xref: /titanic_44/usr/src/uts/common/inet/udp/udp.c (revision 2ec7cc7fc084163eaed884efee9bbd322cc8951b)
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  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/dlpi.h>
30 #include <sys/pattr.h>
31 #include <sys/stropts.h>
32 #include <sys/strlog.h>
33 #include <sys/strsun.h>
34 #include <sys/time.h>
35 #define	_SUN_TPI_VERSION 2
36 #include <sys/tihdr.h>
37 #include <sys/timod.h>
38 #include <sys/ddi.h>
39 #include <sys/sunddi.h>
40 #include <sys/strsubr.h>
41 #include <sys/suntpi.h>
42 #include <sys/xti_inet.h>
43 #include <sys/kmem.h>
44 #include <sys/policy.h>
45 #include <sys/ucred.h>
46 #include <sys/zone.h>
47 
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/sockio.h>
51 #include <sys/vtrace.h>
52 #include <sys/sdt.h>
53 #include <sys/debug.h>
54 #include <sys/isa_defs.h>
55 #include <sys/random.h>
56 #include <netinet/in.h>
57 #include <netinet/ip6.h>
58 #include <netinet/icmp6.h>
59 #include <netinet/udp.h>
60 #include <net/if.h>
61 #include <net/route.h>
62 
63 #include <inet/common.h>
64 #include <inet/ip.h>
65 #include <inet/ip_impl.h>
66 #include <inet/ip6.h>
67 #include <inet/ip_ire.h>
68 #include <inet/ip_if.h>
69 #include <inet/ip_multi.h>
70 #include <inet/ip_ndp.h>
71 #include <inet/proto_set.h>
72 #include <inet/mib2.h>
73 #include <inet/nd.h>
74 #include <inet/optcom.h>
75 #include <inet/snmpcom.h>
76 #include <inet/kstatcom.h>
77 #include <inet/udp_impl.h>
78 #include <inet/ipclassifier.h>
79 #include <inet/ipsec_impl.h>
80 #include <inet/ipp_common.h>
81 #include <sys/squeue_impl.h>
82 #include <inet/ipnet.h>
83 #include <sys/ethernet.h>
84 
85 /*
86  * The ipsec_info.h header file is here since it has the definition for the
87  * M_CTL message types used by IP to convey information to the ULP. The
88  * ipsec_info.h needs the pfkeyv2.h, hence the latter's presence.
89  */
90 #include <net/pfkeyv2.h>
91 #include <inet/ipsec_info.h>
92 
93 #include <sys/tsol/label.h>
94 #include <sys/tsol/tnet.h>
95 #include <rpc/pmap_prot.h>
96 
97 /*
98  * Synchronization notes:
99  *
100  * UDP is MT and uses the usual kernel synchronization primitives. There are 2
101  * locks, the fanout lock (uf_lock) and the udp endpoint lock udp_rwlock.
102  * We also use conn_lock when updating things that affect the IP classifier
103  * lookup.
104  * The lock order is udp_rwlock -> uf_lock and is udp_rwlock -> conn_lock.
105  *
106  * The fanout lock uf_lock:
107  * When a UDP endpoint is bound to a local port, it is inserted into
108  * a bind hash list.  The list consists of an array of udp_fanout_t buckets.
109  * The size of the array is controlled by the udp_bind_fanout_size variable.
110  * This variable can be changed in /etc/system if the default value is
111  * not large enough.  Each bind hash bucket is protected by a per bucket
112  * lock.  It protects the udp_bind_hash and udp_ptpbhn fields in the udp_t
113  * structure and a few other fields in the udp_t. A UDP endpoint is removed
114  * from the bind hash list only when it is being unbound or being closed.
115  * The per bucket lock also protects a UDP endpoint's state changes.
116  *
117  * The udp_rwlock:
118  * This protects most of the other fields in the udp_t. The exact list of
119  * fields which are protected by each of the above locks is documented in
120  * the udp_t structure definition.
121  *
122  * Plumbing notes:
123  * UDP is always a device driver. For compatibility with mibopen() code
124  * it is possible to I_PUSH "udp", but that results in pushing a passthrough
125  * dummy module.
126  *
127  * The above implies that we don't support any intermediate module to
128  * reside in between /dev/ip and udp -- in fact, we never supported such
129  * scenario in the past as the inter-layer communication semantics have
130  * always been private.
131  */
132 
133 /* For /etc/system control */
134 uint_t udp_bind_fanout_size = UDP_BIND_FANOUT_SIZE;
135 
136 /* Option processing attrs */
137 typedef struct udpattrs_s {
138 	union {
139 		ip6_pkt_t	*udpattr_ipp6;	/* For V6 */
140 		ip4_pkt_t 	*udpattr_ipp4;	/* For V4 */
141 	} udpattr_ippu;
142 #define	udpattr_ipp6 udpattr_ippu.udpattr_ipp6
143 #define	udpattr_ipp4 udpattr_ippu.udpattr_ipp4
144 	mblk_t		*udpattr_mb;
145 	boolean_t	udpattr_credset;
146 } udpattrs_t;
147 
148 static void	udp_addr_req(queue_t *q, mblk_t *mp);
149 static void	udp_tpi_bind(queue_t *q, mblk_t *mp);
150 static void	udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp);
151 static void	udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock);
152 static int	udp_build_hdrs(udp_t *udp);
153 static void	udp_capability_req(queue_t *q, mblk_t *mp);
154 static int	udp_tpi_close(queue_t *q, int flags);
155 static void	udp_tpi_connect(queue_t *q, mblk_t *mp);
156 static void	udp_tpi_disconnect(queue_t *q, mblk_t *mp);
157 static void	udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
158 		    int sys_error);
159 static void	udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive,
160 		    t_scalar_t tlierr, int unixerr);
161 static int	udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp,
162 		    cred_t *cr);
163 static int	udp_extra_priv_ports_add(queue_t *q, mblk_t *mp,
164 		    char *value, caddr_t cp, cred_t *cr);
165 static int	udp_extra_priv_ports_del(queue_t *q, mblk_t *mp,
166 		    char *value, caddr_t cp, cred_t *cr);
167 static void	udp_icmp_error(conn_t *, mblk_t *);
168 static void	udp_icmp_error_ipv6(conn_t *, mblk_t *);
169 static void	udp_info_req(queue_t *q, mblk_t *mp);
170 static void	udp_input(void *, mblk_t *, void *);
171 static void	udp_lrput(queue_t *, mblk_t *);
172 static void	udp_lwput(queue_t *, mblk_t *);
173 static int	udp_open(queue_t *q, dev_t *devp, int flag, int sflag,
174 		    cred_t *credp, boolean_t isv6);
175 static int	udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
176 		    cred_t *credp);
177 static int	udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
178 		    cred_t *credp);
179 static  int	udp_unitdata_opt_process(queue_t *q, mblk_t *mp,
180 		    int *errorp, udpattrs_t *udpattrs);
181 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
182 static int	udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
183 static boolean_t udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt);
184 static int	udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
185 		    cred_t *cr);
186 static void	udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp,
187 		    ipha_t *ipha);
188 static void	udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr,
189 		    t_scalar_t destlen, t_scalar_t err);
190 static void	udp_tpi_unbind(queue_t *q, mblk_t *mp);
191 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port,
192     boolean_t random);
193 static mblk_t	*udp_output_v4(conn_t *, mblk_t *, ipaddr_t, uint16_t, uint_t,
194 		    int *, boolean_t, struct nmsghdr *, cred_t *, pid_t);
195 static mblk_t	*udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6,
196 		    int *error, struct nmsghdr *msg, cred_t *cr, pid_t pid);
197 static void	udp_wput_other(queue_t *q, mblk_t *mp);
198 static void	udp_wput_iocdata(queue_t *q, mblk_t *mp);
199 static void	udp_wput_fallback(queue_t *q, mblk_t *mp);
200 static size_t	udp_set_rcv_hiwat(udp_t *udp, size_t size);
201 
202 static void	*udp_stack_init(netstackid_t stackid, netstack_t *ns);
203 static void	udp_stack_fini(netstackid_t stackid, void *arg);
204 
205 static void	*udp_kstat_init(netstackid_t stackid);
206 static void	udp_kstat_fini(netstackid_t stackid, kstat_t *ksp);
207 static void	*udp_kstat2_init(netstackid_t, udp_stat_t *);
208 static void	udp_kstat2_fini(netstackid_t, kstat_t *);
209 static int	udp_kstat_update(kstat_t *kp, int rw);
210 
211 static void	udp_xmit(queue_t *, mblk_t *, ire_t *ire, conn_t *, zoneid_t);
212 
213 static int	udp_send_connected(conn_t *, mblk_t *, struct nmsghdr *,
214 		    cred_t *, pid_t);
215 static void	udp_ulp_recv(conn_t *, mblk_t *);
216 
217 /* Common routine for TPI and socket module */
218 static conn_t	*udp_do_open(cred_t *, boolean_t, int);
219 static void	udp_do_close(conn_t *);
220 static int	udp_do_bind(conn_t *, struct sockaddr *, socklen_t, cred_t *,
221     boolean_t);
222 static int	udp_do_unbind(conn_t *);
223 static int	udp_do_getsockname(udp_t *, struct sockaddr *, uint_t *);
224 static int	udp_do_getpeername(udp_t *, struct sockaddr *, uint_t *);
225 
226 int		udp_getsockname(sock_lower_handle_t,
227     struct sockaddr *, socklen_t *, cred_t *);
228 int		udp_getpeername(sock_lower_handle_t,
229     struct sockaddr *, socklen_t *, cred_t *);
230 static int	udp_do_connect(conn_t *, const struct sockaddr *, socklen_t,
231     cred_t *cr);
232 static int	udp_post_ip_bind_connect(udp_t *, mblk_t *, int);
233 
234 #define	UDP_RECV_HIWATER	(56 * 1024)
235 #define	UDP_RECV_LOWATER	128
236 #define	UDP_XMIT_HIWATER	(56 * 1024)
237 #define	UDP_XMIT_LOWATER	1024
238 
239 /*
240  * The following is defined in tcp.c
241  */
242 extern int	(*cl_inet_connect2)(netstackid_t stack_id,
243 		    uint8_t protocol, boolean_t is_outgoing,
244 		    sa_family_t addr_family,
245 		    uint8_t *laddrp, in_port_t lport,
246 		    uint8_t *faddrp, in_port_t fport, void *args);
247 
248 /*
249  * Checks if the given destination addr/port is allowed out.
250  * If allowed, registers the (dest_addr/port, node_ID) mapping at Cluster.
251  * Called for each connect() and for sendto()/sendmsg() to a different
252  * destination.
253  * For connect(), called in udp_connect().
254  * For sendto()/sendmsg(), called in udp_output_v{4,6}().
255  *
256  * This macro assumes that the cl_inet_connect2 hook is not NULL.
257  * Please check this before calling this macro.
258  *
259  * void
260  * CL_INET_UDP_CONNECT(conn_t cp, udp_t *udp, boolean_t is_outgoing,
261  *     in6_addr_t *faddrp, in_port_t (or uint16_t) fport, int err);
262  */
263 #define	CL_INET_UDP_CONNECT(cp, udp, is_outgoing, faddrp, fport, err) {	\
264 	(err) = 0;							\
265 	/*								\
266 	 * Running in cluster mode - check and register active		\
267 	 * "connection" information					\
268 	 */								\
269 	if ((udp)->udp_ipversion == IPV4_VERSION)			\
270 		(err) = (*cl_inet_connect2)(				\
271 		    (cp)->conn_netstack->netstack_stackid,		\
272 		    IPPROTO_UDP, is_outgoing, AF_INET,			\
273 		    (uint8_t *)&((udp)->udp_v6src._S6_un._S6_u32[3]),	\
274 		    (udp)->udp_port,					\
275 		    (uint8_t *)&((faddrp)->_S6_un._S6_u32[3]),		\
276 		    (in_port_t)(fport), NULL);				\
277 	else								\
278 		(err) = (*cl_inet_connect2)(				\
279 		    (cp)->conn_netstack->netstack_stackid,		\
280 		    IPPROTO_UDP, is_outgoing, AF_INET6,			\
281 		    (uint8_t *)&((udp)->udp_v6src), (udp)->udp_port,	\
282 		    (uint8_t *)(faddrp), (in_port_t)(fport), NULL);	\
283 }
284 
285 static struct module_info udp_mod_info =  {
286 	UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER
287 };
288 
289 /*
290  * Entry points for UDP as a device.
291  * We have separate open functions for the /dev/udp and /dev/udp6 devices.
292  */
293 static struct qinit udp_rinitv4 = {
294 	NULL, NULL, udp_openv4, udp_tpi_close, NULL, &udp_mod_info, NULL
295 };
296 
297 static struct qinit udp_rinitv6 = {
298 	NULL, NULL, udp_openv6, udp_tpi_close, NULL, &udp_mod_info, NULL
299 };
300 
301 static struct qinit udp_winit = {
302 	(pfi_t)udp_wput, (pfi_t)ip_wsrv, NULL, NULL, NULL, &udp_mod_info
303 };
304 
305 /* UDP entry point during fallback */
306 struct qinit udp_fallback_sock_winit = {
307 	(pfi_t)udp_wput_fallback, NULL, NULL, NULL, NULL, &udp_mod_info
308 };
309 
310 /*
311  * UDP needs to handle I_LINK and I_PLINK since ifconfig
312  * likes to use it as a place to hang the various streams.
313  */
314 static struct qinit udp_lrinit = {
315 	(pfi_t)udp_lrput, NULL, udp_openv4, udp_tpi_close, NULL, &udp_mod_info
316 };
317 
318 static struct qinit udp_lwinit = {
319 	(pfi_t)udp_lwput, NULL, udp_openv4, udp_tpi_close, NULL, &udp_mod_info
320 };
321 
322 /* For AF_INET aka /dev/udp */
323 struct streamtab udpinfov4 = {
324 	&udp_rinitv4, &udp_winit, &udp_lrinit, &udp_lwinit
325 };
326 
327 /* For AF_INET6 aka /dev/udp6 */
328 struct streamtab udpinfov6 = {
329 	&udp_rinitv6, &udp_winit, &udp_lrinit, &udp_lwinit
330 };
331 
332 static	sin_t	sin_null;	/* Zero address for quick clears */
333 static	sin6_t	sin6_null;	/* Zero address for quick clears */
334 
335 #define	UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH)
336 
337 /* Default structure copied into T_INFO_ACK messages */
338 static struct T_info_ack udp_g_t_info_ack_ipv4 = {
339 	T_INFO_ACK,
340 	UDP_MAXPACKET_IPV4,	/* TSDU_size. Excl. headers */
341 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
342 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
343 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
344 	sizeof (sin_t),	/* ADDR_size. */
345 	0,		/* OPT_size - not initialized here */
346 	UDP_MAXPACKET_IPV4,	/* TIDU_size.  Excl. headers */
347 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
348 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
349 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
350 };
351 
352 #define	UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN)
353 
354 static	struct T_info_ack udp_g_t_info_ack_ipv6 = {
355 	T_INFO_ACK,
356 	UDP_MAXPACKET_IPV6,	/* TSDU_size.  Excl. headers */
357 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
358 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
359 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
360 	sizeof (sin6_t), /* ADDR_size. */
361 	0,		/* OPT_size - not initialized here */
362 	UDP_MAXPACKET_IPV6,	/* TIDU_size. Excl. headers */
363 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
364 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
365 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
366 };
367 
368 /* largest UDP port number */
369 #define	UDP_MAX_PORT	65535
370 
371 /*
372  * Table of ND variables supported by udp.  These are loaded into us_nd
373  * in udp_open.
374  * All of these are alterable, within the min/max values given, at run time.
375  */
376 /* BEGIN CSTYLED */
377 udpparam_t udp_param_arr[] = {
378  /*min		max		value		name */
379  { 0L,		256,		32,		"udp_wroff_extra" },
380  { 1L,		255,		255,		"udp_ipv4_ttl" },
381  { 0,		IPV6_MAX_HOPS,	IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"},
382  { 1024,	(32 * 1024),	1024,		"udp_smallest_nonpriv_port" },
383  { 0,		1,		1,		"udp_do_checksum" },
384  { 1024,	UDP_MAX_PORT,	(32 * 1024),	"udp_smallest_anon_port" },
385  { 1024,	UDP_MAX_PORT,	UDP_MAX_PORT,	"udp_largest_anon_port" },
386  { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER,	"udp_xmit_hiwat"},
387  { 0,		     (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"},
388  { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER,	"udp_recv_hiwat"},
389  { 65536,	(1<<30),	2*1024*1024,	"udp_max_buf"},
390 };
391 /* END CSTYLED */
392 
393 /* Setable in /etc/system */
394 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */
395 uint32_t udp_random_anon_port = 1;
396 
397 /*
398  * Hook functions to enable cluster networking.
399  * On non-clustered systems these vectors must always be NULL
400  */
401 
402 void (*cl_inet_bind)(netstackid_t stack_id, uchar_t protocol,
403     sa_family_t addr_family, uint8_t *laddrp, in_port_t lport,
404     void *args) = NULL;
405 void (*cl_inet_unbind)(netstackid_t stack_id, uint8_t protocol,
406     sa_family_t addr_family, uint8_t *laddrp, in_port_t lport,
407     void *args) = NULL;
408 
409 typedef union T_primitives *t_primp_t;
410 
411 /*
412  * Return the next anonymous port in the privileged port range for
413  * bind checking.
414  *
415  * Trusted Extension (TX) notes: TX allows administrator to mark or
416  * reserve ports as Multilevel ports (MLP). MLP has special function
417  * on TX systems. Once a port is made MLP, it's not available as
418  * ordinary port. This creates "holes" in the port name space. It
419  * may be necessary to skip the "holes" find a suitable anon port.
420  */
421 static in_port_t
422 udp_get_next_priv_port(udp_t *udp)
423 {
424 	static in_port_t next_priv_port = IPPORT_RESERVED - 1;
425 	in_port_t nextport;
426 	boolean_t restart = B_FALSE;
427 	udp_stack_t *us = udp->udp_us;
428 
429 retry:
430 	if (next_priv_port < us->us_min_anonpriv_port ||
431 	    next_priv_port >= IPPORT_RESERVED) {
432 		next_priv_port = IPPORT_RESERVED - 1;
433 		if (restart)
434 			return (0);
435 		restart = B_TRUE;
436 	}
437 
438 	if (is_system_labeled() &&
439 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
440 	    next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) {
441 		next_priv_port = nextport;
442 		goto retry;
443 	}
444 
445 	return (next_priv_port--);
446 }
447 
448 /*
449  * Hash list removal routine for udp_t structures.
450  */
451 static void
452 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock)
453 {
454 	udp_t	*udpnext;
455 	kmutex_t *lockp;
456 	udp_stack_t *us = udp->udp_us;
457 
458 	if (udp->udp_ptpbhn == NULL)
459 		return;
460 
461 	/*
462 	 * Extract the lock pointer in case there are concurrent
463 	 * hash_remove's for this instance.
464 	 */
465 	ASSERT(udp->udp_port != 0);
466 	if (!caller_holds_lock) {
467 		lockp = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
468 		    us->us_bind_fanout_size)].uf_lock;
469 		ASSERT(lockp != NULL);
470 		mutex_enter(lockp);
471 	}
472 	if (udp->udp_ptpbhn != NULL) {
473 		udpnext = udp->udp_bind_hash;
474 		if (udpnext != NULL) {
475 			udpnext->udp_ptpbhn = udp->udp_ptpbhn;
476 			udp->udp_bind_hash = NULL;
477 		}
478 		*udp->udp_ptpbhn = udpnext;
479 		udp->udp_ptpbhn = NULL;
480 	}
481 	if (!caller_holds_lock) {
482 		mutex_exit(lockp);
483 	}
484 }
485 
486 static void
487 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp)
488 {
489 	udp_t	**udpp;
490 	udp_t	*udpnext;
491 
492 	ASSERT(MUTEX_HELD(&uf->uf_lock));
493 	ASSERT(udp->udp_ptpbhn == NULL);
494 	udpp = &uf->uf_udp;
495 	udpnext = udpp[0];
496 	if (udpnext != NULL) {
497 		/*
498 		 * If the new udp bound to the INADDR_ANY address
499 		 * and the first one in the list is not bound to
500 		 * INADDR_ANY we skip all entries until we find the
501 		 * first one bound to INADDR_ANY.
502 		 * This makes sure that applications binding to a
503 		 * specific address get preference over those binding to
504 		 * INADDR_ANY.
505 		 */
506 		if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) &&
507 		    !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) {
508 			while ((udpnext = udpp[0]) != NULL &&
509 			    !V6_OR_V4_INADDR_ANY(
510 			    udpnext->udp_bound_v6src)) {
511 				udpp = &(udpnext->udp_bind_hash);
512 			}
513 			if (udpnext != NULL)
514 				udpnext->udp_ptpbhn = &udp->udp_bind_hash;
515 		} else {
516 			udpnext->udp_ptpbhn = &udp->udp_bind_hash;
517 		}
518 	}
519 	udp->udp_bind_hash = udpnext;
520 	udp->udp_ptpbhn = udpp;
521 	udpp[0] = udp;
522 }
523 
524 /*
525  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
526  * passed to udp_wput.
527  * It associates a port number and local address with the stream.
528  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP
529  * protocol type (IPPROTO_UDP) placed in the message following the address.
530  * A T_BIND_ACK message is passed upstream when ip acknowledges the request.
531  * (Called as writer.)
532  *
533  * Note that UDP over IPv4 and IPv6 sockets can use the same port number
534  * without setting SO_REUSEADDR. This is needed so that they
535  * can be viewed as two independent transport protocols.
536  * However, anonymouns ports are allocated from the same range to avoid
537  * duplicating the us->us_next_port_to_try.
538  */
539 static void
540 udp_tpi_bind(queue_t *q, mblk_t *mp)
541 {
542 	sin_t		*sin;
543 	sin6_t		*sin6;
544 	mblk_t		*mp1;
545 	struct T_bind_req *tbr;
546 	conn_t		*connp;
547 	udp_t		*udp;
548 	int		error;
549 	struct sockaddr	*sa;
550 	cred_t		*cr;
551 
552 	/*
553 	 * All Solaris components should pass a db_credp
554 	 * for this TPI message, hence we ASSERT.
555 	 * But in case there is some other M_PROTO that looks
556 	 * like a TPI message sent by some other kernel
557 	 * component, we check and return an error.
558 	 */
559 	cr = msg_getcred(mp, NULL);
560 	ASSERT(cr != NULL);
561 	if (cr == NULL) {
562 		udp_err_ack(q, mp, TSYSERR, EINVAL);
563 		return;
564 	}
565 
566 	connp = Q_TO_CONN(q);
567 	udp = connp->conn_udp;
568 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
569 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
570 		    "udp_bind: bad req, len %u",
571 		    (uint_t)(mp->b_wptr - mp->b_rptr));
572 		udp_err_ack(q, mp, TPROTO, 0);
573 		return;
574 	}
575 	if (udp->udp_state != TS_UNBND) {
576 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
577 		    "udp_bind: bad state, %u", udp->udp_state);
578 		udp_err_ack(q, mp, TOUTSTATE, 0);
579 		return;
580 	}
581 	/*
582 	 * Reallocate the message to make sure we have enough room for an
583 	 * address and the protocol type.
584 	 */
585 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
586 	if (!mp1) {
587 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
588 		return;
589 	}
590 
591 	mp = mp1;
592 
593 	/* Reset the message type in preparation for shipping it back. */
594 	DB_TYPE(mp) = M_PCPROTO;
595 
596 	tbr = (struct T_bind_req *)mp->b_rptr;
597 	switch (tbr->ADDR_length) {
598 	case 0:			/* Request for a generic port */
599 		tbr->ADDR_offset = sizeof (struct T_bind_req);
600 		if (udp->udp_family == AF_INET) {
601 			tbr->ADDR_length = sizeof (sin_t);
602 			sin = (sin_t *)&tbr[1];
603 			*sin = sin_null;
604 			sin->sin_family = AF_INET;
605 			mp->b_wptr = (uchar_t *)&sin[1];
606 			sa = (struct sockaddr *)sin;
607 		} else {
608 			ASSERT(udp->udp_family == AF_INET6);
609 			tbr->ADDR_length = sizeof (sin6_t);
610 			sin6 = (sin6_t *)&tbr[1];
611 			*sin6 = sin6_null;
612 			sin6->sin6_family = AF_INET6;
613 			mp->b_wptr = (uchar_t *)&sin6[1];
614 			sa = (struct sockaddr *)sin6;
615 		}
616 		break;
617 
618 	case sizeof (sin_t):	/* Complete IPv4 address */
619 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
620 		    sizeof (sin_t));
621 		if (sa == NULL || !OK_32PTR((char *)sa)) {
622 			udp_err_ack(q, mp, TSYSERR, EINVAL);
623 			return;
624 		}
625 		if (udp->udp_family != AF_INET ||
626 		    sa->sa_family != AF_INET) {
627 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
628 			return;
629 		}
630 		break;
631 
632 	case sizeof (sin6_t):	/* complete IPv6 address */
633 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
634 		    sizeof (sin6_t));
635 		if (sa == NULL || !OK_32PTR((char *)sa)) {
636 			udp_err_ack(q, mp, TSYSERR, EINVAL);
637 			return;
638 		}
639 		if (udp->udp_family != AF_INET6 ||
640 		    sa->sa_family != AF_INET6) {
641 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
642 			return;
643 		}
644 		break;
645 
646 	default:		/* Invalid request */
647 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
648 		    "udp_bind: bad ADDR_length length %u", tbr->ADDR_length);
649 		udp_err_ack(q, mp, TBADADDR, 0);
650 		return;
651 	}
652 
653 	error = udp_do_bind(connp, sa, tbr->ADDR_length, cr,
654 	    tbr->PRIM_type != O_T_BIND_REQ);
655 
656 	if (error != 0) {
657 		if (error > 0) {
658 			udp_err_ack(q, mp, TSYSERR, error);
659 		} else {
660 			udp_err_ack(q, mp, -error, 0);
661 		}
662 	} else {
663 		tbr->PRIM_type = T_BIND_ACK;
664 		qreply(q, mp);
665 	}
666 }
667 
668 /*
669  * This routine handles each T_CONN_REQ message passed to udp.  It
670  * associates a default destination address with the stream.
671  *
672  * This routine sends down a T_BIND_REQ to IP with the following mblks:
673  *	T_BIND_REQ	- specifying local and remote address/port
674  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
675  *	T_OK_ACK	- for the T_CONN_REQ
676  *	T_CONN_CON	- to keep the TPI user happy
677  *
678  * The connect completes in udp_do_connect.
679  * When a T_BIND_ACK is received information is extracted from the IRE
680  * and the two appended messages are sent to the TPI user.
681  * Should udp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
682  * convert it to an error ack for the appropriate primitive.
683  */
684 static void
685 udp_tpi_connect(queue_t *q, mblk_t *mp)
686 {
687 	udp_t	*udp;
688 	conn_t	*connp = Q_TO_CONN(q);
689 	int	error;
690 	socklen_t	len;
691 	struct sockaddr		*sa;
692 	struct T_conn_req	*tcr;
693 	cred_t		*cr;
694 
695 	/*
696 	 * All Solaris components should pass a db_credp
697 	 * for this TPI message, hence we ASSERT.
698 	 * But in case there is some other M_PROTO that looks
699 	 * like a TPI message sent by some other kernel
700 	 * component, we check and return an error.
701 	 */
702 	cr = msg_getcred(mp, NULL);
703 	ASSERT(cr != NULL);
704 	if (cr == NULL) {
705 		udp_err_ack(q, mp, TSYSERR, EINVAL);
706 		return;
707 	}
708 
709 	udp = connp->conn_udp;
710 	tcr = (struct T_conn_req *)mp->b_rptr;
711 
712 	/* A bit of sanity checking */
713 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
714 		udp_err_ack(q, mp, TPROTO, 0);
715 		return;
716 	}
717 
718 	if (tcr->OPT_length != 0) {
719 		udp_err_ack(q, mp, TBADOPT, 0);
720 		return;
721 	}
722 
723 	/*
724 	 * Determine packet type based on type of address passed in
725 	 * the request should contain an IPv4 or IPv6 address.
726 	 * Make sure that address family matches the type of
727 	 * family of the the address passed down
728 	 */
729 	len = tcr->DEST_length;
730 	switch (tcr->DEST_length) {
731 	default:
732 		udp_err_ack(q, mp, TBADADDR, 0);
733 		return;
734 
735 	case sizeof (sin_t):
736 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
737 		    sizeof (sin_t));
738 		break;
739 
740 	case sizeof (sin6_t):
741 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
742 		    sizeof (sin6_t));
743 		break;
744 	}
745 
746 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
747 	if (error != 0) {
748 		udp_err_ack(q, mp, TSYSERR, error);
749 		return;
750 	}
751 
752 	error = udp_do_connect(connp, sa, len, cr);
753 	if (error != 0) {
754 		if (error < 0)
755 			udp_err_ack(q, mp, -error, 0);
756 		else
757 			udp_err_ack(q, mp, TSYSERR, error);
758 	} else {
759 		mblk_t	*mp1;
760 		/*
761 		 * We have to send a connection confirmation to
762 		 * keep TLI happy.
763 		 */
764 		if (udp->udp_family == AF_INET) {
765 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
766 			    sizeof (sin_t), NULL, 0);
767 		} else {
768 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
769 			    sizeof (sin6_t), NULL, 0);
770 		}
771 		if (mp1 == NULL) {
772 			udp_err_ack(q, mp, TSYSERR, ENOMEM);
773 			return;
774 		}
775 
776 		/*
777 		 * Send ok_ack for T_CONN_REQ
778 		 */
779 		mp = mi_tpi_ok_ack_alloc(mp);
780 		if (mp == NULL) {
781 			/* Unable to reuse the T_CONN_REQ for the ack. */
782 			udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
783 			return;
784 		}
785 
786 		putnext(connp->conn_rq, mp);
787 		putnext(connp->conn_rq, mp1);
788 	}
789 }
790 
791 static int
792 udp_tpi_close(queue_t *q, int flags)
793 {
794 	conn_t	*connp;
795 
796 	if (flags & SO_FALLBACK) {
797 		/*
798 		 * stream is being closed while in fallback
799 		 * simply free the resources that were allocated
800 		 */
801 		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
802 		qprocsoff(q);
803 		goto done;
804 	}
805 
806 	connp = Q_TO_CONN(q);
807 	udp_do_close(connp);
808 done:
809 	q->q_ptr = WR(q)->q_ptr = NULL;
810 	return (0);
811 }
812 
813 /*
814  * Called in the close path to quiesce the conn
815  */
816 void
817 udp_quiesce_conn(conn_t *connp)
818 {
819 	udp_t	*udp = connp->conn_udp;
820 
821 	if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) {
822 		/*
823 		 * Running in cluster mode - register unbind information
824 		 */
825 		if (udp->udp_ipversion == IPV4_VERSION) {
826 			(*cl_inet_unbind)(
827 			    connp->conn_netstack->netstack_stackid,
828 			    IPPROTO_UDP, AF_INET,
829 			    (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))),
830 			    (in_port_t)udp->udp_port, NULL);
831 		} else {
832 			(*cl_inet_unbind)(
833 			    connp->conn_netstack->netstack_stackid,
834 			    IPPROTO_UDP, AF_INET6,
835 			    (uint8_t *)(&(udp->udp_v6src)),
836 			    (in_port_t)udp->udp_port, NULL);
837 		}
838 	}
839 
840 	udp_bind_hash_remove(udp, B_FALSE);
841 
842 }
843 
844 void
845 udp_close_free(conn_t *connp)
846 {
847 	udp_t *udp = connp->conn_udp;
848 
849 	/* If there are any options associated with the stream, free them. */
850 	if (udp->udp_ip_snd_options != NULL) {
851 		mi_free((char *)udp->udp_ip_snd_options);
852 		udp->udp_ip_snd_options = NULL;
853 		udp->udp_ip_snd_options_len = 0;
854 	}
855 
856 	if (udp->udp_ip_rcv_options != NULL) {
857 		mi_free((char *)udp->udp_ip_rcv_options);
858 		udp->udp_ip_rcv_options = NULL;
859 		udp->udp_ip_rcv_options_len = 0;
860 	}
861 
862 	/* Free memory associated with sticky options */
863 	if (udp->udp_sticky_hdrs_len != 0) {
864 		kmem_free(udp->udp_sticky_hdrs,
865 		    udp->udp_sticky_hdrs_len);
866 		udp->udp_sticky_hdrs = NULL;
867 		udp->udp_sticky_hdrs_len = 0;
868 	}
869 	if (udp->udp_last_cred != NULL) {
870 		crfree(udp->udp_last_cred);
871 		udp->udp_last_cred = NULL;
872 	}
873 	if (udp->udp_effective_cred != NULL) {
874 		crfree(udp->udp_effective_cred);
875 		udp->udp_effective_cred = NULL;
876 	}
877 
878 	ip6_pkt_free(&udp->udp_sticky_ipp);
879 
880 	/*
881 	 * Clear any fields which the kmem_cache constructor clears.
882 	 * Only udp_connp needs to be preserved.
883 	 * TBD: We should make this more efficient to avoid clearing
884 	 * everything.
885 	 */
886 	ASSERT(udp->udp_connp == connp);
887 	bzero(udp, sizeof (udp_t));
888 	udp->udp_connp = connp;
889 }
890 
891 static int
892 udp_do_disconnect(conn_t *connp)
893 {
894 	udp_t	*udp;
895 	mblk_t	*ire_mp;
896 	udp_fanout_t *udpf;
897 	udp_stack_t *us;
898 	int	error;
899 
900 	udp = connp->conn_udp;
901 	us = udp->udp_us;
902 	rw_enter(&udp->udp_rwlock, RW_WRITER);
903 	if (udp->udp_state != TS_DATA_XFER || udp->udp_pending_op != -1) {
904 		rw_exit(&udp->udp_rwlock);
905 		return (-TOUTSTATE);
906 	}
907 	udp->udp_pending_op = T_DISCON_REQ;
908 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
909 	    us->us_bind_fanout_size)];
910 	mutex_enter(&udpf->uf_lock);
911 	udp->udp_v6src = udp->udp_bound_v6src;
912 	udp->udp_state = TS_IDLE;
913 	mutex_exit(&udpf->uf_lock);
914 
915 	if (udp->udp_family == AF_INET6) {
916 		/* Rebuild the header template */
917 		error = udp_build_hdrs(udp);
918 		if (error != 0) {
919 			udp->udp_pending_op = -1;
920 			rw_exit(&udp->udp_rwlock);
921 			return (error);
922 		}
923 	}
924 
925 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
926 	if (ire_mp == NULL) {
927 		mutex_enter(&udpf->uf_lock);
928 		udp->udp_pending_op = -1;
929 		mutex_exit(&udpf->uf_lock);
930 		rw_exit(&udp->udp_rwlock);
931 		return (ENOMEM);
932 	}
933 
934 	rw_exit(&udp->udp_rwlock);
935 
936 	if (udp->udp_family == AF_INET6) {
937 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, IPPROTO_UDP,
938 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
939 	} else {
940 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, IPPROTO_UDP,
941 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port, B_TRUE);
942 	}
943 
944 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
945 }
946 
947 
948 static void
949 udp_tpi_disconnect(queue_t *q, mblk_t *mp)
950 {
951 	conn_t	*connp = Q_TO_CONN(q);
952 	int	error;
953 
954 	/*
955 	 * Allocate the largest primitive we need to send back
956 	 * T_error_ack is > than T_ok_ack
957 	 */
958 	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
959 	if (mp == NULL) {
960 		/* Unable to reuse the T_DISCON_REQ for the ack. */
961 		udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
962 		return;
963 	}
964 
965 	error = udp_do_disconnect(connp);
966 
967 	if (error != 0) {
968 		if (error < 0) {
969 			udp_err_ack(q, mp, -error, 0);
970 		} else {
971 			udp_err_ack(q, mp, TSYSERR, error);
972 		}
973 	} else {
974 		mp = mi_tpi_ok_ack_alloc(mp);
975 		ASSERT(mp != NULL);
976 		qreply(q, mp);
977 	}
978 }
979 
980 int
981 udp_disconnect(conn_t *connp)
982 {
983 	int error;
984 	udp_t *udp = connp->conn_udp;
985 
986 	udp->udp_dgram_errind = B_FALSE;
987 
988 	error = udp_do_disconnect(connp);
989 
990 	if (error < 0)
991 		error = proto_tlitosyserr(-error);
992 
993 	return (error);
994 }
995 
996 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
997 static void
998 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
999 {
1000 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1001 		qreply(q, mp);
1002 }
1003 
1004 /* Shorthand to generate and send TPI error acks to our client */
1005 static void
1006 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error,
1007     int sys_error)
1008 {
1009 	struct T_error_ack	*teackp;
1010 
1011 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1012 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1013 		teackp = (struct T_error_ack *)mp->b_rptr;
1014 		teackp->ERROR_prim = primitive;
1015 		teackp->TLI_error = t_error;
1016 		teackp->UNIX_error = sys_error;
1017 		qreply(q, mp);
1018 	}
1019 }
1020 
1021 /*ARGSUSED*/
1022 static int
1023 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
1024 {
1025 	int i;
1026 	udp_t		*udp = Q_TO_UDP(q);
1027 	udp_stack_t *us = udp->udp_us;
1028 
1029 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1030 		if (us->us_epriv_ports[i] != 0)
1031 			(void) mi_mpprintf(mp, "%d ", us->us_epriv_ports[i]);
1032 	}
1033 	return (0);
1034 }
1035 
1036 /* ARGSUSED */
1037 static int
1038 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1039     cred_t *cr)
1040 {
1041 	long	new_value;
1042 	int	i;
1043 	udp_t		*udp = Q_TO_UDP(q);
1044 	udp_stack_t *us = udp->udp_us;
1045 
1046 	/*
1047 	 * Fail the request if the new value does not lie within the
1048 	 * port number limits.
1049 	 */
1050 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1051 	    new_value <= 0 || new_value >= 65536) {
1052 		return (EINVAL);
1053 	}
1054 
1055 	/* Check if the value is already in the list */
1056 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1057 		if (new_value == us->us_epriv_ports[i]) {
1058 			return (EEXIST);
1059 		}
1060 	}
1061 	/* Find an empty slot */
1062 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1063 		if (us->us_epriv_ports[i] == 0)
1064 			break;
1065 	}
1066 	if (i == us->us_num_epriv_ports) {
1067 		return (EOVERFLOW);
1068 	}
1069 
1070 	/* Set the new value */
1071 	us->us_epriv_ports[i] = (in_port_t)new_value;
1072 	return (0);
1073 }
1074 
1075 /* ARGSUSED */
1076 static int
1077 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1078     cred_t *cr)
1079 {
1080 	long	new_value;
1081 	int	i;
1082 	udp_t		*udp = Q_TO_UDP(q);
1083 	udp_stack_t *us = udp->udp_us;
1084 
1085 	/*
1086 	 * Fail the request if the new value does not lie within the
1087 	 * port number limits.
1088 	 */
1089 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1090 	    new_value <= 0 || new_value >= 65536) {
1091 		return (EINVAL);
1092 	}
1093 
1094 	/* Check that the value is already in the list */
1095 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1096 		if (us->us_epriv_ports[i] == new_value)
1097 			break;
1098 	}
1099 	if (i == us->us_num_epriv_ports) {
1100 		return (ESRCH);
1101 	}
1102 
1103 	/* Clear the value */
1104 	us->us_epriv_ports[i] = 0;
1105 	return (0);
1106 }
1107 
1108 /* At minimum we need 4 bytes of UDP header */
1109 #define	ICMP_MIN_UDP_HDR	4
1110 
1111 /*
1112  * udp_icmp_error is called by udp_input to process ICMP msgs. passed up by IP.
1113  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1114  * Assumes that IP has pulled up everything up to and including the ICMP header.
1115  */
1116 static void
1117 udp_icmp_error(conn_t *connp, mblk_t *mp)
1118 {
1119 	icmph_t *icmph;
1120 	ipha_t	*ipha;
1121 	int	iph_hdr_length;
1122 	udpha_t	*udpha;
1123 	sin_t	sin;
1124 	sin6_t	sin6;
1125 	mblk_t	*mp1;
1126 	int	error = 0;
1127 	udp_t	*udp = connp->conn_udp;
1128 
1129 	mp1 = NULL;
1130 	ipha = (ipha_t *)mp->b_rptr;
1131 
1132 	ASSERT(OK_32PTR(mp->b_rptr));
1133 
1134 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1135 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1136 		udp_icmp_error_ipv6(connp, mp);
1137 		return;
1138 	}
1139 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
1140 
1141 	/* Skip past the outer IP and ICMP headers */
1142 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1143 	icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length];
1144 	ipha = (ipha_t *)&icmph[1];
1145 
1146 	/* Skip past the inner IP and find the ULP header */
1147 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1148 	udpha = (udpha_t *)((char *)ipha + iph_hdr_length);
1149 
1150 	switch (icmph->icmph_type) {
1151 	case ICMP_DEST_UNREACHABLE:
1152 		switch (icmph->icmph_code) {
1153 		case ICMP_FRAGMENTATION_NEEDED:
1154 			/*
1155 			 * IP has already adjusted the path MTU.
1156 			 */
1157 			break;
1158 		case ICMP_PORT_UNREACHABLE:
1159 		case ICMP_PROTOCOL_UNREACHABLE:
1160 			error = ECONNREFUSED;
1161 			break;
1162 		default:
1163 			/* Transient errors */
1164 			break;
1165 		}
1166 		break;
1167 	default:
1168 		/* Transient errors */
1169 		break;
1170 	}
1171 	if (error == 0) {
1172 		freemsg(mp);
1173 		return;
1174 	}
1175 
1176 	/*
1177 	 * Deliver T_UDERROR_IND when the application has asked for it.
1178 	 * The socket layer enables this automatically when connected.
