xref: /titanic_41/usr/src/uts/common/inet/udp/udp.c (revision 05fc5e5bf1bd330914669ff21940f37885c2b20a)
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 #define	NDD_TOO_QUICK_MSG \
137 	"ndd get info rate too high for non-privileged users, try again " \
138 	"later.\n"
139 #define	NDD_OUT_OF_BUF_MSG	"<< Out of buffer >>\n"
140 
141 /* Option processing attrs */
142 typedef struct udpattrs_s {
143 	union {
144 		ip6_pkt_t	*udpattr_ipp6;	/* For V6 */
145 		ip4_pkt_t 	*udpattr_ipp4;	/* For V4 */
146 	} udpattr_ippu;
147 #define	udpattr_ipp6 udpattr_ippu.udpattr_ipp6
148 #define	udpattr_ipp4 udpattr_ippu.udpattr_ipp4
149 	mblk_t		*udpattr_mb;
150 	boolean_t	udpattr_credset;
151 } udpattrs_t;
152 
153 static void	udp_addr_req(queue_t *q, mblk_t *mp);
154 static void	udp_tpi_bind(queue_t *q, mblk_t *mp);
155 static void	udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp);
156 static void	udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock);
157 static int	udp_build_hdrs(udp_t *udp);
158 static void	udp_capability_req(queue_t *q, mblk_t *mp);
159 static int	udp_tpi_close(queue_t *q, int flags);
160 static void	udp_tpi_connect(queue_t *q, mblk_t *mp);
161 static void	udp_tpi_disconnect(queue_t *q, mblk_t *mp);
162 static void	udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
163 		    int sys_error);
164 static void	udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive,
165 		    t_scalar_t tlierr, int unixerr);
166 static int	udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp,
167 		    cred_t *cr);
168 static int	udp_extra_priv_ports_add(queue_t *q, mblk_t *mp,
169 		    char *value, caddr_t cp, cred_t *cr);
170 static int	udp_extra_priv_ports_del(queue_t *q, mblk_t *mp,
171 		    char *value, caddr_t cp, cred_t *cr);
172 static void	udp_icmp_error(conn_t *, mblk_t *);
173 static void	udp_icmp_error_ipv6(conn_t *, mblk_t *);
174 static void	udp_info_req(queue_t *q, mblk_t *mp);
175 static void	udp_input(void *, mblk_t *, void *);
176 static mblk_t	*udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim,
177 		    t_scalar_t addr_length);
178 static void	udp_lrput(queue_t *, mblk_t *);
179 static void	udp_lwput(queue_t *, mblk_t *);
180 static int	udp_open(queue_t *q, dev_t *devp, int flag, int sflag,
181 		    cred_t *credp, boolean_t isv6);
182 static int	udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
183 		    cred_t *credp);
184 static int	udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
185 		    cred_t *credp);
186 static  int	udp_unitdata_opt_process(queue_t *q, mblk_t *mp,
187 		    int *errorp, udpattrs_t *udpattrs);
188 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
189 static int	udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
190 static boolean_t udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt);
191 static int	udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
192 		    cred_t *cr);
193 static void	udp_report_item(mblk_t *mp, udp_t *udp);
194 static int	udp_rinfop(queue_t *q, infod_t *dp);
195 static int	udp_rrw(queue_t *q, struiod_t *dp);
196 static int	udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
197 		    cred_t *cr);
198 static void	udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp,
199 		    ipha_t *ipha);
200 static void	udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr,
201 		    t_scalar_t destlen, t_scalar_t err);
202 static void	udp_tpi_unbind(queue_t *q, mblk_t *mp);
203 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port,
204     boolean_t random);
205 static mblk_t	*udp_output_v4(conn_t *, mblk_t *, ipaddr_t, uint16_t, uint_t,
206 		    int *, boolean_t, struct nmsghdr *, cred_t *, pid_t);
207 static mblk_t	*udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6,
208 		    int *error, struct nmsghdr *msg, cred_t *cr, pid_t pid);
209 static void	udp_wput_other(queue_t *q, mblk_t *mp);
210 static void	udp_wput_iocdata(queue_t *q, mblk_t *mp);
211 static void	udp_wput_fallback(queue_t *q, mblk_t *mp);
212 static size_t	udp_set_rcv_hiwat(udp_t *udp, size_t size);
213 
214 static void	*udp_stack_init(netstackid_t stackid, netstack_t *ns);
215 static void	udp_stack_fini(netstackid_t stackid, void *arg);
216 
217 static void	*udp_kstat_init(netstackid_t stackid);
218 static void	udp_kstat_fini(netstackid_t stackid, kstat_t *ksp);
219 static void	*udp_kstat2_init(netstackid_t, udp_stat_t *);
220 static void	udp_kstat2_fini(netstackid_t, kstat_t *);
221 static int	udp_kstat_update(kstat_t *kp, int rw);
222 
223 static void	udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp,
224 		    uint_t pkt_len);
225 static void	udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing);
226 static void	udp_xmit(queue_t *, mblk_t *, ire_t *ire, conn_t *, zoneid_t);
227 
228 static int	udp_send_connected(conn_t *, mblk_t *, struct nmsghdr *,
229 		    cred_t *, pid_t);
230 
231 /* Common routine for TPI and socket module */
232 static conn_t	*udp_do_open(cred_t *, boolean_t, int);
233 static void	udp_do_close(conn_t *);
234 static int	udp_do_bind(conn_t *, struct sockaddr *, socklen_t, cred_t *,
235     boolean_t);
236 static int	udp_do_unbind(conn_t *);
237 static int	udp_do_getsockname(udp_t *, struct sockaddr *, uint_t *);
238 static int	udp_do_getpeername(udp_t *, struct sockaddr *, uint_t *);
239 
240 int		udp_getsockname(sock_lower_handle_t,
241     struct sockaddr *, socklen_t *, cred_t *);
242 int		udp_getpeername(sock_lower_handle_t,
243     struct sockaddr *, socklen_t *, cred_t *);
244 static int	udp_do_connect(conn_t *, const struct sockaddr *, socklen_t);
245 static int	udp_post_ip_bind_connect(udp_t *, mblk_t *, int);
246 
247 #define	UDP_RECV_HIWATER	(56 * 1024)
248 #define	UDP_RECV_LOWATER	128
249 #define	UDP_XMIT_HIWATER	(56 * 1024)
250 #define	UDP_XMIT_LOWATER	1024
251 
252 /*
253  * The following is defined in tcp.c
254  */
255 extern int	(*cl_inet_connect2)(netstackid_t stack_id,
256 		    uint8_t protocol, boolean_t is_outgoing,
257 		    sa_family_t addr_family,
258 		    uint8_t *laddrp, in_port_t lport,
259 		    uint8_t *faddrp, in_port_t fport, void *args);
260 
261 /*
262  * Checks if the given destination addr/port is allowed out.
263  * If allowed, registers the (dest_addr/port, node_ID) mapping at Cluster.
264  * Called for each connect() and for sendto()/sendmsg() to a different
265  * destination.
266  * For connect(), called in udp_connect().
267  * For sendto()/sendmsg(), called in udp_output_v{4,6}().
268  *
269  * This macro assumes that the cl_inet_connect2 hook is not NULL.
270  * Please check this before calling this macro.
271  *
272  * void
273  * CL_INET_UDP_CONNECT(conn_t cp, udp_t *udp, boolean_t is_outgoing,
274  *     in6_addr_t *faddrp, in_port_t (or uint16_t) fport, int err);
275  */
276 #define	CL_INET_UDP_CONNECT(cp, udp, is_outgoing, faddrp, fport, err) {	\
277 	(err) = 0;							\
278 	/*								\
279 	 * Running in cluster mode - check and register active		\
280 	 * "connection" information					\
281 	 */								\
282 	if ((udp)->udp_ipversion == IPV4_VERSION)			\
283 		(err) = (*cl_inet_connect2)(				\
284 		    (cp)->conn_netstack->netstack_stackid,		\
285 		    IPPROTO_UDP, is_outgoing, AF_INET,			\
286 		    (uint8_t *)&((udp)->udp_v6src._S6_un._S6_u32[3]),	\
287 		    (udp)->udp_port,					\
288 		    (uint8_t *)&((faddrp)->_S6_un._S6_u32[3]),		\
289 		    (in_port_t)(fport), NULL);				\
290 	else								\
291 		(err) = (*cl_inet_connect2)(				\
292 		    (cp)->conn_netstack->netstack_stackid,		\
293 		    IPPROTO_UDP, is_outgoing, AF_INET6,			\
294 		    (uint8_t *)&((udp)->udp_v6src), (udp)->udp_port,	\
295 		    (uint8_t *)(faddrp), (in_port_t)(fport), NULL);	\
296 }
297 
298 static struct module_info udp_mod_info =  {
299 	UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER
300 };
301 
302 /*
303  * Entry points for UDP as a device.
304  * We have separate open functions for the /dev/udp and /dev/udp6 devices.
305  */
306 static struct qinit udp_rinitv4 = {
307 	NULL, NULL, udp_openv4, udp_tpi_close, NULL,
308 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
309 };
310 
311 static struct qinit udp_rinitv6 = {
312 	NULL, NULL, udp_openv6, udp_tpi_close, NULL,
313 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
314 };
315 
316 static struct qinit udp_winit = {
317 	(pfi_t)udp_wput, (pfi_t)ip_wsrv, NULL, NULL, NULL,
318 	&udp_mod_info, NULL, NULL, NULL, STRUIOT_NONE
319 };
320 
321 /* UDP entry point during fallback */
322 struct qinit udp_fallback_sock_winit = {
323 	(pfi_t)udp_wput_fallback, NULL, NULL, NULL, NULL, &udp_mod_info
324 };
325 
326 /*
327  * UDP needs to handle I_LINK and I_PLINK since ifconfig
328  * likes to use it as a place to hang the various streams.
329  */
330 static struct qinit udp_lrinit = {
331 	(pfi_t)udp_lrput, NULL, udp_openv4, udp_tpi_close, NULL,
332 	&udp_mod_info
333 };
334 
335 static struct qinit udp_lwinit = {
336 	(pfi_t)udp_lwput, NULL, udp_openv4, udp_tpi_close, NULL,
337 	&udp_mod_info
338 };
339 
340 /* For AF_INET aka /dev/udp */
341 struct streamtab udpinfov4 = {
342 	&udp_rinitv4, &udp_winit, &udp_lrinit, &udp_lwinit
343 };
344 
345 /* For AF_INET6 aka /dev/udp6 */
346 struct streamtab udpinfov6 = {
347 	&udp_rinitv6, &udp_winit, &udp_lrinit, &udp_lwinit
348 };
349 
350 static	sin_t	sin_null;	/* Zero address for quick clears */
351 static	sin6_t	sin6_null;	/* Zero address for quick clears */
352 
353 #define	UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH)
354 
355 /* Default structure copied into T_INFO_ACK messages */
356 static struct T_info_ack udp_g_t_info_ack_ipv4 = {
357 	T_INFO_ACK,
358 	UDP_MAXPACKET_IPV4,	/* TSDU_size. Excl. headers */
359 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
360 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
361 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
362 	sizeof (sin_t),	/* ADDR_size. */
363 	0,		/* OPT_size - not initialized here */
364 	UDP_MAXPACKET_IPV4,	/* TIDU_size.  Excl. headers */
365 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
366 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
367 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
368 };
369 
370 #define	UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN)
371 
372 static	struct T_info_ack udp_g_t_info_ack_ipv6 = {
373 	T_INFO_ACK,
374 	UDP_MAXPACKET_IPV6,	/* TSDU_size.  Excl. headers */
375 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
376 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
377 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
378 	sizeof (sin6_t), /* ADDR_size. */
379 	0,		/* OPT_size - not initialized here */
380 	UDP_MAXPACKET_IPV6,	/* TIDU_size. Excl. headers */
381 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
382 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
383 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
384 };
385 
386 /* largest UDP port number */
387 #define	UDP_MAX_PORT	65535
388 
389 /*
390  * Table of ND variables supported by udp.  These are loaded into us_nd
391  * in udp_open.
392  * All of these are alterable, within the min/max values given, at run time.
393  */
394 /* BEGIN CSTYLED */
395 udpparam_t udp_param_arr[] = {
396  /*min		max		value		name */
397  { 0L,		256,		32,		"udp_wroff_extra" },
398  { 1L,		255,		255,		"udp_ipv4_ttl" },
399  { 0,		IPV6_MAX_HOPS,	IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"},
400  { 1024,	(32 * 1024),	1024,		"udp_smallest_nonpriv_port" },
401  { 0,		1,		1,		"udp_do_checksum" },
402  { 1024,	UDP_MAX_PORT,	(32 * 1024),	"udp_smallest_anon_port" },
403  { 1024,	UDP_MAX_PORT,	UDP_MAX_PORT,	"udp_largest_anon_port" },
404  { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER,	"udp_xmit_hiwat"},
405  { 0,		     (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"},
406  { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER,	"udp_recv_hiwat"},
407  { 65536,	(1<<30),	2*1024*1024,	"udp_max_buf"},
408  { 100,		60000,		1000,		"udp_ndd_get_info_interval"},
409 };
410 /* END CSTYLED */
411 
412 /* Setable in /etc/system */
413 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */
414 uint32_t udp_random_anon_port = 1;
415 
416 /*
417  * Hook functions to enable cluster networking.
418  * On non-clustered systems these vectors must always be NULL
419  */
420 
421 void (*cl_inet_bind)(netstackid_t stack_id, uchar_t protocol,
422     sa_family_t addr_family, uint8_t *laddrp, in_port_t lport,
423     void *args) = NULL;
424 void (*cl_inet_unbind)(netstackid_t stack_id, uint8_t protocol,
425     sa_family_t addr_family, uint8_t *laddrp, in_port_t lport,
426     void *args) = NULL;
427 
428 typedef union T_primitives *t_primp_t;
429 
430 /*
431  * Return the next anonymous port in the privileged port range for
432  * bind checking.
433  *
434  * Trusted Extension (TX) notes: TX allows administrator to mark or
435  * reserve ports as Multilevel ports (MLP). MLP has special function
436  * on TX systems. Once a port is made MLP, it's not available as
437  * ordinary port. This creates "holes" in the port name space. It
438  * may be necessary to skip the "holes" find a suitable anon port.
