308 * To avoid for l2tp_transmit() to access sc->l2tp_var after free it.
309 */
310 mutex_enter(&sc->l2tp_lock);
311 var = sc->l2tp_var;
312 l2tp_variant_update(sc, NULL);
313 mutex_exit(&sc->l2tp_lock);
314
315 mutex_enter(&l2tp_softcs.lock);
316 LIST_REMOVE(sc, l2tp_list);
317 mutex_exit(&l2tp_softcs.lock);
318
319 bpf_detach(ifp);
320
321 if_detach(ifp);
322
323 percpu_foreach(sc->l2tp_ro_percpu, l2tp_ro_fini_pc, NULL);
324 percpu_free(sc->l2tp_ro_percpu, sizeof(struct l2tp_ro));
325
326 kmem_free(var, sizeof(struct l2tp_variant));
327 mutex_destroy(&sc->l2tp_lock);
328 kmem_free(sc, sizeof(struct l2tp_softc));
329
330 return 0;
331 }
332
333 static int
334 l2tp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
335 const struct rtentry *rt)
336 {
337 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
338 l2tp_ec.ec_if);
339 struct l2tp_variant *var;
340 struct psref psref;
341 int error = 0;
342
343 var = l2tp_getref_variant(sc, &psref);
344 if (var == NULL) {
345 m_freem(m);
346 return ENETDOWN;
347 }
348
349 IFQ_CLASSIFY(&ifp->if_snd, m, dst->sa_family);
350
351 m->m_flags &= ~(M_BCAST|M_MCAST);
352
353 if ((ifp->if_flags & IFF_UP) == 0) {
354 m_freem(m);
355 error = ENETDOWN;
356 goto end;
357 }
358
359 if (var->lv_psrc == NULL || var->lv_pdst == NULL) {
360 m_freem(m);
361 error = ENETDOWN;
362 goto end;
363 }
364
365 /* XXX should we check if our outer source is legal? */
366
367 /* use DLT_NULL encapsulation here to pass inner af type */
368 M_PREPEND(m, sizeof(int), M_DONTWAIT);
369 if (!m) {
370 error = ENOBUFS;
371 goto end;
372 }
373 *mtod(m, int *) = dst->sa_family;
374
375 IFQ_ENQUEUE(&ifp->if_snd, m, error);
376 if (error)
377 goto end;
378
379 /*
380 * direct call to avoid infinite loop at l2tpintr()
381 */
382 l2tpintr(var);
383
384 error = 0;
385
386 end:
387 l2tp_putref_variant(var, &psref);
388 if (error)
389 ifp->if_oerrors++;
390
391 return error;
392 }
393
394 static void
395 l2tpintr(struct l2tp_variant *var)
396 {
397 struct l2tp_softc *sc;
398 struct ifnet *ifp;
399 struct mbuf *m;
400 int error;
401
402 KASSERT(psref_held(&var->lv_psref, lv_psref_class));
403
404 sc = var->lv_softc;
405 ifp = &sc->l2tp_ec.ec_if;
406
407 /* output processing */
408 if (var->lv_my_sess_id == 0 || var->lv_peer_sess_id == 0) {
409 IFQ_PURGE(&ifp->if_snd);
410 return;
411 }
412
413 for (;;) {
414 IFQ_DEQUEUE(&ifp->if_snd, m);
415 if (m == NULL)
416 break;
417 m->m_flags &= ~(M_BCAST|M_MCAST);
418 bpf_mtap(ifp, m);
419 switch (var->lv_psrc->sa_family) {
420 #ifdef INET
421 case AF_INET:
422 error = in_l2tp_output(var, m);
423 break;
424 #endif
425 #ifdef INET6
426 case AF_INET6:
427 error = in6_l2tp_output(var, m);
428 break;
429 #endif
430 default:
431 m_freem(m);
432 error = ENETDOWN;
433 break;
434 }
435
436 if (error)
437 ifp->if_oerrors++;
438 else {
439 ifp->if_opackets++;
440 /*
441 * obytes is incremented at ether_output() or
442 * bridge_enqueue().
443 */
444 }
445 }
446
447 }
448
449 void
450 l2tp_input(struct mbuf *m, struct ifnet *ifp)
451 {
452
453 KASSERT(ifp != NULL);
454
455 if (0 == (mtod(m, u_long) & 0x03)) {
456 /* copy and align head of payload */
457 struct mbuf *m_head;
458 int copy_length;
459
460 #define L2TP_COPY_LENGTH 60
461 #define L2TP_LINK_HDR_ROOM (MHLEN - L2TP_COPY_LENGTH - 4/*round4(2)*/)
462
463 if (m->m_pkthdr.len < L2TP_COPY_LENGTH) {
464 copy_length = m->m_pkthdr.len;
465 } else {
466 copy_length = L2TP_COPY_LENGTH;
467 }
468
469 if (m->m_len < copy_length) {
470 m = m_pullup(m, copy_length);
471 if (m == NULL)
472 return;
473 }
474
475 MGETHDR(m_head, M_DONTWAIT, MT_HEADER);
476 if (m_head == NULL) {
477 m_freem(m);
478 return;
479 }
480 M_COPY_PKTHDR(m_head, m);
481
482 m_head->m_data += 2 /* align */ + L2TP_LINK_HDR_ROOM;
483 memcpy(m_head->m_data, m->m_data, copy_length);
484 m_head->m_len = copy_length;
485 m->m_data += copy_length;
486 m->m_len -= copy_length;
487
488 /* construct chain */
489 if (m->m_len == 0) {
490 m_head->m_next = m_free(m); /* not m_freem */
491 } else {
492 /*
493 * copyed mtag in previous call M_COPY_PKTHDR
494 * but don't delete mtag in case cutt of M_PKTHDR flag
495 */
496 m_tag_delete_chain(m, NULL);
497 m->m_flags &= ~M_PKTHDR;
498 m_head->m_next = m;
499 }
500
501 /* override m */
502 m = m_head;
503 }
504
505 m_set_rcvif(m, ifp);
506
507 /*
508 * bpf_mtap() and ifp->if_ipackets++ is done in if_input()
509 *
510 * obytes is incremented at ether_output() or bridge_enqueue().