1179 	 */
1180 	if (!udp->udp_dgram_errind) {
1181 		freemsg(mp);
1182 		return;
1183 	}
1184 
1185 
1186 	switch (udp->udp_family) {
1187 	case AF_INET:
1188 		sin = sin_null;
1189 		sin.sin_family = AF_INET;
1190 		sin.sin_addr.s_addr = ipha->ipha_dst;
1191 		sin.sin_port = udpha->uha_dst_port;
1192 		if (IPCL_IS_NONSTR(connp)) {
1193 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1194 			if (udp->udp_state == TS_DATA_XFER) {
1195 				if (sin.sin_port == udp->udp_dstport &&
1196 				    sin.sin_addr.s_addr ==
1197 				    V4_PART_OF_V6(udp->udp_v6dst)) {
1198 					rw_exit(&udp->udp_rwlock);
1199 					(*connp->conn_upcalls->su_set_error)
1200 					    (connp->conn_upper_handle, error);
1201 					goto done;
1202 				}
1203 			} else {
1204 				udp->udp_delayed_error = error;
1205 				*((sin_t *)&udp->udp_delayed_addr) = sin;
1206 			}
1207 			rw_exit(&udp->udp_rwlock);
1208 		} else {
1209 			mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t),
1210 			    NULL, 0, error);
1211 		}
1212 		break;
1213 	case AF_INET6:
1214 		sin6 = sin6_null;
1215 		sin6.sin6_family = AF_INET6;
1216 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
1217 		sin6.sin6_port = udpha->uha_dst_port;
1218 		if (IPCL_IS_NONSTR(connp)) {
1219 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1220 			if (udp->udp_state == TS_DATA_XFER) {
1221 				if (sin6.sin6_port == udp->udp_dstport &&
1222 				    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1223 				    &udp->udp_v6dst)) {
1224 					rw_exit(&udp->udp_rwlock);
1225 					(*connp->conn_upcalls->su_set_error)
1226 					    (connp->conn_upper_handle, error);
1227 					goto done;
1228 				}
1229 			} else {
1230 				udp->udp_delayed_error = error;
1231 				*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1232 			}
1233 			rw_exit(&udp->udp_rwlock);
1234 		} else {
1235 			mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1236 			    NULL, 0, error);
1237 		}
1238 		break;
1239 	}
1240 	if (mp1 != NULL)
1241 		putnext(connp->conn_rq, mp1);
1242 done:
1243 	ASSERT(!RW_ISWRITER(&udp->udp_rwlock));
1244 	freemsg(mp);
1245 }
1246 
1247 /*
1248  * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6.
1249  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1250  * Assumes that IP has pulled up all the extension headers as well as the
1251  * ICMPv6 header.
1252  */
1253 static void
1254 udp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
1255 {
1256 	icmp6_t		*icmp6;
1257 	ip6_t		*ip6h, *outer_ip6h;
1258 	uint16_t	iph_hdr_length;
1259 	uint8_t		*nexthdrp;
1260 	udpha_t		*udpha;
1261 	sin6_t		sin6;
1262 	mblk_t		*mp1;
1263 	int		error = 0;
1264 	udp_t		*udp = connp->conn_udp;
1265 	udp_stack_t	*us = udp->udp_us;
1266 
1267 	outer_ip6h = (ip6_t *)mp->b_rptr;
1268 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1269 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1270 	else
1271 		iph_hdr_length = IPV6_HDR_LEN;
1272 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1273 	ip6h = (ip6_t *)&icmp6[1];
1274 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1275 		freemsg(mp);
1276 		return;
1277 	}
1278 	udpha = (udpha_t *)((char *)ip6h + iph_hdr_length);
1279 
1280 	switch (icmp6->icmp6_type) {
1281 	case ICMP6_DST_UNREACH:
1282 		switch (icmp6->icmp6_code) {
1283 		case ICMP6_DST_UNREACH_NOPORT:
1284 			error = ECONNREFUSED;
1285 			break;
1286 		case ICMP6_DST_UNREACH_ADMIN:
1287 		case ICMP6_DST_UNREACH_NOROUTE:
1288 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1289 		case ICMP6_DST_UNREACH_ADDR:
1290 			/* Transient errors */
1291 			break;
1292 		default:
1293 			break;
1294 		}
1295 		break;
1296 	case ICMP6_PACKET_TOO_BIG: {
1297 		struct T_unitdata_ind	*tudi;
1298 		struct T_opthdr		*toh;
1299 		size_t			udi_size;
1300 		mblk_t			*newmp;
1301 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1302 		    sizeof (struct ip6_mtuinfo);
1303 		sin6_t			*sin6;
1304 		struct ip6_mtuinfo	*mtuinfo;
1305 
1306 		/*
1307 		 * If the application has requested to receive path mtu
1308 		 * information, send up an empty message containing an
1309 		 * IPV6_PATHMTU ancillary data item.
1310 		 */
1311 		if (!udp->udp_ipv6_recvpathmtu)
1312 			break;
1313 
1314 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1315 		    opt_length;
1316 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1317 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
1318 			break;
1319 		}
1320 
1321 		/*
1322 		 * newmp->b_cont is left to NULL on purpose.  This is an
1323 		 * empty message containing only ancillary data.
1324 		 */
1325 		newmp->b_datap->db_type = M_PROTO;
1326 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1327 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1328 		tudi->PRIM_type = T_UNITDATA_IND;
1329 		tudi->SRC_length = sizeof (sin6_t);
1330 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1331 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1332 		tudi->OPT_length = opt_length;
1333 
1334 		sin6 = (sin6_t *)&tudi[1];
1335 		bzero(sin6, sizeof (sin6_t));
1336 		sin6->sin6_family = AF_INET6;
1337 		sin6->sin6_addr = udp->udp_v6dst;
1338 
1339 		toh = (struct T_opthdr *)&sin6[1];
1340 		toh->level = IPPROTO_IPV6;
1341 		toh->name = IPV6_PATHMTU;
1342 		toh->len = opt_length;
1343 		toh->status = 0;
1344 
1345 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1346 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1347 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1348 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1349 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1350 		/*
1351 		 * We've consumed everything we need from the original
1352 		 * message.  Free it, then send our empty message.
1353 		 */
1354 		freemsg(mp);
1355 		udp_ulp_recv(connp, newmp);
1356 
1357 		return;
1358 	}
1359 	case ICMP6_TIME_EXCEEDED:
1360 		/* Transient errors */
1361 		break;
1362 	case ICMP6_PARAM_PROB:
1363 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1364 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1365 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1366 		    (uchar_t *)nexthdrp) {
1367 			error = ECONNREFUSED;
1368 			break;
1369 		}
1370 		break;
1371 	}
1372 	if (error == 0) {
1373 		freemsg(mp);
1374 		return;
1375 	}
1376 
1377 	/*
1378 	 * Deliver T_UDERROR_IND when the application has asked for it.
1379 	 * The socket layer enables this automatically when connected.
1380 	 */
1381 	if (!udp->udp_dgram_errind) {
1382 		freemsg(mp);
1383 		return;
1384 	}
1385 
1386 	sin6 = sin6_null;
1387 	sin6.sin6_family = AF_INET6;
1388 	sin6.sin6_addr = ip6h->ip6_dst;
1389 	sin6.sin6_port = udpha->uha_dst_port;
1390 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1391 
1392 	if (IPCL_IS_NONSTR(connp)) {
1393 		rw_enter(&udp->udp_rwlock, RW_WRITER);
1394 		if (udp->udp_state == TS_DATA_XFER) {
1395 			if (sin6.sin6_port == udp->udp_dstport &&
1396 			    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1397 			    &udp->udp_v6dst)) {
1398 				rw_exit(&udp->udp_rwlock);
1399 				(*connp->conn_upcalls->su_set_error)
1400 				    (connp->conn_upper_handle, error);
1401 				goto done;
1402 			}
1403 		} else {
1404 			udp->udp_delayed_error = error;
1405 			*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1406 		}
1407 		rw_exit(&udp->udp_rwlock);
1408 	} else {
1409 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1410 		    NULL, 0, error);
1411 		if (mp1 != NULL)
1412 			putnext(connp->conn_rq, mp1);
1413 	}
1414 done:
1415 	ASSERT(!RW_ISWRITER(&udp->udp_rwlock));
1416 	freemsg(mp);
1417 }
1418 
1419 /*
1420  * This routine responds to T_ADDR_REQ messages.  It is called by udp_wput.
1421  * The local address is filled in if endpoint is bound. The remote address
1422  * is filled in if remote address has been precified ("connected endpoint")
1423  * (The concept of connected CLTS sockets is alien to published TPI
1424  *  but we support it anyway).
1425  */
1426 static void
1427 udp_addr_req(queue_t *q, mblk_t *mp)
1428 {
1429 	sin_t	*sin;
1430 	sin6_t	*sin6;
1431 	mblk_t	*ackmp;
1432 	struct T_addr_ack *taa;
1433 	udp_t	*udp = Q_TO_UDP(q);
1434 
1435 	/* Make it large enough for worst case */
1436 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1437 	    2 * sizeof (sin6_t), 1);
1438 	if (ackmp == NULL) {
1439 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1440 		return;
1441 	}
1442 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1443 
1444 	bzero(taa, sizeof (struct T_addr_ack));
1445 	ackmp->b_wptr = (uchar_t *)&taa[1];
1446 
1447 	taa->PRIM_type = T_ADDR_ACK;
1448 	ackmp->b_datap->db_type = M_PCPROTO;
1449 	rw_enter(&udp->udp_rwlock, RW_READER);
1450 	/*
1451 	 * Note: Following code assumes 32 bit alignment of basic
1452 	 * data structures like sin_t and struct T_addr_ack.
1453 	 */
1454 	if (udp->udp_state != TS_UNBND) {
1455 		/*
1456 		 * Fill in local address first
1457 		 */
1458 		taa->LOCADDR_offset = sizeof (*taa);
1459 		if (udp->udp_family == AF_INET) {
1460 			taa->LOCADDR_length = sizeof (sin_t);
1461 			sin = (sin_t *)&taa[1];
1462 			/* Fill zeroes and then initialize non-zero fields */
1463 			*sin = sin_null;
1464 			sin->sin_family = AF_INET;
1465 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
1466 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1467 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src,
1468 				    sin->sin_addr.s_addr);
1469 			} else {
1470 				/*
1471 				 * INADDR_ANY
1472 				 * udp_v6src is not set, we might be bound to
1473 				 * broadcast/multicast. Use udp_bound_v6src as
1474 				 * local address instead (that could
1475 				 * also still be INADDR_ANY)
1476 				 */
1477 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src,
1478 				    sin->sin_addr.s_addr);
1479 			}
1480 			sin->sin_port = udp->udp_port;
1481 			ackmp->b_wptr = (uchar_t *)&sin[1];
1482 			if (udp->udp_state == TS_DATA_XFER) {
1483 				/*
1484 				 * connected, fill remote address too
1485 				 */
1486 				taa->REMADDR_length = sizeof (sin_t);
1487 				/* assumed 32-bit alignment */
1488 				taa->REMADDR_offset = taa->LOCADDR_offset +
1489 				    taa->LOCADDR_length;
1490 
1491 				sin = (sin_t *)(ackmp->b_rptr +
1492 				    taa->REMADDR_offset);
1493 				/* initialize */
1494 				*sin = sin_null;
1495 				sin->sin_family = AF_INET;
1496 				sin->sin_addr.s_addr =
1497 				    V4_PART_OF_V6(udp->udp_v6dst);
1498 				sin->sin_port = udp->udp_dstport;
1499 				ackmp->b_wptr = (uchar_t *)&sin[1];
1500 			}
1501 		} else {
1502 			taa->LOCADDR_length = sizeof (sin6_t);
1503 			sin6 = (sin6_t *)&taa[1];
1504 			/* Fill zeroes and then initialize non-zero fields */
1505 			*sin6 = sin6_null;
1506 			sin6->sin6_family = AF_INET6;
1507 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1508 				sin6->sin6_addr = udp->udp_v6src;
1509 			} else {
1510 				/*
1511 				 * UNSPECIFIED
1512 				 * udp_v6src is not set, we might be bound to
1513 				 * broadcast/multicast. Use udp_bound_v6src as
1514 				 * local address instead (that could
1515 				 * also still be UNSPECIFIED)
1516 				 */
1517 				sin6->sin6_addr =
1518 				    udp->udp_bound_v6src;
1519 			}
1520 			sin6->sin6_port = udp->udp_port;
1521 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1522 			if (udp->udp_state == TS_DATA_XFER) {
1523 				/*
1524 				 * connected, fill remote address too
1525 				 */
1526 				taa->REMADDR_length = sizeof (sin6_t);
1527 				/* assumed 32-bit alignment */
1528 				taa->REMADDR_offset = taa->LOCADDR_offset +
1529 				    taa->LOCADDR_length;
1530 
1531 				sin6 = (sin6_t *)(ackmp->b_rptr +
1532 				    taa->REMADDR_offset);
1533 				/* initialize */
1534 				*sin6 = sin6_null;
1535 				sin6->sin6_family = AF_INET6;
1536 				sin6->sin6_addr = udp->udp_v6dst;
1537 				sin6->sin6_port =  udp->udp_dstport;
1538 				ackmp->b_wptr = (uchar_t *)&sin6[1];
1539 			}
1540 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1541 		}
1542 	}
1543 	rw_exit(&udp->udp_rwlock);
1544 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1545 	qreply(q, ackmp);
1546 }
1547 
1548 static void
1549 udp_copy_info(struct T_info_ack *tap, udp_t *udp)
1550 {
1551 	if (udp->udp_family == AF_INET) {
1552 		*tap = udp_g_t_info_ack_ipv4;
1553 	} else {
1554 		*tap = udp_g_t_info_ack_ipv6;
1555 	}
1556 	tap->CURRENT_state = udp->udp_state;
1557 	tap->OPT_size = udp_max_optsize;
1558 }
1559 
1560 static void
1561 udp_do_capability_ack(udp_t *udp, struct T_capability_ack *tcap,
1562     t_uscalar_t cap_bits1)
1563 {
1564 	tcap->CAP_bits1 = 0;
1565 
1566 	if (cap_bits1 & TC1_INFO) {
1567 		udp_copy_info(&tcap->INFO_ack, udp);
1568 		tcap->CAP_bits1 |= TC1_INFO;
1569 	}
1570 }
1571 
1572 /*
1573  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1574  * udp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1575  * udp_g_t_info_ack.  The current state of the stream is copied from
1576  * udp_state.
1577  */
1578 static void
1579 udp_capability_req(queue_t *q, mblk_t *mp)
1580 {
1581 	t_uscalar_t		cap_bits1;
1582 	struct T_capability_ack	*tcap;
1583 	udp_t	*udp = Q_TO_UDP(q);
1584 
1585 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1586 
1587 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1588 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1589 	if (!mp)
1590 		return;
1591 
1592 	tcap = (struct T_capability_ack *)mp->b_rptr;
1593 	udp_do_capability_ack(udp, tcap, cap_bits1);
1594 
1595 	qreply(q, mp);
1596 }
1597 
1598 /*
1599  * This routine responds to T_INFO_REQ messages.  It is called by udp_wput.
1600  * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack.
1601  * The current state of the stream is copied from udp_state.
1602  */
1603 static void
1604 udp_info_req(queue_t *q, mblk_t *mp)
1605 {
1606 	udp_t *udp = Q_TO_UDP(q);
1607 
1608 	/* Create a T_INFO_ACK message. */
1609 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1610 	    T_INFO_ACK);
1611 	if (!mp)
1612 		return;
1613 	udp_copy_info((struct T_info_ack *)mp->b_rptr, udp);
1614 	qreply(q, mp);
1615 }
1616 
1617 /* For /dev/udp aka AF_INET open */
1618 static int
1619 udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1620 {
1621 	return (udp_open(q, devp, flag, sflag, credp, B_FALSE));
1622 }
1623 
1624 /* For /dev/udp6 aka AF_INET6 open */
1625 static int
1626 udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1627 {
1628 	return (udp_open(q, devp, flag, sflag, credp, B_TRUE));
1629 }
1630 
1631 /*
1632  * This is the open routine for udp.  It allocates a udp_t structure for
1633  * the stream and, on the first open of the module, creates an ND table.
1634  */
1635 /*ARGSUSED2*/
1636 static int
1637 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1638     boolean_t isv6)
1639 {
1640 	int		error;
1641 	udp_t		*udp;
1642 	conn_t		*connp;
1643 	dev_t		conn_dev;
1644 	udp_stack_t	*us;
1645 	vmem_t		*minor_arena;
1646 
1647 	TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q);
1648 
1649 	/* If the stream is already open, return immediately. */
1650 	if (q->q_ptr != NULL)
1651 		return (0);
1652 
1653 	if (sflag == MODOPEN)
1654 		return (EINVAL);
1655 
1656 	if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
1657 	    ((conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
1658 		minor_arena = ip_minor_arena_la;
1659 	} else {
1660 		/*
1661 		 * Either minor numbers in the large arena were exhausted
1662 		 * or a non socket application is doing the open.
1663 		 * Try to allocate from the small arena.
1664 		 */
1665 		if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0)
1666 			return (EBUSY);
1667 
1668 		minor_arena = ip_minor_arena_sa;
1669 	}
1670 
1671 	if (flag & SO_FALLBACK) {
1672 		/*
1673 		 * Non streams socket needs a stream to fallback to
1674 		 */
1675 		RD(q)->q_ptr = (void *)conn_dev;
1676 		WR(q)->q_qinfo = &udp_fallback_sock_winit;
1677 		WR(q)->q_ptr = (void *)minor_arena;
1678 		qprocson(q);
1679 		return (0);
1680 	}
1681 
1682 	connp = udp_do_open(credp, isv6, KM_SLEEP);
1683 	if (connp == NULL) {
1684 		inet_minor_free(minor_arena, conn_dev);
1685 		return (ENOMEM);
1686 	}
1687 	udp = connp->conn_udp;
1688 	us = udp->udp_us;
1689 
1690 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1691 	connp->conn_dev = conn_dev;
1692 	connp->conn_minor_arena = minor_arena;
1693 
1694 	/*
1695 	 * Initialize the udp_t structure for this stream.
1696 	 */
1697 	q->q_ptr = connp;
1698 	WR(q)->q_ptr = connp;
1699 	connp->conn_rq = q;
1700 	connp->conn_wq = WR(q);
1701 
1702 	rw_enter(&udp->udp_rwlock, RW_WRITER);
1703 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
1704 	ASSERT(connp->conn_udp == udp);
1705 	ASSERT(udp->udp_connp == connp);
1706 
1707 	if (flag & SO_SOCKSTR) {
1708 		connp->conn_flags |= IPCL_SOCKET;
1709 		udp->udp_issocket = B_TRUE;
1710 	}
1711 
1712 	q->q_hiwat = us->us_recv_hiwat;
1713 	WR(q)->q_hiwat = us->us_xmit_hiwat;
1714 	WR(q)->q_lowat = us->us_xmit_lowat;
1715 
1716 	qprocson(q);
1717 
1718 	if (udp->udp_family == AF_INET6) {
1719 		/* Build initial header template for transmit */
1720 		if ((error = udp_build_hdrs(udp)) != 0) {
1721 			rw_exit(&udp->udp_rwlock);
1722 			qprocsoff(q);
1723 			inet_minor_free(minor_arena, conn_dev);
1724 			ipcl_conn_destroy(connp);
1725 			return (error);
1726 		}
1727 	}
1728 	rw_exit(&udp->udp_rwlock);
1729 
1730 	/* Set the Stream head write offset and high watermark. */
1731 	(void) proto_set_tx_wroff(q, connp,
1732 	    udp->udp_max_hdr_len + us->us_wroff_extra);
1733 	/* XXX udp_set_rcv_hiwat() doesn't hold the lock, is it a bug??? */
1734 	(void) proto_set_rx_hiwat(q, connp, udp_set_rcv_hiwat(udp, q->q_hiwat));
1735 
1736 	mutex_enter(&connp->conn_lock);
1737 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1738 	mutex_exit(&connp->conn_lock);
1739 	return (0);
1740 }
1741 
1742 /*
1743  * Which UDP options OK to set through T_UNITDATA_REQ...
1744  */
1745 /* ARGSUSED */
1746 static boolean_t
1747 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1748 {
1749 	return (B_TRUE);
1750 }
1751 
1752 /*
1753  * This routine gets default values of certain options whose default
1754  * values are maintained by protcol specific code
1755  */
1756 /* ARGSUSED */
1757 int
1758 udp_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
1759 {
1760 	udp_t		*udp = Q_TO_UDP(q);
1761 	udp_stack_t *us = udp->udp_us;
1762 	int *i1 = (int *)ptr;
1763 
1764 	switch (level) {
1765 	case IPPROTO_IP:
1766 		switch (name) {
1767 		case IP_MULTICAST_TTL:
1768 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1769 			return (sizeof (uchar_t));
1770 		case IP_MULTICAST_LOOP:
1771 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1772 			return (sizeof (uchar_t));
1773 		}
1774 		break;
1775 	case IPPROTO_IPV6:
1776 		switch (name) {
1777 		case IPV6_MULTICAST_HOPS:
1778 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1779 			return (sizeof (int));
1780 		case IPV6_MULTICAST_LOOP:
1781 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1782 			return (sizeof (int));
1783 		case IPV6_UNICAST_HOPS:
1784 			*i1 = us->us_ipv6_hoplimit;
1785 			return (sizeof (int));
1786 		}
1787 		break;
1788 	}
1789 	return (-1);
1790 }
1791 
1792 /*
1793  * This routine retrieves the current status of socket options.
1794  * It returns the size of the option retrieved.
1795  */
1796 static int
1797 udp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
1798 {
1799 	udp_t		*udp = connp->conn_udp;
1800 	udp_stack_t	*us = udp->udp_us;
1801 	int		*i1 = (int *)ptr;
1802 	ip6_pkt_t 	*ipp = &udp->udp_sticky_ipp;
1803 	int		len;
1804 
1805 	ASSERT(RW_READ_HELD(&udp->udp_rwlock));
1806 	switch (level) {
1807 	case SOL_SOCKET:
1808 		switch (name) {
1809 		case SO_DEBUG:
1810 			*i1 = udp->udp_debug;
1811 			break;	/* goto sizeof (int) option return */
1812 		case SO_REUSEADDR:
1813 			*i1 = udp->udp_reuseaddr;
1814 			break;	/* goto sizeof (int) option return */
1815 		case SO_TYPE:
1816 			*i1 = SOCK_DGRAM;
1817 			break;	/* goto sizeof (int) option return */
1818 
1819 		/*
1820 		 * The following three items are available here,
1821 		 * but are only meaningful to IP.
1822 		 */
1823 		case SO_DONTROUTE:
1824 			*i1 = udp->udp_dontroute;
1825 			break;	/* goto sizeof (int) option return */
1826 		case SO_USELOOPBACK:
1827 			*i1 = udp->udp_useloopback;
1828 			break;	/* goto sizeof (int) option return */
1829 		case SO_BROADCAST:
1830 			*i1 = udp->udp_broadcast;
1831 			break;	/* goto sizeof (int) option return */
1832 
1833 		case SO_SNDBUF:
1834 			*i1 = udp->udp_xmit_hiwat;
1835 			break;	/* goto sizeof (int) option return */
1836 		case SO_RCVBUF:
1837 			*i1 = udp->udp_rcv_disply_hiwat;
1838 			break;	/* goto sizeof (int) option return */
1839 		case SO_DGRAM_ERRIND:
1840 			*i1 = udp->udp_dgram_errind;
1841 			break;	/* goto sizeof (int) option return */
1842 		case SO_RECVUCRED:
1843 			*i1 = udp->udp_recvucred;
1844 			break;	/* goto sizeof (int) option return */
1845 		case SO_TIMESTAMP:
1846 			*i1 = udp->udp_timestamp;
1847 			break;	/* goto sizeof (int) option return */
1848 		case SO_ANON_MLP:
1849 			*i1 = connp->conn_anon_mlp;
1850 			break;	/* goto sizeof (int) option return */
1851 		case SO_MAC_EXEMPT:
1852 			*i1 = connp->conn_mac_exempt;
1853 			break;	/* goto sizeof (int) option return */
1854 		case SO_ALLZONES:
1855 			*i1 = connp->conn_allzones;
1856 			break;	/* goto sizeof (int) option return */
1857 		case SO_EXCLBIND:
1858 			*i1 = udp->udp_exclbind ? SO_EXCLBIND : 0;
1859 			break;
1860 		case SO_PROTOTYPE:
1861 			*i1 = IPPROTO_UDP;
1862 			break;
1863 		case SO_DOMAIN:
1864 			*i1 = udp->udp_family;
1865 			break;
1866 		default:
1867 			return (-1);
1868 		}
1869 		break;
1870 	case IPPROTO_IP:
1871 		if (udp->udp_family != AF_INET)
1872 			return (-1);
1873 		switch (name) {
1874 		case IP_OPTIONS:
1875 		case T_IP_OPTIONS:
1876 			len = udp->udp_ip_rcv_options_len - udp->udp_label_len;
1877 			if (len > 0) {
1878 				bcopy(udp->udp_ip_rcv_options +
1879 				    udp->udp_label_len, ptr, len);
1880 			}
1881 			return (len);
1882 		case IP_TOS:
1883 		case T_IP_TOS:
1884 			*i1 = (int)udp->udp_type_of_service;
1885 			break;	/* goto sizeof (int) option return */
1886 		case IP_TTL:
1887 			*i1 = (int)udp->udp_ttl;
1888 			break;	/* goto sizeof (int) option return */
1889 		case IP_DHCPINIT_IF:
1890 			return (-EINVAL);
1891 		case IP_NEXTHOP:
1892 		case IP_RECVPKTINFO:
1893 			/*
1894 			 * This also handles IP_PKTINFO.
1895 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
1896 			 * Differentiation is based on the size of the argument
1897 			 * passed in.
1898 			 * This option is handled in IP which will return an
1899 			 * error for IP_PKTINFO as it's not supported as a
1900 			 * sticky option.
1901 			 */
1902 			return (-EINVAL);
1903 		case IP_MULTICAST_IF:
1904 			/* 0 address if not set */
1905 			*(ipaddr_t *)ptr = udp->udp_multicast_if_addr;
1906 			return (sizeof (ipaddr_t));
1907 		case IP_MULTICAST_TTL:
1908 			*(uchar_t *)ptr = udp->udp_multicast_ttl;
1909 			return (sizeof (uchar_t));
1910 		case IP_MULTICAST_LOOP:
1911 			*ptr = connp->conn_multicast_loop;
1912 			return (sizeof (uint8_t));
1913 		case IP_RECVOPTS:
1914 			*i1 = udp->udp_recvopts;
1915 			break;	/* goto sizeof (int) option return */
1916 		case IP_RECVDSTADDR:
1917 			*i1 = udp->udp_recvdstaddr;
1918 			break;	/* goto sizeof (int) option return */
1919 		case IP_RECVIF:
1920 			*i1 = udp->udp_recvif;
1921 			break;	/* goto sizeof (int) option return */
1922 		case IP_RECVSLLA:
1923 			*i1 = udp->udp_recvslla;
1924 			break;	/* goto sizeof (int) option return */
1925 		case IP_RECVTTL:
1926 			*i1 = udp->udp_recvttl;
1927 			break;	/* goto sizeof (int) option return */
1928 		case IP_ADD_MEMBERSHIP:
1929 		case IP_DROP_MEMBERSHIP:
1930 		case IP_BLOCK_SOURCE:
1931 		case IP_UNBLOCK_SOURCE:
1932 		case IP_ADD_SOURCE_MEMBERSHIP:
1933 		case IP_DROP_SOURCE_MEMBERSHIP:
1934 		case MCAST_JOIN_GROUP:
1935 		case MCAST_LEAVE_GROUP:
1936 		case MCAST_BLOCK_SOURCE:
1937 		case MCAST_UNBLOCK_SOURCE:
1938 		case MCAST_JOIN_SOURCE_GROUP:
1939 		case MCAST_LEAVE_SOURCE_GROUP:
1940 			/* cannot "get" the value for these */
1941 			return (-1);
1942 		case IP_BOUND_IF:
1943 			/* Zero if not set */
1944 			*i1 = udp->udp_bound_if;
1945 			break;	/* goto sizeof (int) option return */
1946 		case IP_UNSPEC_SRC:
1947 			*i1 = udp->udp_unspec_source;
1948 			break;	/* goto sizeof (int) option return */
1949 		case IP_BROADCAST_TTL:
1950 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
1951 			return (sizeof (uchar_t));
1952 		default:
1953 			return (-1);
1954 		}
1955 		break;
1956 	case IPPROTO_IPV6:
1957 		if (udp->udp_family != AF_INET6)
1958 			return (-1);
1959 		switch (name) {
1960 		case IPV6_UNICAST_HOPS:
1961 			*i1 = (unsigned int)udp->udp_ttl;
1962 			break;	/* goto sizeof (int) option return */
1963 		case IPV6_MULTICAST_IF:
1964 			/* 0 index if not set */
1965 			*i1 = udp->udp_multicast_if_index;
1966 			break;	/* goto sizeof (int) option return */
1967 		case IPV6_MULTICAST_HOPS:
1968 			*i1 = udp->udp_multicast_ttl;
1969 			break;	/* goto sizeof (int) option return */
1970 		case IPV6_MULTICAST_LOOP:
1971 			*i1 = connp->conn_multicast_loop;
1972 			break;	/* goto sizeof (int) option return */
1973 		case IPV6_JOIN_GROUP:
1974 		case IPV6_LEAVE_GROUP:
1975 		case MCAST_JOIN_GROUP:
1976 		case MCAST_LEAVE_GROUP:
1977 		case MCAST_BLOCK_SOURCE:
1978 		case MCAST_UNBLOCK_SOURCE:
1979 		case MCAST_JOIN_SOURCE_GROUP:
1980 		case MCAST_LEAVE_SOURCE_GROUP:
1981 			/* cannot "get" the value for these */
1982 			return (-1);
1983 		case IPV6_BOUND_IF:
1984 			/* Zero if not set */
1985 			*i1 = udp->udp_bound_if;
1986 			break;	/* goto sizeof (int) option return */
1987 		case IPV6_UNSPEC_SRC:
1988 			*i1 = udp->udp_unspec_source;
1989 			break;	/* goto sizeof (int) option return */
1990 		case IPV6_RECVPKTINFO:
1991 			*i1 = udp->udp_ip_recvpktinfo;
1992 			break;	/* goto sizeof (int) option return */
1993 		case IPV6_RECVTCLASS:
1994 			*i1 = udp->udp_ipv6_recvtclass;
1995 			break;	/* goto sizeof (int) option return */
1996 		case IPV6_RECVPATHMTU:
1997 			*i1 = udp->udp_ipv6_recvpathmtu;
1998 			break;	/* goto sizeof (int) option return */
1999 		case IPV6_RECVHOPLIMIT:
2000 			*i1 = udp->udp_ipv6_recvhoplimit;
2001 			break;	/* goto sizeof (int) option return */
2002 		case IPV6_RECVHOPOPTS:
2003 			*i1 = udp->udp_ipv6_recvhopopts;
2004 			break;	/* goto sizeof (int) option return */
2005 		case IPV6_RECVDSTOPTS:
2006 			*i1 = udp->udp_ipv6_recvdstopts;
2007 			break;	/* goto sizeof (int) option return */
2008 		case _OLD_IPV6_RECVDSTOPTS:
2009 			*i1 = udp->udp_old_ipv6_recvdstopts;
2010 			break;	/* goto sizeof (int) option return */
2011 		case IPV6_RECVRTHDRDSTOPTS:
2012 			*i1 = udp->udp_ipv6_recvrthdrdstopts;
2013 			break;	/* goto sizeof (int) option return */
2014 		case IPV6_RECVRTHDR:
2015 			*i1 = udp->udp_ipv6_recvrthdr;
2016 			break;	/* goto sizeof (int) option return */
2017 		case IPV6_PKTINFO: {
2018 			/* XXX assumes that caller has room for max size! */
2019 			struct in6_pktinfo *pkti;
2020 
2021 			pkti = (struct in6_pktinfo *)ptr;
2022 			if (ipp->ipp_fields & IPPF_IFINDEX)
2023 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
2024 			else
2025 				pkti->ipi6_ifindex = 0;
2026 			if (ipp->ipp_fields & IPPF_ADDR)
2027 				pkti->ipi6_addr = ipp->ipp_addr;
2028 			else
2029 				pkti->ipi6_addr = ipv6_all_zeros;
2030 			return (sizeof (struct in6_pktinfo));
2031 		}
2032 		case IPV6_TCLASS:
2033 			if (ipp->ipp_fields & IPPF_TCLASS)
2034 				*i1 = ipp->ipp_tclass;
2035 			else
2036 				*i1 = IPV6_FLOW_TCLASS(
2037 				    IPV6_DEFAULT_VERS_AND_FLOW);
2038 			break;	/* goto sizeof (int) option return */
2039 		case IPV6_NEXTHOP: {
2040 			sin6_t *sin6 = (sin6_t *)ptr;
2041 
2042 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2043 				return (0);
2044 			*sin6 = sin6_null;
2045 			sin6->sin6_family = AF_INET6;
2046 			sin6->sin6_addr = ipp->ipp_nexthop;
2047 			return (sizeof (sin6_t));
2048 		}
2049 		case IPV6_HOPOPTS:
2050 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2051 				return (0);
2052 			if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6)
2053 				return (0);
2054 			/*
2055 			 * The cipso/label option is added by kernel.
2056 			 * User is not usually aware of this option.
2057 			 * We copy out the hbh opt after the label option.
2058 			 */
2059 			bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6,
2060 			    ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2061 			if (udp->udp_label_len_v6 > 0) {
2062 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2063 				ptr[1] = (ipp->ipp_hopoptslen -
2064 				    udp->udp_label_len_v6 + 7) / 8 - 1;
2065 			}
2066 			return (ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2067 		case IPV6_RTHDRDSTOPTS:
2068 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2069 				return (0);
2070 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2071 			return (ipp->ipp_rtdstoptslen);
2072 		case IPV6_RTHDR:
2073 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2074 				return (0);
2075 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2076 			return (ipp->ipp_rthdrlen);
2077 		case IPV6_DSTOPTS:
2078 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
2079 				return (0);
2080 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2081 			return (ipp->ipp_dstoptslen);
2082 		case IPV6_PATHMTU:
2083 			return (ip_fill_mtuinfo(&udp->udp_v6dst,
2084 			    udp->udp_dstport, (struct ip6_mtuinfo *)ptr,
2085 			    us->us_netstack));
2086 		default:
2087 			return (-1);
2088 		}
2089 		break;
2090 	case IPPROTO_UDP:
2091 		switch (name) {
2092 		case UDP_ANONPRIVBIND:
2093 			*i1 = udp->udp_anon_priv_bind;
2094 			break;
2095 		case UDP_EXCLBIND:
2096 			*i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0;
2097 			break;
2098 		case UDP_RCVHDR:
2099 			*i1 = udp->udp_rcvhdr ? 1 : 0;
2100 			break;
2101 		case UDP_NAT_T_ENDPOINT:
2102 			*i1 = udp->udp_nat_t_endpoint;
2103 			break;
2104 		default:
2105 			return (-1);
2106 		}
2107 		break;
2108 	default:
2109 		return (-1);
2110 	}
2111 	return (sizeof (int));
2112 }
2113 
2114 int
2115 udp_tpi_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2116 {
2117 	udp_t   *udp;
2118 	int	err;
2119 
2120 	udp = Q_TO_UDP(q);
2121 
2122 	rw_enter(&udp->udp_rwlock, RW_READER);
2123 	err = udp_opt_get(Q_TO_CONN(q), level, name, ptr);
2124 	rw_exit(&udp->udp_rwlock);
2125 	return (err);
2126 }
2127 
2128 /*
2129  * This routine sets socket options.
2130  */
2131 /* ARGSUSED */
2132 static int
2133 udp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
2134     uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
2135     void *thisdg_attrs, boolean_t checkonly)
2136 {
2137 	udpattrs_t *attrs = thisdg_attrs;
2138 	int	*i1 = (int *)invalp;
2139 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2140 	udp_t	*udp = connp->conn_udp;
2141 	udp_stack_t	*us = udp->udp_us;
2142 	int	error;
2143 	uint_t	newlen;
2144 	size_t	sth_wroff;
2145 
2146 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
2147 	/*
2148 	 * For fixed length options, no sanity check
2149 	 * of passed in length is done. It is assumed *_optcom_req()
2150 	 * routines do the right thing.
2151 	 */
2152 	switch (level) {
2153 	case SOL_SOCKET:
2154 		switch (name) {
2155 		case SO_REUSEADDR:
2156 			if (!checkonly) {
2157 				udp->udp_reuseaddr = onoff;
2158 				PASS_OPT_TO_IP(connp);
2159 			}
2160 			break;
2161 		case SO_DEBUG:
2162 			if (!checkonly)
2163 				udp->udp_debug = onoff;
2164 			break;
2165 		/*
2166 		 * The following three items are available here,
2167 		 * but are only meaningful to IP.
2168 		 */
2169 		case SO_DONTROUTE:
2170 			if (!checkonly) {
2171 				udp->udp_dontroute = onoff;
2172 				PASS_OPT_TO_IP(connp);
2173 			}
2174 			break;
2175 		case SO_USELOOPBACK:
2176 			if (!checkonly) {
2177 				udp->udp_useloopback = onoff;
2178 				PASS_OPT_TO_IP(connp);
2179 			}
2180 			break;
2181 		case SO_BROADCAST:
2182 			if (!checkonly) {
2183 				udp->udp_broadcast = onoff;
2184 				PASS_OPT_TO_IP(connp);
2185 			}
2186 			break;
2187 
2188 		case SO_SNDBUF:
2189 			if (*i1 > us->us_max_buf) {
2190 				*outlenp = 0;
2191 				return (ENOBUFS);
2192 			}
2193 			if (!checkonly) {
2194 				udp->udp_xmit_hiwat = *i1;
2195 				connp->conn_wq->q_hiwat = *i1;
2196 			}
2197 			break;
2198 		case SO_RCVBUF:
2199 			if (*i1 > us->us_max_buf) {
2200 				*outlenp = 0;
2201 				return (ENOBUFS);
2202 			}
2203 			if (!checkonly) {
2204 				int size;
2205 
2206 				udp->udp_rcv_disply_hiwat = *i1;
2207 				size = udp_set_rcv_hiwat(udp, *i1);
2208 				rw_exit(&udp->udp_rwlock);
2209 				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
2210 				    size);
2211 				rw_enter(&udp->udp_rwlock, RW_WRITER);
2212 			}
2213 			break;
2214 		case SO_DGRAM_ERRIND:
2215 			if (!checkonly)
2216 				udp->udp_dgram_errind = onoff;
2217 			break;
2218 		case SO_RECVUCRED:
2219 			if (!checkonly)
2220 				udp->udp_recvucred = onoff;
2221 			break;
2222 		case SO_ALLZONES:
2223 			/*
2224 			 * "soft" error (negative)
2225 			 * option not handled at this level
2226 			 * Do not modify *outlenp.
2227 			 */
2228 			return (-EINVAL);
2229 		case SO_TIMESTAMP:
2230 			if (!checkonly)
2231 				udp->udp_timestamp = onoff;
2232 			break;
2233 		case SO_ANON_MLP:
2234 			if (!checkonly) {
2235 				connp->conn_anon_mlp = onoff;
2236 				PASS_OPT_TO_IP(connp);
2237 			}
2238 			break;
2239 		case SO_MAC_EXEMPT:
2240 			if (secpolicy_net_mac_aware(cr) != 0 ||
2241 			    udp->udp_state != TS_UNBND)
2242 				return (EACCES);
2243 			if (!checkonly) {
2244 				connp->conn_mac_exempt = onoff;
2245 				PASS_OPT_TO_IP(connp);
2246 			}
2247 			break;
2248 		case SCM_UCRED: {
2249 			struct ucred_s *ucr;
2250 			cred_t *cr, *newcr;
2251 			ts_label_t *tsl;
2252 
2253 			/*
2254 			 * Only sockets that have proper privileges and are
2255 			 * bound to MLPs will have any other value here, so
2256 			 * this implicitly tests for privilege to set label.
2257 			 */
2258 			if (connp->conn_mlp_type == mlptSingle)
2259 				break;
2260 			ucr = (struct ucred_s *)invalp;
2261 			if (inlen != ucredsize ||
2262 			    ucr->uc_labeloff < sizeof (*ucr) ||
2263 			    ucr->uc_labeloff + sizeof (bslabel_t) > inlen)
2264 				return (EINVAL);
2265 			if (!checkonly) {
2266 				mblk_t *mb;
2267 				pid_t  cpid;
2268 
2269 				if (attrs == NULL ||
2270 				    (mb = attrs->udpattr_mb) == NULL)
2271 					return (EINVAL);
2272 				if ((cr = msg_getcred(mb, &cpid)) == NULL)
2273 					cr = udp->udp_connp->conn_cred;
2274 				ASSERT(cr != NULL);
2275 				if ((tsl = crgetlabel(cr)) == NULL)
2276 					return (EINVAL);
2277 				newcr = copycred_from_bslabel(cr, UCLABEL(ucr),
2278 				    tsl->tsl_doi, KM_NOSLEEP);
2279 				if (newcr == NULL)
2280 					return (ENOSR);
2281 				mblk_setcred(mb, newcr, cpid);
2282 				attrs->udpattr_credset = B_TRUE;
2283 				crfree(newcr);
2284 			}
2285 			break;
2286 		}
2287 		case SO_EXCLBIND:
2288 			if (!checkonly)
2289 				udp->udp_exclbind = onoff;
2290 			break;
2291 		case SO_RCVTIMEO:
2292 		case SO_SNDTIMEO:
2293 			/*
2294 			 * Pass these two options in order for third part
2295 			 * protocol usage. Here just return directly.
2296 			 */
2297 			return (0);
2298 		default:
2299 			*outlenp = 0;
2300 			return (EINVAL);
2301 		}
2302 		break;
2303 	case IPPROTO_IP:
2304 		if (udp->udp_family != AF_INET) {
2305 			*outlenp = 0;
2306 			return (ENOPROTOOPT);
2307 		}
2308 		switch (name) {
2309 		case IP_OPTIONS:
2310 		case T_IP_OPTIONS:
2311 			/* Save options for use by IP. */
2312 			newlen = inlen + udp->udp_label_len;
2313 			if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
2314 				*outlenp = 0;
2315 				return (EINVAL);
2316 			}
2317 			if (checkonly)
2318 				break;
2319 
2320 			/*
2321 			 * Update the stored options taking into account
2322 			 * any CIPSO option which we should not overwrite.
2323 			 */
2324 			if (!tsol_option_set(&udp->udp_ip_snd_options,
2325 			    &udp->udp_ip_snd_options_len,
2326 			    udp->udp_label_len, invalp, inlen)) {
2327 				*outlenp = 0;
2328 				return (ENOMEM);
2329 			}
2330 
2331 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2332 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
2333 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
2334 			rw_exit(&udp->udp_rwlock);
2335 			(void) proto_set_tx_wroff(connp->conn_rq, connp,
2336 			    sth_wroff);
2337 			rw_enter(&udp->udp_rwlock, RW_WRITER);
2338 			break;
2339 
2340 		case IP_TTL:
2341 			if (!checkonly) {
2342 				udp->udp_ttl = (uchar_t)*i1;
2343 			}
2344 			break;
2345 		case IP_TOS:
2346 		case T_IP_TOS:
2347 			if (!checkonly) {
2348 				udp->udp_type_of_service = (uchar_t)*i1;
2349 			}
2350 			break;
2351 		case IP_MULTICAST_IF: {
2352 			/*
2353 			 * TODO should check OPTMGMT reply and undo this if
2354 			 * there is an error.