439  */
440 static in_port_t
441 udp_get_next_priv_port(udp_t *udp)
442 {
443 	static in_port_t next_priv_port = IPPORT_RESERVED - 1;
444 	in_port_t nextport;
445 	boolean_t restart = B_FALSE;
446 	udp_stack_t *us = udp->udp_us;
447 
448 retry:
449 	if (next_priv_port < us->us_min_anonpriv_port ||
450 	    next_priv_port >= IPPORT_RESERVED) {
451 		next_priv_port = IPPORT_RESERVED - 1;
452 		if (restart)
453 			return (0);
454 		restart = B_TRUE;
455 	}
456 
457 	if (is_system_labeled() &&
458 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
459 	    next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) {
460 		next_priv_port = nextport;
461 		goto retry;
462 	}
463 
464 	return (next_priv_port--);
465 }
466 
467 /* UDP bind hash report triggered via the Named Dispatch mechanism. */
468 /* ARGSUSED */
469 static int
470 udp_bind_hash_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
471 {
472 	udp_fanout_t	*udpf;
473 	int		i;
474 	zoneid_t	zoneid;
475 	conn_t		*connp;
476 	udp_t		*udp;
477 	udp_stack_t	*us;
478 
479 	connp = Q_TO_CONN(q);
480 	udp = connp->conn_udp;
481 	us = udp->udp_us;
482 
483 	/* Refer to comments in udp_status_report(). */
484 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
485 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
486 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
487 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
488 			return (0);
489 		}
490 	}
491 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
492 		/* The following may work even if we cannot get a large buf. */
493 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
494 		return (0);
495 	}
496 
497 	(void) mi_mpprintf(mp,
498 	    "UDP     " MI_COL_HDRPAD_STR
499 	/*   12345678[89ABCDEF] */
500 	    " zone lport src addr        dest addr       port  state");
501 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
502 
503 	zoneid = connp->conn_zoneid;
504 
505 	for (i = 0; i < us->us_bind_fanout_size; i++) {
506 		udpf = &us->us_bind_fanout[i];
507 		mutex_enter(&udpf->uf_lock);
508 
509 		/* Print the hash index. */
510 		udp = udpf->uf_udp;
511 		if (zoneid != GLOBAL_ZONEID) {
512 			/* skip to first entry in this zone; might be none */
513 			while (udp != NULL &&
514 			    udp->udp_connp->conn_zoneid != zoneid)
515 				udp = udp->udp_bind_hash;
516 		}
517 		if (udp != NULL) {
518 			uint_t print_len, buf_len;
519 
520 			buf_len = mp->b_cont->b_datap->db_lim -
521 			    mp->b_cont->b_wptr;
522 			print_len = snprintf((char *)mp->b_cont->b_wptr,
523 			    buf_len, "%d\n", i);
524 			if (print_len < buf_len) {
525 				mp->b_cont->b_wptr += print_len;
526 			} else {
527 				mp->b_cont->b_wptr += buf_len;
528 			}
529 			for (; udp != NULL; udp = udp->udp_bind_hash) {
530 				if (zoneid == GLOBAL_ZONEID ||
531 				    zoneid == udp->udp_connp->conn_zoneid)
532 					udp_report_item(mp->b_cont, udp);
533 			}
534 		}
535 		mutex_exit(&udpf->uf_lock);
536 	}
537 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
538 	return (0);
539 }
540 
541 /*
542  * Hash list removal routine for udp_t structures.
543  */
544 static void
545 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock)
546 {
547 	udp_t	*udpnext;
548 	kmutex_t *lockp;
549 	udp_stack_t *us = udp->udp_us;
550 
551 	if (udp->udp_ptpbhn == NULL)
552 		return;
553 
554 	/*
555 	 * Extract the lock pointer in case there are concurrent
556 	 * hash_remove's for this instance.
557 	 */
558 	ASSERT(udp->udp_port != 0);
559 	if (!caller_holds_lock) {
560 		lockp = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
561 		    us->us_bind_fanout_size)].uf_lock;
562 		ASSERT(lockp != NULL);
563 		mutex_enter(lockp);
564 	}
565 	if (udp->udp_ptpbhn != NULL) {
566 		udpnext = udp->udp_bind_hash;
567 		if (udpnext != NULL) {
568 			udpnext->udp_ptpbhn = udp->udp_ptpbhn;
569 			udp->udp_bind_hash = NULL;
570 		}
571 		*udp->udp_ptpbhn = udpnext;
572 		udp->udp_ptpbhn = NULL;
573 	}
574 	if (!caller_holds_lock) {
575 		mutex_exit(lockp);
576 	}
577 }
578 
579 static void
580 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp)
581 {
582 	udp_t	**udpp;
583 	udp_t	*udpnext;
584 
585 	ASSERT(MUTEX_HELD(&uf->uf_lock));
586 	ASSERT(udp->udp_ptpbhn == NULL);
587 	udpp = &uf->uf_udp;
588 	udpnext = udpp[0];
589 	if (udpnext != NULL) {
590 		/*
591 		 * If the new udp bound to the INADDR_ANY address
592 		 * and the first one in the list is not bound to
593 		 * INADDR_ANY we skip all entries until we find the
594 		 * first one bound to INADDR_ANY.
595 		 * This makes sure that applications binding to a
596 		 * specific address get preference over those binding to
597 		 * INADDR_ANY.
598 		 */
599 		if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) &&
600 		    !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) {
601 			while ((udpnext = udpp[0]) != NULL &&
602 			    !V6_OR_V4_INADDR_ANY(
603 			    udpnext->udp_bound_v6src)) {
604 				udpp = &(udpnext->udp_bind_hash);
605 			}
606 			if (udpnext != NULL)
607 				udpnext->udp_ptpbhn = &udp->udp_bind_hash;
608 		} else {
609 			udpnext->udp_ptpbhn = &udp->udp_bind_hash;
610 		}
611 	}
612 	udp->udp_bind_hash = udpnext;
613 	udp->udp_ptpbhn = udpp;
614 	udpp[0] = udp;
615 }
616 
617 /*
618  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
619  * passed to udp_wput.
620  * It associates a port number and local address with the stream.
621  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP
622  * protocol type (IPPROTO_UDP) placed in the message following the address.
623  * A T_BIND_ACK message is passed upstream when ip acknowledges the request.
624  * (Called as writer.)
625  *
626  * Note that UDP over IPv4 and IPv6 sockets can use the same port number
627  * without setting SO_REUSEADDR. This is needed so that they
628  * can be viewed as two independent transport protocols.
629  * However, anonymouns ports are allocated from the same range to avoid
630  * duplicating the us->us_next_port_to_try.
631  */
632 static void
633 udp_tpi_bind(queue_t *q, mblk_t *mp)
634 {
635 	sin_t		*sin;
636 	sin6_t		*sin6;
637 	mblk_t		*mp1;
638 	struct T_bind_req *tbr;
639 	conn_t		*connp;
640 	udp_t		*udp;
641 	int		error;
642 	struct sockaddr	*sa;
643 
644 	connp = Q_TO_CONN(q);
645 	udp = connp->conn_udp;
646 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
647 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
648 		    "udp_bind: bad req, len %u",
649 		    (uint_t)(mp->b_wptr - mp->b_rptr));
650 		udp_err_ack(q, mp, TPROTO, 0);
651 		return;
652 	}
653 	if (udp->udp_state != TS_UNBND) {
654 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
655 		    "udp_bind: bad state, %u", udp->udp_state);
656 		udp_err_ack(q, mp, TOUTSTATE, 0);
657 		return;
658 	}
659 	/*
660 	 * Reallocate the message to make sure we have enough room for an
661 	 * address and the protocol type.
662 	 */
663 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
664 	if (!mp1) {
665 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
666 		return;
667 	}
668 
669 	mp = mp1;
670 
671 	/* Reset the message type in preparation for shipping it back. */
672 	DB_TYPE(mp) = M_PCPROTO;
673 
674 	tbr = (struct T_bind_req *)mp->b_rptr;
675 	switch (tbr->ADDR_length) {
676 	case 0:			/* Request for a generic port */
677 		tbr->ADDR_offset = sizeof (struct T_bind_req);
678 		if (udp->udp_family == AF_INET) {
679 			tbr->ADDR_length = sizeof (sin_t);
680 			sin = (sin_t *)&tbr[1];
681 			*sin = sin_null;
682 			sin->sin_family = AF_INET;
683 			mp->b_wptr = (uchar_t *)&sin[1];
684 			sa = (struct sockaddr *)sin;
685 		} else {
686 			ASSERT(udp->udp_family == AF_INET6);
687 			tbr->ADDR_length = sizeof (sin6_t);
688 			sin6 = (sin6_t *)&tbr[1];
689 			*sin6 = sin6_null;
690 			sin6->sin6_family = AF_INET6;
691 			mp->b_wptr = (uchar_t *)&sin6[1];
692 			sa = (struct sockaddr *)sin6;
693 		}
694 		break;
695 
696 	case sizeof (sin_t):	/* Complete IPv4 address */
697 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
698 		    sizeof (sin_t));
699 		if (sa == NULL || !OK_32PTR((char *)sa)) {
700 			udp_err_ack(q, mp, TSYSERR, EINVAL);
701 			return;
702 		}
703 		if (udp->udp_family != AF_INET ||
704 		    sa->sa_family != AF_INET) {
705 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
706 			return;
707 		}
708 		break;
709 
710 	case sizeof (sin6_t):	/* complete IPv6 address */
711 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
712 		    sizeof (sin6_t));
713 		if (sa == NULL || !OK_32PTR((char *)sa)) {
714 			udp_err_ack(q, mp, TSYSERR, EINVAL);
715 			return;
716 		}
717 		if (udp->udp_family != AF_INET6 ||
718 		    sa->sa_family != AF_INET6) {
719 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
720 			return;
721 		}
722 		break;
723 
724 	default:		/* Invalid request */
725 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
726 		    "udp_bind: bad ADDR_length length %u", tbr->ADDR_length);
727 		udp_err_ack(q, mp, TBADADDR, 0);
728 		return;
729 	}
730 
731 
732 	cred_t *cr = DB_CREDDEF(mp, connp->conn_cred);
733 	error = udp_do_bind(connp, sa, tbr->ADDR_length, cr,
734 	    tbr->PRIM_type != O_T_BIND_REQ);
735 
736 	if (error != 0) {
737 		if (error > 0) {
738 			udp_err_ack(q, mp, TSYSERR, error);
739 		} else {
740 			udp_err_ack(q, mp, -error, 0);
741 		}
742 	} else {
743 		tbr->PRIM_type = T_BIND_ACK;
744 		qreply(q, mp);
745 	}
746 }
747 
748 /*
749  * This routine handles each T_CONN_REQ message passed to udp.  It
750  * associates a default destination address with the stream.
751  *
752  * This routine sends down a T_BIND_REQ to IP with the following mblks:
753  *	T_BIND_REQ	- specifying local and remote address/port
754  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
755  *	T_OK_ACK	- for the T_CONN_REQ
756  *	T_CONN_CON	- to keep the TPI user happy
757  *
758  * The connect completes in udp_do_connect.
759  * When a T_BIND_ACK is received information is extracted from the IRE
760  * and the two appended messages are sent to the TPI user.
761  * Should udp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
762  * convert it to an error ack for the appropriate primitive.
763  */
764 static void
765 udp_tpi_connect(queue_t *q, mblk_t *mp)
766 {
767 	mblk_t	*mp1;
768 	udp_t	*udp;
769 	conn_t	*connp = Q_TO_CONN(q);
770 	int	error;
771 	socklen_t	len;
772 	struct sockaddr		*sa;
773 	struct T_conn_req	*tcr;
774 
775 	udp = connp->conn_udp;
776 	tcr = (struct T_conn_req *)mp->b_rptr;
777 
778 	/* A bit of sanity checking */
779 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
780 		udp_err_ack(q, mp, TPROTO, 0);
781 		return;
782 	}
783 
784 	if (tcr->OPT_length != 0) {
785 		udp_err_ack(q, mp, TBADOPT, 0);
786 		return;
787 	}
788 
789 	/*
790 	 * Determine packet type based on type of address passed in
791 	 * the request should contain an IPv4 or IPv6 address.
792 	 * Make sure that address family matches the type of
793 	 * family of the the address passed down
794 	 */
795 	len = tcr->DEST_length;
796 	switch (tcr->DEST_length) {
797 	default:
798 		udp_err_ack(q, mp, TBADADDR, 0);
799 		return;
800 
801 	case sizeof (sin_t):
802 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
803 		    sizeof (sin_t));
804 		break;
805 
806 	case sizeof (sin6_t):
807 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
808 		    sizeof (sin6_t));
809 		break;
810 	}
811 
812 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
813 	if (error != 0) {
814 		udp_err_ack(q, mp, TSYSERR, error);
815 		return;
816 	}
817 
818 	/*
819 	 * We have to send a connection confirmation to
820 	 * keep TLI happy.
821 	 */
822 	if (udp->udp_family == AF_INET) {
823 		mp1 = mi_tpi_conn_con(NULL, (char *)sa,
824 		    sizeof (sin_t), NULL, 0);
825 	} else {
826 		mp1 = mi_tpi_conn_con(NULL, (char *)sa,
827 		    sizeof (sin6_t), NULL, 0);
828 	}
829 	if (mp1 == NULL) {
830 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
831 		return;
832 	}
833 
834 	/*
835 	 * ok_ack for T_CONN_REQ
836 	 */
837 	mp = mi_tpi_ok_ack_alloc(mp);
838 	if (mp == NULL) {
839 		/* Unable to reuse the T_CONN_REQ for the ack. */
840 		freemsg(mp1);
841 		udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
842 		return;
843 	}
844 
845 	error = udp_do_connect(connp, sa, len);
846 	if (error != 0) {
847 		freeb(mp1);
848 		if (error < 0)
849 			udp_err_ack(q, mp, -error, 0);
850 		else
851 			udp_err_ack(q, mp, TSYSERR, error);
852 	} else {
853 		putnext(connp->conn_rq, mp);
854 		putnext(connp->conn_rq, mp1);
855 	}
856 }
857 
858 static int
859 udp_tpi_close(queue_t *q, int flags)
860 {
861 	conn_t	*connp;
862 
863 	if (flags & SO_FALLBACK) {
864 		/*
865 		 * stream is being closed while in fallback
866 		 * simply free the resources that were allocated
867 		 */
868 		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
869 		qprocsoff(q);
870 		goto done;
871 	}
872 
873 	connp = Q_TO_CONN(q);
874 	udp_do_close(connp);
875 done:
876 	q->q_ptr = WR(q)->q_ptr = NULL;
877 	return (0);
878 }
879 
880 /*
881  * Called in the close path to quiesce the conn
882  */
883 void
884 udp_quiesce_conn(conn_t *connp)
885 {
886 	udp_t	*udp = connp->conn_udp;
887 
888 	if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) {
889 		/*
890 		 * Running in cluster mode - register unbind information
891 		 */
892 		if (udp->udp_ipversion == IPV4_VERSION) {
893 			(*cl_inet_unbind)(
894 			    connp->conn_netstack->netstack_stackid,
895 			    IPPROTO_UDP, AF_INET,
896 			    (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))),
897 			    (in_port_t)udp->udp_port, NULL);
898 		} else {
899 			(*cl_inet_unbind)(
900 			    connp->conn_netstack->netstack_stackid,
901 			    IPPROTO_UDP, AF_INET6,
902 			    (uint8_t *)(&(udp->udp_v6src)),
903 			    (in_port_t)udp->udp_port, NULL);
904 		}
905 	}
906 
907 	udp_bind_hash_remove(udp, B_FALSE);
908 
909 }
910 
911 void
912 udp_close_free(conn_t *connp)
913 {
914 	udp_t *udp = connp->conn_udp;
915 
916 	/* If there are any options associated with the stream, free them. */
917 	if (udp->udp_ip_snd_options != NULL) {
918 		mi_free((char *)udp->udp_ip_snd_options);
919 		udp->udp_ip_snd_options = NULL;
920 		udp->udp_ip_snd_options_len = 0;
921 	}
922 
923 	if (udp->udp_ip_rcv_options != NULL) {
924 		mi_free((char *)udp->udp_ip_rcv_options);
925 		udp->udp_ip_rcv_options = NULL;
926 		udp->udp_ip_rcv_options_len = 0;
927 	}
928 
929 	/* Free memory associated with sticky options */
930 	if (udp->udp_sticky_hdrs_len != 0) {
931 		kmem_free(udp->udp_sticky_hdrs,
932 		    udp->udp_sticky_hdrs_len);
933 		udp->udp_sticky_hdrs = NULL;
934 		udp->udp_sticky_hdrs_len = 0;
935 	}
936 
937 	ip6_pkt_free(&udp->udp_sticky_ipp);
938 
939 	/*
940 	 * Clear any fields which the kmem_cache constructor clears.
