xref: /titanic_44/usr/src/uts/common/inet/ip/conn_opt.c (revision ac3e5be0db2e7e4fabda0a139e10da4702077a9d)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/strsun.h>
30 #define	_SUN_TPI_VERSION 2
31 #include <sys/tihdr.h>
32 #include <sys/xti_inet.h>
33 #include <sys/ucred.h>
34 #include <sys/zone.h>
35 #include <sys/ddi.h>
36 #include <sys/sunddi.h>
37 #include <sys/cmn_err.h>
38 #include <sys/debug.h>
39 #include <sys/atomic.h>
40 #include <sys/policy.h>
41 
42 #include <sys/systm.h>
43 #include <sys/param.h>
44 #include <sys/kmem.h>
45 #include <sys/sdt.h>
46 #include <sys/socket.h>
47 #include <sys/ethernet.h>
48 #include <sys/mac.h>
49 #include <net/if.h>
50 #include <net/if_types.h>
51 #include <net/if_arp.h>
52 #include <net/route.h>
53 #include <sys/sockio.h>
54 #include <netinet/in.h>
55 #include <net/if_dl.h>
56 
57 #include <inet/common.h>
58 #include <inet/mi.h>
59 #include <inet/mib2.h>
60 #include <inet/nd.h>
61 #include <inet/arp.h>
62 #include <inet/snmpcom.h>
63 #include <inet/kstatcom.h>
64 
65 #include <netinet/igmp_var.h>
66 #include <netinet/ip6.h>
67 #include <netinet/icmp6.h>
68 #include <netinet/sctp.h>
69 
70 #include <inet/ip.h>
71 #include <inet/ip_impl.h>
72 #include <inet/ip6.h>
73 #include <inet/ip6_asp.h>
74 #include <inet/tcp.h>
75 #include <inet/ip_multi.h>
76 #include <inet/ip_if.h>
77 #include <inet/ip_ire.h>
78 #include <inet/ip_ftable.h>
79 #include <inet/ip_rts.h>
80 #include <inet/optcom.h>
81 #include <inet/ip_ndp.h>
82 #include <inet/ip_listutils.h>
83 #include <netinet/igmp.h>
84 #include <netinet/ip_mroute.h>
85 #include <netinet/udp.h>
86 #include <inet/ipp_common.h>
87 
88 #include <net/pfkeyv2.h>
89 #include <inet/sadb.h>
90 #include <inet/ipsec_impl.h>
91 #include <inet/ipdrop.h>
92 #include <inet/ip_netinfo.h>
93 
94 #include <inet/ipclassifier.h>
95 #include <inet/sctp_ip.h>
96 #include <inet/sctp/sctp_impl.h>
97 #include <inet/udp_impl.h>
98 #include <sys/sunddi.h>
99 
100 #include <sys/tsol/label.h>
101 #include <sys/tsol/tnet.h>
102 
103 /*
104  * Return how much size is needed for the different ancillary data items
105  */
106 uint_t
conn_recvancillary_size(conn_t * connp,crb_t recv_ancillary,ip_recv_attr_t * ira,mblk_t * mp,ip_pkt_t * ipp)107 conn_recvancillary_size(conn_t *connp, crb_t recv_ancillary,
108     ip_recv_attr_t *ira, mblk_t *mp, ip_pkt_t *ipp)
109 {
110 	uint_t		ancil_size;
111 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
112 
113 	/*
114 	 * If IP_RECVDSTADDR is set we include the destination IP
115 	 * address as an option. With IP_RECVOPTS we include all
116 	 * the IP options.
117 	 */
118 	ancil_size = 0;
119 	if (recv_ancillary.crb_recvdstaddr &&
120 	    (ira->ira_flags & IRAF_IS_IPV4)) {
121 		ancil_size += sizeof (struct T_opthdr) +
122 		    sizeof (struct in_addr);
123 		IP_STAT(ipst, conn_in_recvdstaddr);
124 	}
125 
126 	/*
127 	 * ip_recvpktinfo is used for both AF_INET and AF_INET6 but
128 	 * are different
129 	 */
130 	if (recv_ancillary.crb_ip_recvpktinfo &&
131 	    connp->conn_family == AF_INET) {
132 		ancil_size += sizeof (struct T_opthdr) +
133 		    sizeof (struct in_pktinfo);
134 		IP_STAT(ipst, conn_in_recvpktinfo);
135 	}
136 
137 	if ((recv_ancillary.crb_recvopts) &&
138 	    (ipp->ipp_fields & IPPF_IPV4_OPTIONS)) {
139 		ancil_size += sizeof (struct T_opthdr) +
140 		    ipp->ipp_ipv4_options_len;
141 		IP_STAT(ipst, conn_in_recvopts);
142 	}
143 
144 	if (recv_ancillary.crb_recvslla) {
145 		ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
146 		ill_t *ill;
147 
148 		/* Make sure ira_l2src is setup if not already */
149 		if (!(ira->ira_flags & IRAF_L2SRC_SET)) {
150 			ill = ill_lookup_on_ifindex(ira->ira_rifindex, B_FALSE,
151 			    ipst);
152 			if (ill != NULL) {
153 				ip_setl2src(mp, ira, ill);
154 				ill_refrele(ill);
155 			}
156 		}
157 		ancil_size += sizeof (struct T_opthdr) +
158 		    sizeof (struct sockaddr_dl);
159 		IP_STAT(ipst, conn_in_recvslla);
160 	}
161 
162 	if (recv_ancillary.crb_recvif) {
163 		ancil_size += sizeof (struct T_opthdr) + sizeof (uint_t);
164 		IP_STAT(ipst, conn_in_recvif);
165 	}
166 
167 	/*
168 	 * ip_recvpktinfo is used for both AF_INET and AF_INET6 but
169 	 * are different
170 	 */
171 	if (recv_ancillary.crb_ip_recvpktinfo &&
172 	    connp->conn_family == AF_INET6) {
173 		ancil_size += sizeof (struct T_opthdr) +
174 		    sizeof (struct in6_pktinfo);
175 		IP_STAT(ipst, conn_in_recvpktinfo);
176 	}
177 
178 	if (recv_ancillary.crb_ipv6_recvhoplimit) {
179 		ancil_size += sizeof (struct T_opthdr) + sizeof (int);
180 		IP_STAT(ipst, conn_in_recvhoplimit);
181 	}
182 
183 	if (recv_ancillary.crb_ipv6_recvtclass) {
184 		ancil_size += sizeof (struct T_opthdr) + sizeof (int);
185 		IP_STAT(ipst, conn_in_recvtclass);
186 	}
187 
188 	if (recv_ancillary.crb_ipv6_recvhopopts &&
189 	    (ipp->ipp_fields & IPPF_HOPOPTS)) {
190 		ancil_size += sizeof (struct T_opthdr) + ipp->ipp_hopoptslen;
191 		IP_STAT(ipst, conn_in_recvhopopts);
192 	}
193 	/*
194 	 * To honor RFC3542 when an application asks for both IPV6_RECVDSTOPTS
195 	 * and IPV6_RECVRTHDR, we pass up the item rthdrdstopts (the destination
196 	 * options that appear before a routing header.
197 	 * We also pass them up if IPV6_RECVRTHDRDSTOPTS is set.
198 	 */
199 	if (ipp->ipp_fields & IPPF_RTHDRDSTOPTS) {
200 		if (recv_ancillary.crb_ipv6_recvrthdrdstopts ||
201 		    (recv_ancillary.crb_ipv6_recvdstopts &&
202 		    recv_ancillary.crb_ipv6_recvrthdr)) {
203 			ancil_size += sizeof (struct T_opthdr) +
204 			    ipp->ipp_rthdrdstoptslen;
205 			IP_STAT(ipst, conn_in_recvrthdrdstopts);
206 		}
207 	}
208 	if ((recv_ancillary.crb_ipv6_recvrthdr) &&
209 	    (ipp->ipp_fields & IPPF_RTHDR)) {
210 		ancil_size += sizeof (struct T_opthdr) + ipp->ipp_rthdrlen;
211 		IP_STAT(ipst, conn_in_recvrthdr);
212 	}
213 	if ((recv_ancillary.crb_ipv6_recvdstopts ||
214 	    recv_ancillary.crb_old_ipv6_recvdstopts) &&
215 	    (ipp->ipp_fields & IPPF_DSTOPTS)) {
216 		ancil_size += sizeof (struct T_opthdr) + ipp->ipp_dstoptslen;
217 		IP_STAT(ipst, conn_in_recvdstopts);
218 	}
219 	if (recv_ancillary.crb_recvucred && ira->ira_cred != NULL) {
220 		ancil_size += sizeof (struct T_opthdr) +
221 		    ucredminsize(ira->ira_cred);
222 		IP_STAT(ipst, conn_in_recvucred);
223 	}
224 
225 	/*
226 	 * If SO_TIMESTAMP is set allocate the appropriate sized
227 	 * buffer. Since gethrestime() expects a pointer aligned
228 	 * argument, we allocate space necessary for extra
229 	 * alignment (even though it might not be used).
230 	 */
231 	if (recv_ancillary.crb_timestamp) {
232 		ancil_size += sizeof (struct T_opthdr) +
233 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
234 		IP_STAT(ipst, conn_in_timestamp);
235 	}
236 
237 	/*
238 	 * If IP_RECVTTL is set allocate the appropriate sized buffer
239 	 */
240 	if (recv_ancillary.crb_recvttl &&
241 	    (ira->ira_flags & IRAF_IS_IPV4)) {
242 		ancil_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
243 		IP_STAT(ipst, conn_in_recvttl);
244 	}
245 
246 	return (ancil_size);
247 }
248 
249 /*
250  * Lay down the ancillary data items at "ancil_buf".
251  * Assumes caller has used conn_recvancillary_size to allocate a sufficiently
252  * large buffer - ancil_size.
253  */
254 void
conn_recvancillary_add(conn_t * connp,crb_t recv_ancillary,ip_recv_attr_t * ira,ip_pkt_t * ipp,uchar_t * ancil_buf,uint_t ancil_size)255 conn_recvancillary_add(conn_t *connp, crb_t recv_ancillary,
256     ip_recv_attr_t *ira, ip_pkt_t *ipp, uchar_t *ancil_buf, uint_t ancil_size)
257 {
258 	/*
259 	 * Copy in destination address before options to avoid
260 	 * any padding issues.
261 	 */
262 	if (recv_ancillary.crb_recvdstaddr &&
263 	    (ira->ira_flags & IRAF_IS_IPV4)) {
264 		struct T_opthdr *toh;
265 		ipaddr_t *dstptr;
266 
267 		toh = (struct T_opthdr *)ancil_buf;
268 		toh->level = IPPROTO_IP;
269 		toh->name = IP_RECVDSTADDR;
270 		toh->len = sizeof (struct T_opthdr) + sizeof (ipaddr_t);
271 		toh->status = 0;
272 		ancil_buf += sizeof (struct T_opthdr);
273 		dstptr = (ipaddr_t *)ancil_buf;
274 		*dstptr = ipp->ipp_addr_v4;
275 		ancil_buf += sizeof (ipaddr_t);
276 		ancil_size -= toh->len;
277 	}
278 
279 	/*
280 	 * ip_recvpktinfo is used for both AF_INET and AF_INET6 but
281 	 * are different
282 	 */
283 	if (recv_ancillary.crb_ip_recvpktinfo &&
284 	    connp->conn_family == AF_INET) {
285 		ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
286 		struct T_opthdr *toh;
287 		struct in_pktinfo *pktinfop;
288 		ill_t *ill;
289 		ipif_t *ipif;
290 
291 		toh = (struct T_opthdr *)ancil_buf;
292 		toh->level = IPPROTO_IP;
293 		toh->name = IP_PKTINFO;
294 		toh->len = sizeof (struct T_opthdr) + sizeof (*pktinfop);
295 		toh->status = 0;
296 		ancil_buf += sizeof (struct T_opthdr);
297 		pktinfop = (struct in_pktinfo *)ancil_buf;
298 
299 		pktinfop->ipi_ifindex = ira->ira_ruifindex;
300 		pktinfop->ipi_spec_dst.s_addr = INADDR_ANY;
301 
302 		/* Find a good address to report */
303 		ill = ill_lookup_on_ifindex(ira->ira_ruifindex, B_FALSE, ipst);
304 		if (ill != NULL) {
305 			ipif = ipif_good_addr(ill, IPCL_ZONEID(connp));
306 			if (ipif != NULL) {
307 				pktinfop->ipi_spec_dst.s_addr =
308 				    ipif->ipif_lcl_addr;
309 				ipif_refrele(ipif);
310 			}
311 			ill_refrele(ill);
312 		}
313 		pktinfop->ipi_addr.s_addr = ipp->ipp_addr_v4;
314 		ancil_buf += sizeof (struct in_pktinfo);
315 		ancil_size -= toh->len;
316 	}
317 
318 	if ((recv_ancillary.crb_recvopts) &&
319 	    (ipp->ipp_fields & IPPF_IPV4_OPTIONS)) {
320 		struct T_opthdr *toh;
321 
322 		toh = (struct T_opthdr *)ancil_buf;
323 		toh->level = IPPROTO_IP;
324 		toh->name = IP_RECVOPTS;
325 		toh->len = sizeof (struct T_opthdr) + ipp->ipp_ipv4_options_len;
326 		toh->status = 0;
327 		ancil_buf += sizeof (struct T_opthdr);
328 		bcopy(ipp->ipp_ipv4_options, ancil_buf,
329 		    ipp->ipp_ipv4_options_len);
330 		ancil_buf += ipp->ipp_ipv4_options_len;
331 		ancil_size -= toh->len;
332 	}
333 
334 	if (recv_ancillary.crb_recvslla) {
335 		ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
336 		struct T_opthdr *toh;
337 		struct sockaddr_dl *dstptr;
338 		ill_t *ill;
339 		int alen = 0;
340 
341 		ill = ill_lookup_on_ifindex(ira->ira_rifindex, B_FALSE, ipst);
342 		if (ill != NULL)
343 			alen = ill->ill_phys_addr_length;
344 
345 		/*
346 		 * For loopback multicast and broadcast the packet arrives
347 		 * with ira_ruifdex being the physical interface, but
348 		 * ira_l2src is all zero since ip_postfrag_loopback doesn't
349 		 * know our l2src. We don't report the address in that case.