511 */
512 if_percpuq_enqueue(ifp->if_percpuq, m);
513 }
514
515 void
516 l2tp_start(struct ifnet *ifp)
517 {
518 struct psref psref;
519 struct l2tp_variant *var;
520 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
521 l2tp_ec.ec_if);
522
523 var = l2tp_getref_variant(sc, &psref);
524 if (var == NULL)
525 return;
526
527 if (var->lv_psrc == NULL || var->lv_pdst == NULL)
528 return;
529
530 l2tpintr(var);
531 l2tp_putref_variant(var, &psref);
532 }
533
534 int
535 l2tp_transmit(struct ifnet *ifp, struct mbuf *m)
536 {
537 int error;
538 struct psref psref;
539 struct l2tp_variant *var;
540 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
541 l2tp_ec.ec_if);
542
543 var = l2tp_getref_variant(sc, &psref);
544 if (var == NULL) {
545 m_freem(m);
546 return ENETDOWN;
547 }
548
549 if (var->lv_psrc == NULL || var->lv_pdst == NULL) {
550 m_freem(m);
551 error = ENETDOWN;
552 goto out;
553 }
554
555 m->m_flags &= ~(M_BCAST|M_MCAST);
556 bpf_mtap(ifp, m);
557 switch (var->lv_psrc->sa_family) {
558 #ifdef INET
559 case AF_INET:
560 error = in_l2tp_output(var, m);
561 break;
562 #endif
563 #ifdef INET6
564 case AF_INET6:
565 error = in6_l2tp_output(var, m);
566 break;
567 #endif
568 default:
569 m_freem(m);
570 error = ENETDOWN;
571 break;
572 }
573
574 if (error)
575 ifp->if_oerrors++;
576 else {
577 ifp->if_opackets++;
578 /*
579 * obytes is incremented at ether_output() or bridge_enqueue().
580 */
581 }
582
583 out:
584 l2tp_putref_variant(var, &psref);
585 return error;
586 }
587
588 /* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
589 int
590 l2tp_ioctl(struct ifnet *ifp, u_long cmd, void *data)
591 {
592 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
593 l2tp_ec.ec_if);
594 struct l2tp_variant *var, *var_tmp;
595 struct ifreq *ifr = data;
596 int error = 0, size;
597 struct sockaddr *dst, *src;
598 struct l2tp_req l2tpr;
599 u_long mtu;
600 int bound;
601 struct psref psref;
602
603 switch (cmd) {
604 case SIOCSIFADDR:
605 ifp->if_flags |= IFF_UP;
606 break;
607
608 case SIOCSIFDSTADDR:
609 break;
610
611 case SIOCADDMULTI:
612 case SIOCDELMULTI:
613 switch (ifr->ifr_addr.sa_family) {
614 #ifdef INET
615 case AF_INET: /* IP supports Multicast */
616 break;
617 #endif /* INET */
618 #ifdef INET6
619 case AF_INET6: /* IP6 supports Multicast */
620 break;
621 #endif /* INET6 */
622 default: /* Other protocols doesn't support Multicast */
623 error = EAFNOSUPPORT;
624 break;
625 }
626 break;
627
628 case SIOCSIFMTU:
629 mtu = ifr->ifr_mtu;
630 if (mtu < L2TP_MTU_MIN || mtu > L2TP_MTU_MAX)
631 return (EINVAL);
632 ifp->if_mtu = mtu;
633 break;
634
635 #ifdef INET
636 case SIOCSIFPHYADDR:
637 src = (struct sockaddr *)
638 &(((struct in_aliasreq *)data)->ifra_addr);
639 dst = (struct sockaddr *)
640 &(((struct in_aliasreq *)data)->ifra_dstaddr);
641 if (src->sa_family != AF_INET || dst->sa_family != AF_INET)
642 return EAFNOSUPPORT;
643 else if (src->sa_len != sizeof(struct sockaddr_in)
644 || dst->sa_len != sizeof(struct sockaddr_in))
645 return EINVAL;
646
647 error = l2tp_set_tunnel(&sc->l2tp_ec.ec_if, src, dst);
648 break;
649
650 #endif /* INET */
651 #ifdef INET6
652 case SIOCSIFPHYADDR_IN6:
653 src = (struct sockaddr *)
654 &(((struct in6_aliasreq *)data)->ifra_addr);
655 dst = (struct sockaddr *)
656 &(((struct in6_aliasreq *)data)->ifra_dstaddr);
657 if (src->sa_family != AF_INET6 || dst->sa_family != AF_INET6)
658 return EAFNOSUPPORT;