2355 			 */
2356 			struct in_addr *inap = (struct in_addr *)invalp;
2357 			if (!checkonly) {
2358 				udp->udp_multicast_if_addr =
2359 				    inap->s_addr;
2360 				PASS_OPT_TO_IP(connp);
2361 			}
2362 			break;
2363 		}
2364 		case IP_MULTICAST_TTL:
2365 			if (!checkonly)
2366 				udp->udp_multicast_ttl = *invalp;
2367 			break;
2368 		case IP_MULTICAST_LOOP:
2369 			if (!checkonly) {
2370 				connp->conn_multicast_loop = *invalp;
2371 				PASS_OPT_TO_IP(connp);
2372 			}
2373 			break;
2374 		case IP_RECVOPTS:
2375 			if (!checkonly)
2376 				udp->udp_recvopts = onoff;
2377 			break;
2378 		case IP_RECVDSTADDR:
2379 			if (!checkonly)
2380 				udp->udp_recvdstaddr = onoff;
2381 			break;
2382 		case IP_RECVIF:
2383 			if (!checkonly) {
2384 				udp->udp_recvif = onoff;
2385 				PASS_OPT_TO_IP(connp);
2386 			}
2387 			break;
2388 		case IP_RECVSLLA:
2389 			if (!checkonly) {
2390 				udp->udp_recvslla = onoff;
2391 				PASS_OPT_TO_IP(connp);
2392 			}
2393 			break;
2394 		case IP_RECVTTL:
2395 			if (!checkonly)
2396 				udp->udp_recvttl = onoff;
2397 			break;
2398 		case IP_PKTINFO: {
2399 			/*
2400 			 * This also handles IP_RECVPKTINFO.
2401 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
2402 			 * Differentiation is based on the size of the
2403 			 * argument passed in.
2404 			 */
2405 			struct in_pktinfo *pktinfop;
2406 			ip4_pkt_t *attr_pktinfop;
2407 
2408 			if (checkonly)
2409 				break;
2410 
2411 			if (inlen == sizeof (int)) {
2412 				/*
2413 				 * This is IP_RECVPKTINFO option.
2414 				 * Keep a local copy of whether this option is
2415 				 * set or not and pass it down to IP for
2416 				 * processing.
2417 				 */
2418 
2419 				udp->udp_ip_recvpktinfo = onoff;
2420 				return (-EINVAL);
2421 			}
2422 
2423 			if (attrs == NULL ||
2424 			    (attr_pktinfop = attrs->udpattr_ipp4) == NULL) {
2425 				/*
2426 				 * sticky option or no buffer to return
2427 				 * the results.
2428 				 */
2429 				return (EINVAL);
2430 			}
2431 
2432 			if (inlen != sizeof (struct in_pktinfo))
2433 				return (EINVAL);
2434 
2435 			pktinfop = (struct in_pktinfo *)invalp;
2436 
2437 			/*
2438 			 * At least one of the values should be specified
2439 			 */
2440 			if (pktinfop->ipi_ifindex == 0 &&
2441 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2442 				return (EINVAL);
2443 			}
2444 
2445 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2446 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2447 
2448 			break;
2449 		}
2450 		case IP_ADD_MEMBERSHIP:
2451 		case IP_DROP_MEMBERSHIP:
2452 		case IP_BLOCK_SOURCE:
2453 		case IP_UNBLOCK_SOURCE:
2454 		case IP_ADD_SOURCE_MEMBERSHIP:
2455 		case IP_DROP_SOURCE_MEMBERSHIP:
2456 		case MCAST_JOIN_GROUP:
2457 		case MCAST_LEAVE_GROUP:
2458 		case MCAST_BLOCK_SOURCE:
2459 		case MCAST_UNBLOCK_SOURCE:
2460 		case MCAST_JOIN_SOURCE_GROUP:
2461 		case MCAST_LEAVE_SOURCE_GROUP:
2462 		case IP_SEC_OPT:
2463 		case IP_NEXTHOP:
2464 		case IP_DHCPINIT_IF:
2465 			/*
2466 			 * "soft" error (negative)
2467 			 * option not handled at this level
2468 			 * Do not modify *outlenp.
2469 			 */
2470 			return (-EINVAL);
2471 		case IP_BOUND_IF:
2472 			if (!checkonly) {
2473 				udp->udp_bound_if = *i1;
2474 				PASS_OPT_TO_IP(connp);
2475 			}
2476 			break;
2477 		case IP_UNSPEC_SRC:
2478 			if (!checkonly) {
2479 				udp->udp_unspec_source = onoff;
2480 				PASS_OPT_TO_IP(connp);
2481 			}
2482 			break;
2483 		case IP_BROADCAST_TTL:
2484 			if (!checkonly)
2485 				connp->conn_broadcast_ttl = *invalp;
2486 			break;
2487 		default:
2488 			*outlenp = 0;
2489 			return (EINVAL);
2490 		}
2491 		break;
2492 	case IPPROTO_IPV6: {
2493 		ip6_pkt_t		*ipp;
2494 		boolean_t		sticky;
2495 
2496 		if (udp->udp_family != AF_INET6) {
2497 			*outlenp = 0;
2498 			return (ENOPROTOOPT);
2499 		}
2500 		/*
2501 		 * Deal with both sticky options and ancillary data
2502 		 */
2503 		sticky = B_FALSE;
2504 		if (attrs == NULL || (ipp = attrs->udpattr_ipp6) ==
2505 		    NULL) {
2506 			/* sticky options, or none */
2507 			ipp = &udp->udp_sticky_ipp;
2508 			sticky = B_TRUE;
2509 		}
2510 
2511 		switch (name) {
2512 		case IPV6_MULTICAST_IF:
2513 			if (!checkonly) {
2514 				udp->udp_multicast_if_index = *i1;
2515 				PASS_OPT_TO_IP(connp);
2516 			}
2517 			break;
2518 		case IPV6_UNICAST_HOPS:
2519 			/* -1 means use default */
2520 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2521 				*outlenp = 0;
2522 				return (EINVAL);
2523 			}
2524 			if (!checkonly) {
2525 				if (*i1 == -1) {
2526 					udp->udp_ttl = ipp->ipp_unicast_hops =
2527 					    us->us_ipv6_hoplimit;
2528 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2529 					/* Pass modified value to IP. */
2530 					*i1 = udp->udp_ttl;
2531 				} else {
2532 					udp->udp_ttl = ipp->ipp_unicast_hops =
2533 					    (uint8_t)*i1;
2534 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2535 				}
2536 				/* Rebuild the header template */
2537 				error = udp_build_hdrs(udp);
2538 				if (error != 0) {
2539 					*outlenp = 0;
2540 					return (error);
2541 				}
2542 			}
2543 			break;
2544 		case IPV6_MULTICAST_HOPS:
2545 			/* -1 means use default */
2546 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2547 				*outlenp = 0;
2548 				return (EINVAL);
2549 			}
2550 			if (!checkonly) {
2551 				if (*i1 == -1) {
2552 					udp->udp_multicast_ttl =
2553 					    ipp->ipp_multicast_hops =
2554 					    IP_DEFAULT_MULTICAST_TTL;
2555 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2556 					/* Pass modified value to IP. */
2557 					*i1 = udp->udp_multicast_ttl;
2558 				} else {
2559 					udp->udp_multicast_ttl =
2560 					    ipp->ipp_multicast_hops =
2561 					    (uint8_t)*i1;
2562 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2563 				}
2564 			}
2565 			break;
2566 		case IPV6_MULTICAST_LOOP:
2567 			if (*i1 != 0 && *i1 != 1) {
2568 				*outlenp = 0;
2569 				return (EINVAL);
2570 			}
2571 			if (!checkonly) {
2572 				connp->conn_multicast_loop = *i1;
2573 				PASS_OPT_TO_IP(connp);
2574 			}
2575 			break;
2576 		case IPV6_JOIN_GROUP:
2577 		case IPV6_LEAVE_GROUP:
2578 		case MCAST_JOIN_GROUP:
2579 		case MCAST_LEAVE_GROUP:
2580 		case MCAST_BLOCK_SOURCE:
2581 		case MCAST_UNBLOCK_SOURCE:
2582 		case MCAST_JOIN_SOURCE_GROUP:
2583 		case MCAST_LEAVE_SOURCE_GROUP:
2584 			/*
2585 			 * "soft" error (negative)
2586 			 * option not handled at this level
2587 			 * Note: Do not modify *outlenp
2588 			 */
2589 			return (-EINVAL);
2590 		case IPV6_BOUND_IF:
2591 			if (!checkonly) {
2592 				udp->udp_bound_if = *i1;
2593 				PASS_OPT_TO_IP(connp);
2594 			}
2595 			break;
2596 		case IPV6_UNSPEC_SRC:
2597 			if (!checkonly) {
2598 				udp->udp_unspec_source = onoff;
2599 				PASS_OPT_TO_IP(connp);
2600 			}
2601 			break;
2602 		/*
2603 		 * Set boolean switches for ancillary data delivery
2604 		 */
2605 		case IPV6_RECVPKTINFO:
2606 			if (!checkonly) {
2607 				udp->udp_ip_recvpktinfo = onoff;
2608 				PASS_OPT_TO_IP(connp);
2609 			}
2610 			break;
2611 		case IPV6_RECVTCLASS:
2612 			if (!checkonly) {
2613 				udp->udp_ipv6_recvtclass = onoff;
2614 				PASS_OPT_TO_IP(connp);
2615 			}
2616 			break;
2617 		case IPV6_RECVPATHMTU:
2618 			if (!checkonly) {
2619 				udp->udp_ipv6_recvpathmtu = onoff;
2620 				PASS_OPT_TO_IP(connp);
2621 			}
2622 			break;
2623 		case IPV6_RECVHOPLIMIT:
2624 			if (!checkonly) {
2625 				udp->udp_ipv6_recvhoplimit = onoff;
2626 				PASS_OPT_TO_IP(connp);
2627 			}
2628 			break;
2629 		case IPV6_RECVHOPOPTS:
2630 			if (!checkonly) {
2631 				udp->udp_ipv6_recvhopopts = onoff;
2632 				PASS_OPT_TO_IP(connp);
2633 			}
2634 			break;
2635 		case IPV6_RECVDSTOPTS:
2636 			if (!checkonly) {
2637 				udp->udp_ipv6_recvdstopts = onoff;
2638 				PASS_OPT_TO_IP(connp);
2639 			}
2640 			break;
2641 		case _OLD_IPV6_RECVDSTOPTS:
2642 			if (!checkonly)
2643 				udp->udp_old_ipv6_recvdstopts = onoff;
2644 			break;
2645 		case IPV6_RECVRTHDRDSTOPTS:
2646 			if (!checkonly) {
2647 				udp->udp_ipv6_recvrthdrdstopts = onoff;
2648 				PASS_OPT_TO_IP(connp);
2649 			}
2650 			break;
2651 		case IPV6_RECVRTHDR:
2652 			if (!checkonly) {
2653 				udp->udp_ipv6_recvrthdr = onoff;
2654 				PASS_OPT_TO_IP(connp);
2655 			}
2656 			break;
2657 		/*
2658 		 * Set sticky options or ancillary data.
2659 		 * If sticky options, (re)build any extension headers
2660 		 * that might be needed as a result.
2661 		 */
2662 		case IPV6_PKTINFO:
2663 			/*
2664 			 * The source address and ifindex are verified
2665 			 * in ip_opt_set(). For ancillary data the
2666 			 * source address is checked in ip_wput_v6.
2667 			 */
2668 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2669 				return (EINVAL);
2670 			if (checkonly)
2671 				break;
2672 
2673 			if (inlen == 0) {
2674 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2675 				ipp->ipp_sticky_ignored |=
2676 				    (IPPF_IFINDEX|IPPF_ADDR);
2677 			} else {
2678 				struct in6_pktinfo *pkti;
2679 
2680 				pkti = (struct in6_pktinfo *)invalp;
2681 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2682 				ipp->ipp_addr = pkti->ipi6_addr;
2683 				if (ipp->ipp_ifindex != 0)
2684 					ipp->ipp_fields |= IPPF_IFINDEX;
2685 				else
2686 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2687 				if (!IN6_IS_ADDR_UNSPECIFIED(
2688 				    &ipp->ipp_addr))
2689 					ipp->ipp_fields |= IPPF_ADDR;
2690 				else
2691 					ipp->ipp_fields &= ~IPPF_ADDR;
2692 			}
2693 			if (sticky) {
2694 				error = udp_build_hdrs(udp);
2695 				if (error != 0)
2696 					return (error);
2697 				PASS_OPT_TO_IP(connp);
2698 			}
2699 			break;
2700 		case IPV6_HOPLIMIT:
2701 			if (sticky)
2702 				return (EINVAL);
2703 			if (inlen != 0 && inlen != sizeof (int))
2704 				return (EINVAL);
2705 			if (checkonly)
2706 				break;
2707 
2708 			if (inlen == 0) {
2709 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2710 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2711 			} else {
2712 				if (*i1 > 255 || *i1 < -1)
2713 					return (EINVAL);
2714 				if (*i1 == -1)
2715 					ipp->ipp_hoplimit =
2716 					    us->us_ipv6_hoplimit;
2717 				else
2718 					ipp->ipp_hoplimit = *i1;
2719 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2720 			}
2721 			break;
2722 		case IPV6_TCLASS:
2723 			if (inlen != 0 && inlen != sizeof (int))
2724 				return (EINVAL);
2725 			if (checkonly)
2726 				break;
2727 
2728 			if (inlen == 0) {
2729 				ipp->ipp_fields &= ~IPPF_TCLASS;
2730 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2731 			} else {
2732 				if (*i1 > 255 || *i1 < -1)
2733 					return (EINVAL);
2734 				if (*i1 == -1)
2735 					ipp->ipp_tclass = 0;
2736 				else
2737 					ipp->ipp_tclass = *i1;
2738 				ipp->ipp_fields |= IPPF_TCLASS;
2739 			}
2740 			if (sticky) {
2741 				error = udp_build_hdrs(udp);
2742 				if (error != 0)
2743 					return (error);
2744 			}
2745 			break;
2746 		case IPV6_NEXTHOP:
2747 			/*
2748 			 * IP will verify that the nexthop is reachable
2749 			 * and fail for sticky options.
2750 			 */
2751 			if (inlen != 0 && inlen != sizeof (sin6_t))
2752 				return (EINVAL);
2753 			if (checkonly)
2754 				break;
2755 
2756 			if (inlen == 0) {
2757 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2758 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2759 			} else {
2760 				sin6_t *sin6 = (sin6_t *)invalp;
2761 
2762 				if (sin6->sin6_family != AF_INET6) {
2763 					return (EAFNOSUPPORT);
2764 				}
2765 				if (IN6_IS_ADDR_V4MAPPED(
2766 				    &sin6->sin6_addr))
2767 					return (EADDRNOTAVAIL);
2768 				ipp->ipp_nexthop = sin6->sin6_addr;
2769 				if (!IN6_IS_ADDR_UNSPECIFIED(
2770 				    &ipp->ipp_nexthop))
2771 					ipp->ipp_fields |= IPPF_NEXTHOP;
2772 				else
2773 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2774 			}
2775 			if (sticky) {
2776 				error = udp_build_hdrs(udp);
2777 				if (error != 0)
2778 					return (error);
2779 				PASS_OPT_TO_IP(connp);
2780 			}
2781 			break;
2782 		case IPV6_HOPOPTS: {
2783 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2784 			/*
2785 			 * Sanity checks - minimum size, size a multiple of
2786 			 * eight bytes, and matching size passed in.
2787 			 */
2788 			if (inlen != 0 &&
2789 			    inlen != (8 * (hopts->ip6h_len + 1)))
2790 				return (EINVAL);
2791 
2792 			if (checkonly)
2793 				break;
2794 
2795 			error = optcom_pkt_set(invalp, inlen, sticky,
2796 			    (uchar_t **)&ipp->ipp_hopopts,
2797 			    &ipp->ipp_hopoptslen,
2798 			    sticky ? udp->udp_label_len_v6 : 0);
2799 			if (error != 0)
2800 				return (error);
2801 			if (ipp->ipp_hopoptslen == 0) {
2802 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2803 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2804 			} else {
2805 				ipp->ipp_fields |= IPPF_HOPOPTS;
2806 			}
2807 			if (sticky) {
2808 				error = udp_build_hdrs(udp);
2809 				if (error != 0)
2810 					return (error);
2811 			}
2812 			break;
2813 		}
2814 		case IPV6_RTHDRDSTOPTS: {
2815 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2816 
2817 			/*
2818 			 * Sanity checks - minimum size, size a multiple of
2819 			 * eight bytes, and matching size passed in.
2820 			 */
2821 			if (inlen != 0 &&
2822 			    inlen != (8 * (dopts->ip6d_len + 1)))
2823 				return (EINVAL);
2824 
2825 			if (checkonly)
2826 				break;
2827 
2828 			if (inlen == 0) {
2829 				if (sticky &&
2830 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2831 					kmem_free(ipp->ipp_rtdstopts,
2832 					    ipp->ipp_rtdstoptslen);
2833 					ipp->ipp_rtdstopts = NULL;
2834 					ipp->ipp_rtdstoptslen = 0;
2835 				}
2836 
2837 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2838 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2839 			} else {
2840 				error = optcom_pkt_set(invalp, inlen, sticky,
2841 				    (uchar_t **)&ipp->ipp_rtdstopts,
2842 				    &ipp->ipp_rtdstoptslen, 0);
2843 				if (error != 0)
2844 					return (error);
2845 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2846 			}
2847 			if (sticky) {
2848 				error = udp_build_hdrs(udp);
2849 				if (error != 0)
2850 					return (error);
2851 			}
2852 			break;
2853 		}
2854 		case IPV6_DSTOPTS: {
2855 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2856 
2857 			/*
2858 			 * Sanity checks - minimum size, size a multiple of
2859 			 * eight bytes, and matching size passed in.
2860 			 */
2861 			if (inlen != 0 &&
2862 			    inlen != (8 * (dopts->ip6d_len + 1)))
2863 				return (EINVAL);
2864 
2865 			if (checkonly)
2866 				break;
2867 
2868 			if (inlen == 0) {
2869 				if (sticky &&
2870 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2871 					kmem_free(ipp->ipp_dstopts,
2872 					    ipp->ipp_dstoptslen);
2873 					ipp->ipp_dstopts = NULL;
2874 					ipp->ipp_dstoptslen = 0;
2875 				}
2876 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2877 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2878 			} else {
2879 				error = optcom_pkt_set(invalp, inlen, sticky,
2880 				    (uchar_t **)&ipp->ipp_dstopts,
2881 				    &ipp->ipp_dstoptslen, 0);
2882 				if (error != 0)
2883 					return (error);
2884 				ipp->ipp_fields |= IPPF_DSTOPTS;
2885 			}
2886 			if (sticky) {
2887 				error = udp_build_hdrs(udp);
2888 				if (error != 0)
2889 					return (error);
2890 			}
2891 			break;
2892 		}
2893 		case IPV6_RTHDR: {
2894 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2895 
2896 			/*
2897 			 * Sanity checks - minimum size, size a multiple of
2898 			 * eight bytes, and matching size passed in.
2899 			 */
2900 			if (inlen != 0 &&
2901 			    inlen != (8 * (rt->ip6r_len + 1)))
2902 				return (EINVAL);
2903 
2904 			if (checkonly)
2905 				break;
2906 
2907 			if (inlen == 0) {
2908 				if (sticky &&
2909 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2910 					kmem_free(ipp->ipp_rthdr,
2911 					    ipp->ipp_rthdrlen);
2912 					ipp->ipp_rthdr = NULL;
2913 					ipp->ipp_rthdrlen = 0;
2914 				}
2915 				ipp->ipp_fields &= ~IPPF_RTHDR;
2916 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2917 			} else {
2918 				error = optcom_pkt_set(invalp, inlen, sticky,
2919 				    (uchar_t **)&ipp->ipp_rthdr,
2920 				    &ipp->ipp_rthdrlen, 0);
2921 				if (error != 0)
2922 					return (error);
2923 				ipp->ipp_fields |= IPPF_RTHDR;
2924 			}
2925 			if (sticky) {
2926 				error = udp_build_hdrs(udp);
2927 				if (error != 0)
2928 					return (error);
2929 			}
2930 			break;
2931 		}
2932 
2933 		case IPV6_DONTFRAG:
2934 			if (checkonly)
2935 				break;
2936 
2937 			if (onoff) {
2938 				ipp->ipp_fields |= IPPF_DONTFRAG;
2939 			} else {
2940 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2941 			}
2942 			break;
2943 
2944 		case IPV6_USE_MIN_MTU:
2945 			if (inlen != sizeof (int))
2946 				return (EINVAL);
2947 
2948 			if (*i1 < -1 || *i1 > 1)
2949 				return (EINVAL);
2950 
2951 			if (checkonly)
2952 				break;
2953 
2954 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
2955 			ipp->ipp_use_min_mtu = *i1;
2956 			break;
2957 
2958 		case IPV6_SEC_OPT:
2959 		case IPV6_SRC_PREFERENCES:
2960 		case IPV6_V6ONLY:
2961 			/* Handled at the IP level */
2962 			return (-EINVAL);
2963 		default:
2964 			*outlenp = 0;
2965 			return (EINVAL);
2966 		}
2967 		break;
2968 		}		/* end IPPROTO_IPV6 */
2969 	case IPPROTO_UDP:
2970 		switch (name) {
2971 		case UDP_ANONPRIVBIND:
2972 			if ((error = secpolicy_net_privaddr(cr, 0,
2973 			    IPPROTO_UDP)) != 0) {
2974 				*outlenp = 0;
2975 				return (error);
2976 			}
2977 			if (!checkonly) {
2978 				udp->udp_anon_priv_bind = onoff;
2979 			}
2980 			break;
2981 		case UDP_EXCLBIND:
2982 			if (!checkonly)
2983 				udp->udp_exclbind = onoff;
2984 			break;
2985 		case UDP_RCVHDR:
2986 			if (!checkonly)
2987 				udp->udp_rcvhdr = onoff;
2988 			break;
2989 		case UDP_NAT_T_ENDPOINT:
2990 			if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
2991 				*outlenp = 0;
2992 				return (error);
2993 			}
2994 
2995 			/*
2996 			 * Use udp_family instead so we can avoid ambiguitites
2997 			 * with AF_INET6 sockets that may switch from IPv4
2998 			 * to IPv6.
2999 			 */
3000 			if (udp->udp_family != AF_INET) {
3001 				*outlenp = 0;
3002 				return (EAFNOSUPPORT);
3003 			}
3004 
3005 			if (!checkonly) {
3006 				int size;
3007 
3008 				udp->udp_nat_t_endpoint = onoff;
3009 
3010 				udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3011 				    UDPH_SIZE + udp->udp_ip_snd_options_len;
3012 
3013 				/* Also, adjust wroff */
3014 				if (onoff) {
3015 					udp->udp_max_hdr_len +=
3016 					    sizeof (uint32_t);
3017 				}
3018 				size = udp->udp_max_hdr_len +
3019 				    us->us_wroff_extra;
3020 				(void) proto_set_tx_wroff(connp->conn_rq, connp,
3021 				    size);
3022 			}
3023 			break;
3024 		default:
3025 			*outlenp = 0;
3026 			return (EINVAL);
3027 		}
3028 		break;
3029 	default:
3030 		*outlenp = 0;
3031 		return (EINVAL);
3032 	}
3033 	/*
3034 	 * Common case of OK return with outval same as inval.
3035 	 */
3036 	if (invalp != outvalp) {
3037 		/* don't trust bcopy for identical src/dst */
3038 		(void) bcopy(invalp, outvalp, inlen);
3039 	}
3040 	*outlenp = inlen;
3041 	return (0);
3042 }
3043 
3044 int
3045 udp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3046     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3047     void *thisdg_attrs, cred_t *cr)
3048 {
3049 	int		error;
3050 	boolean_t	checkonly;
3051 
3052 	error = 0;
3053 	switch (optset_context) {
3054 	case SETFN_OPTCOM_CHECKONLY:
3055 		checkonly = B_TRUE;
3056 		/*
3057 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3058 		 * inlen != 0 implies value supplied and
3059 		 * 	we have to "pretend" to set it.
3060 		 * inlen == 0 implies that there is no
3061 		 * 	value part in T_CHECK request and just validation
3062 		 * done elsewhere should be enough, we just return here.
3063 		 */
3064 		if (inlen == 0) {
3065 			*outlenp = 0;
3066 			goto done;
3067 		}
3068 		break;
3069 	case SETFN_OPTCOM_NEGOTIATE:
3070 		checkonly = B_FALSE;
3071 		break;
3072 	case SETFN_UD_NEGOTIATE:
3073 	case SETFN_CONN_NEGOTIATE:
3074 		checkonly = B_FALSE;
3075 		/*
3076 		 * Negotiating local and "association-related" options
3077 		 * through T_UNITDATA_REQ.
3078 		 *
3079 		 * Following routine can filter out ones we do not
3080 		 * want to be "set" this way.
3081 		 */
3082 		if (!udp_opt_allow_udr_set(level, name)) {
3083 			*outlenp = 0;
3084 			error = EINVAL;
3085 			goto done;
3086 		}
3087 		break;
3088 	default:
3089 		/*
3090 		 * We should never get here
3091 		 */
3092 		*outlenp = 0;
3093 		error = EINVAL;
3094 		goto done;
3095 	}
3096 
3097 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3098 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3099 
3100 	error = udp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3101 	    outvalp, cr, thisdg_attrs, checkonly);
3102 done:
3103 	return (error);
3104 }
3105 
3106 /* ARGSUSED */
3107 int
3108 udp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3109     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3110     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3111 {
3112 	conn_t  *connp =  Q_TO_CONN(q);
3113 	int error;
3114 	udp_t	*udp = connp->conn_udp;
3115 
3116 	rw_enter(&udp->udp_rwlock, RW_WRITER);
3117 	error = udp_opt_set(connp, optset_context, level, name, inlen, invalp,
3118 	    outlenp, outvalp, thisdg_attrs, cr);
3119 	rw_exit(&udp->udp_rwlock);
3120 	return (error);
3121 }
3122 
3123 /*
3124  * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl.
3125  * The headers include ip6i_t (if needed), ip6_t, any sticky extension
3126  * headers, and the udp header.
3127  * Returns failure if can't allocate memory.
3128  */
3129 static int
3130 udp_build_hdrs(udp_t *udp)
3131 {
3132 	udp_stack_t *us = udp->udp_us;
3133 	uchar_t	*hdrs;
3134 	uint_t	hdrs_len;
3135 	ip6_t	*ip6h;
3136 	ip6i_t	*ip6i;
3137 	udpha_t	*udpha;
3138 	ip6_pkt_t *ipp = &udp->udp_sticky_ipp;
3139 	size_t	sth_wroff;
3140 	conn_t	*connp = udp->udp_connp;
3141 
3142 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
3143 	ASSERT(connp != NULL);
3144 
3145 	hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE;
3146 	ASSERT(hdrs_len != 0);
3147 	if (hdrs_len != udp->udp_sticky_hdrs_len) {
3148 		/* Need to reallocate */
3149 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3150 		if (hdrs == NULL)
3151 			return (ENOMEM);
3152 
3153 		if (udp->udp_sticky_hdrs_len != 0) {
3154 			kmem_free(udp->udp_sticky_hdrs,
3155 			    udp->udp_sticky_hdrs_len);
3156 		}
3157 		udp->udp_sticky_hdrs = hdrs;
3158 		udp->udp_sticky_hdrs_len = hdrs_len;
3159 	}
3160 	ip_build_hdrs_v6(udp->udp_sticky_hdrs,
3161 	    udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP);
3162 
3163 	/* Set header fields not in ipp */
3164 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3165 		ip6i = (ip6i_t *)udp->udp_sticky_hdrs;
3166 		ip6h = (ip6_t *)&ip6i[1];
3167 	} else {
3168 		ip6h = (ip6_t *)udp->udp_sticky_hdrs;
3169 	}
3170 
3171 	if (!(ipp->ipp_fields & IPPF_ADDR))
3172 		ip6h->ip6_src = udp->udp_v6src;
3173 
3174 	udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE);
3175 	udpha->uha_src_port = udp->udp_port;
3176 
3177 	/* Try to get everything in a single mblk */
3178 	if (hdrs_len > udp->udp_max_hdr_len) {
3179 		udp->udp_max_hdr_len = hdrs_len;
3180 		sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
3181 		rw_exit(&udp->udp_rwlock);
3182 		(void) proto_set_tx_wroff(udp->udp_connp->conn_rq,
3183 		    udp->udp_connp, sth_wroff);
3184 		rw_enter(&udp->udp_rwlock, RW_WRITER);
3185 	}
3186 	return (0);
3187 }
3188 
3189 /*
3190  * This routine retrieves the value of an ND variable in a udpparam_t
3191  * structure.  It is called through nd_getset when a user reads the
3192  * variable.
3193  */
3194 /* ARGSUSED */
3195 static int
3196 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3197 {
3198 	udpparam_t *udppa = (udpparam_t *)cp;
3199 
3200 	(void) mi_mpprintf(mp, "%d", udppa->udp_param_value);
3201 	return (0);
3202 }
3203 
3204 /*
3205  * Walk through the param array specified registering each element with the
3206  * named dispatch (ND) handler.
3207  */
3208 static boolean_t
3209 udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt)
3210 {
3211 	for (; cnt-- > 0; udppa++) {
3212 		if (udppa->udp_param_name && udppa->udp_param_name[0]) {
3213 			if (!nd_load(ndp, udppa->udp_param_name,
3214 			    udp_param_get, udp_param_set,
3215 			    (caddr_t)udppa)) {
3216 				nd_free(ndp);
3217 				return (B_FALSE);
3218 			}
3219 		}
3220 	}
3221 	if (!nd_load(ndp, "udp_extra_priv_ports",
3222 	    udp_extra_priv_ports_get, NULL, NULL)) {
3223 		nd_free(ndp);
3224 		return (B_FALSE);
3225 	}
3226 	if (!nd_load(ndp, "udp_extra_priv_ports_add",
3227 	    NULL, udp_extra_priv_ports_add, NULL)) {
3228 		nd_free(ndp);
3229 		return (B_FALSE);
3230 	}
3231 	if (!nd_load(ndp, "udp_extra_priv_ports_del",
3232 	    NULL, udp_extra_priv_ports_del, NULL)) {
3233 		nd_free(ndp);
3234 		return (B_FALSE);
3235 	}
3236 	return (B_TRUE);
3237 }
3238 
3239 /* This routine sets an ND variable in a udpparam_t structure. */
3240 /* ARGSUSED */
3241 static int
3242 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3243 {
3244 	long		new_value;
3245 	udpparam_t	*udppa = (udpparam_t *)cp;
3246 
3247 	/*
3248 	 * Fail the request if the new value does not lie within the
3249 	 * required bounds.
3250 	 */
3251 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3252 	    new_value < udppa->udp_param_min ||
3253 	    new_value > udppa->udp_param_max) {
3254 		return (EINVAL);
3255 	}
3256 
3257 	/* Set the new value */
3258 	udppa->udp_param_value = new_value;
3259 	return (0);
3260 }
3261 
3262 /*
3263  * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with
3264  * T_opthdr) and return the number of bytes copied.  'dbuf' may be NULL to
3265  * just count the length needed for allocation.  If 'dbuf' is non-NULL,
3266  * then it's assumed to be allocated to be large enough.
3267  *
3268  * Returns zero if trimming of the security option causes all options to go
3269  * away.
3270  */
3271 static size_t
3272 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf)
3273 {
3274 	struct T_opthdr *toh;
3275 	size_t hol = ipp->ipp_hopoptslen;
3276 	ip6_hbh_t *dstopt = NULL;
3277 	const ip6_hbh_t *srcopt = ipp->ipp_hopopts;
3278 	size_t tlen, olen, plen;
3279 	boolean_t deleting;
3280 	const struct ip6_opt *sopt, *lastpad;
3281 	struct ip6_opt *dopt;
3282 
3283 	if ((toh = (struct T_opthdr *)dbuf) != NULL) {
3284 		toh->level = IPPROTO_IPV6;
3285 		toh->name = IPV6_HOPOPTS;
3286 		toh->status = 0;
3287 		dstopt = (ip6_hbh_t *)(toh + 1);
3288 	}
3289 
3290 	/*
3291 	 * If labeling is enabled, then skip the label option
3292 	 * but get other options if there are any.
3293 	 */
3294 	if (is_system_labeled()) {
3295 		dopt = NULL;
3296 		if (dstopt != NULL) {
3297 			/* will fill in ip6h_len later */
3298 			dstopt->ip6h_nxt = srcopt->ip6h_nxt;
3299 			dopt = (struct ip6_opt *)(dstopt + 1);
3300 		}
3301 		sopt = (const struct ip6_opt *)(srcopt + 1);
3302 		hol -= sizeof (*srcopt);
3303 		tlen = sizeof (*dstopt);
3304 		lastpad = NULL;
3305 		deleting = B_FALSE;
3306 		/*
3307 		 * This loop finds the first (lastpad pointer) of any number of
3308 		 * pads that preceeds the security option, then treats the
3309 		 * security option as though it were a pad, and then finds the
3310 		 * next non-pad option (or end of list).
3311 		 *
3312 		 * It then treats the entire block as one big pad.  To preserve
3313 		 * alignment of any options that follow, or just the end of the
3314 		 * list, it computes a minimal new padding size that keeps the
3315 		 * same alignment for the next option.
3316 		 *
3317 		 * If it encounters just a sequence of pads with no security
3318 		 * option, those are copied as-is rather than collapsed.
3319 		 *
3320 		 * Note that to handle the end of list case, the code makes one
3321 		 * loop with 'hol' set to zero.
3322 		 */
3323 		for (;;) {
3324 			if (hol > 0) {
3325 				if (sopt->ip6o_type == IP6OPT_PAD1) {
3326 					if (lastpad == NULL)
3327 						lastpad = sopt;
3328 					sopt = (const struct ip6_opt *)
3329 					    &sopt->ip6o_len;
3330 					hol--;
3331 					continue;
3332 				}
3333 				olen = sopt->ip6o_len + sizeof (*sopt);
3334 				if (olen > hol)
3335 					olen = hol;
3336 				if (sopt->ip6o_type == IP6OPT_PADN ||
3337 				    sopt->ip6o_type == ip6opt_ls) {
3338 					if (sopt->ip6o_type == ip6opt_ls)
3339 						deleting = B_TRUE;
3340 					if (lastpad == NULL)
3341 						lastpad = sopt;
3342 					sopt = (const struct ip6_opt *)
3343 					    ((const char *)sopt + olen);
3344 					hol -= olen;
3345 					continue;
3346 				}
3347 			} else {
3348 				/* if nothing was copied at all, then delete */
3349 				if (tlen == sizeof (*dstopt))
3350 					return (0);
3351 				/* last pass; pick up any trailing padding */
3352 				olen = 0;
3353 			}
3354 			if (deleting) {
3355 				/*
3356 				 * compute aligning effect of deleted material
3357 				 * to reproduce with pad.
3358 				 */
3359 				plen = ((const char *)sopt -
3360 				    (const char *)lastpad) & 7;
3361 				tlen += plen;
3362 				if (dopt != NULL) {
3363 					if (plen == 1) {
3364 						dopt->ip6o_type = IP6OPT_PAD1;
3365 					} else if (plen > 1) {
3366 						plen -= sizeof (*dopt);
3367 						dopt->ip6o_type = IP6OPT_PADN;
3368 						dopt->ip6o_len = plen;
3369 						if (plen > 0)
3370 							bzero(dopt + 1, plen);
3371 					}
3372 					dopt = (struct ip6_opt *)
3373 					    ((char *)dopt + plen);
3374 				}
3375 				deleting = B_FALSE;
3376 				lastpad = NULL;
3377 			}
3378 			/* if there's uncopied padding, then copy that now */
3379 			if (lastpad != NULL) {
3380 				olen += (const char *)sopt -
3381 				    (const char *)lastpad;
3382 				sopt = lastpad;
3383 				lastpad = NULL;
3384 			}
3385 			if (dopt != NULL && olen > 0) {
3386 				bcopy(sopt, dopt, olen);
3387 				dopt = (struct ip6_opt *)((char *)dopt + olen);
3388 			}
3389 			if (hol == 0)
3390 				break;
3391 			tlen += olen;
3392 			sopt = (const struct ip6_opt *)
3393 			    ((const char *)sopt + olen);
3394 			hol -= olen;
3395 		}
3396 		/* go back and patch up the length value, rounded upward */
3397 		if (dstopt != NULL)
3398 			dstopt->ip6h_len = (tlen - 1) >> 3;
3399 	} else {
3400 		tlen = hol;
3401 		if (dstopt != NULL)
3402 			bcopy(srcopt, dstopt, hol);
3403 	}
3404 
3405 	tlen += sizeof (*toh);
3406 	if (toh != NULL)
3407 		toh->len = tlen;
3408 
3409 	return (tlen);
3410 }
3411 
3412 /*
3413  * Update udp_rcv_opt_len from the packet.
3414  * Called when options received, and when no options received but
3415  * udp_ip_recv_opt_len has previously recorded options.
3416  */
3417 static void
3418 udp_save_ip_rcv_opt(udp_t *udp, void *opt, int opt_len)
3419 {
3420 	/* Save the options if any */
3421 	if (opt_len > 0) {
3422 		if (opt_len > udp->udp_ip_rcv_options_len) {
3423 			/* Need to allocate larger buffer */
3424 			if (udp->udp_ip_rcv_options_len != 0)
3425 				mi_free((char *)udp->udp_ip_rcv_options);
3426 			udp->udp_ip_rcv_options_len = 0;
3427 			udp->udp_ip_rcv_options =
3428 			    (uchar_t *)mi_alloc(opt_len, BPRI_HI);
3429 			if (udp->udp_ip_rcv_options != NULL)
3430 				udp->udp_ip_rcv_options_len = opt_len;
3431 		}
3432 		if (udp->udp_ip_rcv_options_len != 0) {
3433 			bcopy(opt, udp->udp_ip_rcv_options, opt_len);
3434 			/* Adjust length if we are resusing the space */
3435 			udp->udp_ip_rcv_options_len = opt_len;
3436 		}
3437 	} else if (udp->udp_ip_rcv_options_len != 0) {
3438 		/* Clear out previously recorded options */
3439 		mi_free((char *)udp->udp_ip_rcv_options);
3440 		udp->udp_ip_rcv_options = NULL;
3441 		udp->udp_ip_rcv_options_len = 0;
3442 	}
3443 }
3444 
3445 static mblk_t *
3446 udp_queue_fallback(udp_t *udp, mblk_t *mp)
3447 {
3448 	ASSERT(MUTEX_HELD(&udp->udp_recv_lock));
3449 	if (IPCL_IS_NONSTR(udp->udp_connp)) {
3450 		/*
3451 		 * fallback has started but messages have not been moved yet
3452 		 */
3453 		if (udp->udp_fallback_queue_head == NULL) {
3454 			ASSERT(udp->udp_fallback_queue_tail == NULL);
3455 			udp->udp_fallback_queue_head = mp;
3456 			udp->udp_fallback_queue_tail = mp;
3457 		} else {
3458 			ASSERT(udp->udp_fallback_queue_tail != NULL);
3459 			udp->udp_fallback_queue_tail->b_next = mp;
3460 			udp->udp_fallback_queue_tail = mp;
3461 		}
3462 		return (NULL);
3463 	} else {
3464 		/*
3465 		 * Fallback completed, let the caller putnext() the mblk.
3466 		 */
3467 		return (mp);
3468 	}
3469 }
3470 
3471 /*
3472  * Deliver data to ULP. In case we have a socket, and it's falling back to
3473  * TPI, then we'll queue the mp for later processing.
3474  */
3475 static void
3476 udp_ulp_recv(conn_t *connp, mblk_t *mp)
3477 {
3478 	if (IPCL_IS_NONSTR(connp)) {
3479 		udp_t *udp = connp->conn_udp;
3480 		int error;
3481 
3482 		if ((*connp->conn_upcalls->su_recv)
3483 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
3484 		    NULL) < 0) {
3485 			mutex_enter(&udp->udp_recv_lock);
3486 			if (error == ENOSPC) {
3487 				/*
3488 				 * let's confirm while holding the lock
3489 				 */
3490 				if ((*connp->conn_upcalls->su_recv)
3491 				    (connp->conn_upper_handle, NULL, 0, 0,
3492 				    &error, NULL) < 0) {
3493 					ASSERT(error == ENOSPC);
3494 					if (error == ENOSPC) {
3495 						connp->conn_flow_cntrld =
3496 						    B_TRUE;
3497 					}
3498 				}
3499 				mutex_exit(&udp->udp_recv_lock);
3500 			} else {
3501 				ASSERT(error == EOPNOTSUPP);
3502 				mp = udp_queue_fallback(udp, mp);
3503 				mutex_exit(&udp->udp_recv_lock);
3504 				if (mp != NULL)
3505 					putnext(connp->conn_rq, mp);
3506 			}
3507 		}
3508 		ASSERT(MUTEX_NOT_HELD(&udp->udp_recv_lock));
3509 	} else {
3510 		putnext(connp->conn_rq, mp);
3511 	}
3512 }
3513 
3514 /* ARGSUSED2 */
3515 static void
3516 udp_input(void *arg1, mblk_t *mp, void *arg2)
3517 {
3518 	conn_t *connp = (conn_t *)arg1;
3519 	struct T_unitdata_ind	*tudi;
3520 	uchar_t			*rptr;		/* Pointer to IP header */
3521 	int			hdr_length;	/* Length of IP+UDP headers */
3522 	int			opt_len;
3523 	int			udi_size;	/* Size of T_unitdata_ind */
3524 	int			mp_len;
3525 	udp_t			*udp;
3526 	udpha_t			*udpha;
3527 	int			ipversion;
3528 	ip6_pkt_t		ipp;
3529 	ip6_t			*ip6h;
3530 	ip6i_t			*ip6i;
3531 	mblk_t			*mp1;
3532 	mblk_t			*options_mp = NULL;
3533 	ip_pktinfo_t		*pinfo = NULL;
3534 	cred_t			*cr = NULL;
3535 	pid_t			cpid;
3536 	uint32_t		udp_ip_rcv_options_len;
3537 	udp_bits_t		udp_bits;
3538 	cred_t			*rcr = connp->conn_cred;
3539 	udp_stack_t *us;
3540 
3541 	ASSERT(connp->conn_flags & IPCL_UDPCONN);
3542 
3543 	udp = connp->conn_udp;
3544 	us = udp->udp_us;
3545 	rptr = mp->b_rptr;
3546 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3547 	ASSERT(OK_32PTR(rptr));
3548 
3549 	/*
3550 	 * IP should have prepended the options data in an M_CTL
3551 	 * Check M_CTL "type" to make sure are not here bcos of
3552 	 * a valid ICMP message
3553 	 */
3554 	if (DB_TYPE(mp) == M_CTL) {
3555 		if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3556 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3557 		    IN_PKTINFO) {
3558 			/*
3559 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3560 			 * has been prepended to the packet by IP. We need to
3561 			 * extract the mblk and adjust the rptr
3562 			 */
3563 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3564 			options_mp = mp;
3565 			mp = mp->b_cont;
3566 			rptr = mp->b_rptr;
3567 			UDP_STAT(us, udp_in_pktinfo);
3568 		} else {
3569 			/*
3570 			 * ICMP messages.
