xref: /netbsd-src/sys/net/if_gre.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /*	$NetBSD: if_gre.c,v 1.118 2007/11/07 00:19:54 ad Exp $ */
2 
3 /*
4  * Copyright (c) 1998 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Heiko W.Rupp <hwr@pilhuhn.de>
9  *
10  * IPv6-over-GRE contributed by Gert Doering <gert@greenie.muc.de>
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *        This product includes software developed by the NetBSD
23  *        Foundation, Inc. and its contributors.
24  * 4. Neither the name of The NetBSD Foundation nor the names of its
25  *    contributors may be used to endorse or promote products derived
26  *    from this software without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
29  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
30  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
31  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
32  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
33  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
34  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
35  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
36  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
37  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
38  * POSSIBILITY OF SUCH DAMAGE.
39  */
40 
41 /*
42  * Encapsulate L3 protocols into IP
43  * See RFC 1701 and 1702 for more details.
44  * If_gre is compatible with Cisco GRE tunnels, so you can
45  * have a NetBSD box as the other end of a tunnel interface of a Cisco
46  * router. See gre(4) for more details.
47  */
48 
49 #include <sys/cdefs.h>
50 __KERNEL_RCSID(0, "$NetBSD: if_gre.c,v 1.118 2007/11/07 00:19:54 ad Exp $");
51 
52 #include "opt_gre.h"
53 #include "opt_inet.h"
54 #include "bpfilter.h"
55 
56 #ifdef INET
57 #include <sys/param.h>
58 #include <sys/file.h>
59 #include <sys/filedesc.h>
60 #include <sys/malloc.h>
61 #include <sys/mallocvar.h>
62 #include <sys/mbuf.h>
63 #include <sys/proc.h>
64 #include <sys/domain.h>
65 #include <sys/protosw.h>
66 #include <sys/socket.h>
67 #include <sys/socketvar.h>
68 #include <sys/ioctl.h>
69 #include <sys/queue.h>
70 #include <sys/intr.h>
71 #if __NetBSD__
72 #include <sys/systm.h>
73 #include <sys/sysctl.h>
74 #include <sys/kauth.h>
75 #endif
76 
77 #include <sys/kernel.h>
78 #include <sys/mutex.h>
79 #include <sys/condvar.h>
80 #include <sys/kthread.h>
81 
82 #include <sys/cpu.h>
83 
84 #include <net/ethertypes.h>
85 #include <net/if.h>
86 #include <net/if_types.h>
87 #include <net/netisr.h>
88 #include <net/route.h>
89 
90 #ifdef INET
91 #include <netinet/in.h>
92 #include <netinet/in_systm.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip.h>
95 #include <netinet/ip_var.h>
96 #else
97 #error "Huh? if_gre without inet?"
98 #endif
99 
100 
101 #ifdef NETATALK
102 #include <netatalk/at.h>
103 #include <netatalk/at_var.h>
104 #include <netatalk/at_extern.h>
105 #endif
106 
107 #if NBPFILTER > 0
108 #include <sys/time.h>
109 #include <net/bpf.h>
110 #endif
111 
112 #include <net/if_gre.h>
113 
114 #include <compat/sys/socket.h>
115 #include <compat/sys/sockio.h>
116 /*
117  * It is not easy to calculate the right value for a GRE MTU.
118  * We leave this task to the admin and use the same default that
119  * other vendors use.
120  */
121 #define GREMTU 1476
122 
123 #ifdef GRE_DEBUG
124 int gre_debug = 0;
125 #define	GRE_DPRINTF(__sc, __fmt, ...)				\
126 	do {							\
127 		if (__predict_false(gre_debug ||		\
128 		    ((__sc)->sc_if.if_flags & IFF_DEBUG) != 0))	\
129 			printf(__fmt, __VA_ARGS__);		\
130 	} while (/*CONSTCOND*/0)
131 #else
132 #define	GRE_DPRINTF(__sc, __fmt, ...)	do { } while (/*CONSTCOND*/0)
133 #endif /* GRE_DEBUG */
134 
135 int ip_gre_ttl = GRE_TTL;
136 MALLOC_DEFINE(M_GRE_BUFQ, "gre_bufq", "gre mbuf queue");
137 
138 static int gre_clone_create(struct if_clone *, int);
139 static int gre_clone_destroy(struct ifnet *);
140 
141 static struct if_clone gre_cloner =
142     IF_CLONE_INITIALIZER("gre", gre_clone_create, gre_clone_destroy);
143 
144 static int gre_input(struct gre_softc *, struct mbuf *, int,
145     const struct gre_h *);
146 static bool gre_is_nullconf(const struct gre_soparm *);
147 static int gre_output(struct ifnet *, struct mbuf *,
148 			   const struct sockaddr *, struct rtentry *);
149 static int gre_ioctl(struct ifnet *, u_long, void *);
150 static void gre_closef(struct file **, struct lwp *);
151 static int gre_getsockname(struct socket *, struct mbuf *, struct lwp *);
152 static int gre_getpeername(struct socket *, struct mbuf *, struct lwp *);
153 static int gre_getnames(struct socket *, struct lwp *,
154     struct sockaddr_storage *, struct sockaddr_storage *);
155 static void gre_clearconf(struct gre_soparm *, bool);
156 static int gre_soreceive(struct socket *, struct mbuf **);
157 static int gre_sosend(struct socket *, struct mbuf *, struct lwp *);
158 static struct socket *gre_reconf(struct gre_softc *, struct socket *, lwp_t *,
159     const struct gre_soparm *);
160 
161 static int
162 nearest_pow2(size_t len0)
163 {
164 	size_t len, mid;
165 
166 	if (len0 == 0)
167 		return 1;
168 
169 	for (len = len0; (len & (len - 1)) != 0; len &= len - 1)
170 		;
171 
172 	mid = len | (len >> 1);
173 
174 	/* avoid overflow */
175 	if ((len << 1) < len)
176 		return len;
177 	if (len0 >= mid)
178 		return len << 1;
179 	return len;
180 }
181 
182 static struct gre_bufq *
183 gre_bufq_init(struct gre_bufq *bq, size_t len0)
184 {
185 	size_t len;
186 
187 	len = nearest_pow2(len0);
188 
189 	memset(bq, 0, sizeof(*bq));
190 	bq->bq_buf = malloc(len * sizeof(struct mbuf *), M_GRE_BUFQ, M_WAITOK);
191 	bq->bq_len = len;
192 	bq->bq_lenmask = len - 1;
193 
194 	return bq;
195 }
196 
197 static bool