941 	 * Only udp_connp needs to be preserved.
942 	 * TBD: We should make this more efficient to avoid clearing
943 	 * everything.
944 	 */
945 	ASSERT(udp->udp_connp == connp);
946 	bzero(udp, sizeof (udp_t));
947 	udp->udp_connp = connp;
948 }
949 
950 static int
951 udp_do_disconnect(conn_t *connp)
952 {
953 	udp_t	*udp;
954 	mblk_t	*ire_mp;
955 	udp_fanout_t *udpf;
956 	udp_stack_t *us;
957 	int	error;
958 
959 	udp = connp->conn_udp;
960 	us = udp->udp_us;
961 	rw_enter(&udp->udp_rwlock, RW_WRITER);
962 	if (udp->udp_state != TS_DATA_XFER || udp->udp_pending_op != -1) {
963 		rw_exit(&udp->udp_rwlock);
964 		return (-TOUTSTATE);
965 	}
966 	udp->udp_pending_op = T_DISCON_REQ;
967 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
968 	    us->us_bind_fanout_size)];
969 	mutex_enter(&udpf->uf_lock);
970 	udp->udp_v6src = udp->udp_bound_v6src;
971 	udp->udp_state = TS_IDLE;
972 	mutex_exit(&udpf->uf_lock);
973 
974 	if (udp->udp_family == AF_INET6) {
975 		/* Rebuild the header template */
976 		error = udp_build_hdrs(udp);
977 		if (error != 0) {
978 			udp->udp_pending_op = -1;
979 			rw_exit(&udp->udp_rwlock);
980 			return (error);
981 		}
982 	}
983 
984 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
985 	if (ire_mp == NULL) {
986 		mutex_enter(&udpf->uf_lock);
987 		udp->udp_pending_op = -1;
988 		mutex_exit(&udpf->uf_lock);
989 		rw_exit(&udp->udp_rwlock);
990 		return (ENOMEM);
991 	}
992 
993 	rw_exit(&udp->udp_rwlock);
994 
995 	if (udp->udp_family == AF_INET6) {
996 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, IPPROTO_UDP,
997 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
998 	} else {
999 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, IPPROTO_UDP,
1000 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port, B_TRUE);
1001 	}
1002 
1003 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
1004 }
1005 
1006 
1007 static void
1008 udp_tpi_disconnect(queue_t *q, mblk_t *mp)
1009 {
1010 	conn_t	*connp = Q_TO_CONN(q);
1011 	int	error;
1012 
1013 	/*
1014 	 * Allocate the largest primitive we need to send back
1015 	 * T_error_ack is > than T_ok_ack
1016 	 */
1017 	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
1018 	if (mp == NULL) {
1019 		/* Unable to reuse the T_DISCON_REQ for the ack. */
1020 		udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
1021 		return;
1022 	}
1023 
1024 	error = udp_do_disconnect(connp);
1025 
1026 	if (error != 0) {
1027 		if (error < 0) {
1028 			udp_err_ack(q, mp, -error, 0);
1029 		} else {
1030 			udp_err_ack(q, mp, TSYSERR, error);
1031 		}
1032 	} else {
1033 		mp = mi_tpi_ok_ack_alloc(mp);
1034 		ASSERT(mp != NULL);
1035 		qreply(q, mp);
1036 	}
1037 }
1038 
1039 int
1040 udp_disconnect(conn_t *connp)
1041 {
1042 	int error;
1043 	udp_t *udp = connp->conn_udp;
1044 
1045 	udp->udp_dgram_errind = B_FALSE;
1046 
1047 	error = udp_do_disconnect(connp);
1048 
1049 	if (error < 0)
1050 		error = proto_tlitosyserr(-error);
1051 
1052 	return (error);
1053 }
1054 
1055 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
1056 static void
1057 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
1058 {
1059 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1060 		qreply(q, mp);
1061 }
1062 
1063 /* Shorthand to generate and send TPI error acks to our client */
1064 static void
1065 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error,
1066     int sys_error)
1067 {
1068 	struct T_error_ack	*teackp;
1069 
1070 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1071 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1072 		teackp = (struct T_error_ack *)mp->b_rptr;
1073 		teackp->ERROR_prim = primitive;
1074 		teackp->TLI_error = t_error;
1075 		teackp->UNIX_error = sys_error;
1076 		qreply(q, mp);
1077 	}
1078 }
1079 
1080 /*ARGSUSED*/
1081 static int
1082 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
1083 {
1084 	int i;
1085 	udp_t		*udp = Q_TO_UDP(q);
1086 	udp_stack_t *us = udp->udp_us;
1087 
1088 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1089 		if (us->us_epriv_ports[i] != 0)
1090 			(void) mi_mpprintf(mp, "%d ", us->us_epriv_ports[i]);
1091 	}
1092 	return (0);
1093 }
1094 
1095 /* ARGSUSED */
1096 static int
1097 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1098     cred_t *cr)
1099 {
1100 	long	new_value;
1101 	int	i;
1102 	udp_t		*udp = Q_TO_UDP(q);
1103 	udp_stack_t *us = udp->udp_us;
1104 
1105 	/*
1106 	 * Fail the request if the new value does not lie within the
1107 	 * port number limits.
1108 	 */
1109 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1110 	    new_value <= 0 || new_value >= 65536) {
1111 		return (EINVAL);
1112 	}
1113 
1114 	/* Check if the value is already in the list */
1115 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1116 		if (new_value == us->us_epriv_ports[i]) {
1117 			return (EEXIST);
1118 		}
1119 	}
1120 	/* Find an empty slot */
1121 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1122 		if (us->us_epriv_ports[i] == 0)
1123 			break;
1124 	}
1125 	if (i == us->us_num_epriv_ports) {
1126 		return (EOVERFLOW);
1127 	}
1128 
1129 	/* Set the new value */
1130 	us->us_epriv_ports[i] = (in_port_t)new_value;
1131 	return (0);
1132 }
1133 
1134 /* ARGSUSED */
1135 static int
1136 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1137     cred_t *cr)
1138 {
1139 	long	new_value;
1140 	int	i;
1141 	udp_t		*udp = Q_TO_UDP(q);
1142 	udp_stack_t *us = udp->udp_us;
1143 
1144 	/*
1145 	 * Fail the request if the new value does not lie within the
1146 	 * port number limits.
1147 	 */
1148 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1149 	    new_value <= 0 || new_value >= 65536) {
1150 		return (EINVAL);
1151 	}
1152 
1153 	/* Check that the value is already in the list */
1154 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1155 		if (us->us_epriv_ports[i] == new_value)
1156 			break;
1157 	}
1158 	if (i == us->us_num_epriv_ports) {
1159 		return (ESRCH);
1160 	}
1161 
1162 	/* Clear the value */
1163 	us->us_epriv_ports[i] = 0;
1164 	return (0);
1165 }
1166 
1167 /* At minimum we need 4 bytes of UDP header */
1168 #define	ICMP_MIN_UDP_HDR	4
1169 
1170 /*
1171  * udp_icmp_error is called by udp_input to process ICMP msgs. passed up by IP.
1172  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1173  * Assumes that IP has pulled up everything up to and including the ICMP header.
1174  */
1175 static void
1176 udp_icmp_error(conn_t *connp, mblk_t *mp)
1177 			    {
1178 	icmph_t *icmph;
1179 	ipha_t	*ipha;
1180 	int	iph_hdr_length;
1181 	udpha_t	*udpha;
1182 	sin_t	sin;
1183 	sin6_t	sin6;
1184 	mblk_t	*mp1;
1185 	int	error = 0;
1186 	udp_t	*udp = connp->conn_udp;
1187 
1188 	mp1 = NULL;
1189 	ipha = (ipha_t *)mp->b_rptr;
1190 
1191 	ASSERT(OK_32PTR(mp->b_rptr));
1192 
1193 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1194 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1195 		udp_icmp_error_ipv6(connp, mp);
1196 		return;
1197 	}
1198 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
1199 
1200 	/* Skip past the outer IP and ICMP headers */
1201 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1202 	icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length];
1203 	ipha = (ipha_t *)&icmph[1];
1204 
1205 	/* Skip past the inner IP and find the ULP header */
1206 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1207 	udpha = (udpha_t *)((char *)ipha + iph_hdr_length);
1208 
1209 	switch (icmph->icmph_type) {
1210 	case ICMP_DEST_UNREACHABLE:
1211 		switch (icmph->icmph_code) {
1212 		case ICMP_FRAGMENTATION_NEEDED:
1213 			/*
1214 			 * IP has already adjusted the path MTU.
1215 			 */
1216 			break;
1217 		case ICMP_PORT_UNREACHABLE:
1218 		case ICMP_PROTOCOL_UNREACHABLE:
1219 			error = ECONNREFUSED;
1220 			break;
1221 		default:
1222 			/* Transient errors */
1223 			break;
1224 		}
1225 		break;
1226 	default:
1227 		/* Transient errors */
1228 		break;
1229 	}
1230 	if (error == 0) {
1231 		freemsg(mp);
1232 		return;
1233 	}
1234 
1235 	/*
1236 	 * Deliver T_UDERROR_IND when the application has asked for it.
1237 	 * The socket layer enables this automatically when connected.
1238 	 */
1239 	if (!udp->udp_dgram_errind) {
1240 		freemsg(mp);
1241 		return;
1242 	}
1243 
1244 
1245 	switch (udp->udp_family) {
1246 	case AF_INET:
1247 		sin = sin_null;
1248 		sin.sin_family = AF_INET;
1249 		sin.sin_addr.s_addr = ipha->ipha_dst;
1250 		sin.sin_port = udpha->uha_dst_port;
1251 		if (IPCL_IS_NONSTR(connp)) {
1252 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1253 			if (udp->udp_state == TS_DATA_XFER) {
1254 				if (sin.sin_port == udp->udp_dstport &&
1255 				    sin.sin_addr.s_addr ==
1256 				    V4_PART_OF_V6(udp->udp_v6dst)) {
1257 
1258 					rw_exit(&udp->udp_rwlock);
1259 					(*connp->conn_upcalls->su_set_error)
1260 					    (connp->conn_upper_handle, error);
1261 					goto done;
1262 				}
1263 			} else {
1264 				udp->udp_delayed_error = error;
1265 				*((sin_t *)&udp->udp_delayed_addr) = sin;
1266 			}
1267 			rw_exit(&udp->udp_rwlock);
1268 		} else {
1269 			mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t),
1270 			    NULL, 0, error);
1271 		}
1272 		break;
1273 	case AF_INET6:
1274 		sin6 = sin6_null;
1275 		sin6.sin6_family = AF_INET6;
1276 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
1277 		sin6.sin6_port = udpha->uha_dst_port;
1278 		if (IPCL_IS_NONSTR(connp)) {
1279 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1280 			if (udp->udp_state == TS_DATA_XFER) {
1281 				if (sin6.sin6_port == udp->udp_dstport &&
1282 				    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1283 				    &udp->udp_v6dst)) {
1284 					rw_exit(&udp->udp_rwlock);
1285 					(*connp->conn_upcalls->su_set_error)
1286 					    (connp->conn_upper_handle, error);
1287 					goto done;
1288 				}
1289 			} else {
1290 				udp->udp_delayed_error = error;
1291 				*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1292 			}
1293 			rw_exit(&udp->udp_rwlock);
1294 		} else {
1295 
1296 			mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1297 			    NULL, 0, error);
1298 		}
1299 		break;
1300 	}
1301 	if (mp1 != NULL)
1302 		putnext(connp->conn_rq, mp1);
1303 done:
1304 	freemsg(mp);
1305 }
1306 
1307 /*
1308  * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6.
1309  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1310  * Assumes that IP has pulled up all the extension headers as well as the
1311  * ICMPv6 header.
1312  */
1313 static void
1314 udp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
1315 {
1316 	icmp6_t		*icmp6;
1317 	ip6_t		*ip6h, *outer_ip6h;
1318 	uint16_t	iph_hdr_length;
1319 	uint8_t		*nexthdrp;
1320 	udpha_t		*udpha;
1321 	sin6_t		sin6;
1322 	mblk_t		*mp1;
1323 	int		error = 0;
1324 	udp_t		*udp = connp->conn_udp;
1325 	udp_stack_t	*us = udp->udp_us;
1326 
1327 	outer_ip6h = (ip6_t *)mp->b_rptr;
1328 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1329 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1330 	else
1331 		iph_hdr_length = IPV6_HDR_LEN;
1332 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1333 	ip6h = (ip6_t *)&icmp6[1];
1334 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1335 		freemsg(mp);
1336 		return;
1337 	}
1338 	udpha = (udpha_t *)((char *)ip6h + iph_hdr_length);
1339 
1340 	switch (icmp6->icmp6_type) {
1341 	case ICMP6_DST_UNREACH:
1342 		switch (icmp6->icmp6_code) {
1343 		case ICMP6_DST_UNREACH_NOPORT:
1344 			error = ECONNREFUSED;
1345 			break;
1346 		case ICMP6_DST_UNREACH_ADMIN:
1347 		case ICMP6_DST_UNREACH_NOROUTE:
1348 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1349 		case ICMP6_DST_UNREACH_ADDR:
1350 			/* Transient errors */
1351 			break;
1352 		default:
1353 			break;
1354 		}
1355 		break;
1356 	case ICMP6_PACKET_TOO_BIG: {
1357 		struct T_unitdata_ind	*tudi;
1358 		struct T_opthdr		*toh;
1359 		size_t			udi_size;
1360 		mblk_t			*newmp;
1361 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1362 		    sizeof (struct ip6_mtuinfo);
1363 		sin6_t			*sin6;
1364 		struct ip6_mtuinfo	*mtuinfo;
1365 
1366 		/*
1367 		 * If the application has requested to receive path mtu
1368 		 * information, send up an empty message containing an
1369 		 * IPV6_PATHMTU ancillary data item.
1370 		 */
1371 		if (!udp->udp_ipv6_recvpathmtu)
1372 			break;
1373 
1374 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1375 		    opt_length;
1376 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1377 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
1378 			break;
1379 		}
1380 
1381 		/*
1382 		 * newmp->b_cont is left to NULL on purpose.  This is an
1383 		 * empty message containing only ancillary data.