659 else if (src->sa_len != sizeof(struct sockaddr_in6)
660 || dst->sa_len != sizeof(struct sockaddr_in6))
661 return EINVAL;
662
663 error = l2tp_set_tunnel(&sc->l2tp_ec.ec_if, src, dst);
664 break;
665
666 #endif /* INET6 */
667 case SIOCSLIFPHYADDR:
668 src = (struct sockaddr *)
669 &(((struct if_laddrreq *)data)->addr);
670 dst = (struct sockaddr *)
671 &(((struct if_laddrreq *)data)->dstaddr);
672 if (src->sa_family != dst->sa_family)
673 return EINVAL;
674 else if (src->sa_family == AF_INET
675 && src->sa_len != sizeof(struct sockaddr_in))
676 return EINVAL;
677 else if (src->sa_family == AF_INET6
678 && src->sa_len != sizeof(struct sockaddr_in6))
679 return EINVAL;
680 else if (dst->sa_family == AF_INET
681 && dst->sa_len != sizeof(struct sockaddr_in))
682 return EINVAL;
683 else if (dst->sa_family == AF_INET6
684 && dst->sa_len != sizeof(struct sockaddr_in6))
685 return EINVAL;
686
687 error = l2tp_set_tunnel(&sc->l2tp_ec.ec_if, src, dst);
688 break;
689
690 case SIOCDIFPHYADDR:
691 l2tp_delete_tunnel(&sc->l2tp_ec.ec_if);
692 break;
693
694 case SIOCGIFPSRCADDR:
695 #ifdef INET6
696 case SIOCGIFPSRCADDR_IN6:
697 #endif /* INET6 */
698 bound = curlwp_bind();
699 var = l2tp_getref_variant(sc, &psref);
700 if (var == NULL) {
701 curlwp_bindx(bound);
702 error = EADDRNOTAVAIL;
703 goto bad;
704 }
705 if (var->lv_psrc == NULL) {
706 l2tp_putref_variant(var, &psref);
707 curlwp_bindx(bound);
708 error = EADDRNOTAVAIL;
709 goto bad;
710 }
711 src = var->lv_psrc;
712 switch (cmd) {
713 #ifdef INET
714 case SIOCGIFPSRCADDR:
715 dst = &ifr->ifr_addr;
716 size = sizeof(ifr->ifr_addr);
717 break;
718 #endif /* INET */
719 #ifdef INET6
720 case SIOCGIFPSRCADDR_IN6:
721 dst = (struct sockaddr *)
722 &(((struct in6_ifreq *)data)->ifr_addr);
723 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
724 break;
725 #endif /* INET6 */
726 default:
727 l2tp_putref_variant(var, &psref);
728 curlwp_bindx(bound);
729 error = EADDRNOTAVAIL;
730 goto bad;
731 }
732 if (src->sa_len > size) {
733 l2tp_putref_variant(var, &psref);
734 curlwp_bindx(bound);
735 return EINVAL;
736 }
737 sockaddr_copy(dst, src->sa_len, src);
738 l2tp_putref_variant(var, &psref);
739 curlwp_bindx(bound);
740 break;
741
742 case SIOCGIFPDSTADDR:
743 #ifdef INET6
744 case SIOCGIFPDSTADDR_IN6:
745 #endif /* INET6 */
746 bound = curlwp_bind();
747 var = l2tp_getref_variant(sc, &psref);
748 if (var == NULL) {
749 curlwp_bindx(bound);
750 error = EADDRNOTAVAIL;
751 goto bad;
752 }
753 if (var->lv_pdst == NULL) {
754 l2tp_putref_variant(var, &psref);
755 curlwp_bindx(bound);
756 error = EADDRNOTAVAIL;
757 goto bad;
758 }
759 src = var->lv_pdst;
760 switch (cmd) {
761 #ifdef INET
762 case SIOCGIFPDSTADDR:
763 dst = &ifr->ifr_addr;
764 size = sizeof(ifr->ifr_addr);
765 break;
766 #endif /* INET */
767 #ifdef INET6
768 case SIOCGIFPDSTADDR_IN6:
769 dst = (struct sockaddr *)
770 &(((struct in6_ifreq *)data)->ifr_addr);
771 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
772 break;
773 #endif /* INET6 */
774 default:
775 l2tp_putref_variant(var, &psref);
776 curlwp_bindx(bound);
777 error = EADDRNOTAVAIL;
778 goto bad;
779 }
780 if (src->sa_len > size) {
781 l2tp_putref_variant(var, &psref);
782 curlwp_bindx(bound);
783 return EINVAL;
784 }
785 sockaddr_copy(dst, src->sa_len, src);
786 l2tp_putref_variant(var, &psref);
787 curlwp_bindx(bound);
788 break;
789
790 case SIOCGLIFPHYADDR:
791 bound = curlwp_bind();