198 gre_bufq_empty(struct gre_bufq *bq)
199 {
200 	return bq->bq_prodidx == bq->bq_considx;
201 }
202 
203 static struct mbuf *
204 gre_bufq_dequeue(struct gre_bufq *bq)
205 {
206 	struct mbuf *m;
207 
208 	if (gre_bufq_empty(bq))
209 		return NULL;
210 
211 	m = bq->bq_buf[bq->bq_considx];
212 	bq->bq_considx = (bq->bq_considx + 1) & bq->bq_lenmask;
213 
214 	return m;
215 }
216 
217 static void
218 gre_bufq_purge(struct gre_bufq *bq)
219 {
220 	struct mbuf *m;
221 
222 	while ((m = gre_bufq_dequeue(bq)) != NULL)
223 		m_freem(m);
224 }
225 
226 static int
227 gre_bufq_enqueue(struct gre_bufq *bq, struct mbuf *m)
228 {
229 	int next;
230 
231 	next = (bq->bq_prodidx + 1) & bq->bq_lenmask;
232 
233 	if (next == bq->bq_considx) {
234 		bq->bq_drops++;
235 		return ENOBUFS;
236 	}
237 
238 	bq->bq_buf[bq->bq_prodidx] = m;
239 	bq->bq_prodidx = next;
240 	return 0;
241 }
242 
243 static void
244 greintr(void *arg)
245 {
246 	struct gre_softc *sc = (struct gre_softc *)arg;
247 	struct socket *so = sc->sc_so;
248 	lwp_t *l = curlwp;
249 	int rc;
250 	struct mbuf *m;
251 
252 	KASSERT(sc->sc_so != NULL);
253 
254 	sc->sc_send_ev.ev_count++;
255 	GRE_DPRINTF(sc, "%s: enter\n", __func__);
256 	while ((m = gre_bufq_dequeue(&sc->sc_snd)) != NULL) {
257 		/* XXX handle ENOBUFS? */
258 		if ((rc = gre_sosend(so, m, l)) != 0) {
259 			GRE_DPRINTF(sc, "%s: gre_sosend failed %d\n", __func__,
260 			    rc);
261 		}
262 	}
263 }
264 
265 /* Caller must hold sc->sc_mtx. */
266 static void
267 gre_wait(struct gre_softc *sc)
268 {
269 	sc->sc_waiters++;
270 	cv_wait(&sc->sc_condvar, &sc->sc_mtx);
271 	sc->sc_waiters--;
272 }
273 
274 /* Caller must hold sc->sc_mtx. */
275 static void
276 gre_join(struct gre_softc *sc)
277 {
278 	while (sc->sc_waiters > 0)
279 		cv_wait(&sc->sc_condvar, &sc->sc_mtx);
280 }
281 
282 static void
283 gre_evcnt_detach(struct gre_softc *sc)
284 {
285 	evcnt_detach(&sc->sc_unsupp_ev);
286 	evcnt_detach(&sc->sc_pullup_ev);
287 	evcnt_detach(&sc->sc_error_ev);
288 	evcnt_detach(&sc->sc_block_ev);
289 	evcnt_detach(&sc->sc_recv_ev);
290 
291 	evcnt_detach(&sc->sc_oflow_ev);
292 	evcnt_detach(&sc->sc_send_ev);
293 }
294 
295 static void
296 gre_evcnt_attach(struct gre_softc *sc)
297 {
298 	evcnt_attach_dynamic(&sc->sc_recv_ev, EVCNT_TYPE_MISC,
299 	    NULL, sc->sc_if.if_xname, "recv");
300 	evcnt_attach_dynamic(&sc->sc_block_ev, EVCNT_TYPE_MISC,
301 	    &sc->sc_recv_ev, sc->sc_if.if_xname, "would block");
302 	evcnt_attach_dynamic(&sc->sc_error_ev, EVCNT_TYPE_MISC,
303 	    &sc->sc_recv_ev, sc->sc_if.if_xname, "error");
304 	evcnt_attach_dynamic(&sc->sc_pullup_ev, EVCNT_TYPE_MISC,
305 	    &sc->sc_recv_ev, sc->sc_if.if_xname, "pullup failed");
306 	evcnt_attach_dynamic(&sc->sc_unsupp_ev, EVCNT_TYPE_MISC,
307 	    &sc->sc_recv_ev, sc->sc_if.if_xname, "unsupported");
308 
309 	evcnt_attach_dynamic(&sc->sc_send_ev, EVCNT_TYPE_MISC,
310 	    NULL, sc->sc_if.if_xname, "send");
311 	evcnt_attach_dynamic(&sc->sc_oflow_ev, EVCNT_TYPE_MISC,
312 	    &sc->sc_send_ev, sc->sc_if.if_xname, "overflow");
313 }
314 
315 static int
316 gre_clone_create(struct if_clone *ifc, int unit)
317 {
318 	struct gre_soparm sp;
319 	struct gre_softc *sc;
320 
321 	sc = malloc(sizeof(struct gre_softc), M_DEVBUF, M_WAITOK|M_ZERO);
322 	mutex_init(&sc->sc_mtx, MUTEX_DRIVER, IPL_SOFTNET);
323 	cv_init(&sc->sc_condvar, "gre wait");
324 
325 	snprintf(sc->sc_if.if_xname, sizeof(sc->sc_if.if_xname), "%s%d",
326 	    ifc->ifc_name, unit);
327 	sc->sc_if.if_softc = sc;
328 	sc->sc_if.if_type = IFT_TUNNEL;
329 	sc->sc_if.if_addrlen = 0;
330 	sc->sc_if.if_hdrlen = sizeof(struct ip) + sizeof(struct gre_h);
331 	sc->sc_if.if_dlt = DLT_NULL;
332 	sc->sc_if.if_mtu = GREMTU;
333 	sc->sc_if.if_flags = IFF_POINTOPOINT|IFF_MULTICAST;
334 	sc->sc_if.if_output = gre_output;
335 	sc->sc_if.if_ioctl = gre_ioctl;
336 	sockaddr_copy(sstosa(&sp.sp_dst), sizeof(sp.sp_dst), sintocsa(&in_any));
337 	sockaddr_copy(sstosa(&sp.sp_src), sizeof(sp.sp_src), sintocsa(&in_any));
338 	sp.sp_proto = IPPROTO_GRE;
339 	sp.sp_type = SOCK_RAW;
340 	sp.sp_bysock = 0;
341 	sp.sp_fd = -1;
342 	sc->sc_soparm = sp;
343 
344 	gre_evcnt_attach(sc);
345 
346 	gre_bufq_init(&sc->sc_snd, 17);
347 	sc->sc_if.if_flags |= IFF_LINK0;
348 	if_attach(&sc->sc_if);
349 	if_alloc_sadl(&sc->sc_if);
350 #if NBPFILTER > 0
351 	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int32_t));
352 #endif
353 	sc->sc_lwp = &lwp0;
354 	sc->sc_state = GRE_S_IDLE;
355 	return 0;
356 }
357 
358 static int
359 gre_clone_destroy(struct ifnet *ifp)
360 {
361 	struct gre_softc *sc = ifp->if_softc;
362 
363 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
364 
365 #if NBPFILTER > 0
366 	bpfdetach(ifp);
367 #endif
368 	if_detach(ifp);
369 
370 	mutex_enter(&sc->sc_mtx);
371 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
372 	while (sc->sc_state != GRE_S_IDLE)
373 		gre_wait(sc);
374 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
375 	sc->sc_state = GRE_S_DIE;
376 	cv_broadcast(&sc->sc_condvar);
377 	gre_join(sc);
378 	sc->sc_so = gre_reconf(sc, sc->sc_so, sc->sc_lwp, NULL);
379 	mutex_exit(&sc->sc_mtx);
380 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
381 
382 	cv_destroy(&sc->sc_condvar);
383 	mutex_destroy(&sc->sc_mtx);
384 	gre_evcnt_detach(sc);
385 	free(sc, M_DEVBUF);
386 
387 	return 0;
388 }
389 
390 static void
391 gre_receive(struct socket *so, void *arg, int waitflag)
392 {
393 	struct gre_softc *sc = (struct gre_softc *)arg;
394 	int rc;
395 	const struct gre_h *gh;
396 	struct mbuf *m;
397 
398 	GRE_DPRINTF(sc, "%s: enter\n", __func__);
399 
400 	sc->sc_recv_ev.ev_count++;
401 
402 	rc = gre_soreceive(so, &m);
403 	/* TBD Back off if ECONNREFUSED (indicates
404 	 * ICMP Port Unreachable)?