3571 			 */
3572 			udp_icmp_error(connp, mp);
3573 			return;
3574 		}
3575 	}
3576 
3577 	mp_len = msgdsize(mp);
3578 	/*
3579 	 * This is the inbound data path.
3580 	 * First, we check to make sure the IP version number is correct,
3581 	 * and then pull the IP and UDP headers into the first mblk.
3582 	 */
3583 
3584 	/* Initialize regardless if ipversion is IPv4 or IPv6 */
3585 	ipp.ipp_fields = 0;
3586 
3587 	ipversion = IPH_HDR_VERSION(rptr);
3588 
3589 	rw_enter(&udp->udp_rwlock, RW_READER);
3590 	udp_ip_rcv_options_len = udp->udp_ip_rcv_options_len;
3591 	udp_bits = udp->udp_bits;
3592 	rw_exit(&udp->udp_rwlock);
3593 
3594 	switch (ipversion) {
3595 	case IPV4_VERSION:
3596 		ASSERT(MBLKL(mp) >= sizeof (ipha_t));
3597 		ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP);
3598 		hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
3599 		opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE);
3600 		if ((opt_len > 0 || udp_ip_rcv_options_len > 0) &&
3601 		    udp->udp_family == AF_INET) {
3602 			/*
3603 			 * Record/update udp_ip_rcv_options with the lock
3604 			 * held. Not needed for AF_INET6 sockets
3605 			 * since they don't support a getsockopt of IP_OPTIONS.
3606 			 */
3607 			rw_enter(&udp->udp_rwlock, RW_WRITER);
3608 			udp_save_ip_rcv_opt(udp, rptr + IP_SIMPLE_HDR_LENGTH,
3609 			    opt_len);
3610 			rw_exit(&udp->udp_rwlock);
3611 		}
3612 		/* Handle IPV6_RECVPKTINFO even for IPv4 packet. */
3613 		if ((udp->udp_family == AF_INET6) && (pinfo != NULL) &&
3614 		    udp->udp_ip_recvpktinfo) {
3615 			if (pinfo->ip_pkt_flags & IPF_RECVIF) {
3616 				ipp.ipp_fields |= IPPF_IFINDEX;
3617 				ipp.ipp_ifindex = pinfo->ip_pkt_ifindex;
3618 			}
3619 		}
3620 		break;
3621 	case IPV6_VERSION:
3622 		/*
3623 		 * IPv6 packets can only be received by applications
3624 		 * that are prepared to receive IPv6 addresses.
3625 		 * The IP fanout must ensure this.
3626 		 */
3627 		ASSERT(udp->udp_family == AF_INET6);
3628 
3629 		ip6h = (ip6_t *)rptr;
3630 		ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr);
3631 
3632 		if (ip6h->ip6_nxt != IPPROTO_UDP) {
3633 			uint8_t nexthdrp;
3634 			/* Look for ifindex information */
3635 			if (ip6h->ip6_nxt == IPPROTO_RAW) {
3636 				ip6i = (ip6i_t *)ip6h;
3637 				if ((uchar_t *)&ip6i[1] > mp->b_wptr)
3638 					goto tossit;
3639 
3640 				if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3641 					ASSERT(ip6i->ip6i_ifindex != 0);
3642 					ipp.ipp_fields |= IPPF_IFINDEX;
3643 					ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3644 				}
3645 				rptr = (uchar_t *)&ip6i[1];
3646 				mp->b_rptr = rptr;
3647 				if (rptr == mp->b_wptr) {
3648 					mp1 = mp->b_cont;
3649 					freeb(mp);
3650 					mp = mp1;
3651 					rptr = mp->b_rptr;
3652 				}
3653 				if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE))
3654 					goto tossit;
3655 				ip6h = (ip6_t *)rptr;
3656 				mp_len = msgdsize(mp);
3657 			}
3658 			/*
3659 			 * Find any potentially interesting extension headers
3660 			 * as well as the length of the IPv6 + extension
3661 			 * headers.
3662 			 */
3663 			hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) +
3664 			    UDPH_SIZE;
3665 			ASSERT(nexthdrp == IPPROTO_UDP);
3666 		} else {
3667 			hdr_length = IPV6_HDR_LEN + UDPH_SIZE;
3668 			ip6i = NULL;
3669 		}
3670 		break;
3671 	default:
3672 		ASSERT(0);
3673 	}
3674 
3675 	/*
3676 	 * IP inspected the UDP header thus all of it must be in the mblk.
3677 	 * UDP length check is performed for IPv6 packets and IPv4 packets
3678 	 * to check if the size of the packet as specified
3679 	 * by the header is the same as the physical size of the packet.
3680 	 * FIXME? Didn't IP already check this?
3681 	 */
3682 	udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
3683 	if ((MBLKL(mp) < hdr_length) ||
3684 	    (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) {
3685 		goto tossit;
3686 	}
3687 
3688 
3689 	/* Walk past the headers unless UDP_RCVHDR was set. */
3690 	if (!udp_bits.udpb_rcvhdr) {
3691 		mp->b_rptr = rptr + hdr_length;
3692 		mp_len -= hdr_length;
3693 	}
3694 
3695 	/*
3696 	 * This is the inbound data path.  Packets are passed upstream as
3697 	 * T_UNITDATA_IND messages with full IP headers still attached.
3698 	 */
3699 	if (udp->udp_family == AF_INET) {
3700 		sin_t *sin;
3701 
3702 		ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
3703 
3704 		/*
3705 		 * Normally only send up the source address.
3706 		 * If IP_RECVDSTADDR is set we include the destination IP
3707 		 * address as an option. With IP_RECVOPTS we include all
3708 		 * the IP options.
3709 		 */
3710 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3711 		if (udp_bits.udpb_recvdstaddr) {
3712 			udi_size += sizeof (struct T_opthdr) +
3713 			    sizeof (struct in_addr);
3714 			UDP_STAT(us, udp_in_recvdstaddr);
3715 		}
3716 
3717 		if (udp_bits.udpb_ip_recvpktinfo && (pinfo != NULL) &&
3718 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3719 			udi_size += sizeof (struct T_opthdr) +
3720 			    sizeof (struct in_pktinfo);
3721 			UDP_STAT(us, udp_ip_rcvpktinfo);
3722 		}
3723 
3724 		if ((udp_bits.udpb_recvopts) && opt_len > 0) {
3725 			udi_size += sizeof (struct T_opthdr) + opt_len;
3726 			UDP_STAT(us, udp_in_recvopts);
3727 		}
3728 
3729 		/*
3730 		 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
3731 		 * space accordingly
3732 		 */
3733 		if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
3734 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3735 			udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
3736 			UDP_STAT(us, udp_in_recvif);
3737 		}
3738 
3739 		if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
3740 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
3741 			udi_size += sizeof (struct T_opthdr) +
3742 			    sizeof (struct sockaddr_dl);
3743 			UDP_STAT(us, udp_in_recvslla);
3744 		}
3745 
3746 		if ((udp_bits.udpb_recvucred) &&
3747 		    (cr = msg_getcred(mp, &cpid)) != NULL) {
3748 			udi_size += sizeof (struct T_opthdr) + ucredsize;
3749 			UDP_STAT(us, udp_in_recvucred);
3750 		}
3751 
3752 		/*
3753 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3754 		 * buffer. Since gethrestime() expects a pointer aligned
3755 		 * argument, we allocate space necessary for extra
3756 		 * alignment (even though it might not be used).
3757 		 */
3758 		if (udp_bits.udpb_timestamp) {
3759 			udi_size += sizeof (struct T_opthdr) +
3760 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3761 			UDP_STAT(us, udp_in_timestamp);
3762 		}
3763 
3764 		/*
3765 		 * If IP_RECVTTL is set allocate the appropriate sized buffer
3766 		 */
3767 		if (udp_bits.udpb_recvttl) {
3768 			udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
3769 			UDP_STAT(us, udp_in_recvttl);
3770 		}
3771 
3772 		/* Allocate a message block for the T_UNITDATA_IND structure. */
3773 		mp1 = allocb(udi_size, BPRI_MED);
3774 		if (mp1 == NULL) {
3775 			freemsg(mp);
3776 			if (options_mp != NULL)
3777 				freeb(options_mp);
3778 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
3779 			return;
3780 		}
3781 		mp1->b_cont = mp;
3782 		mp = mp1;
3783 		mp->b_datap->db_type = M_PROTO;
3784 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3785 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3786 		tudi->PRIM_type = T_UNITDATA_IND;
3787 		tudi->SRC_length = sizeof (sin_t);
3788 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3789 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
3790 		    sizeof (sin_t);
3791 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3792 		tudi->OPT_length = udi_size;
3793 		sin = (sin_t *)&tudi[1];
3794 		sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
3795 		sin->sin_port =	udpha->uha_src_port;
3796 		sin->sin_family = udp->udp_family;
3797 		*(uint32_t *)&sin->sin_zero[0] = 0;
3798 		*(uint32_t *)&sin->sin_zero[4] = 0;
3799 
3800 		/*
3801 		 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
3802 		 * IP_RECVTTL has been set.
3803 		 */
3804 		if (udi_size != 0) {
3805 			/*
3806 			 * Copy in destination address before options to avoid
3807 			 * any padding issues.
3808 			 */
3809 			char *dstopt;
3810 
3811 			dstopt = (char *)&sin[1];
3812 			if (udp_bits.udpb_recvdstaddr) {
3813 				struct T_opthdr *toh;
3814 				ipaddr_t *dstptr;
3815 
3816 				toh = (struct T_opthdr *)dstopt;
3817 				toh->level = IPPROTO_IP;
3818 				toh->name = IP_RECVDSTADDR;
3819 				toh->len = sizeof (struct T_opthdr) +
3820 				    sizeof (ipaddr_t);
3821 				toh->status = 0;
3822 				dstopt += sizeof (struct T_opthdr);
3823 				dstptr = (ipaddr_t *)dstopt;
3824 				*dstptr = ((ipha_t *)rptr)->ipha_dst;
3825 				dstopt += sizeof (ipaddr_t);
3826 				udi_size -= toh->len;
3827 			}
3828 
3829 			if (udp_bits.udpb_recvopts && opt_len > 0) {
3830 				struct T_opthdr *toh;
3831 
3832 				toh = (struct T_opthdr *)dstopt;
3833 				toh->level = IPPROTO_IP;
3834 				toh->name = IP_RECVOPTS;
3835 				toh->len = sizeof (struct T_opthdr) + opt_len;
3836 				toh->status = 0;
3837 				dstopt += sizeof (struct T_opthdr);
3838 				bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt,
3839 				    opt_len);
3840 				dstopt += opt_len;
3841 				udi_size -= toh->len;
3842 			}
3843 
3844 			if ((udp_bits.udpb_ip_recvpktinfo) && (pinfo != NULL) &&
3845 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3846 				struct T_opthdr *toh;
3847 				struct in_pktinfo *pktinfop;
3848 
3849 				toh = (struct T_opthdr *)dstopt;
3850 				toh->level = IPPROTO_IP;
3851 				toh->name = IP_PKTINFO;
3852 				toh->len = sizeof (struct T_opthdr) +
3853 				    sizeof (*pktinfop);
3854 				toh->status = 0;
3855 				dstopt += sizeof (struct T_opthdr);
3856 				pktinfop = (struct in_pktinfo *)dstopt;
3857 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3858 				pktinfop->ipi_spec_dst =
3859 				    pinfo->ip_pkt_match_addr;
3860 				pktinfop->ipi_addr.s_addr =
3861 				    ((ipha_t *)rptr)->ipha_dst;
3862 
3863 				dstopt += sizeof (struct in_pktinfo);
3864 				udi_size -= toh->len;
3865 			}
3866 
3867 			if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
3868 			    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
3869 
3870 				struct T_opthdr *toh;
3871 				struct sockaddr_dl	*dstptr;
3872 
3873 				toh = (struct T_opthdr *)dstopt;
3874 				toh->level = IPPROTO_IP;
3875 				toh->name = IP_RECVSLLA;
3876 				toh->len = sizeof (struct T_opthdr) +
3877 				    sizeof (struct sockaddr_dl);
3878 				toh->status = 0;
3879 				dstopt += sizeof (struct T_opthdr);
3880 				dstptr = (struct sockaddr_dl *)dstopt;
3881 				bcopy(&pinfo->ip_pkt_slla, dstptr,
3882 				    sizeof (struct sockaddr_dl));
3883 				dstopt += sizeof (struct sockaddr_dl);
3884 				udi_size -= toh->len;
3885 			}
3886 
3887 			if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
3888 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3889 
3890 				struct T_opthdr *toh;
3891 				uint_t		*dstptr;
3892 
3893 				toh = (struct T_opthdr *)dstopt;
3894 				toh->level = IPPROTO_IP;
3895 				toh->name = IP_RECVIF;
3896 				toh->len = sizeof (struct T_opthdr) +
3897 				    sizeof (uint_t);
3898 				toh->status = 0;
3899 				dstopt += sizeof (struct T_opthdr);
3900 				dstptr = (uint_t *)dstopt;
3901 				*dstptr = pinfo->ip_pkt_ifindex;
3902 				dstopt += sizeof (uint_t);
3903 				udi_size -= toh->len;
3904 			}
3905 
3906 			if (cr != NULL) {
3907 				struct T_opthdr *toh;
3908 
3909 				toh = (struct T_opthdr *)dstopt;
3910 				toh->level = SOL_SOCKET;
3911 				toh->name = SCM_UCRED;
3912 				toh->len = sizeof (struct T_opthdr) + ucredsize;
3913 				toh->status = 0;
3914 				dstopt += sizeof (struct T_opthdr);
3915 				(void) cred2ucred(cr, cpid, dstopt, rcr);
3916 				dstopt += ucredsize;
3917 				udi_size -= toh->len;
3918 			}
3919 
3920 			if (udp_bits.udpb_timestamp) {
3921 				struct	T_opthdr *toh;
3922 
3923 				toh = (struct T_opthdr *)dstopt;
3924 				toh->level = SOL_SOCKET;
3925 				toh->name = SCM_TIMESTAMP;
3926 				toh->len = sizeof (struct T_opthdr) +
3927 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3928 				toh->status = 0;
3929 				dstopt += sizeof (struct T_opthdr);
3930 				/* Align for gethrestime() */
3931 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3932 				    sizeof (intptr_t));
3933 				gethrestime((timestruc_t *)dstopt);
3934 				dstopt = (char *)toh + toh->len;
3935 				udi_size -= toh->len;
3936 			}
3937 
3938 			/*
3939 			 * CAUTION:
3940 			 * Due to aligment issues
3941 			 * Processing of IP_RECVTTL option
3942 			 * should always be the last. Adding
3943 			 * any option processing after this will
3944 			 * cause alignment panic.
3945 			 */
3946 			if (udp_bits.udpb_recvttl) {
3947 				struct	T_opthdr *toh;
3948 				uint8_t	*dstptr;
3949 
3950 				toh = (struct T_opthdr *)dstopt;
3951 				toh->level = IPPROTO_IP;
3952 				toh->name = IP_RECVTTL;
3953 				toh->len = sizeof (struct T_opthdr) +
3954 				    sizeof (uint8_t);
3955 				toh->status = 0;
3956 				dstopt += sizeof (struct T_opthdr);
3957 				dstptr = (uint8_t *)dstopt;
3958 				*dstptr = ((ipha_t *)rptr)->ipha_ttl;
3959 				dstopt += sizeof (uint8_t);
3960 				udi_size -= toh->len;
3961 			}
3962 
3963 			/* Consumed all of allocated space */
3964 			ASSERT(udi_size == 0);
3965 		}
3966 	} else {
3967 		sin6_t *sin6;
3968 
3969 		/*
3970 		 * Handle both IPv4 and IPv6 packets for IPv6 sockets.
3971 		 *
3972 		 * Normally we only send up the address. If receiving of any
3973 		 * optional receive side information is enabled, we also send
3974 		 * that up as options.
3975 		 */
3976 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3977 
3978 		if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3979 		    IPPF_RTHDR|IPPF_IFINDEX)) {
3980 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
3981 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
3982 				size_t hlen;
3983 
3984 				UDP_STAT(us, udp_in_recvhopopts);
3985 				hlen = copy_hop_opts(&ipp, NULL);
3986 				if (hlen == 0)
3987 					ipp.ipp_fields &= ~IPPF_HOPOPTS;
3988 				udi_size += hlen;
3989 			}
3990 			if (((udp_bits.udpb_ipv6_recvdstopts) ||
3991 			    udp_bits.udpb_old_ipv6_recvdstopts) &&
3992 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3993 				udi_size += sizeof (struct T_opthdr) +
3994 				    ipp.ipp_dstoptslen;
3995 				UDP_STAT(us, udp_in_recvdstopts);
3996 			}
3997 			if ((((udp_bits.udpb_ipv6_recvdstopts) &&
3998 			    udp_bits.udpb_ipv6_recvrthdr &&
3999 			    (ipp.ipp_fields & IPPF_RTHDR)) ||
4000 			    (udp_bits.udpb_ipv6_recvrthdrdstopts)) &&
4001 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4002 				udi_size += sizeof (struct T_opthdr) +
4003 				    ipp.ipp_rtdstoptslen;
4004 				UDP_STAT(us, udp_in_recvrtdstopts);
4005 			}
4006 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4007 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4008 				udi_size += sizeof (struct T_opthdr) +
4009 				    ipp.ipp_rthdrlen;
4010 				UDP_STAT(us, udp_in_recvrthdr);
4011 			}
4012 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4013 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4014 				udi_size += sizeof (struct T_opthdr) +
4015 				    sizeof (struct in6_pktinfo);
4016 				UDP_STAT(us, udp_in_recvpktinfo);
4017 			}
4018 
4019 		}
4020 		if ((udp_bits.udpb_recvucred) &&
4021 		    (cr = msg_getcred(mp, &cpid)) != NULL) {
4022 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4023 			UDP_STAT(us, udp_in_recvucred);
4024 		}
4025 
4026 		/*
4027 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4028 		 * buffer. Since gethrestime() expects a pointer aligned
4029 		 * argument, we allocate space necessary for extra
4030 		 * alignment (even though it might not be used).
4031 		 */
4032 		if (udp_bits.udpb_timestamp) {
4033 			udi_size += sizeof (struct T_opthdr) +
4034 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4035 			UDP_STAT(us, udp_in_timestamp);
4036 		}
4037 
4038 		if (udp_bits.udpb_ipv6_recvhoplimit) {
4039 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4040 			UDP_STAT(us, udp_in_recvhoplimit);
4041 		}
4042 
4043 		if (udp_bits.udpb_ipv6_recvtclass) {
4044 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4045 			UDP_STAT(us, udp_in_recvtclass);
4046 		}
4047 
4048 		mp1 = allocb(udi_size, BPRI_MED);
4049 		if (mp1 == NULL) {
4050 			freemsg(mp);
4051 			if (options_mp != NULL)
4052 				freeb(options_mp);
4053 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
4054 			return;
4055 		}
4056 		mp1->b_cont = mp;
4057 		mp = mp1;
4058 		mp->b_datap->db_type = M_PROTO;
4059 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4060 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4061 		tudi->PRIM_type = T_UNITDATA_IND;
4062 		tudi->SRC_length = sizeof (sin6_t);
4063 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4064 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4065 		    sizeof (sin6_t);
4066 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
4067 		tudi->OPT_length = udi_size;
4068 		sin6 = (sin6_t *)&tudi[1];
4069 		if (ipversion == IPV4_VERSION) {
4070 			in6_addr_t v6dst;
4071 
4072 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src,
4073 			    &sin6->sin6_addr);
4074 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst,
4075 			    &v6dst);
4076 			sin6->sin6_flowinfo = 0;
4077 			sin6->sin6_scope_id = 0;
4078 			sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst,
4079 			    connp->conn_zoneid, us->us_netstack);
4080 		} else {
4081 			sin6->sin6_addr = ip6h->ip6_src;
4082 			/* No sin6_flowinfo per API */
4083 			sin6->sin6_flowinfo = 0;
4084 			/* For link-scope source pass up scope id */
4085 			if ((ipp.ipp_fields & IPPF_IFINDEX) &&
4086 			    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
4087 				sin6->sin6_scope_id = ipp.ipp_ifindex;
4088 			else
4089 				sin6->sin6_scope_id = 0;
4090 			sin6->__sin6_src_id = ip_srcid_find_addr(
4091 			    &ip6h->ip6_dst, connp->conn_zoneid,
4092 			    us->us_netstack);
4093 		}
4094 		sin6->sin6_port = udpha->uha_src_port;
4095 		sin6->sin6_family = udp->udp_family;
4096 
4097 		if (udi_size != 0) {
4098 			uchar_t *dstopt;
4099 
4100 			dstopt = (uchar_t *)&sin6[1];
4101 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4102 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4103 				struct T_opthdr *toh;
4104 				struct in6_pktinfo *pkti;
4105 
4106 				toh = (struct T_opthdr *)dstopt;
4107 				toh->level = IPPROTO_IPV6;
4108 				toh->name = IPV6_PKTINFO;
4109 				toh->len = sizeof (struct T_opthdr) +
4110 				    sizeof (*pkti);
4111 				toh->status = 0;
4112 				dstopt += sizeof (struct T_opthdr);
4113 				pkti = (struct in6_pktinfo *)dstopt;
4114 				if (ipversion == IPV6_VERSION)
4115 					pkti->ipi6_addr = ip6h->ip6_dst;
4116 				else
4117 					IN6_IPADDR_TO_V4MAPPED(
4118 					    ((ipha_t *)rptr)->ipha_dst,
4119 					    &pkti->ipi6_addr);
4120 				pkti->ipi6_ifindex = ipp.ipp_ifindex;
4121 				dstopt += sizeof (*pkti);
4122 				udi_size -= toh->len;
4123 			}
4124 			if (udp_bits.udpb_ipv6_recvhoplimit) {
4125 				struct T_opthdr *toh;
4126 
4127 				toh = (struct T_opthdr *)dstopt;
4128 				toh->level = IPPROTO_IPV6;
4129 				toh->name = IPV6_HOPLIMIT;
4130 				toh->len = sizeof (struct T_opthdr) +
4131 				    sizeof (uint_t);
4132 				toh->status = 0;
4133 				dstopt += sizeof (struct T_opthdr);
4134 				if (ipversion == IPV6_VERSION)
4135 					*(uint_t *)dstopt = ip6h->ip6_hops;
4136 				else
4137 					*(uint_t *)dstopt =
4138 					    ((ipha_t *)rptr)->ipha_ttl;
4139 				dstopt += sizeof (uint_t);
4140 				udi_size -= toh->len;
4141 			}
4142 			if (udp_bits.udpb_ipv6_recvtclass) {
4143 				struct T_opthdr *toh;
4144 
4145 				toh = (struct T_opthdr *)dstopt;
4146 				toh->level = IPPROTO_IPV6;
4147 				toh->name = IPV6_TCLASS;
4148 				toh->len = sizeof (struct T_opthdr) +
4149 				    sizeof (uint_t);
4150 				toh->status = 0;
4151 				dstopt += sizeof (struct T_opthdr);
4152 				if (ipversion == IPV6_VERSION) {
4153 					*(uint_t *)dstopt =
4154 					    IPV6_FLOW_TCLASS(ip6h->ip6_flow);
4155 				} else {
4156 					ipha_t *ipha = (ipha_t *)rptr;
4157 					*(uint_t *)dstopt =
4158 					    ipha->ipha_type_of_service;
4159 				}
4160 				dstopt += sizeof (uint_t);
4161 				udi_size -= toh->len;
4162 			}
4163 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4164 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4165 				size_t hlen;
4166 
4167 				hlen = copy_hop_opts(&ipp, dstopt);
4168 				dstopt += hlen;
4169 				udi_size -= hlen;
4170 			}
4171 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4172 			    (udp_bits.udpb_ipv6_recvrthdr) &&
4173 			    (ipp.ipp_fields & IPPF_RTHDR) &&
4174 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4175 				struct T_opthdr *toh;
4176 
4177 				toh = (struct T_opthdr *)dstopt;
4178 				toh->level = IPPROTO_IPV6;
4179 				toh->name = IPV6_DSTOPTS;
4180 				toh->len = sizeof (struct T_opthdr) +
4181 				    ipp.ipp_rtdstoptslen;
4182 				toh->status = 0;
4183 				dstopt += sizeof (struct T_opthdr);
4184 				bcopy(ipp.ipp_rtdstopts, dstopt,
4185 				    ipp.ipp_rtdstoptslen);
4186 				dstopt += ipp.ipp_rtdstoptslen;
4187 				udi_size -= toh->len;
4188 			}
4189 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4190 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4191 				struct T_opthdr *toh;
4192 
4193 				toh = (struct T_opthdr *)dstopt;
4194 				toh->level = IPPROTO_IPV6;
4195 				toh->name = IPV6_RTHDR;
4196 				toh->len = sizeof (struct T_opthdr) +
4197 				    ipp.ipp_rthdrlen;
4198 				toh->status = 0;
4199 				dstopt += sizeof (struct T_opthdr);
4200 				bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
4201 				dstopt += ipp.ipp_rthdrlen;
4202 				udi_size -= toh->len;
4203 			}
4204 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4205 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4206 				struct T_opthdr *toh;
4207 
4208 				toh = (struct T_opthdr *)dstopt;
4209 				toh->level = IPPROTO_IPV6;
4210 				toh->name = IPV6_DSTOPTS;
4211 				toh->len = sizeof (struct T_opthdr) +
4212 				    ipp.ipp_dstoptslen;
4213 				toh->status = 0;
4214 				dstopt += sizeof (struct T_opthdr);
4215 				bcopy(ipp.ipp_dstopts, dstopt,
4216 				    ipp.ipp_dstoptslen);
4217 				dstopt += ipp.ipp_dstoptslen;
4218 				udi_size -= toh->len;
4219 			}
4220 			if (cr != NULL) {
4221 				struct T_opthdr *toh;
4222 
4223 				toh = (struct T_opthdr *)dstopt;
4224 				toh->level = SOL_SOCKET;
4225 				toh->name = SCM_UCRED;
4226 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4227 				toh->status = 0;
4228 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
4229 				dstopt += toh->len;
4230 				udi_size -= toh->len;
4231 			}
4232 			if (udp_bits.udpb_timestamp) {
4233 				struct	T_opthdr *toh;
4234 
4235 				toh = (struct T_opthdr *)dstopt;
4236 				toh->level = SOL_SOCKET;
4237 				toh->name = SCM_TIMESTAMP;
4238 				toh->len = sizeof (struct T_opthdr) +
4239 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4240 				toh->status = 0;
4241 				dstopt += sizeof (struct T_opthdr);
4242 				/* Align for gethrestime() */
4243 				dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
4244 				    sizeof (intptr_t));
4245 				gethrestime((timestruc_t *)dstopt);
4246 				dstopt = (uchar_t *)toh + toh->len;
4247 				udi_size -= toh->len;
4248 			}
4249 
4250 			/* Consumed all of allocated space */
4251 			ASSERT(udi_size == 0);
4252 		}
4253 #undef	sin6
4254 		/* No IP_RECVDSTADDR for IPv6. */
4255 	}
4256 
4257 	BUMP_MIB(&us->us_udp_mib, udpHCInDatagrams);
4258 	if (options_mp != NULL)
4259 		freeb(options_mp);
4260 
4261 	udp_ulp_recv(connp, mp);
4262 
4263 	return;
4264 
4265 tossit:
4266 	freemsg(mp);
4267 	if (options_mp != NULL)
4268 		freeb(options_mp);
4269 	BUMP_MIB(&us->us_udp_mib, udpInErrors);
4270 }
4271 
4272 /*
4273  * return SNMP stuff in buffer in mpdata. We don't hold any lock and report
4274  * information that can be changing beneath us.
4275  */
4276 mblk_t *
4277 udp_snmp_get(queue_t *q, mblk_t *mpctl)
4278 {
4279 	mblk_t			*mpdata;
4280 	mblk_t			*mp_conn_ctl;
4281 	mblk_t			*mp_attr_ctl;
4282 	mblk_t			*mp6_conn_ctl;
4283 	mblk_t			*mp6_attr_ctl;
4284 	mblk_t			*mp_conn_tail;
4285 	mblk_t			*mp_attr_tail;
4286 	mblk_t			*mp6_conn_tail;
4287 	mblk_t			*mp6_attr_tail;
4288 	struct opthdr		*optp;
4289 	mib2_udpEntry_t		ude;
4290 	mib2_udp6Entry_t	ude6;
4291 	mib2_transportMLPEntry_t mlp;
4292 	int			state;
4293 	zoneid_t		zoneid;
4294 	int			i;
4295 	connf_t			*connfp;
4296 	conn_t			*connp = Q_TO_CONN(q);
4297 	int			v4_conn_idx;
4298 	int			v6_conn_idx;
4299 	boolean_t		needattr;
4300 	udp_t			*udp;
4301 	ip_stack_t		*ipst = connp->conn_netstack->netstack_ip;
4302 	udp_stack_t		*us = connp->conn_netstack->netstack_udp;
4303 	mblk_t			*mp2ctl;
4304 
4305 	/*
4306 	 * make a copy of the original message
4307 	 */
4308 	mp2ctl = copymsg(mpctl);
4309 
4310 	mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL;
4311 	if (mpctl == NULL ||
4312 	    (mpdata = mpctl->b_cont) == NULL ||
4313 	    (mp_conn_ctl = copymsg(mpctl)) == NULL ||
4314 	    (mp_attr_ctl = copymsg(mpctl)) == NULL ||
4315 	    (mp6_conn_ctl = copymsg(mpctl)) == NULL ||
4316 	    (mp6_attr_ctl = copymsg(mpctl)) == NULL) {
4317 		freemsg(mp_conn_ctl);
4318 		freemsg(mp_attr_ctl);
4319 		freemsg(mp6_conn_ctl);
4320 		freemsg(mpctl);
4321 		freemsg(mp2ctl);
4322 		return (0);
4323 	}
4324 
4325 	zoneid = connp->conn_zoneid;
4326 
4327 	/* fixed length structure for IPv4 and IPv6 counters */
4328 	SET_MIB(us->us_udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t));
4329 	SET_MIB(us->us_udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t));
4330 	/* synchronize 64- and 32-bit counters */
4331 	SYNC32_MIB(&us->us_udp_mib, udpInDatagrams, udpHCInDatagrams);
4332 	SYNC32_MIB(&us->us_udp_mib, udpOutDatagrams, udpHCOutDatagrams);
4333 
4334 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4335 	optp->level = MIB2_UDP;
4336 	optp->name = 0;
4337 	(void) snmp_append_data(mpdata, (char *)&us->us_udp_mib,
4338 	    sizeof (us->us_udp_mib));
4339 	optp->len = msgdsize(mpdata);
4340 	qreply(q, mpctl);
4341 
4342 	mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL;
4343 	v4_conn_idx = v6_conn_idx = 0;
4344 
4345 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4346 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4347 		connp = NULL;
4348 
4349 		while ((connp = ipcl_get_next_conn(connfp, connp,
4350 		    IPCL_UDPCONN))) {
4351 			udp = connp->conn_udp;
4352 			if (zoneid != connp->conn_zoneid)
4353 				continue;
4354 
4355 			/*
4356 			 * Note that the port numbers are sent in
4357 			 * host byte order
4358 			 */
4359 
4360 			if (udp->udp_state == TS_UNBND)
4361 				state = MIB2_UDP_unbound;
4362 			else if (udp->udp_state == TS_IDLE)
4363 				state = MIB2_UDP_idle;
4364 			else if (udp->udp_state == TS_DATA_XFER)
4365 				state = MIB2_UDP_connected;
4366 			else
4367 				state = MIB2_UDP_unknown;
4368 
4369 			needattr = B_FALSE;
4370 			bzero(&mlp, sizeof (mlp));
4371 			if (connp->conn_mlp_type != mlptSingle) {
4372 				if (connp->conn_mlp_type == mlptShared ||
4373 				    connp->conn_mlp_type == mlptBoth)
4374 					mlp.tme_flags |= MIB2_TMEF_SHARED;
4375 				if (connp->conn_mlp_type == mlptPrivate ||
4376 				    connp->conn_mlp_type == mlptBoth)
4377 					mlp.tme_flags |= MIB2_TMEF_PRIVATE;
4378 				needattr = B_TRUE;
4379 			}
4380 			if (connp->conn_anon_mlp) {
4381 				mlp.tme_flags |= MIB2_TMEF_ANONMLP;
4382 				needattr = B_TRUE;
4383 			}
4384 			if (connp->conn_mac_exempt) {
4385 				mlp.tme_flags |= MIB2_TMEF_MACEXEMPT;
4386 				needattr = B_TRUE;
4387 			}
4388 
4389 			/*
4390 			 * Create an IPv4 table entry for IPv4 entries and also
4391 			 * any IPv6 entries which are bound to in6addr_any
4392 			 * (i.e. anything a IPv4 peer could connect/send to).
4393 			 */
4394 			if (udp->udp_ipversion == IPV4_VERSION ||
4395 			    (udp->udp_state <= TS_IDLE &&
4396 			    IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) {
4397 				ude.udpEntryInfo.ue_state = state;
4398 				/*
4399 				 * If in6addr_any this will set it to
4400 				 * INADDR_ANY
4401 				 */
4402 				ude.udpLocalAddress =
4403 				    V4_PART_OF_V6(udp->udp_v6src);
4404 				ude.udpLocalPort = ntohs(udp->udp_port);
4405 				if (udp->udp_state == TS_DATA_XFER) {
4406 					/*
4407 					 * Can potentially get here for
4408 					 * v6 socket if another process
4409 					 * (say, ping) has just done a
4410 					 * sendto(), changing the state
4411 					 * from the TS_IDLE above to
4412 					 * TS_DATA_XFER by the time we hit
4413 					 * this part of the code.
4414 					 */
4415 					ude.udpEntryInfo.ue_RemoteAddress =
4416 					    V4_PART_OF_V6(udp->udp_v6dst);
4417 					ude.udpEntryInfo.ue_RemotePort =
4418 					    ntohs(udp->udp_dstport);
4419 				} else {
4420 					ude.udpEntryInfo.ue_RemoteAddress = 0;
4421 					ude.udpEntryInfo.ue_RemotePort = 0;
4422 				}
4423 
4424 				/*
4425 				 * We make the assumption that all udp_t
4426 				 * structs will be created within an address
4427 				 * region no larger than 32-bits.
4428 				 */
4429 				ude.udpInstance = (uint32_t)(uintptr_t)udp;
4430 				ude.udpCreationProcess =
4431 				    (udp->udp_open_pid < 0) ?
4432 				    MIB2_UNKNOWN_PROCESS :
4433 				    udp->udp_open_pid;
4434 				ude.udpCreationTime = udp->udp_open_time;
4435 
4436 				(void) snmp_append_data2(mp_conn_ctl->b_cont,
4437 				    &mp_conn_tail, (char *)&ude, sizeof (ude));
4438 				mlp.tme_connidx = v4_conn_idx++;
4439 				if (needattr)
4440 					(void) snmp_append_data2(
4441 					    mp_attr_ctl->b_cont, &mp_attr_tail,
4442 					    (char *)&mlp, sizeof (mlp));
4443 			}
4444 			if (udp->udp_ipversion == IPV6_VERSION) {
4445 				ude6.udp6EntryInfo.ue_state  = state;
4446 				ude6.udp6LocalAddress = udp->udp_v6src;
4447 				ude6.udp6LocalPort = ntohs(udp->udp_port);
4448 				ude6.udp6IfIndex = udp->udp_bound_if;
4449 				if (udp->udp_state == TS_DATA_XFER) {
4450 					ude6.udp6EntryInfo.ue_RemoteAddress =
4451 					    udp->udp_v6dst;
4452 					ude6.udp6EntryInfo.ue_RemotePort =
4453 					    ntohs(udp->udp_dstport);
4454 				} else {
4455 					ude6.udp6EntryInfo.ue_RemoteAddress =
4456 					    sin6_null.sin6_addr;
4457 					ude6.udp6EntryInfo.ue_RemotePort = 0;
4458 				}
4459 				/*
4460 				 * We make the assumption that all udp_t
4461 				 * structs will be created within an address
4462 				 * region no larger than 32-bits.
4463 				 */
4464 				ude6.udp6Instance = (uint32_t)(uintptr_t)udp;
4465 				ude6.udp6CreationProcess =
4466 				    (udp->udp_open_pid < 0) ?
4467 				    MIB2_UNKNOWN_PROCESS :
4468 				    udp->udp_open_pid;
4469 				ude6.udp6CreationTime = udp->udp_open_time;
4470 
4471 				(void) snmp_append_data2(mp6_conn_ctl->b_cont,
4472 				    &mp6_conn_tail, (char *)&ude6,
4473 				    sizeof (ude6));
4474 				mlp.tme_connidx = v6_conn_idx++;
4475 				if (needattr)
4476 					(void) snmp_append_data2(
4477 					    mp6_attr_ctl->b_cont,
4478 					    &mp6_attr_tail, (char *)&mlp,
4479 					    sizeof (mlp));
4480 			}
4481 		}
4482 	}
4483 
4484 	/* IPv4 UDP endpoints */
4485 	optp = (struct opthdr *)&mp_conn_ctl->b_rptr[
4486 	    sizeof (struct T_optmgmt_ack)];
4487 	optp->level = MIB2_UDP;
4488 	optp->name = MIB2_UDP_ENTRY;
4489 	optp->len = msgdsize(mp_conn_ctl->b_cont);
4490 	qreply(q, mp_conn_ctl);
4491 
4492 	/* table of MLP attributes... */
4493 	optp = (struct opthdr *)&mp_attr_ctl->b_rptr[
4494 	    sizeof (struct T_optmgmt_ack)];
4495 	optp->level = MIB2_UDP;
4496 	optp->name = EXPER_XPORT_MLP;
4497 	optp->len = msgdsize(mp_attr_ctl->b_cont);
4498 	if (optp->len == 0)
4499 		freemsg(mp_attr_ctl);
4500 	else
4501 		qreply(q, mp_attr_ctl);
4502 
4503 	/* IPv6 UDP endpoints */
4504 	optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[
4505 	    sizeof (struct T_optmgmt_ack)];
4506 	optp->level = MIB2_UDP6;
4507 	optp->name = MIB2_UDP6_ENTRY;
4508 	optp->len = msgdsize(mp6_conn_ctl->b_cont);
4509 	qreply(q, mp6_conn_ctl);
4510 
4511 	/* table of MLP attributes... */
4512 	optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[
4513 	    sizeof (struct T_optmgmt_ack)];
4514 	optp->level = MIB2_UDP6;
4515 	optp->name = EXPER_XPORT_MLP;
4516 	optp->len = msgdsize(mp6_attr_ctl->b_cont);
4517 	if (optp->len == 0)
4518 		freemsg(mp6_attr_ctl);
4519 	else
4520 		qreply(q, mp6_attr_ctl);
4521 
4522 	return (mp2ctl);
4523 }
4524 
4525 /*
4526  * Return 0 if invalid set request, 1 otherwise, including non-udp requests.
4527  * NOTE: Per MIB-II, UDP has no writable data.
4528  * TODO:  If this ever actually tries to set anything, it needs to be
4529  * to do the appropriate locking.
4530  */
4531 /* ARGSUSED */
4532 int
4533 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4534     uchar_t *ptr, int len)
4535 {
4536 	switch (level) {
4537 	case MIB2_UDP:
4538 		return (0);
4539 	default:
4540 		return (1);
4541 	}
4542 }
4543 
4544 /*
4545  * This routine creates a T_UDERROR_IND message and passes it upstream.
4546  * The address and options are copied from the T_UNITDATA_REQ message
4547  * passed in mp.  This message is freed.
4548  */
4549 static void
4550 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen,
4551     t_scalar_t err)
4552 {
4553 	struct T_unitdata_req *tudr;
4554 	mblk_t	*mp1;
4555 	uchar_t	*optaddr;
4556 	t_scalar_t optlen;
4557 
4558 	if (DB_TYPE(mp) == M_DATA) {
4559 		ASSERT(destaddr != NULL && destlen != 0);
4560 		optaddr = NULL;
4561 		optlen = 0;
4562 	} else {
4563 		if ((mp->b_wptr < mp->b_rptr) ||
4564 		    (MBLKL(mp)) < sizeof (struct T_unitdata_req)) {
4565 			goto done;
4566 		}
4567 		tudr = (struct T_unitdata_req *)mp->b_rptr;
4568 		destaddr = mp->b_rptr + tudr->DEST_offset;
4569 		if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr ||
4570 		    destaddr + tudr->DEST_length < mp->b_rptr ||
4571 		    destaddr + tudr->DEST_length > mp->b_wptr) {
4572 			goto done;
4573 		}
4574 		optaddr = mp->b_rptr + tudr->OPT_offset;
4575 		if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr ||
4576 		    optaddr + tudr->OPT_length < mp->b_rptr ||
4577 		    optaddr + tudr->OPT_length > mp->b_wptr) {
4578 			goto done;
4579 		}
4580 		destlen = tudr->DEST_length;
4581 		optlen = tudr->OPT_length;
4582 	}
4583 
4584 	mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen,
4585 	    (char *)optaddr, optlen, err);
4586 	if (mp1 != NULL)
4587 		qreply(q, mp1);
4588 
4589 done:
4590 	freemsg(mp);
4591 }
4592 
4593 /*
4594  * This routine removes a port number association from a stream.  It
4595  * is called by udp_wput to handle T_UNBIND_REQ messages.
4596  */
4597 static void
4598 udp_tpi_unbind(queue_t *q, mblk_t *mp)
4599 {
4600 	conn_t	*connp = Q_TO_CONN(q);
4601 	int	error;
4602 
4603 	error = udp_do_unbind(connp);
4604 	if (error) {
4605 		if (error < 0)
4606 			udp_err_ack(q, mp, -error, 0);
4607 		else
4608 			udp_err_ack(q, mp, TSYSERR, error);
4609 		return;
4610 	}
4611 
4612 	mp = mi_tpi_ok_ack_alloc(mp);
4613 	ASSERT(mp != NULL);
4614 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4615 	qreply(q, mp);
4616 }
4617 
4618 /*
4619  * Don't let port fall into the privileged range.
4620  * Since the extra privileged ports can be arbitrary we also
4621  * ensure that we exclude those from consideration.
4622  * us->us_epriv_ports is not sorted thus we loop over it until
4623  * there are no changes.