792 var = l2tp_getref_variant(sc, &psref);
793 if (var == NULL) {
794 curlwp_bindx(bound);
795 error = EADDRNOTAVAIL;
796 goto bad;
797 }
798 if (var->lv_psrc == NULL || var->lv_pdst == NULL) {
799 l2tp_putref_variant(var, &psref);
800 curlwp_bindx(bound);
801 error = EADDRNOTAVAIL;
802 goto bad;
803 }
804
805 /* copy src */
806 src = var->lv_psrc;
807 dst = (struct sockaddr *)
808 &(((struct if_laddrreq *)data)->addr);
809 size = sizeof(((struct if_laddrreq *)data)->addr);
810 if (src->sa_len > size) {
811 l2tp_putref_variant(var, &psref);
812 curlwp_bindx(bound);
813 return EINVAL;
814 }
815 sockaddr_copy(dst, src->sa_len, src);
816
817 /* copy dst */
818 src = var->lv_pdst;
819 dst = (struct sockaddr *)
820 &(((struct if_laddrreq *)data)->dstaddr);
821 size = sizeof(((struct if_laddrreq *)data)->dstaddr);
822 if (src->sa_len > size) {
823 l2tp_putref_variant(var, &psref);
824 curlwp_bindx(bound);
825 return EINVAL;
826 }
827 sockaddr_copy(dst, src->sa_len, src);
828 l2tp_putref_variant(var, &psref);
829 curlwp_bindx(bound);
830 break;
831
832 case SIOCSL2TPSESSION:
833 if ((error = copyin(ifr->ifr_data, &l2tpr, sizeof(l2tpr))) != 0)
834 break;
835
836 /* session id must not zero */
837 if (l2tpr.my_sess_id == 0 || l2tpr.peer_sess_id == 0)
838 return EINVAL;
839
840 bound = curlwp_bind();
841 var_tmp = l2tp_lookup_session_ref(l2tpr.my_sess_id, &psref);
842 if (var_tmp != NULL) {
843 /* duplicate session id */
844 log(LOG_WARNING, "%s: duplicate session id %" PRIu32 " of %s\n",
845 sc->l2tp_ec.ec_if.if_xname, l2tpr.my_sess_id,
846 var_tmp->lv_softc->l2tp_ec.ec_if.if_xname);
847 psref_release(&psref, &var_tmp->lv_psref,
848 lv_psref_class);
849 curlwp_bindx(bound);
850 return EINVAL;
851 }
852 curlwp_bindx(bound);
853
854 error = l2tp_set_session(sc, l2tpr.my_sess_id, l2tpr.peer_sess_id);
855 break;
856 case SIOCDL2TPSESSION:
857 l2tp_clear_session(sc);
858 break;
859 case SIOCSL2TPCOOKIE:
860 if ((error = copyin(ifr->ifr_data, &l2tpr, sizeof(l2tpr))) != 0)
861 break;
862
863 error = l2tp_set_cookie(sc, l2tpr.my_cookie, l2tpr.my_cookie_len,
864 l2tpr.peer_cookie, l2tpr.peer_cookie_len);
865 break;
866 case SIOCDL2TPCOOKIE:
867 l2tp_clear_cookie(sc);
868 break;
869 case SIOCSL2TPSTATE:
870 if ((error = copyin(ifr->ifr_data, &l2tpr, sizeof(l2tpr))) != 0)
871 break;
872
873 l2tp_set_state(sc, l2tpr.state);
874 break;
875 case SIOCGL2TP:
876 /* get L2TPV3 session info */
877 memset(&l2tpr, 0, sizeof(l2tpr));
878
879 bound = curlwp_bind();
880 var = l2tp_getref_variant(sc, &psref);
881 if (var == NULL) {
882 curlwp_bindx(bound);
883 error = EADDRNOTAVAIL;
884 goto bad;
885 }
886
887 l2tpr.state = var->lv_state;
888 l2tpr.my_sess_id = var->lv_my_sess_id;
889 l2tpr.peer_sess_id = var->lv_peer_sess_id;
890 l2tpr.my_cookie = var->lv_my_cookie;
891 l2tpr.my_cookie_len = var->lv_my_cookie_len;
892 l2tpr.peer_cookie = var->lv_peer_cookie;
893 l2tpr.peer_cookie_len = var->lv_peer_cookie_len;
894 l2tp_putref_variant(var, &psref);
895 curlwp_bindx(bound);
896
897 error = copyout(&l2tpr, ifr->ifr_data, sizeof(l2tpr));
898 break;
899
900 default:
901 error = ifioctl_common(ifp, cmd, data);
902 break;
903 }
904 bad:
905 return error;
906 }
907
908 static int
909 l2tp_set_tunnel(struct ifnet *ifp, struct sockaddr *src, struct sockaddr *dst)
910 {
911 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
912 l2tp_ec.ec_if);
913 struct sockaddr *osrc, *odst;