405 	 */
406 	if (rc == EWOULDBLOCK) {
407 		GRE_DPRINTF(sc, "%s: EWOULDBLOCK\n", __func__);
408 		sc->sc_block_ev.ev_count++;
409 		return;
410 	} else if (rc != 0 || m == NULL) {
411 		GRE_DPRINTF(sc, "%s: rc %d m %p\n",
412 		    sc->sc_if.if_xname, rc, (void *)m);
413 		sc->sc_error_ev.ev_count++;
414 		return;
415 	}
416 	if (m->m_len < sizeof(*gh) &&
417 	    (m = m_pullup(m, sizeof(*gh))) == NULL) {
418 		GRE_DPRINTF(sc, "%s: m_pullup failed\n", __func__);
419 		sc->sc_pullup_ev.ev_count++;
420 		return;
421 	}
422 	gh = mtod(m, const struct gre_h *);
423 
424 	if (gre_input(sc, m, 0, gh) == 0) {
425 		sc->sc_unsupp_ev.ev_count++;
426 		GRE_DPRINTF(sc, "%s: dropping unsupported\n", __func__);
427 		m_freem(m);
428 	}
429 }
430 
431 static void
432 gre_upcall_add(struct socket *so, void *arg)
433 {
434 	/* XXX What if the kernel already set an upcall? */
435 	KASSERT((so->so_rcv.sb_flags & SB_UPCALL) == 0);
436 	so->so_upcallarg = arg;
437 	so->so_upcall = gre_receive;
438 	so->so_rcv.sb_flags |= SB_UPCALL;
439 }
440 
441 static void
442 gre_upcall_remove(struct socket *so)
443 {
444 	so->so_rcv.sb_flags &= ~SB_UPCALL;
445 	so->so_upcallarg = NULL;
446 	so->so_upcall = NULL;
447 }
448 
449 static int
450 gre_socreate(struct gre_softc *sc, struct lwp *l,
451     struct gre_soparm *sp, int *fdout)
452 {
453 	const struct protosw *pr;
454 	int fd, rc;
455 	struct mbuf *m;
456 	struct sockaddr *sa;
457 	sa_family_t af;
458 	struct socket *so;
459 
460 	GRE_DPRINTF(sc, "%s: enter\n", __func__);
461 
462 	af = sp->sp_src.ss_family;
463 	rc = fsocreate(af, &so, sp->sp_type, sp->sp_proto, l, &fd);
464 	if (rc != 0) {
465 		GRE_DPRINTF(sc, "%s: fsocreate failed\n", __func__);
466 		return rc;
467 	}
468 
469 	if ((m = getsombuf(so)) == NULL) {
470 		rc = ENOBUFS;
471 		goto out;
472 	}
473 	sa = mtod(m, struct sockaddr *);
474 	sockaddr_copy(sa, MIN(MLEN, sizeof(sp->sp_src)), sstosa(&sp->sp_src));
475 	m->m_len = sp->sp_src.ss_len;
476 
477 #if 0
478 	/* XXX */
479 	GRE_DPRINTF(sc, "%s: bind 0x%08" PRIx32 " port %d\n", __func__,
480 	    sin->sin_addr.s_addr, ntohs(sin->sin_port));
481 #endif
482 	if ((rc = sobind(so, m, l)) != 0) {
483 		GRE_DPRINTF(sc, "%s: sobind failed\n", __func__);
484 		goto out;
485 	}
486 
487 	if ((rc = gre_getsockname(so, m, l)) != 0) {
488 		GRE_DPRINTF(sc, "%s: gre_getsockname\n", __func__);
489 		goto out;
490 	}
491 	sockaddr_copy(sstosa(&sp->sp_src), sizeof(sp->sp_src), sa);
492 
493 	sockaddr_copy(sa, MIN(MLEN, sizeof(sp->sp_dst)), sstosa(&sp->sp_dst));
494 	m->m_len = sp->sp_dst.ss_len;
495 
496 	if ((rc = soconnect(so, m, l)) != 0) {
497 		GRE_DPRINTF(sc, "%s: soconnect failed\n", __func__);
498 		goto out;
499 	}
500 
501 	/* XXX convert to a (new) SOL_SOCKET call */
502 	*mtod(m, int *) = ip_gre_ttl;
503 	m->m_len = sizeof(int);
504 	pr = so->so_proto;
505 	KASSERT(pr != NULL);
506 	rc = sosetopt(so, IPPROTO_IP, IP_TTL, m);
507 	m = NULL;
508 	if (rc != 0) {
509 		GRE_DPRINTF(sc, "%s: sosetopt ttl failed\n", __func__);
510 		rc = 0;
511 	}
512 	rc = sosetopt(so, SOL_SOCKET, SO_NOHEADER, m_intopt(so, 1));
513 	if (rc != 0) {
514 		GRE_DPRINTF(sc, "%s: sosetopt SO_NOHEADER failed\n", __func__);
515 		rc = 0;
516 	}
517 out:
518 	m_freem(m);
519 
520 	if (rc != 0)
521 		fdrelease(l, fd);
522 	else
523 		*fdout = fd;
524 
525 	return rc;
526 }
527 
528 static int
529 gre_sosend(struct socket *so, struct mbuf *top, struct lwp *l)
530 {
531 	struct mbuf	**mp;
532 	struct proc	*p;
533 	long		space, resid;
534 	int		error, s;
535 
536 	p = l->l_proc;
537 
538 	resid = top->m_pkthdr.len;
539 	if (p)
540 		p->p_stats->p_ru.ru_msgsnd++;
541 #define	snderr(errno)	{ error = errno; splx(s); goto release; }
542 
543 	if ((error = sblock(&so->so_snd, M_NOWAIT)) != 0)
544 		goto out;
545 	s = splsoftnet();
546 	if (so->so_state & SS_CANTSENDMORE)
547 		snderr(EPIPE);
548 	if (so->so_error) {
549 		error = so->so_error;
550 		so->so_error = 0;
551 		splx(s);
552 		goto release;
553 	}
554 	if ((so->so_state & SS_ISCONNECTED) == 0) {
555 		if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
556 			if ((so->so_state & SS_ISCONFIRMING) == 0)
557 				snderr(ENOTCONN);
558 		} else
559 			snderr(EDESTADDRREQ);
560 	}
561 	space = sbspace(&so->so_snd);
562 	if (resid > so->so_snd.sb_hiwat)
563 		snderr(EMSGSIZE);
564 	if (space < resid)
565 		snderr(EWOULDBLOCK);
566 	splx(s);
567 	mp = &top;
568 	/*
569 	 * Data is prepackaged in "top".