4624  */
4625 static in_port_t
4626 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random)
4627 {
4628 	int i;
4629 	in_port_t nextport;
4630 	boolean_t restart = B_FALSE;
4631 	udp_stack_t *us = udp->udp_us;
4632 
4633 	if (random && udp_random_anon_port != 0) {
4634 		(void) random_get_pseudo_bytes((uint8_t *)&port,
4635 		    sizeof (in_port_t));
4636 		/*
4637 		 * Unless changed by a sys admin, the smallest anon port
4638 		 * is 32768 and the largest anon port is 65535.  It is
4639 		 * very likely (50%) for the random port to be smaller
4640 		 * than the smallest anon port.  When that happens,
4641 		 * add port % (anon port range) to the smallest anon
4642 		 * port to get the random port.  It should fall into the
4643 		 * valid anon port range.
4644 		 */
4645 		if (port < us->us_smallest_anon_port) {
4646 			port = us->us_smallest_anon_port +
4647 			    port % (us->us_largest_anon_port -
4648 			    us->us_smallest_anon_port);
4649 		}
4650 	}
4651 
4652 retry:
4653 	if (port < us->us_smallest_anon_port)
4654 		port = us->us_smallest_anon_port;
4655 
4656 	if (port > us->us_largest_anon_port) {
4657 		port = us->us_smallest_anon_port;
4658 		if (restart)
4659 			return (0);
4660 		restart = B_TRUE;
4661 	}
4662 
4663 	if (port < us->us_smallest_nonpriv_port)
4664 		port = us->us_smallest_nonpriv_port;
4665 
4666 	for (i = 0; i < us->us_num_epriv_ports; i++) {
4667 		if (port == us->us_epriv_ports[i]) {
4668 			port++;
4669 			/*
4670 			 * Make sure that the port is in the
4671 			 * valid range.
4672 			 */
4673 			goto retry;
4674 		}
4675 	}
4676 
4677 	if (is_system_labeled() &&
4678 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
4679 	    port, IPPROTO_UDP, B_TRUE)) != 0) {
4680 		port = nextport;
4681 		goto retry;
4682 	}
4683 
4684 	return (port);
4685 }
4686 
4687 static int
4688 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst)
4689 {
4690 	int err;
4691 	cred_t *cred;
4692 	cred_t *orig_cred = NULL;
4693 	cred_t *effective_cred = NULL;
4694 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4695 	udp_t *udp = Q_TO_UDP(wq);
4696 	udp_stack_t	*us = udp->udp_us;
4697 
4698 	/*
4699 	 * All Solaris components should pass a db_credp
4700 	 * for this message, hence we ASSERT.
4701 	 * On production kernels we return an error to be robust against
4702 	 * random streams modules sitting on top of us.
4703 	 */
4704 	cred = orig_cred = msg_getcred(mp, NULL);
4705 	ASSERT(cred != NULL);
4706 	if (cred == NULL)
4707 		return (EINVAL);
4708 
4709 	/*
4710 	 * Verify the destination is allowed to receive packets at
4711 	 * the security label of the message data. tsol_check_dest()
4712 	 * may create a new effective cred for this message with a
4713 	 * modified label or label flags. Note that we use the cred/label
4714 	 * from the message to handle MLP
4715 	 */
4716 	if ((err = tsol_check_dest(cred, &dst, IPV4_VERSION,
4717 	    udp->udp_connp->conn_mac_exempt, &effective_cred)) != 0)
4718 		goto done;
4719 	if (effective_cred != NULL)
4720 		cred = effective_cred;
4721 
4722 	/*
4723 	 * Calculate the security label to be placed in the text
4724 	 * of the message (if any).
4725 	 */
4726 	if ((err = tsol_compute_label(cred, dst, opt_storage,
4727 	    us->us_netstack->netstack_ip)) != 0)
4728 		goto done;
4729 
4730 	/*
4731 	 * Insert the security label in the cached ip options,
4732 	 * removing any old label that may exist.
4733 	 */
4734 	if ((err = tsol_update_options(&udp->udp_ip_snd_options,
4735 	    &udp->udp_ip_snd_options_len, &udp->udp_label_len,
4736 	    opt_storage)) != 0)
4737 		goto done;
4738 
4739 	/*
4740 	 * Save the destination address and creds we used to
4741 	 * generate the security label text.
4742 	 */
4743 	if (cred != udp->udp_effective_cred) {
4744 		if (udp->udp_effective_cred != NULL)
4745 			crfree(udp->udp_effective_cred);
4746 		crhold(cred);
4747 		udp->udp_effective_cred = cred;
4748 	}
4749 	if (orig_cred != udp->udp_last_cred) {
4750 		if (udp->udp_last_cred != NULL)
4751 			crfree(udp->udp_last_cred);
4752 		crhold(orig_cred);
4753 		udp->udp_last_cred = orig_cred;
4754 	}
4755 done:
4756 	if (effective_cred != NULL)
4757 		crfree(effective_cred);
4758 
4759 	if (err != 0) {
4760 		DTRACE_PROBE4(
4761 		    tx__ip__log__info__updatelabel__udp,
4762 		    char *, "queue(1) failed to update options(2) on mp(3)",
4763 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4764 	}
4765 	return (err);
4766 }
4767 
4768 static mblk_t *
4769 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port,
4770     uint_t srcid, int *error, boolean_t insert_spi, struct nmsghdr *msg,
4771     cred_t *cr, pid_t pid)
4772 {
4773 	udp_t		*udp = connp->conn_udp;
4774 	mblk_t		*mp1 = mp;
4775 	mblk_t		*mp2;
4776 	ipha_t		*ipha;
4777 	int		ip_hdr_length;
4778 	uint32_t 	ip_len;
4779 	udpha_t		*udpha;
4780 	boolean_t 	lock_held = B_FALSE;
4781 	in_port_t	uha_src_port;
4782 	udpattrs_t	attrs;
4783 	uchar_t		ip_snd_opt[IP_MAX_OPT_LENGTH];
4784 	uint32_t	ip_snd_opt_len = 0;
4785 	ip4_pkt_t  	pktinfo;
4786 	ip4_pkt_t  	*pktinfop = &pktinfo;
4787 	ip_opt_info_t	optinfo;
4788 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
4789 	udp_stack_t	*us = udp->udp_us;
4790 	ipsec_stack_t	*ipss = ipst->ips_netstack->netstack_ipsec;
4791 	queue_t		*q = connp->conn_wq;
4792 	ire_t		*ire;
4793 	in6_addr_t	v6dst;
4794 	boolean_t	update_lastdst = B_FALSE;
4795 
4796 	*error = 0;
4797 	pktinfop->ip4_ill_index = 0;
4798 	pktinfop->ip4_addr = INADDR_ANY;
4799 	optinfo.ip_opt_flags = 0;
4800 	optinfo.ip_opt_ill_index = 0;
4801 
4802 	if (v4dst == INADDR_ANY)
4803 		v4dst = htonl(INADDR_LOOPBACK);
4804 
4805 	/*
4806 	 * If options passed in, feed it for verification and handling
4807 	 */
4808 	attrs.udpattr_credset = B_FALSE;
4809 	if (IPCL_IS_NONSTR(connp)) {
4810 		if (msg->msg_controllen != 0) {
4811 			attrs.udpattr_ipp4 = pktinfop;
4812 			attrs.udpattr_mb = mp;
4813 
4814 			rw_enter(&udp->udp_rwlock, RW_WRITER);
4815 			*error = process_auxiliary_options(connp,
4816 			    msg->msg_control, msg->msg_controllen,
4817 			    &attrs, &udp_opt_obj, udp_opt_set, cr);
4818 			rw_exit(&udp->udp_rwlock);
4819 			if (*error)
4820 				goto done;
4821 		}
4822 	} else {
4823 		if (DB_TYPE(mp) != M_DATA) {
4824 			mp1 = mp->b_cont;
4825 			if (((struct T_unitdata_req *)
4826 			    mp->b_rptr)->OPT_length != 0) {
4827 				attrs.udpattr_ipp4 = pktinfop;
4828 				attrs.udpattr_mb = mp;
4829 				if (udp_unitdata_opt_process(q, mp, error,
4830 				    &attrs) < 0)
4831 					goto done;
4832 				/*
4833 				 * Note: success in processing options.
4834 				 * mp option buffer represented by
4835 				 * OPT_length/offset now potentially modified
4836 				 * and contain option setting results
4837 				 */
4838 				ASSERT(*error == 0);
4839 			}
4840 		}
4841 	}
4842 
4843 	/* mp1 points to the M_DATA mblk carrying the packet */
4844 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
4845 
4846 	/*
4847 	 * Determine whether we need to mark the mblk with the user's
4848 	 * credentials.
4849 	 * If labeled then sockfs would have already done this.
4850 	 */
4851 	ASSERT(!is_system_labeled() || msg_getcred(mp, NULL) != NULL);
4852 
4853 	ire = connp->conn_ire_cache;
4854 	if (CLASSD(v4dst) || (ire == NULL) || (ire->ire_addr != v4dst) ||
4855 	    (ire->ire_type & (IRE_BROADCAST | IRE_LOCAL | IRE_LOOPBACK))) {
4856 		if (cr != NULL && msg_getcred(mp, NULL) == NULL)
4857 			mblk_setcred(mp, cr, pid);
4858 	}
4859 
4860 	rw_enter(&udp->udp_rwlock, RW_READER);
4861 	lock_held = B_TRUE;
4862 
4863 	/*
4864 	 * Cluster and TSOL note:
4865 	 *    udp.udp_v6lastdst		is shared by Cluster and TSOL
4866 	 *    udp.udp_lastdstport	is used by Cluster
4867 	 *
4868 	 * Both Cluster and TSOL need to update the dest addr and/or port.
4869 	 * Updating is done after both Cluster and TSOL checks, protected
4870 	 * by conn_lock.
4871 	 */
4872 	mutex_enter(&connp->conn_lock);
4873 
4874 	if (cl_inet_connect2 != NULL &&
4875 	    (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
4876 	    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
4877 	    udp->udp_lastdstport != port)) {
4878 		mutex_exit(&connp->conn_lock);
4879 		*error = 0;
4880 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
4881 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &v6dst, port, *error);
4882 		if (*error != 0) {
4883 			*error = EHOSTUNREACH;
4884 			goto done;
4885 		}
4886 		update_lastdst = B_TRUE;
4887 		mutex_enter(&connp->conn_lock);
4888 	}
4889 
4890 	/*
4891 	 * Check if our saved options are valid; update if not.
4892 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
4893 	 * to different destination may require different labels,
4894 	 * or worse, UDP packets to same IP address may require
4895 	 * different labels due to use of shared all-zones address.
4896 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
4897 	 * and ip_snd_options_len are consistent for the current
4898 	 * destination and are updated atomically.
4899 	 */
4900 	if (is_system_labeled()) {
4901 		cred_t	*credp;
4902 		pid_t	cpid;
4903 
4904 		/* Using UDP MLP requires SCM_UCRED from user */
4905 		if (connp->conn_mlp_type != mlptSingle &&
4906 		    !attrs.udpattr_credset) {
4907 			mutex_exit(&connp->conn_lock);
4908 			DTRACE_PROBE4(
4909 			    tx__ip__log__info__output__udp,
4910 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
4911 			    mblk_t *, mp, udpattrs_t *, &attrs, queue_t *, q);
4912 			*error = EINVAL;
4913 			goto done;
4914 		}
4915 		/*
4916 		 * Update label option for this UDP socket if
4917 		 * - the destination has changed,
4918 		 * - the UDP socket is MLP, or
4919 		 * - the cred attached to the mblk changed.
4920 		 */
4921 		credp = msg_getcred(mp, &cpid);
4922 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
4923 		    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
4924 		    connp->conn_mlp_type != mlptSingle ||
4925 		    credp != udp->udp_last_cred) {
4926 			if ((*error = udp_update_label(q, mp, v4dst)) != 0) {
4927 				mutex_exit(&connp->conn_lock);
4928 				goto done;
4929 			}
4930 			update_lastdst = B_TRUE;
4931 		}
4932 
4933 		/*
4934 		 * Attach the effective cred to the mblk to ensure future
4935 		 * routing decisions will be based on it's label.
4936 		 */
4937 		mblk_setcred(mp, udp->udp_effective_cred, cpid);
4938 	}
4939 	if (update_lastdst) {
4940 		IN6_IPADDR_TO_V4MAPPED(v4dst, &udp->udp_v6lastdst);
4941 		udp->udp_lastdstport = port;
4942 	}
4943 	if (udp->udp_ip_snd_options_len > 0) {
4944 		ip_snd_opt_len = udp->udp_ip_snd_options_len;
4945 		bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
4946 	}
4947 	mutex_exit(&connp->conn_lock);
4948 
4949 	/* Add an IP header */
4950 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len +
4951 	    (insert_spi ? sizeof (uint32_t) : 0);
4952 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
4953 	if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) ||
4954 	    !OK_32PTR(ipha)) {
4955 		mp2 = allocb(ip_hdr_length + us->us_wroff_extra, BPRI_LO);
4956 		if (mp2 == NULL) {
4957 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
4958 			    "udp_wput_end: q %p (%S)", q, "allocbfail2");
4959 			*error = ENOMEM;
4960 			goto done;
4961 		}
4962 		mp2->b_wptr = DB_LIM(mp2);
4963 		mp2->b_cont = mp1;
4964 		mp1 = mp2;
4965 		if (DB_TYPE(mp) != M_DATA)
4966 			mp->b_cont = mp1;
4967 		else
4968 			mp = mp1;
4969 		ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length);
4970 	}
4971 	ip_hdr_length -= (UDPH_SIZE + (insert_spi ? sizeof (uint32_t) : 0));
4972 #ifdef	_BIG_ENDIAN
4973 	/* Set version, header length, and tos */
4974 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4975 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4976 	    udp->udp_type_of_service);
4977 	/* Set ttl and protocol */
4978 	*(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP;
4979 #else
4980 	/* Set version, header length, and tos */
4981 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4982 	    ((udp->udp_type_of_service << 8) |
4983 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4984 	/* Set ttl and protocol */
4985 	*(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl;
4986 #endif
4987 	if (pktinfop->ip4_addr != INADDR_ANY) {
4988 		ipha->ipha_src = pktinfop->ip4_addr;
4989 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4990 	} else {
4991 		/*
4992 		 * Copy our address into the packet.  If this is zero,
4993 		 * first look at __sin6_src_id for a hint. If we leave the
4994 		 * source as INADDR_ANY then ip will fill in the real source
4995 		 * address.
4996 		 */
4997 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src);
4998 		if (srcid != 0 && ipha->ipha_src == INADDR_ANY) {
4999 			in6_addr_t v6src;
5000 
5001 			ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid,
5002 			    us->us_netstack);
5003 			IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src);
5004 		}
5005 	}
5006 	uha_src_port = udp->udp_port;
5007 	if (ip_hdr_length == IP_SIMPLE_HDR_LENGTH) {
5008 		rw_exit(&udp->udp_rwlock);
5009 		lock_held = B_FALSE;
5010 	}
5011 
5012 	if (pktinfop->ip4_ill_index != 0) {
5013 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
5014 	}
5015 
5016 	ipha->ipha_fragment_offset_and_flags = 0;
5017 	ipha->ipha_ident = 0;
5018 
5019 	mp1->b_rptr = (uchar_t *)ipha;
5020 
5021 	ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <=
5022 	    (uintptr_t)UINT_MAX);
5023 
5024 	/* Determine length of packet */
5025 	ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha);
5026 	if ((mp2 = mp1->b_cont) != NULL) {
5027 		do {
5028 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
5029 			ip_len += (uint32_t)MBLKL(mp2);
5030 		} while ((mp2 = mp2->b_cont) != NULL);
5031 	}
5032 	/*
5033 	 * If the size of the packet is greater than the maximum allowed by
5034 	 * ip, return an error. Passing this down could cause panics because
5035 	 * the size will have wrapped and be inconsistent with the msg size.
5036 	 */
5037 	if (ip_len > IP_MAXPACKET) {
5038 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5039 		    "udp_wput_end: q %p (%S)", q, "IP length exceeded");
5040 		*error = EMSGSIZE;
5041 		goto done;
5042 	}
5043 	ipha->ipha_length = htons((uint16_t)ip_len);
5044 	ip_len -= ip_hdr_length;
5045 	ip_len = htons((uint16_t)ip_len);
5046 	udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length);
5047 
5048 	/* Insert all-0s SPI now. */
5049 	if (insert_spi)
5050 		*((uint32_t *)(udpha + 1)) = 0;
5051 
5052 	/*
5053 	 * Copy in the destination address
5054 	 */
5055 	ipha->ipha_dst = v4dst;
5056 
5057 	/*
5058 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
5059 	 */
5060 	if (CLASSD(v4dst))
5061 		ipha->ipha_ttl = udp->udp_multicast_ttl;
5062 
5063 	udpha->uha_dst_port = port;
5064 	udpha->uha_src_port = uha_src_port;
5065 
5066 	if (ip_snd_opt_len > 0) {
5067 		uint32_t	cksum;
5068 
5069 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
5070 		lock_held = B_FALSE;
5071 		rw_exit(&udp->udp_rwlock);
5072 		/*
5073 		 * Massage source route putting first source route in ipha_dst.
5074 		 * Ignore the destination in T_unitdata_req.
5075 		 * Create a checksum adjustment for a source route, if any.
5076 		 */
5077 		cksum = ip_massage_options(ipha, us->us_netstack);
5078 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5079 		cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) +
5080 		    (ipha->ipha_dst & 0xFFFF);
5081 		if ((int)cksum < 0)
5082 			cksum--;
5083 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5084 		/*
5085 		 * IP does the checksum if uha_checksum is non-zero,
5086 		 * We make it easy for IP to include our pseudo header
5087 		 * by putting our length in uha_checksum.
5088 		 */
5089 		cksum += ip_len;
5090 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5091 		/* There might be a carry. */
5092 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5093 #ifdef _LITTLE_ENDIAN
5094 		if (us->us_do_checksum)
5095 			ip_len = (cksum << 16) | ip_len;
5096 #else
5097 		if (us->us_do_checksum)
5098 			ip_len = (ip_len << 16) | cksum;
5099 		else
5100 			ip_len <<= 16;
5101 #endif
5102 	} else {
5103 		/*
5104 		 * IP does the checksum if uha_checksum is non-zero,
5105 		 * We make it easy for IP to include our pseudo header
5106 		 * by putting our length in uha_checksum.
5107 		 */
5108 		if (us->us_do_checksum)
5109 			ip_len |= (ip_len << 16);
5110 #ifndef _LITTLE_ENDIAN
5111 		else
5112 			ip_len <<= 16;
5113 #endif
5114 	}
5115 	ASSERT(!lock_held);
5116 	/* Set UDP length and checksum */
5117 	*((uint32_t *)&udpha->uha_length) = ip_len;
5118 
5119 	if (DB_TYPE(mp) != M_DATA) {
5120 		cred_t *cr;
5121 		pid_t cpid;
5122 
5123 		/* Move any cred from the T_UNITDATA_REQ to the packet */
5124 		cr = msg_extractcred(mp, &cpid);
5125 		if (cr != NULL) {
5126 			if (mp1->b_datap->db_credp != NULL)
5127 				crfree(mp1->b_datap->db_credp);
5128 			mp1->b_datap->db_credp = cr;
5129 			mp1->b_datap->db_cpid = cpid;
5130 		}
5131 		ASSERT(mp != mp1);
5132 		freeb(mp);
5133 	}
5134 
5135 	/* mp has been consumed and we'll return success */
5136 	ASSERT(*error == 0);
5137 	mp = NULL;
5138 
5139 	/* We're done.  Pass the packet to ip. */
5140 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
5141 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5142 	    "udp_wput_end: q %p (%S)", q, "end");
5143 
5144 	if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 ||
5145 	    CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) ||
5146 	    connp->conn_dontroute ||
5147 	    connp->conn_outgoing_ill != NULL || optinfo.ip_opt_flags != 0 ||
5148 	    optinfo.ip_opt_ill_index != 0 ||
5149 	    ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION ||
5150 	    IPP_ENABLED(IPP_LOCAL_OUT, ipst) ||
5151 	    ipst->ips_ip_g_mrouter != NULL) {
5152 		UDP_STAT(us, udp_ip_send);
5153 		ip_output_options(connp, mp1, connp->conn_wq, IP_WPUT,
5154 		    &optinfo);
5155 	} else {
5156 		udp_send_data(udp, connp->conn_wq, mp1, ipha);
5157 	}
5158 
5159 done:
5160 	if (lock_held)
5161 		rw_exit(&udp->udp_rwlock);
5162 	if (*error != 0) {
5163 		ASSERT(mp != NULL);
5164 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5165 	}
5166 	return (mp);
5167 }
5168 
5169 static void
5170 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha)
5171 {
5172 	conn_t	*connp = udp->udp_connp;
5173 	ipaddr_t src, dst;
5174 	ire_t	*ire;
5175 	ipif_t	*ipif = NULL;
5176 	mblk_t	*ire_fp_mp;
5177 	boolean_t retry_caching;
5178 	udp_stack_t *us = udp->udp_us;
5179 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5180 
5181 	dst = ipha->ipha_dst;
5182 	src = ipha->ipha_src;
5183 	ASSERT(ipha->ipha_ident == 0);
5184 
5185 	if (CLASSD(dst)) {
5186 		int err;
5187 
5188 		ipif = conn_get_held_ipif(connp,
5189 		    &connp->conn_multicast_ipif, &err);
5190 
5191 		if (ipif == NULL || ipif->ipif_isv6 ||
5192 		    (ipif->ipif_ill->ill_phyint->phyint_flags &
5193 		    PHYI_LOOPBACK)) {
5194 			if (ipif != NULL)
5195 				ipif_refrele(ipif);
5196 			UDP_STAT(us, udp_ip_send);
5197 			ip_output(connp, mp, q, IP_WPUT);
5198 			return;
5199 		}
5200 	}
5201 
5202 	retry_caching = B_FALSE;
5203 	mutex_enter(&connp->conn_lock);
5204 	ire = connp->conn_ire_cache;
5205 	ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT));
5206 
5207 	if (ire == NULL || ire->ire_addr != dst ||
5208 	    (ire->ire_marks & IRE_MARK_CONDEMNED)) {
5209 		retry_caching = B_TRUE;
5210 	} else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) {
5211 		ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr;
5212 
5213 		ASSERT(ipif != NULL);
5214 		if (!IS_ON_SAME_LAN(stq_ill, ipif->ipif_ill))
5215 			retry_caching = B_TRUE;
5216 	}
5217 
5218 	if (!retry_caching) {
5219 		ASSERT(ire != NULL);
5220 		IRE_REFHOLD(ire);
5221 		mutex_exit(&connp->conn_lock);
5222 	} else {
5223 		boolean_t cached = B_FALSE;
5224 
5225 		connp->conn_ire_cache = NULL;
5226 		mutex_exit(&connp->conn_lock);
5227 
5228 		/* Release the old ire */
5229 		if (ire != NULL) {
5230 			IRE_REFRELE_NOTR(ire);
5231 			ire = NULL;
5232 		}
5233 
5234 		if (CLASSD(dst)) {
5235 			ASSERT(ipif != NULL);
5236 			ire = ire_ctable_lookup(dst, 0, 0, ipif,
5237 			    connp->conn_zoneid, msg_getlabel(mp),
5238 			    MATCH_IRE_ILL, ipst);
5239 		} else {
5240 			ASSERT(ipif == NULL);
5241 			ire = ire_cache_lookup(dst, connp->conn_zoneid,
5242 			    msg_getlabel(mp), ipst);
5243 		}
5244 
5245 		if (ire == NULL) {
5246 			if (ipif != NULL)
5247 				ipif_refrele(ipif);
5248 			UDP_STAT(us, udp_ire_null);
5249 			ip_output(connp, mp, q, IP_WPUT);
5250 			return;
5251 		}
5252 		IRE_REFHOLD_NOTR(ire);
5253 
5254 		mutex_enter(&connp->conn_lock);
5255 		if (CONN_CACHE_IRE(connp) && connp->conn_ire_cache == NULL &&
5256 		    !(ire->ire_marks & IRE_MARK_CONDEMNED)) {
5257 			irb_t		*irb = ire->ire_bucket;
5258 
5259 			/*
5260 			 * IRE's created for non-connection oriented transports
5261 			 * are normally initialized with IRE_MARK_TEMPORARY set
5262 			 * in the ire_marks. These IRE's are preferentially
5263 			 * reaped when the hash chain length in the cache
5264 			 * bucket exceeds the maximum value specified in
5265 			 * ip[6]_ire_max_bucket_cnt. This can severely affect
5266 			 * UDP performance if IRE cache entries that we need
5267 			 * to reuse are continually removed. To remedy this,
5268 			 * when we cache the IRE in the conn_t, we remove the
5269 			 * IRE_MARK_TEMPORARY bit from the ire_marks if it was
5270 			 * set.
5271 			 */
5272 			if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5273 				rw_enter(&irb->irb_lock, RW_WRITER);
5274 				if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5275 					ire->ire_marks &= ~IRE_MARK_TEMPORARY;
5276 					irb->irb_tmp_ire_cnt--;
5277 				}
5278 				rw_exit(&irb->irb_lock);
5279 			}
5280 			connp->conn_ire_cache = ire;
5281 			cached = B_TRUE;
5282 		}
5283 		mutex_exit(&connp->conn_lock);
5284 
5285 		/*
5286 		 * We can continue to use the ire but since it was not
5287 		 * cached, we should drop the extra reference.
5288 		 */
5289 		if (!cached)
5290 			IRE_REFRELE_NOTR(ire);
5291 	}
5292 	ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION);
5293 	ASSERT(!CLASSD(dst) || ipif != NULL);
5294 
5295 	/*
5296 	 * Check if we can take the fast-path.
5297 	 * Note that "incomplete" ire's (where the link-layer for next hop
5298 	 * is not resolved, or where the fast-path header in nce_fp_mp is not
5299 	 * available yet) are sent down the legacy (slow) path
5300 	 */
5301 	if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) ||
5302 	    (ire->ire_flags & RTF_MULTIRT) || (ire->ire_stq == NULL) ||
5303 	    (ire->ire_max_frag < ntohs(ipha->ipha_length)) ||
5304 	    ((ire->ire_nce == NULL) ||
5305 	    ((ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL)) ||
5306 	    connp->conn_nexthop_set || (MBLKL(ire_fp_mp) > MBLKHEAD(mp))) {
5307 		if (ipif != NULL)
5308 			ipif_refrele(ipif);
5309 		UDP_STAT(us, udp_ip_ire_send);
5310 		IRE_REFRELE(ire);
5311 		ip_output(connp, mp, q, IP_WPUT);
5312 		return;
5313 	}
5314 
5315 	if (src == INADDR_ANY && !connp->conn_unspec_src) {
5316 		if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC))
5317 			ipha->ipha_src = ipif->ipif_src_addr;
5318 		else
5319 			ipha->ipha_src = ire->ire_src_addr;
5320 	}
5321 
5322 	if (ipif != NULL)
5323 		ipif_refrele(ipif);
5324 
5325 	udp_xmit(connp->conn_wq, mp, ire, connp, connp->conn_zoneid);
5326 }
5327 
5328 static void
5329 udp_xmit(queue_t *q, mblk_t *mp, ire_t *ire, conn_t *connp, zoneid_t zoneid)
5330 {
5331 	ipaddr_t src, dst;
5332 	ill_t	*ill;
5333 	mblk_t	*ire_fp_mp;
5334 	uint_t	ire_fp_mp_len;
5335 	uint16_t *up;
5336 	uint32_t cksum, hcksum_txflags;
5337 	queue_t	*dev_q;
5338 	udp_t	*udp = connp->conn_udp;
5339 	ipha_t	*ipha = (ipha_t *)mp->b_rptr;
5340 	udp_stack_t	*us = udp->udp_us;
5341 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5342 	boolean_t	ll_multicast = B_FALSE;
5343 	boolean_t	direct_send;
5344 
5345 	dev_q = ire->ire_stq->q_next;
5346 	ASSERT(dev_q != NULL);
5347 
5348 	ill = ire_to_ill(ire);
5349 	ASSERT(ill != NULL);
5350 
5351 	/*
5352 	 * For the direct send case, if resetting of conn_direct_blocked
5353 	 * was missed, it is still ok because the putq() would enable
5354 	 * the queue and write service will drain it out.
5355 	 */
5356 	direct_send = ILL_DIRECT_CAPABLE(ill);
5357 
5358 	/* is queue flow controlled? */
5359 	if ((!direct_send) && (q->q_first != NULL || connp->conn_draining ||
5360 	    DEV_Q_FLOW_BLOCKED(dev_q))) {
5361 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
5362 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
5363 		if (ipst->ips_ip_output_queue) {
5364 			DTRACE_PROBE1(udp__xmit__putq, conn_t *, connp);
5365 			(void) putq(connp->conn_wq, mp);
5366 		} else {
5367 			freemsg(mp);
5368 		}
5369 		ire_refrele(ire);
5370 		return;
5371 	}
5372 
5373 	ire_fp_mp = ire->ire_nce->nce_fp_mp;
5374 	ire_fp_mp_len = MBLKL(ire_fp_mp);
5375 	ASSERT(MBLKHEAD(mp) >= ire_fp_mp_len);
5376 
5377 	dst = ipha->ipha_dst;
5378 	src = ipha->ipha_src;
5379 
5380 
5381 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
5382 
5383 	ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1);
5384 #ifndef _BIG_ENDIAN
5385 	ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8);
5386 #endif
5387 
5388 	if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
5389 		ASSERT(ill->ill_hcksum_capab != NULL);
5390 		hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags;
5391 	} else {
5392 		hcksum_txflags = 0;
5393 	}
5394 
5395 	/* pseudo-header checksum (do it in parts for IP header checksum) */
5396 	cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF);
5397 
5398 	ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION);
5399 	up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH);
5400 	if (*up != 0) {
5401 		IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags,
5402 		    mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH,
5403 		    ntohs(ipha->ipha_length), cksum);
5404 
5405 		/* Software checksum? */
5406 		if (DB_CKSUMFLAGS(mp) == 0) {
5407 			UDP_STAT(us, udp_out_sw_cksum);
5408 			UDP_STAT_UPDATE(us, udp_out_sw_cksum_bytes,
5409 			    ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH);
5410 		}
5411 	}
5412 
5413 	if (!CLASSD(dst)) {
5414 		ipha->ipha_fragment_offset_and_flags |=
5415 		    (uint32_t)htons(ire->ire_frag_flag);
5416 	}
5417 
5418 	/* Calculate IP header checksum if hardware isn't capable */
5419 	if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) {
5420 		IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0],
5421 		    ((uint16_t *)ipha)[4]);
5422 	}
5423 
5424 	if (CLASSD(dst)) {
5425 		if (ilm_lookup_ill(ill, dst, ALL_ZONES) != NULL) {
5426 			ip_multicast_loopback(q, ill, mp,
5427 			    connp->conn_multicast_loop ? 0 :
5428 			    IP_FF_NO_MCAST_LOOP, zoneid);
5429 		}
5430 
5431 		/* If multicast TTL is 0 then we are done */
5432 		if (ipha->ipha_ttl == 0) {
5433 			freemsg(mp);
5434 			ire_refrele(ire);
5435 			return;
5436 		}
5437 		ll_multicast = B_TRUE;
5438 	}
5439 
5440 	ASSERT(DB_TYPE(ire_fp_mp) == M_DATA);
5441 	mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len;
5442 	bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len);
5443 
5444 	UPDATE_OB_PKT_COUNT(ire);
5445 	ire->ire_last_used_time = lbolt;
5446 
5447 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
5448 	UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
5449 	    ntohs(ipha->ipha_length));
5450 
5451 	DTRACE_PROBE4(ip4__physical__out__start,
5452 	    ill_t *, NULL, ill_t *, ill, ipha_t *, ipha, mblk_t *, mp);
5453 	FW_HOOKS(ipst->ips_ip4_physical_out_event,
5454 	    ipst->ips_ipv4firewall_physical_out, NULL, ill, ipha, mp, mp,
5455 	    ll_multicast, ipst);
5456 	DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);
5457 	if (ipst->ips_ipobs_enabled && mp != NULL) {
5458 		zoneid_t szone;
5459 
5460 		szone = ip_get_zoneid_v4(ipha->ipha_src, mp,
5461 		    ipst, ALL_ZONES);
5462 		ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone,
5463 		    ALL_ZONES, ill, IPV4_VERSION, ire_fp_mp_len, ipst);
5464 	}
5465 
5466 	if (mp == NULL)
5467 		goto bail;
5468 
5469 	DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
5470 	    void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
5471 	    ipha_t *, ipha, ip6_t *, NULL, int, 0);
5472 
5473 	if (direct_send) {
5474 		uintptr_t cookie;
5475 		ill_dld_direct_t *idd = &ill->ill_dld_capab->idc_direct;
5476 
5477 		cookie = idd->idd_tx_df(idd->idd_tx_dh, mp,
5478 		    (uintptr_t)connp, 0);
5479 		if (cookie != NULL) {
5480 			idl_tx_list_t *idl_txl;
5481 
5482 			/*
5483 			 * Flow controlled.
5484 			 */
5485 			DTRACE_PROBE2(non__null__cookie, uintptr_t,
5486 			    cookie, conn_t *, connp);
5487 			idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
5488 			mutex_enter(&idl_txl->txl_lock);
5489 			/*
5490 			 * Check again after holding txl_lock to see if Tx
5491 			 * ring is still blocked and only then insert the
5492 			 * connp into the drain list.
5493 			 */
5494 			if (connp->conn_direct_blocked ||
5495 			    (idd->idd_tx_fctl_df(idd->idd_tx_fctl_dh,
5496 			    cookie) == 0)) {
5497 				mutex_exit(&idl_txl->txl_lock);
5498 				goto bail;
5499 			}
5500 			if (idl_txl->txl_cookie != NULL &&
5501 			    idl_txl->txl_cookie != cookie) {
5502 				DTRACE_PROBE2(udp__xmit__collision,
5503 				    uintptr_t, cookie,
5504 				    uintptr_t, idl_txl->txl_cookie);
5505 				UDP_STAT(us, udp_cookie_coll);
5506 			} else {
5507 				connp->conn_direct_blocked = B_TRUE;
5508 				idl_txl->txl_cookie = cookie;
5509 				conn_drain_insert(connp, idl_txl);
5510 				DTRACE_PROBE1(udp__xmit__insert,
5511 				    conn_t *, connp);
5512 			}
5513 			mutex_exit(&idl_txl->txl_lock);
5514 		}
5515 	} else {
5516 		DTRACE_PROBE1(udp__xmit__putnext, mblk_t *, mp);
5517 		putnext(ire->ire_stq, mp);
5518 	}
5519 bail:
5520 	IRE_REFRELE(ire);
5521 }
5522 
5523 static boolean_t
5524 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst)
5525 {
5526 	udp_t *udp = Q_TO_UDP(wq);
5527 	int err;
5528 	cred_t *cred;
5529 	cred_t *orig_cred;
5530 	cred_t *effective_cred = NULL;
5531 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
5532 	udp_stack_t		*us = udp->udp_us;
5533 
5534 	/*
5535 	 * All Solaris components should pass a db_credp
5536 	 * for this message, hence we ASSERT.
5537 	 * On production kernels we return an error to be robust against
5538 	 * random streams modules sitting on top of us.
5539 	 */
5540 	cred = orig_cred = msg_getcred(mp, NULL);
5541 	ASSERT(cred != NULL);
5542 	if (cred == NULL)
5543 		return (EINVAL);
5544 
5545 	/*
5546 	 * Verify the destination is allowed to receive packets at
5547 	 * the security label of the message data. tsol_check_dest()
5548 	 * may create a new effective cred for this message with a
5549 	 * modified label or label flags. Note that we use the
5550 	 * cred/label from the message to handle MLP.
5551 	 */
5552 	if ((err = tsol_check_dest(cred, dst, IPV6_VERSION,
5553 	    udp->udp_connp->conn_mac_exempt, &effective_cred)) != 0)
5554 		goto done;
5555 	if (effective_cred != NULL)
5556 		cred = effective_cred;
5557 
5558 	/*
5559 	 * Calculate the security label to be placed in the text
5560 	 * of the message (if any).
5561 	 */
5562 	if ((err = tsol_compute_label_v6(cred, dst, opt_storage,
5563 	    us->us_netstack->netstack_ip)) != 0)
5564 		goto done;
5565 
5566 	/*
5567 	 * Insert the security label in the cached ip options,
5568 	 * removing any old label that may exist.
5569 	 */
5570 	if ((err = tsol_update_sticky(&udp->udp_sticky_ipp,
5571 	    &udp->udp_label_len_v6, opt_storage)) != 0)
5572 		goto done;
5573 
5574 	/*
5575 	 * Save the destination address and cred we used to
5576 	 * generate the security label text.
5577 	 */
5578 	if (cred != udp->udp_effective_cred) {
5579 		if (udp->udp_effective_cred != NULL)
5580 			crfree(udp->udp_effective_cred);
5581 		crhold(cred);
5582 		udp->udp_effective_cred = cred;
5583 	}
5584 	if (orig_cred != udp->udp_last_cred) {
5585 		if (udp->udp_last_cred != NULL)
5586 			crfree(udp->udp_last_cred);
5587 		crhold(orig_cred);
5588 		udp->udp_last_cred = orig_cred;
5589 	}
5590 
5591 done:
5592 	if (effective_cred != NULL)
5593 		crfree(effective_cred);
5594 
5595 	if (err != 0) {
5596 		DTRACE_PROBE4(
5597 		    tx__ip__log__drop__updatelabel__udp6,
5598 		    char *, "queue(1) failed to update options(2) on mp(3)",
5599 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
5600 	}
5601 	return (err);
5602 }
5603 
5604 static int
5605 udp_send_connected(conn_t *connp, mblk_t *mp, struct nmsghdr *msg, cred_t *cr,
5606     pid_t pid)
5607 {
5608 	udp_t		*udp = connp->conn_udp;
5609 	udp_stack_t	*us = udp->udp_us;
5610 	ipaddr_t	v4dst;
5611 	in_port_t	dstport;
5612 	boolean_t	mapped_addr;
5613 	struct sockaddr_storage ss;
5614 	sin_t		*sin;
5615 	sin6_t		*sin6;
5616 	struct sockaddr	*addr;
5617 	socklen_t	addrlen;
5618 	int		error;
5619 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5620 
5621 	/* M_DATA for connected socket */
5622 
5623 	ASSERT(udp->udp_issocket);
5624 	UDP_DBGSTAT(us, udp_data_conn);
5625 
5626 	mutex_enter(&connp->conn_lock);
5627 	if (udp->udp_state != TS_DATA_XFER) {
5628 		mutex_exit(&connp->conn_lock);
5629 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5630 		UDP_STAT(us, udp_out_err_notconn);
5631 		freemsg(mp);
5632 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5633 		    "udp_wput_end: connp %p (%S)", connp,
5634 		    "not-connected; address required");
5635 		return (EDESTADDRREQ);
5636 	}
5637 
5638 	mapped_addr = IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst);
5639 	if (mapped_addr)
5640 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst);
5641 
5642 	/* Initialize addr and addrlen as if they're passed in */
5643 	if (udp->udp_family == AF_INET) {
5644 		sin = (sin_t *)&ss;
5645 		sin->sin_family = AF_INET;
5646 		dstport = sin->sin_port = udp->udp_dstport;
5647 		ASSERT(mapped_addr);
5648 		sin->sin_addr.s_addr = v4dst;
5649 		addr = (struct sockaddr *)sin;
5650 		addrlen = sizeof (*sin);
5651 	} else {
5652 		sin6 = (sin6_t *)&ss;
5653 		sin6->sin6_family = AF_INET6;
5654 		dstport = sin6->sin6_port = udp->udp_dstport;
5655 		sin6->sin6_flowinfo = udp->udp_flowinfo;
5656 		sin6->sin6_addr = udp->udp_v6dst;
5657 		sin6->sin6_scope_id = 0;
5658 		sin6->__sin6_src_id = 0;
5659 		addr = (struct sockaddr *)sin6;
5660 		addrlen = sizeof (*sin6);
5661 	}
5662 	mutex_exit(&connp->conn_lock);
5663 
5664 	if (mapped_addr) {
5665 		/*
5666 		 * Handle both AF_INET and AF_INET6; the latter
5667 		 * for IPV4 mapped destination addresses.  Note
5668 		 * here that both addr and addrlen point to the
5669 		 * corresponding struct depending on the address
5670 		 * family of the socket.
5671 		 */
5672 		mp = udp_output_v4(connp, mp, v4dst, dstport, 0, &error,
5673 		    insert_spi, msg, cr, pid);
5674 	} else {
5675 		mp = udp_output_v6(connp, mp, sin6, &error, msg, cr, pid);
5676 	}
5677 	if (error == 0) {
5678 		ASSERT(mp == NULL);
5679 		return (0);
5680 	}
5681 
5682 	UDP_STAT(us, udp_out_err_output);
5683 	ASSERT(mp != NULL);
5684 	if (IPCL_IS_NONSTR(connp)) {
5685 		freemsg(mp);
5686 		return (error);
5687 	} else {
5688 		/* mp is freed by the following routine */
5689 		udp_ud_err(connp->conn_wq, mp, (uchar_t *)addr,
5690 		    (t_scalar_t)addrlen, (t_scalar_t)error);
5691 		return (0);
5692 	}
5693 }
5694 
5695 /* ARGSUSED */
5696 static int
5697 udp_send_not_connected(conn_t *connp,  mblk_t *mp, struct sockaddr *addr,
5698     socklen_t addrlen, struct nmsghdr *msg, cred_t *cr, pid_t pid)
5699 {
5700 
5701 	udp_t		*udp = connp->conn_udp;
5702 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5703 	int		error = 0;
5704 	sin6_t		*sin6;
5705 	sin_t		*sin;
5706 	uint_t		srcid;
5707 	uint16_t	port;
5708 	ipaddr_t	v4dst;
5709 
5710 
5711 	ASSERT(addr != NULL);
5712 
5713 	switch (udp->udp_family) {
5714 	case AF_INET6:
5715 		sin6 = (sin6_t *)addr;
5716 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
5717 			/*
5718 			 * Destination is a non-IPv4-compatible IPv6 address.
5719 			 * Send out an IPv6 format packet.
5720 			 */
5721 			mp = udp_output_v6(connp, mp, sin6, &error, msg, cr,
5722 			    pid);
5723 			if (error != 0)
5724 				goto ud_error;
5725 
5726 			return (0);
5727 		}
5728 		/*
5729 		 * If the local address is not zero or a mapped address
5730 		 * return an error.  It would be possible to send an IPv4
5731 		 * packet but the response would never make it back to the
5732 		 * application since it is bound to a non-mapped address.