350 		 */
351 		if (ira->ira_flags & IRAF_LOOPBACK)
352 			alen = 0;
353 
354 		toh = (struct T_opthdr *)ancil_buf;
355 		toh->level = IPPROTO_IP;
356 		toh->name = IP_RECVSLLA;
357 		toh->len = sizeof (struct T_opthdr) +
358 		    sizeof (struct sockaddr_dl);
359 		toh->status = 0;
360 		ancil_buf += sizeof (struct T_opthdr);
361 		dstptr = (struct sockaddr_dl *)ancil_buf;
362 		dstptr->sdl_family = AF_LINK;
363 		dstptr->sdl_index = ira->ira_ruifindex;
364 		if (ill != NULL)
365 			dstptr->sdl_type = ill->ill_type;
366 		else
367 			dstptr->sdl_type = 0;
368 		dstptr->sdl_nlen = 0;
369 		dstptr->sdl_alen = alen;
370 		dstptr->sdl_slen = 0;
371 		bcopy(ira->ira_l2src, dstptr->sdl_data, alen);
372 		ancil_buf += sizeof (struct sockaddr_dl);
373 		ancil_size -= toh->len;
374 		if (ill != NULL)
375 			ill_refrele(ill);
376 	}
377 
378 	if (recv_ancillary.crb_recvif) {
379 		struct T_opthdr *toh;
380 		uint_t		*dstptr;
381 
382 		toh = (struct T_opthdr *)ancil_buf;
383 		toh->level = IPPROTO_IP;
384 		toh->name = IP_RECVIF;
385 		toh->len = sizeof (struct T_opthdr) + sizeof (uint_t);
386 		toh->status = 0;
387 		ancil_buf += sizeof (struct T_opthdr);
388 		dstptr = (uint_t *)ancil_buf;
389 		*dstptr = ira->ira_ruifindex;
390 		ancil_buf += sizeof (uint_t);
391 		ancil_size -= toh->len;
392 	}
393 
394 	/*
395 	 * ip_recvpktinfo is used for both AF_INET and AF_INET6 but
396 	 * are different
397 	 */
398 	if (recv_ancillary.crb_ip_recvpktinfo &&
399 	    connp->conn_family == AF_INET6) {
400 		struct T_opthdr *toh;
401 		struct in6_pktinfo *pkti;
402 
403 		toh = (struct T_opthdr *)ancil_buf;
404 		toh->level = IPPROTO_IPV6;
405 		toh->name = IPV6_PKTINFO;
406 		toh->len = sizeof (struct T_opthdr) + sizeof (*pkti);
407 		toh->status = 0;
408 		ancil_buf += sizeof (struct T_opthdr);
409 		pkti = (struct in6_pktinfo *)ancil_buf;
410 		if (ira->ira_flags & IRAF_IS_IPV4) {
411 			IN6_IPADDR_TO_V4MAPPED(ipp->ipp_addr_v4,
412 			    &pkti->ipi6_addr);
413 		} else {
414 			pkti->ipi6_addr = ipp->ipp_addr;
415 		}
416 		pkti->ipi6_ifindex = ira->ira_ruifindex;
417 
418 		ancil_buf += sizeof (*pkti);
419 		ancil_size -= toh->len;
420 	}
421 	if (recv_ancillary.crb_ipv6_recvhoplimit) {
422 		struct T_opthdr *toh;
423 
424 		toh = (struct T_opthdr *)ancil_buf;
425 		toh->level = IPPROTO_IPV6;
426 		toh->name = IPV6_HOPLIMIT;
427 		toh->len = sizeof (struct T_opthdr) + sizeof (uint_t);
428 		toh->status = 0;
429 		ancil_buf += sizeof (struct T_opthdr);
430 		*(uint_t *)ancil_buf = ipp->ipp_hoplimit;
431 		ancil_buf += sizeof (uint_t);
432 		ancil_size -= toh->len;
433 	}
434 	if (recv_ancillary.crb_ipv6_recvtclass) {
435 		struct T_opthdr *toh;
436 
437 		toh = (struct T_opthdr *)ancil_buf;
438 		toh->level = IPPROTO_IPV6;
439 		toh->name = IPV6_TCLASS;
440 		toh->len = sizeof (struct T_opthdr) + sizeof (uint_t);
441 		toh->status = 0;
442 		ancil_buf += sizeof (struct T_opthdr);
443 
444 		if (ira->ira_flags & IRAF_IS_IPV4)
445 			*(uint_t *)ancil_buf = ipp->ipp_type_of_service;
446 		else
447 			*(uint_t *)ancil_buf = ipp->ipp_tclass;
448 		ancil_buf += sizeof (uint_t);
449 		ancil_size -= toh->len;
450 	}
451 	if (recv_ancillary.crb_ipv6_recvhopopts &&
452 	    (ipp->ipp_fields & IPPF_HOPOPTS)) {
453 		struct T_opthdr *toh;
454 
455 		toh = (struct T_opthdr *)ancil_buf;
456 		toh->level = IPPROTO_IPV6;
457 		toh->name = IPV6_HOPOPTS;
458 		toh->len = sizeof (struct T_opthdr) + ipp->ipp_hopoptslen;
459 		toh->status = 0;
460 		ancil_buf += sizeof (struct T_opthdr);
461 		bcopy(ipp->ipp_hopopts, ancil_buf, ipp->ipp_hopoptslen);
462 		ancil_buf += ipp->ipp_hopoptslen;
463 		ancil_size -= toh->len;
464 	}
465 	/*
466 	 * To honor RFC3542 when an application asks for both IPV6_RECVDSTOPTS
467 	 * and IPV6_RECVRTHDR, we pass up the item rthdrdstopts (the destination
468 	 * options that appear before a routing header.
469 	 * We also pass them up if IPV6_RECVRTHDRDSTOPTS is set.
470 	 */
471 	if (ipp->ipp_fields & IPPF_RTHDRDSTOPTS) {
472 		if (recv_ancillary.crb_ipv6_recvrthdrdstopts ||
473 		    (recv_ancillary.crb_ipv6_recvdstopts &&
474 		    recv_ancillary.crb_ipv6_recvrthdr)) {
475 			struct T_opthdr *toh;
476 
477 			toh = (struct T_opthdr *)ancil_buf;
478 			toh->level = IPPROTO_IPV6;
479 			toh->name = IPV6_DSTOPTS;
480 			toh->len = sizeof (struct T_opthdr) +
481 			    ipp->ipp_rthdrdstoptslen;
482 			toh->status = 0;
483 			ancil_buf += sizeof (struct T_opthdr);
484 			bcopy(ipp->ipp_rthdrdstopts, ancil_buf,
485 			    ipp->ipp_rthdrdstoptslen);
486 			ancil_buf += ipp->ipp_rthdrdstoptslen;
487 			ancil_size -= toh->len;
488 		}
489 	}
490 	if (recv_ancillary.crb_ipv6_recvrthdr &&
491 	    (ipp->ipp_fields & IPPF_RTHDR)) {
492 		struct T_opthdr *toh;
493 
494 		toh = (struct T_opthdr *)ancil_buf;
495 		toh->level = IPPROTO_IPV6;
496 		toh->name = IPV6_RTHDR;
497 		toh->len = sizeof (struct T_opthdr) + ipp->ipp_rthdrlen;
498 		toh->status = 0;
499 		ancil_buf += sizeof (struct T_opthdr);
500 		bcopy(ipp->ipp_rthdr, ancil_buf, ipp->ipp_rthdrlen);
501 		ancil_buf += ipp->ipp_rthdrlen;
502 		ancil_size -= toh->len;
503 	}
504 	if ((recv_ancillary.crb_ipv6_recvdstopts ||
505 	    recv_ancillary.crb_old_ipv6_recvdstopts) &&
506 	    (ipp->ipp_fields & IPPF_DSTOPTS)) {
507 		struct T_opthdr *toh;
508 
509 		toh = (struct T_opthdr *)ancil_buf;
510 		toh->level = IPPROTO_IPV6;
511 		toh->name = IPV6_DSTOPTS;
512 		toh->len = sizeof (struct T_opthdr) + ipp->ipp_dstoptslen;
513 		toh->status = 0;
514 		ancil_buf += sizeof (struct T_opthdr);
515 		bcopy(ipp->ipp_dstopts, ancil_buf, ipp->ipp_dstoptslen);
516 		ancil_buf += ipp->ipp_dstoptslen;
517 		ancil_size -= toh->len;
518 	}
519 
520 	if (recv_ancillary.crb_recvucred && ira->ira_cred != NULL) {
521 		struct T_opthdr *toh;
522 		cred_t		*rcr = connp->conn_cred;
523 
524 		toh = (struct T_opthdr *)ancil_buf;
525 		toh->level = SOL_SOCKET;
526 		toh->name = SCM_UCRED;
527 		toh->len = sizeof (struct T_opthdr) +
528 		    ucredminsize(ira->ira_cred);
529 		toh->status = 0;
530 		(void) cred2ucred(ira->ira_cred, ira->ira_cpid, &toh[1], rcr);
531 		ancil_buf += toh->len;
532 		ancil_size -= toh->len;
533 	}
534 	if (recv_ancillary.crb_timestamp) {
535 		struct	T_opthdr *toh;
536 
537 		toh = (struct T_opthdr *)ancil_buf;
538 		toh->level = SOL_SOCKET;
539 		toh->name = SCM_TIMESTAMP;
540 		toh->len = sizeof (struct T_opthdr) +
541 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
542 		toh->status = 0;
543 		ancil_buf += sizeof (struct T_opthdr);
544 		/* Align for gethrestime() */
545 		ancil_buf = (uchar_t *)P2ROUNDUP((intptr_t)ancil_buf,
546 		    sizeof (intptr_t));
547 		gethrestime((timestruc_t *)ancil_buf);
548 		ancil_buf = (uchar_t *)toh + toh->len;
549 		ancil_size -= toh->len;
550 	}
551 
552 	/*
553 	 * CAUTION:
554 	 * Due to aligment issues
555 	 * Processing of IP_RECVTTL option
556 	 * should always be the last. Adding
557 	 * any option processing after this will
558 	 * cause alignment panic.
559 	 */
560 	if (recv_ancillary.crb_recvttl &&
561 	    (ira->ira_flags & IRAF_IS_IPV4)) {
562 		struct	T_opthdr *toh;
563 		uint8_t	*dstptr;
564 
565 		toh = (struct T_opthdr *)ancil_buf;
566 		toh->level = IPPROTO_IP;
567 		toh->name = IP_RECVTTL;
568 		toh->len = sizeof (struct T_opthdr) + sizeof (uint8_t);
569 		toh->status = 0;
570 		ancil_buf += sizeof (struct T_opthdr);
571 		dstptr = (uint8_t *)ancil_buf;
572 		*dstptr = ipp->ipp_hoplimit;
573 		ancil_buf += sizeof (uint8_t);
574 		ancil_size -= toh->len;
575 	}
576 
577 	/* Consumed all of allocated space */
578 	ASSERT(ancil_size == 0);
579 
580 }
581 
582 /*
583  * This routine retrieves the current status of socket options.
584  * It returns the size of the option retrieved, or -1.
585  */
586 int
conn_opt_get(conn_opt_arg_t * coa,t_scalar_t level,t_scalar_t name,uchar_t * ptr)587 conn_opt_get(conn_opt_arg_t *coa, t_scalar_t level, t_scalar_t name,
588     uchar_t *ptr)
589 {
590 	int		*i1 = (int *)ptr;
591 	conn_t		*connp = coa->coa_connp;
592 	ip_xmit_attr_t	*ixa = coa->coa_ixa;
593 	ip_pkt_t	*ipp = coa->coa_ipp;
594 	ip_stack_t	*ipst = ixa->ixa_ipst;
595 	uint_t		len;
596 
597 	ASSERT(MUTEX_HELD(&coa->coa_connp->conn_lock));
598 
599 	switch (level) {
600 	case SOL_SOCKET:
601 		switch (name) {
602 		case SO_DEBUG:
603 			*i1 = connp->conn_debug ? SO_DEBUG : 0;
604 			break;	/* goto sizeof (int) option return */
605 		case SO_KEEPALIVE:
606 			*i1 = connp->conn_keepalive ? SO_KEEPALIVE : 0;
607 			break;
608 		case SO_LINGER:	{
609 			struct linger *lgr = (struct linger *)ptr;
610 
611 			lgr->l_onoff = connp->conn_linger ? SO_LINGER : 0;
612 			lgr->l_linger = connp->conn_lingertime;
613 			}
614 			return (sizeof (struct linger));
615 
616 		case SO_OOBINLINE:
617 			*i1 = connp->conn_oobinline ? SO_OOBINLINE : 0;
618 			break;
619 		case SO_REUSEADDR:
620 			*i1 = connp->conn_reuseaddr ? SO_REUSEADDR : 0;
621 			break;	/* goto sizeof (int) option return */
622 		case SO_TYPE:
623 			*i1 = connp->conn_so_type;
624 			break;	/* goto sizeof (int) option return */
625 		case SO_DONTROUTE:
626 			*i1 = (ixa->ixa_flags & IXAF_DONTROUTE) ?
627 			    SO_DONTROUTE : 0;
628 			break;	/* goto sizeof (int) option return */
629 		case SO_USELOOPBACK:
630 			*i1 = connp->conn_useloopback ? SO_USELOOPBACK : 0;
631 			break;	/* goto sizeof (int) option return */
632 		case SO_BROADCAST:
633 			*i1 = connp->conn_broadcast ? SO_BROADCAST : 0;
634 			break;	/* goto sizeof (int) option return */
635 
636 		case SO_SNDBUF:
637 			*i1 = connp->conn_sndbuf;
638 			break;	/* goto sizeof (int) option return */
639 		case SO_RCVBUF:
640 			*i1 = connp->conn_rcvbuf;
641 			break;	/* goto sizeof (int) option return */
642 		case SO_RCVTIMEO:
643 		case SO_SNDTIMEO:
644 			/*
645 			 * Pass these two options in order for third part
646 			 * protocol usage. Here just return directly.
647 			 */
648 			*i1 = 0;
649 			break;
650 		case SO_DGRAM_ERRIND:
651 			*i1 = connp->conn_dgram_errind ? SO_DGRAM_ERRIND : 0;
652 			break;	/* goto sizeof (int) option return */
653 		case SO_RECVUCRED:
654 			*i1 = connp->conn_recv_ancillary.crb_recvucred;
655 			break;	/* goto sizeof (int) option return */
656 		case SO_TIMESTAMP:
657 			*i1 = connp->conn_recv_ancillary.crb_timestamp;
658 			break;	/* goto sizeof (int) option return */
659 		case SO_VRRP:
660 			*i1 = connp->conn_isvrrp;
661 			break;	/* goto sizeof (int) option return */
662 		case SO_ANON_MLP:
663 			*i1 = connp->conn_anon_mlp;
664 			break;	/* goto sizeof (int) option return */
665 		case SO_MAC_EXEMPT:
666 			*i1 = (connp->conn_mac_mode == CONN_MAC_AWARE);
667 			break;	/* goto sizeof (int) option return */
668 		case SO_MAC_IMPLICIT:
669 			*i1 = (connp->conn_mac_mode == CONN_MAC_IMPLICIT);
670 			break;	/* goto sizeof (int) option return */
671 		case SO_ALLZONES:
672 			*i1 = connp->conn_allzones;
673 			break;	/* goto sizeof (int) option return */
674 		case SO_EXCLBIND:
675 			*i1 = connp->conn_exclbind ? SO_EXCLBIND : 0;
676 			break;
677 		case SO_PROTOTYPE:
678 			*i1 = connp->conn_proto;
679 			break;
680 
681 		case SO_DOMAIN:
682 			*i1 = connp->conn_family;
683 			break;
684 		default:
685 			return (-1);
686 		}
687 		break;
688 	case IPPROTO_IP:
689 		if (connp->conn_family != AF_INET)
690 			return (-1);
691 		switch (name) {
692 		case IP_OPTIONS:
693 		case T_IP_OPTIONS:
694 			if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
695 				return (0);
696 
697 			len = ipp->ipp_ipv4_options_len;
698 			if (len > 0) {
699 				bcopy(ipp->ipp_ipv4_options, ptr, len);
700 			}
701 			return (len);
702 
703 		case IP_PKTINFO: {
704 			/*
705 			 * This also handles IP_RECVPKTINFO.
706 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
707 			 * Differentiation is based on the size of the
708 			 * argument passed in.
709 			 */
710 			struct in_pktinfo *pktinfo;
711 
712 #ifdef notdef
713 			/* optcom doesn't provide a length with "get" */
714 			if (inlen == sizeof (int)) {
715 				/* This is IP_RECVPKTINFO option. */
716 				*i1 = connp->conn_recv_ancillary.