570 	 */
571 	s = splsoftnet();
572 
573 	if (so->so_state & SS_CANTSENDMORE)
574 		snderr(EPIPE);
575 
576 	error = (*so->so_proto->pr_usrreq)(so, PRU_SEND, top, NULL, NULL, l);
577 	splx(s);
578 
579 	top = NULL;
580 	mp = &top;
581 	if (error != 0)
582 		goto release;
583 
584  release:
585 	sbunlock(&so->so_snd);
586  out:
587 	if (top != NULL)
588 		m_freem(top);
589 	return error;
590 }
591 
592 /* This is a stripped-down version of soreceive() that will never
593  * block.  It will support SOCK_DGRAM sockets.  It may also support
594  * SOCK_SEQPACKET sockets.
595  */
596 static int
597 gre_soreceive(struct socket *so, struct mbuf **mp0)
598 {
599 	struct lwp *l = curlwp;
600 	struct mbuf *m, **mp;
601 	int flags, len, error, s, type;
602 	const struct protosw	*pr;
603 	struct mbuf *nextrecord;
604 
605 	KASSERT(mp0 != NULL);
606 
607 	flags = MSG_DONTWAIT;
608 	pr = so->so_proto;
609 	mp = mp0;
610 	type = 0;
611 
612 	*mp = NULL;
613 
614 	KASSERT(pr->pr_flags & PR_ATOMIC);
615 
616 	if (so->so_state & SS_ISCONFIRMING)
617 		(*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
618 
619  restart:
620 	if ((error = sblock(&so->so_rcv, M_NOWAIT)) != 0)
621 		return error;
622 	s = splsoftnet();
623 
624 	m = so->so_rcv.sb_mb;
625 	/*
626 	 * If we have less data than requested, do not block awaiting more.
627 	 */
628 	if (m == NULL) {
629 #ifdef DIAGNOSTIC
630 		if (so->so_rcv.sb_cc)
631 			panic("receive 1");
632 #endif
633 		if (so->so_error) {
634 			error = so->so_error;
635 			so->so_error = 0;
636 		} else if (so->so_state & SS_CANTRCVMORE)
637 			;
638 		else if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0
639 		      && (so->so_proto->pr_flags & PR_CONNREQUIRED))
640 			error = ENOTCONN;
641 		else
642 			error = EWOULDBLOCK;
643 		goto release;
644 	}
645 	/*
646 	 * On entry here, m points to the first record of the socket buffer.
647 	 * While we process the initial mbufs containing address and control
648 	 * info, we save a copy of m->m_nextpkt into nextrecord.
649 	 */
650 	if (l != NULL)
651 		l->l_proc->p_stats->p_ru.ru_msgrcv++;
652 	KASSERT(m == so->so_rcv.sb_mb);
653 	SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
654 	SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
655 	nextrecord = m->m_nextpkt;
656 	if (pr->pr_flags & PR_ADDR) {
657 #ifdef DIAGNOSTIC
658 		if (m->m_type != MT_SONAME)
659 			panic("receive 1a");
660 #endif
661 		sbfree(&so->so_rcv, m);
662 		MFREE(m, so->so_rcv.sb_mb);
663 		m = so->so_rcv.sb_mb;
664 	}
665 	while (m != NULL && m->m_type == MT_CONTROL && error == 0) {
666 		sbfree(&so->so_rcv, m);
667 		/*
668 		 * Dispose of any SCM_RIGHTS message that went
669 		 * through the read path rather than recv.
670 		 */
671 		if (pr->pr_domain->dom_dispose &&
672 		    mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS)
673 			(*pr->pr_domain->dom_dispose)(m);
674 		MFREE(m, so->so_rcv.sb_mb);
675 		m = so->so_rcv.sb_mb;
676 	}
677 
678 	/*
679 	 * If m is non-NULL, we have some data to read.  From now on,
680 	 * make sure to keep sb_lastrecord consistent when working on
681 	 * the last packet on the chain (nextrecord == NULL) and we
682 	 * change m->m_nextpkt.
683 	 */
684 	if (m != NULL) {
685 		m->m_nextpkt = nextrecord;
686 		/*
687 		 * If nextrecord == NULL (this is a single chain),
688 		 * then sb_lastrecord may not be valid here if m
689 		 * was changed earlier.
690 		 */
691 		if (nextrecord == NULL) {
692 			KASSERT(so->so_rcv.sb_mb == m);
693 			so->so_rcv.sb_lastrecord = m;
694 		}
695 		type = m->m_type;
696 		if (type == MT_OOBDATA)
697 			flags |= MSG_OOB;
698 	} else {
699 		KASSERT(so->so_rcv.sb_mb == m);
700 		so->so_rcv.sb_mb = nextrecord;
701 		SB_EMPTY_FIXUP(&so->so_rcv);
702 	}
703 	SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
704 	SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
705 
706 	while (m != NULL) {
707 		if (m->m_type == MT_OOBDATA) {
708 			if (type != MT_OOBDATA)
709 				break;
710 		} else if (type == MT_OOBDATA)
711 			break;
712 #ifdef DIAGNOSTIC
713 		else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
714 			panic("receive 3");
715 #endif
716 		so->so_state &= ~SS_RCVATMARK;
717 		if (so->so_oobmark != 0 && so->so_oobmark < m->m_len)
718 			break;
719 		len = m->m_len;
720 		/*
721 		 * mp is set, just pass back the mbufs.
722 		 * Sockbuf must be consistent here (points to current mbuf,
723 		 * it points to next record) when we drop priority;
724 		 * we must note any additions to the sockbuf when we
725 		 * block interrupts again.
726 		 */
727 		if (m->m_flags & M_EOR)
728 			flags |= MSG_EOR;
729 		nextrecord = m->m_nextpkt;
730 		sbfree(&so->so_rcv, m);
731 		*mp = m;
732 		mp = &m->m_next;
733 		so->so_rcv.sb_mb = m = m->m_next;
734 		*mp = NULL;
735 		/*
736 		 * If m != NULL, we also know that
737 		 * so->so_rcv.sb_mb != NULL.