1384 		 */
1385 		newmp->b_datap->db_type = M_PROTO;
1386 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1387 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1388 		tudi->PRIM_type = T_UNITDATA_IND;
1389 		tudi->SRC_length = sizeof (sin6_t);
1390 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1391 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1392 		tudi->OPT_length = opt_length;
1393 
1394 		sin6 = (sin6_t *)&tudi[1];
1395 		bzero(sin6, sizeof (sin6_t));
1396 		sin6->sin6_family = AF_INET6;
1397 		sin6->sin6_addr = udp->udp_v6dst;
1398 
1399 		toh = (struct T_opthdr *)&sin6[1];
1400 		toh->level = IPPROTO_IPV6;
1401 		toh->name = IPV6_PATHMTU;
1402 		toh->len = opt_length;
1403 		toh->status = 0;
1404 
1405 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1406 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1407 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1408 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1409 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1410 		/*
1411 		 * We've consumed everything we need from the original
1412 		 * message.  Free it, then send our empty message.
1413 		 */
1414 		freemsg(mp);
1415 		if (!IPCL_IS_NONSTR(connp)) {
1416 			putnext(connp->conn_rq, newmp);
1417 		} else {
1418 			(*connp->conn_upcalls->su_recv)
1419 			    (connp->conn_upper_handle, newmp, 0, 0, &error,
1420 			    NULL);
1421 		}
1422 		return;
1423 	}
1424 	case ICMP6_TIME_EXCEEDED:
1425 		/* Transient errors */
1426 		break;
1427 	case ICMP6_PARAM_PROB:
1428 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1429 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1430 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1431 		    (uchar_t *)nexthdrp) {
1432 			error = ECONNREFUSED;
1433 			break;
1434 		}
1435 		break;
1436 	}
1437 	if (error == 0) {
1438 		freemsg(mp);
1439 		return;
1440 	}
1441 
1442 	/*
1443 	 * Deliver T_UDERROR_IND when the application has asked for it.
1444 	 * The socket layer enables this automatically when connected.
1445 	 */
1446 	if (!udp->udp_dgram_errind) {
1447 		freemsg(mp);
1448 		return;
1449 	}
1450 
1451 	sin6 = sin6_null;
1452 	sin6.sin6_family = AF_INET6;
1453 	sin6.sin6_addr = ip6h->ip6_dst;
1454 	sin6.sin6_port = udpha->uha_dst_port;
1455 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1456 
1457 	if (IPCL_IS_NONSTR(connp)) {
1458 		rw_enter(&udp->udp_rwlock, RW_WRITER);
1459 		if (udp->udp_state == TS_DATA_XFER) {
1460 			if (sin6.sin6_port == udp->udp_dstport &&
1461 			    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1462 			    &udp->udp_v6dst)) {
1463 				rw_exit(&udp->udp_rwlock);
1464 				(*connp->conn_upcalls->su_set_error)
1465 				    (connp->conn_upper_handle, error);
1466 				goto done;
1467 			}
1468 		} else {
1469 			udp->udp_delayed_error = error;
1470 			*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1471 		}
1472 		rw_exit(&udp->udp_rwlock);
1473 	} else {
1474 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1475 		    NULL, 0, error);
1476 		if (mp1 != NULL)
1477 			putnext(connp->conn_rq, mp1);
1478 	}
1479 
1480 done:
1481 	freemsg(mp);
1482 }
1483 
1484 /*
1485  * This routine responds to T_ADDR_REQ messages.  It is called by udp_wput.
1486  * The local address is filled in if endpoint is bound. The remote address
1487  * is filled in if remote address has been precified ("connected endpoint")
1488  * (The concept of connected CLTS sockets is alien to published TPI
1489  *  but we support it anyway).
1490  */
1491 static void
1492 udp_addr_req(queue_t *q, mblk_t *mp)
1493 {
1494 	sin_t	*sin;
1495 	sin6_t	*sin6;
1496 	mblk_t	*ackmp;
1497 	struct T_addr_ack *taa;
1498 	udp_t	*udp = Q_TO_UDP(q);
1499 
1500 	/* Make it large enough for worst case */
1501 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1502 	    2 * sizeof (sin6_t), 1);
1503 	if (ackmp == NULL) {
1504 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1505 		return;
1506 	}
1507 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1508 
1509 	bzero(taa, sizeof (struct T_addr_ack));
1510 	ackmp->b_wptr = (uchar_t *)&taa[1];
1511 
1512 	taa->PRIM_type = T_ADDR_ACK;
1513 	ackmp->b_datap->db_type = M_PCPROTO;
1514 	rw_enter(&udp->udp_rwlock, RW_READER);
1515 	/*
1516 	 * Note: Following code assumes 32 bit alignment of basic
1517 	 * data structures like sin_t and struct T_addr_ack.
1518 	 */
1519 	if (udp->udp_state != TS_UNBND) {
1520 		/*
1521 		 * Fill in local address first
1522 		 */
1523 		taa->LOCADDR_offset = sizeof (*taa);
1524 		if (udp->udp_family == AF_INET) {
1525 			taa->LOCADDR_length = sizeof (sin_t);
1526 			sin = (sin_t *)&taa[1];
1527 			/* Fill zeroes and then initialize non-zero fields */
1528 			*sin = sin_null;
1529 			sin->sin_family = AF_INET;
1530 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
1531 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1532 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src,
1533 				    sin->sin_addr.s_addr);
1534 			} else {
1535 				/*
1536 				 * INADDR_ANY
1537 				 * udp_v6src is not set, we might be bound to
1538 				 * broadcast/multicast. Use udp_bound_v6src as
1539 				 * local address instead (that could
1540 				 * also still be INADDR_ANY)
1541 				 */
1542 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src,
1543 				    sin->sin_addr.s_addr);
1544 			}
1545 			sin->sin_port = udp->udp_port;
1546 			ackmp->b_wptr = (uchar_t *)&sin[1];
1547 			if (udp->udp_state == TS_DATA_XFER) {
1548 				/*
1549 				 * connected, fill remote address too
1550 				 */
1551 				taa->REMADDR_length = sizeof (sin_t);
1552 				/* assumed 32-bit alignment */
1553 				taa->REMADDR_offset = taa->LOCADDR_offset +
1554 				    taa->LOCADDR_length;
1555 
1556 				sin = (sin_t *)(ackmp->b_rptr +
1557 				    taa->REMADDR_offset);
1558 				/* initialize */
1559 				*sin = sin_null;
1560 				sin->sin_family = AF_INET;
1561 				sin->sin_addr.s_addr =
1562 				    V4_PART_OF_V6(udp->udp_v6dst);
1563 				sin->sin_port = udp->udp_dstport;
1564 				ackmp->b_wptr = (uchar_t *)&sin[1];
1565 			}
1566 		} else {
1567 			taa->LOCADDR_length = sizeof (sin6_t);
1568 			sin6 = (sin6_t *)&taa[1];
1569 			/* Fill zeroes and then initialize non-zero fields */
1570 			*sin6 = sin6_null;
1571 			sin6->sin6_family = AF_INET6;
1572 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1573 				sin6->sin6_addr = udp->udp_v6src;
1574 			} else {
1575 				/*
1576 				 * UNSPECIFIED
1577 				 * udp_v6src is not set, we might be bound to
1578 				 * broadcast/multicast. Use udp_bound_v6src as
1579 				 * local address instead (that could
1580 				 * also still be UNSPECIFIED)
1581 				 */
1582 				sin6->sin6_addr =
1583 				    udp->udp_bound_v6src;
1584 			}
1585 			sin6->sin6_port = udp->udp_port;
1586 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1587 			if (udp->udp_state == TS_DATA_XFER) {
1588 				/*
1589 				 * connected, fill remote address too
1590 				 */
1591 				taa->REMADDR_length = sizeof (sin6_t);
1592 				/* assumed 32-bit alignment */
1593 				taa->REMADDR_offset = taa->LOCADDR_offset +
1594 				    taa->LOCADDR_length;
1595 
1596 				sin6 = (sin6_t *)(ackmp->b_rptr +
1597 				    taa->REMADDR_offset);
1598 				/* initialize */
1599 				*sin6 = sin6_null;
1600 				sin6->sin6_family = AF_INET6;
1601 				sin6->sin6_addr = udp->udp_v6dst;
1602 				sin6->sin6_port =  udp->udp_dstport;
1603 				ackmp->b_wptr = (uchar_t *)&sin6[1];
1604 			}
1605 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1606 		}
1607 	}
1608 	rw_exit(&udp->udp_rwlock);
1609 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1610 	qreply(q, ackmp);
1611 }
1612 
1613 static void
1614 udp_copy_info(struct T_info_ack *tap, udp_t *udp)
1615 {
1616 	if (udp->udp_family == AF_INET) {
1617 		*tap = udp_g_t_info_ack_ipv4;
1618 	} else {
1619 		*tap = udp_g_t_info_ack_ipv6;
1620 	}
1621 	tap->CURRENT_state = udp->udp_state;
1622 	tap->OPT_size = udp_max_optsize;
1623 }
1624 
1625 static void
1626 udp_do_capability_ack(udp_t *udp, struct T_capability_ack *tcap,
1627     t_uscalar_t cap_bits1)
1628 {
1629 	tcap->CAP_bits1 = 0;
1630 
1631 	if (cap_bits1 & TC1_INFO) {
1632 		udp_copy_info(&tcap->INFO_ack, udp);
1633 		tcap->CAP_bits1 |= TC1_INFO;
1634 	}
1635 }
1636 
1637 /*
1638  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1639  * udp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1640  * udp_g_t_info_ack.  The current state of the stream is copied from
1641  * udp_state.
1642  */
1643 static void
1644 udp_capability_req(queue_t *q, mblk_t *mp)
1645 {
1646 	t_uscalar_t		cap_bits1;
1647 	struct T_capability_ack	*tcap;
1648 	udp_t	*udp = Q_TO_UDP(q);
1649 
1650 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1651 
1652 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1653 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1654 	if (!mp)
1655 		return;
1656 
1657 	tcap = (struct T_capability_ack *)mp->b_rptr;
1658 	udp_do_capability_ack(udp, tcap, cap_bits1);
1659 
1660 	qreply(q, mp);
1661 }
1662 
1663 /*
1664  * This routine responds to T_INFO_REQ messages.  It is called by udp_wput.
1665  * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack.
1666  * The current state of the stream is copied from udp_state.
1667  */
1668 static void
1669 udp_info_req(queue_t *q, mblk_t *mp)
1670 {
1671 	udp_t *udp = Q_TO_UDP(q);
1672 
1673 	/* Create a T_INFO_ACK message. */
1674 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1675 	    T_INFO_ACK);
1676 	if (!mp)
1677 		return;
1678 	udp_copy_info((struct T_info_ack *)mp->b_rptr, udp);
1679 	qreply(q, mp);
1680 }
1681 
1682 /*
1683  * IP recognizes seven kinds of bind requests:
1684  *
1685  * - A zero-length address binds only to the protocol number.
1686  *
1687  * - A 4-byte address is treated as a request to
1688  * validate that the address is a valid local IPv4
1689  * address, appropriate for an application to bind to.
1690  * IP does the verification, but does not make any note
1691  * of the address at this time.
1692  *
1693  * - A 16-byte address contains is treated as a request
1694  * to validate a local IPv6 address, as the 4-byte
1695  * address case above.
1696  *
1697  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
1698  * use it for the inbound fanout of packets.
1699  *
1700  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
1701  * use it for the inbound fanout of packets.
1702  *
1703  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
1704  * information consisting of local and remote addresses
1705  * and ports.  In this case, the addresses are both
1706  * validated as appropriate for this operation, and, if
1707  * so, the information is retained for use in the
1708  * inbound fanout.
1709  *
1710  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
1711  * fanout information, like the 12-byte case above.
1712  *
1713  * IP will also fill in the IRE request mblk with information
1714  * regarding our peer.  In all cases, we notify IP of our protocol
1715  * type by appending a single protocol byte to the bind request.
1716  */
1717 static mblk_t *
1718 udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, t_scalar_t addr_length)
1719 {
1720 	char	*cp;
1721 	mblk_t	*mp;
1722 	struct T_bind_req *tbr;
1723 	ipa_conn_t	*ac;
1724 	ipa6_conn_t	*ac6;
1725 	sin_t		*sin;
1726 	sin6_t		*sin6;
1727 
1728 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
1729 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
1730 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
1731 	if (!mp)
1732 		return (mp);
1733 	mp->b_datap->db_type = M_PROTO;
1734 	tbr = (struct T_bind_req *)mp->b_rptr;
1735 	tbr->PRIM_type = bind_prim;
1736 	tbr->ADDR_offset = sizeof (*tbr);
1737 	tbr->CONIND_number = 0;
1738 	tbr->ADDR_length = addr_length;
1739 	cp = (char *)&tbr[1];
1740 	switch (addr_length) {
1741 	case sizeof (ipa_conn_t):
1742 		ASSERT(udp->udp_family == AF_INET);
1743 		/* Append a request for an IRE */
1744 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1745 		if (!mp->b_cont) {
1746 			freemsg(mp);
1747 			return (NULL);
1748 		}
1749 		mp->b_cont->b_wptr += sizeof (ire_t);
1750 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1751 
1752 		/* cp known to be 32 bit aligned */
1753 		ac = (ipa_conn_t *)cp;
1754 		ac->ac_laddr = V4_PART_OF_V6(udp->udp_v6src);
1755 		ac->ac_faddr = V4_PART_OF_V6(udp->udp_v6dst);
1756 		ac->ac_fport = udp->udp_dstport;
1757 		ac->ac_lport = udp->udp_port;
1758 		break;
1759 
1760 	case sizeof (ipa6_conn_t):
1761 		ASSERT(udp->udp_family == AF_INET6);
1762 		/* Append a request for an IRE */
1763 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1764 		if (!mp->b_cont) {
1765 			freemsg(mp);
1766 			return (NULL);
1767 		}
1768 		mp->b_cont->b_wptr += sizeof (ire_t);
1769 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1770 
1771 		/* cp known to be 32 bit aligned */
1772 		ac6 = (ipa6_conn_t *)cp;
1773 		ac6->ac6_laddr = udp->udp_v6src;
1774 		ac6->ac6_faddr = udp->udp_v6dst;
1775 		ac6->ac6_fport = udp->udp_dstport;
1776 		ac6->ac6_lport = udp->udp_port;
1777 		break;
1778 
1779 	case sizeof (sin_t):
1780 		ASSERT(udp->udp_family == AF_INET);
1781 		/* Append a request for an IRE */
1782 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1783 		if (!mp->b_cont) {
1784 			freemsg(mp);
1785 			return (NULL);
1786 		}
1787 		mp->b_cont->b_wptr += sizeof (ire_t);
1788 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1789 
1790 		sin = (sin_t *)cp;
1791 		*sin = sin_null;
1792 		sin->sin_family = AF_INET;
1793 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_bound_v6src);
1794 		sin->sin_port = udp->udp_port;
1795 		break;
1796 
1797 	case sizeof (sin6_t):
1798 		ASSERT(udp->udp_family == AF_INET6);
1799 		/* Append a request for an IRE */
1800 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1801 		if (!mp->b_cont) {
1802 			freemsg(mp);
1803 			return (NULL);
1804 		}
1805 		mp->b_cont->b_wptr += sizeof (ire_t);
1806 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1807 
1808 		sin6 = (sin6_t *)cp;
1809 		*sin6 = sin6_null;
1810 		sin6->sin6_family = AF_INET6;
1811 		sin6->sin6_addr = udp->udp_bound_v6src;
1812 		sin6->sin6_port = udp->udp_port;
1813 		break;
1814 	}
1815 	/* Add protocol number to end */
1816 	cp[addr_length] = (char)IPPROTO_UDP;
1817 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
1818 	return (mp);
1819 }
1820 
1821 /* For /dev/udp aka AF_INET open */
1822 static int
1823 udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1824 {
1825 	return (udp_open(q, devp, flag, sflag, credp, B_FALSE));
1826 }
1827 
1828 /* For /dev/udp6 aka AF_INET6 open */
1829 static int
1830 udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1831 {
1832 	return (udp_open(q, devp, flag, sflag, credp, B_TRUE));
1833 }
1834 
1835 /*
1836  * This is the open routine for udp.  It allocates a udp_t structure for
1837  * the stream and, on the first open of the module, creates an ND table.