717 				    crb_ip_recvpktinfo;
718 				return (sizeof (int));
719 			}
720 #endif
721 			/* XXX assumes that caller has room for max size! */
722 
723 			pktinfo = (struct in_pktinfo *)ptr;
724 			pktinfo->ipi_ifindex = ixa->ixa_ifindex;
725 			if (ipp->ipp_fields & IPPF_ADDR)
726 				pktinfo->ipi_spec_dst.s_addr = ipp->ipp_addr_v4;
727 			else
728 				pktinfo->ipi_spec_dst.s_addr = INADDR_ANY;
729 			return (sizeof (struct in_pktinfo));
730 		}
731 		case IP_DONTFRAG:
732 			*i1 = (ixa->ixa_flags & IXAF_DONTFRAG) != 0;
733 			return (sizeof (int));
734 		case IP_TOS:
735 		case T_IP_TOS:
736 			*i1 = (int)ipp->ipp_type_of_service;
737 			break;	/* goto sizeof (int) option return */
738 		case IP_TTL:
739 			*i1 = (int)ipp->ipp_unicast_hops;
740 			break;	/* goto sizeof (int) option return */
741 		case IP_DHCPINIT_IF:
742 			return (-1);
743 		case IP_NEXTHOP:
744 			if (ixa->ixa_flags & IXAF_NEXTHOP_SET) {
745 				*(ipaddr_t *)ptr = ixa->ixa_nexthop_v4;
746 				return (sizeof (ipaddr_t));
747 			} else {
748 				return (0);
749 			}
750 
751 		case IP_MULTICAST_IF:
752 			/* 0 address if not set */
753 			*(ipaddr_t *)ptr = ixa->ixa_multicast_ifaddr;
754 			return (sizeof (ipaddr_t));
755 		case IP_MULTICAST_TTL:
756 			*(uchar_t *)ptr = ixa->ixa_multicast_ttl;
757 			return (sizeof (uchar_t));
758 		case IP_MULTICAST_LOOP:
759 			*ptr = (ixa->ixa_flags & IXAF_MULTICAST_LOOP) ? 1 : 0;
760 			return (sizeof (uint8_t));
761 		case IP_RECVOPTS:
762 			*i1 = connp->conn_recv_ancillary.crb_recvopts;
763 			break;	/* goto sizeof (int) option return */
764 		case IP_RECVDSTADDR:
765 			*i1 = connp->conn_recv_ancillary.crb_recvdstaddr;
766 			break;	/* goto sizeof (int) option return */
767 		case IP_RECVIF:
768 			*i1 = connp->conn_recv_ancillary.crb_recvif;
769 			break;	/* goto sizeof (int) option return */
770 		case IP_RECVSLLA:
771 			*i1 = connp->conn_recv_ancillary.crb_recvslla;
772 			break;	/* goto sizeof (int) option return */
773 		case IP_RECVTTL:
774 			*i1 = connp->conn_recv_ancillary.crb_recvttl;
775 			break;	/* goto sizeof (int) option return */
776 		case IP_ADD_MEMBERSHIP:
777 		case IP_DROP_MEMBERSHIP:
778 		case MCAST_JOIN_GROUP:
779 		case MCAST_LEAVE_GROUP:
780 		case IP_BLOCK_SOURCE:
781 		case IP_UNBLOCK_SOURCE:
782 		case IP_ADD_SOURCE_MEMBERSHIP:
783 		case IP_DROP_SOURCE_MEMBERSHIP:
784 		case MCAST_BLOCK_SOURCE:
785 		case MCAST_UNBLOCK_SOURCE:
786 		case MCAST_JOIN_SOURCE_GROUP:
787 		case MCAST_LEAVE_SOURCE_GROUP:
788 		case MRT_INIT:
789 		case MRT_DONE:
790 		case MRT_ADD_VIF:
791 		case MRT_DEL_VIF:
792 		case MRT_ADD_MFC:
793 		case MRT_DEL_MFC:
794 			/* cannot "get" the value for these */
795 			return (-1);
796 		case MRT_VERSION:
797 		case MRT_ASSERT:
798 			(void) ip_mrouter_get(name, connp, ptr);
799 			return (sizeof (int));
800 		case IP_SEC_OPT:
801 			return (ipsec_req_from_conn(connp, (ipsec_req_t	*)ptr,
802 			    IPSEC_AF_V4));
803 		case IP_BOUND_IF:
804 			/* Zero if not set */
805 			*i1 = connp->conn_bound_if;
806 			break;	/* goto sizeof (int) option return */
807 		case IP_UNSPEC_SRC:
808 			*i1 = connp->conn_unspec_src;
809 			break;	/* goto sizeof (int) option return */
810 		case IP_BROADCAST_TTL:
811 			if (ixa->ixa_flags & IXAF_BROADCAST_TTL_SET)
812 				*(uchar_t *)ptr = ixa->ixa_broadcast_ttl;
813 			else
814 				*(uchar_t *)ptr = ipst->ips_ip_broadcast_ttl;
815 			return (sizeof (uchar_t));
816 		default:
817 			return (-1);
818 		}
819 		break;
820 	case IPPROTO_IPV6:
821 		if (connp->conn_family != AF_INET6)
822 			return (-1);
823 		switch (name) {
824 		case IPV6_UNICAST_HOPS:
825 			*i1 = (int)ipp->ipp_unicast_hops;
826 			break;	/* goto sizeof (int) option return */
827 		case IPV6_MULTICAST_IF:
828 			/* 0 index if not set */
829 			*i1 = ixa->ixa_multicast_ifindex;
830 			break;	/* goto sizeof (int) option return */
831 		case IPV6_MULTICAST_HOPS:
832 			*i1 = ixa->ixa_multicast_ttl;
833 			break;	/* goto sizeof (int) option return */
834 		case IPV6_MULTICAST_LOOP:
835 			*i1 = (ixa->ixa_flags & IXAF_MULTICAST_LOOP) ? 1 : 0;
836 			break;	/* goto sizeof (int) option return */
837 		case IPV6_JOIN_GROUP:
838 		case IPV6_LEAVE_GROUP:
839 		case MCAST_JOIN_GROUP:
840 		case MCAST_LEAVE_GROUP:
841 		case MCAST_BLOCK_SOURCE:
842 		case MCAST_UNBLOCK_SOURCE:
843 		case MCAST_JOIN_SOURCE_GROUP:
844 		case MCAST_LEAVE_SOURCE_GROUP:
845 			/* cannot "get" the value for these */
846 			return (-1);
847 		case IPV6_BOUND_IF:
848 			/* Zero if not set */
849 			*i1 = connp->conn_bound_if;
850 			break;	/* goto sizeof (int) option return */
851 		case IPV6_UNSPEC_SRC:
852 			*i1 = connp->conn_unspec_src;
853 			break;	/* goto sizeof (int) option return */
854 		case IPV6_RECVPKTINFO:
855 			*i1 = connp->conn_recv_ancillary.crb_ip_recvpktinfo;
856 			break;	/* goto sizeof (int) option return */
857 		case IPV6_RECVTCLASS:
858 			*i1 = connp->conn_recv_ancillary.crb_ipv6_recvtclass;
859 			break;	/* goto sizeof (int) option return */
860 		case IPV6_RECVPATHMTU:
861 			*i1 = connp->conn_ipv6_recvpathmtu;
862 			break;	/* goto sizeof (int) option return */
863 		case IPV6_RECVHOPLIMIT:
864 			*i1 = connp->conn_recv_ancillary.crb_ipv6_recvhoplimit;
865 			break;	/* goto sizeof (int) option return */
866 		case IPV6_RECVHOPOPTS:
867 			*i1 = connp->conn_recv_ancillary.crb_ipv6_recvhopopts;
868 			break;	/* goto sizeof (int) option return */
869 		case IPV6_RECVDSTOPTS:
870 			*i1 = connp->conn_recv_ancillary.crb_ipv6_recvdstopts;
871 			break;	/* goto sizeof (int) option return */
872 		case _OLD_IPV6_RECVDSTOPTS:
873 			*i1 =
874 			    connp->conn_recv_ancillary.crb_old_ipv6_recvdstopts;
875 			break;	/* goto sizeof (int) option return */
876 		case IPV6_RECVRTHDRDSTOPTS:
877 			*i1 = connp->conn_recv_ancillary.
878 			    crb_ipv6_recvrthdrdstopts;
879 			break;	/* goto sizeof (int) option return */
880 		case IPV6_RECVRTHDR:
881 			*i1 = connp->conn_recv_ancillary.crb_ipv6_recvrthdr;
882 			break;	/* goto sizeof (int) option return */
883 		case IPV6_PKTINFO: {
884 			/* XXX assumes that caller has room for max size! */
885 			struct in6_pktinfo *pkti;
886 
887 			pkti = (struct in6_pktinfo *)ptr;
888 			pkti->ipi6_ifindex = ixa->ixa_ifindex;
889 			if (ipp->ipp_fields & IPPF_ADDR)
890 				pkti->ipi6_addr = ipp->ipp_addr;
891 			else
892 				pkti->ipi6_addr = ipv6_all_zeros;
893 			return (sizeof (struct in6_pktinfo));
894 		}
895 		case IPV6_TCLASS:
896 			*i1 = ipp->ipp_tclass;
897 			break;	/* goto sizeof (int) option return */
898 		case IPV6_NEXTHOP: {
899 			sin6_t *sin6 = (sin6_t *)ptr;
900 
901 			if (ixa->ixa_flags & IXAF_NEXTHOP_SET)
902 				return (0);
903 
904 			*sin6 = sin6_null;
905 			sin6->sin6_family = AF_INET6;
906 			sin6->sin6_addr = ixa->ixa_nexthop_v6;
907 
908 			return (sizeof (sin6_t));
909 		}
910 		case IPV6_HOPOPTS:
911 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
912 				return (0);
913 			bcopy(ipp->ipp_hopopts, ptr,
914 			    ipp->ipp_hopoptslen);
915 			return (ipp->ipp_hopoptslen);
916 		case IPV6_RTHDRDSTOPTS:
917 			if (!(ipp->ipp_fields & IPPF_RTHDRDSTOPTS))
918 				return (0);
919 			bcopy(ipp->ipp_rthdrdstopts, ptr,
920 			    ipp->ipp_rthdrdstoptslen);
921 			return (ipp->ipp_rthdrdstoptslen);
922 		case IPV6_RTHDR:
923 			if (!(ipp->ipp_fields & IPPF_RTHDR))
924 				return (0);
925 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
926 			return (ipp->ipp_rthdrlen);
927 		case IPV6_DSTOPTS:
928 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
929 				return (0);
930 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
931 			return (ipp->ipp_dstoptslen);
932 		case IPV6_PATHMTU:
933 			return (ip_fill_mtuinfo(connp, ixa,
934 			    (struct ip6_mtuinfo *)ptr));
935 		case IPV6_SEC_OPT:
936 			return (ipsec_req_from_conn(connp, (ipsec_req_t	*)ptr,
937 			    IPSEC_AF_V6));
938 		case IPV6_SRC_PREFERENCES:
939 			return (ip6_get_src_preferences(ixa, (uint32_t *)ptr));
940 		case IPV6_DONTFRAG:
941 			*i1 = (ixa->ixa_flags & IXAF_DONTFRAG) != 0;
942 			return (sizeof (int));
943 		case IPV6_USE_MIN_MTU:
944 			if (ixa->ixa_flags & IXAF_USE_MIN_MTU)
945 				*i1 = ixa->ixa_use_min_mtu;
946 			else
947 				*i1 = IPV6_USE_MIN_MTU_MULTICAST;
948 			break;
949 		case IPV6_V6ONLY:
950 			*i1 = connp->conn_ipv6_v6only;
951 			return (sizeof (int));
952 		default:
953 			return (-1);
954 		}
955 		break;
956 	case IPPROTO_UDP:
957 		switch (name) {
958 		case UDP_ANONPRIVBIND:
959 			*i1 = connp->conn_anon_priv_bind;
960 			break;
961 		case UDP_EXCLBIND:
962 			*i1 = connp->conn_exclbind ? UDP_EXCLBIND : 0;
963 			break;
964 		default:
965 			return (-1);
966 		}
967 		break;
968 	case IPPROTO_TCP:
969 		switch (name) {
970 		case TCP_RECVDSTADDR:
971 			*i1 = connp->conn_recv_ancillary.crb_recvdstaddr;
972 			break;
973 		case TCP_ANONPRIVBIND:
974 			*i1 = connp->conn_anon_priv_bind;
975 			break;
976 		case TCP_EXCLBIND:
977 			*i1 = connp->conn_exclbind ? TCP_EXCLBIND : 0;
978 			break;
979 		default:
980 			return (-1);
981 		}
982 		break;
983 	default:
984 		return (-1);
985 	}
986 	return (sizeof (int));
987 }
988 
989 static int conn_opt_set_socket(conn_opt_arg_t *coa, t_scalar_t name,
990     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr);
991 static int conn_opt_set_ip(conn_opt_arg_t *coa, t_scalar_t name,
992     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr);
993 static int conn_opt_set_ipv6(conn_opt_arg_t *coa, t_scalar_t name,
994     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr);
995 static int conn_opt_set_udp(conn_opt_arg_t *coa, t_scalar_t name,
996     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr);
997 static int conn_opt_set_tcp(conn_opt_arg_t *coa, t_scalar_t name,
998     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr);
999 
1000 /*
1001  * This routine sets the most common socket options including some
1002  * that are transport/ULP specific.
1003  * It returns errno or zero.
1004  *
1005  * For fixed length options, there is no sanity check
1006  * of passed in length is done. It is assumed *_optcom_req()
1007  * routines do the right thing.
1008  */
1009 int
conn_opt_set(conn_opt_arg_t * coa,t_scalar_t level,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)1010 conn_opt_set(conn_opt_arg_t *coa, t_scalar_t level, t_scalar_t name,
1011     uint_t inlen, uchar_t *invalp, boolean_t checkonly, cred_t *cr)
1012 {
1013 	ASSERT(MUTEX_NOT_HELD(&coa->coa_connp->conn_lock));
1014 
1015 	/* We have different functions for different levels */
1016 	switch (level) {
1017 	case SOL_SOCKET:
1018 		return (conn_opt_set_socket(coa, name, inlen, invalp,
1019 		    checkonly, cr));
1020 	case IPPROTO_IP:
1021 		return (conn_opt_set_ip(coa, name, inlen, invalp,
1022 		    checkonly, cr));
1023 	case IPPROTO_IPV6:
1024 		return (conn_opt_set_ipv6(coa, name, inlen, invalp,
1025 		    checkonly, cr));
1026 	case IPPROTO_UDP:
1027 		return (conn_opt_set_udp(coa, name, inlen, invalp,
1028 		    checkonly, cr));
1029 	case IPPROTO_TCP:
1030 		return (conn_opt_set_tcp(coa, name, inlen, invalp,
1031 		    checkonly, cr));
1032 	default:
1033 		return (0);
1034 	}
1035 }
1036 
1037 /*
1038  * Handle SOL_SOCKET
1039  * Note that we do not handle SO_PROTOTYPE here. The ULPs that support
1040  * it implement their own checks and setting of conn_proto.
1041  */
1042 /* ARGSUSED1 */
1043 static int
conn_opt_set_socket(conn_opt_arg_t * coa,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)1044 conn_opt_set_socket(conn_opt_arg_t *coa, t_scalar_t name, uint_t inlen,
1045     uchar_t *invalp, boolean_t checkonly, cred_t *cr)
1046 {
1047 	conn_t		*connp = coa->coa_connp;
1048 	ip_xmit_attr_t	*ixa = coa->coa_ixa;
1049 	int		*i1 = (int *)invalp;
1050 	boolean_t	onoff = (*i1 == 0) ? 0 : 1;
1051 
1052 	switch (name) {
1053 	case SO_ALLZONES:
1054 		if (IPCL_IS_BOUND(connp))
1055 			return (EINVAL);
1056 		break;
1057 	case SO_VRRP:
1058 		if (secpolicy_ip_config(cr, checkonly) != 0)
1059 			return (EACCES);
1060 		break;
1061 	case SO_MAC_EXEMPT:
1062 		if (secpolicy_net_mac_aware(cr) != 0)
1063 			return (EACCES);
1064 		if (IPCL_IS_BOUND(connp))
1065 			return (EINVAL);
1066 		break;
1067 	case SO_MAC_IMPLICIT:
1068 		if (secpolicy_net_mac_implicit(cr) != 0)
1069 			return (EACCES);
1070 		break;
1071 	}
1072 	if (checkonly)
1073 		return (0);
1074 
1075 	mutex_enter(&connp->conn_lock);
1076 	/* Here we set the actual option value */
1077 	switch (name) {
1078 	case SO_DEBUG:
1079 		connp->conn_debug = onoff;
1080 		break;
1081 	case SO_KEEPALIVE:
1082 		connp->conn_keepalive = onoff;
1083 		break;
1084 	case SO_LINGER: {
1085 		struct linger *lgr = (struct linger *)invalp;
1086 
1087 		if (lgr->l_onoff) {
1088 			connp->conn_linger = 1;
1089 			connp->conn_lingertime = lgr->l_linger;
1090 		} else {
1091 			connp->conn_linger = 0;
1092 			connp->conn_lingertime = 0;
1093 		}
1094 		break;
1095 	}
1096 	case SO_OOBINLINE:
1097 		connp->conn_oobinline = onoff;
1098 		coa->coa_changed |= COA_OOBINLINE_CHANGED;
1099 		break;
1100 	case SO_REUSEADDR:
1101 		connp->conn_reuseaddr = onoff;
1102 		break;
1103 	case SO_DONTROUTE:
1104 		if (onoff)
1105 			ixa->ixa_flags |= IXAF_DONTROUTE;
1106 		else
1107 			ixa->ixa_flags &= ~IXAF_DONTROUTE;
1108 		coa->coa_changed |= COA_ROUTE_CHANGED;
1109 		break;
1110 	case SO_USELOOPBACK:
1111 		connp->conn_useloopback = onoff;
1112 		break;
1113 	case SO_BROADCAST:
1114 		connp->conn_broadcast = onoff;
1115 		break;
1116 	case SO_SNDBUF:
1117 		/* ULP has range checked the value */
1118 		connp->conn_sndbuf = *i1;
1119 		coa->coa_changed |= COA_SNDBUF_CHANGED;
1120 		break;
1121 	case SO_RCVBUF:
1122 		/* ULP has range checked the value */
1123 		connp->conn_rcvbuf = *i1;
1124 		coa->coa_changed |= COA_RCVBUF_CHANGED;
1125 		break;
1126 	case SO_RCVTIMEO:
1127 	case SO_SNDTIMEO:
1128 		/*
1129 		 * Pass these two options in order for third part
1130 		 * protocol usage.
1131 		 */
1132 		break;
1133 	case SO_DGRAM_ERRIND:
1134 		connp->conn_dgram_errind = onoff;
1135 		break;
1136 	case SO_RECVUCRED:
1137 		connp->conn_recv_ancillary.crb_recvucred = onoff;
1138 		break;
1139 	case SO_ALLZONES:
1140 		connp->conn_allzones = onoff;
1141 		coa->coa_changed |= COA_ROUTE_CHANGED;
1142 		if (onoff)
1143 			ixa->ixa_zoneid = ALL_ZONES;
1144 		else
1145 			ixa->ixa_zoneid = connp->conn_zoneid;
1146 		break;
1147 	case SO_TIMESTAMP:
1148 		connp->conn_recv_ancillary.crb_timestamp = onoff;
1149 		break;
1150 	case SO_VRRP:
1151 		connp->conn_isvrrp = onoff;
1152 		break;
1153 	case SO_ANON_MLP:
1154 		connp->conn_anon_mlp = onoff;
1155 		break;
1156 	case SO_MAC_EXEMPT:
1157 		connp->conn_mac_mode = onoff ?