738 		 */
739 		KASSERT(so->so_rcv.sb_mb == m);
740 		if (m) {
741 			m->m_nextpkt = nextrecord;
742 			if (nextrecord == NULL)
743 				so->so_rcv.sb_lastrecord = m;
744 		} else {
745 			so->so_rcv.sb_mb = nextrecord;
746 			SB_EMPTY_FIXUP(&so->so_rcv);
747 		}
748 		SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
749 		SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
750 		if (so->so_oobmark) {
751 			so->so_oobmark -= len;
752 			if (so->so_oobmark == 0) {
753 				so->so_state |= SS_RCVATMARK;
754 				break;
755 			}
756 		}
757 		if (flags & MSG_EOR)
758 			break;
759 	}
760 
761 	if (m != NULL) {
762 		m_freem(*mp);
763 		*mp = NULL;
764 		error = ENOMEM;
765 		(void) sbdroprecord(&so->so_rcv);
766 	} else {
767 		/*
768 		 * First part is an inline SB_EMPTY_FIXUP().  Second
769 		 * part makes sure sb_lastrecord is up-to-date if
770 		 * there is still data in the socket buffer.
771 		 */
772 		so->so_rcv.sb_mb = nextrecord;
773 		if (so->so_rcv.sb_mb == NULL) {
774 			so->so_rcv.sb_mbtail = NULL;
775 			so->so_rcv.sb_lastrecord = NULL;
776 		} else if (nextrecord->m_nextpkt == NULL)
777 			so->so_rcv.sb_lastrecord = nextrecord;
778 	}
779 	SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
780 	SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
781 	if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
782 		(*pr->pr_usrreq)(so, PRU_RCVD, NULL,
783 		    (struct mbuf *)(long)flags, NULL, l);
784 	if (*mp0 == NULL && (flags & MSG_EOR) == 0 &&
785 	    (so->so_state & SS_CANTRCVMORE) == 0) {
786 		sbunlock(&so->so_rcv);
787 		splx(s);
788 		goto restart;
789 	}
790 
791  release:
792 	sbunlock(&so->so_rcv);
793 	splx(s);
794 	return error;
795 }
796 
797 static struct socket *
798 gre_reconf(struct gre_softc *sc, struct socket *so, lwp_t *l,
799     const struct gre_soparm *newsoparm)
800 {
801 	int rc;
802 	struct file *fp;
803 	struct ifnet *ifp = &sc->sc_if;
804 
805 	GRE_DPRINTF(sc, "%s: enter\n", __func__);
806 
807 shutdown:
808 	if (sc->sc_soparm.sp_fd != -1) {
809 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
810 		gre_upcall_remove(so);
811 		softint_disestablish(sc->sc_si);
812 		sc->sc_si = NULL;
813 		mutex_exit(&sc->sc_mtx);
814 		fdrelease(l, sc->sc_soparm.sp_fd);
815 		mutex_enter(&sc->sc_mtx);
816 		gre_clearconf(&sc->sc_soparm, false);
817 		sc->sc_soparm.sp_fd = -1;
818 		so = NULL;
819 	}
820 
821 	if (newsoparm != NULL) {
822 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
823 		sc->sc_soparm = *newsoparm;
824 	}
825 
826 	if (sc->sc_soparm.sp_fd != -1) {
827 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
828 		rc = getsock(l->l_proc->p_fd, sc->sc_soparm.sp_fd, &fp);
829 		if (rc != 0)
830 			goto shutdown;
831 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
832 		FILE_UNUSE(fp, NULL);
833 		so = (struct socket *)fp->f_data;
834 		sc->sc_si = softint_establish(SOFTINT_NET, greintr, sc);
835 		gre_upcall_add(so, sc);
836 		if ((ifp->if_flags & IFF_UP) == 0) {
837 			GRE_DPRINTF(sc, "%s: down\n", __func__);
838 			goto shutdown;
839 		}
840 	}
841 
842 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
843 	if (so != NULL)
844 		sc->sc_if.if_flags |= IFF_RUNNING;
845 	else {
846 		gre_bufq_purge(&sc->sc_snd);
847 		sc->sc_if.if_flags &= ~IFF_RUNNING;
848 	}
849 	return so;
850 }
851 
852 static int
853 gre_input(struct gre_softc *sc, struct mbuf *m, int hlen,
854     const struct gre_h *gh)
855 {
856 	u_int16_t flags;
857 	u_int32_t af;		/* af passed to BPF tap */
858 	int isr, s;
859 	struct ifqueue *ifq;
860 
861 	sc->sc_if.if_ipackets++;
862 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
863 
864 	hlen += sizeof(struct gre_h);
865 
866 	/* process GRE flags as packet can be of variable len */
867 	flags = ntohs(gh->flags);
868 
869 	/* Checksum & Offset are present */
870 	if ((flags & GRE_CP) | (flags & GRE_RP))
871 		hlen += 4;
872 	/* We don't support routing fields (variable length) */
873 	if (flags & GRE_RP) {
874 		sc->sc_if.if_ierrors++;
875 		return 0;
876 	}
877 	if (flags & GRE_KP)
878 		hlen += 4;
879 	if (flags & GRE_SP)
880 		hlen += 4;
881 
882 	switch (ntohs(gh->ptype)) { /* ethertypes */
883 	case ETHERTYPE_IP:
884 		ifq = &ipintrq;
885 		isr = NETISR_IP;
886 		af = AF_INET;
887 		break;
888 #ifdef NETATALK
889 	case ETHERTYPE_ATALK:
890 		ifq = &atintrq1;
891 		isr = NETISR_ATALK;
892 		af = AF_APPLETALK;
893 		break;
894 #endif
895 #ifdef INET6
896 	case ETHERTYPE_IPV6:
897 		ifq = &ip6intrq;
898 		isr = NETISR_IPV6;
899 		af = AF_INET6;
900 		break;
901 #endif
902 	default:	   /* others not yet supported */
903 		GRE_DPRINTF(sc, "%s: unhandled ethertype 0x%04x\n", __func__,
904 		    ntohs(gh->ptype));
905 		sc->sc_if.if_noproto++;
906 		return 0;
907 	}
908 
909 	if (hlen > m->m_pkthdr.len) {
910 		m_freem(m);
911 		sc->sc_if.if_ierrors++;
912 		return EINVAL;
913 	}
914 	m_adj(m, hlen);
915 
916 #if NBPFILTER > 0