1838  */
1839 /*ARGSUSED2*/
1840 static int
1841 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1842     boolean_t isv6)
1843 {
1844 	int		error;
1845 	udp_t		*udp;
1846 	conn_t		*connp;
1847 	dev_t		conn_dev;
1848 	udp_stack_t	*us;
1849 	vmem_t		*minor_arena;
1850 
1851 	TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q);
1852 
1853 	/* If the stream is already open, return immediately. */
1854 	if (q->q_ptr != NULL)
1855 		return (0);
1856 
1857 	if (sflag == MODOPEN)
1858 		return (EINVAL);
1859 
1860 	if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
1861 	    ((conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
1862 		minor_arena = ip_minor_arena_la;
1863 	} else {
1864 		/*
1865 		 * Either minor numbers in the large arena were exhausted
1866 		 * or a non socket application is doing the open.
1867 		 * Try to allocate from the small arena.
1868 		 */
1869 		if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0)
1870 			return (EBUSY);
1871 
1872 		minor_arena = ip_minor_arena_sa;
1873 	}
1874 
1875 	if (flag & SO_FALLBACK) {
1876 		/*
1877 		 * Non streams socket needs a stream to fallback to
1878 		 */
1879 		RD(q)->q_ptr = (void *)conn_dev;
1880 		WR(q)->q_qinfo = &udp_fallback_sock_winit;
1881 		WR(q)->q_ptr = (void *)minor_arena;
1882 		qprocson(q);
1883 		return (0);
1884 	}
1885 
1886 	connp = udp_do_open(credp, isv6, KM_SLEEP);
1887 	if (connp == NULL) {
1888 		inet_minor_free(minor_arena, conn_dev);
1889 		return (ENOMEM);
1890 	}
1891 	udp = connp->conn_udp;
1892 	us = udp->udp_us;
1893 
1894 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1895 	connp->conn_dev = conn_dev;
1896 	connp->conn_minor_arena = minor_arena;
1897 
1898 	/*
1899 	 * Initialize the udp_t structure for this stream.
1900 	 */
1901 	q->q_ptr = connp;
1902 	WR(q)->q_ptr = connp;
1903 	connp->conn_rq = q;
1904 	connp->conn_wq = WR(q);
1905 
1906 	rw_enter(&udp->udp_rwlock, RW_WRITER);
1907 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
1908 	ASSERT(connp->conn_udp == udp);
1909 	ASSERT(udp->udp_connp == connp);
1910 
1911 	if (flag & SO_SOCKSTR) {
1912 		connp->conn_flags |= IPCL_SOCKET;
1913 		udp->udp_issocket = B_TRUE;
1914 		udp->udp_direct_sockfs = B_TRUE;
1915 	}
1916 
1917 	q->q_hiwat = us->us_recv_hiwat;
1918 	WR(q)->q_hiwat = us->us_xmit_hiwat;
1919 	WR(q)->q_lowat = us->us_xmit_lowat;
1920 
1921 	qprocson(q);
1922 
1923 	if (udp->udp_family == AF_INET6) {
1924 		/* Build initial header template for transmit */
1925 		if ((error = udp_build_hdrs(udp)) != 0) {
1926 			rw_exit(&udp->udp_rwlock);
1927 			qprocsoff(q);
1928 			inet_minor_free(minor_arena, conn_dev);
1929 			ipcl_conn_destroy(connp);
1930 			return (error);
1931 		}
1932 	}
1933 	rw_exit(&udp->udp_rwlock);
1934 
1935 	/* Set the Stream head write offset and high watermark. */
1936 	(void) proto_set_tx_wroff(q, connp,
1937 	    udp->udp_max_hdr_len + us->us_wroff_extra);
1938 	/* XXX udp_set_rcv_hiwat() doesn't hold the lock, is it a bug??? */
1939 	(void) proto_set_rx_hiwat(q, connp, udp_set_rcv_hiwat(udp, q->q_hiwat));
1940 
1941 	mutex_enter(&connp->conn_lock);
1942 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1943 	mutex_exit(&connp->conn_lock);
1944 	return (0);
1945 }
1946 
1947 /*
1948  * Which UDP options OK to set through T_UNITDATA_REQ...
1949  */
1950 /* ARGSUSED */
1951 static boolean_t
1952 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1953 {
1954 	return (B_TRUE);
1955 }
1956 
1957 /*
1958  * This routine gets default values of certain options whose default
1959  * values are maintained by protcol specific code
1960  */
1961 /* ARGSUSED */
1962 int
1963 udp_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
1964 {
1965 	udp_t		*udp = Q_TO_UDP(q);
1966 	udp_stack_t *us = udp->udp_us;
1967 	int *i1 = (int *)ptr;
1968 
1969 	switch (level) {
1970 	case IPPROTO_IP:
1971 		switch (name) {
1972 		case IP_MULTICAST_TTL:
1973 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1974 			return (sizeof (uchar_t));
1975 		case IP_MULTICAST_LOOP:
1976 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1977 			return (sizeof (uchar_t));
1978 		}
1979 		break;
1980 	case IPPROTO_IPV6:
1981 		switch (name) {
1982 		case IPV6_MULTICAST_HOPS:
1983 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1984 			return (sizeof (int));
1985 		case IPV6_MULTICAST_LOOP:
1986 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1987 			return (sizeof (int));
1988 		case IPV6_UNICAST_HOPS:
1989 			*i1 = us->us_ipv6_hoplimit;
1990 			return (sizeof (int));
1991 		}
1992 		break;
1993 	}
1994 	return (-1);
1995 }
1996 
1997 /*
1998  * This routine retrieves the current status of socket options.
1999  * It returns the size of the option retrieved.
2000  */
2001 static int
2002 udp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
2003 {
2004 	udp_t		*udp = connp->conn_udp;
2005 	udp_stack_t	*us = udp->udp_us;
2006 	int		*i1 = (int *)ptr;
2007 	ip6_pkt_t 	*ipp = &udp->udp_sticky_ipp;
2008 	int		len;
2009 
2010 	ASSERT(RW_READ_HELD(&udp->udp_rwlock));
2011 	switch (level) {
2012 	case SOL_SOCKET:
2013 		switch (name) {
2014 		case SO_DEBUG:
2015 			*i1 = udp->udp_debug;
2016 			break;	/* goto sizeof (int) option return */
2017 		case SO_REUSEADDR:
2018 			*i1 = udp->udp_reuseaddr;
2019 			break;	/* goto sizeof (int) option return */
2020 		case SO_TYPE:
2021 			*i1 = SOCK_DGRAM;
2022 			break;	/* goto sizeof (int) option return */
2023 
2024 		/*
2025 		 * The following three items are available here,
2026 		 * but are only meaningful to IP.
2027 		 */
2028 		case SO_DONTROUTE:
2029 			*i1 = udp->udp_dontroute;
2030 			break;	/* goto sizeof (int) option return */
2031 		case SO_USELOOPBACK:
2032 			*i1 = udp->udp_useloopback;
2033 			break;	/* goto sizeof (int) option return */
2034 		case SO_BROADCAST:
2035 			*i1 = udp->udp_broadcast;
2036 			break;	/* goto sizeof (int) option return */
2037 
2038 		case SO_SNDBUF:
2039 			*i1 = udp->udp_xmit_hiwat;
2040 			break;	/* goto sizeof (int) option return */
2041 		case SO_RCVBUF:
2042 			*i1 = udp->udp_rcv_disply_hiwat;
2043 			break;	/* goto sizeof (int) option return */
2044 		case SO_DGRAM_ERRIND:
2045 			*i1 = udp->udp_dgram_errind;
2046 			break;	/* goto sizeof (int) option return */
2047 		case SO_RECVUCRED:
2048 			*i1 = udp->udp_recvucred;
2049 			break;	/* goto sizeof (int) option return */
2050 		case SO_TIMESTAMP:
2051 			*i1 = udp->udp_timestamp;
2052 			break;	/* goto sizeof (int) option return */
2053 		case SO_ANON_MLP:
2054 			*i1 = connp->conn_anon_mlp;
2055 			break;	/* goto sizeof (int) option return */
2056 		case SO_MAC_EXEMPT:
2057 			*i1 = connp->conn_mac_exempt;
2058 			break;	/* goto sizeof (int) option return */
2059 		case SO_ALLZONES:
2060 			*i1 = connp->conn_allzones;
2061 			break;	/* goto sizeof (int) option return */
2062 		case SO_EXCLBIND:
2063 			*i1 = udp->udp_exclbind ? SO_EXCLBIND : 0;
2064 			break;
2065 		case SO_PROTOTYPE:
2066 			*i1 = IPPROTO_UDP;
2067 			break;
2068 		case SO_DOMAIN:
2069 			*i1 = udp->udp_family;
2070 			break;
2071 		default:
2072 			return (-1);
2073 		}
2074 		break;
2075 	case IPPROTO_IP:
2076 		if (udp->udp_family != AF_INET)
2077 			return (-1);
2078 		switch (name) {
2079 		case IP_OPTIONS:
2080 		case T_IP_OPTIONS:
2081 			len = udp->udp_ip_rcv_options_len - udp->udp_label_len;
2082 			if (len > 0) {
2083 				bcopy(udp->udp_ip_rcv_options +
2084 				    udp->udp_label_len, ptr, len);
2085 			}
2086 			return (len);
2087 		case IP_TOS:
2088 		case T_IP_TOS:
2089 			*i1 = (int)udp->udp_type_of_service;
2090 			break;	/* goto sizeof (int) option return */
2091 		case IP_TTL:
2092 			*i1 = (int)udp->udp_ttl;
2093 			break;	/* goto sizeof (int) option return */
2094 		case IP_DHCPINIT_IF:
2095 			return (-EINVAL);
2096 		case IP_NEXTHOP:
2097 		case IP_RECVPKTINFO:
2098 			/*
2099 			 * This also handles IP_PKTINFO.
2100 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2101 			 * Differentiation is based on the size of the argument
2102 			 * passed in.
2103 			 * This option is handled in IP which will return an
2104 			 * error for IP_PKTINFO as it's not supported as a
2105 			 * sticky option.
2106 			 */
2107 			return (-EINVAL);
2108 		case IP_MULTICAST_IF:
2109 			/* 0 address if not set */
2110 			*(ipaddr_t *)ptr = udp->udp_multicast_if_addr;
2111 			return (sizeof (ipaddr_t));
2112 		case IP_MULTICAST_TTL:
2113 			*(uchar_t *)ptr = udp->udp_multicast_ttl;
2114 			return (sizeof (uchar_t));
2115 		case IP_MULTICAST_LOOP:
2116 			*ptr = connp->conn_multicast_loop;
2117 			return (sizeof (uint8_t));
2118 		case IP_RECVOPTS:
2119 			*i1 = udp->udp_recvopts;
2120 			break;	/* goto sizeof (int) option return */
2121 		case IP_RECVDSTADDR:
2122 			*i1 = udp->udp_recvdstaddr;
2123 			break;	/* goto sizeof (int) option return */
2124 		case IP_RECVIF:
2125 			*i1 = udp->udp_recvif;
2126 			break;	/* goto sizeof (int) option return */
2127 		case IP_RECVSLLA:
2128 			*i1 = udp->udp_recvslla;
2129 			break;	/* goto sizeof (int) option return */
2130 		case IP_RECVTTL:
2131 			*i1 = udp->udp_recvttl;
2132 			break;	/* goto sizeof (int) option return */
2133 		case IP_ADD_MEMBERSHIP:
2134 		case IP_DROP_MEMBERSHIP:
2135 		case IP_BLOCK_SOURCE:
2136 		case IP_UNBLOCK_SOURCE:
2137 		case IP_ADD_SOURCE_MEMBERSHIP:
2138 		case IP_DROP_SOURCE_MEMBERSHIP:
2139 		case MCAST_JOIN_GROUP:
2140 		case MCAST_LEAVE_GROUP:
2141 		case MCAST_BLOCK_SOURCE:
2142 		case MCAST_UNBLOCK_SOURCE:
2143 		case MCAST_JOIN_SOURCE_GROUP:
2144 		case MCAST_LEAVE_SOURCE_GROUP:
2145 			/* cannot "get" the value for these */
2146 			return (-1);
2147 		case IP_BOUND_IF:
2148 			/* Zero if not set */
2149 			*i1 = udp->udp_bound_if;
2150 			break;	/* goto sizeof (int) option return */
2151 		case IP_UNSPEC_SRC:
2152 			*i1 = udp->udp_unspec_source;
2153 			break;	/* goto sizeof (int) option return */
2154 		case IP_BROADCAST_TTL:
2155 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
2156 			return (sizeof (uchar_t));
2157 		default:
2158 			return (-1);
2159 		}
2160 		break;
2161 	case IPPROTO_IPV6:
2162 		if (udp->udp_family != AF_INET6)
2163 			return (-1);
2164 		switch (name) {
2165 		case IPV6_UNICAST_HOPS:
2166 			*i1 = (unsigned int)udp->udp_ttl;
2167 			break;	/* goto sizeof (int) option return */
2168 		case IPV6_MULTICAST_IF:
2169 			/* 0 index if not set */
2170 			*i1 = udp->udp_multicast_if_index;
2171 			break;	/* goto sizeof (int) option return */
2172 		case IPV6_MULTICAST_HOPS:
2173 			*i1 = udp->udp_multicast_ttl;
2174 			break;	/* goto sizeof (int) option return */
2175 		case IPV6_MULTICAST_LOOP:
2176 			*i1 = connp->conn_multicast_loop;
2177 			break;	/* goto sizeof (int) option return */
2178 		case IPV6_JOIN_GROUP:
2179 		case IPV6_LEAVE_GROUP:
2180 		case MCAST_JOIN_GROUP:
2181 		case MCAST_LEAVE_GROUP:
2182 		case MCAST_BLOCK_SOURCE:
2183 		case MCAST_UNBLOCK_SOURCE:
2184 		case MCAST_JOIN_SOURCE_GROUP:
2185 		case MCAST_LEAVE_SOURCE_GROUP:
2186 			/* cannot "get" the value for these */
2187 			return (-1);
2188 		case IPV6_BOUND_IF:
2189 			/* Zero if not set */
2190 			*i1 = udp->udp_bound_if;
2191 			break;	/* goto sizeof (int) option return */
2192 		case IPV6_UNSPEC_SRC:
2193 			*i1 = udp->udp_unspec_source;
2194 			break;	/* goto sizeof (int) option return */
2195 		case IPV6_RECVPKTINFO:
2196 			*i1 = udp->udp_ip_recvpktinfo;
2197 			break;	/* goto sizeof (int) option return */
2198 		case IPV6_RECVTCLASS:
2199 			*i1 = udp->udp_ipv6_recvtclass;
2200 			break;	/* goto sizeof (int) option return */
2201 		case IPV6_RECVPATHMTU:
2202 			*i1 = udp->udp_ipv6_recvpathmtu;
2203 			break;	/* goto sizeof (int) option return */
2204 		case IPV6_RECVHOPLIMIT:
2205 			*i1 = udp->udp_ipv6_recvhoplimit;
2206 			break;	/* goto sizeof (int) option return */
2207 		case IPV6_RECVHOPOPTS:
2208 			*i1 = udp->udp_ipv6_recvhopopts;
2209 			break;	/* goto sizeof (int) option return */
2210 		case IPV6_RECVDSTOPTS:
2211 			*i1 = udp->udp_ipv6_recvdstopts;
2212 			break;	/* goto sizeof (int) option return */
2213 		case _OLD_IPV6_RECVDSTOPTS:
2214 			*i1 = udp->udp_old_ipv6_recvdstopts;
2215 			break;	/* goto sizeof (int) option return */
2216 		case IPV6_RECVRTHDRDSTOPTS:
2217 			*i1 = udp->udp_ipv6_recvrthdrdstopts;
2218 			break;	/* goto sizeof (int) option return */
2219 		case IPV6_RECVRTHDR:
2220 			*i1 = udp->udp_ipv6_recvrthdr;
2221 			break;	/* goto sizeof (int) option return */
2222 		case IPV6_PKTINFO: {
2223 			/* XXX assumes that caller has room for max size! */
2224 			struct in6_pktinfo *pkti;
2225 
2226 			pkti = (struct in6_pktinfo *)ptr;
2227 			if (ipp->ipp_fields & IPPF_IFINDEX)
2228 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
2229 			else
2230 				pkti->ipi6_ifindex = 0;
2231 			if (ipp->ipp_fields & IPPF_ADDR)
2232 				pkti->ipi6_addr = ipp->ipp_addr;
2233 			else
2234 				pkti->ipi6_addr = ipv6_all_zeros;
2235 			return (sizeof (struct in6_pktinfo));
2236 		}
2237 		case IPV6_TCLASS:
2238 			if (ipp->ipp_fields & IPPF_TCLASS)
2239 				*i1 = ipp->ipp_tclass;
2240 			else
2241 				*i1 = IPV6_FLOW_TCLASS(
2242 				    IPV6_DEFAULT_VERS_AND_FLOW);
2243 			break;	/* goto sizeof (int) option return */
2244 		case IPV6_NEXTHOP: {
2245 			sin6_t *sin6 = (sin6_t *)ptr;
2246 
2247 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2248 				return (0);
2249 			*sin6 = sin6_null;
2250 			sin6->sin6_family = AF_INET6;
2251 			sin6->sin6_addr = ipp->ipp_nexthop;
2252 			return (sizeof (sin6_t));
2253 		}
2254 		case IPV6_HOPOPTS:
2255 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2256 				return (0);
2257 			if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6)
2258 				return (0);
2259 			/*
2260 			 * The cipso/label option is added by kernel.