914 struct sockaddr *nsrc, *ndst;
915 struct l2tp_variant *ovar, *nvar;
916 int error;
917
918 nsrc = sockaddr_dup(src, M_WAITOK);
919 ndst = sockaddr_dup(dst, M_WAITOK);
920
921 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
922
923 error = encap_lock_enter();
924 if (error)
925 goto error;
926
927 mutex_enter(&sc->l2tp_lock);
928
929 ovar = sc->l2tp_var;
930 osrc = ovar->lv_psrc;
931 odst = ovar->lv_pdst;
932 *nvar = *ovar;
933 psref_target_init(&nvar->lv_psref, lv_psref_class);
934 nvar->lv_psrc = nsrc;
935 nvar->lv_pdst = ndst;
936 error = l2tp_encap_attach(nvar);
937 if (error) {
938 mutex_exit(&sc->l2tp_lock);
939 encap_lock_exit();
940 goto error;
941 }
942 membar_producer();
943 l2tp_variant_update(sc, nvar);
944
945 mutex_exit(&sc->l2tp_lock);
946
947 (void)l2tp_encap_detach(ovar);
948 encap_lock_exit();
949
950 if (osrc)
951 sockaddr_free(osrc);
952 if (odst)
953 sockaddr_free(odst);
954 kmem_free(ovar, sizeof(*ovar));
955
956 return 0;
957
958 error:
959 sockaddr_free(nsrc);
960 sockaddr_free(ndst);
961 kmem_free(nvar, sizeof(*nvar));
962
963 return error;
964 }
965
966 static void
967 l2tp_delete_tunnel(struct ifnet *ifp)
968 {
969 struct l2tp_softc *sc = container_of(ifp, struct l2tp_softc,
970 l2tp_ec.ec_if);
971 struct sockaddr *osrc, *odst;
972 struct l2tp_variant *ovar, *nvar;
973 int error;
974
975 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
976
977 error = encap_lock_enter();
978 if (error) {
979 kmem_free(nvar, sizeof(*nvar));
980 return;
981 }
982 mutex_enter(&sc->l2tp_lock);
983
984 ovar = sc->l2tp_var;
985 osrc = ovar->lv_psrc;
986 odst = ovar->lv_pdst;
987 *nvar = *ovar;
988 psref_target_init(&nvar->lv_psref, lv_psref_class);
989 nvar->lv_psrc = NULL;
990 nvar->lv_pdst = NULL;
991 membar_producer();
992 l2tp_variant_update(sc, nvar);
993
994 mutex_exit(&sc->l2tp_lock);
995
996 (void)l2tp_encap_detach(ovar);
997 encap_lock_exit();
998
999 if (osrc)
1000 sockaddr_free(osrc);
1001 if (odst)
1002 sockaddr_free(odst);
1003 kmem_free(ovar, sizeof(*ovar));
1004 }
1005
1006 static int
1007 id_hash_func(uint32_t id, u_long mask)
1008 {
1009 uint32_t hash;
1010
1011 hash = (id >> 16) ^ id;
1012 hash = (hash >> 4) ^ hash;
1013
1014 return hash & mask;
1015 }
1016
1017 static void
1018 l2tp_hash_init(void)
1019 {
1020
1021 l2tp_hash.lists = hashinit(L2TP_ID_HASH_SIZE, HASH_PSLIST, true,
1022 &l2tp_hash.mask);
1023 }
1024
1025 static int
1026 l2tp_hash_fini(void)
1027 {
1028 int i;
1029
1030 mutex_enter(&l2tp_hash.lock);
1031
1032 for (i = 0; i < l2tp_hash.mask + 1; i++) {
1033 if (PSLIST_WRITER_FIRST(&l2tp_hash.lists[i], struct l2tp_softc,
1034 l2tp_hash) != NULL) {
1035 mutex_exit(&l2tp_hash.lock);
1036 return EBUSY;
1037 }
1038 }
1039 for (i = 0; i < l2tp_hash.mask + 1; i++)
1040 PSLIST_DESTROY(&l2tp_hash.lists[i]);
1041
1042 mutex_exit(&l2tp_hash.lock);
1043
1044 hashdone(l2tp_hash.lists, HASH_PSLIST, l2tp_hash.mask);
1045
1046 return 0;
1047 }
1048
1049 static int
1050 l2tp_set_session(struct l2tp_softc *sc, uint32_t my_sess_id,
1051 uint32_t peer_sess_id)
1052 {
1053 uint32_t idx;
1054 struct l2tp_variant *nvar;
1055 struct l2tp_variant *ovar;
1056 struct ifnet *ifp = &sc->l2tp_ec.ec_if;
1057
1058 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
1059
1060 mutex_enter(&sc->l2tp_lock);
1061 ovar = sc->l2tp_var;
1062 *nvar = *ovar;
1063 psref_target_init(&nvar->lv_psref, lv_psref_class);
1064 nvar->lv_my_sess_id = my_sess_id;