917 	if (sc->sc_if.if_bpf != NULL)
918 		bpf_mtap_af(sc->sc_if.if_bpf, af, m);
919 #endif /*NBPFILTER > 0*/
920 
921 	m->m_pkthdr.rcvif = &sc->sc_if;
922 
923 	s = splnet();
924 	if (IF_QFULL(ifq)) {
925 		IF_DROP(ifq);
926 		m_freem(m);
927 	} else {
928 		IF_ENQUEUE(ifq, m);
929 	}
930 	/* we need schednetisr since the address family may change */
931 	schednetisr(isr);
932 	splx(s);
933 
934 	return 1;	/* packet is done, no further processing needed */
935 }
936 
937 /*
938  * The output routine. Takes a packet and encapsulates it in the protocol
939  * given by sc->sc_soparm.sp_proto. See also RFC 1701 and RFC 2004
940  */
941 static int
942 gre_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
943 	   struct rtentry *rt)
944 {
945 	int error = 0;
946 	struct gre_softc *sc = ifp->if_softc;
947 	struct gre_h *gh;
948 	struct ip *ip;
949 	u_int8_t ip_tos = 0;
950 	u_int16_t etype = 0;
951 
952 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) {
953 		m_freem(m);
954 		error = ENETDOWN;
955 		goto end;
956 	}
957 
958 #if NBPFILTER > 0
959 	if (ifp->if_bpf != NULL)
960 		bpf_mtap_af(ifp->if_bpf, dst->sa_family, m);
961 #endif
962 
963 	m->m_flags &= ~(M_BCAST|M_MCAST);
964 
965 	GRE_DPRINTF(sc, "%s: dst->sa_family=%d\n", __func__, dst->sa_family);
966 	switch (dst->sa_family) {
967 	case AF_INET:
968 		ip = mtod(m, struct ip *);
969 		ip_tos = ip->ip_tos;
970 		etype = htons(ETHERTYPE_IP);
971 		break;
972 #ifdef NETATALK
973 	case AF_APPLETALK:
974 		etype = htons(ETHERTYPE_ATALK);
975 		break;
976 #endif
977 #ifdef INET6
978 	case AF_INET6:
979 		etype = htons(ETHERTYPE_IPV6);
980 		break;
981 #endif
982 	default:
983 		IF_DROP(&ifp->if_snd);
984 		m_freem(m);
985 		error = EAFNOSUPPORT;
986 		goto end;
987 	}
988 
989 	M_PREPEND(m, sizeof(*gh), M_DONTWAIT);
990 
991 	if (m == NULL) {
992 		IF_DROP(&ifp->if_snd);
993 		error = ENOBUFS;
994 		goto end;
995 	}
996 
997 	gh = mtod(m, struct gre_h *);
998 	gh->flags = 0;
999 	gh->ptype = etype;
1000 	/* XXX Need to handle IP ToS.  Look at how I handle IP TTL. */
1001 
1002 	ifp->if_opackets++;
1003 	ifp->if_obytes += m->m_pkthdr.len;
1004 
1005 	/* send it off */
1006 	if ((error = gre_bufq_enqueue(&sc->sc_snd, m)) != 0) {
1007 		sc->sc_oflow_ev.ev_count++;
1008 		m_freem(m);
1009 	} else
1010 		softint_schedule(sc->sc_si);
1011   end:
1012 	if (error)
1013 		ifp->if_oerrors++;
1014 	return error;
1015 }
1016 
1017 /* Caller must hold sc->sc_mtx. */
1018 static int
1019 gre_getname(struct socket *so, int req, struct mbuf *nam, struct lwp *l)
1020 {
1021 	return (*so->so_proto->pr_usrreq)(so, req, NULL, nam, NULL, l);
1022 }
1023 
1024 /* Caller must hold sc->sc_mtx. */
1025 static int
1026 gre_getsockname(struct socket *so, struct mbuf *nam, struct lwp *l)
1027 {
1028 	return gre_getname(so, PRU_SOCKADDR, nam, l);
1029 }
1030 
1031 /* Caller must hold sc->sc_mtx. */
1032 static int
1033 gre_getpeername(struct socket *so, struct mbuf *nam, struct lwp *l)
1034 {
1035 	return gre_getname(so, PRU_PEERADDR, nam, l);
1036 }
1037 
1038 /* Caller must hold sc->sc_mtx. */
1039 static int
1040 gre_getnames(struct socket *so, struct lwp *l, struct sockaddr_storage *src,
1041     struct sockaddr_storage *dst)
1042 {
1043 	struct mbuf *m;
1044 	struct sockaddr_storage *ss;
1045 	int rc;
1046 
1047 	if ((m = getsombuf(so)) == NULL)
1048 		return ENOBUFS;
1049 
1050 	ss = mtod(m, struct sockaddr_storage *);
1051 
1052 	if ((rc = gre_getsockname(so, m, l)) != 0)
1053 		goto out;
1054 	*src = *ss;
1055 
1056 	if ((rc = gre_getpeername(so, m, l)) != 0)
1057 		goto out;
1058 	*dst = *ss;
1059 
1060 out:
1061 	m_freem(m);
1062 	return rc;
1063 }
1064 
1065 static void
1066 gre_closef(struct file **fpp, struct lwp *l)
1067 {
1068 	struct file *fp = *fpp;
1069 
1070 	mutex_enter(&fp->f_lock);
1071 	FILE_USE(fp);
1072 	closef(fp, l);
1073 	*fpp = NULL;
1074 }
1075 
1076 /* Caller must hold sc->sc_mtx. */
1077 static int
1078 gre_ssock(struct ifnet *ifp, struct gre_soparm *sp, int fd)
1079 {
1080 	int error, kfd;
1081 	const struct protosw *pr;
1082 	struct file *fp;
1083 	struct filedesc	*fdp;
1084 	struct gre_softc *sc = ifp->if_softc;
1085 	struct lwp *l = curlwp;
1086 	struct proc *kp, *p = curproc;
1087 	struct socket *so;
1088 	struct sockaddr_storage dst, src;
1089 
1090 	/* getsock() will FILE_USE() and unlock the descriptor for us */
1091 	if ((error = getsock(p->p_fd, fd, &fp)) != 0) {
1092 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1093 		return EINVAL;
1094 	}
1095 
1096 	/* Increase reference count.  Now that our reference to
1097 	 * the file descriptor is counted, this thread can release
1098 	 * our "use" of the descriptor, but it will not be destroyed
1099 	 * by some other thread's action.  This thread needs to
1100 	 * release its use, too, because one and only one thread
1101 	 * can have use of the descriptor at once.  The kernel
1102 	 * thread will pick up the use if it needs it.