1158 		    CONN_MAC_AWARE : CONN_MAC_DEFAULT;
1159 		break;
1160 	case SO_MAC_IMPLICIT:
1161 		connp->conn_mac_mode = onoff ?
1162 		    CONN_MAC_IMPLICIT : CONN_MAC_DEFAULT;
1163 		break;
1164 	case SO_EXCLBIND:
1165 		connp->conn_exclbind = onoff;
1166 		break;
1167 	}
1168 	mutex_exit(&connp->conn_lock);
1169 	return (0);
1170 }
1171 
1172 /* Handle IPPROTO_IP */
1173 static int
conn_opt_set_ip(conn_opt_arg_t * coa,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)1174 conn_opt_set_ip(conn_opt_arg_t *coa, t_scalar_t name, uint_t inlen,
1175     uchar_t *invalp, boolean_t checkonly, cred_t *cr)
1176 {
1177 	conn_t		*connp = coa->coa_connp;
1178 	ip_xmit_attr_t	*ixa = coa->coa_ixa;
1179 	ip_pkt_t	*ipp = coa->coa_ipp;
1180 	int		*i1 = (int *)invalp;
1181 	boolean_t	onoff = (*i1 == 0) ? 0 : 1;
1182 	ipaddr_t	addr = (ipaddr_t)*i1;
1183 	uint_t		ifindex;
1184 	zoneid_t	zoneid = IPCL_ZONEID(connp);
1185 	ipif_t		*ipif;
1186 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
1187 	int		error;
1188 
1189 	if (connp->conn_family != AF_INET)
1190 		return (EINVAL);
1191 
1192 	switch (name) {
1193 	case IP_TTL:
1194 		/* Don't allow zero */
1195 		if (*i1 < 1 || *i1 > 255)
1196 			return (EINVAL);
1197 		break;
1198 	case IP_MULTICAST_IF:
1199 		if (addr == INADDR_ANY) {
1200 			/* Clear */
1201 			ifindex = 0;
1202 			break;
1203 		}
1204 		ipif = ipif_lookup_addr(addr, NULL, zoneid, ipst);
1205 		if (ipif == NULL)
1206 			return (EHOSTUNREACH);
1207 		/* not supported by the virtual network iface */
1208 		if (IS_VNI(ipif->ipif_ill)) {
1209 			ipif_refrele(ipif);
1210 			return (EINVAL);
1211 		}
1212 		ifindex = ipif->ipif_ill->ill_phyint->phyint_ifindex;
1213 		ipif_refrele(ipif);
1214 		break;
1215 	case IP_NEXTHOP: {
1216 		ire_t	*ire;
1217 
1218 		if (addr == INADDR_ANY) {
1219 			/* Clear */
1220 			break;
1221 		}
1222 		/* Verify that the next-hop is on-link */
1223 		ire = ire_ftable_lookup_v4(addr, 0, 0, IRE_ONLINK, NULL, zoneid,
1224 		    NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
1225 		if (ire == NULL)
1226 			return (EHOSTUNREACH);
1227 		ire_refrele(ire);
1228 		break;
1229 	}
1230 	case IP_OPTIONS:
1231 	case T_IP_OPTIONS: {
1232 		uint_t newlen;
1233 
1234 		if (ipp->ipp_fields & IPPF_LABEL_V4)
1235 			newlen = inlen + (ipp->ipp_label_len_v4 + 3) & ~3;
1236 		else
1237 			newlen = inlen;
1238 		if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
1239 			return (EINVAL);
1240 		}
1241 		break;
1242 	}
1243 	case IP_PKTINFO: {
1244 		struct in_pktinfo *pktinfo;
1245 
1246 		/* Two different valid lengths */
1247 		if (inlen != sizeof (int) &&
1248 		    inlen != sizeof (struct in_pktinfo))
1249 			return (EINVAL);
1250 		if (inlen == sizeof (int))
1251 			break;
1252 
1253 		pktinfo = (struct in_pktinfo *)invalp;
1254 		if (pktinfo->ipi_spec_dst.s_addr != INADDR_ANY) {
1255 			switch (ip_laddr_verify_v4(pktinfo->ipi_spec_dst.s_addr,
1256 			    zoneid, ipst, B_FALSE)) {
1257 			case IPVL_UNICAST_UP:
1258 			case IPVL_UNICAST_DOWN:
1259 				break;
1260 			default:
1261 				return (EADDRNOTAVAIL);
1262 			}
1263 		}
1264 		if (!ip_xmit_ifindex_valid(pktinfo->ipi_ifindex, zoneid,
1265 		    B_FALSE, ipst))
1266 			return (ENXIO);
1267 		break;
1268 	}
1269 	case IP_BOUND_IF:
1270 		ifindex = *(uint_t *)i1;
1271 
1272 		/* Just check it is ok. */
1273 		if (!ip_xmit_ifindex_valid(ifindex, zoneid, B_FALSE, ipst))
1274 			return (ENXIO);
1275 		break;
1276 	}
1277 	if (checkonly)
1278 		return (0);
1279 
1280 	/* Here we set the actual option value */
1281 	/*
1282 	 * conn_lock protects the bitfields, and is used to
1283 	 * set the fields atomically. Not needed for ixa settings since
1284 	 * the caller has an exclusive copy of the ixa.
1285 	 * We can not hold conn_lock across the multicast options though.
1286 	 */
1287 	switch (name) {
1288 	case IP_OPTIONS:
1289 	case T_IP_OPTIONS:
1290 		/* Save options for use by IP. */
1291 		mutex_enter(&connp->conn_lock);
1292 		error = optcom_pkt_set(invalp, inlen,
1293 		    (uchar_t **)&ipp->ipp_ipv4_options,
1294 		    &ipp->ipp_ipv4_options_len);
1295 		if (error != 0) {
1296 			mutex_exit(&connp->conn_lock);
1297 			return (error);
1298 		}
1299 		if (ipp->ipp_ipv4_options_len == 0) {
1300 			ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
1301 		} else {
1302 			ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
1303 		}
1304 		mutex_exit(&connp->conn_lock);
1305 		coa->coa_changed |= COA_HEADER_CHANGED;
1306 		coa->coa_changed |= COA_WROFF_CHANGED;
1307 		break;
1308 
1309 	case IP_TTL:
1310 		mutex_enter(&connp->conn_lock);
1311 		ipp->ipp_unicast_hops = *i1;
1312 		mutex_exit(&connp->conn_lock);
1313 		coa->coa_changed |= COA_HEADER_CHANGED;
1314 		break;
1315 	case IP_TOS:
1316 	case T_IP_TOS:
1317 		mutex_enter(&connp->conn_lock);
1318 		if (*i1 == -1) {
1319 			ipp->ipp_type_of_service = 0;
1320 		} else {
1321 			ipp->ipp_type_of_service = *i1;
1322 		}
1323 		mutex_exit(&connp->conn_lock);
1324 		coa->coa_changed |= COA_HEADER_CHANGED;
1325 		break;
1326 	case IP_MULTICAST_IF:
1327 		ixa->ixa_multicast_ifindex = ifindex;
1328 		ixa->ixa_multicast_ifaddr = addr;
1329 		coa->coa_changed |= COA_ROUTE_CHANGED;
1330 		break;
1331 	case IP_MULTICAST_TTL:
1332 		ixa->ixa_multicast_ttl = *invalp;
1333 		/* Handled automatically by ip_output */
1334 		break;
1335 	case IP_MULTICAST_LOOP:
1336 		if (*invalp != 0)
1337 			ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
1338 		else
1339 			ixa->ixa_flags &= ~IXAF_MULTICAST_LOOP;
1340 		/* Handled automatically by ip_output */
1341 		break;
1342 	case IP_RECVOPTS:
1343 		mutex_enter(&connp->conn_lock);
1344 		connp->conn_recv_ancillary.crb_recvopts = onoff;
1345 		mutex_exit(&connp->conn_lock);
1346 		break;
1347 	case IP_RECVDSTADDR:
1348 		mutex_enter(&connp->conn_lock);
1349 		connp->conn_recv_ancillary.crb_recvdstaddr = onoff;
1350 		mutex_exit(&connp->conn_lock);
1351 		break;
1352 	case IP_RECVIF:
1353 		mutex_enter(&connp->conn_lock);
1354 		connp->conn_recv_ancillary.crb_recvif = onoff;
1355 		mutex_exit(&connp->conn_lock);
1356 		break;
1357 	case IP_RECVSLLA:
1358 		mutex_enter(&connp->conn_lock);
1359 		connp->conn_recv_ancillary.crb_recvslla = onoff;
1360 		mutex_exit(&connp->conn_lock);
1361 		break;
1362 	case IP_RECVTTL:
1363 		mutex_enter(&connp->conn_lock);
1364 		connp->conn_recv_ancillary.crb_recvttl = onoff;
1365 		mutex_exit(&connp->conn_lock);
1366 		break;
1367 	case IP_PKTINFO: {
1368 		/*
1369 		 * This also handles IP_RECVPKTINFO.
1370 		 * IP_PKTINFO and IP_RECVPKTINFO have same value.
1371 		 * Differentiation is based on the size of the
1372 		 * argument passed in.
1373 		 */
1374 		struct in_pktinfo *pktinfo;
1375 
1376 		if (inlen == sizeof (int)) {
1377 			/* This is IP_RECVPKTINFO option. */
1378 			mutex_enter(&connp->conn_lock);
1379 			connp->conn_recv_ancillary.crb_ip_recvpktinfo =
1380 			    onoff;
1381 			mutex_exit(&connp->conn_lock);
1382 			break;
1383 		}
1384 
1385 		/* This is IP_PKTINFO option. */
1386 		mutex_enter(&connp->conn_lock);
1387 		pktinfo = (struct in_pktinfo *)invalp;
1388 		if (pktinfo->ipi_spec_dst.s_addr != INADDR_ANY) {
1389 			ipp->ipp_fields |= IPPF_ADDR;
1390 			IN6_INADDR_TO_V4MAPPED(&pktinfo->ipi_spec_dst,
1391 			    &ipp->ipp_addr);
1392 		} else {
1393 			ipp->ipp_fields &= ~IPPF_ADDR;
1394 			ipp->ipp_addr = ipv6_all_zeros;
1395 		}
1396 		mutex_exit(&connp->conn_lock);
1397 		ixa->ixa_ifindex = pktinfo->ipi_ifindex;
1398 		coa->coa_changed |= COA_ROUTE_CHANGED;
1399 		coa->coa_changed |= COA_HEADER_CHANGED;
1400 		break;
1401 	}
1402 	case IP_DONTFRAG:
1403 		if (onoff) {
1404 			ixa->ixa_flags |= (IXAF_DONTFRAG | IXAF_PMTU_IPV4_DF);
1405 			ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
1406 		} else {
1407 			ixa->ixa_flags &= ~(IXAF_DONTFRAG | IXAF_PMTU_IPV4_DF);
1408 			ixa->ixa_flags |= IXAF_PMTU_DISCOVERY;
1409 		}
1410 		/* Need to redo ip_attr_connect */
1411 		coa->coa_changed |= COA_ROUTE_CHANGED;
1412 		break;
1413 	case IP_ADD_MEMBERSHIP:
1414 	case IP_DROP_MEMBERSHIP:
1415 	case MCAST_JOIN_GROUP:
1416 	case MCAST_LEAVE_GROUP:
1417 		return (ip_opt_set_multicast_group(connp, name,
1418 		    invalp, B_FALSE, checkonly));
1419 
1420 	case IP_BLOCK_SOURCE:
1421 	case IP_UNBLOCK_SOURCE:
1422 	case IP_ADD_SOURCE_MEMBERSHIP:
1423 	case IP_DROP_SOURCE_MEMBERSHIP:
1424 	case MCAST_BLOCK_SOURCE:
1425 	case MCAST_UNBLOCK_SOURCE:
1426 	case MCAST_JOIN_SOURCE_GROUP:
1427 	case MCAST_LEAVE_SOURCE_GROUP:
1428 		return (ip_opt_set_multicast_sources(connp, name,
1429 		    invalp, B_FALSE, checkonly));
1430 
1431 	case IP_SEC_OPT:
1432 		mutex_enter(&connp->conn_lock);
1433 		error = ipsec_set_req(cr, connp, (ipsec_req_t *)invalp);
1434 		mutex_exit(&connp->conn_lock);
1435 		if (error != 0) {
1436 			return (error);
1437 		}
1438 		/* This is an IPsec policy change - redo ip_attr_connect */
1439 		coa->coa_changed |= COA_ROUTE_CHANGED;
1440 		break;
1441 	case IP_NEXTHOP:
1442 		ixa->ixa_nexthop_v4 = addr;
1443 		if (addr != INADDR_ANY)
1444 			ixa->ixa_flags |= IXAF_NEXTHOP_SET;
1445 		else
1446 			ixa->ixa_flags &= ~IXAF_NEXTHOP_SET;
1447 		coa->coa_changed |= COA_ROUTE_CHANGED;
1448 		break;
1449 
1450 	case IP_BOUND_IF:
1451 		ixa->ixa_ifindex = ifindex;		/* Send */
1452 		mutex_enter(&connp->conn_lock);
1453 		connp->conn_incoming_ifindex = ifindex;	/* Receive */
1454 		connp->conn_bound_if = ifindex;		/* getsockopt */
1455 		mutex_exit(&connp->conn_lock);
1456 		coa->coa_changed |= COA_ROUTE_CHANGED;
1457 		break;
1458 	case IP_UNSPEC_SRC:
1459 		mutex_enter(&connp->conn_lock);
1460 		connp->conn_unspec_src = onoff;
1461 		if (onoff)
1462 			ixa->ixa_flags &= ~IXAF_VERIFY_SOURCE;
1463 		else
1464 			ixa->ixa_flags |= IXAF_VERIFY_SOURCE;
1465 
1466 		mutex_exit(&connp->conn_lock);
1467 		break;
1468 	case IP_BROADCAST_TTL:
1469 		ixa->ixa_broadcast_ttl = *invalp;
1470 		ixa->ixa_flags |= IXAF_BROADCAST_TTL_SET;
1471 		/* Handled automatically by ip_output */
1472 		break;
1473 	case MRT_INIT:
1474 	case MRT_DONE:
1475 	case MRT_ADD_VIF:
1476 	case MRT_DEL_VIF:
1477 	case MRT_ADD_MFC:
1478 	case MRT_DEL_MFC:
1479 	case MRT_ASSERT:
1480 		if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
1481 			return (error);
1482 		}
1483 		error = ip_mrouter_set((int)name, connp, checkonly,
1484 		    (uchar_t *)invalp, inlen);
1485 		if (error) {
1486 			return (error);
1487 		}
1488 		return (0);
1489 
1490 	}
1491 	return (0);
1492 }
1493 
1494 /* Handle IPPROTO_IPV6 */
1495 static int
conn_opt_set_ipv6(conn_opt_arg_t * coa,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)1496 conn_opt_set_ipv6(conn_opt_arg_t *coa, t_scalar_t name, uint_t inlen,
1497     uchar_t *invalp, boolean_t checkonly, cred_t *cr)
1498 {
1499 	conn_t		*connp = coa->coa_connp;
1500 	ip_xmit_attr_t	*ixa = coa->coa_ixa;
1501 	ip_pkt_t	*ipp = coa->coa_ipp;
1502 	int		*i1 = (int *)invalp;
1503 	boolean_t	onoff = (*i1 == 0) ? 0 : 1;
1504 	uint_t		ifindex;
1505 	zoneid_t	zoneid = IPCL_ZONEID(connp);
1506 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
1507 	int		error;
1508 
1509 	if (connp->conn_family != AF_INET6)
1510 		return (EINVAL);
1511 
1512 	switch (name) {
1513 	case IPV6_MULTICAST_IF:
1514 		/*
1515 		 * The only possible error is EINVAL.
1516 		 * We call this option on both V4 and V6
1517 		 * If both fail, then this call returns
1518 		 * EINVAL. If at least one of them succeeds we
1519 		 * return success.