1065 nvar->lv_peer_sess_id = peer_sess_id;
1066 membar_producer();
1067
1068 mutex_enter(&l2tp_hash.lock);
1069 if (ovar->lv_my_sess_id > 0 && ovar->lv_peer_sess_id > 0) {
1070 PSLIST_WRITER_REMOVE(sc, l2tp_hash);
1071 pserialize_perform(l2tp_psz);
1072 }
1073 mutex_exit(&l2tp_hash.lock);
1074
1075 l2tp_variant_update(sc, nvar);
1076 mutex_exit(&sc->l2tp_lock);
1077
1078 idx = id_hash_func(nvar->lv_my_sess_id, l2tp_hash.mask);
1079 if ((ifp->if_flags & IFF_DEBUG) != 0)
1080 log(LOG_DEBUG, "%s: add hash entry: sess_id=%" PRIu32 ", idx=%" PRIu32 "\n",
1081 sc->l2tp_ec.ec_if.if_xname, nvar->lv_my_sess_id, idx);
1082
1083 mutex_enter(&l2tp_hash.lock);
1084 PSLIST_WRITER_INSERT_HEAD(&l2tp_hash.lists[idx], sc, l2tp_hash);
1085 mutex_exit(&l2tp_hash.lock);
1086
1087 kmem_free(ovar, sizeof(*ovar));
1088 return 0;
1089 }
1090
1091 static int
1092 l2tp_clear_session(struct l2tp_softc *sc)
1093 {
1094 struct l2tp_variant *nvar;
1095 struct l2tp_variant *ovar;
1096
1097 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
1098
1099 mutex_enter(&sc->l2tp_lock);
1100 ovar = sc->l2tp_var;
1101 *nvar = *ovar;
1102 psref_target_init(&nvar->lv_psref, lv_psref_class);
1103 nvar->lv_my_sess_id = 0;
1104 nvar->lv_peer_sess_id = 0;
1105 membar_producer();
1106
1107 mutex_enter(&l2tp_hash.lock);
1108 if (ovar->lv_my_sess_id > 0 && ovar->lv_peer_sess_id > 0) {
1109 PSLIST_WRITER_REMOVE(sc, l2tp_hash);
1110 pserialize_perform(l2tp_psz);
1111 }
1112 mutex_exit(&l2tp_hash.lock);
1113
1114 l2tp_variant_update(sc, nvar);
1115 mutex_exit(&sc->l2tp_lock);
1116 kmem_free(ovar, sizeof(*ovar));
1117 return 0;
1118 }
1119
1120 struct l2tp_variant *
1121 l2tp_lookup_session_ref(uint32_t id, struct psref *psref)
1122 {
1123 int idx;
1124 int s;
1125 struct l2tp_softc *sc;
1126
1127 idx = id_hash_func(id, l2tp_hash.mask);
1128
1129 s = pserialize_read_enter();
1130 PSLIST_READER_FOREACH(sc, &l2tp_hash.lists[idx], struct l2tp_softc,
1131 l2tp_hash) {
1132 struct l2tp_variant *var = sc->l2tp_var;
1133 if (var == NULL)
1134 continue;
1135 if (var->lv_my_sess_id != id)
1136 continue;
1137 psref_acquire(psref, &var->lv_psref, lv_psref_class);
1138 pserialize_read_exit(s);
1139 return var;
1140 }
1141 pserialize_read_exit(s);
1142 return NULL;
1143 }
1144
1145 /*
1146 * l2tp_variant update API.
1147 *
1148 * Assumption:
1149 * reader side dereferences sc->l2tp_var in reader critical section only,
1150 * that is, all of reader sides do not reader the sc->l2tp_var after
1151 * pserialize_perform().
1152 */
1153 static void
1154 l2tp_variant_update(struct l2tp_softc *sc, struct l2tp_variant *nvar)
1155 {
1156 struct ifnet *ifp = &sc->l2tp_ec.ec_if;
1157 struct l2tp_variant *ovar = sc->l2tp_var;
1158
1159 KASSERT(mutex_owned(&sc->l2tp_lock));
1160
1161 sc->l2tp_var = nvar;
1162 pserialize_perform(l2tp_psz);
1163 psref_target_destroy(&ovar->lv_psref, lv_psref_class);
1164
1165 /*
1166 * In the manual of atomic_swap_ptr(3), there is no mention if 2nd
1167 * argument is rewrite or not. So, use sc->l2tp_var instead of nvar.
1168 */
1169 if (sc->l2tp_var != NULL) {
1170 if (sc->l2tp_var->lv_psrc != NULL
1171 && sc->l2tp_var->lv_pdst != NULL)
1172 ifp->if_flags |= IFF_RUNNING;
1173 else
1174 ifp->if_flags &= ~IFF_RUNNING;
1175 }
1176 }
1177
1178 static int
1179 l2tp_set_cookie(struct l2tp_softc *sc, uint64_t my_cookie, u_int my_cookie_len,
1180 uint64_t peer_cookie, u_int peer_cookie_len)
1181 {
1182 struct l2tp_variant *nvar;