1103 	 */
1104 	fp->f_count++;
1105 	GRE_DPRINTF(sc, "%s: l.%d f_count %d\n", __func__, __LINE__,
1106 	    fp->f_count);
1107 	FILE_UNUSE(fp, NULL);
1108 
1109 	kp = sc->sc_lwp->l_proc;
1110 	while ((error = fdalloc(kp, 0, &kfd)) != 0 && error == ENOSPC)
1111 		fdexpand(kp);
1112 	if (error != 0)
1113 		goto closef;
1114 	fdp = kp->p_fd;
1115 	rw_enter(&fdp->fd_lock, RW_WRITER);
1116 	fdp->fd_ofiles[kfd] = fp;
1117 	rw_exit(&fdp->fd_lock);
1118 
1119 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1120 
1121 	so = (struct socket *)fp->f_data;
1122 	pr = so->so_proto;
1123 	if ((pr->pr_flags & PR_ATOMIC) == 0 ||
1124 	    (sp->sp_type != 0 && pr->pr_type != sp->sp_type) ||
1125 	    (sp->sp_proto != 0 && pr->pr_protocol != 0 &&
1126 	     pr->pr_protocol != sp->sp_proto)) {
1127 		GRE_DPRINTF(sc, "%s: l.%d, type %d, proto %d\n", __func__,
1128 		    __LINE__, pr->pr_type, pr->pr_protocol);
1129 		error = EINVAL;
1130 		goto release;
1131 	}
1132 
1133 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1134 
1135 	/* check address */
1136 	if ((error = gre_getnames(so, l, &src, &dst)) != 0)
1137 		goto release;
1138 
1139 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1140 
1141 	if (error != 0)
1142 		goto release;
1143 
1144 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1145 
1146 	sp->sp_src = src;
1147 	sp->sp_dst = dst;
1148 	/* fp does not any longer belong to this thread. */
1149 	sp->sp_fd = kfd;
1150 
1151 	/* XXX print src & dst */
1152 
1153 	return 0;
1154 release:
1155 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1156 	fdrelease(sc->sc_lwp, kfd);
1157 	return error;
1158 closef:
1159 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1160 	gre_closef(&fp, l);
1161 	return error;
1162 }
1163 
1164 static bool
1165 sockaddr_is_anyaddr(const struct sockaddr *sa)
1166 {
1167 	socklen_t anylen, salen;
1168 	const void *anyaddr, *addr;
1169 
1170 	if ((anyaddr = sockaddr_anyaddr(sa, &anylen)) == NULL ||
1171 	    (addr = sockaddr_const_addr(sa, &salen)) == NULL)
1172 		return false;
1173 
1174 	if (salen > anylen)
1175 		return false;
1176 
1177 	return memcmp(anyaddr, addr, MIN(anylen, salen)) == 0;
1178 }
1179 
1180 static bool
1181 gre_is_nullconf(const struct gre_soparm *sp)
1182 {
1183 	return sockaddr_is_anyaddr(sstocsa(&sp->sp_src)) ||
1184 	       sockaddr_is_anyaddr(sstocsa(&sp->sp_dst));
1185 }
1186 
1187 static void
1188 gre_clearconf(struct gre_soparm *sp, bool force)
1189 {
1190 	if (sp->sp_bysock || force) {
1191 		sockaddr_copy(sstosa(&sp->sp_src), sizeof(sp->sp_src),
1192 		    sockaddr_any(sstosa(&sp->sp_src)));
1193 		sockaddr_copy(sstosa(&sp->sp_dst), sizeof(sp->sp_dst),
1194 		    sockaddr_any(sstosa(&sp->sp_dst)));
1195 		sp->sp_bysock = 0;
1196 	}
1197 	sp->sp_fd = -1;
1198 }
1199 
1200 static int
1201 gre_ioctl(struct ifnet *ifp, const u_long cmd, void *data)
1202 {
1203 	struct lwp *l = curlwp;
1204 	struct ifreq *ifr;
1205 	struct if_laddrreq *lifr = (struct if_laddrreq *)data;
1206 	struct gre_softc *sc = ifp->if_softc;
1207 	struct gre_soparm *sp;
1208 	int fd, error = 0, oproto, otype;
1209 	struct gre_soparm sp0;
1210 
1211 	ifr = data;
1212 
1213 	GRE_DPRINTF(sc, "%s: l.%d, cmd %lu\n", __func__, __LINE__, cmd);
1214 
1215 	switch (cmd) {
1216 	case SIOCSIFFLAGS:
1217 	case SIOCSIFMTU:
1218 	case GRESPROTO:
1219 	case GRESADDRD:
1220 	case GRESADDRS:
1221 	case GRESSOCK:
1222 	case GREDSOCK:
1223 	case SIOCSLIFPHYADDR:
1224 	case SIOCDIFPHYADDR:
1225 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
1226 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
1227 		    NULL) != 0)
1228 			return EPERM;
1229 		break;
1230 	default:
1231 		break;
1232 	}
1233 
1234 	mutex_enter(&sc->sc_mtx);
1235 
1236 	while (sc->sc_state == GRE_S_IOCTL)
1237 		gre_wait(sc);
1238 
1239 	if (sc->sc_state != GRE_S_IDLE) {
1240 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1241 		error = ENXIO;
1242 		goto out;
1243 	}
1244 
1245 	sc->sc_state = GRE_S_IOCTL;
1246 	sp0 = sc->sc_soparm;
1247 	sp0.sp_fd = -1;
1248 	sp = &sp0;
1249 
1250 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1251 
1252 	switch (cmd) {
1253 	case SIOCSIFADDR:
1254 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1255 		if ((ifp->if_flags & IFF_UP) != 0)
1256 			break;
1257 		gre_clearconf(sp, false);
1258 		ifp->if_flags |= IFF_UP;
1259 		goto mksocket;
1260 	case SIOCSIFDSTADDR:
1261 		break;
1262 	case SIOCSIFFLAGS:
1263 		oproto = sp->sp_proto;
1264 		otype = sp->sp_type;
1265 		switch (ifr->ifr_flags & (IFF_LINK0|IFF_LINK2)) {
1266 		case IFF_LINK0|IFF_LINK2:
1267 			sp->sp_proto = IPPROTO_UDP;
1268 			sp->sp_type = SOCK_DGRAM;
1269 			break;
1270 		case IFF_LINK2:
1271 			sp->sp_proto = 0;
1272 			sp->sp_type = 0;
1273 			break;
1274 		case IFF_LINK0:
1275 			sp->sp_proto = IPPROTO_GRE;
1276 			sp->sp_type = SOCK_RAW;
1277 			break;
1278 		default:
1279 			GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1280 			error = EINVAL;
1281 			goto out;
1282 		}
1283 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1284 		gre_clearconf(sp, false);
1285 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
1286 		    (IFF_UP|IFF_RUNNING) &&
1287 		    (oproto == sp->sp_proto || sp->sp_proto == 0) &&
1288 		    (otype == sp->sp_type || sp->sp_type == 0))
1289 			break;
1290 		switch (sp->sp_proto) {
1291 		case IPPROTO_UDP:
1292 		case IPPROTO_GRE:
1293 			goto mksocket;
1294 		default:
1295 			break;
1296 		}
1297 		break;
1298 	case SIOCSIFMTU:
1299 		/* XXX determine MTU automatically by probing w/
1300 		 * XXX do-not-fragment packets?