1520 		 */
1521 		ifindex = *(uint_t *)i1;
1522 
1523 		if (!ip_xmit_ifindex_valid(ifindex, zoneid, B_TRUE, ipst) &&
1524 		    !ip_xmit_ifindex_valid(ifindex, zoneid, B_FALSE, ipst))
1525 			return (EINVAL);
1526 		break;
1527 	case IPV6_UNICAST_HOPS:
1528 		/* Don't allow zero. -1 means to use default */
1529 		if (*i1 < -1 || *i1 == 0 || *i1 > IPV6_MAX_HOPS)
1530 			return (EINVAL);
1531 		break;
1532 	case IPV6_MULTICAST_HOPS:
1533 		/* -1 means use default */
1534 		if (*i1 < -1 || *i1 > IPV6_MAX_HOPS)
1535 			return (EINVAL);
1536 		break;
1537 	case IPV6_MULTICAST_LOOP:
1538 		if (*i1 != 0 && *i1 != 1)
1539 			return (EINVAL);
1540 		break;
1541 	case IPV6_BOUND_IF:
1542 		ifindex = *(uint_t *)i1;
1543 
1544 		if (!ip_xmit_ifindex_valid(ifindex, zoneid, B_TRUE, ipst))
1545 			return (ENXIO);
1546 		break;
1547 	case IPV6_PKTINFO: {
1548 		struct in6_pktinfo *pkti;
1549 		boolean_t isv6;
1550 
1551 		if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
1552 			return (EINVAL);
1553 		if (inlen == 0)
1554 			break;	/* Clear values below */
1555 
1556 		/*
1557 		 * Verify the source address and ifindex. Privileged users
1558 		 * can use any source address.
1559 		 */
1560 		pkti = (struct in6_pktinfo *)invalp;
1561 
1562 		/*
1563 		 * For link-local addresses we use the ipi6_ifindex when
1564 		 * we verify the local address.
1565 		 * If net_rawaccess then any source address can be used.
1566 		 */
1567 		if (!IN6_IS_ADDR_UNSPECIFIED(&pkti->ipi6_addr) &&
1568 		    secpolicy_net_rawaccess(cr) != 0) {
1569 			uint_t scopeid = 0;
1570 			in6_addr_t *v6src = &pkti->ipi6_addr;
1571 			ipaddr_t v4src;
1572 			ip_laddr_t laddr_type = IPVL_UNICAST_UP;
1573 
1574 			if (IN6_IS_ADDR_V4MAPPED(v6src)) {
1575 				IN6_V4MAPPED_TO_IPADDR(v6src, v4src);
1576 				if (v4src != INADDR_ANY) {
1577 					laddr_type = ip_laddr_verify_v4(v4src,
1578 					    zoneid, ipst, B_FALSE);
1579 				}
1580 			} else {
1581 				if (IN6_IS_ADDR_LINKSCOPE(v6src))
1582 					scopeid = pkti->ipi6_ifindex;
1583 
1584 				laddr_type = ip_laddr_verify_v6(v6src, zoneid,
1585 				    ipst, B_FALSE, scopeid);
1586 			}
1587 			switch (laddr_type) {
1588 			case IPVL_UNICAST_UP:
1589 			case IPVL_UNICAST_DOWN:
1590 				break;
1591 			default:
1592 				return (EADDRNOTAVAIL);
1593 			}
1594 			ixa->ixa_flags |= IXAF_VERIFY_SOURCE;
1595 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&pkti->ipi6_addr)) {
1596 			/* Allow any source */
1597 			ixa->ixa_flags &= ~IXAF_VERIFY_SOURCE;
1598 		}
1599 		isv6 = !(IN6_IS_ADDR_V4MAPPED(&pkti->ipi6_addr));
1600 		if (!ip_xmit_ifindex_valid(pkti->ipi6_ifindex, zoneid, isv6,
1601 		    ipst))
1602 			return (ENXIO);
1603 		break;
1604 	}
1605 	case IPV6_HOPLIMIT:
1606 		/* It is only allowed as ancilary data */
1607 		if (!coa->coa_ancillary)
1608 			return (EINVAL);
1609 
1610 		if (inlen != 0 && inlen != sizeof (int))
1611 			return (EINVAL);
1612 		if (inlen == sizeof (int)) {
1613 			if (*i1 > 255 || *i1 < -1 || *i1 == 0)
1614 				return (EINVAL);
1615 		}
1616 		break;
1617 	case IPV6_TCLASS:
1618 		if (inlen != 0 && inlen != sizeof (int))
1619 			return (EINVAL);
1620 		if (inlen == sizeof (int)) {
1621 			if (*i1 > 255 || *i1 < -1)
1622 				return (EINVAL);
1623 		}
1624 		break;
1625 	case IPV6_NEXTHOP:
1626 		if (inlen != 0 && inlen != sizeof (sin6_t))
1627 			return (EINVAL);
1628 		if (inlen == sizeof (sin6_t)) {
1629 			sin6_t *sin6 = (sin6_t *)invalp;
1630 			ire_t	*ire;
1631 
1632 			if (sin6->sin6_family != AF_INET6)
1633 				return (EAFNOSUPPORT);
1634 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
1635 				return (EADDRNOTAVAIL);
1636 
1637 			/* Verify that the next-hop is on-link */
1638 			ire = ire_ftable_lookup_v6(&sin6->sin6_addr,
1639 			    0, 0, IRE_ONLINK, NULL, zoneid,
1640 			    NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
1641 			if (ire == NULL)
1642 				return (EHOSTUNREACH);
1643 			ire_refrele(ire);
1644 			break;
1645 		}
1646 		break;
1647 	case IPV6_RTHDR:
1648 	case IPV6_DSTOPTS:
1649 	case IPV6_RTHDRDSTOPTS:
1650 	case IPV6_HOPOPTS: {
1651 		/* All have the length field in the same place */
1652 		ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
1653 		/*
1654 		 * Sanity checks - minimum size, size a multiple of
1655 		 * eight bytes, and matching size passed in.
1656 		 */
1657 		if (inlen != 0 &&
1658 		    inlen != (8 * (hopts->ip6h_len + 1)))
1659 			return (EINVAL);
1660 		break;
1661 	}
1662 	case IPV6_PATHMTU:
1663 		/* Can't be set */
1664 		return (EINVAL);
1665 
1666 	case IPV6_USE_MIN_MTU:
1667 		if (inlen != sizeof (int))
1668 			return (EINVAL);
1669 		if (*i1 < -1 || *i1 > 1)
1670 			return (EINVAL);
1671 		break;
1672 	case IPV6_SRC_PREFERENCES:
1673 		if (inlen != sizeof (uint32_t))
1674 			return (EINVAL);
1675 		break;
1676 	case IPV6_V6ONLY:
1677 		if (*i1 < 0 || *i1 > 1) {
1678 			return (EINVAL);
1679 		}
1680 		break;
1681 	}
1682 	if (checkonly)
1683 		return (0);
1684 
1685 	/* Here we set the actual option value */
1686 	/*
1687 	 * conn_lock protects the bitfields, and is used to
1688 	 * set the fields atomically. Not needed for ixa settings since
1689 	 * the caller has an exclusive copy of the ixa.
1690 	 * We can not hold conn_lock across the multicast options though.
1691 	 */
1692 	ASSERT(MUTEX_NOT_HELD(&coa->coa_connp->conn_lock));
1693 	switch (name) {
1694 	case IPV6_MULTICAST_IF:
1695 		ixa->ixa_multicast_ifindex = ifindex;
1696 		/* Need to redo ip_attr_connect */
1697 		coa->coa_changed |= COA_ROUTE_CHANGED;
1698 		break;
1699 	case IPV6_UNICAST_HOPS:
1700 		/* -1 means use default */
1701 		mutex_enter(&connp->conn_lock);
1702 		if (*i1 == -1) {
1703 			ipp->ipp_unicast_hops = connp->conn_default_ttl;
1704 		} else {
1705 			ipp->ipp_unicast_hops = (uint8_t)*i1;
1706 		}
1707 		mutex_exit(&connp->conn_lock);
1708 		coa->coa_changed |= COA_HEADER_CHANGED;
1709 		break;
1710 	case IPV6_MULTICAST_HOPS:
1711 		/* -1 means use default */
1712 		if (*i1 == -1) {
1713 			ixa->ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1714 		} else {
1715 			ixa->ixa_multicast_ttl = (uint8_t)*i1;
1716 		}
1717 		/* Handled automatically by ip_output */
1718 		break;
1719 	case IPV6_MULTICAST_LOOP:
1720 		if (*i1 != 0)
1721 			ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
1722 		else
1723 			ixa->ixa_flags &= ~IXAF_MULTICAST_LOOP;
1724 		/* Handled automatically by ip_output */
1725 		break;
1726 	case IPV6_JOIN_GROUP:
1727 	case IPV6_LEAVE_GROUP:
1728 	case MCAST_JOIN_GROUP:
1729 	case MCAST_LEAVE_GROUP:
1730 		return (ip_opt_set_multicast_group(connp, name,
1731 		    invalp, B_TRUE, checkonly));
1732 
1733 	case MCAST_BLOCK_SOURCE:
1734 	case MCAST_UNBLOCK_SOURCE:
1735 	case MCAST_JOIN_SOURCE_GROUP:
1736 	case MCAST_LEAVE_SOURCE_GROUP:
1737 		return (ip_opt_set_multicast_sources(connp, name,
1738 		    invalp, B_TRUE, checkonly));
1739 
1740 	case IPV6_BOUND_IF:
1741 		ixa->ixa_ifindex = ifindex;		/* Send */
1742 		mutex_enter(&connp->conn_lock);
1743 		connp->conn_incoming_ifindex = ifindex;	/* Receive */
1744 		connp->conn_bound_if = ifindex;		/* getsockopt */
1745 		mutex_exit(&connp->conn_lock);
1746 		coa->coa_changed |= COA_ROUTE_CHANGED;
1747 		break;
1748 	case IPV6_UNSPEC_SRC:
1749 		mutex_enter(&connp->conn_lock);
1750 		connp->conn_unspec_src = onoff;
1751 		if (onoff)
1752 			ixa->ixa_flags &= ~IXAF_VERIFY_SOURCE;
1753 		else
1754 			ixa->ixa_flags |= IXAF_VERIFY_SOURCE;
1755 		mutex_exit(&connp->conn_lock);
1756 		break;
1757 	case IPV6_RECVPKTINFO:
1758 		mutex_enter(&connp->conn_lock);
1759 		connp->conn_recv_ancillary.crb_ip_recvpktinfo = onoff;
1760 		mutex_exit(&connp->conn_lock);
1761 		break;
1762 	case IPV6_RECVTCLASS:
1763 		mutex_enter(&connp->conn_lock);
1764 		connp->conn_recv_ancillary.crb_ipv6_recvtclass = onoff;
1765 		mutex_exit(&connp->conn_lock);
1766 		break;
1767 	case IPV6_RECVPATHMTU:
1768 		mutex_enter(&connp->conn_lock);
1769 		connp->conn_ipv6_recvpathmtu = onoff;
1770 		mutex_exit(&connp->conn_lock);
1771 		break;
1772 	case IPV6_RECVHOPLIMIT:
1773 		mutex_enter(&connp->conn_lock);
1774 		connp->conn_recv_ancillary.crb_ipv6_recvhoplimit =
1775 		    onoff;
1776 		mutex_exit(&connp->conn_lock);
1777 		break;
1778 	case IPV6_RECVHOPOPTS:
1779 		mutex_enter(&connp->conn_lock);
1780 		connp->conn_recv_ancillary.crb_ipv6_recvhopopts = onoff;
1781 		mutex_exit(&connp->conn_lock);
1782 		break;
1783 	case IPV6_RECVDSTOPTS:
1784 		mutex_enter(&connp->conn_lock);
1785 		connp->conn_recv_ancillary.crb_ipv6_recvdstopts = onoff;
1786 		mutex_exit(&connp->conn_lock);
1787 		break;
1788 	case _OLD_IPV6_RECVDSTOPTS:
1789 		mutex_enter(&connp->conn_lock);
1790 		connp->conn_recv_ancillary.crb_old_ipv6_recvdstopts =
1791 		    onoff;
1792 		mutex_exit(&connp->conn_lock);
1793 		break;
1794 	case IPV6_RECVRTHDRDSTOPTS:
1795 		mutex_enter(&connp->conn_lock);
1796 		connp->conn_recv_ancillary.crb_ipv6_recvrthdrdstopts =
1797 		    onoff;
1798 		mutex_exit(&connp->conn_lock);
1799 		break;
1800 	case IPV6_RECVRTHDR:
1801 		mutex_enter(&connp->conn_lock);
1802 		connp->conn_recv_ancillary.crb_ipv6_recvrthdr = onoff;
1803 		mutex_exit(&connp->conn_lock);
1804 		break;
1805 	case IPV6_PKTINFO:
1806 		mutex_enter(&connp->conn_lock);
1807 		if (inlen == 0) {
1808 			ipp->ipp_fields &= ~IPPF_ADDR;
1809 			ipp->ipp_addr = ipv6_all_zeros;
1810 			ixa->ixa_ifindex = 0;
1811 		} else {
1812 			struct in6_pktinfo *pkti;
1813 
1814 			pkti = (struct in6_pktinfo *)invalp;
1815 			ipp->ipp_addr = pkti->ipi6_addr;
1816 			if (!IN6_IS_ADDR_UNSPECIFIED(&ipp->ipp_addr))
1817 				ipp->ipp_fields |= IPPF_ADDR;
1818 			else
1819 				ipp->ipp_fields &= ~IPPF_ADDR;
1820 			ixa->ixa_ifindex = pkti->ipi6_ifindex;
1821 		}
1822 		mutex_exit(&connp->conn_lock);
1823 		/* Source and ifindex might have changed */
1824 		coa->coa_changed |= COA_HEADER_CHANGED;
1825 		coa->coa_changed |= COA_ROUTE_CHANGED;
1826 		break;
1827 	case IPV6_HOPLIMIT:
1828 		mutex_enter(&connp->conn_lock);
1829 		if (inlen == 0 || *i1 == -1) {
1830 			/* Revert to default */
1831 			ipp->ipp_fields &= ~IPPF_HOPLIMIT;
1832 			ixa->ixa_flags &= ~IXAF_NO_TTL_CHANGE;
1833 		} else {
1834 			ipp->ipp_hoplimit = *i1;
1835 			ipp->ipp_fields |= IPPF_HOPLIMIT;
1836 			/* Ensure that it sticks for multicast packets */
1837 			ixa->ixa_flags |= IXAF_NO_TTL_CHANGE;
1838 		}
1839 		mutex_exit(&connp->conn_lock);
1840 		coa->coa_changed |= COA_HEADER_CHANGED;
1841 		break;
1842 	case IPV6_TCLASS:
1843 		/*
1844 		 * IPV6_TCLASS accepts -1 as use kernel default
1845 		 * and [0, 255] as the actualy traffic class.