1183
1184 if (my_cookie == 0 || peer_cookie == 0)
1185 return EINVAL;
1186
1187 if (my_cookie_len != 4 && my_cookie_len != 8
1188 && peer_cookie_len != 4 && peer_cookie_len != 8)
1189 return EINVAL;
1190
1191 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
1192
1193 mutex_enter(&sc->l2tp_lock);
1194
1195 *nvar = *sc->l2tp_var;
1196 psref_target_init(&nvar->lv_psref, lv_psref_class);
1197 nvar->lv_my_cookie = my_cookie;
1198 nvar->lv_my_cookie_len = my_cookie_len;
1199 nvar->lv_peer_cookie = peer_cookie;
1200 nvar->lv_peer_cookie_len = peer_cookie_len;
1201 nvar->lv_use_cookie = L2TP_COOKIE_ON;
1202 membar_producer();
1203 l2tp_variant_update(sc, nvar);
1204
1205 mutex_exit(&sc->l2tp_lock);
1206
1207 struct ifnet *ifp = &sc->l2tp_ec.ec_if;
1208 if ((ifp->if_flags & IFF_DEBUG) != 0) {
1209 log(LOG_DEBUG,
1210 "%s: set cookie: "
1211 "local cookie_len=%u local cookie=%" PRIu64 ", "
1212 "remote cookie_len=%u remote cookie=%" PRIu64 "\n",
1213 ifp->if_xname, my_cookie_len, my_cookie,
1214 peer_cookie_len, peer_cookie);
1215 }
1216
1217 return 0;
1218 }
1219
1220 static void
1221 l2tp_clear_cookie(struct l2tp_softc *sc)
1222 {
1223 struct l2tp_variant *nvar;
1224
1225 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
1226
1227 mutex_enter(&sc->l2tp_lock);
1228
1229 *nvar = *sc->l2tp_var;
1230 psref_target_init(&nvar->lv_psref, lv_psref_class);
1231 nvar->lv_my_cookie = 0;
1232 nvar->lv_my_cookie_len = 0;
1233 nvar->lv_peer_cookie = 0;
1234 nvar->lv_peer_cookie_len = 0;
1235 nvar->lv_use_cookie = L2TP_COOKIE_OFF;
1236 membar_producer();
1237 l2tp_variant_update(sc, nvar);
1238
1239 mutex_exit(&sc->l2tp_lock);
1240 }
1241
1242 static void
1243 l2tp_set_state(struct l2tp_softc *sc, int state)
1244 {
1245 struct ifnet *ifp = &sc->l2tp_ec.ec_if;
1246 struct l2tp_variant *nvar;
1247
1248 nvar = kmem_alloc(sizeof(*nvar), KM_SLEEP);
1249
1250 mutex_enter(&sc->l2tp_lock);
1251
1252 *nvar = *sc->l2tp_var;
1253 psref_target_init(&nvar->lv_psref, lv_psref_class);
1254 nvar->lv_state = state;
1255 membar_producer();
1256 l2tp_variant_update(sc, nvar);
1257
1258 if (nvar->lv_state == L2TP_STATE_UP) {
1259 ifp->if_link_state = LINK_STATE_UP;
1260 } else {
1261 ifp->if_link_state = LINK_STATE_DOWN;
1262 }
1263
1264 mutex_exit(&sc->l2tp_lock);
1265
1266 #ifdef NOTYET
1267 vlan_linkstate_notify(ifp, ifp->if_link_state);
1268 #endif
1269 }
1270
1271 static int
1272 l2tp_encap_attach(struct l2tp_variant *var)
1273 {
1274 int error;
1275
1276 if (var == NULL || var->lv_psrc == NULL)
1277 return EINVAL;
1278
1279 switch (var->lv_psrc->sa_family) {
1280 #ifdef INET
1281 case AF_INET:
1282 error = in_l2tp_attach(var);
1283 break;
1284 #endif
1285 #ifdef INET6
1286 case AF_INET6:
1287 error = in6_l2tp_attach(var);
1288 break;
1289 #endif
1290 default:
1291 error = EINVAL;
1292 break;
1293 }
1294
1295 return error;
1296 }
1297
1298 static int
1299 l2tp_encap_detach(struct l2tp_variant *var)
1300 {
1301 int error;
1302
1303 if (var == NULL || var->lv_psrc == NULL)
1304 return EINVAL;
1305
1306 switch (var->lv_psrc->sa_family) {
1307 #ifdef INET
1308 case AF_INET:
1309 error = in_l2tp_detach(var);
1310 break;
1311 #endif
1312 #ifdef INET6
1313 case AF_INET6:
1314 error = in6_l2tp_detach(var);
1315 break;
1316 #endif
1317 default:
1318 error = EINVAL;
1319 break;
1320 }
1321
1322 return error;
1323 }
1324
1325 /*
1326 * TODO:
1327 * unify with gif_check_nesting().