5733 		 */
5734 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) &&
5735 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
5736 			error = EADDRNOTAVAIL;
5737 			goto ud_error;
5738 		}
5739 		/* Send IPv4 packet without modifying udp_ipversion */
5740 		/* Extract port and ipaddr */
5741 		port = sin6->sin6_port;
5742 		IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst);
5743 		srcid = sin6->__sin6_src_id;
5744 		break;
5745 
5746 	case AF_INET:
5747 		sin = (sin_t *)addr;
5748 		/* Extract port and ipaddr */
5749 		port = sin->sin_port;
5750 		v4dst = sin->sin_addr.s_addr;
5751 		srcid = 0;
5752 		break;
5753 	}
5754 
5755 	mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error, insert_spi,
5756 	    msg, cr, pid);
5757 
5758 	if (error == 0) {
5759 		ASSERT(mp == NULL);
5760 		return (0);
5761 	}
5762 
5763 ud_error:
5764 	ASSERT(mp != NULL);
5765 
5766 	return (error);
5767 }
5768 
5769 /*
5770  * This routine handles all messages passed downstream.  It either
5771  * consumes the message or passes it downstream; it never queues a
5772  * a message.
5773  *
5774  * Also entry point for sockfs when udp is in "direct sockfs" mode.  This mode
5775  * is valid when we are directly beneath the stream head, and thus sockfs
5776  * is able to bypass STREAMS and directly call us, passing along the sockaddr
5777  * structure without the cumbersome T_UNITDATA_REQ interface for the case of
5778  * connected endpoints.
5779  */
5780 void
5781 udp_wput(queue_t *q, mblk_t *mp)
5782 {
5783 	conn_t		*connp = Q_TO_CONN(q);
5784 	udp_t		*udp = connp->conn_udp;
5785 	int		error = 0;
5786 	struct sockaddr	*addr;
5787 	socklen_t	addrlen;
5788 	udp_stack_t	*us = udp->udp_us;
5789 
5790 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START,
5791 	    "udp_wput_start: queue %p mp %p", q, mp);
5792 
5793 	/*
5794 	 * We directly handle several cases here: T_UNITDATA_REQ message
5795 	 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for connected
5796 	 * socket.
5797 	 */
5798 	switch (DB_TYPE(mp)) {
5799 	case M_DATA:
5800 		/*
5801 		 * Quick check for error cases. Checks will be done again
5802 		 * under the lock later on
5803 		 */
5804 		if (!udp->udp_issocket || udp->udp_state != TS_DATA_XFER) {
5805 			/* Not connected; address is required */
5806 			BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5807 			UDP_STAT(us, udp_out_err_notconn);
5808 			freemsg(mp);
5809 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5810 			    "udp_wput_end: connp %p (%S)", connp,
5811 			    "not-connected; address required");
5812 			return;
5813 		}
5814 		(void) udp_send_connected(connp, mp, NULL, NULL, -1);
5815 		return;
5816 
5817 	case M_PROTO:
5818 	case M_PCPROTO: {
5819 		struct T_unitdata_req *tudr;
5820 
5821 		ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX);
5822 		tudr = (struct T_unitdata_req *)mp->b_rptr;
5823 
5824 		/* Handle valid T_UNITDATA_REQ here */
5825 		if (MBLKL(mp) >= sizeof (*tudr) &&
5826 		    ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) {
5827 			if (mp->b_cont == NULL) {
5828 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5829 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5830 				error = EPROTO;
5831 				goto ud_error;
5832 			}
5833 
5834 			if (!MBLKIN(mp, 0, tudr->DEST_offset +
5835 			    tudr->DEST_length)) {
5836 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5837 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5838 				error = EADDRNOTAVAIL;
5839 				goto ud_error;
5840 			}
5841 			/*
5842 			 * If a port has not been bound to the stream, fail.
5843 			 * This is not a problem when sockfs is directly
5844 			 * above us, because it will ensure that the socket
5845 			 * is first bound before allowing data to be sent.
5846 			 */
5847 			if (udp->udp_state == TS_UNBND) {
5848 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5849 				    "udp_wput_end: q %p (%S)", q, "outstate");
5850 				error = EPROTO;
5851 				goto ud_error;
5852 			}
5853 			addr = (struct sockaddr *)
5854 			    &mp->b_rptr[tudr->DEST_offset];
5855 			addrlen = tudr->DEST_length;
5856 			if (tudr->OPT_length != 0)
5857 				UDP_STAT(us, udp_out_opt);
5858 			break;
5859 		}
5860 		/* FALLTHRU */
5861 	}
5862 	default:
5863 		udp_wput_other(q, mp);
5864 		return;
5865 	}
5866 	ASSERT(addr != NULL);
5867 
5868 	error = udp_send_not_connected(connp,  mp, addr, addrlen, NULL, NULL,
5869 	    -1);
5870 	if (error != 0) {
5871 ud_error:
5872 		UDP_STAT(us, udp_out_err_output);
5873 		ASSERT(mp != NULL);
5874 		/* mp is freed by the following routine */
5875 		udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
5876 		    (t_scalar_t)error);
5877 	}
5878 }
5879 
5880 /* ARGSUSED */
5881 static void
5882 udp_wput_fallback(queue_t *wq, mblk_t *mp)
5883 {
5884 #ifdef DEBUG
5885 	cmn_err(CE_CONT, "udp_wput_fallback: Message in fallback \n");
5886 #endif
5887 	freemsg(mp);
5888 }
5889 
5890 
5891 /*
5892  * udp_output_v6():
5893  * Assumes that udp_wput did some sanity checking on the destination
5894  * address.
5895  */
5896 static mblk_t *
5897 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error,
5898     struct nmsghdr *msg, cred_t *cr, pid_t pid)
5899 {
5900 	ip6_t		*ip6h;
5901 	ip6i_t		*ip6i;	/* mp1->b_rptr even if no ip6i_t */
5902 	mblk_t		*mp1 = mp;
5903 	mblk_t		*mp2;
5904 	int		udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
5905 	size_t		ip_len;
5906 	udpha_t		*udph;
5907 	udp_t		*udp = connp->conn_udp;
5908 	udp_stack_t	*us = udp->udp_us;
5909 	queue_t		*q = connp->conn_wq;
5910 	ip6_pkt_t	ipp_s;	/* For ancillary data options */
5911 	ip6_pkt_t	*ipp = &ipp_s;
5912 	ip6_pkt_t	*tipp;	/* temporary ipp */
5913 	uint32_t	csum = 0;
5914 	uint_t		ignore = 0;
5915 	uint_t		option_exists = 0, is_sticky = 0;
5916 	uint8_t		*cp;
5917 	uint8_t		*nxthdr_ptr;
5918 	in6_addr_t	ip6_dst;
5919 	in_port_t	port;
5920 	udpattrs_t	attrs;
5921 	boolean_t	opt_present;
5922 	ip6_hbh_t	*hopoptsptr = NULL;
5923 	uint_t		hopoptslen = 0;
5924 	boolean_t	is_ancillary = B_FALSE;
5925 	size_t		sth_wroff = 0;
5926 	ire_t		*ire;
5927 	boolean_t	update_lastdst = B_FALSE;
5928 
5929 	*error = 0;
5930 
5931 	/*
5932 	 * If the local address is a mapped address return
5933 	 * an error.
5934 	 * It would be possible to send an IPv6 packet but the
5935 	 * response would never make it back to the application
5936 	 * since it is bound to a mapped address.
5937 	 */
5938 	if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) {
5939 		*error = EADDRNOTAVAIL;
5940 		goto done;
5941 	}
5942 
5943 	ipp->ipp_fields = 0;
5944 	ipp->ipp_sticky_ignored = 0;
5945 
5946 	/*
5947 	 * If TPI options passed in, feed it for verification and handling
5948 	 */
5949 	attrs.udpattr_credset = B_FALSE;
5950 	opt_present = B_FALSE;
5951 	if (IPCL_IS_NONSTR(connp)) {
5952 		if (msg->msg_controllen != 0) {
5953 			attrs.udpattr_ipp6 = ipp;
5954 			attrs.udpattr_mb = mp;
5955 
5956 			rw_enter(&udp->udp_rwlock, RW_WRITER);
5957 			*error = process_auxiliary_options(connp,
5958 			    msg->msg_control, msg->msg_controllen,
5959 			    &attrs, &udp_opt_obj, udp_opt_set, cr);
5960 			rw_exit(&udp->udp_rwlock);
5961 			if (*error)
5962 				goto done;
5963 			ASSERT(*error == 0);
5964 			opt_present = B_TRUE;
5965 		}
5966 	} else {
5967 		if (DB_TYPE(mp) != M_DATA) {
5968 			mp1 = mp->b_cont;
5969 			if (((struct T_unitdata_req *)
5970 			    mp->b_rptr)->OPT_length != 0) {
5971 				attrs.udpattr_ipp6 = ipp;
5972 				attrs.udpattr_mb = mp;
5973 				if (udp_unitdata_opt_process(q, mp, error,
5974 				    &attrs) < 0) {
5975 					goto done;
5976 				}
5977 				ASSERT(*error == 0);
5978 				opt_present = B_TRUE;
5979 			}
5980 		}
5981 	}
5982 
5983 	/*
5984 	 * Determine whether we need to mark the mblk with the user's
5985 	 * credentials.
5986 	 * If labeled then sockfs would have already done this.
5987 	 */
5988 	ASSERT(!is_system_labeled() || msg_getcred(mp, NULL) != NULL);
5989 	ire = connp->conn_ire_cache;
5990 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) || (ire == NULL) ||
5991 	    (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, &sin6->sin6_addr)) ||
5992 	    (ire->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) {
5993 		if (cr != NULL && msg_getcred(mp, NULL) == NULL)
5994 			mblk_setcred(mp, cr, pid);
5995 	}
5996 
5997 	rw_enter(&udp->udp_rwlock, RW_READER);
5998 	ignore = ipp->ipp_sticky_ignored;
5999 
6000 	/* mp1 points to the M_DATA mblk carrying the packet */
6001 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
6002 
6003 	if (sin6->sin6_scope_id != 0 &&
6004 	    IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
6005 		/*
6006 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
6007 		 * option nor ancillary data.  It needs to be
6008 		 * explicitly set in options_exists.
6009 		 */
6010 		option_exists |= IPPF_SCOPE_ID;
6011 	}
6012 
6013 	/*
6014 	 * Compute the destination address
6015 	 */
6016 	ip6_dst = sin6->sin6_addr;
6017 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
6018 		ip6_dst = ipv6_loopback;
6019 
6020 	port = sin6->sin6_port;
6021 
6022 	/*
6023 	 * Cluster and TSOL notes, Cluster check:
6024 	 * see comments in udp_output_v4().
6025 	 */
6026 	mutex_enter(&connp->conn_lock);
6027 
6028 	if (cl_inet_connect2 != NULL &&
6029 	    (!IN6_ARE_ADDR_EQUAL(&ip6_dst, &udp->udp_v6lastdst) ||
6030 	    port != udp->udp_lastdstport)) {
6031 		mutex_exit(&connp->conn_lock);
6032 		*error = 0;
6033 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &ip6_dst, port, *error);
6034 		if (*error != 0) {
6035 			*error = EHOSTUNREACH;
6036 			rw_exit(&udp->udp_rwlock);
6037 			goto done;
6038 		}
6039 		update_lastdst = B_TRUE;
6040 		mutex_enter(&connp->conn_lock);
6041 	}
6042 
6043 	/*
6044 	 * If we're not going to the same destination as last time, then
6045 	 * recompute the label required.  This is done in a separate routine to
6046 	 * avoid blowing up our stack here.
6047 	 *
6048 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
6049 	 * to different destination may require different labels,
6050 	 * or worse, UDP packets to same IP address may require
6051 	 * different labels due to use of shared all-zones address.
6052 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
6053 	 * and sticky ipp_hopoptslen are consistent for the current
6054 	 * destination and are updated atomically.
6055 	 */
6056 	if (is_system_labeled()) {
6057 		cred_t  *credp;
6058 		pid_t   cpid;
6059 
6060 		/* Using UDP MLP requires SCM_UCRED from user */
6061 		if (connp->conn_mlp_type != mlptSingle &&
6062 		    !attrs.udpattr_credset) {
6063 			DTRACE_PROBE4(
6064 			    tx__ip__log__info__output__udp6,
6065 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
6066 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
6067 			*error = EINVAL;
6068 			rw_exit(&udp->udp_rwlock);
6069 			mutex_exit(&connp->conn_lock);
6070 			goto done;
6071 		}
6072 		/*
6073 		 * update label option for this UDP socket if
6074 		 * - the destination has changed,
6075 		 * - the UDP socket is MLP, or
6076 		 * - the cred attached to the mblk changed.
6077 		 */
6078 		credp = msg_getcred(mp, &cpid);
6079 		if (opt_present ||
6080 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst) ||
6081 		    connp->conn_mlp_type != mlptSingle ||
6082 		    credp != udp->udp_last_cred) {
6083 			if ((*error = udp_update_label_v6(q, mp, &ip6_dst))
6084 			    != 0) {
6085 				rw_exit(&udp->udp_rwlock);
6086 				mutex_exit(&connp->conn_lock);
6087 				goto done;
6088 			}
6089 			update_lastdst = B_TRUE;
6090 		}
6091 		/*
6092 		 * Attach the effective cred to the mblk to ensure future
6093 		 * routing decisions will be based on it's label.
6094 		 */
6095 		mblk_setcred(mp, udp->udp_effective_cred, cpid);
6096 	}
6097 
6098 	if (update_lastdst) {
6099 		udp->udp_v6lastdst = ip6_dst;
6100 		udp->udp_lastdstport = port;
6101 	}
6102 
6103 	/*
6104 	 * If there's a security label here, then we ignore any options the
6105 	 * user may try to set.  We keep the peer's label as a hidden sticky
6106 	 * option. We make a private copy of this label before releasing the
6107 	 * lock so that label is kept consistent with the destination addr.
6108 	 */
6109 	if (udp->udp_label_len_v6 > 0) {
6110 		ignore &= ~IPPF_HOPOPTS;
6111 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
6112 	}
6113 
6114 	if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) {
6115 		/* No sticky options nor ancillary data. */
6116 		mutex_exit(&connp->conn_lock);
6117 		goto no_options;
6118 	}
6119 
6120 	/*
6121 	 * Go through the options figuring out where each is going to
6122 	 * come from and build two masks.  The first mask indicates if
6123 	 * the option exists at all.  The second mask indicates if the
6124 	 * option is sticky or ancillary.
6125 	 */
6126 	if (!(ignore & IPPF_HOPOPTS)) {
6127 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
6128 			option_exists |= IPPF_HOPOPTS;
6129 			udp_ip_hdr_len += ipp->ipp_hopoptslen;
6130 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
6131 			option_exists |= IPPF_HOPOPTS;
6132 			is_sticky |= IPPF_HOPOPTS;
6133 			ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0);
6134 			hopoptsptr = kmem_alloc(
6135 			    udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
6136 			if (hopoptsptr == NULL) {
6137 				*error = ENOMEM;
6138 				mutex_exit(&connp->conn_lock);
6139 				goto done;
6140 			}
6141 			hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen;
6142 			bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr,
6143 			    hopoptslen);
6144 			udp_ip_hdr_len += hopoptslen;
6145 		}
6146 	}
6147 	mutex_exit(&connp->conn_lock);
6148 
6149 	if (!(ignore & IPPF_RTHDR)) {
6150 		if (ipp->ipp_fields & IPPF_RTHDR) {
6151 			option_exists |= IPPF_RTHDR;
6152 			udp_ip_hdr_len += ipp->ipp_rthdrlen;
6153 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
6154 			option_exists |= IPPF_RTHDR;
6155 			is_sticky |= IPPF_RTHDR;
6156 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen;
6157 		}
6158 	}
6159 
6160 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
6161 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
6162 			option_exists |= IPPF_RTDSTOPTS;
6163 			udp_ip_hdr_len += ipp->ipp_rtdstoptslen;
6164 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
6165 			option_exists |= IPPF_RTDSTOPTS;
6166 			is_sticky |= IPPF_RTDSTOPTS;
6167 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen;
6168 		}
6169 	}
6170 
6171 	if (!(ignore & IPPF_DSTOPTS)) {
6172 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
6173 			option_exists |= IPPF_DSTOPTS;
6174 			udp_ip_hdr_len += ipp->ipp_dstoptslen;
6175 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
6176 			option_exists |= IPPF_DSTOPTS;
6177 			is_sticky |= IPPF_DSTOPTS;
6178 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen;
6179 		}
6180 	}
6181 
6182 	if (!(ignore & IPPF_IFINDEX)) {
6183 		if (ipp->ipp_fields & IPPF_IFINDEX) {
6184 			option_exists |= IPPF_IFINDEX;
6185 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
6186 			option_exists |= IPPF_IFINDEX;
6187 			is_sticky |= IPPF_IFINDEX;
6188 		}
6189 	}
6190 
6191 	if (!(ignore & IPPF_ADDR)) {
6192 		if (ipp->ipp_fields & IPPF_ADDR) {
6193 			option_exists |= IPPF_ADDR;
6194 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) {
6195 			option_exists |= IPPF_ADDR;
6196 			is_sticky |= IPPF_ADDR;
6197 		}
6198 	}
6199 
6200 	if (!(ignore & IPPF_DONTFRAG)) {
6201 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
6202 			option_exists |= IPPF_DONTFRAG;
6203 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
6204 			option_exists |= IPPF_DONTFRAG;
6205 			is_sticky |= IPPF_DONTFRAG;
6206 		}
6207 	}
6208 
6209 	if (!(ignore & IPPF_USE_MIN_MTU)) {
6210 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
6211 			option_exists |= IPPF_USE_MIN_MTU;
6212 		} else if (udp->udp_sticky_ipp.ipp_fields &
6213 		    IPPF_USE_MIN_MTU) {
6214 			option_exists |= IPPF_USE_MIN_MTU;
6215 			is_sticky |= IPPF_USE_MIN_MTU;
6216 		}
6217 	}
6218 
6219 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
6220 		option_exists |= IPPF_HOPLIMIT;
6221 	/* IPV6_HOPLIMIT can never be sticky */
6222 	ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
6223 
6224 	if (!(ignore & IPPF_UNICAST_HOPS) &&
6225 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
6226 		option_exists |= IPPF_UNICAST_HOPS;
6227 		is_sticky |= IPPF_UNICAST_HOPS;
6228 	}
6229 
6230 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
6231 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
6232 		option_exists |= IPPF_MULTICAST_HOPS;
6233 		is_sticky |= IPPF_MULTICAST_HOPS;
6234 	}
6235 
6236 	if (!(ignore & IPPF_TCLASS)) {
6237 		if (ipp->ipp_fields & IPPF_TCLASS) {
6238 			option_exists |= IPPF_TCLASS;
6239 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
6240 			option_exists |= IPPF_TCLASS;
6241 			is_sticky |= IPPF_TCLASS;
6242 		}
6243 	}
6244 
6245 	if (!(ignore & IPPF_NEXTHOP) &&
6246 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_NEXTHOP)) {
6247 		option_exists |= IPPF_NEXTHOP;
6248 		is_sticky |= IPPF_NEXTHOP;
6249 	}
6250 
6251 no_options:
6252 
6253 	/*
6254 	 * If any options carried in the ip6i_t were specified, we
6255 	 * need to account for the ip6i_t in the data we'll be sending
6256 	 * down.
6257 	 */
6258 	if (option_exists & IPPF_HAS_IP6I)
6259 		udp_ip_hdr_len += sizeof (ip6i_t);
6260 
6261 	/* check/fix buffer config, setup pointers into it */
6262 	ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len];
6263 	if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) ||
6264 	    !OK_32PTR(ip6h)) {
6265 
6266 		/* Try to get everything in a single mblk next time */
6267 		if (udp_ip_hdr_len > udp->udp_max_hdr_len) {
6268 			udp->udp_max_hdr_len = udp_ip_hdr_len;
6269 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
6270 		}
6271 
6272 		mp2 = allocb(udp_ip_hdr_len + us->us_wroff_extra, BPRI_LO);
6273 		if (mp2 == NULL) {
6274 			*error = ENOMEM;
6275 			rw_exit(&udp->udp_rwlock);
6276 			goto done;
6277 		}
6278 		mp2->b_wptr = DB_LIM(mp2);
6279 		mp2->b_cont = mp1;
6280 		mp1 = mp2;
6281 		if (DB_TYPE(mp) != M_DATA)
6282 			mp->b_cont = mp1;
6283 		else
6284 			mp = mp1;
6285 
6286 		ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len);
6287 	}
6288 	mp1->b_rptr = (unsigned char *)ip6h;
6289 	ip6i = (ip6i_t *)ip6h;
6290 
6291 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp)
6292 	if (option_exists & IPPF_HAS_IP6I) {
6293 		ip6h = (ip6_t *)&ip6i[1];
6294 		ip6i->ip6i_flags = 0;
6295 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6296 
6297 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
6298 		if (option_exists & IPPF_SCOPE_ID) {
6299 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6300 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
6301 		} else if (option_exists & IPPF_IFINDEX) {
6302 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
6303 			ASSERT(tipp->ipp_ifindex != 0);
6304 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6305 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
6306 		}
6307 
6308 		if (option_exists & IPPF_ADDR) {
6309 			/*
6310 			 * Enable per-packet source address verification if
6311 			 * IPV6_PKTINFO specified the source address.
6312 			 * ip6_src is set in the transport's _wput function.
6313 			 */
6314 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
6315 		}
6316 
6317 		if (option_exists & IPPF_DONTFRAG) {
6318 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
6319 		}
6320 
6321 		if (option_exists & IPPF_USE_MIN_MTU) {
6322 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
6323 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
6324 		}
6325 
6326 		if (option_exists & IPPF_NEXTHOP) {
6327 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
6328 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
6329 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
6330 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
6331 		}
6332 
6333 		/*
6334 		 * tell IP this is an ip6i_t private header
6335 		 */
6336 		ip6i->ip6i_nxt = IPPROTO_RAW;
6337 	}
6338 
6339 	/* Initialize IPv6 header */
6340 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6341 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
6342 
6343 	/* Set the hoplimit of the outgoing packet. */
6344 	if (option_exists & IPPF_HOPLIMIT) {
6345 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
6346 		ip6h->ip6_hops = ipp->ipp_hoplimit;
6347 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6348 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
6349 		ip6h->ip6_hops = udp->udp_multicast_ttl;
6350 		if (option_exists & IPPF_MULTICAST_HOPS)
6351 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6352 	} else {
6353 		ip6h->ip6_hops = udp->udp_ttl;
6354 		if (option_exists & IPPF_UNICAST_HOPS)
6355 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6356 	}
6357 
6358 	if (option_exists & IPPF_ADDR) {
6359 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
6360 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
6361 		ip6h->ip6_src = tipp->ipp_addr;
6362 	} else {
6363 		/*
6364 		 * The source address was not set using IPV6_PKTINFO.
6365 		 * First look at the bound source.
6366 		 * If unspecified fallback to __sin6_src_id.
6367 		 */
6368 		ip6h->ip6_src = udp->udp_v6src;
6369 		if (sin6->__sin6_src_id != 0 &&
6370 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
6371 			ip_srcid_find_id(sin6->__sin6_src_id,
6372 			    &ip6h->ip6_src, connp->conn_zoneid,
6373 			    us->us_netstack);
6374 		}
6375 	}
6376 
6377 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
6378 	cp = (uint8_t *)&ip6h[1];
6379 
6380 	/*
6381 	 * Here's where we have to start stringing together
6382 	 * any extension headers in the right order:
6383 	 * Hop-by-hop, destination, routing, and final destination opts.
6384 	 */
6385 	if (option_exists & IPPF_HOPOPTS) {
6386 		/* Hop-by-hop options */
6387 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
6388 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
6389 		if (hopoptslen == 0) {
6390 			hopoptsptr = tipp->ipp_hopopts;
6391 			hopoptslen = tipp->ipp_hopoptslen;
6392 			is_ancillary = B_TRUE;
6393 		}
6394 
6395 		*nxthdr_ptr = IPPROTO_HOPOPTS;
6396 		nxthdr_ptr = &hbh->ip6h_nxt;
6397 
6398 		bcopy(hopoptsptr, cp, hopoptslen);
6399 		cp += hopoptslen;
6400 
6401 		if (hopoptsptr != NULL && !is_ancillary) {
6402 			kmem_free(hopoptsptr, hopoptslen);
6403 			hopoptsptr = NULL;
6404 			hopoptslen = 0;
6405 		}
6406 	}
6407 	/*
6408 	 * En-route destination options
6409 	 * Only do them if there's a routing header as well
6410 	 */
6411 	if (option_exists & IPPF_RTDSTOPTS) {
6412 		ip6_dest_t *dst = (ip6_dest_t *)cp;
6413 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
6414 
6415 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6416 		nxthdr_ptr = &dst->ip6d_nxt;
6417 
6418 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
6419 		cp += tipp->ipp_rtdstoptslen;
6420 	}
6421 	/*
6422 	 * Routing header next
6423 	 */
6424 	if (option_exists & IPPF_RTHDR) {
6425 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
6426 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
6427 
6428 		*nxthdr_ptr = IPPROTO_ROUTING;
6429 		nxthdr_ptr = &rt->ip6r_nxt;
6430 
6431 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
6432 		cp += tipp->ipp_rthdrlen;
6433 	}
6434 	/*
6435 	 * Do ultimate destination options
6436 	 */
6437 	if (option_exists & IPPF_DSTOPTS) {
6438 		ip6_dest_t *dest = (ip6_dest_t *)cp;
6439 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
6440 
6441 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6442 		nxthdr_ptr = &dest->ip6d_nxt;
6443 
6444 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
6445 		cp += tipp->ipp_dstoptslen;
6446 	}
6447 	/*
6448 	 * Now set the last header pointer to the proto passed in
6449 	 */
6450 	ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE));
6451 	*nxthdr_ptr = IPPROTO_UDP;
6452 
6453 	/* Update UDP header */
6454 	udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE);
6455 	udph->uha_dst_port = sin6->sin6_port;
6456 	udph->uha_src_port = udp->udp_port;
6457 
6458 	/*
6459 	 * Copy in the destination address
6460 	 */
6461 	ip6h->ip6_dst = ip6_dst;
6462 
6463 	ip6h->ip6_vcf =
6464 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
6465 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
6466 
6467 	if (option_exists & IPPF_TCLASS) {
6468 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
6469 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
6470 		    tipp->ipp_tclass);
6471 	}
6472 	rw_exit(&udp->udp_rwlock);
6473 
6474 	if (option_exists & IPPF_RTHDR) {
6475 		ip6_rthdr_t	*rth;
6476 
6477 		/*
6478 		 * Perform any processing needed for source routing.
6479 		 * We know that all extension headers will be in the same mblk
6480 		 * as the IPv6 header.
6481 		 */
6482 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
6483 		if (rth != NULL && rth->ip6r_segleft != 0) {
6484 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
6485 				/*
6486 				 * Drop packet - only support Type 0 routing.
6487 				 * Notify the application as well.
6488 				 */
6489 				*error = EPROTO;
6490 				goto done;
6491 			}
6492 
6493 			/*
6494 			 * rth->ip6r_len is twice the number of
6495 			 * addresses in the header. Thus it must be even.
6496 			 */
6497 			if (rth->ip6r_len & 0x1) {
6498 				*error = EPROTO;
6499 				goto done;
6500 			}
6501 			/*
6502 			 * Shuffle the routing header and ip6_dst
6503 			 * addresses, and get the checksum difference
6504 			 * between the first hop (in ip6_dst) and
6505 			 * the destination (in the last routing hdr entry).
6506 			 */
6507 			csum = ip_massage_options_v6(ip6h, rth,
6508 			    us->us_netstack);
6509 			/*
6510 			 * Verify that the first hop isn't a mapped address.
6511 			 * Routers along the path need to do this verification
6512 			 * for subsequent hops.
6513 			 */
6514 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
6515 				*error = EADDRNOTAVAIL;
6516 				goto done;
6517 			}
6518 
6519 			cp += (rth->ip6r_len + 1)*8;
6520 		}
6521 	}
6522 
6523 	/* count up length of UDP packet */
6524 	ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN;
6525 	if ((mp2 = mp1->b_cont) != NULL) {
6526 		do {
6527 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
6528 			ip_len += (uint32_t)MBLKL(mp2);
6529 		} while ((mp2 = mp2->b_cont) != NULL);
6530 	}
6531 
6532 	/*
6533 	 * If the size of the packet is greater than the maximum allowed by
6534 	 * ip, return an error. Passing this down could cause panics because
6535 	 * the size will have wrapped and be inconsistent with the msg size.
6536 	 */
6537 	if (ip_len > IP_MAXPACKET) {
6538 		*error = EMSGSIZE;
6539 		goto done;
6540 	}
6541 
6542 	/* Store the UDP length. Subtract length of extension hdrs */
6543 	udph->uha_length = htons(ip_len + IPV6_HDR_LEN -
6544 	    (int)((uchar_t *)udph - (uchar_t *)ip6h));
6545 
6546 	/*
6547 	 * We make it easy for IP to include our pseudo header
6548 	 * by putting our length in uh_checksum, modified (if
6549 	 * we have a routing header) by the checksum difference
6550 	 * between the ultimate destination and first hop addresses.
6551 	 * Note: UDP over IPv6 must always checksum the packet.
6552 	 */
6553 	csum += udph->uha_length;
6554 	csum = (csum & 0xFFFF) + (csum >> 16);
6555 	udph->uha_checksum = (uint16_t)csum;
6556 
6557 #ifdef _LITTLE_ENDIAN
6558 	ip_len = htons(ip_len);
6559 #endif
6560 	ip6h->ip6_plen = ip_len;
6561 
6562 	if (DB_TYPE(mp) != M_DATA) {
6563 		cred_t *cr;
6564 		pid_t cpid;
6565 
6566 		/* Move any cred from the T_UNITDATA_REQ to the packet */
6567 		cr = msg_extractcred(mp, &cpid);
6568 		if (cr != NULL) {
6569 			if (mp1->b_datap->db_credp != NULL)
6570 				crfree(mp1->b_datap->db_credp);
6571 			mp1->b_datap->db_credp = cr;
6572 			mp1->b_datap->db_cpid = cpid;
6573 		}
6574 
6575 		ASSERT(mp != mp1);
6576 		freeb(mp);
6577 	}
6578 
6579 	/* mp has been consumed and we'll return success */
6580 	ASSERT(*error == 0);
6581 	mp = NULL;
6582 
6583 	/* We're done. Pass the packet to IP */
6584 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
6585 	ip_output_v6(connp, mp1, q, IP_WPUT);
6586 
6587 done:
6588 	if (sth_wroff != 0) {
6589 		(void) proto_set_tx_wroff(RD(q), connp,
6590 		    udp->udp_max_hdr_len + us->us_wroff_extra);
6591 	}
6592 	if (hopoptsptr != NULL && !is_ancillary) {
6593 		kmem_free(hopoptsptr, hopoptslen);
6594 		hopoptsptr = NULL;
6595 	}
6596 	if (*error != 0) {
6597 		ASSERT(mp != NULL);
6598 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6599 	}
6600 	return (mp);
6601 }
6602 
6603 
6604 static int
6605 i_udp_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6606 {
6607 	sin_t *sin = (sin_t *)sa;
6608 	sin6_t *sin6 = (sin6_t *)sa;
6609 
6610 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6611 
6612 	if (udp->udp_state != TS_DATA_XFER)
6613 		return (ENOTCONN);
6614 
6615 	switch (udp->udp_family) {
6616 	case AF_INET:
6617 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6618 
6619 		if (*salenp < sizeof (sin_t))
6620 			return (EINVAL);
6621 
6622 		*salenp = sizeof (sin_t);
6623 		*sin = sin_null;
6624 		sin->sin_family = AF_INET;
6625 		sin->sin_port = udp->udp_dstport;
6626 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
6627 		break;
6628 
6629 	case AF_INET6:
6630 		if (*salenp < sizeof (sin6_t))
6631 			return (EINVAL);
6632 
6633 		*salenp = sizeof (sin6_t);
6634 		*sin6 = sin6_null;
6635 		sin6->sin6_family = AF_INET6;
6636 		sin6->sin6_port = udp->udp_dstport;
6637 		sin6->sin6_addr = udp->udp_v6dst;
6638 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6639 		break;
6640 	}
6641 
6642 	return (0);
6643 }
6644 
6645 static int
6646 udp_getmyname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6647 {
6648 	sin_t *sin = (sin_t *)sa;
6649 	sin6_t *sin6 = (sin6_t *)sa;
6650 
6651 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6652 
6653 	switch (udp->udp_family) {
6654 	case AF_INET:
6655 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6656 
6657 		if (*salenp < sizeof (sin_t))
6658 			return (EINVAL);
6659 
6660 		*salenp = sizeof (sin_t);
6661 		*sin = sin_null;
6662 		sin->sin_family = AF_INET;
6663 		sin->sin_port = udp->udp_port;
6664 
6665 		/*
6666 		 * If udp_v6src is unspecified, we might be bound to broadcast
6667 		 * / multicast.  Use udp_bound_v6src as local address instead
6668 		 * (that could also still be unspecified).
6669 		 */
6670 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
6671 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
6672 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
6673 		} else {
6674 			sin->sin_addr.s_addr =
6675 			    V4_PART_OF_V6(udp->udp_bound_v6src);
6676 		}
6677 		break;
6678 
6679 	case AF_INET6:
6680 		if (*salenp < sizeof (sin6_t))
6681 			return (EINVAL);
6682 
6683 		*salenp = sizeof (sin6_t);
6684 		*sin6 = sin6_null;
6685 		sin6->sin6_family = AF_INET6;
6686 		sin6->sin6_port = udp->udp_port;
6687 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6688 
6689 		/*
6690 		 * If udp_v6src is unspecified, we might be bound to broadcast
6691 		 * / multicast.  Use udp_bound_v6src as local address instead
6692 		 * (that could also still be unspecified).
6693 		 */
6694 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))
6695 			sin6->sin6_addr = udp->udp_v6src;
6696 		else
6697 			sin6->sin6_addr = udp->udp_bound_v6src;
6698 		break;
6699 	}
6700 
6701 	return (0);
6702 }
6703 
6704 /*
6705  * Handle special out-of-band ioctl requests (see PSARC/2008/265).
6706  */
6707 static void
6708 udp_wput_cmdblk(queue_t *q, mblk_t *mp)
6709 {
6710 	void	*data;
6711 	mblk_t	*datamp = mp->b_cont;
6712 	udp_t	*udp = Q_TO_UDP(q);
6713 	cmdblk_t *cmdp = (cmdblk_t *)mp->b_rptr;
6714 
6715 	if (datamp == NULL || MBLKL(datamp) < cmdp->cb_len) {
6716 		cmdp->cb_error = EPROTO;
6717 		qreply(q, mp);
6718 		return;
6719 	}
6720 	data = datamp->b_rptr;
6721 
6722 	rw_enter(&udp->udp_rwlock, RW_READER);
6723 	switch (cmdp->cb_cmd) {
6724 	case TI_GETPEERNAME:
6725 		cmdp->cb_error = i_udp_getpeername(udp, data, &cmdp->cb_len);
6726 		break;
6727 	case TI_GETMYNAME:
6728 		cmdp->cb_error = udp_getmyname(udp, data, &cmdp->cb_len);
6729 		break;
6730 	default:
6731 		cmdp->cb_error = EINVAL;
6732 		break;
6733 	}
6734 	rw_exit(&udp->udp_rwlock);
6735 
6736 	qreply(q, mp);
6737 }
6738 
6739 static void
6740 udp_use_pure_tpi(udp_t *udp)
6741 {
6742 	rw_enter(&udp->udp_rwlock, RW_WRITER);
6743 	udp->udp_issocket = B_FALSE;
6744 	rw_exit(&udp->udp_rwlock);
6745 
6746 	UDP_STAT(udp->udp_us, udp_sock_fallback);
6747 }
6748 
6749 static void
6750 udp_wput_other(queue_t *q, mblk_t *mp)
6751 {
6752 	uchar_t	*rptr = mp->b_rptr;
6753 	struct datab *db;
6754 	struct iocblk *iocp;
6755 	cred_t	*cr;
6756 	conn_t	*connp = Q_TO_CONN(q);
6757 	udp_t	*udp = connp->conn_udp;
6758 	udp_stack_t *us;
6759 
6760 	TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START,
6761 	    "udp_wput_other_start: q %p", q);
6762 
6763 	us = udp->udp_us;
6764 	db = mp->b_datap;
6765 
6766 	switch (db->db_type) {
6767 	case M_CMD:
6768 		udp_wput_cmdblk(q, mp);
6769 		return;
6770 
6771 	case M_PROTO:
6772 	case M_PCPROTO:
6773 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
6774 			freemsg(mp);
6775 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6776 			    "udp_wput_other_end: q %p (%S)", q, "protoshort");
6777 			return;
6778 		}
6779 		switch (((t_primp_t)rptr)->type) {
6780 		case T_ADDR_REQ:
6781 			udp_addr_req(q, mp);
6782 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6783 			    "udp_wput_other_end: q %p (%S)", q, "addrreq");
6784 			return;
6785 		case O_T_BIND_REQ:
6786 		case T_BIND_REQ:
6787 			udp_tpi_bind(q, mp);
6788 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6789 			    "udp_wput_other_end: q %p (%S)", q, "bindreq");
6790 			return;
6791 		case T_CONN_REQ:
6792 			udp_tpi_connect(q, mp);
6793 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6794 			    "udp_wput_other_end: q %p (%S)", q, "connreq");
6795 			return;
6796 		case T_CAPABILITY_REQ:
6797 			udp_capability_req(q, mp);
6798 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6799 			    "udp_wput_other_end: q %p (%S)", q, "capabreq");
6800 			return;
6801 		case T_INFO_REQ:
6802 			udp_info_req(q, mp);
6803 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6804 			    "udp_wput_other_end: q %p (%S)", q, "inforeq");
6805 			return;
6806 		case T_UNITDATA_REQ:
6807 			/*
6808 			 * If a T_UNITDATA_REQ gets here, the address must
6809 			 * be bad.  Valid T_UNITDATA_REQs are handled
6810 			 * in udp_wput.
6811 			 */
6812 			udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL);
6813 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6814 			    "udp_wput_other_end: q %p (%S)", q, "unitdatareq");
6815 			return;
6816 		case T_UNBIND_REQ:
6817 			udp_tpi_unbind(q, mp);
6818 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6819 			    "udp_wput_other_end: q %p (%S)", q, "unbindreq");
6820 			return;
6821 		case T_SVR4_OPTMGMT_REQ:
6822 			/*
6823 			 * All Solaris components should pass a db_credp
6824 			 * for this TPI message, hence we ASSERT.
6825 			 * But in case there is some other M_PROTO that looks
6826 			 * like a TPI message sent by some other kernel
6827 			 * component, we check and return an error.
6828 			 */
6829 			cr = msg_getcred(mp, NULL);
6830 			ASSERT(cr != NULL);
6831 			if (cr == NULL) {
6832 				udp_err_ack(q, mp, TSYSERR, EINVAL);
6833 				return;
6834 			}
6835 			if (!snmpcom_req(q, mp, udp_snmp_set, ip_snmp_get,
6836 			    cr)) {
6837 				(void) svr4_optcom_req(q,
6838 				    mp, cr, &udp_opt_obj, B_TRUE);
6839 			}
6840 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6841 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6842 			return;
6843 
6844 		case T_OPTMGMT_REQ:
6845 			/*
6846 			 * All Solaris components should pass a db_credp
6847 			 * for this TPI message, hence we ASSERT.
6848 			 * But in case there is some other M_PROTO that looks
6849 			 * like a TPI message sent by some other kernel
6850 			 * component, we check and return an error.
6851 			 */
6852 			cr = msg_getcred(mp, NULL);
6853 			ASSERT(cr != NULL);
6854 			if (cr == NULL) {
6855 				udp_err_ack(q, mp, TSYSERR, EINVAL);
6856 				return;
6857 			}
6858 			(void) tpi_optcom_req(q, mp, cr, &udp_opt_obj, B_TRUE);
6859 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6860 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6861 			return;
6862 
6863 		case T_DISCON_REQ:
6864 			udp_tpi_disconnect(q, mp);
6865 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6866 			    "udp_wput_other_end: q %p (%S)", q, "disconreq");
6867 			return;
6868 
6869 		/* The following TPI message is not supported by udp. */
6870 		case O_T_CONN_RES:
6871 		case T_CONN_RES:
6872 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6873 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6874 			    "udp_wput_other_end: q %p (%S)", q,
6875 			    "connres/disconreq");
6876 			return;
6877 
6878 		/* The following 3 TPI messages are illegal for udp. */
6879 		case T_DATA_REQ:
6880 		case T_EXDATA_REQ:
6881 		case T_ORDREL_REQ:
6882 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6883 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6884 			    "udp_wput_other_end: q %p (%S)", q,
6885 			    "data/exdata/ordrel");
6886 			return;
6887 		default:
6888 			break;
6889 		}
6890 		break;
6891 	case M_FLUSH:
6892 		if (*rptr & FLUSHW)
6893 			flushq(q, FLUSHDATA);
6894 		break;
6895 	case M_IOCTL:
6896 		iocp = (struct iocblk *)mp->b_rptr;
6897 		switch (iocp->ioc_cmd) {
6898 		case TI_GETPEERNAME:
6899 			if (udp->udp_state != TS_DATA_XFER) {
6900 				/*
6901 				 * If a default destination address has not
6902 				 * been associated with the stream, then we
6903 				 * don't know the peer's name.
6904 				 */
6905 				iocp->ioc_error = ENOTCONN;
6906 				iocp->ioc_count = 0;
6907 				mp->b_datap->db_type = M_IOCACK;
6908 				qreply(q, mp);
6909 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6910 				    "udp_wput_other_end: q %p (%S)", q,
6911 				    "getpeername");
6912 				return;
6913 			}
6914 			/* FALLTHRU */
6915 		case TI_GETMYNAME: {
6916 			/*
6917 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
6918 			 * need to copyin the user's strbuf structure.
6919 			 * Processing will continue in the M_IOCDATA case
6920 			 * below.
6921 			 */
6922 			mi_copyin(q, mp, NULL,
6923 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
6924 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6925 			    "udp_wput_other_end: q %p (%S)", q, "getmyname");
6926 			return;
6927 			}
6928 		case ND_SET:
6929 			/* nd_getset performs the necessary checking */
6930 		case ND_GET:
6931 			if (nd_getset(q, us->us_nd, mp)) {
6932 				qreply(q, mp);
6933 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6934 				    "udp_wput_other_end: q %p (%S)", q, "get");
6935 				return;
6936 			}
6937 			break;
6938 		case _SIOCSOCKFALLBACK:
6939 			/*
6940 			 * Either sockmod is about to be popped and the
6941 			 * socket would now be treated as a plain stream,
6942 			 * or a module is about to be pushed so we have
6943 			 * to follow pure TPI semantics.