1846 		 */
1847 		mutex_enter(&connp->conn_lock);
1848 		if (inlen == 0 || *i1 == -1) {
1849 			ipp->ipp_tclass = 0;
1850 			ipp->ipp_fields &= ~IPPF_TCLASS;
1851 		} else {
1852 			ipp->ipp_tclass = *i1;
1853 			ipp->ipp_fields |= IPPF_TCLASS;
1854 		}
1855 		mutex_exit(&connp->conn_lock);
1856 		coa->coa_changed |= COA_HEADER_CHANGED;
1857 		break;
1858 	case IPV6_NEXTHOP:
1859 		if (inlen == 0) {
1860 			ixa->ixa_flags &= ~IXAF_NEXTHOP_SET;
1861 		} else {
1862 			sin6_t *sin6 = (sin6_t *)invalp;
1863 
1864 			ixa->ixa_nexthop_v6 = sin6->sin6_addr;
1865 			if (!IN6_IS_ADDR_UNSPECIFIED(&ixa->ixa_nexthop_v6))
1866 				ixa->ixa_flags |= IXAF_NEXTHOP_SET;
1867 			else
1868 				ixa->ixa_flags &= ~IXAF_NEXTHOP_SET;
1869 		}
1870 		coa->coa_changed |= COA_ROUTE_CHANGED;
1871 		break;
1872 	case IPV6_HOPOPTS:
1873 		mutex_enter(&connp->conn_lock);
1874 		error = optcom_pkt_set(invalp, inlen,
1875 		    (uchar_t **)&ipp->ipp_hopopts, &ipp->ipp_hopoptslen);
1876 		if (error != 0) {
1877 			mutex_exit(&connp->conn_lock);
1878 			return (error);
1879 		}
1880 		if (ipp->ipp_hopoptslen == 0) {
1881 			ipp->ipp_fields &= ~IPPF_HOPOPTS;
1882 		} else {
1883 			ipp->ipp_fields |= IPPF_HOPOPTS;
1884 		}
1885 		mutex_exit(&connp->conn_lock);
1886 		coa->coa_changed |= COA_HEADER_CHANGED;
1887 		coa->coa_changed |= COA_WROFF_CHANGED;
1888 		break;
1889 	case IPV6_RTHDRDSTOPTS:
1890 		mutex_enter(&connp->conn_lock);
1891 		error = optcom_pkt_set(invalp, inlen,
1892 		    (uchar_t **)&ipp->ipp_rthdrdstopts,
1893 		    &ipp->ipp_rthdrdstoptslen);
1894 		if (error != 0) {
1895 			mutex_exit(&connp->conn_lock);
1896 			return (error);
1897 		}
1898 		if (ipp->ipp_rthdrdstoptslen == 0) {
1899 			ipp->ipp_fields &= ~IPPF_RTHDRDSTOPTS;
1900 		} else {
1901 			ipp->ipp_fields |= IPPF_RTHDRDSTOPTS;
1902 		}
1903 		mutex_exit(&connp->conn_lock);
1904 		coa->coa_changed |= COA_HEADER_CHANGED;
1905 		coa->coa_changed |= COA_WROFF_CHANGED;
1906 		break;
1907 	case IPV6_DSTOPTS:
1908 		mutex_enter(&connp->conn_lock);
1909 		error = optcom_pkt_set(invalp, inlen,
1910 		    (uchar_t **)&ipp->ipp_dstopts, &ipp->ipp_dstoptslen);
1911 		if (error != 0) {
1912 			mutex_exit(&connp->conn_lock);
1913 			return (error);
1914 		}
1915 		if (ipp->ipp_dstoptslen == 0) {
1916 			ipp->ipp_fields &= ~IPPF_DSTOPTS;
1917 		} else {
1918 			ipp->ipp_fields |= IPPF_DSTOPTS;
1919 		}
1920 		mutex_exit(&connp->conn_lock);
1921 		coa->coa_changed |= COA_HEADER_CHANGED;
1922 		coa->coa_changed |= COA_WROFF_CHANGED;
1923 		break;
1924 	case IPV6_RTHDR:
1925 		mutex_enter(&connp->conn_lock);
1926 		error = optcom_pkt_set(invalp, inlen,
1927 		    (uchar_t **)&ipp->ipp_rthdr, &ipp->ipp_rthdrlen);
1928 		if (error != 0) {
1929 			mutex_exit(&connp->conn_lock);
1930 			return (error);
1931 		}
1932 		if (ipp->ipp_rthdrlen == 0) {
1933 			ipp->ipp_fields &= ~IPPF_RTHDR;
1934 		} else {
1935 			ipp->ipp_fields |= IPPF_RTHDR;
1936 		}
1937 		mutex_exit(&connp->conn_lock);
1938 		coa->coa_changed |= COA_HEADER_CHANGED;
1939 		coa->coa_changed |= COA_WROFF_CHANGED;
1940 		break;
1941 
1942 	case IPV6_DONTFRAG:
1943 		if (onoff) {
1944 			ixa->ixa_flags |= IXAF_DONTFRAG;
1945 			ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
1946 		} else {
1947 			ixa->ixa_flags &= ~IXAF_DONTFRAG;
1948 			ixa->ixa_flags |= IXAF_PMTU_DISCOVERY;
1949 		}
1950 		/* Need to redo ip_attr_connect */
1951 		coa->coa_changed |= COA_ROUTE_CHANGED;
1952 		break;
1953 
1954 	case IPV6_USE_MIN_MTU:
1955 		ixa->ixa_flags |= IXAF_USE_MIN_MTU;
1956 		ixa->ixa_use_min_mtu = *i1;
1957 		/* Need to redo ip_attr_connect */
1958 		coa->coa_changed |= COA_ROUTE_CHANGED;
1959 		break;
1960 
1961 	case IPV6_SEC_OPT:
1962 		mutex_enter(&connp->conn_lock);
1963 		error = ipsec_set_req(cr, connp, (ipsec_req_t *)invalp);
1964 		mutex_exit(&connp->conn_lock);
1965 		if (error != 0) {
1966 			return (error);
1967 		}
1968 		/* This is an IPsec policy change - redo ip_attr_connect */
1969 		coa->coa_changed |= COA_ROUTE_CHANGED;
1970 		break;
1971 	case IPV6_SRC_PREFERENCES:
1972 		/*
1973 		 * This socket option only affects connected
1974 		 * sockets that haven't already bound to a specific
1975 		 * IPv6 address.  In other words, sockets that
1976 		 * don't call bind() with an address other than the
1977 		 * unspecified address and that call connect().
1978 		 * ip_set_destination_v6() passes these preferences
1979 		 * to the ipif_select_source_v6() function.
1980 		 */
1981 		mutex_enter(&connp->conn_lock);
1982 		error = ip6_set_src_preferences(ixa, *(uint32_t *)invalp);
1983 		mutex_exit(&connp->conn_lock);
1984 		if (error != 0) {
1985 			return (error);
1986 		}
1987 		break;
1988 	case IPV6_V6ONLY:
1989 		mutex_enter(&connp->conn_lock);
1990 		connp->conn_ipv6_v6only = onoff;
1991 		mutex_exit(&connp->conn_lock);
1992 		break;
1993 	}
1994 	return (0);
1995 }
1996 
1997 /* Handle IPPROTO_UDP */
1998 /* ARGSUSED1 */
1999 static int
conn_opt_set_udp(conn_opt_arg_t * coa,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)2000 conn_opt_set_udp(conn_opt_arg_t *coa, t_scalar_t name, uint_t inlen,
2001     uchar_t *invalp, boolean_t checkonly, cred_t *cr)
2002 {
2003 	conn_t		*connp = coa->coa_connp;
2004 	int		*i1 = (int *)invalp;
2005 	boolean_t	onoff = (*i1 == 0) ? 0 : 1;
2006 	int		error;
2007 
2008 	switch (name) {
2009 	case UDP_ANONPRIVBIND:
2010 		if ((error = secpolicy_net_privaddr(cr, 0, IPPROTO_UDP)) != 0) {
2011 			return (error);
2012 		}
2013 		break;
2014 	}
2015 	if (checkonly)
2016 		return (0);
2017 
2018 	/* Here we set the actual option value */
2019 	mutex_enter(&connp->conn_lock);
2020 	switch (name) {
2021 	case UDP_ANONPRIVBIND:
2022 		connp->conn_anon_priv_bind = onoff;
2023 		break;
2024 	case UDP_EXCLBIND:
2025 		connp->conn_exclbind = onoff;
2026 		break;
2027 	}
2028 	mutex_exit(&connp->conn_lock);
2029 	return (0);
2030 }
2031 
2032 /* Handle IPPROTO_TCP */
2033 /* ARGSUSED1 */
2034 static int
conn_opt_set_tcp(conn_opt_arg_t * coa,t_scalar_t name,uint_t inlen,uchar_t * invalp,boolean_t checkonly,cred_t * cr)2035 conn_opt_set_tcp(conn_opt_arg_t *coa, t_scalar_t name, uint_t inlen,
2036     uchar_t *invalp, boolean_t checkonly, cred_t *cr)
2037 {
2038 	conn_t		*connp = coa->coa_connp;
2039 	int		*i1 = (int *)invalp;
2040 	boolean_t	onoff = (*i1 == 0) ? 0 : 1;
2041 	int		error;
2042 
2043 	switch (name) {
2044 	case TCP_ANONPRIVBIND:
2045 		if ((error = secpolicy_net_privaddr(cr, 0, IPPROTO_TCP)) != 0) {
2046 			return (error);
2047 		}
2048 		break;
2049 	}
2050 	if (checkonly)
2051 		return (0);
2052 
2053 	/* Here we set the actual option value */
2054 	mutex_enter(&connp->conn_lock);
2055 	switch (name) {
2056 	case TCP_ANONPRIVBIND:
2057 		connp->conn_anon_priv_bind = onoff;
2058 		break;
2059 	case TCP_EXCLBIND:
2060 		connp->conn_exclbind = onoff;
2061 		break;
2062 	case TCP_RECVDSTADDR:
2063 		connp->conn_recv_ancillary.crb_recvdstaddr = onoff;
2064 		break;
2065 	}
2066 	mutex_exit(&connp->conn_lock);
2067 	return (0);
2068 }
2069 
2070 int
conn_getsockname(conn_t * connp,struct sockaddr * sa,uint_t * salenp)2071 conn_getsockname(conn_t *connp, struct sockaddr *sa, uint_t *salenp)
2072 {
2073 	sin_t		*sin;
2074 	sin6_t		*sin6;
2075 
2076 	if (connp->conn_family == AF_INET) {
2077 		if (*salenp < sizeof (sin_t))
2078 			return (EINVAL);
2079 
2080 		*salenp = sizeof (sin_t);
2081 		/* Fill zeroes and then initialize non-zero fields */
2082 		sin = (sin_t *)sa;
2083 		*sin = sin_null;
2084 		sin->sin_family = AF_INET;
2085 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&connp->conn_saddr_v6) &&
2086 		    !IN6_IS_ADDR_UNSPECIFIED(&connp->conn_saddr_v6)) {
2087 			sin->sin_addr.s_addr = connp->conn_saddr_v4;
2088 		} else {
2089 			/*
2090 			 * INADDR_ANY
2091 			 * conn_saddr is not set, we might be bound to
2092 			 * broadcast/multicast. Use conn_bound_addr as
2093 			 * local address instead (that could
2094 			 * also still be INADDR_ANY)
2095 			 */
2096 			sin->sin_addr.s_addr = connp->conn_bound_addr_v4;
2097 		}
2098 		sin->sin_port = connp->conn_lport;
2099 	} else {
2100 		if (*salenp < sizeof (sin6_t))
2101 			return (EINVAL);
2102 
2103 		*salenp = sizeof (sin6_t);
2104 		/* Fill zeroes and then initialize non-zero fields */
2105 		sin6 = (sin6_t *)sa;
2106 		*sin6 = sin6_null;
2107 		sin6->sin6_family = AF_INET6;
2108 		if (!IN6_IS_ADDR_UNSPECIFIED(&connp->conn_saddr_v6)) {
2109 			sin6->sin6_addr = connp->conn_saddr_v6;
2110 		} else {
2111 			/*
2112 			 * conn_saddr is not set, we might be bound to
2113 			 * broadcast/multicast. Use conn_bound_addr as
2114 			 * local address instead (which could
2115 			 * also still be unspecified)
2116 			 */
2117 			sin6->sin6_addr = connp->conn_bound_addr_v6;
2118 		}
2119 		sin6->sin6_port = connp->conn_lport;
2120 		if (IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr) &&
2121 		    (connp->conn_ixa->ixa_flags & IXAF_SCOPEID_SET))
2122 			sin6->sin6_scope_id = connp->conn_ixa->ixa_scopeid;
2123 	}
2124 	return (0);
2125 }
2126 
2127 int
conn_getpeername(conn_t * connp,struct sockaddr * sa,uint_t * salenp)2128 conn_getpeername(conn_t *connp, struct sockaddr *sa, uint_t *salenp)
2129 {
2130 	struct sockaddr_in	*sin;
2131 	struct sockaddr_in6	*sin6;
2132 
2133 	if (connp->conn_family == AF_INET) {
2134 		if (*salenp < sizeof (sin_t))
2135 			return (EINVAL);
2136 
2137 		*salenp = sizeof (sin_t);
2138 		/* initialize */
2139 		sin = (sin_t *)sa;
2140 		*sin = sin_null;
2141 		sin->sin_family = AF_INET;
2142 		sin->sin_addr.s_addr = connp->conn_faddr_v4;
2143 		sin->sin_port = connp->conn_fport;
2144 	} else {
2145 		if (*salenp < sizeof (sin6_t))
2146 			return (EINVAL);
2147 
2148 		*salenp = sizeof (sin6_t);
2149 		/* initialize */
2150 		sin6 = (sin6_t *)sa;
2151 		*sin6 = sin6_null;
2152 		sin6->sin6_family = AF_INET6;
2153 		sin6->sin6_addr = connp->conn_faddr_v6;
2154 		sin6->sin6_port =  connp->conn_fport;
2155 		sin6->sin6_flowinfo = connp->conn_flowinfo;
2156 		if (IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr) &&
2157 		    (connp->conn_ixa->ixa_flags & IXAF_SCOPEID_SET))
2158 			sin6->sin6_scope_id = connp->conn_ixa->ixa_scopeid;
2159 	}
2160 	return (0);
2161 }
2162 
2163 static uint32_t	cksum_massage_options_v4(ipha_t *, netstack_t *);
2164 static uint32_t cksum_massage_options_v6(ip6_t *, uint_t, netstack_t *);
2165 
2166 /*
2167  * Allocate and fill in conn_ht_iphc based on the current information
2168  * in the conn.
2169  * Normally used when we bind() and connect().
2170  * Returns failure if can't allocate memory, or if there is a problem
2171  * with a routing header/option.
2172  *
2173  * We allocate space for the transport header (ulp_hdr_len + extra) and
2174  * indicate the offset of the ulp header by setting ixa_ip_hdr_length.
2175  * The extra is there for transports that want some spare room for future
2176  * options. conn_ht_iphc_allocated is what was allocated; conn_ht_iphc_len
2177  * excludes the extra part.
2178  *
2179  * We massage an routing option/header and store the ckecksum difference
2180  * in conn_sum.
2181  *
2182  * Caller needs to update conn_wroff if desired.
2183  */
2184 int
conn_build_hdr_template(conn_t * connp,uint_t ulp_hdr_length,uint_t extra,const in6_addr_t * v6src,const in6_addr_t * v6dst,uint32_t flowinfo)2185 conn_build_hdr_template(conn_t *connp, uint_t ulp_hdr_length, uint_t extra,
2186     const in6_addr_t *v6src, const in6_addr_t *v6dst, uint32_t flowinfo)
2187 {
2188 	ip_xmit_attr_t	*ixa = connp->conn_ixa;
2189 	ip_pkt_t	*ipp = &connp->conn_xmit_ipp;
2190 	uint_t		ip_hdr_length;
2191 	uchar_t		*hdrs;
2192 	uint_t		hdrs_len;
2193 
2194 	ASSERT(MUTEX_HELD(&connp->conn_lock));
2195 
2196 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2197 		ip_hdr_length = ip_total_hdrs_len_v4(ipp);
2198 		/* In case of TX label and IP options it can be too much */
2199 		if (ip_hdr_length > IP_MAX_HDR_LENGTH) {
2200 			/* Preserves existing TX errno for this */
2201 			return (EHOSTUNREACH);
2202 		}
2203 	} else {
2204 		ip_hdr_length = ip_total_hdrs_len_v6(ipp);
2205 	}
2206 	ixa->ixa_ip_hdr_length = ip_hdr_length;
2207 	hdrs_len = ip_hdr_length + ulp_hdr_length + extra;
2208 	ASSERT(hdrs_len != 0);
2209 
2210 	if (hdrs_len != connp->conn_ht_iphc_allocated) {
2211 		/* Allocate new before we free any old */
2212 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
2213 		if (hdrs == NULL)
2214 			return (ENOMEM);
2215 
2216 		if (connp->conn_ht_iphc != NULL) {
2217 			kmem_free(connp->conn_ht_iphc,
2218 			    connp->conn_ht_iphc_allocated);
2219 		}
2220 		connp->conn_ht_iphc = hdrs;
2221 		connp->conn_ht_iphc_allocated = hdrs_len;
2222 	} else {
2223 		hdrs = connp->conn_ht_iphc;
2224 	}
2225 	hdrs_len -= extra;
2226 	connp->conn_ht_iphc_len = hdrs_len;
2227 
2228 	connp->conn_ht_ulp = hdrs + ip_hdr_length;
2229 	connp->conn_ht_ulp_len = ulp_hdr_length;
2230 
2231 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2232 		ipha_t	*ipha = (ipha_t *)hdrs;
2233 
2234 		IN6_V4MAPPED_TO_IPADDR(v6src, ipha->ipha_src);
2235 		IN6_V4MAPPED_TO_IPADDR(v6dst, ipha->ipha_dst);
2236 		ip_build_hdrs_v4(hdrs, ip_hdr_length, ipp, connp->conn_proto);
2237 		ipha->ipha_length = htons(hdrs_len);
2238 		if (ixa->ixa_flags & IXAF_PMTU_IPV4_DF)
2239 			ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
2240 		else
2241 			ipha->ipha_fragment_offset_and_flags &= ~IPH_DF_HTONS;
2242 
2243 		if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
2244 			connp->conn_sum = cksum_massage_options_v4(ipha,
2245 			    connp->conn_netstack);
2246 		} else {
2247 			connp->conn_sum = 0;
2248 		}
2249 	} else {
2250 		ip6_t	*ip6h = (ip6_t *)hdrs;
2251 
2252 		ip6h->ip6_src = *v6src;
2253 		ip6h->ip6_dst = *v6dst;
2254 		ip_build_hdrs_v6(hdrs, ip_hdr_length, ipp, connp->conn_proto,
2255 		    flowinfo);
2256 		ip6h->ip6_plen = htons(hdrs_len - IPV6_HDR_LEN);
2257 
2258 		if (ipp->ipp_fields & IPPF_RTHDR) {
2259 			connp->conn_sum = cksum_massage_options_v6(ip6h,
2260 			    ip_hdr_length, connp->conn_netstack);
2261 
2262 			/*
2263 			 * Verify that the first hop isn't a mapped address.