1328 */
1329 int
1330 l2tp_check_nesting(struct ifnet *ifp, struct mbuf *m)
1331 {
1332 struct m_tag *mtag;
1333 int *count;
1334
1335 mtag = m_tag_find(m, PACKET_TAG_TUNNEL_INFO, NULL);
1336 if (mtag != NULL) {
1337 count = (int *)(mtag + 1);
1338 if (++(*count) > max_l2tp_nesting) {
1339 log(LOG_NOTICE,
1340 "%s: recursively called too many times(%d)\n",
1341 if_name(ifp),
1342 *count);
1343 return EIO;
1344 }
1345 } else {
1346 mtag = m_tag_get(PACKET_TAG_TUNNEL_INFO, sizeof(*count),
1347 M_NOWAIT);
1348 if (mtag != NULL) {
1349 m_tag_prepend(m, mtag);
1350 count = (int *)(mtag + 1);
1351 *count = 0;
1352 }
1353 #ifdef L2TP_DEBUG
1354 else {
1355 log(LOG_DEBUG,
1356 "%s: m_tag_get() failed, recursion calls are not prevented.\n",
1357 if_name(ifp));
1358 }
1359 #endif
1360 }
1361
1362 return 0;
1363 }
1364
1365 /*
1366 * Module infrastructure
1367 */
1368 #include "if_module.h"
1369
1370 IF_MODULE(MODULE_CLASS_DRIVER, l2tp, "")
1371
1372
1373 /* TODO: IP_TCPMSS support */
1374 #ifdef IP_TCPMSS
1375 static int l2tp_need_tcpmss_clamp(struct ifnet *);
1376 #ifdef INET
1377 static struct mbuf *l2tp_tcpmss4_clamp(struct ifnet *, struct mbuf *);
1378 #endif
1379 #ifdef INET6
1380 static struct mbuf *l2tp_tcpmss6_clamp(struct ifnet *, struct mbuf *);
1381 #endif
1382
1383 struct mbuf *
1384 l2tp_tcpmss_clamp(struct ifnet *ifp, struct mbuf *m)
1385 {
1386
1387 if (l2tp_need_tcpmss_clamp(ifp)) {
1388 struct ether_header *eh;
1389 struct ether_vlan_header evh;
1390
1391 /* save ether header */
1392 m_copydata(m, 0, sizeof(evh), (void *)&evh);
1393 eh = (struct ether_header *)&evh;
1394
1395 switch (ntohs(eh->ether_type)) {
1396 case ETHERTYPE_VLAN: /* Ether + VLAN */
1397 if (m->m_pkthdr.len <= sizeof(struct ether_vlan_header))
1398 break;
1399 m_adj(m, sizeof(struct ether_vlan_header));
1400 switch (ntohs(evh.evl_proto)) {
1401 #ifdef INET
1402 case ETHERTYPE_IP: /* Ether + VLAN + IPv4 */
1403 m = l2tp_tcpmss4_clamp(ifp, m);
1404 if (m == NULL)
1405 return NULL;
1406 break;
1407 #endif /* INET */
1408 #ifdef INET6
1409 case ETHERTYPE_IPV6: /* Ether + VLAN + IPv6 */
1410 m = l2tp_tcpmss6_clamp(ifp, m);
1411 if (m == NULL)
1412 return NULL;
1413 break;
1414 #endif /* INET6 */
1415 default:
1416 break;
1417 }
1418 /* restore ether header */
1419 M_PREPEND(m, sizeof(struct ether_vlan_header),
1420 M_DONTWAIT);
1421 if (m == NULL)
1422 return NULL;
1423 *mtod(m, struct ether_vlan_header *) = evh;
1424 break;
1425 #ifdef INET
1426 case ETHERTYPE_IP: /* Ether + IPv4 */
1427 if (m->m_pkthdr.len <= sizeof(struct ether_header))
1428 break;
1429 m_adj(m, sizeof(struct ether_header));
1430 m = l2tp_tcpmss4_clamp(ifp, m);
1431 if (m == NULL)
1432 return NULL;
1433 /* restore ether header */
1434 M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
1435 if (m == NULL)
1436 return NULL;
1437 *mtod(m, struct ether_header *) = *eh;
1438 break;
1439 #endif /* INET */
1440 #ifdef INET6
1441 case ETHERTYPE_IPV6: /* Ether + IPv6 */
1442 if (m->m_pkthdr.len <= sizeof(struct ether_header))
1443 break;
1444 m_adj(m, sizeof(struct ether_header));
1445 m = l2tp_tcpmss6_clamp(ifp, m);
1446 if (m == NULL)
1447 return NULL;
1448 /* restore ether header */
1449 M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
1450 if (m == NULL)
1451 return NULL;
1452 *mtod(m, struct ether_header *) = *eh;
1453 break;
1454 #endif /* INET6 */
1455 default:
1456 break;
1457 }
1458 }
1459
1460 return m;
1461 }
1462
1463 static int
1464 l2tp_need_tcpmss_clamp(struct ifnet *ifp)
1465 {
1466 int ret = 0;
1467
1468 #ifdef INET
1469 if (ifp->if_tcpmss != 0)
1470 ret = 1;
1471 #endif /* INET */
1472
1473 #ifdef INET6
1474 if (ifp->if_tcpmss6 != 0)
1475 ret = 1;
1476 #endif /* INET6 */
1477
1478 return ret;
1479 }
1480
1481 #ifdef INET
1482 static struct mbuf *
1483 l2tp_tcpmss4_clamp(struct ifnet *ifp, struct mbuf *m)
1484 {
1485
1486 if (ifp->if_tcpmss != 0) {
1487 return ip_tcpmss(m, (ifp->if_tcpmss < 0) ?
1488 ifp->if_mtu - IP_TCPMSS_EXTLEN :
1489 ifp->if_tcpmss);
1490 }
1491 return m;
1492 }
1493 #endif /* INET */
1494
1495 #ifdef INET6
1496 static struct mbuf *
1497 l2tp_tcpmss6_clamp(struct ifnet *ifp, struct mbuf *m)
1498 {
1499 int ip6hdrlen;
1500
1501 if (ifp->if_tcpmss6 != 0 &&
1502 ip6_tcpmss_applicable(m, &ip6hdrlen)) {
1503 return ip6_tcpmss(m, ip6hdrlen,
1504 (ifp->if_tcpmss6 < 0) ?
1505 ifp->if_mtu - IP6_TCPMSS_EXTLEN :
1506 ifp->if_tcpmss6);
1507 }
1508 return m;
1509 }
1510 #endif /* INET6 */
1511
1512 #endif /* IP_TCPMSS */
1513