2261 			 * User is not usually aware of this option.
2262 			 * We copy out the hbh opt after the label option.
2263 			 */
2264 			bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6,
2265 			    ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2266 			if (udp->udp_label_len_v6 > 0) {
2267 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2268 				ptr[1] = (ipp->ipp_hopoptslen -
2269 				    udp->udp_label_len_v6 + 7) / 8 - 1;
2270 			}
2271 			return (ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2272 		case IPV6_RTHDRDSTOPTS:
2273 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2274 				return (0);
2275 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2276 			return (ipp->ipp_rtdstoptslen);
2277 		case IPV6_RTHDR:
2278 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2279 				return (0);
2280 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2281 			return (ipp->ipp_rthdrlen);
2282 		case IPV6_DSTOPTS:
2283 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
2284 				return (0);
2285 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2286 			return (ipp->ipp_dstoptslen);
2287 		case IPV6_PATHMTU:
2288 			return (ip_fill_mtuinfo(&udp->udp_v6dst,
2289 			    udp->udp_dstport, (struct ip6_mtuinfo *)ptr,
2290 			    us->us_netstack));
2291 		default:
2292 			return (-1);
2293 		}
2294 		break;
2295 	case IPPROTO_UDP:
2296 		switch (name) {
2297 		case UDP_ANONPRIVBIND:
2298 			*i1 = udp->udp_anon_priv_bind;
2299 			break;
2300 		case UDP_EXCLBIND:
2301 			*i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0;
2302 			break;
2303 		case UDP_RCVHDR:
2304 			*i1 = udp->udp_rcvhdr ? 1 : 0;
2305 			break;
2306 		case UDP_NAT_T_ENDPOINT:
2307 			*i1 = udp->udp_nat_t_endpoint;
2308 			break;
2309 		default:
2310 			return (-1);
2311 		}
2312 		break;
2313 	default:
2314 		return (-1);
2315 	}
2316 	return (sizeof (int));
2317 }
2318 
2319 int
2320 udp_tpi_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2321 {
2322 	udp_t   *udp;
2323 	int	err;
2324 
2325 	udp = Q_TO_UDP(q);
2326 
2327 	rw_enter(&udp->udp_rwlock, RW_READER);
2328 	err = udp_opt_get(Q_TO_CONN(q), level, name, ptr);
2329 	rw_exit(&udp->udp_rwlock);
2330 	return (err);
2331 }
2332 
2333 /*
2334  * This routine sets socket options.
2335  */
2336 /* ARGSUSED */
2337 static int
2338 udp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
2339     uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
2340     void *thisdg_attrs, boolean_t checkonly)
2341 {
2342 	udpattrs_t *attrs = thisdg_attrs;
2343 	int	*i1 = (int *)invalp;
2344 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2345 	udp_t	*udp = connp->conn_udp;
2346 	udp_stack_t	*us = udp->udp_us;
2347 	int	error;
2348 	uint_t	newlen;
2349 	size_t	sth_wroff;
2350 
2351 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
2352 	/*
2353 	 * For fixed length options, no sanity check
2354 	 * of passed in length is done. It is assumed *_optcom_req()
2355 	 * routines do the right thing.
2356 	 */
2357 	switch (level) {
2358 	case SOL_SOCKET:
2359 		switch (name) {
2360 		case SO_REUSEADDR:
2361 			if (!checkonly) {
2362 				udp->udp_reuseaddr = onoff;
2363 				PASS_OPT_TO_IP(connp);
2364 			}
2365 			break;
2366 		case SO_DEBUG:
2367 			if (!checkonly)
2368 				udp->udp_debug = onoff;
2369 			break;
2370 		/*
2371 		 * The following three items are available here,
2372 		 * but are only meaningful to IP.
2373 		 */
2374 		case SO_DONTROUTE:
2375 			if (!checkonly) {
2376 				udp->udp_dontroute = onoff;
2377 				PASS_OPT_TO_IP(connp);
2378 			}
2379 			break;
2380 		case SO_USELOOPBACK:
2381 			if (!checkonly) {
2382 				udp->udp_useloopback = onoff;
2383 				PASS_OPT_TO_IP(connp);
2384 			}
2385 			break;
2386 		case SO_BROADCAST:
2387 			if (!checkonly) {
2388 				udp->udp_broadcast = onoff;
2389 				PASS_OPT_TO_IP(connp);
2390 			}
2391 			break;
2392 
2393 		case SO_SNDBUF:
2394 			if (*i1 > us->us_max_buf) {
2395 				*outlenp = 0;
2396 				return (ENOBUFS);
2397 			}
2398 			if (!checkonly) {
2399 				udp->udp_xmit_hiwat = *i1;
2400 				connp->conn_wq->q_hiwat = *i1;
2401 			}
2402 			break;
2403 		case SO_RCVBUF:
2404 			if (*i1 > us->us_max_buf) {
2405 				*outlenp = 0;
2406 				return (ENOBUFS);
2407 			}
2408 			if (!checkonly) {
2409 				int size;
2410 
2411 				udp->udp_rcv_disply_hiwat = *i1;
2412 				size = udp_set_rcv_hiwat(udp, *i1);
2413 				rw_exit(&udp->udp_rwlock);
2414 				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
2415 				    size);
2416 				rw_enter(&udp->udp_rwlock, RW_WRITER);
2417 			}
2418 			break;
2419 		case SO_DGRAM_ERRIND:
2420 			if (!checkonly)
2421 				udp->udp_dgram_errind = onoff;
2422 			break;
2423 		case SO_RECVUCRED:
2424 			if (!checkonly)
2425 				udp->udp_recvucred = onoff;
2426 			break;
2427 		case SO_ALLZONES:
2428 			/*
2429 			 * "soft" error (negative)
2430 			 * option not handled at this level
2431 			 * Do not modify *outlenp.
2432 			 */
2433 			return (-EINVAL);
2434 		case SO_TIMESTAMP:
2435 			if (!checkonly)
2436 				udp->udp_timestamp = onoff;
2437 			break;
2438 		case SO_ANON_MLP:
2439 			if (!checkonly) {
2440 				connp->conn_anon_mlp = onoff;
2441 				PASS_OPT_TO_IP(connp);
2442 			}
2443 			break;
2444 		case SO_MAC_EXEMPT:
2445 			if (secpolicy_net_mac_aware(cr) != 0 ||
2446 			    udp->udp_state != TS_UNBND)
2447 				return (EACCES);
2448 			if (!checkonly) {
2449 				connp->conn_mac_exempt = onoff;
2450 				PASS_OPT_TO_IP(connp);
2451 			}
2452 			break;
2453 		case SCM_UCRED: {
2454 			struct ucred_s *ucr;
2455 			cred_t *cr, *newcr;
2456 			ts_label_t *tsl;
2457 
2458 			/*
2459 			 * Only sockets that have proper privileges and are
2460 			 * bound to MLPs will have any other value here, so
2461 			 * this implicitly tests for privilege to set label.
2462 			 */
2463 			if (connp->conn_mlp_type == mlptSingle)
2464 				break;
2465 			ucr = (struct ucred_s *)invalp;
2466 			if (inlen != ucredsize ||
2467 			    ucr->uc_labeloff < sizeof (*ucr) ||
2468 			    ucr->uc_labeloff + sizeof (bslabel_t) > inlen)
2469 				return (EINVAL);
2470 			if (!checkonly) {
2471 				mblk_t *mb;
2472 
2473 				if (attrs == NULL ||
2474 				    (mb = attrs->udpattr_mb) == NULL)
2475 					return (EINVAL);
2476 				if ((cr = DB_CRED(mb)) == NULL)
2477 					cr = udp->udp_connp->conn_cred;
2478 				ASSERT(cr != NULL);
2479 				if ((tsl = crgetlabel(cr)) == NULL)
2480 					return (EINVAL);
2481 				newcr = copycred_from_bslabel(cr, UCLABEL(ucr),
2482 				    tsl->tsl_doi, KM_NOSLEEP);
2483 				if (newcr == NULL)
2484 					return (ENOSR);
2485 				mblk_setcred(mb, newcr);
2486 				attrs->udpattr_credset = B_TRUE;
2487 				crfree(newcr);
2488 			}
2489 			break;
2490 		}
2491 		case SO_EXCLBIND:
2492 			if (!checkonly)
2493 				udp->udp_exclbind = onoff;
2494 			break;
2495 		default:
2496 			*outlenp = 0;
2497 			return (EINVAL);
2498 		}
2499 		break;
2500 	case IPPROTO_IP:
2501 		if (udp->udp_family != AF_INET) {
2502 			*outlenp = 0;
2503 			return (ENOPROTOOPT);
2504 		}
2505 		switch (name) {
2506 		case IP_OPTIONS:
2507 		case T_IP_OPTIONS:
2508 			/* Save options for use by IP. */
2509 			newlen = inlen + udp->udp_label_len;
2510 			if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
2511 				*outlenp = 0;
2512 				return (EINVAL);
2513 			}
2514 			if (checkonly)
2515 				break;
2516 
2517 			/*
2518 			 * Update the stored options taking into account
2519 			 * any CIPSO option which we should not overwrite.
2520 			 */
2521 			if (!tsol_option_set(&udp->udp_ip_snd_options,
2522 			    &udp->udp_ip_snd_options_len,
2523 			    udp->udp_label_len, invalp, inlen)) {
2524 				*outlenp = 0;
2525 				return (ENOMEM);
2526 			}
2527 
2528 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2529 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
2530 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
2531 			rw_exit(&udp->udp_rwlock);
2532 			(void) proto_set_tx_wroff(connp->conn_rq, connp,
2533 			    sth_wroff);
2534 			rw_enter(&udp->udp_rwlock, RW_WRITER);
2535 			break;
2536 
2537 		case IP_TTL:
2538 			if (!checkonly) {
2539 				udp->udp_ttl = (uchar_t)*i1;
2540 			}
2541 			break;
2542 		case IP_TOS:
2543 		case T_IP_TOS:
2544 			if (!checkonly) {
2545 				udp->udp_type_of_service = (uchar_t)*i1;
2546 			}
2547 			break;
2548 		case IP_MULTICAST_IF: {
2549 			/*
2550 			 * TODO should check OPTMGMT reply and undo this if
2551 			 * there is an error.
2552 			 */
2553 			struct in_addr *inap = (struct in_addr *)invalp;
2554 			if (!checkonly) {
2555 				udp->udp_multicast_if_addr =
2556 				    inap->s_addr;
2557 				PASS_OPT_TO_IP(connp);
2558 			}
2559 			break;
2560 		}
2561 		case IP_MULTICAST_TTL:
2562 			if (!checkonly)
2563 				udp->udp_multicast_ttl = *invalp;
2564 			break;
2565 		case IP_MULTICAST_LOOP:
2566 			if (!checkonly) {
2567 				connp->conn_multicast_loop = *invalp;
2568 				PASS_OPT_TO_IP(connp);
2569 			}
2570 			break;
2571 		case IP_RECVOPTS:
2572 			if (!checkonly)
2573 				udp->udp_recvopts = onoff;
2574 			break;
2575 		case IP_RECVDSTADDR:
2576 			if (!checkonly)
2577 				udp->udp_recvdstaddr = onoff;
2578 			break;
2579 		case IP_RECVIF:
2580 			if (!checkonly) {
2581 				udp->udp_recvif = onoff;
2582 				PASS_OPT_TO_IP(connp);
2583 			}
2584 			break;
2585 		case IP_RECVSLLA:
2586 			if (!checkonly) {
2587 				udp->udp_recvslla = onoff;
2588 				PASS_OPT_TO_IP(connp);
2589 			}
2590 			break;
2591 		case IP_RECVTTL:
2592 			if (!checkonly)
2593 				udp->udp_recvttl = onoff;
2594 			break;
2595 		case IP_PKTINFO: {
2596 			/*
2597 			 * This also handles IP_RECVPKTINFO.
2598 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
2599 			 * Differentiation is based on the size of the
2600 			 * argument passed in.
2601 			 */
2602 			struct in_pktinfo *pktinfop;
2603 			ip4_pkt_t *attr_pktinfop;
2604 
2605 			if (checkonly)
2606 				break;
2607 
2608 			if (inlen == sizeof (int)) {
2609 				/*
2610 				 * This is IP_RECVPKTINFO option.