2264 			 * Routers along the path need to do this verification
2265 			 * for subsequent hops.
2266 			 */
2267 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst))
2268 				return (EADDRNOTAVAIL);
2269 
2270 		} else {
2271 			connp->conn_sum = 0;
2272 		}
2273 	}
2274 	return (0);
2275 }
2276 
2277 /*
2278  * Prepend a header template to data_mp based on the ip_pkt_t
2279  * and the passed in source, destination and protocol.
2280  *
2281  * Returns failure if can't allocate memory, in which case data_mp is freed.
2282  * We allocate space for the transport header (ulp_hdr_len) and
2283  * indicate the offset of the ulp header by setting ixa_ip_hdr_length.
2284  *
2285  * We massage an routing option/header and return the ckecksum difference
2286  * in *sump. This is in host byte order.
2287  *
2288  * Caller needs to update conn_wroff if desired.
2289  */
2290 mblk_t *
conn_prepend_hdr(ip_xmit_attr_t * ixa,const ip_pkt_t * ipp,const in6_addr_t * v6src,const in6_addr_t * v6dst,uint8_t protocol,uint32_t flowinfo,uint_t ulp_hdr_length,mblk_t * data_mp,uint_t data_length,uint_t wroff_extra,uint32_t * sump,int * errorp)2291 conn_prepend_hdr(ip_xmit_attr_t *ixa, const ip_pkt_t *ipp,
2292     const in6_addr_t *v6src, const in6_addr_t *v6dst,
2293     uint8_t protocol, uint32_t flowinfo, uint_t ulp_hdr_length, mblk_t *data_mp,
2294     uint_t data_length, uint_t wroff_extra, uint32_t *sump, int *errorp)
2295 {
2296 	uint_t		ip_hdr_length;
2297 	uchar_t		*hdrs;
2298 	uint_t		hdrs_len;
2299 	mblk_t		*mp;
2300 
2301 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2302 		ip_hdr_length = ip_total_hdrs_len_v4(ipp);
2303 		ASSERT(ip_hdr_length <= IP_MAX_HDR_LENGTH);
2304 	} else {
2305 		ip_hdr_length = ip_total_hdrs_len_v6(ipp);
2306 	}
2307 	hdrs_len = ip_hdr_length + ulp_hdr_length;
2308 	ASSERT(hdrs_len != 0);
2309 
2310 	ixa->ixa_ip_hdr_length = ip_hdr_length;
2311 
2312 	/* Can we prepend to data_mp? */
2313 	if (data_mp != NULL &&
2314 	    data_mp->b_rptr - data_mp->b_datap->db_base >= hdrs_len &&
2315 	    data_mp->b_datap->db_ref == 1) {
2316 		hdrs = data_mp->b_rptr - hdrs_len;
2317 		data_mp->b_rptr = hdrs;
2318 		mp = data_mp;
2319 	} else {
2320 		mp = allocb(hdrs_len + wroff_extra, BPRI_MED);
2321 		if (mp == NULL) {
2322 			freemsg(data_mp);
2323 			*errorp = ENOMEM;
2324 			return (NULL);
2325 		}
2326 		mp->b_wptr = mp->b_datap->db_lim;
2327 		hdrs = mp->b_rptr = mp->b_wptr - hdrs_len;
2328 		mp->b_cont = data_mp;
2329 	}
2330 
2331 	/*
2332 	 * Set the source in the header. ip_build_hdrs_v4/v6 will overwrite it
2333 	 * if PKTINFO (aka IPPF_ADDR) was set.
2334 	 */
2335 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2336 		ipha_t *ipha = (ipha_t *)hdrs;
2337 
2338 		ASSERT(IN6_IS_ADDR_V4MAPPED(v6dst));
2339 		IN6_V4MAPPED_TO_IPADDR(v6src, ipha->ipha_src);
2340 		IN6_V4MAPPED_TO_IPADDR(v6dst, ipha->ipha_dst);
2341 		ip_build_hdrs_v4(hdrs, ip_hdr_length, ipp, protocol);
2342 		ipha->ipha_length = htons(hdrs_len + data_length);
2343 		if (ixa->ixa_flags & IXAF_PMTU_IPV4_DF)
2344 			ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
2345 		else
2346 			ipha->ipha_fragment_offset_and_flags &= ~IPH_DF_HTONS;
2347 
2348 		if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
2349 			*sump = cksum_massage_options_v4(ipha,
2350 			    ixa->ixa_ipst->ips_netstack);
2351 		} else {
2352 			*sump = 0;
2353 		}
2354 	} else {
2355 		ip6_t *ip6h = (ip6_t *)hdrs;
2356 
2357 		ip6h->ip6_src = *v6src;
2358 		ip6h->ip6_dst = *v6dst;
2359 		ip_build_hdrs_v6(hdrs, ip_hdr_length, ipp, protocol, flowinfo);
2360 		ip6h->ip6_plen = htons(hdrs_len + data_length - IPV6_HDR_LEN);
2361 
2362 		if (ipp->ipp_fields & IPPF_RTHDR) {
2363 			*sump = cksum_massage_options_v6(ip6h,
2364 			    ip_hdr_length, ixa->ixa_ipst->ips_netstack);
2365 
2366 			/*
2367 			 * Verify that the first hop isn't a mapped address.
2368 			 * Routers along the path need to do this verification
2369 			 * for subsequent hops.
2370 			 */
2371 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
2372 				*errorp = EADDRNOTAVAIL;
2373 				freemsg(mp);
2374 				return (NULL);
2375 			}
2376 		} else {
2377 			*sump = 0;
2378 		}
2379 	}
2380 	return (mp);
2381 }
2382 
2383 /*
2384  * Massage a source route if any putting the first hop
2385  * in ipha_dst. Compute a starting value for the checksum which
2386  * takes into account that the original ipha_dst should be
2387  * included in the checksum but that IP will include the
2388  * first hop from the source route in the tcp checksum.
2389  */
2390 static uint32_t
cksum_massage_options_v4(ipha_t * ipha,netstack_t * ns)2391 cksum_massage_options_v4(ipha_t *ipha, netstack_t *ns)
2392 {
2393 	in_addr_t	dst;
2394 	uint32_t	cksum;
2395 
2396 	/* Get last hop then diff against first hop */
2397 	cksum = ip_massage_options(ipha, ns);
2398 	cksum = (cksum & 0xFFFF) + (cksum >> 16);
2399 	dst = ipha->ipha_dst;
2400 	cksum -= ((dst >> 16) + (dst & 0xffff));
2401 	if ((int)cksum < 0)
2402 		cksum--;
2403 	cksum = (cksum & 0xFFFF) + (cksum >> 16);
2404 	cksum = (cksum & 0xFFFF) + (cksum >> 16);
2405 	ASSERT(cksum < 0x10000);
2406 	return (ntohs(cksum));
2407 }
2408 
2409 static uint32_t
cksum_massage_options_v6(ip6_t * ip6h,uint_t ip_hdr_len,netstack_t * ns)2410 cksum_massage_options_v6(ip6_t *ip6h, uint_t ip_hdr_len, netstack_t *ns)
2411 {
2412 	uint8_t		*end;
2413 	ip6_rthdr_t	*rth;
2414 	uint32_t	cksum;
2415 
2416 	end = (uint8_t *)ip6h + ip_hdr_len;
2417 	rth = ip_find_rthdr_v6(ip6h, end);
2418 	if (rth == NULL)
2419 		return (0);
2420 
2421 	cksum = ip_massage_options_v6(ip6h, rth, ns);
2422 	cksum = (cksum & 0xFFFF) + (cksum >> 16);
2423 	ASSERT(cksum < 0x10000);
2424 	return (ntohs(cksum));
2425 }
2426 
2427 /*
2428  * ULPs that change the destination address need to call this for each
2429  * change to discard any state about a previous destination that might
2430  * have been multicast or multirt.
2431  */
2432 void
ip_attr_newdst(ip_xmit_attr_t * ixa)2433 ip_attr_newdst(ip_xmit_attr_t *ixa)
2434 {
2435 	ixa->ixa_flags &= ~(IXAF_LOOPBACK_COPY | IXAF_NO_HW_CKSUM |
2436 	    IXAF_NO_TTL_CHANGE | IXAF_IPV6_ADD_FRAGHDR |
2437 	    IXAF_NO_LOOP_ZONEID_SET);
2438 }
2439 
2440 /*
2441  * Determine the nexthop which will be used.
2442  * Normally this is just the destination, but if a IPv4 source route, or
2443  * IPv6 routing header, is in the ip_pkt_t then we extract the nexthop from
2444  * there.
2445  */
2446 void
ip_attr_nexthop(const ip_pkt_t * ipp,const ip_xmit_attr_t * ixa,const in6_addr_t * dst,in6_addr_t * nexthop)2447 ip_attr_nexthop(const ip_pkt_t *ipp, const ip_xmit_attr_t *ixa,
2448     const in6_addr_t *dst, in6_addr_t *nexthop)
2449 {
2450 	if (!(ipp->ipp_fields & (IPPF_IPV4_OPTIONS|IPPF_RTHDR))) {
2451 		*nexthop = *dst;
2452 		return;
2453 	}
2454 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2455 		ipaddr_t v4dst;
2456 		ipaddr_t v4nexthop;
2457 
2458 		IN6_V4MAPPED_TO_IPADDR(dst, v4dst);
2459 		v4nexthop = ip_pkt_source_route_v4(ipp);
2460 		if (v4nexthop == INADDR_ANY)
2461 			v4nexthop = v4dst;
2462 
2463 		IN6_IPADDR_TO_V4MAPPED(v4nexthop, nexthop);
2464 	} else {
2465 		const in6_addr_t *v6nexthop;
2466 
2467 		v6nexthop = ip_pkt_source_route_v6(ipp);
2468 		if (v6nexthop == NULL)
2469 			v6nexthop = dst;
2470 
2471 		*nexthop = *v6nexthop;
2472 	}
2473 }
2474 
2475 /*
2476  * Update the ip_xmit_attr_t based the addresses, conn_xmit_ipp and conn_ixa.
2477  * If IPDF_IPSEC is set we cache the IPsec policy to handle the unconnected
2478  * case (connected latching is done in conn_connect).
2479  * Note that IPsec policy lookup requires conn_proto and conn_laddr to be
2480  * set, but doesn't otherwise use the conn_t.
2481  *
2482  * Caller must set/clear IXAF_IS_IPV4 as appropriately.
2483  * Caller must use ip_attr_nexthop() to determine the nexthop argument.
2484  *
2485  * The caller must NOT hold conn_lock (to avoid problems with ill_refrele
2486  * causing the squeue to run doing ipcl_walk grabbing conn_lock.)
2487  *
2488  * Updates laddrp and uinfo if they are non-NULL.
2489  *
2490  * TSOL notes: The callers if ip_attr_connect must check if the destination
2491  * is different than before and in that case redo conn_update_label.
2492  * The callers of conn_connect do not need that since conn_connect
2493  * performs the conn_update_label.
2494  */
2495 int
ip_attr_connect(const conn_t * connp,ip_xmit_attr_t * ixa,const in6_addr_t * v6src,const in6_addr_t * v6dst,const in6_addr_t * v6nexthop,in_port_t dstport,in6_addr_t * laddrp,iulp_t * uinfo,uint32_t flags)2496 ip_attr_connect(const conn_t *connp, ip_xmit_attr_t *ixa,
2497     const in6_addr_t *v6src, const in6_addr_t *v6dst,
2498     const in6_addr_t *v6nexthop, in_port_t dstport, in6_addr_t *laddrp,
2499     iulp_t *uinfo, uint32_t flags)
2500 {
2501 	in6_addr_t		laddr = *v6src;
2502 	int			error;
2503 
2504 	ASSERT(MUTEX_NOT_HELD(&connp->conn_lock));
2505 
2506 	if (connp->conn_zone_is_global)
2507 		flags |= IPDF_ZONE_IS_GLOBAL;
2508 	else
2509 		flags &= ~IPDF_ZONE_IS_GLOBAL;
2510 
2511 	/*
2512 	 * Lookup the route to determine a source address and the uinfo.
2513 	 * If the ULP has a source route option then the caller will
2514 	 * have set v6nexthop to be the first hop.
2515 	 */
2516 	if (ixa->ixa_flags & IXAF_IS_IPV4) {
2517 		ipaddr_t v4dst;
2518 		ipaddr_t v4src, v4nexthop;
2519 
2520 		IN6_V4MAPPED_TO_IPADDR(v6dst, v4dst);
2521 		IN6_V4MAPPED_TO_IPADDR(v6nexthop, v4nexthop);
2522 		IN6_V4MAPPED_TO_IPADDR(v6src, v4src);
2523 
2524 		if (connp->conn_unspec_src || v4src != INADDR_ANY)
2525 			flags &= ~IPDF_SELECT_SRC;
2526 		else
2527 			flags |= IPDF_SELECT_SRC;
2528 
2529 		error = ip_set_destination_v4(&v4src, v4dst, v4nexthop, ixa,
2530 		    uinfo, flags, connp->conn_mac_mode);
2531 		IN6_IPADDR_TO_V4MAPPED(v4src, &laddr);
2532 	} else {
2533 		if (connp->conn_unspec_src || !IN6_IS_ADDR_UNSPECIFIED(v6src))
2534 			flags &= ~IPDF_SELECT_SRC;
2535 		else
2536 			flags |= IPDF_SELECT_SRC;
2537 
2538 		error = ip_set_destination_v6(&laddr, v6dst, v6nexthop, ixa,
2539 		    uinfo, flags, connp->conn_mac_mode);
2540 	}
2541 	/* Pass out some address even if we hit a RTF_REJECT etc */
2542 	if (laddrp != NULL)
2543 		*laddrp = laddr;
2544 
2545 	if (error != 0)
2546 		return (error);
2547 
2548 	if (flags & IPDF_IPSEC) {
2549 		/*
2550 		 * Set any IPsec policy in ixa. Routine also looks at ULP
2551 		 * ports.
2552 		 */
2553 		ipsec_cache_outbound_policy(connp, v6src, v6dst, dstport, ixa);
2554 	}
2555 	return (0);
2556 }
2557 
2558 /*
2559  * Connect the conn based on the addresses, conn_xmit_ipp and conn_ixa.
2560  * Assumes that conn_faddr and conn_fport are already set. As such it is not
2561  * usable for SCTP, since SCTP has multiple faddrs.
2562  *
2563  * Caller must hold conn_lock to provide atomic constency between the
2564  * conn_t's addresses and the ixa.
2565  * NOTE: this function drops and reaquires conn_lock since it can't be
2566  * held across ip_attr_connect/ip_set_destination.
2567  *
2568  * The caller needs to handle inserting in the receive-side fanout when
2569  * appropriate after conn_connect returns.
2570  */
2571 int
conn_connect(conn_t * connp,iulp_t * uinfo,uint32_t flags)2572 conn_connect(conn_t *connp, iulp_t *uinfo, uint32_t flags)
2573 {
2574 	ip_xmit_attr_t	*ixa = connp->conn_ixa;
2575 	in6_addr_t	nexthop;
2576 	in6_addr_t	saddr, faddr;
2577 	in_port_t	fport;
2578 	int		error;
2579 
2580 	ASSERT(MUTEX_HELD(&connp->conn_lock));
2581 
2582 	if (connp->conn_ipversion == IPV4_VERSION)
2583 		ixa->ixa_flags |= IXAF_IS_IPV4;
2584 	else
2585 		ixa->ixa_flags &= ~IXAF_IS_IPV4;
2586 
2587 	/* We do IPsec latching below - hence no caching in ip_attr_connect */
2588 	flags &= ~IPDF_IPSEC;
2589 
2590 	/* In case we had previously done an ip_attr_connect */
2591 	ip_attr_newdst(ixa);
2592 
2593 	/*
2594 	 * Determine the nexthop and copy the addresses before dropping
2595 	 * conn_lock.
2596 	 */
2597 	ip_attr_nexthop(&connp->conn_xmit_ipp, connp->conn_ixa,
2598 	    &connp->conn_faddr_v6, &nexthop);
2599 	saddr = connp->conn_saddr_v6;
2600 	faddr = connp->conn_faddr_v6;
2601 	fport = connp->conn_fport;
2602 
2603 	mutex_exit(&connp->conn_lock);
2604 	error = ip_attr_connect(connp, ixa, &saddr, &faddr, &nexthop, fport,
2605 	    &saddr, uinfo, flags | IPDF_VERIFY_DST);
2606 	mutex_enter(&connp->conn_lock);
2607 
2608 	/* Could have changed even if an error */
2609 	connp->conn_saddr_v6 = saddr;
2610 	if (error != 0)
2611 		return (error);
2612 
2613 	/*
2614 	 * Check whether Trusted Solaris policy allows communication with this
2615 	 * host, and pretend that the destination is unreachable if not.