1301 		 */
1302 		if (ifr->ifr_mtu < 576) {
1303 			error = EINVAL;
1304 			break;
1305 		}
1306 		ifp->if_mtu = ifr->ifr_mtu;
1307 		break;
1308 	case SIOCGIFMTU:
1309 		ifr->ifr_mtu = sc->sc_if.if_mtu;
1310 		break;
1311 	case SIOCADDMULTI:
1312 	case SIOCDELMULTI:
1313 		if (ifr == NULL) {
1314 			error = EAFNOSUPPORT;
1315 			break;
1316 		}
1317 		switch (ifreq_getaddr(cmd, ifr)->sa_family) {
1318 #ifdef INET
1319 		case AF_INET:
1320 			break;
1321 #endif
1322 #ifdef INET6
1323 		case AF_INET6:
1324 			break;
1325 #endif
1326 		default:
1327 			error = EAFNOSUPPORT;
1328 			break;
1329 		}
1330 		break;
1331 	case GRESPROTO:
1332 		gre_clearconf(sp, false);
1333 		oproto = sp->sp_proto;
1334 		otype = sp->sp_type;
1335 		sp->sp_proto = ifr->ifr_flags;
1336 		switch (sp->sp_proto) {
1337 		case IPPROTO_UDP:
1338 			ifp->if_flags |= IFF_LINK0|IFF_LINK2;
1339 			sp->sp_type = SOCK_DGRAM;
1340 			break;
1341 		case IPPROTO_GRE:
1342 			ifp->if_flags |= IFF_LINK0;
1343 			ifp->if_flags &= ~IFF_LINK2;
1344 			sp->sp_type = SOCK_RAW;
1345 			break;
1346 		case 0:
1347 			ifp->if_flags &= ~IFF_LINK0;
1348 			ifp->if_flags |= IFF_LINK2;
1349 			sp->sp_type = 0;
1350 			break;
1351 		default:
1352 			error = EPROTONOSUPPORT;
1353 			break;
1354 		}
1355 		if ((oproto == sp->sp_proto || sp->sp_proto == 0) &&
1356 		    (otype == sp->sp_type || sp->sp_type == 0))
1357 			break;
1358 		switch (sp->sp_proto) {
1359 		case IPPROTO_UDP:
1360 		case IPPROTO_GRE:
1361 			goto mksocket;
1362 		default:
1363 			break;
1364 		}
1365 		break;
1366 	case GREGPROTO:
1367 		ifr->ifr_flags = sp->sp_proto;
1368 		break;
1369 	case GRESADDRS:
1370 	case GRESADDRD:
1371 		gre_clearconf(sp, false);
1372 		/*
1373 		 * set tunnel endpoints, compute a less specific route
1374 		 * to the remote end and mark if as up
1375 		 */
1376 		switch (cmd) {
1377 		case GRESADDRS:
1378 			sockaddr_copy(sstosa(&sp->sp_src),
1379 			    sizeof(sp->sp_src), ifreq_getaddr(cmd, ifr));
1380 			break;
1381 		case GRESADDRD:
1382 			sockaddr_copy(sstosa(&sp->sp_dst),
1383 			    sizeof(sp->sp_dst), ifreq_getaddr(cmd, ifr));
1384 			break;
1385 		}
1386 	checkaddr:
1387 		if (sockaddr_any(sstosa(&sp->sp_src)) == NULL ||
1388 		    sockaddr_any(sstosa(&sp->sp_dst)) == NULL) {
1389 			error = EINVAL;
1390 			break;
1391 		}
1392 		/* let gre_socreate() check the rest */
1393 	mksocket:
1394 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1395 		/* If we're administratively down, or the configuration
1396 		 * is empty, there's no use creating a socket.
1397 		 */
1398 		if ((ifp->if_flags & IFF_UP) == 0 || gre_is_nullconf(sp))
1399 			goto sendconf;
1400 
1401 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1402 		mutex_exit(&sc->sc_mtx);
1403 		error = gre_socreate(sc, l, sp, &fd);
1404 		mutex_enter(&sc->sc_mtx);
1405 
1406 		if (error != 0)
1407 			break;
1408 
1409 	setsock:
1410 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1411 
1412 		mutex_exit(&sc->sc_mtx);
1413 		error = gre_ssock(ifp, sp, fd);
1414 
1415 		if (cmd != GRESSOCK) {
1416 			GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1417 			fdrelease(l, fd);
1418 		}
1419 
1420 		mutex_enter(&sc->sc_mtx);
1421 
1422 		if (error == 0) {
1423 	sendconf:
1424 			GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1425 			ifp->if_flags &= ~IFF_RUNNING;
1426 			sc->sc_so = gre_reconf(sc, sc->sc_so, sc->sc_lwp, sp);
1427 		}
1428 
1429 		break;
1430 	case GREGADDRS:
1431 		ifreq_setaddr(cmd, ifr, sstosa(&sp->sp_src));
1432 		break;
1433 	case GREGADDRD:
1434 		ifreq_setaddr(cmd, ifr, sstosa(&sp->sp_dst));
1435 		break;
1436 	case GREDSOCK:
1437 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1438 		if (sp->sp_bysock)
1439 			ifp->if_flags &= ~IFF_UP;
1440 		gre_clearconf(sp, false);
1441 		goto mksocket;
1442 	case GRESSOCK:
1443 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1444 		gre_clearconf(sp, true);
1445 		fd = (int)ifr->ifr_value;
1446 		sp->sp_bysock = 1;
1447 		ifp->if_flags |= IFF_UP;
1448 		goto setsock;
1449 	case SIOCSLIFPHYADDR:
1450 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1451 		if (lifr->addr.ss_family != lifr->dstaddr.ss_family) {
1452 			error = EAFNOSUPPORT;
1453 			break;
1454 		}
1455 		sockaddr_copy(sstosa(&sp->sp_src), sizeof(sp->sp_src),
1456 		    sstosa(&lifr->addr));
1457 		sockaddr_copy(sstosa(&sp->sp_dst), sizeof(sp->sp_dst),
1458 		    sstosa(&lifr->dstaddr));
1459 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1460 		goto checkaddr;
1461 	case SIOCDIFPHYADDR:
1462 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1463 		gre_clearconf(sp, true);
1464 		ifp->if_flags &= ~IFF_UP;
1465 		goto mksocket;
1466 	case SIOCGLIFPHYADDR:
1467 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1468 		if (gre_is_nullconf(sp)) {
1469 			error = EADDRNOTAVAIL;
1470 			break;
1471 		}
1472 		sockaddr_copy(sstosa(&lifr->addr), sizeof(lifr->addr),
1473 		    sstosa(&sp->sp_src));
1474 		sockaddr_copy(sstosa(&lifr->dstaddr), sizeof(lifr->dstaddr),
1475 		    sstosa(&sp->sp_dst));
1476 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1477 		break;
1478 	default:
1479 		error = EINVAL;
1480 		break;
1481 	}
1482 out:
1483 	GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1484 	if (sc->sc_state == GRE_S_IOCTL) {
1485 		GRE_DPRINTF(sc, "%s: l.%d\n", __func__, __LINE__);
1486 		sc->sc_state = GRE_S_IDLE;
1487 	}
1488 	cv_signal(&sc->sc_condvar);
1489 	mutex_exit(&sc->sc_mtx);
1490 	return error;
1491 }
1492 
1493 #endif
1494 
1495 void	greattach(int);
1496 
1497 /* ARGSUSED */
1498 void
1499 greattach(int count)
1500 {
1501 #ifdef INET
1502 	if_clone_attach(&gre_cloner);
1503 #endif
1504 }
1505