6944 			 */
6945 			if (!udp->udp_issocket) {
6946 				DB_TYPE(mp) = M_IOCNAK;
6947 				iocp->ioc_error = EINVAL;
6948 			} else {
6949 				udp_use_pure_tpi(udp);
6950 
6951 				DB_TYPE(mp) = M_IOCACK;
6952 				iocp->ioc_error = 0;
6953 			}
6954 			iocp->ioc_count = 0;
6955 			iocp->ioc_rval = 0;
6956 			qreply(q, mp);
6957 			return;
6958 		default:
6959 			break;
6960 		}
6961 		break;
6962 	case M_IOCDATA:
6963 		udp_wput_iocdata(q, mp);
6964 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6965 		    "udp_wput_other_end: q %p (%S)", q, "iocdata");
6966 		return;
6967 	default:
6968 		/* Unrecognized messages are passed through without change. */
6969 		break;
6970 	}
6971 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6972 	    "udp_wput_other_end: q %p (%S)", q, "end");
6973 	ip_output(connp, mp, q, IP_WPUT);
6974 }
6975 
6976 /*
6977  * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA
6978  * messages.
6979  */
6980 static void
6981 udp_wput_iocdata(queue_t *q, mblk_t *mp)
6982 {
6983 	mblk_t		*mp1;
6984 	struct	iocblk *iocp = (struct iocblk *)mp->b_rptr;
6985 	STRUCT_HANDLE(strbuf, sb);
6986 	udp_t		*udp = Q_TO_UDP(q);
6987 	int		error;
6988 	uint_t		addrlen;
6989 
6990 	/* Make sure it is one of ours. */
6991 	switch (iocp->ioc_cmd) {
6992 	case TI_GETMYNAME:
6993 	case TI_GETPEERNAME:
6994 		break;
6995 	default:
6996 		ip_output(udp->udp_connp, mp, q, IP_WPUT);
6997 		return;
6998 	}
6999 
7000 	switch (mi_copy_state(q, mp, &mp1)) {
7001 	case -1:
7002 		return;
7003 	case MI_COPY_CASE(MI_COPY_IN, 1):
7004 		break;
7005 	case MI_COPY_CASE(MI_COPY_OUT, 1):
7006 		/*
7007 		 * The address has been copied out, so now
7008 		 * copyout the strbuf.
7009 		 */
7010 		mi_copyout(q, mp);
7011 		return;
7012 	case MI_COPY_CASE(MI_COPY_OUT, 2):
7013 		/*
7014 		 * The address and strbuf have been copied out.
7015 		 * We're done, so just acknowledge the original
7016 		 * M_IOCTL.
7017 		 */
7018 		mi_copy_done(q, mp, 0);
7019 		return;
7020 	default:
7021 		/*
7022 		 * Something strange has happened, so acknowledge
7023 		 * the original M_IOCTL with an EPROTO error.
7024 		 */
7025 		mi_copy_done(q, mp, EPROTO);
7026 		return;
7027 	}
7028 
7029 	/*
7030 	 * Now we have the strbuf structure for TI_GETMYNAME
7031 	 * and TI_GETPEERNAME.  Next we copyout the requested
7032 	 * address and then we'll copyout the strbuf.
7033 	 */
7034 	STRUCT_SET_HANDLE(sb, iocp->ioc_flag, (void *)mp1->b_rptr);
7035 	addrlen = udp->udp_family == AF_INET ? sizeof (sin_t) : sizeof (sin6_t);
7036 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
7037 		mi_copy_done(q, mp, EINVAL);
7038 		return;
7039 	}
7040 
7041 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
7042 
7043 	if (mp1 == NULL)
7044 		return;
7045 
7046 	rw_enter(&udp->udp_rwlock, RW_READER);
7047 	switch (iocp->ioc_cmd) {
7048 	case TI_GETMYNAME:
7049 		error = udp_do_getsockname(udp, (void *)mp1->b_rptr, &addrlen);
7050 		break;
7051 	case TI_GETPEERNAME:
7052 		error = udp_do_getpeername(udp, (void *)mp1->b_rptr, &addrlen);
7053 		break;
7054 	}
7055 	rw_exit(&udp->udp_rwlock);
7056 
7057 	if (error != 0) {
7058 		mi_copy_done(q, mp, error);
7059 	} else {
7060 		mp1->b_wptr += addrlen;
7061 		STRUCT_FSET(sb, len, addrlen);
7062 
7063 		/* Copy out the address */
7064 		mi_copyout(q, mp);
7065 	}
7066 }
7067 
7068 static int
7069 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
7070     udpattrs_t *udpattrs)
7071 {
7072 	struct T_unitdata_req *udreqp;
7073 	int is_absreq_failure;
7074 	cred_t *cr;
7075 
7076 	ASSERT(((t_primp_t)mp->b_rptr)->type);
7077 
7078 	/*
7079 	 * All Solaris components should pass a db_credp
7080 	 * for this TPI message, hence we should ASSERT.
7081 	 * However, RPC (svc_clts_ksend) does this odd thing where it
7082 	 * passes the options from a T_UNITDATA_IND unchanged in a
7083 	 * T_UNITDATA_REQ. While that is the right thing to do for
7084 	 * some options, SCM_UCRED being the key one, this also makes it
7085 	 * pass down IP_RECVDSTADDR. Hence we can't ASSERT here.
7086 	 */
7087 	cr = msg_getcred(mp, NULL);
7088 	if (cr == NULL) {
7089 		cr = Q_TO_CONN(q)->conn_cred;
7090 	}
7091 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
7092 
7093 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
7094 	    udreqp->OPT_offset, cr, &udp_opt_obj,
7095 	    udpattrs, &is_absreq_failure);
7096 
7097 	if (*errorp != 0) {
7098 		/*
7099 		 * Note: No special action needed in this
7100 		 * module for "is_absreq_failure"
7101 		 */
7102 		return (-1);		/* failure */
7103 	}
7104 	ASSERT(is_absreq_failure == 0);
7105 	return (0);	/* success */
7106 }
7107 
7108 void
7109 udp_ddi_g_init(void)
7110 {
7111 	udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr,
7112 	    udp_opt_obj.odb_opt_arr_cnt);
7113 
7114 	/*
7115 	 * We want to be informed each time a stack is created or
7116 	 * destroyed in the kernel, so we can maintain the
7117 	 * set of udp_stack_t's.
7118 	 */
7119 	netstack_register(NS_UDP, udp_stack_init, NULL, udp_stack_fini);
7120 }
7121 
7122 void
7123 udp_ddi_g_destroy(void)
7124 {
7125 	netstack_unregister(NS_UDP);
7126 }
7127 
7128 #define	INET_NAME	"ip"
7129 
7130 /*
7131  * Initialize the UDP stack instance.
7132  */
7133 static void *
7134 udp_stack_init(netstackid_t stackid, netstack_t *ns)
7135 {
7136 	udp_stack_t	*us;
7137 	udpparam_t	*pa;
7138 	int		i;
7139 	int		error = 0;
7140 	major_t		major;
7141 
7142 	us = (udp_stack_t *)kmem_zalloc(sizeof (*us), KM_SLEEP);
7143 	us->us_netstack = ns;
7144 
7145 	us->us_num_epriv_ports = UDP_NUM_EPRIV_PORTS;
7146 	us->us_epriv_ports[0] = 2049;
7147 	us->us_epriv_ports[1] = 4045;
7148 
7149 	/*
7150 	 * The smallest anonymous port in the priviledged port range which UDP
7151 	 * looks for free port.  Use in the option UDP_ANONPRIVBIND.
7152 	 */
7153 	us->us_min_anonpriv_port = 512;
7154 
7155 	us->us_bind_fanout_size = udp_bind_fanout_size;
7156 
7157 	/* Roundup variable that might have been modified in /etc/system */
7158 	if (us->us_bind_fanout_size & (us->us_bind_fanout_size - 1)) {
7159 		/* Not a power of two. Round up to nearest power of two */
7160 		for (i = 0; i < 31; i++) {
7161 			if (us->us_bind_fanout_size < (1 << i))
7162 				break;
7163 		}
7164 		us->us_bind_fanout_size = 1 << i;
7165 	}
7166 	us->us_bind_fanout = kmem_zalloc(us->us_bind_fanout_size *
7167 	    sizeof (udp_fanout_t), KM_SLEEP);
7168 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7169 		mutex_init(&us->us_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT,
7170 		    NULL);
7171 	}
7172 
7173 	pa = (udpparam_t *)kmem_alloc(sizeof (udp_param_arr), KM_SLEEP);
7174 
7175 	us->us_param_arr = pa;
7176 	bcopy(udp_param_arr, us->us_param_arr, sizeof (udp_param_arr));
7177 
7178 	(void) udp_param_register(&us->us_nd,
7179 	    us->us_param_arr, A_CNT(udp_param_arr));
7180 
7181 	us->us_kstat = udp_kstat2_init(stackid, &us->us_statistics);
7182 	us->us_mibkp = udp_kstat_init(stackid);
7183 
7184 	major = mod_name_to_major(INET_NAME);
7185 	error = ldi_ident_from_major(major, &us->us_ldi_ident);
7186 	ASSERT(error == 0);
7187 	return (us);
7188 }
7189 
7190 /*
7191  * Free the UDP stack instance.
7192  */
7193 static void
7194 udp_stack_fini(netstackid_t stackid, void *arg)
7195 {
7196 	udp_stack_t *us = (udp_stack_t *)arg;
7197 	int i;
7198 
7199 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7200 		mutex_destroy(&us->us_bind_fanout[i].uf_lock);
7201 	}
7202 
7203 	kmem_free(us->us_bind_fanout, us->us_bind_fanout_size *
7204 	    sizeof (udp_fanout_t));
7205 
7206 	us->us_bind_fanout = NULL;
7207 
7208 	nd_free(&us->us_nd);
7209 	kmem_free(us->us_param_arr, sizeof (udp_param_arr));
7210 	us->us_param_arr = NULL;
7211 
7212 	udp_kstat_fini(stackid, us->us_mibkp);
7213 	us->us_mibkp = NULL;
7214 
7215 	udp_kstat2_fini(stackid, us->us_kstat);
7216 	us->us_kstat = NULL;
7217 	bzero(&us->us_statistics, sizeof (us->us_statistics));
7218 
7219 	ldi_ident_release(us->us_ldi_ident);
7220 	kmem_free(us, sizeof (*us));
7221 }
7222 
7223 static void *
7224 udp_kstat2_init(netstackid_t stackid, udp_stat_t *us_statisticsp)
7225 {
7226 	kstat_t *ksp;
7227 
7228 	udp_stat_t template = {
7229 		{ "udp_ip_send",		KSTAT_DATA_UINT64 },
7230 		{ "udp_ip_ire_send",		KSTAT_DATA_UINT64 },
7231 		{ "udp_ire_null",		KSTAT_DATA_UINT64 },
7232 		{ "udp_sock_fallback",		KSTAT_DATA_UINT64 },
7233 		{ "udp_out_sw_cksum",		KSTAT_DATA_UINT64 },
7234 		{ "udp_out_sw_cksum_bytes",	KSTAT_DATA_UINT64 },
7235 		{ "udp_out_opt",		KSTAT_DATA_UINT64 },
7236 		{ "udp_out_err_notconn",	KSTAT_DATA_UINT64 },
7237 		{ "udp_out_err_output",		KSTAT_DATA_UINT64 },
7238 		{ "udp_out_err_tudr",		KSTAT_DATA_UINT64 },
7239 		{ "udp_in_pktinfo",		KSTAT_DATA_UINT64 },
7240 		{ "udp_in_recvdstaddr",		KSTAT_DATA_UINT64 },
7241 		{ "udp_in_recvopts",		KSTAT_DATA_UINT64 },
7242 		{ "udp_in_recvif",		KSTAT_DATA_UINT64 },
7243 		{ "udp_in_recvslla",		KSTAT_DATA_UINT64 },
7244 		{ "udp_in_recvucred",		KSTAT_DATA_UINT64 },
7245 		{ "udp_in_recvttl",		KSTAT_DATA_UINT64 },
7246 		{ "udp_in_recvhopopts",		KSTAT_DATA_UINT64 },
7247 		{ "udp_in_recvhoplimit",	KSTAT_DATA_UINT64 },
7248 		{ "udp_in_recvdstopts",		KSTAT_DATA_UINT64 },
7249 		{ "udp_in_recvrtdstopts",	KSTAT_DATA_UINT64 },
7250 		{ "udp_in_recvrthdr",		KSTAT_DATA_UINT64 },
7251 		{ "udp_in_recvpktinfo",		KSTAT_DATA_UINT64 },
7252 		{ "udp_in_recvtclass",		KSTAT_DATA_UINT64 },
7253 		{ "udp_in_timestamp",		KSTAT_DATA_UINT64 },
7254 #ifdef DEBUG
7255 		{ "udp_data_conn",		KSTAT_DATA_UINT64 },
7256 		{ "udp_data_notconn",		KSTAT_DATA_UINT64 },
7257 #endif
7258 	};
7259 
7260 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, "udpstat", "net",
7261 	    KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
7262 	    KSTAT_FLAG_VIRTUAL, stackid);
7263 
7264 	if (ksp == NULL)
7265 		return (NULL);
7266 
7267 	bcopy(&template, us_statisticsp, sizeof (template));
7268 	ksp->ks_data = (void *)us_statisticsp;
7269 	ksp->ks_private = (void *)(uintptr_t)stackid;
7270 
7271 	kstat_install(ksp);
7272 	return (ksp);
7273 }
7274 
7275 static void
7276 udp_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
7277 {
7278 	if (ksp != NULL) {
7279 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7280 		kstat_delete_netstack(ksp, stackid);
7281 	}
7282 }
7283 
7284 static void *
7285 udp_kstat_init(netstackid_t stackid)
7286 {
7287 	kstat_t	*ksp;
7288 
7289 	udp_named_kstat_t template = {
7290 		{ "inDatagrams",	KSTAT_DATA_UINT64, 0 },
7291 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
7292 		{ "outDatagrams",	KSTAT_DATA_UINT64, 0 },
7293 		{ "entrySize",		KSTAT_DATA_INT32, 0 },
7294 		{ "entry6Size",		KSTAT_DATA_INT32, 0 },
7295 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
7296 	};
7297 
7298 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, UDP_MOD_NAME, "mib2",
7299 	    KSTAT_TYPE_NAMED,
7300 	    NUM_OF_FIELDS(udp_named_kstat_t), 0, stackid);
7301 
7302 	if (ksp == NULL || ksp->ks_data == NULL)
7303 		return (NULL);
7304 
7305 	template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t);
7306 	template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t);
7307 
7308 	bcopy(&template, ksp->ks_data, sizeof (template));
7309 	ksp->ks_update = udp_kstat_update;
7310 	ksp->ks_private = (void *)(uintptr_t)stackid;
7311 
7312 	kstat_install(ksp);
7313 	return (ksp);
7314 }
7315 
7316 static void
7317 udp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
7318 {
7319 	if (ksp != NULL) {
7320 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7321 		kstat_delete_netstack(ksp, stackid);
7322 	}
7323 }
7324 
7325 static int
7326 udp_kstat_update(kstat_t *kp, int rw)
7327 {
7328 	udp_named_kstat_t *udpkp;
7329 	netstackid_t	stackid = (netstackid_t)(uintptr_t)kp->ks_private;
7330 	netstack_t	*ns;
7331 	udp_stack_t	*us;
7332 
7333 	if ((kp == NULL) || (kp->ks_data == NULL))
7334 		return (EIO);
7335 
7336 	if (rw == KSTAT_WRITE)
7337 		return (EACCES);
7338 
7339 	ns = netstack_find_by_stackid(stackid);
7340 	if (ns == NULL)
7341 		return (-1);
7342 	us = ns->netstack_udp;
7343 	if (us == NULL) {
7344 		netstack_rele(ns);
7345 		return (-1);
7346 	}
7347 	udpkp = (udp_named_kstat_t *)kp->ks_data;
7348 
7349 	udpkp->inDatagrams.value.ui64 =	us->us_udp_mib.udpHCInDatagrams;
7350 	udpkp->inErrors.value.ui32 =	us->us_udp_mib.udpInErrors;
7351 	udpkp->outDatagrams.value.ui64 = us->us_udp_mib.udpHCOutDatagrams;
7352 	udpkp->outErrors.value.ui32 =	us->us_udp_mib.udpOutErrors;
7353 	netstack_rele(ns);
7354 	return (0);
7355 }
7356 
7357 static size_t
7358 udp_set_rcv_hiwat(udp_t *udp, size_t size)
7359 {
7360 	udp_stack_t *us = udp->udp_us;
7361 
7362 	/* We add a bit of extra buffering */
7363 	size += size >> 1;
7364 	if (size > us->us_max_buf)
7365 		size = us->us_max_buf;
7366 
7367 	udp->udp_rcv_hiwat = size;
7368 	return (size);
7369 }
7370 
7371 /*
7372  * For the lower queue so that UDP can be a dummy mux.
7373  * Nobody should be sending
7374  * packets up this stream
7375  */
7376 static void
7377 udp_lrput(queue_t *q, mblk_t *mp)
7378 {
7379 	mblk_t *mp1;
7380 
7381 	switch (mp->b_datap->db_type) {
7382 	case M_FLUSH:
7383 		/* Turn around */
7384 		if (*mp->b_rptr & FLUSHW) {
7385 			*mp->b_rptr &= ~FLUSHR;
7386 			qreply(q, mp);
7387 			return;
7388 		}
7389 		break;
7390 	}
7391 	/* Could receive messages that passed through ar_rput */
7392 	for (mp1 = mp; mp1; mp1 = mp1->b_cont)
7393 		mp1->b_prev = mp1->b_next = NULL;
7394 	freemsg(mp);
7395 }
7396 
7397 /*
7398  * For the lower queue so that UDP can be a dummy mux.
7399  * Nobody should be sending packets down this stream.
7400  */
7401 /* ARGSUSED */
7402 void
7403 udp_lwput(queue_t *q, mblk_t *mp)
7404 {
7405 	freemsg(mp);
7406 }
7407 
7408 /*
7409  * Below routines for UDP socket module.
7410  */
7411 
7412 static conn_t *
7413 udp_do_open(cred_t *credp, boolean_t isv6, int flags)
7414 {
7415 	udp_t		*udp;
7416 	conn_t		*connp;
7417 	zoneid_t 	zoneid;
7418 	netstack_t 	*ns;
7419 	udp_stack_t 	*us;
7420 
7421 	ns = netstack_find_by_cred(credp);
7422 	ASSERT(ns != NULL);
7423 	us = ns->netstack_udp;
7424 	ASSERT(us != NULL);
7425 
7426 	/*
7427 	 * For exclusive stacks we set the zoneid to zero
7428 	 * to make UDP operate as if in the global zone.
7429 	 */
7430 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
7431 		zoneid = GLOBAL_ZONEID;
7432 	else
7433 		zoneid = crgetzoneid(credp);
7434 
7435 	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);
7436 
7437 	connp = ipcl_conn_create(IPCL_UDPCONN, flags, ns);
7438 	if (connp == NULL) {
7439 		netstack_rele(ns);
7440 		return (NULL);
7441 	}
7442 	udp = connp->conn_udp;
7443 
7444 	/*
7445 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
7446 	 * done by netstack_find_by_cred()
7447 	 */
7448 	netstack_rele(ns);
7449 
7450 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7451 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
7452 	ASSERT(connp->conn_udp == udp);
7453 	ASSERT(udp->udp_connp == connp);
7454 
7455 	/* Set the initial state of the stream and the privilege status. */
7456 	udp->udp_state = TS_UNBND;
7457 	if (isv6) {
7458 		udp->udp_family = AF_INET6;
7459 		udp->udp_ipversion = IPV6_VERSION;
7460 		udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
7461 		udp->udp_ttl = us->us_ipv6_hoplimit;
7462 		connp->conn_af_isv6 = B_TRUE;
7463 		connp->conn_flags |= IPCL_ISV6;
7464 	} else {
7465 		udp->udp_family = AF_INET;
7466 		udp->udp_ipversion = IPV4_VERSION;
7467 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE;
7468 		udp->udp_ttl = us->us_ipv4_ttl;
7469 		connp->conn_af_isv6 = B_FALSE;
7470 		connp->conn_flags &= ~IPCL_ISV6;
7471 	}
7472 
7473 	udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
7474 	udp->udp_pending_op = -1;
7475 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
7476 	connp->conn_zoneid = zoneid;
7477 
7478 	udp->udp_open_time = lbolt64;
7479 	udp->udp_open_pid = curproc->p_pid;
7480 
7481 	/*
7482 	 * If the caller has the process-wide flag set, then default to MAC
7483 	 * exempt mode.  This allows read-down to unlabeled hosts.
7484 	 */
7485 	if (getpflags(NET_MAC_AWARE, credp) != 0)
7486 		connp->conn_mac_exempt = B_TRUE;
7487 
7488 	connp->conn_ulp_labeled = is_system_labeled();
7489 
7490 	udp->udp_us = us;
7491 
7492 	connp->conn_recv = udp_input;
7493 	crhold(credp);
7494 	connp->conn_cred = credp;
7495 
7496 	*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
7497 
7498 	rw_exit(&udp->udp_rwlock);
7499 
7500 	return (connp);
7501 }
7502 
7503 /* ARGSUSED */
7504 sock_lower_handle_t
7505 udp_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
7506     uint_t *smodep, int *errorp, int flags, cred_t *credp)
7507 {
7508 	udp_t		*udp = NULL;
7509 	udp_stack_t	*us;
7510 	conn_t		*connp;
7511 	boolean_t	isv6;
7512 
7513 	if (type != SOCK_DGRAM || (family != AF_INET && family != AF_INET6) ||
7514 	    (proto != 0 && proto != IPPROTO_UDP)) {
7515 		*errorp = EPROTONOSUPPORT;
7516 		return (NULL);
7517 	}
7518 
7519 	if (family == AF_INET6)
7520 		isv6 = B_TRUE;
7521 	else
7522 		isv6 = B_FALSE;
7523 
7524 	connp = udp_do_open(credp, isv6, flags);
7525 	if (connp == NULL) {
7526 		*errorp = ENOMEM;
7527 		return (NULL);
7528 	}
7529 
7530 	udp = connp->conn_udp;
7531 	ASSERT(udp != NULL);
7532 	us = udp->udp_us;
7533 	ASSERT(us != NULL);
7534 
7535 	udp->udp_issocket = B_TRUE;
7536 	connp->conn_flags |= IPCL_NONSTR | IPCL_SOCKET;
7537 
7538 	/* Set flow control */
7539 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7540 	(void) udp_set_rcv_hiwat(udp, us->us_recv_hiwat);
7541 	udp->udp_rcv_disply_hiwat = us->us_recv_hiwat;
7542 	udp->udp_rcv_lowat = udp_mod_info.mi_lowat;
7543 	udp->udp_xmit_hiwat = us->us_xmit_hiwat;
7544 	udp->udp_xmit_lowat = us->us_xmit_lowat;
7545 
7546 	if (udp->udp_family == AF_INET6) {
7547 		/* Build initial header template for transmit */
7548 		if ((*errorp = udp_build_hdrs(udp)) != 0) {
7549 			rw_exit(&udp->udp_rwlock);
7550 			ipcl_conn_destroy(connp);
7551 			return (NULL);
7552 		}
7553 	}
7554 	rw_exit(&udp->udp_rwlock);
7555 
7556 	connp->conn_flow_cntrld = B_FALSE;
7557 
7558 	ASSERT(us->us_ldi_ident != NULL);
7559 
7560 	if ((*errorp = ip_create_helper_stream(connp, us->us_ldi_ident)) != 0) {
7561 		ip1dbg(("udp_create: create of IP helper stream failed\n"));
7562 		udp_do_close(connp);
7563 		return (NULL);
7564 	}
7565 
7566 	/* Set the send flow control */
7567 	connp->conn_wq->q_hiwat = us->us_xmit_hiwat;
7568 	connp->conn_wq->q_lowat = us->us_xmit_lowat;
7569 
7570 	mutex_enter(&connp->conn_lock);
7571 	connp->conn_state_flags &= ~CONN_INCIPIENT;
7572 	mutex_exit(&connp->conn_lock);
7573 
7574 	*errorp = 0;
7575 	*smodep = SM_ATOMIC;
7576 	*sock_downcalls = &sock_udp_downcalls;
7577 	return ((sock_lower_handle_t)connp);
7578 }
7579 
7580 /* ARGSUSED */
7581 void
7582 udp_activate(sock_lower_handle_t proto_handle, sock_upper_handle_t sock_handle,
7583     sock_upcalls_t *sock_upcalls, int flags, cred_t *cr)
7584 {
7585 	conn_t 		*connp = (conn_t *)proto_handle;
7586 	udp_t 		*udp = connp->conn_udp;
7587 	udp_stack_t	*us = udp->udp_us;
7588 	struct sock_proto_props sopp;
7589 
7590 	/* All Solaris components should pass a cred for this operation. */
7591 	ASSERT(cr != NULL);
7592 
7593 	connp->conn_upcalls = sock_upcalls;
7594 	connp->conn_upper_handle = sock_handle;
7595 
7596 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT |
7597 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
7598 	sopp.sopp_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
7599 	sopp.sopp_maxblk = INFPSZ;
7600 	sopp.sopp_rxhiwat = udp->udp_rcv_hiwat;
7601 	sopp.sopp_maxaddrlen = sizeof (sin6_t);
7602 	sopp.sopp_maxpsz =
7603 	    (udp->udp_family == AF_INET) ? UDP_MAXPACKET_IPV4 :
7604 	    UDP_MAXPACKET_IPV6;
7605 	sopp.sopp_minpsz = (udp_mod_info.mi_minpsz == 1) ? 0 :
7606 	    udp_mod_info.mi_minpsz;
7607 
7608 	(*connp->conn_upcalls->su_set_proto_props)(connp->conn_upper_handle,
7609 	    &sopp);
7610 }
7611 
7612 static void
7613 udp_do_close(conn_t *connp)
7614 {
7615 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
7616 
7617 	udp_quiesce_conn(connp);
7618 	ip_quiesce_conn(connp);
7619 
7620 	if (!IPCL_IS_NONSTR(connp)) {
7621 		ASSERT(connp->conn_wq != NULL);
7622 		ASSERT(connp->conn_rq != NULL);
7623 		qprocsoff(connp->conn_rq);
7624 	}
7625 
7626 	udp_close_free(connp);
7627 
7628 	/*
7629 	 * Now we are truly single threaded on this stream, and can
7630 	 * delete the things hanging off the connp, and finally the connp.
7631 	 * We removed this connp from the fanout list, it cannot be
7632 	 * accessed thru the fanouts, and we already waited for the
7633 	 * conn_ref to drop to 0. We are already in close, so
7634 	 * there cannot be any other thread from the top. qprocsoff
7635 	 * has completed, and service has completed or won't run in
7636 	 * future.
7637 	 */
7638 	ASSERT(connp->conn_ref == 1);
7639 	if (!IPCL_IS_NONSTR(connp)) {
7640 		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
7641 	} else {
7642 		ip_free_helper_stream(connp);
7643 	}
7644 
7645 	connp->conn_ref--;
7646 	ipcl_conn_destroy(connp);
7647 }
7648 
7649 /* ARGSUSED */
7650 int
7651 udp_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
7652 {
7653 	conn_t	*connp = (conn_t *)proto_handle;
7654 
7655 	/* All Solaris components should pass a cred for this operation. */
7656 	ASSERT(cr != NULL);
7657 
7658 	udp_do_close(connp);
7659 	return (0);
7660 }
7661 
7662 static int
7663 udp_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len, cred_t *cr,
7664     boolean_t bind_to_req_port_only)
7665 {
7666 	sin_t		*sin;
7667 	sin6_t		*sin6;
7668 	sin6_t		sin6addr;
7669 	in_port_t	port;		/* Host byte order */
7670 	in_port_t	requested_port;	/* Host byte order */
7671 	int		count;
7672 	in6_addr_t	v6src;
7673 	int		loopmax;
7674 	udp_fanout_t	*udpf;
7675 	in_port_t	lport;		/* Network byte order */
7676 	zoneid_t	zoneid;
7677 	udp_t		*udp;
7678 	boolean_t	is_inaddr_any;
7679 	mlp_type_t	addrtype, mlptype;
7680 	udp_stack_t	*us;
7681 	int		error = 0;
7682 	mblk_t		*mp = NULL;
7683 
7684 	udp = connp->conn_udp;
7685 	us = udp->udp_us;
7686 
7687 	if (udp->udp_state != TS_UNBND) {
7688 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7689 		    "udp_bind: bad state, %u", udp->udp_state);
7690 		return (-TOUTSTATE);
7691 	}
7692 
7693 	switch (len) {
7694 	case 0:
7695 		if (udp->udp_family == AF_INET) {
7696 			sin = (sin_t *)&sin6addr;
7697 			*sin = sin_null;
7698 			sin->sin_family = AF_INET;
7699 			sin->sin_addr.s_addr = INADDR_ANY;
7700 			udp->udp_ipversion = IPV4_VERSION;
7701 		} else {
7702 			ASSERT(udp->udp_family == AF_INET6);
7703 			sin6 = (sin6_t *)&sin6addr;
7704 			*sin6 = sin6_null;
7705 			sin6->sin6_family = AF_INET6;
7706 			V6_SET_ZERO(sin6->sin6_addr);
7707 			udp->udp_ipversion = IPV6_VERSION;
7708 		}
7709 		port = 0;
7710 		break;
7711 
7712 	case sizeof (sin_t):	/* Complete IPv4 address */
7713 		sin = (sin_t *)sa;
7714 
7715 		if (sin == NULL || !OK_32PTR((char *)sin))
7716 			return (EINVAL);
7717 
7718 		if (udp->udp_family != AF_INET ||
7719 		    sin->sin_family != AF_INET) {
7720 			return (EAFNOSUPPORT);
7721 		}
7722 		port = ntohs(sin->sin_port);
7723 		break;
7724 
7725 	case sizeof (sin6_t):	/* complete IPv6 address */
7726 		sin6 = (sin6_t *)sa;
7727 
7728 		if (sin6 == NULL || !OK_32PTR((char *)sin6))
7729 			return (EINVAL);
7730 
7731 		if (udp->udp_family != AF_INET6 ||
7732 		    sin6->sin6_family != AF_INET6) {
7733 			return (EAFNOSUPPORT);
7734 		}
7735 		port = ntohs(sin6->sin6_port);
7736 		break;
7737 
7738 	default:		/* Invalid request */
7739 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7740 		    "udp_bind: bad ADDR_length length %u", len);
7741 		return (-TBADADDR);
7742 	}
7743 
7744 	requested_port = port;
7745 
7746 	if (requested_port == 0 || !bind_to_req_port_only)
7747 		bind_to_req_port_only = B_FALSE;
7748 	else		/* T_BIND_REQ and requested_port != 0 */
7749 		bind_to_req_port_only = B_TRUE;
7750 
7751 	if (requested_port == 0) {
7752 		/*
7753 		 * If the application passed in zero for the port number, it
7754 		 * doesn't care which port number we bind to. Get one in the
7755 		 * valid range.
7756 		 */
7757 		if (udp->udp_anon_priv_bind) {
7758 			port = udp_get_next_priv_port(udp);
7759 		} else {
7760 			port = udp_update_next_port(udp,
7761 			    us->us_next_port_to_try, B_TRUE);
7762 		}
7763 	} else {
7764 		/*
7765 		 * If the port is in the well-known privileged range,
7766 		 * make sure the caller was privileged.
7767 		 */
7768 		int i;
7769 		boolean_t priv = B_FALSE;
7770 
7771 		if (port < us->us_smallest_nonpriv_port) {
7772 			priv = B_TRUE;
7773 		} else {
7774 			for (i = 0; i < us->us_num_epriv_ports; i++) {
7775 				if (port == us->us_epriv_ports[i]) {
7776 					priv = B_TRUE;
7777 					break;
7778 				}
7779 			}
7780 		}
7781 
7782 		if (priv) {
7783 			if (secpolicy_net_privaddr(cr, port, IPPROTO_UDP) != 0)
7784 				return (-TACCES);
7785 		}
7786 	}
7787 
7788 	if (port == 0)
7789 		return (-TNOADDR);
7790 
7791 	/*
7792 	 * The state must be TS_UNBND. TPI mandates that users must send
7793 	 * TPI primitives only 1 at a time and wait for the response before
7794 	 * sending the next primitive.
7795 	 */
7796 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7797 	if (udp->udp_state != TS_UNBND || udp->udp_pending_op != -1) {
7798 		rw_exit(&udp->udp_rwlock);
7799 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7800 		    "udp_bind: bad state, %u", udp->udp_state);
7801 		return (-TOUTSTATE);
7802 	}
7803 	/* XXX how to remove the T_BIND_REQ? Should set it before calling */
7804 	udp->udp_pending_op = T_BIND_REQ;
7805 	/*
7806 	 * Copy the source address into our udp structure. This address
7807 	 * may still be zero; if so, IP will fill in the correct address
7808 	 * each time an outbound packet is passed to it. Since the udp is
7809 	 * not yet in the bind hash list, we don't grab the uf_lock to
7810 	 * change udp_ipversion
7811 	 */
7812 	if (udp->udp_family == AF_INET) {
7813 		ASSERT(sin != NULL);
7814 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
7815 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
7816 		    udp->udp_ip_snd_options_len;
7817 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src);
7818 	} else {
7819 		ASSERT(sin6 != NULL);
7820 		v6src = sin6->sin6_addr;
7821 		if (IN6_IS_ADDR_V4MAPPED(&v6src)) {
7822 			/*
7823 			 * no need to hold the uf_lock to set the udp_ipversion
7824 			 * since we are not yet in the fanout list
7825 			 */
7826 			udp->udp_ipversion = IPV4_VERSION;
7827 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
7828 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
7829 		} else {
7830 			udp->udp_ipversion = IPV6_VERSION;
7831 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
7832 		}
7833 	}
7834 
7835 	/*
7836 	 * If udp_reuseaddr is not set, then we have to make sure that
7837 	 * the IP address and port number the application requested
7838 	 * (or we selected for the application) is not being used by
7839 	 * another stream.  If another stream is already using the
7840 	 * requested IP address and port, the behavior depends on
7841 	 * "bind_to_req_port_only". If set the bind fails; otherwise we
7842 	 * search for any an unused port to bind to the the stream.
7843 	 *
7844 	 * As per the BSD semantics, as modified by the Deering multicast
7845 	 * changes, if udp_reuseaddr is set, then we allow multiple binds
7846 	 * to the same port independent of the local IP address.
7847 	 *
7848 	 * This is slightly different than in SunOS 4.X which did not
7849 	 * support IP multicast. Note that the change implemented by the
7850 	 * Deering multicast code effects all binds - not only binding
7851 	 * to IP multicast addresses.
7852 	 *
7853 	 * Note that when binding to port zero we ignore SO_REUSEADDR in
7854 	 * order to guarantee a unique port.
7855 	 */
7856 
7857 	count = 0;
7858 	if (udp->udp_anon_priv_bind) {
7859 		/*
7860 		 * loopmax = (IPPORT_RESERVED-1) -
7861 		 *    us->us_min_anonpriv_port + 1
7862 		 */
7863 		loopmax = IPPORT_RESERVED - us->us_min_anonpriv_port;
7864 	} else {
7865 		loopmax = us->us_largest_anon_port -
7866 		    us->us_smallest_anon_port + 1;
7867 	}
7868 
7869 	is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src);
7870 	zoneid = connp->conn_zoneid;
7871 
7872 	for (;;) {
7873 		udp_t		*udp1;
7874 		boolean_t	found_exclbind = B_FALSE;
7875 
7876 		/*
7877 		 * Walk through the list of udp streams bound to
7878 		 * requested port with the same IP address.
7879 		 */
7880 		lport = htons(port);
7881 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(lport,
7882 		    us->us_bind_fanout_size)];
7883 		mutex_enter(&udpf->uf_lock);
7884 		for (udp1 = udpf->uf_udp; udp1 != NULL;
7885 		    udp1 = udp1->udp_bind_hash) {
7886 			if (lport != udp1->udp_port)
7887 				continue;
7888 
7889 			/*
7890 			 * On a labeled system, we must treat bindings to ports
7891 			 * on shared IP addresses by sockets with MAC exemption
7892 			 * privilege as being in all zones, as there's
7893 			 * otherwise no way to identify the right receiver.
7894 			 */
7895 			if (!(IPCL_ZONE_MATCH(udp1->udp_connp, zoneid) ||
7896 			    IPCL_ZONE_MATCH(connp,
7897 			    udp1->udp_connp->conn_zoneid)) &&
7898 			    !connp->conn_mac_exempt && \
7899 			    !udp1->udp_connp->conn_mac_exempt)
7900 				continue;
7901 
7902 			/*
7903 			 * If UDP_EXCLBIND is set for either the bound or
7904 			 * binding endpoint, the semantics of bind
7905 			 * is changed according to the following chart.
7906 			 *
7907 			 * spec = specified address (v4 or v6)
7908 			 * unspec = unspecified address (v4 or v6)
7909 			 * A = specified addresses are different for endpoints
7910 			 *
7911 			 * bound	bind to		allowed?
7912 			 * -------------------------------------
7913 			 * unspec	unspec		no
7914 			 * unspec	spec		no
7915 			 * spec		unspec		no
7916 			 * spec		spec		yes if A
7917 			 *
7918 			 * For labeled systems, SO_MAC_EXEMPT behaves the same
7919 			 * as UDP_EXCLBIND, except that zoneid is ignored.
7920 			 */
7921 			if (udp1->udp_exclbind || udp->udp_exclbind ||
7922 			    udp1->udp_connp->conn_mac_exempt ||
7923 			    connp->conn_mac_exempt) {
7924 				if (V6_OR_V4_INADDR_ANY(
7925 				    udp1->udp_bound_v6src) ||
7926 				    is_inaddr_any ||
7927 				    IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
7928 				    &v6src)) {
7929 					found_exclbind = B_TRUE;
7930 					break;
7931 				}
7932 				continue;
7933 			}
7934 
7935 			/*
7936 			 * Check ipversion to allow IPv4 and IPv6 sockets to
7937 			 * have disjoint port number spaces.
7938 			 */
7939 			if (udp->udp_ipversion != udp1->udp_ipversion) {
7940 
7941 				/*
7942 				 * On the first time through the loop, if the
7943 				 * the user intentionally specified a
7944 				 * particular port number, then ignore any
7945 				 * bindings of the other protocol that may
7946 				 * conflict. This allows the user to bind IPv6
7947 				 * alone and get both v4 and v6, or bind both
7948 				 * both and get each seperately. On subsequent
7949 				 * times through the loop, we're checking a
7950 				 * port that we chose (not the user) and thus
7951 				 * we do not allow casual duplicate bindings.
7952 				 */
7953 				if (count == 0 && requested_port != 0)
7954 					continue;
7955 			}
7956 
7957 			/*
7958 			 * No difference depending on SO_REUSEADDR.
7959 			 *
7960 			 * If existing port is bound to a
7961 			 * non-wildcard IP address and
7962 			 * the requesting stream is bound to
7963 			 * a distinct different IP addresses
7964 			 * (non-wildcard, also), keep going.
7965 			 */
7966 			if (!is_inaddr_any &&
7967 			    !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) &&
7968 			    !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
7969 			    &v6src)) {
7970 				continue;
7971 			}
7972 			break;
7973 		}
7974 
7975 		if (!found_exclbind &&
7976 		    (udp->udp_reuseaddr && requested_port != 0)) {
7977 			break;
7978 		}
7979 
7980 		if (udp1 == NULL) {
7981 			/*
7982 			 * No other stream has this IP address
7983 			 * and port number. We can use it.
7984 			 */
7985 			break;
7986 		}
7987 		mutex_exit(&udpf->uf_lock);
7988 		if (bind_to_req_port_only) {
7989 			/*
7990 			 * We get here only when requested port
7991 			 * is bound (and only first  of the for()
7992 			 * loop iteration).
7993 			 *
7994 			 * The semantics of this bind request
7995 			 * require it to fail so we return from
7996 			 * the routine (and exit the loop).
7997 			 *
7998 			 */
7999 			udp->udp_pending_op = -1;
8000 			rw_exit(&udp->udp_rwlock);
8001 			return (-TADDRBUSY);
8002 		}
8003 
8004 		if (udp->udp_anon_priv_bind) {
8005 			port = udp_get_next_priv_port(udp);
8006 		} else {
8007 			if ((count == 0) && (requested_port != 0)) {
8008 				/*
8009 				 * If the application wants us to find
8010 				 * a port, get one to start with. Set
8011 				 * requested_port to 0, so that we will
8012 				 * update us->us_next_port_to_try below.
8013 				 */
8014 				port = udp_update_next_port(udp,
8015 				    us->us_next_port_to_try, B_TRUE);
8016 				requested_port = 0;
8017 			} else {
8018 				port = udp_update_next_port(udp, port + 1,
8019 				    B_FALSE);
8020 			}
8021 		}
8022 
8023 		if (port == 0 || ++count >= loopmax) {
8024 			/*
8025 			 * We've tried every possible port number and
8026 			 * there are none available, so send an error
8027 			 * to the user.
8028 			 */
8029 			udp->udp_pending_op = -1;
8030 			rw_exit(&udp->udp_rwlock);
8031 			return (-TNOADDR);
8032 		}
8033 	}
8034 
8035 	/*
8036 	 * Copy the source address into our udp structure.  This address
8037 	 * may still be zero; if so, ip will fill in the correct address
8038 	 * each time an outbound packet is passed to it.
8039 	 * If we are binding to a broadcast or multicast address then
8040 	 * udp_post_ip_bind_connect will clear the source address
8041 	 * when udp_do_bind success.
8042 	 */
8043 	udp->udp_v6src = udp->udp_bound_v6src = v6src;
8044 	udp->udp_port = lport;
8045 	/*
8046 	 * Now reset the the next anonymous port if the application requested
8047 	 * an anonymous port, or we handed out the next anonymous port.