2611 				 * Keep a local copy of whether this option is
2612 				 * set or not and pass it down to IP for
2613 				 * processing.
2614 				 */
2615 
2616 				udp->udp_ip_recvpktinfo = onoff;
2617 				return (-EINVAL);
2618 			}
2619 
2620 			if (attrs == NULL ||
2621 			    (attr_pktinfop = attrs->udpattr_ipp4) == NULL) {
2622 				/*
2623 				 * sticky option or no buffer to return
2624 				 * the results.
2625 				 */
2626 				return (EINVAL);
2627 			}
2628 
2629 			if (inlen != sizeof (struct in_pktinfo))
2630 				return (EINVAL);
2631 
2632 			pktinfop = (struct in_pktinfo *)invalp;
2633 
2634 			/*
2635 			 * At least one of the values should be specified
2636 			 */
2637 			if (pktinfop->ipi_ifindex == 0 &&
2638 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2639 				return (EINVAL);
2640 			}
2641 
2642 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2643 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2644 
2645 			break;
2646 		}
2647 		case IP_ADD_MEMBERSHIP:
2648 		case IP_DROP_MEMBERSHIP:
2649 		case IP_BLOCK_SOURCE:
2650 		case IP_UNBLOCK_SOURCE:
2651 		case IP_ADD_SOURCE_MEMBERSHIP:
2652 		case IP_DROP_SOURCE_MEMBERSHIP:
2653 		case MCAST_JOIN_GROUP:
2654 		case MCAST_LEAVE_GROUP:
2655 		case MCAST_BLOCK_SOURCE:
2656 		case MCAST_UNBLOCK_SOURCE:
2657 		case MCAST_JOIN_SOURCE_GROUP:
2658 		case MCAST_LEAVE_SOURCE_GROUP:
2659 		case IP_SEC_OPT:
2660 		case IP_NEXTHOP:
2661 		case IP_DHCPINIT_IF:
2662 			/*
2663 			 * "soft" error (negative)
2664 			 * option not handled at this level
2665 			 * Do not modify *outlenp.
2666 			 */
2667 			return (-EINVAL);
2668 		case IP_BOUND_IF:
2669 			if (!checkonly) {
2670 				udp->udp_bound_if = *i1;
2671 				PASS_OPT_TO_IP(connp);
2672 			}
2673 			break;
2674 		case IP_UNSPEC_SRC:
2675 			if (!checkonly) {
2676 				udp->udp_unspec_source = onoff;
2677 				PASS_OPT_TO_IP(connp);
2678 			}
2679 			break;
2680 		case IP_BROADCAST_TTL:
2681 			if (!checkonly)
2682 				connp->conn_broadcast_ttl = *invalp;
2683 			break;
2684 		default:
2685 			*outlenp = 0;
2686 			return (EINVAL);
2687 		}
2688 		break;
2689 	case IPPROTO_IPV6: {
2690 		ip6_pkt_t		*ipp;
2691 		boolean_t		sticky;
2692 
2693 		if (udp->udp_family != AF_INET6) {
2694 			*outlenp = 0;
2695 			return (ENOPROTOOPT);
2696 		}
2697 		/*
2698 		 * Deal with both sticky options and ancillary data
2699 		 */
2700 		sticky = B_FALSE;
2701 		if (attrs == NULL || (ipp = attrs->udpattr_ipp6) ==
2702 		    NULL) {
2703 			/* sticky options, or none */
2704 			ipp = &udp->udp_sticky_ipp;
2705 			sticky = B_TRUE;
2706 		}
2707 
2708 		switch (name) {
2709 		case IPV6_MULTICAST_IF:
2710 			if (!checkonly) {
2711 				udp->udp_multicast_if_index = *i1;
2712 				PASS_OPT_TO_IP(connp);
2713 			}
2714 			break;
2715 		case IPV6_UNICAST_HOPS:
2716 			/* -1 means use default */
2717 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2718 				*outlenp = 0;
2719 				return (EINVAL);
2720 			}
2721 			if (!checkonly) {
2722 				if (*i1 == -1) {
2723 					udp->udp_ttl = ipp->ipp_unicast_hops =
2724 					    us->us_ipv6_hoplimit;
2725 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2726 					/* Pass modified value to IP. */
2727 					*i1 = udp->udp_ttl;
2728 				} else {
2729 					udp->udp_ttl = ipp->ipp_unicast_hops =
2730 					    (uint8_t)*i1;
2731 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2732 				}
2733 				/* Rebuild the header template */
2734 				error = udp_build_hdrs(udp);
2735 				if (error != 0) {
2736 					*outlenp = 0;
2737 					return (error);
2738 				}
2739 			}
2740 			break;
2741 		case IPV6_MULTICAST_HOPS:
2742 			/* -1 means use default */
2743 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2744 				*outlenp = 0;
2745 				return (EINVAL);
2746 			}
2747 			if (!checkonly) {
2748 				if (*i1 == -1) {
2749 					udp->udp_multicast_ttl =
2750 					    ipp->ipp_multicast_hops =
2751 					    IP_DEFAULT_MULTICAST_TTL;
2752 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2753 					/* Pass modified value to IP. */
2754 					*i1 = udp->udp_multicast_ttl;
2755 				} else {
2756 					udp->udp_multicast_ttl =
2757 					    ipp->ipp_multicast_hops =
2758 					    (uint8_t)*i1;
2759 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2760 				}
2761 			}
2762 			break;
2763 		case IPV6_MULTICAST_LOOP:
2764 			if (*i1 != 0 && *i1 != 1) {
2765 				*outlenp = 0;
2766 				return (EINVAL);
2767 			}
2768 			if (!checkonly) {
2769 				connp->conn_multicast_loop = *i1;
2770 				PASS_OPT_TO_IP(connp);
2771 			}
2772 			break;
2773 		case IPV6_JOIN_GROUP:
2774 		case IPV6_LEAVE_GROUP:
2775 		case MCAST_JOIN_GROUP:
2776 		case MCAST_LEAVE_GROUP:
2777 		case MCAST_BLOCK_SOURCE:
2778 		case MCAST_UNBLOCK_SOURCE:
2779 		case MCAST_JOIN_SOURCE_GROUP:
2780 		case MCAST_LEAVE_SOURCE_GROUP:
2781 			/*
2782 			 * "soft" error (negative)
2783 			 * option not handled at this level
2784 			 * Note: Do not modify *outlenp
2785 			 */
2786 			return (-EINVAL);
2787 		case IPV6_BOUND_IF:
2788 			if (!checkonly) {
2789 				udp->udp_bound_if = *i1;
2790 				PASS_OPT_TO_IP(connp);
2791 			}
2792 			break;
2793 		case IPV6_UNSPEC_SRC:
2794 			if (!checkonly) {
2795 				udp->udp_unspec_source = onoff;
2796 				PASS_OPT_TO_IP(connp);
2797 			}
2798 			break;
2799 		/*
2800 		 * Set boolean switches for ancillary data delivery
2801 		 */
2802 		case IPV6_RECVPKTINFO:
2803 			if (!checkonly) {
2804 				udp->udp_ip_recvpktinfo = onoff;
2805 				PASS_OPT_TO_IP(connp);
2806 			}
2807 			break;
2808 		case IPV6_RECVTCLASS:
2809 			if (!checkonly) {
2810 				udp->udp_ipv6_recvtclass = onoff;
2811 				PASS_OPT_TO_IP(connp);
2812 			}
2813 			break;
2814 		case IPV6_RECVPATHMTU:
2815 			if (!checkonly) {
2816 				udp->udp_ipv6_recvpathmtu = onoff;
2817 				PASS_OPT_TO_IP(connp);
2818 			}
2819 			break;
2820 		case IPV6_RECVHOPLIMIT:
2821 			if (!checkonly) {
2822 				udp->udp_ipv6_recvhoplimit = onoff;
2823 				PASS_OPT_TO_IP(connp);
2824 			}
2825 			break;
2826 		case IPV6_RECVHOPOPTS:
2827 			if (!checkonly) {
2828 				udp->udp_ipv6_recvhopopts = onoff;
2829 				PASS_OPT_TO_IP(connp);
2830 			}
2831 			break;
2832 		case IPV6_RECVDSTOPTS:
2833 			if (!checkonly) {
2834 				udp->udp_ipv6_recvdstopts = onoff;
2835 				PASS_OPT_TO_IP(connp);
2836 			}
2837 			break;
2838 		case _OLD_IPV6_RECVDSTOPTS:
2839 			if (!checkonly)
2840 				udp->udp_old_ipv6_recvdstopts = onoff;
2841 			break;
2842 		case IPV6_RECVRTHDRDSTOPTS:
2843 			if (!checkonly) {
2844 				udp->udp_ipv6_recvrthdrdstopts = onoff;
2845 				PASS_OPT_TO_IP(connp);
2846 			}
2847 			break;
2848 		case IPV6_RECVRTHDR:
2849 			if (!checkonly) {
2850 				udp->udp_ipv6_recvrthdr = onoff;
2851 				PASS_OPT_TO_IP(connp);
2852 			}
2853 			break;
2854 		/*
2855 		 * Set sticky options or ancillary data.
2856 		 * If sticky options, (re)build any extension headers
2857 		 * that might be needed as a result.
2858 		 */
2859 		case IPV6_PKTINFO:
2860 			/*
2861 			 * The source address and ifindex are verified
2862 			 * in ip_opt_set(). For ancillary data the
2863 			 * source address is checked in ip_wput_v6.
2864 			 */
2865 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2866 				return (EINVAL);
2867 			if (checkonly)
2868 				break;
2869 
2870 			if (inlen == 0) {
2871 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2872 				ipp->ipp_sticky_ignored |=
2873 				    (IPPF_IFINDEX|IPPF_ADDR);
2874 			} else {
2875 				struct in6_pktinfo *pkti;
2876 
2877 				pkti = (struct in6_pktinfo *)invalp;
2878 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2879 				ipp->ipp_addr = pkti->ipi6_addr;
2880 				if (ipp->ipp_ifindex != 0)
2881 					ipp->ipp_fields |= IPPF_IFINDEX;
2882 				else
2883 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2884 				if (!IN6_IS_ADDR_UNSPECIFIED(
2885 				    &ipp->ipp_addr))
2886 					ipp->ipp_fields |= IPPF_ADDR;
2887 				else
2888 					ipp->ipp_fields &= ~IPPF_ADDR;
2889 			}
2890 			if (sticky) {
2891 				error = udp_build_hdrs(udp);
2892 				if (error != 0)
2893 					return (error);
2894 				PASS_OPT_TO_IP(connp);
2895 			}
2896 			break;
2897 		case IPV6_HOPLIMIT:
2898 			if (sticky)
2899 				return (EINVAL);
2900 			if (inlen != 0 && inlen != sizeof (int))
2901 				return (EINVAL);
2902 			if (checkonly)
2903 				break;
2904 
2905 			if (inlen == 0) {
2906 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2907 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2908 			} else {
2909 				if (*i1 > 255 || *i1 < -1)
2910 					return (EINVAL);
2911 				if (*i1 == -1)
2912 					ipp->ipp_hoplimit =
2913 					    us->us_ipv6_hoplimit;
2914 				else
2915 					ipp->ipp_hoplimit = *i1;
2916 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2917 			}
2918 			break;
2919 		case IPV6_TCLASS:
2920 			if (inlen != 0 && inlen != sizeof (int))
2921 				return (EINVAL);
2922 			if (checkonly)
2923 				break;
2924 
2925 			if (inlen == 0) {
2926 				ipp->ipp_fields &= ~IPPF_TCLASS;
2927 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2928 			} else {
2929 				if (*i1 > 255 || *i1 < -1)
2930 					return (EINVAL);
2931 				if (*i1 == -1)
2932 					ipp->ipp_tclass = 0;
2933 				else
2934 					ipp->ipp_tclass = *i1;
2935 				ipp->ipp_fields |= IPPF_TCLASS;
2936 			}
2937 			if (sticky) {
2938 				error = udp_build_hdrs(udp);
2939 				if (error != 0)
2940 					return (error);
2941 			}
2942 			break;
2943 		case IPV6_NEXTHOP:
2944 			/*
2945 			 * IP will verify that the nexthop is reachable
2946 			 * and fail for sticky options.
2947 			 */
2948 			if (inlen != 0 && inlen != sizeof (sin6_t))
2949 				return (EINVAL);
2950 			if (checkonly)
2951 				break;
2952 
2953 			if (inlen == 0) {
2954 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2955 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2956 			} else {
2957 				sin6_t *sin6 = (sin6_t *)invalp;
2958 
2959 				if (sin6->sin6_family != AF_INET6) {
2960 					return (EAFNOSUPPORT);
2961 				}
2962 				if (IN6_IS_ADDR_V4MAPPED(
2963 				    &sin6->sin6_addr))
2964 					return (EADDRNOTAVAIL);
2965 				ipp->ipp_nexthop = sin6->sin6_addr;
2966 				if (!IN6_IS_ADDR_UNSPECIFIED(
2967 				    &ipp->ipp_nexthop))
2968 					ipp->ipp_fields |= IPPF_NEXTHOP;
2969 				else
2970 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2971 			}
2972 			if (sticky) {
2973 				error = udp_build_hdrs(udp);
2974 				if (error != 0)
2975 					return (error);
2976 				PASS_OPT_TO_IP(connp);
2977 			}
2978 			break;
2979 		case IPV6_HOPOPTS: {
2980 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2981 			/*
2982 			 * Sanity checks - minimum size, size a multiple of
2983 			 * eight bytes, and matching size passed in.
2984 			 */
2985 			if (inlen != 0 &&
2986 			    inlen != (8 * (hopts->ip6h_len + 1)))
2987 				return (EINVAL);
2988 
2989 			if (checkonly)
2990 				break;
2991 
2992 			error = optcom_pkt_set(invalp, inlen, sticky,
2993 			    (uchar_t **)&ipp->ipp_hopopts,
2994 			    &ipp->ipp_hopoptslen,
2995 			    sticky ? udp->udp_label_len_v6 : 0);
2996 			if (error != 0)
2997 				return (error);
2998 			if (ipp->ipp_hopoptslen == 0) {
2999 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
3000 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
3001 			} else {
3002 				ipp->ipp_fields |= IPPF_HOPOPTS;
3003 			}
3004 			if (sticky) {
3005 				error = udp_build_hdrs(udp);
3006 				if (error != 0)
3007 					return (error);
3008 			}
3009 			break;
3010 		}
3011 		case IPV6_RTHDRDSTOPTS: {
3012 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
3013 
3014 			/*
3015 			 * Sanity checks - minimum size, size a multiple of
3016 			 * eight bytes, and matching size passed in.