2616 	 * Compute any needed label and place it in ipp_label_v4/v6.
2617 	 *
2618 	 * Later conn_build_hdr_template() takes ipp_label_v4/v6 to form
2619 	 * the packet.
2620 	 *
2621 	 * TSOL Note: Any concurrent threads would pick a different ixa
2622 	 * (and ipp if they are to change the ipp)  so we
2623 	 * don't have to worry about concurrent threads.
2624 	 */
2625 	if (is_system_labeled()) {
2626 		if (connp->conn_mlp_type != mlptSingle)
2627 			return (ECONNREFUSED);
2628 
2629 		/*
2630 		 * conn_update_label will set ipp_label* which will later
2631 		 * be used by conn_build_hdr_template.
2632 		 */
2633 		error = conn_update_label(connp, ixa,
2634 		    &connp->conn_faddr_v6, &connp->conn_xmit_ipp);
2635 		if (error != 0)
2636 			return (error);
2637 	}
2638 
2639 	/*
2640 	 * Ensure that we match on the selected local address.
2641 	 * This overrides conn_laddr in the case we had earlier bound to a
2642 	 * multicast or broadcast address.
2643 	 */
2644 	connp->conn_laddr_v6 = connp->conn_saddr_v6;
2645 
2646 	/*
2647 	 * Allow setting new policies.
2648 	 * The addresses/ports are already set, thus the IPsec policy calls
2649 	 * can handle their passed-in conn's.
2650 	 */
2651 	connp->conn_policy_cached = B_FALSE;
2652 
2653 	/*
2654 	 * Cache IPsec policy in this conn.  If we have per-socket policy,
2655 	 * we'll cache that.  If we don't, we'll inherit global policy.
2656 	 *
2657 	 * This is done before the caller inserts in the receive-side fanout.
2658 	 * Note that conn_policy_cached is set by ipsec_conn_cache_policy() even
2659 	 * for connections where we don't have a policy. This is to prevent
2660 	 * global policy lookups in the inbound path.
2661 	 *
2662 	 * If we insert before we set conn_policy_cached,
2663 	 * CONN_INBOUND_POLICY_PRESENT() check can still evaluate true
2664 	 * because global policy cound be non-empty. We normally call
2665 	 * ipsec_check_policy() for conn_policy_cached connections only if
2666 	 * conn_in_enforce_policy is set. But in this case,
2667 	 * conn_policy_cached can get set anytime since we made the
2668 	 * CONN_INBOUND_POLICY_PRESENT() check and ipsec_check_policy() is
2669 	 * called, which will make the above assumption false.  Thus, we
2670 	 * need to insert after we set conn_policy_cached.
2671 	 */
2672 	error = ipsec_conn_cache_policy(connp,
2673 	    connp->conn_ipversion == IPV4_VERSION);
2674 	if (error != 0)
2675 		return (error);
2676 
2677 	/*
2678 	 * We defer to do LSO check until here since now we have better idea
2679 	 * whether IPsec is present. If the underlying ill is LSO capable,
2680 	 * copy its capability in so the ULP can decide whether to enable LSO
2681 	 * on this connection. So far, only TCP/IPv4 is implemented, so won't
2682 	 * claim LSO for IPv6.
2683 	 *
2684 	 * Currently, won't enable LSO for IRE_LOOPBACK or IRE_LOCAL, because
2685 	 * the receiver can not handle it. Also not to enable LSO for MULTIRT.
2686 	 */
2687 	ixa->ixa_flags &= ~IXAF_LSO_CAPAB;
2688 
2689 	ASSERT(ixa->ixa_ire != NULL);
2690 	if (ixa->ixa_ipst->ips_ip_lso_outbound && (flags & IPDF_LSO) &&
2691 	    !(ixa->ixa_flags & IXAF_IPSEC_SECURE) &&
2692 	    !(ixa->ixa_ire->ire_type & (IRE_LOCAL | IRE_LOOPBACK)) &&
2693 	    !(ixa->ixa_ire->ire_flags & RTF_MULTIRT) &&
2694 	    (ixa->ixa_nce != NULL) &&
2695 	    ((ixa->ixa_flags & IXAF_IS_IPV4) ?
2696 	    ILL_LSO_TCP_IPV4_USABLE(ixa->ixa_nce->nce_ill) :
2697 	    ILL_LSO_TCP_IPV6_USABLE(ixa->ixa_nce->nce_ill))) {
2698 		ixa->ixa_lso_capab = *ixa->ixa_nce->nce_ill->ill_lso_capab;
2699 		ixa->ixa_flags |= IXAF_LSO_CAPAB;
2700 	}
2701 
2702 	/* Check whether ZEROCOPY capability is usable for this connection. */
2703 	ixa->ixa_flags &= ~IXAF_ZCOPY_CAPAB;
2704 
2705 	if ((flags & IPDF_ZCOPY) &&
2706 	    !(ixa->ixa_flags & IXAF_IPSEC_SECURE) &&
2707 	    !(ixa->ixa_ire->ire_type & (IRE_LOCAL | IRE_LOOPBACK)) &&
2708 	    !(ixa->ixa_ire->ire_flags & RTF_MULTIRT) &&
2709 	    (ixa->ixa_nce != NULL) &&
2710 	    ILL_ZCOPY_USABLE(ixa->ixa_nce->nce_ill)) {
2711 		ixa->ixa_flags |= IXAF_ZCOPY_CAPAB;
2712 	}
2713 	return (0);
2714 }
2715 
2716 /*
2717  * Predicates to check if the addresses match conn_last*
2718  */
2719 
2720 /*
2721  * Compare the conn against an address.
2722  * If using mapped addresses on AF_INET6 sockets, use the _v6 function
2723  */
2724 boolean_t
conn_same_as_last_v4(conn_t * connp,sin_t * sin)2725 conn_same_as_last_v4(conn_t *connp, sin_t *sin)
2726 {
2727 	ASSERT(connp->conn_family == AF_INET);
2728 	return (sin->sin_addr.s_addr == connp->conn_v4lastdst &&
2729 	    sin->sin_port == connp->conn_lastdstport);
2730 }
2731 
2732 /*
2733  * Compare, including for mapped addresses
2734  */
2735 boolean_t
conn_same_as_last_v6(conn_t * connp,sin6_t * sin6)2736 conn_same_as_last_v6(conn_t *connp, sin6_t *sin6)
2737 {
2738 	return (IN6_ARE_ADDR_EQUAL(&connp->conn_v6lastdst, &sin6->sin6_addr) &&
2739 	    sin6->sin6_port == connp->conn_lastdstport &&
2740 	    sin6->sin6_flowinfo == connp->conn_lastflowinfo &&
2741 	    sin6->sin6_scope_id == connp->conn_lastscopeid);
2742 }
2743 
2744 /*
2745  * Compute a label and place it in the ip_packet_t.
2746  * Handles IPv4 and IPv6.
2747  * The caller should have a correct ixa_tsl and ixa_zoneid and have
2748  * already called conn_connect or ip_attr_connect to ensure that tsol_check_dest
2749  * has been called.
2750  */
2751 int
conn_update_label(const conn_t * connp,const ip_xmit_attr_t * ixa,const in6_addr_t * v6dst,ip_pkt_t * ipp)2752 conn_update_label(const conn_t *connp, const ip_xmit_attr_t *ixa,
2753     const in6_addr_t *v6dst, ip_pkt_t *ipp)
2754 {
2755 	int		err;
2756 	ipaddr_t	v4dst;
2757 
2758 	if (IN6_IS_ADDR_V4MAPPED(v6dst)) {
2759 		uchar_t		opt_storage[IP_MAX_OPT_LENGTH];
2760 
2761 		IN6_V4MAPPED_TO_IPADDR(v6dst, v4dst);
2762 
2763 		err = tsol_compute_label_v4(ixa->ixa_tsl, ixa->ixa_zoneid,
2764 		    v4dst, opt_storage, ixa->ixa_ipst);
2765 		if (err == 0) {
2766 			/* Length contained in opt_storage[IPOPT_OLEN] */
2767 			err = optcom_pkt_set(opt_storage,
2768 			    opt_storage[IPOPT_OLEN],
2769 			    (uchar_t **)&ipp->ipp_label_v4,
2770 			    &ipp->ipp_label_len_v4);
2771 		}
2772 		if (err != 0) {
2773 			DTRACE_PROBE4(tx__ip__log__info__updatelabel,
2774 			    char *, "conn(1) failed to update options(2) "
2775 			    "on ixa(3)",
2776 			    conn_t *, connp, char *, opt_storage,
2777 			    ip_xmit_attr_t *, ixa);
2778 		}
2779 		if (ipp->ipp_label_len_v4 != 0)
2780 			ipp->ipp_fields |= IPPF_LABEL_V4;
2781 		else
2782 			ipp->ipp_fields &= ~IPPF_LABEL_V4;
2783 	} else {
2784 		uchar_t		opt_storage[TSOL_MAX_IPV6_OPTION];
2785 		uint_t		optlen;
2786 
2787 		err = tsol_compute_label_v6(ixa->ixa_tsl, ixa->ixa_zoneid,
2788 		    v6dst, opt_storage, ixa->ixa_ipst);
2789 		if (err == 0) {
2790 			/*
2791 			 * Note that ipp_label_v6 is just the option - not
2792 			 * the hopopts extension header.
2793 			 *
2794 			 * Length contained in opt_storage[IPOPT_OLEN], but
2795 			 * that doesn't include the two byte options header.
2796 			 */
2797 			optlen = opt_storage[IPOPT_OLEN];
2798 			if (optlen != 0)
2799 				optlen += 2;
2800 
2801 			err = optcom_pkt_set(opt_storage, optlen,
2802 			    (uchar_t **)&ipp->ipp_label_v6,
2803 			    &ipp->ipp_label_len_v6);
2804 		}
2805 		if (err != 0) {
2806 			DTRACE_PROBE4(tx__ip__log__info__updatelabel,
2807 			    char *, "conn(1) failed to update options(2) "
2808 			    "on ixa(3)",
2809 			    conn_t *, connp, char *, opt_storage,
2810 			    ip_xmit_attr_t *, ixa);
2811 		}
2812 		if (ipp->ipp_label_len_v6 != 0)
2813 			ipp->ipp_fields |= IPPF_LABEL_V6;
2814 		else
2815 			ipp->ipp_fields &= ~IPPF_LABEL_V6;
2816 	}
2817 	return (err);
2818 }
2819 
2820 /*
2821  * Inherit all options settings from the parent/listener to the eager.
2822  * Returns zero on success; ENOMEM if memory allocation failed.
2823  *
2824  * We assume that the eager has not had any work done i.e., the conn_ixa
2825  * and conn_xmit_ipp are all zero.
2826  * Furthermore we assume that no other thread can access the eager (because
2827  * it isn't inserted in any fanout list).
2828  */
2829 int
conn_inherit_parent(conn_t * lconnp,conn_t * econnp)2830 conn_inherit_parent(conn_t *lconnp, conn_t *econnp)
2831 {
2832 	cred_t	*credp;
2833 	int	err;
2834 	void	*notify_cookie;
2835 	uint32_t xmit_hint;
2836 
2837 	econnp->conn_family = lconnp->conn_family;
2838 	econnp->conn_ipv6_v6only = lconnp->conn_ipv6_v6only;
2839 	econnp->conn_wq = lconnp->conn_wq;
2840 	econnp->conn_rq = lconnp->conn_rq;
2841 
2842 	/*
2843 	 * Make a safe copy of the transmit attributes.
2844 	 * conn_connect will later be used by the caller to setup the ire etc.
2845 	 */
2846 	ASSERT(econnp->conn_ixa->ixa_refcnt == 1);
2847 	ASSERT(econnp->conn_ixa->ixa_ire == NULL);
2848 	ASSERT(econnp->conn_ixa->ixa_dce == NULL);
2849 	ASSERT(econnp->conn_ixa->ixa_nce == NULL);
2850 
2851 	/* Preserve ixa_notify_cookie and xmit_hint */
2852 	notify_cookie = econnp->conn_ixa->ixa_notify_cookie;
2853 	xmit_hint = econnp->conn_ixa->ixa_xmit_hint;
2854 	ixa_safe_copy(lconnp->conn_ixa, econnp->conn_ixa);
2855 	econnp->conn_ixa->ixa_notify_cookie = notify_cookie;
2856 	econnp->conn_ixa->ixa_xmit_hint = xmit_hint;
2857 
2858 	econnp->conn_bound_if = lconnp->conn_bound_if;
2859 	econnp->conn_incoming_ifindex = lconnp->conn_incoming_ifindex;
2860 
2861 	/* Inherit all RECV options */
2862 	econnp->conn_recv_ancillary = lconnp->conn_recv_ancillary;
2863 
2864 	err = ip_pkt_copy(&lconnp->conn_xmit_ipp, &econnp->conn_xmit_ipp,
2865 	    KM_NOSLEEP);
2866 	if (err != 0)
2867 		return (err);
2868 
2869 	econnp->conn_zoneid = lconnp->conn_zoneid;
2870 	econnp->conn_allzones = lconnp->conn_allzones;
2871 
2872 	/* This is odd. Pick a flowlabel for each connection instead? */
2873 	econnp->conn_flowinfo = lconnp->conn_flowinfo;
2874 
2875 	econnp->conn_default_ttl = lconnp->conn_default_ttl;
2876 
2877 	/*
2878 	 * TSOL: tsol_input_proc() needs the eager's cred before the
2879 	 * eager is accepted
2880 	 */
2881 	ASSERT(lconnp->conn_cred != NULL);
2882 	econnp->conn_cred = credp = lconnp->conn_cred;
2883 	crhold(credp);
2884 	econnp->conn_cpid = lconnp->conn_cpid;
2885 	econnp->conn_open_time = ddi_get_lbolt64();
2886 
2887 	/*
2888 	 * Cache things in the ixa without any refhold.
2889 	 * Listener might not have set up ixa_cred
2890 	 */
2891 	ASSERT(!(econnp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
2892 	econnp->conn_ixa->ixa_cred = econnp->conn_cred;
2893 	econnp->conn_ixa->ixa_cpid = econnp->conn_cpid;
2894 	if (is_system_labeled())
2895 		econnp->conn_ixa->ixa_tsl = crgetlabel(econnp->conn_cred);
2896 
2897 	/*
2898 	 * If the caller has the process-wide flag set, then default to MAC
2899 	 * exempt mode.  This allows read-down to unlabeled hosts.
2900 	 */
2901 	if (getpflags(NET_MAC_AWARE, credp) != 0)
2902 		econnp->conn_mac_mode = CONN_MAC_AWARE;
2903 
2904 	econnp->conn_zone_is_global = lconnp->conn_zone_is_global;
2905 
2906 	/*
2907 	 * We eliminate the need for sockfs to send down a T_SVR4_OPTMGMT_REQ
2908 	 * via soaccept()->soinheritoptions() which essentially applies
2909 	 * all the listener options to the new connection. The options that we
2910 	 * need to take care of are:
2911 	 * SO_DEBUG, SO_REUSEADDR, SO_KEEPALIVE, SO_DONTROUTE, SO_BROADCAST,
2912 	 * SO_USELOOPBACK, SO_OOBINLINE, SO_DGRAM_ERRIND, SO_LINGER,
2913 	 * SO_SNDBUF, SO_RCVBUF.
2914 	 *
2915 	 * SO_RCVBUF:	conn_rcvbuf is set.
2916 	 * SO_SNDBUF:	conn_sndbuf is set.
2917 	 */
2918 
2919 	/* Could we define a struct and use a struct copy for this? */
2920 	econnp->conn_sndbuf = lconnp->conn_sndbuf;
2921 	econnp->conn_rcvbuf = lconnp->conn_rcvbuf;
2922 	econnp->conn_sndlowat = lconnp->conn_sndlowat;
2923 	econnp->conn_rcvlowat = lconnp->conn_rcvlowat;
2924 	econnp->conn_dgram_errind = lconnp->conn_dgram_errind;
2925 	econnp->conn_oobinline = lconnp->conn_oobinline;
2926 	econnp->conn_debug = lconnp->conn_debug;
2927 	econnp->conn_keepalive = lconnp->conn_keepalive;
2928 	econnp->conn_linger = lconnp->conn_linger;
2929 	econnp->conn_lingertime = lconnp->conn_lingertime;
2930 
2931 	/* Set the IP options */
2932 	econnp->conn_broadcast = lconnp->conn_broadcast;
2933 	econnp->conn_useloopback = lconnp->conn_useloopback;
2934 	econnp->conn_reuseaddr = lconnp->conn_reuseaddr;
2935 	return (0);
2936 }
2937