8048 	 */
8049 	if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) {
8050 		us->us_next_port_to_try = port + 1;
8051 	}
8052 
8053 	/* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */
8054 	if (udp->udp_family == AF_INET) {
8055 		sin->sin_port = udp->udp_port;
8056 	} else {
8057 		sin6->sin6_port = udp->udp_port;
8058 		/* Rebuild the header template */
8059 		error = udp_build_hdrs(udp);
8060 		if (error != 0) {
8061 			udp->udp_pending_op = -1;
8062 			rw_exit(&udp->udp_rwlock);
8063 			mutex_exit(&udpf->uf_lock);
8064 			return (error);
8065 		}
8066 	}
8067 	udp->udp_state = TS_IDLE;
8068 	udp_bind_hash_insert(udpf, udp);
8069 	mutex_exit(&udpf->uf_lock);
8070 	rw_exit(&udp->udp_rwlock);
8071 
8072 	if (cl_inet_bind) {
8073 		/*
8074 		 * Running in cluster mode - register bind information
8075 		 */
8076 		if (udp->udp_ipversion == IPV4_VERSION) {
8077 			(*cl_inet_bind)(connp->conn_netstack->netstack_stackid,
8078 			    IPPROTO_UDP, AF_INET,
8079 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8080 			    (in_port_t)udp->udp_port, NULL);
8081 		} else {
8082 			(*cl_inet_bind)(connp->conn_netstack->netstack_stackid,
8083 			    IPPROTO_UDP, AF_INET6,
8084 			    (uint8_t *)&(udp->udp_v6src),
8085 			    (in_port_t)udp->udp_port, NULL);
8086 		}
8087 	}
8088 
8089 	connp->conn_anon_port = (is_system_labeled() && requested_port == 0);
8090 	if (is_system_labeled() && (!connp->conn_anon_port ||
8091 	    connp->conn_anon_mlp)) {
8092 		uint16_t mlpport;
8093 		zone_t *zone;
8094 
8095 		zone = crgetzone(cr);
8096 		connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth :
8097 		    mlptSingle;
8098 		addrtype = tsol_mlp_addr_type(
8099 		    connp->conn_allzones ? ALL_ZONES : zone->zone_id,
8100 		    IPV6_VERSION, &v6src, us->us_netstack->netstack_ip);
8101 		if (addrtype == mlptSingle) {
8102 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8103 			udp->udp_pending_op = -1;
8104 			rw_exit(&udp->udp_rwlock);
8105 			connp->conn_anon_port = B_FALSE;
8106 			connp->conn_mlp_type = mlptSingle;
8107 			return (-TNOADDR);
8108 		}
8109 		mlpport = connp->conn_anon_port ? PMAPPORT : port;
8110 		mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport,
8111 		    addrtype);
8112 
8113 		/*
8114 		 * It is a coding error to attempt to bind an MLP port
8115 		 * without first setting SOL_SOCKET/SCM_UCRED.
8116 		 */
8117 		if (mlptype != mlptSingle &&
8118 		    connp->conn_mlp_type == mlptSingle) {
8119 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8120 			udp->udp_pending_op = -1;
8121 			rw_exit(&udp->udp_rwlock);
8122 			connp->conn_anon_port = B_FALSE;
8123 			connp->conn_mlp_type = mlptSingle;
8124 			return (EINVAL);
8125 		}
8126 
8127 		/*
8128 		 * It is an access violation to attempt to bind an MLP port
8129 		 * without NET_BINDMLP privilege.
8130 		 */
8131 		if (mlptype != mlptSingle &&
8132 		    secpolicy_net_bindmlp(cr) != 0) {
8133 			if (udp->udp_debug) {
8134 				(void) strlog(UDP_MOD_ID, 0, 1,
8135 				    SL_ERROR|SL_TRACE,
8136 				    "udp_bind: no priv for multilevel port %d",
8137 				    mlpport);
8138 			}
8139 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8140 			udp->udp_pending_op = -1;
8141 			rw_exit(&udp->udp_rwlock);
8142 			connp->conn_anon_port = B_FALSE;
8143 			connp->conn_mlp_type = mlptSingle;
8144 			return (-TACCES);
8145 		}
8146 
8147 		/*
8148 		 * If we're specifically binding a shared IP address and the
8149 		 * port is MLP on shared addresses, then check to see if this
8150 		 * zone actually owns the MLP.  Reject if not.
8151 		 */
8152 		if (mlptype == mlptShared && addrtype == mlptShared) {
8153 			/*
8154 			 * No need to handle exclusive-stack zones since
8155 			 * ALL_ZONES only applies to the shared stack.
8156 			 */
8157 			zoneid_t mlpzone;
8158 
8159 			mlpzone = tsol_mlp_findzone(IPPROTO_UDP,
8160 			    htons(mlpport));
8161 			if (connp->conn_zoneid != mlpzone) {
8162 				if (udp->udp_debug) {
8163 					(void) strlog(UDP_MOD_ID, 0, 1,
8164 					    SL_ERROR|SL_TRACE,
8165 					    "udp_bind: attempt to bind port "
8166 					    "%d on shared addr in zone %d "
8167 					    "(should be %d)",
8168 					    mlpport, connp->conn_zoneid,
8169 					    mlpzone);
8170 				}
8171 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8172 				udp->udp_pending_op = -1;
8173 				rw_exit(&udp->udp_rwlock);
8174 				connp->conn_anon_port = B_FALSE;
8175 				connp->conn_mlp_type = mlptSingle;
8176 				return (-TACCES);
8177 			}
8178 		}
8179 		if (connp->conn_anon_port) {
8180 			error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp,
8181 			    port, B_TRUE);
8182 			if (error != 0) {
8183 				if (udp->udp_debug) {
8184 					(void) strlog(UDP_MOD_ID, 0, 1,
8185 					    SL_ERROR|SL_TRACE,
8186 					    "udp_bind: cannot establish anon "
8187 					    "MLP for port %d", port);
8188 				}
8189 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8190 				udp->udp_pending_op = -1;
8191 				rw_exit(&udp->udp_rwlock);
8192 				connp->conn_anon_port = B_FALSE;
8193 				connp->conn_mlp_type = mlptSingle;
8194 				return (-TACCES);
8195 			}
8196 		}
8197 		connp->conn_mlp_type = mlptype;
8198 	}
8199 
8200 	if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8201 		/*
8202 		 * Append a request for an IRE if udp_v6src not
8203 		 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address).
8204 		 */
8205 		mp = allocb(sizeof (ire_t), BPRI_HI);
8206 		if (!mp) {
8207 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8208 			udp->udp_pending_op = -1;
8209 			rw_exit(&udp->udp_rwlock);
8210 			return (ENOMEM);
8211 		}
8212 		mp->b_wptr += sizeof (ire_t);
8213 		mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8214 	}
8215 	if (udp->udp_family == AF_INET6) {
8216 		ASSERT(udp->udp_connp->conn_af_isv6);
8217 		error = ip_proto_bind_laddr_v6(connp, &mp, IPPROTO_UDP,
8218 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
8219 	} else {
8220 		ASSERT(!udp->udp_connp->conn_af_isv6);
8221 		error = ip_proto_bind_laddr_v4(connp, &mp, IPPROTO_UDP,
8222 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port,
8223 		    B_TRUE);
8224 	}
8225 
8226 	(void) udp_post_ip_bind_connect(udp, mp, error);
8227 	return (error);
8228 }
8229 
8230 int
8231 udp_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
8232     socklen_t len, cred_t *cr)
8233 {
8234 	int		error;
8235 	conn_t		*connp;
8236 
8237 	/* All Solaris components should pass a cred for this operation. */
8238 	ASSERT(cr != NULL);
8239 
8240 	connp = (conn_t *)proto_handle;
8241 
8242 	if (sa == NULL)
8243 		error = udp_do_unbind(connp);
8244 	else
8245 		error = udp_do_bind(connp, sa, len, cr, B_TRUE);
8246 
8247 	if (error < 0) {
8248 		if (error == -TOUTSTATE)
8249 			error = EINVAL;
8250 		else
8251 			error = proto_tlitosyserr(-error);
8252 	}
8253 
8254 	return (error);
8255 }
8256 
8257 static int
8258 udp_implicit_bind(conn_t *connp, cred_t *cr)
8259 {
8260 	int error;
8261 
8262 	/* All Solaris components should pass a cred for this operation. */
8263 	ASSERT(cr != NULL);
8264 
8265 	error = udp_do_bind(connp, NULL, 0, cr, B_FALSE);
8266 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
8267 }
8268 
8269 /*
8270  * This routine removes a port number association from a stream. It
8271  * is called by udp_unbind and udp_tpi_unbind.
8272  */
8273 static int
8274 udp_do_unbind(conn_t *connp)
8275 {
8276 	udp_t 		*udp = connp->conn_udp;
8277 	udp_fanout_t	*udpf;
8278 	udp_stack_t	*us = udp->udp_us;
8279 
8280 	if (cl_inet_unbind != NULL) {
8281 		/*
8282 		 * Running in cluster mode - register unbind information
8283 		 */
8284 		if (udp->udp_ipversion == IPV4_VERSION) {
8285 			(*cl_inet_unbind)(
8286 			    connp->conn_netstack->netstack_stackid,
8287 			    IPPROTO_UDP, AF_INET,
8288 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8289 			    (in_port_t)udp->udp_port, NULL);
8290 		} else {
8291 			(*cl_inet_unbind)(
8292 			    connp->conn_netstack->netstack_stackid,
8293 			    IPPROTO_UDP, AF_INET6,
8294 			    (uint8_t *)&(udp->udp_v6src),
8295 			    (in_port_t)udp->udp_port, NULL);
8296 		}
8297 	}
8298 
8299 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8300 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8301 		rw_exit(&udp->udp_rwlock);
8302 		return (-TOUTSTATE);
8303 	}
8304 	udp->udp_pending_op = T_UNBIND_REQ;
8305 	rw_exit(&udp->udp_rwlock);
8306 
8307 	/*
8308 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
8309 	 * and therefore ip_unbind must never return NULL.
8310 	 */
8311 	ip_unbind(connp);
8312 
8313 	/*
8314 	 * Once we're unbound from IP, the pending operation may be cleared
8315 	 * here.
8316 	 */
8317 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8318 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8319 	    us->us_bind_fanout_size)];
8320 
8321 	mutex_enter(&udpf->uf_lock);
8322 	udp_bind_hash_remove(udp, B_TRUE);
8323 	V6_SET_ZERO(udp->udp_v6src);
8324 	V6_SET_ZERO(udp->udp_bound_v6src);
8325 	udp->udp_port = 0;
8326 	mutex_exit(&udpf->uf_lock);
8327 
8328 	udp->udp_pending_op = -1;
8329 	udp->udp_state = TS_UNBND;
8330 	if (udp->udp_family == AF_INET6)
8331 		(void) udp_build_hdrs(udp);
8332 	rw_exit(&udp->udp_rwlock);
8333 
8334 	return (0);
8335 }
8336 
8337 static int
8338 udp_post_ip_bind_connect(udp_t *udp, mblk_t *ire_mp, int error)
8339 {
8340 	ire_t		*ire;
8341 	udp_fanout_t	*udpf;
8342 	udp_stack_t	*us = udp->udp_us;
8343 
8344 	ASSERT(udp->udp_pending_op != -1);
8345 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8346 	if (error == 0) {
8347 		/* For udp_do_connect() success */
8348 		/* udp_do_bind() success will do nothing in here */
8349 		/*
8350 		 * If a broadcast/multicast address was bound, set
8351 		 * the source address to 0.
8352 		 * This ensures no datagrams with broadcast address
8353 		 * as source address are emitted (which would violate
8354 		 * RFC1122 - Hosts requirements)
8355 		 *
8356 		 * Note that when connecting the returned IRE is
8357 		 * for the destination address and we only perform
8358 		 * the broadcast check for the source address (it
8359 		 * is OK to connect to a broadcast/multicast address.)
8360 		 */
8361 		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
8362 			ire = (ire_t *)ire_mp->b_rptr;
8363 
8364 			/*
8365 			 * Note: we get IRE_BROADCAST for IPv6 to "mark" a
8366 			 * multicast local address.
8367 			 */
8368 			udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8369 			    us->us_bind_fanout_size)];
8370 			if (ire->ire_type == IRE_BROADCAST &&
8371 			    udp->udp_state != TS_DATA_XFER) {
8372 				ASSERT(udp->udp_pending_op == T_BIND_REQ ||
8373 				    udp->udp_pending_op == O_T_BIND_REQ);
8374 				/*
8375 				 * This was just a local bind to a broadcast
8376 				 * addr.
8377 				 */
8378 				mutex_enter(&udpf->uf_lock);
8379 				V6_SET_ZERO(udp->udp_v6src);
8380 				mutex_exit(&udpf->uf_lock);
8381 				if (udp->udp_family == AF_INET6)
8382 					(void) udp_build_hdrs(udp);
8383 			} else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8384 				if (udp->udp_family == AF_INET6)
8385 					(void) udp_build_hdrs(udp);
8386 			}
8387 		}
8388 	} else {
8389 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8390 		    us->us_bind_fanout_size)];
8391 		mutex_enter(&udpf->uf_lock);
8392 
8393 		if (udp->udp_state == TS_DATA_XFER) {
8394 			/* Connect failed */
8395 			/* Revert back to the bound source */
8396 			udp->udp_v6src = udp->udp_bound_v6src;
8397 			udp->udp_state = TS_IDLE;
8398 		} else {
8399 			/* For udp_do_bind() failed */
8400 			V6_SET_ZERO(udp->udp_v6src);
8401 			V6_SET_ZERO(udp->udp_bound_v6src);
8402 			udp->udp_state = TS_UNBND;
8403 			udp_bind_hash_remove(udp, B_TRUE);
8404 			udp->udp_port = 0;
8405 		}
8406 		mutex_exit(&udpf->uf_lock);
8407 		if (udp->udp_family == AF_INET6)
8408 			(void) udp_build_hdrs(udp);
8409 	}
8410 	udp->udp_pending_op = -1;
8411 	rw_exit(&udp->udp_rwlock);
8412 	if (ire_mp != NULL)
8413 		freeb(ire_mp);
8414 	return (error);
8415 }
8416 
8417 /*
8418  * It associates a default destination address with the stream.
8419  */
8420 static int
8421 udp_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len,
8422     cred_t *cr)
8423 {
8424 	sin6_t		*sin6;
8425 	sin_t		*sin;
8426 	in6_addr_t 	v6dst;
8427 	ipaddr_t 	v4dst;
8428 	uint16_t 	dstport;
8429 	uint32_t 	flowinfo;
8430 	mblk_t		*ire_mp;
8431 	udp_fanout_t	*udpf;
8432 	udp_t		*udp, *udp1;
8433 	ushort_t	ipversion;
8434 	udp_stack_t	*us;
8435 	int		error;
8436 
8437 	udp = connp->conn_udp;
8438 	us = udp->udp_us;
8439 
8440 	/*
8441 	 * Address has been verified by the caller
8442 	 */
8443 	switch (len) {
8444 	default:
8445 		/*
8446 		 * Should never happen
8447 		 */
8448 		return (EINVAL);
8449 
8450 	case sizeof (sin_t):
8451 		sin = (sin_t *)sa;
8452 		v4dst = sin->sin_addr.s_addr;
8453 		dstport = sin->sin_port;
8454 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8455 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8456 		ipversion = IPV4_VERSION;
8457 		break;
8458 
8459 	case sizeof (sin6_t):
8460 		sin6 = (sin6_t *)sa;
8461 		v6dst = sin6->sin6_addr;
8462 		dstport = sin6->sin6_port;
8463 		if (IN6_IS_ADDR_V4MAPPED(&v6dst)) {
8464 			IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst);
8465 			ipversion = IPV4_VERSION;
8466 			flowinfo = 0;
8467 		} else {
8468 			ipversion = IPV6_VERSION;
8469 			flowinfo = sin6->sin6_flowinfo;
8470 		}
8471 		break;
8472 	}
8473 
8474 	if (dstport == 0)
8475 		return (-TBADADDR);
8476 
8477 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8478 
8479 	/*
8480 	 * This UDP must have bound to a port already before doing a connect.
8481 	 * TPI mandates that users must send TPI primitives only 1 at a time
8482 	 * and wait for the response before sending the next primitive.
8483 	 */
8484 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8485 		rw_exit(&udp->udp_rwlock);
8486 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
8487 		    "udp_connect: bad state, %u", udp->udp_state);
8488 		return (-TOUTSTATE);
8489 	}
8490 	udp->udp_pending_op = T_CONN_REQ;
8491 	ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL);
8492 
8493 	if (ipversion == IPV4_VERSION) {
8494 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
8495 		    udp->udp_ip_snd_options_len;
8496 	} else {
8497 		udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
8498 	}
8499 
8500 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8501 	    us->us_bind_fanout_size)];
8502 
8503 	mutex_enter(&udpf->uf_lock);
8504 	if (udp->udp_state == TS_DATA_XFER) {
8505 		/* Already connected - clear out state */
8506 		udp->udp_v6src = udp->udp_bound_v6src;
8507 		udp->udp_state = TS_IDLE;
8508 	}
8509 
8510 	/*
8511 	 * Create a default IP header with no IP options.
8512 	 */
8513 	udp->udp_dstport = dstport;
8514 	udp->udp_ipversion = ipversion;
8515 	if (ipversion == IPV4_VERSION) {
8516 		/*
8517 		 * Interpret a zero destination to mean loopback.
8518 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8519 		 * generate the T_CONN_CON.
8520 		 */
8521 		if (v4dst == INADDR_ANY) {
8522 			v4dst = htonl(INADDR_LOOPBACK);
8523 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8524 			if (udp->udp_family == AF_INET) {
8525 				sin->sin_addr.s_addr = v4dst;
8526 			} else {
8527 				sin6->sin6_addr = v6dst;
8528 			}
8529 		}
8530 		udp->udp_v6dst = v6dst;
8531 		udp->udp_flowinfo = 0;
8532 
8533 		/*
8534 		 * If the destination address is multicast and
8535 		 * an outgoing multicast interface has been set,
8536 		 * use the address of that interface as our
8537 		 * source address if no source address has been set.
8538 		 */
8539 		if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY &&
8540 		    CLASSD(v4dst) &&
8541 		    udp->udp_multicast_if_addr != INADDR_ANY) {
8542 			IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr,
8543 			    &udp->udp_v6src);
8544 		}
8545 	} else {
8546 		ASSERT(udp->udp_ipversion == IPV6_VERSION);
8547 		/*
8548 		 * Interpret a zero destination to mean loopback.
8549 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8550 		 * generate the T_CONN_CON.
8551 		 */
8552 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
8553 			v6dst = ipv6_loopback;
8554 			sin6->sin6_addr = v6dst;
8555 		}
8556 		udp->udp_v6dst = v6dst;
8557 		udp->udp_flowinfo = flowinfo;
8558 		/*
8559 		 * If the destination address is multicast and
8560 		 * an outgoing multicast interface has been set,
8561 		 * then the ip bind logic will pick the correct source
8562 		 * address (i.e. matching the outgoing multicast interface).
8563 		 */
8564 	}
8565 
8566 	/*
8567 	 * Verify that the src/port/dst/port is unique for all
8568 	 * connections in TS_DATA_XFER
8569 	 */
8570 	for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) {
8571 		if (udp1->udp_state != TS_DATA_XFER)
8572 			continue;
8573 		if (udp->udp_port != udp1->udp_port ||
8574 		    udp->udp_ipversion != udp1->udp_ipversion ||
8575 		    dstport != udp1->udp_dstport ||
8576 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) ||
8577 		    !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst) ||
8578 		    !(IPCL_ZONE_MATCH(udp->udp_connp,
8579 		    udp1->udp_connp->conn_zoneid) ||
8580 		    IPCL_ZONE_MATCH(udp1->udp_connp,
8581 		    udp->udp_connp->conn_zoneid)))
8582 			continue;
8583 		mutex_exit(&udpf->uf_lock);
8584 		udp->udp_pending_op = -1;
8585 		rw_exit(&udp->udp_rwlock);
8586 		return (-TBADADDR);
8587 	}
8588 
8589 	if (cl_inet_connect2 != NULL) {
8590 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &v6dst, dstport, error);
8591 		if (error != 0) {
8592 			mutex_exit(&udpf->uf_lock);
8593 			udp->udp_pending_op = -1;
8594 			rw_exit(&udp->udp_rwlock);
8595 			return (-TBADADDR);
8596 		}
8597 	}
8598 
8599 	udp->udp_state = TS_DATA_XFER;
8600 	mutex_exit(&udpf->uf_lock);
8601 
8602 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
8603 	if (ire_mp == NULL) {
8604 		mutex_enter(&udpf->uf_lock);
8605 		udp->udp_state = TS_IDLE;
8606 		udp->udp_pending_op = -1;
8607 		mutex_exit(&udpf->uf_lock);
8608 		rw_exit(&udp->udp_rwlock);
8609 		return (ENOMEM);
8610 	}
8611 
8612 	rw_exit(&udp->udp_rwlock);
8613 
8614 	ire_mp->b_wptr += sizeof (ire_t);
8615 	ire_mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8616 
8617 	if (udp->udp_family == AF_INET) {
8618 		error = ip_proto_bind_connected_v4(connp, &ire_mp, IPPROTO_UDP,
8619 		    &V4_PART_OF_V6(udp->udp_v6src), udp->udp_port,
8620 		    V4_PART_OF_V6(udp->udp_v6dst), udp->udp_dstport,
8621 		    B_TRUE, B_TRUE, cr);
8622 	} else {
8623 		error = ip_proto_bind_connected_v6(connp, &ire_mp, IPPROTO_UDP,
8624 		    &udp->udp_v6src, udp->udp_port, &udp->udp_v6dst,
8625 		    &udp->udp_sticky_ipp, udp->udp_dstport, B_TRUE, B_TRUE, cr);
8626 	}
8627 
8628 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
8629 }
8630 
8631 /* ARGSUSED */
8632 static int
8633 udp_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
8634     socklen_t len, sock_connid_t *id, cred_t *cr)
8635 {
8636 	conn_t	*connp = (conn_t *)proto_handle;
8637 	udp_t	*udp = connp->conn_udp;
8638 	int	error;
8639 	boolean_t did_bind = B_FALSE;
8640 
8641 	/* All Solaris components should pass a cred for this operation. */
8642 	ASSERT(cr != NULL);
8643 
8644 	if (sa == NULL) {
8645 		/*
8646 		 * Disconnect
8647 		 * Make sure we are connected
8648 		 */
8649 		if (udp->udp_state != TS_DATA_XFER)
8650 			return (EINVAL);
8651 
8652 		error = udp_disconnect(connp);
8653 		return (error);
8654 	}
8655 
8656 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
8657 	if (error != 0)
8658 		goto done;
8659 
8660 	/* do an implicit bind if necessary */
8661 	if (udp->udp_state == TS_UNBND) {
8662 		error = udp_implicit_bind(connp, cr);
8663 		/*
8664 		 * We could be racing with an actual bind, in which case
8665 		 * we would see EPROTO. We cross our fingers and try
8666 		 * to connect.
8667 		 */
8668 		if (!(error == 0 || error == EPROTO))
8669 			goto done;
8670 		did_bind = B_TRUE;
8671 	}
8672 	/*
8673 	 * set SO_DGRAM_ERRIND
8674 	 */
8675 	udp->udp_dgram_errind = B_TRUE;
8676 
8677 	error = udp_do_connect(connp, sa, len, cr);
8678 
8679 	if (error != 0 && did_bind) {
8680 		int unbind_err;
8681 
8682 		unbind_err = udp_do_unbind(connp);
8683 		ASSERT(unbind_err == 0);
8684 	}
8685 
8686 	if (error == 0) {
8687 		*id = 0;
8688 		(*connp->conn_upcalls->su_connected)
8689 		    (connp->conn_upper_handle, 0, NULL, -1);
8690 	} else if (error < 0) {
8691 		error = proto_tlitosyserr(-error);
8692 	}
8693 
8694 done:
8695 	if (error != 0 && udp->udp_state == TS_DATA_XFER) {
8696 		/*
8697 		 * No need to hold locks to set state
8698 		 * after connect failure socket state is undefined
8699 		 * We set the state only to imitate old sockfs behavior
8700 		 */
8701 		udp->udp_state = TS_IDLE;
8702 	}
8703 	return (error);
8704 }
8705 
8706 /* ARGSUSED */
8707 int
8708 udp_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
8709     cred_t *cr)
8710 {
8711 	conn_t		*connp = (conn_t *)proto_handle;
8712 	udp_t		*udp = connp->conn_udp;
8713 	udp_stack_t	*us = udp->udp_us;
8714 	int		error = 0;
8715 
8716 	ASSERT(DB_TYPE(mp) == M_DATA);
8717 
8718 	/* All Solaris components should pass a cred for this operation. */
8719 	ASSERT(cr != NULL);
8720 
8721 	/* If labeled then sockfs should have already set db_credp */
8722 	ASSERT(!is_system_labeled() || msg_getcred(mp, NULL) != NULL);
8723 
8724 	/*
8725 	 * If the socket is connected and no change in destination
8726 	 */
8727 	if (msg->msg_namelen == 0) {
8728 		error = udp_send_connected(connp, mp, msg, cr, curproc->p_pid);
8729 		if (error == EDESTADDRREQ)
8730 			return (error);
8731 		else
8732 			return (udp->udp_dgram_errind ? error : 0);
8733 	}
8734 
8735 	/*
8736 	 * Do an implicit bind if necessary.
8737 	 */
8738 	if (udp->udp_state == TS_UNBND) {
8739 		error = udp_implicit_bind(connp, cr);
8740 		/*
8741 		 * We could be racing with an actual bind, in which case
8742 		 * we would see EPROTO. We cross our fingers and try
8743 		 * to send.
8744 		 */
8745 		if (!(error == 0 || error == EPROTO)) {
8746 			freemsg(mp);
8747 			return (error);
8748 		}
8749 	}
8750 
8751 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8752 
8753 	if (msg->msg_name != NULL && udp->udp_state == TS_DATA_XFER) {
8754 		rw_exit(&udp->udp_rwlock);
8755 		freemsg(mp);
8756 		return (EISCONN);
8757 	}
8758 
8759 
8760 	if (udp->udp_delayed_error != 0) {
8761 		boolean_t	match;
8762 
8763 		error = udp->udp_delayed_error;
8764 		match = B_FALSE;
8765 		udp->udp_delayed_error = 0;
8766 		switch (udp->udp_family) {
8767 		case AF_INET: {
8768 			/* Compare just IP address and port */
8769 			sin_t *sin1 = (sin_t *)msg->msg_name;
8770 			sin_t *sin2 = (sin_t *)&udp->udp_delayed_addr;
8771 
8772 			if (msg->msg_namelen == sizeof (sin_t) &&
8773 			    sin1->sin_port == sin2->sin_port &&
8774 			    sin1->sin_addr.s_addr == sin2->sin_addr.s_addr)
8775 				match = B_TRUE;
8776 
8777 			break;
8778 		}
8779 		case AF_INET6: {
8780 			sin6_t	*sin1 = (sin6_t *)msg->msg_name;
8781 			sin6_t	*sin2 = (sin6_t *)&udp->udp_delayed_addr;
8782 
8783 			if (msg->msg_namelen == sizeof (sin6_t) &&
8784 			    sin1->sin6_port == sin2->sin6_port &&
8785 			    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
8786 			    &sin2->sin6_addr))
8787 				match = B_TRUE;
8788 			break;
8789 		}
8790 		default:
8791 			ASSERT(0);
8792 		}
8793 
8794 		*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
8795 
8796 		if (match) {
8797 			rw_exit(&udp->udp_rwlock);
8798 			freemsg(mp);
8799 			return (error);
8800 		}
8801 	}
8802 
8803 	error = proto_verify_ip_addr(udp->udp_family,
8804 	    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
8805 	rw_exit(&udp->udp_rwlock);
8806 
8807 	if (error != 0) {
8808 		freemsg(mp);
8809 		return (error);
8810 	}
8811 
8812 	error = udp_send_not_connected(connp, mp,
8813 	    (struct sockaddr  *)msg->msg_name, msg->msg_namelen, msg, cr,
8814 	    curproc->p_pid);
8815 	if (error != 0) {
8816 		UDP_STAT(us, udp_out_err_output);
8817 		freemsg(mp);
8818 	}
8819 	return (udp->udp_dgram_errind ? error : 0);
8820 }
8821 
8822 int
8823 udp_fallback(sock_lower_handle_t proto_handle, queue_t *q,
8824     boolean_t issocket, so_proto_quiesced_cb_t quiesced_cb)
8825 {
8826 	conn_t 	*connp = (conn_t *)proto_handle;
8827 	udp_t	*udp;
8828 	struct T_capability_ack tca;
8829 	struct sockaddr_in6 laddr, faddr;
8830 	socklen_t laddrlen, faddrlen;
8831 	short opts;
8832 	struct stroptions *stropt;
8833 	mblk_t *stropt_mp;
8834 	int error;
8835 
8836 	udp = connp->conn_udp;
8837 
8838 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
8839 
8840 	/*
8841 	 * setup the fallback stream that was allocated
8842 	 */
8843 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
8844 	connp->conn_minor_arena = WR(q)->q_ptr;
8845 
8846 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
8847 
8848 	WR(q)->q_qinfo = &udp_winit;
8849 
8850 	connp->conn_rq = RD(q);
8851 	connp->conn_wq = WR(q);
8852 
8853 	/* Notify stream head about options before sending up data */
8854 	stropt_mp->b_datap->db_type = M_SETOPTS;
8855 	stropt_mp->b_wptr += sizeof (*stropt);
8856 	stropt = (struct stroptions *)stropt_mp->b_rptr;
8857 	stropt->so_flags = SO_WROFF | SO_HIWAT;
8858 	stropt->so_wroff =
8859 	    (ushort_t)(udp->udp_max_hdr_len + udp->udp_us->us_wroff_extra);
8860 	stropt->so_hiwat = udp->udp_rcv_disply_hiwat;
8861 	putnext(RD(q), stropt_mp);
8862 
8863 	/*
8864 	 * Free the helper stream
8865 	 */
8866 	ip_free_helper_stream(connp);
8867 
8868 	if (!issocket)
8869 		udp_use_pure_tpi(udp);
8870 
8871 	/*
8872 	 * Collect the information needed to sync with the sonode
8873 	 */
8874 	udp_do_capability_ack(udp, &tca, TC1_INFO);
8875 
8876 	laddrlen = faddrlen = sizeof (sin6_t);
8877 	(void) udp_getsockname((sock_lower_handle_t)connp,
8878 	    (struct sockaddr *)&laddr, &laddrlen, CRED());
8879 	error = udp_getpeername((sock_lower_handle_t)connp,
8880 	    (struct sockaddr *)&faddr, &faddrlen, CRED());
8881 	if (error != 0)
8882 		faddrlen = 0;
8883 
8884 	opts = 0;
8885 	if (udp->udp_dgram_errind)
8886 		opts |= SO_DGRAM_ERRIND;
8887 	if (udp->udp_dontroute)
8888 		opts |= SO_DONTROUTE;
8889 
8890 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
8891 	    (struct sockaddr *)&laddr, laddrlen,
8892 	    (struct sockaddr *)&faddr, faddrlen, opts);
8893 
8894 	mutex_enter(&udp->udp_recv_lock);
8895 	/*
8896 	 * Attempts to send data up during fallback will result in it being
8897 	 * queued in udp_t. Now we push up any queued packets.
8898 	 */
8899 	while (udp->udp_fallback_queue_head != NULL) {
8900 		mblk_t *mp;
8901 		mp = udp->udp_fallback_queue_head;
8902 		udp->udp_fallback_queue_head = mp->b_next;
8903 		mutex_exit(&udp->udp_recv_lock);
8904 		mp->b_next = NULL;
8905 		putnext(RD(q), mp);
8906 		mutex_enter(&udp->udp_recv_lock);
8907 	}
8908 	udp->udp_fallback_queue_tail = udp->udp_fallback_queue_head;
8909 	/*
8910 	 * No longer a streams less socket
8911 	 */
8912 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8913 	connp->conn_flags &= ~IPCL_NONSTR;
8914 	rw_exit(&udp->udp_rwlock);
8915 
8916 	mutex_exit(&udp->udp_recv_lock);
8917 
8918 	ASSERT(connp->conn_ref >= 1);
8919 
8920 	return (0);
8921 }
8922 
8923 static int
8924 udp_do_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
8925 {
8926 	sin_t	*sin = (sin_t *)sa;
8927 	sin6_t	*sin6 = (sin6_t *)sa;
8928 
8929 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
8930 	ASSERT(udp != NULL);
8931 
8932 	if (udp->udp_state != TS_DATA_XFER)
8933 		return (ENOTCONN);
8934 
8935 	switch (udp->udp_family) {
8936 	case AF_INET:
8937 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8938 
8939 		if (*salenp < sizeof (sin_t))
8940 			return (EINVAL);
8941 
8942 		*salenp = sizeof (sin_t);
8943 		*sin = sin_null;
8944 		sin->sin_family = AF_INET;
8945 		sin->sin_port = udp->udp_dstport;
8946 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
8947 		break;
8948 	case AF_INET6:
8949 		if (*salenp < sizeof (sin6_t))
8950 			return (EINVAL);
8951 
8952 		*salenp = sizeof (sin6_t);
8953 		*sin6 = sin6_null;
8954 		sin6->sin6_family = AF_INET6;
8955 		sin6->sin6_port = udp->udp_dstport;
8956 		sin6->sin6_addr = udp->udp_v6dst;
8957 		sin6->sin6_flowinfo = udp->udp_flowinfo;
8958 		break;
8959 	}
8960 
8961 	return (0);
8962 }
8963 
8964 /* ARGSUSED */
8965 int
8966 udp_getpeername(sock_lower_handle_t  proto_handle, struct sockaddr *sa,
8967     socklen_t *salenp, cred_t *cr)
8968 {
8969 	conn_t	*connp = (conn_t *)proto_handle;
8970 	udp_t	*udp = connp->conn_udp;
8971 	int error;
8972 
8973 	/* All Solaris components should pass a cred for this operation. */
8974 	ASSERT(cr != NULL);
8975 
8976 	ASSERT(udp != NULL);
8977 
8978 	rw_enter(&udp->udp_rwlock, RW_READER);
8979 
8980 	error = udp_do_getpeername(udp, sa, salenp);
8981 
8982 	rw_exit(&udp->udp_rwlock);
8983 
8984 	return (error);
8985 }
8986 
8987 static int
8988 udp_do_getsockname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
8989 {
8990 	sin_t	*sin = (sin_t *)sa;
8991 	sin6_t	*sin6 = (sin6_t *)sa;
8992 
8993 	ASSERT(udp != NULL);
8994 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
8995 
8996 	switch (udp->udp_family) {
8997 	case AF_INET:
8998 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8999 
9000 		if (*salenp < sizeof (sin_t))
9001 			return (EINVAL);
9002 
9003 		*salenp = sizeof (sin_t);
9004 		*sin = sin_null;
9005 		sin->sin_family = AF_INET;
9006 		if (udp->udp_state == TS_UNBND) {
9007 			break;
9008 		}
9009 		sin->sin_port = udp->udp_port;
9010 
9011 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
9012 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9013 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
9014 		} else {
9015 			/*
9016 			 * INADDR_ANY
9017 			 * udp_v6src is not set, we might be bound to
9018 			 * broadcast/multicast. Use udp_bound_v6src as
9019 			 * local address instead (that could
9020 			 * also still be INADDR_ANY)
9021 			 */
9022 			sin->sin_addr.s_addr =
9023 			    V4_PART_OF_V6(udp->udp_bound_v6src);
9024 		}
9025 		break;
9026 
9027 	case AF_INET6:
9028 		if (*salenp < sizeof (sin6_t))
9029 			return (EINVAL);
9030 
9031 		*salenp = sizeof (sin6_t);
9032 		*sin6 = sin6_null;
9033 		sin6->sin6_family = AF_INET6;
9034 		if (udp->udp_state == TS_UNBND) {
9035 			break;
9036 		}
9037 		sin6->sin6_port = udp->udp_port;
9038 
9039 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9040 			sin6->sin6_addr = udp->udp_v6src;
9041 		} else {
9042 			/*
9043 			 * UNSPECIFIED
9044 			 * udp_v6src is not set, we might be bound to
9045 			 * broadcast/multicast. Use udp_bound_v6src as
9046 			 * local address instead (that could
9047 			 * also still be UNSPECIFIED)
9048 			 */
9049 			sin6->sin6_addr = udp->udp_bound_v6src;
9050 		}
9051 	}
9052 	return (0);
9053 }
9054 
9055 /* ARGSUSED */
9056 int
9057 udp_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
9058     socklen_t *salenp, cred_t *cr)
9059 {
9060 	conn_t	*connp = (conn_t *)proto_handle;
9061 	udp_t	*udp = connp->conn_udp;
9062 	int error;
9063 
9064 	/* All Solaris components should pass a cred for this operation. */
9065 	ASSERT(cr != NULL);
9066 
9067 	ASSERT(udp != NULL);
9068 	rw_enter(&udp->udp_rwlock, RW_READER);
9069 
9070 	error = udp_do_getsockname(udp, sa, salenp);
9071 
9072 	rw_exit(&udp->udp_rwlock);
9073 
9074 	return (error);
9075 }
9076 
9077 int
9078 udp_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9079     void *optvalp, socklen_t *optlen, cred_t *cr)
9080 {
9081 	conn_t		*connp = (conn_t *)proto_handle;
9082 	udp_t		*udp = connp->conn_udp;
9083 	int		error;
9084 	t_uscalar_t	max_optbuf_len;
9085 	void		*optvalp_buf;
9086 	int		len;
9087 
9088 	/* All Solaris components should pass a cred for this operation. */
9089 	ASSERT(cr != NULL);
9090 
9091 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
9092 	    udp_opt_obj.odb_opt_des_arr,
9093 	    udp_opt_obj.odb_opt_arr_cnt,
9094 	    udp_opt_obj.odb_topmost_tpiprovider,
9095 	    B_FALSE, B_TRUE, cr);
9096 	if (error != 0) {
9097 		if (error < 0)
9098 			error = proto_tlitosyserr(-error);
9099 		return (error);
9100 	}
9101 
9102 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
9103 	rw_enter(&udp->udp_rwlock, RW_READER);
9104 	len = udp_opt_get(connp, level, option_name, optvalp_buf);
9105 	rw_exit(&udp->udp_rwlock);
9106 
9107 	if (len < 0) {
9108 		/*
9109 		 * Pass on to IP
9110 		 */
9111 		kmem_free(optvalp_buf, max_optbuf_len);
9112 		return (ip_get_options(connp, level, option_name,
9113 		    optvalp, optlen, cr));
9114 	} else {
9115 		/*
9116 		 * update optlen and copy option value
9117 		 */
9118 		t_uscalar_t size = MIN(len, *optlen);
9119 		bcopy(optvalp_buf, optvalp, size);
9120 		bcopy(&size, optlen, sizeof (size));
9121 
9122 		kmem_free(optvalp_buf, max_optbuf_len);
9123 		return (0);
9124 	}
9125 }
9126 
9127 int
9128 udp_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9129     const void *optvalp, socklen_t optlen, cred_t *cr)
9130 {
9131 	conn_t		*connp = (conn_t *)proto_handle;
9132 	udp_t		*udp = connp->conn_udp;
9133 	int		error;
9134 
9135 	/* All Solaris components should pass a cred for this operation. */
9136 	ASSERT(cr != NULL);
9137 
9138 	error = proto_opt_check(level, option_name, optlen, NULL,
9139 	    udp_opt_obj.odb_opt_des_arr,
9140 	    udp_opt_obj.odb_opt_arr_cnt,
9141 	    udp_opt_obj.odb_topmost_tpiprovider,
9142 	    B_TRUE, B_FALSE, cr);
9143 
9144 	if (error != 0) {
9145 		if (error < 0)
9146 			error = proto_tlitosyserr(-error);
9147 		return (error);
9148 	}
9149 
9150 	rw_enter(&udp->udp_rwlock, RW_WRITER);
9151 	error = udp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level, option_name,
9152 	    optlen, (uchar_t *)optvalp, (uint_t *)&optlen, (uchar_t *)optvalp,
9153 	    NULL, cr);
9154 	rw_exit(&udp->udp_rwlock);
9155 
9156 	if (error < 0) {
9157 		/*
9158 		 * Pass on to ip
9159 		 */
9160 		error = ip_set_options(connp, level, option_name, optvalp,
9161 		    optlen, cr);
9162 	}
9163 
9164 	return (error);
9165 }
9166 
9167 void
9168 udp_clr_flowctrl(sock_lower_handle_t proto_handle)
9169 {
9170 	conn_t	*connp = (conn_t *)proto_handle;
9171 	udp_t	*udp = connp->conn_udp;
9172 
9173 	mutex_enter(&udp->udp_recv_lock);
9174 	connp->conn_flow_cntrld = B_FALSE;
9175 	mutex_exit(&udp->udp_recv_lock);
9176 }
9177 
9178 /* ARGSUSED */
9179 int
9180 udp_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
9181 {
9182 	conn_t	*connp = (conn_t *)proto_handle;
9183 
9184 	/* All Solaris components should pass a cred for this operation. */
9185 	ASSERT(cr != NULL);
9186 
9187 	/* shut down the send side */
9188 	if (how != SHUT_RD)
9189 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9190 		    SOCK_OPCTL_SHUT_SEND, 0);
9191 	/* shut down the recv side */
9192 	if (how != SHUT_WR)
9193 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9194 		    SOCK_OPCTL_SHUT_RECV, 0);
9195 	return (0);
9196 }
9197 
9198 int
9199 udp_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
9200     int mode, int32_t *rvalp, cred_t *cr)
9201 {
9202 	conn_t  	*connp = (conn_t *)proto_handle;
9203 	int		error;
9204 
9205 	/* All Solaris components should pass a cred for this operation. */
9206 	ASSERT(cr != NULL);
9207 
9208 	switch (cmd) {
9209 		case ND_SET:
9210 		case ND_GET:
9211 		case _SIOCSOCKFALLBACK:
9212 		case TI_GETPEERNAME:
9213 		case TI_GETMYNAME:
9214 			ip1dbg(("udp_ioctl: cmd 0x%x on non streams socket",
9215 			    cmd));
9216 			error = EINVAL;
9217 			break;
9218 		default:
9219 			/*
9220 			 * Pass on to IP using helper stream
9221 			 */
9222 			error = ldi_ioctl(connp->conn_helper_info->iphs_handle,
9223 			    cmd, arg, mode, cr, rvalp);
9224 			break;
9225 	}
9226 	return (error);
9227 }
9228 
9229 /* ARGSUSED */
9230 int
9231 udp_accept(sock_lower_handle_t lproto_handle,
9232     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
9233     cred_t *cr)
9234 {
9235 	return (EOPNOTSUPP);
9236 }
9237 
9238 /* ARGSUSED */
9239 int
9240 udp_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
9241 {
9242 	return (EOPNOTSUPP);
9243 }
9244 
9245 sock_downcalls_t sock_udp_downcalls = {
9246 	udp_activate,		/* sd_activate */
9247 	udp_accept,		/* sd_accept */
9248 	udp_bind,		/* sd_bind */
9249 	udp_listen,		/* sd_listen */
9250 	udp_connect,		/* sd_connect */
9251 	udp_getpeername,	/* sd_getpeername */
9252 	udp_getsockname,	/* sd_getsockname */
9253 	udp_getsockopt,		/* sd_getsockopt */
9254 	udp_setsockopt,		/* sd_setsockopt */
9255 	udp_send,		/* sd_send */
9256 	NULL,			/* sd_send_uio */
9257 	NULL,			/* sd_recv_uio */
9258 	NULL,			/* sd_poll */
9259 	udp_shutdown,		/* sd_shutdown */
9260 	udp_clr_flowctrl,	/* sd_setflowctrl */
9261 	udp_ioctl,		/* sd_ioctl */
9262 	udp_close		/* sd_close */
9263 };
9264