3017 			 */
3018 			if (inlen != 0 &&
3019 			    inlen != (8 * (dopts->ip6d_len + 1)))
3020 				return (EINVAL);
3021 
3022 			if (checkonly)
3023 				break;
3024 
3025 			if (inlen == 0) {
3026 				if (sticky &&
3027 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
3028 					kmem_free(ipp->ipp_rtdstopts,
3029 					    ipp->ipp_rtdstoptslen);
3030 					ipp->ipp_rtdstopts = NULL;
3031 					ipp->ipp_rtdstoptslen = 0;
3032 				}
3033 
3034 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
3035 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
3036 			} else {
3037 				error = optcom_pkt_set(invalp, inlen, sticky,
3038 				    (uchar_t **)&ipp->ipp_rtdstopts,
3039 				    &ipp->ipp_rtdstoptslen, 0);
3040 				if (error != 0)
3041 					return (error);
3042 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
3043 			}
3044 			if (sticky) {
3045 				error = udp_build_hdrs(udp);
3046 				if (error != 0)
3047 					return (error);
3048 			}
3049 			break;
3050 		}
3051 		case IPV6_DSTOPTS: {
3052 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
3053 
3054 			/*
3055 			 * Sanity checks - minimum size, size a multiple of
3056 			 * eight bytes, and matching size passed in.
3057 			 */
3058 			if (inlen != 0 &&
3059 			    inlen != (8 * (dopts->ip6d_len + 1)))
3060 				return (EINVAL);
3061 
3062 			if (checkonly)
3063 				break;
3064 
3065 			if (inlen == 0) {
3066 				if (sticky &&
3067 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
3068 					kmem_free(ipp->ipp_dstopts,
3069 					    ipp->ipp_dstoptslen);
3070 					ipp->ipp_dstopts = NULL;
3071 					ipp->ipp_dstoptslen = 0;
3072 				}
3073 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
3074 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
3075 			} else {
3076 				error = optcom_pkt_set(invalp, inlen, sticky,
3077 				    (uchar_t **)&ipp->ipp_dstopts,
3078 				    &ipp->ipp_dstoptslen, 0);
3079 				if (error != 0)
3080 					return (error);
3081 				ipp->ipp_fields |= IPPF_DSTOPTS;
3082 			}
3083 			if (sticky) {
3084 				error = udp_build_hdrs(udp);
3085 				if (error != 0)
3086 					return (error);
3087 			}
3088 			break;
3089 		}
3090 		case IPV6_RTHDR: {
3091 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
3092 
3093 			/*
3094 			 * Sanity checks - minimum size, size a multiple of
3095 			 * eight bytes, and matching size passed in.
3096 			 */
3097 			if (inlen != 0 &&
3098 			    inlen != (8 * (rt->ip6r_len + 1)))
3099 				return (EINVAL);
3100 
3101 			if (checkonly)
3102 				break;
3103 
3104 			if (inlen == 0) {
3105 				if (sticky &&
3106 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
3107 					kmem_free(ipp->ipp_rthdr,
3108 					    ipp->ipp_rthdrlen);
3109 					ipp->ipp_rthdr = NULL;
3110 					ipp->ipp_rthdrlen = 0;
3111 				}
3112 				ipp->ipp_fields &= ~IPPF_RTHDR;
3113 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
3114 			} else {
3115 				error = optcom_pkt_set(invalp, inlen, sticky,
3116 				    (uchar_t **)&ipp->ipp_rthdr,
3117 				    &ipp->ipp_rthdrlen, 0);
3118 				if (error != 0)
3119 					return (error);
3120 				ipp->ipp_fields |= IPPF_RTHDR;
3121 			}
3122 			if (sticky) {
3123 				error = udp_build_hdrs(udp);
3124 				if (error != 0)
3125 					return (error);
3126 			}
3127 			break;
3128 		}
3129 
3130 		case IPV6_DONTFRAG:
3131 			if (checkonly)
3132 				break;
3133 
3134 			if (onoff) {
3135 				ipp->ipp_fields |= IPPF_DONTFRAG;
3136 			} else {
3137 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
3138 			}
3139 			break;
3140 
3141 		case IPV6_USE_MIN_MTU:
3142 			if (inlen != sizeof (int))
3143 				return (EINVAL);
3144 
3145 			if (*i1 < -1 || *i1 > 1)
3146 				return (EINVAL);
3147 
3148 			if (checkonly)
3149 				break;
3150 
3151 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3152 			ipp->ipp_use_min_mtu = *i1;
3153 			break;
3154 
3155 		case IPV6_SEC_OPT:
3156 		case IPV6_SRC_PREFERENCES:
3157 		case IPV6_V6ONLY:
3158 			/* Handled at the IP level */
3159 			return (-EINVAL);
3160 		default:
3161 			*outlenp = 0;
3162 			return (EINVAL);
3163 		}
3164 		break;
3165 		}		/* end IPPROTO_IPV6 */
3166 	case IPPROTO_UDP:
3167 		switch (name) {
3168 		case UDP_ANONPRIVBIND:
3169 			if ((error = secpolicy_net_privaddr(cr, 0,
3170 			    IPPROTO_UDP)) != 0) {
3171 				*outlenp = 0;
3172 				return (error);
3173 			}
3174 			if (!checkonly) {
3175 				udp->udp_anon_priv_bind = onoff;
3176 			}
3177 			break;
3178 		case UDP_EXCLBIND:
3179 			if (!checkonly)
3180 				udp->udp_exclbind = onoff;
3181 			break;
3182 		case UDP_RCVHDR:
3183 			if (!checkonly)
3184 				udp->udp_rcvhdr = onoff;
3185 			break;
3186 		case UDP_NAT_T_ENDPOINT:
3187 			if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
3188 				*outlenp = 0;
3189 				return (error);
3190 			}
3191 
3192 			/*
3193 			 * Use udp_family instead so we can avoid ambiguitites
3194 			 * with AF_INET6 sockets that may switch from IPv4
3195 			 * to IPv6.
3196 			 */
3197 			if (udp->udp_family != AF_INET) {
3198 				*outlenp = 0;
3199 				return (EAFNOSUPPORT);
3200 			}
3201 
3202 			if (!checkonly) {
3203 				int size;
3204 
3205 				udp->udp_nat_t_endpoint = onoff;
3206 
3207 				udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3208 				    UDPH_SIZE + udp->udp_ip_snd_options_len;
3209 
3210 				/* Also, adjust wroff */
3211 				if (onoff) {
3212 					udp->udp_max_hdr_len +=
3213 					    sizeof (uint32_t);
3214 				}
3215 				size = udp->udp_max_hdr_len +
3216 				    us->us_wroff_extra;
3217 				(void) proto_set_tx_wroff(connp->conn_rq, connp,
3218 				    size);
3219 			}
3220 			break;
3221 		default:
3222 			*outlenp = 0;
3223 			return (EINVAL);
3224 		}
3225 		break;
3226 	default:
3227 		*outlenp = 0;
3228 		return (EINVAL);
3229 	}
3230 	/*
3231 	 * Common case of OK return with outval same as inval.
3232 	 */
3233 	if (invalp != outvalp) {
3234 		/* don't trust bcopy for identical src/dst */
3235 		(void) bcopy(invalp, outvalp, inlen);
3236 	}
3237 	*outlenp = inlen;
3238 	return (0);
3239 }
3240 
3241 int
3242 udp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3243     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3244     void *thisdg_attrs, cred_t *cr)
3245 {
3246 	int		error;
3247 	boolean_t	checkonly;
3248 
3249 	error = 0;
3250 	switch (optset_context) {
3251 	case SETFN_OPTCOM_CHECKONLY:
3252 		checkonly = B_TRUE;
3253 		/*
3254 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3255 		 * inlen != 0 implies value supplied and
3256 		 * 	we have to "pretend" to set it.
3257 		 * inlen == 0 implies that there is no
3258 		 * 	value part in T_CHECK request and just validation
3259 		 * done elsewhere should be enough, we just return here.
3260 		 */
3261 		if (inlen == 0) {
3262 			*outlenp = 0;
3263 			goto done;
3264 		}
3265 		break;
3266 	case SETFN_OPTCOM_NEGOTIATE:
3267 		checkonly = B_FALSE;
3268 		break;
3269 	case SETFN_UD_NEGOTIATE:
3270 	case SETFN_CONN_NEGOTIATE:
3271 		checkonly = B_FALSE;
3272 		/*
3273 		 * Negotiating local and "association-related" options
3274 		 * through T_UNITDATA_REQ.
3275 		 *
3276 		 * Following routine can filter out ones we do not
3277 		 * want to be "set" this way.
3278 		 */
3279 		if (!udp_opt_allow_udr_set(level, name)) {
3280 			*outlenp = 0;
3281 			error = EINVAL;
3282 			goto done;
3283 		}
3284 		break;
3285 	default:
3286 		/*
3287 		 * We should never get here
3288 		 */
3289 		*outlenp = 0;
3290 		error = EINVAL;
3291 		goto done;
3292 	}
3293 
3294 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3295 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3296 
3297 	error = udp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3298 	    outvalp, cr, thisdg_attrs, checkonly);
3299 done:
3300 	return (error);
3301 }
3302 
3303 /* ARGSUSED */
3304 int
3305 udp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3306     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3307     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3308 {
3309 	conn_t  *connp =  Q_TO_CONN(q);
3310 	int error;
3311 	udp_t	*udp = connp->conn_udp;
3312 
3313 	rw_enter(&udp->udp_rwlock, RW_WRITER);
3314 	error = udp_opt_set(connp, optset_context, level, name, inlen, invalp,
3315 	    outlenp, outvalp, thisdg_attrs, cr);
3316 	rw_exit(&udp->udp_rwlock);
3317 	return (error);
3318 }
3319 
3320 /*
3321  * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl.
3322  * The headers include ip6i_t (if needed), ip6_t, any sticky extension
3323  * headers, and the udp header.
3324  * Returns failure if can't allocate memory.
3325  */
3326 static int
3327 udp_build_hdrs(udp_t *udp)
3328 {
3329 	udp_stack_t *us = udp->udp_us;
3330 	uchar_t	*hdrs;
3331 	uint_t	hdrs_len;
3332 	ip6_t	*ip6h;
3333 	ip6i_t	*ip6i;
3334 	udpha_t	*udpha;
3335 	ip6_pkt_t *ipp = &udp->udp_sticky_ipp;
3336 	size_t	sth_wroff;
3337 	conn_t	*connp = udp->udp_connp;
3338 
3339 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
3340 	ASSERT(connp != NULL);
3341 
3342 	hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE;
3343 	ASSERT(hdrs_len != 0);
3344 	if (hdrs_len != udp->udp_sticky_hdrs_len) {
3345 		/* Need to reallocate */
3346 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3347 		if (hdrs == NULL)
3348 			return (ENOMEM);
3349 
3350 		if (udp->udp_sticky_hdrs_len != 0) {
3351 			kmem_free(udp->udp_sticky_hdrs,
3352 			    udp->udp_sticky_hdrs_len);
3353 		}
3354 		udp->udp_sticky_hdrs = hdrs;
3355 		udp->udp_sticky_hdrs_len = hdrs_len;
3356 	}
3357 	ip_build_hdrs_v6(udp->udp_sticky_hdrs,
3358 	    udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP);
3359 
3360 	/* Set header fields not in ipp */
3361 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3362 		ip6i = (ip6i_t *)udp->udp_sticky_hdrs;
3363 		ip6h = (ip6_t *)&ip6i[1];
3364 	} else {
3365 		ip6h = (ip6_t *)udp->udp_sticky_hdrs;
3366 	}
3367 
3368 	if (!(ipp->ipp_fields & IPPF_ADDR))
3369 		ip6h->ip6_src = udp->udp_v6src;
3370 
3371 	udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE);
3372 	udpha->uha_src_port = udp->udp_port;
3373 
3374 	/* Try to get everything in a single mblk */
3375 	if (hdrs_len > udp->udp_max_hdr_len) {
3376 		udp->udp_max_hdr_len = hdrs_len;
3377 		sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
3378 		rw_exit(&udp->udp_rwlock);
3379 		(void) proto_set_tx_wroff(udp->udp_connp->conn_rq,
3380 		    udp->udp_connp, sth_wroff);
3381 		rw_enter(&udp->udp_rwlock, RW_WRITER);
3382 	}
3383 	return (0);
3384 }
3385 
3386 /*
3387  * This routine retrieves the value of an ND variable in a udpparam_t
3388  * structure.  It is called through nd_getset when a user reads the
3389  * variable.
3390  */
3391 /* ARGSUSED */
3392 static int
3393 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3394 {
3395 	udpparam_t *udppa = (udpparam_t *)cp;
3396 
3397 	(void) mi_mpprintf(mp, "%d", udppa->udp_param_value);
3398 	return (0);
3399 }
3400 
3401 /*
3402  * Walk through the param array specified registering each element with the
3403  * named dispatch (ND) handler.
3404  */
3405 static boolean_t
3406 udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt)
3407 {
3408 	for (; cnt-- > 0; udppa++) {
3409 		if (udppa->udp_param_name && udppa->udp_param_name[0]) {
3410 			if (!nd_load(ndp, udppa->udp_param_name,
3411 			    udp_param_get, udp_param_set,
3412 			    (caddr_t)udppa)) {
3413 				nd_free(ndp);
3414 				return (B_FALSE);
3415 			}
3416 		}
3417 	}
3418 	if (!nd_load(ndp, "udp_extra_priv_ports",
3419 	    udp_extra_priv_ports_get, NULL, NULL)) {
3420 		nd_free(ndp);
3421 		return (B_FALSE);
3422 	}
3423 	if (!nd_load(ndp, "udp_extra_priv_ports_add",
3424 	    NULL, udp_extra_priv_ports_add, NULL)) {
3425 		nd_free(ndp);
3426 		return (B_FALSE);
3427 	}
3428 	if (!nd_load(ndp, "udp_extra_priv_ports_del",
3429 	    NULL, udp_extra_priv_ports_del, NULL)) {
3430 		nd_free(ndp);
3431 		return (B_FALSE);
3432 	}
3433 	if (!nd_load(ndp, "udp_status", udp_status_report, NULL,
3434 	    NULL)) {
3435 		nd_free(ndp);
3436 		return (B_FALSE);
3437 	}
3438 	if (!nd_load(ndp, "udp_bind_hash", udp_bind_hash_report, NULL,
3439 	    NULL)) {
3440 		nd_free(ndp);
3441 		return (B_FALSE);
3442 	}
3443 	return (B_TRUE);
3444 }
3445 
3446 /* This routine sets an ND variable in a udpparam_t structure. */
3447 /* ARGSUSED */
3448 static int
3449 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3450 {
3451 	long		new_value;
3452 	udpparam_t	*udppa = (udpparam_t *)cp;
3453 
3454 	/*
3455 	 * Fail the request if the new value does not lie within the
3456 	 * required bounds.
3457 	 */
3458 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3459 	    new_value < udppa->udp_param_min ||
3460 	    new_value > udppa->udp_param_max) {
3461 		return (EINVAL);
3462 	}
3463 
3464 	/* Set the new value */
3465 	udppa->udp_param_value = new_value;
3466 	return (0);
3467 }
3468 
3469 /*
3470  * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with
3471  * T_opthdr) and return the number of bytes copied.  'dbuf' may be NULL to
3472  * just count the length needed for allocation.  If 'dbuf' is non-NULL,
3473  * then it's assumed to be allocated to be large enough.
3474  *
3475  * Returns zero if trimming of the security option causes all options to go
3476  * away.
3477  */
3478 static size_t
3479 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf)
3480 {
3481 	struct T_opthdr *toh;
3482 	size_t hol = ipp->ipp_hopoptslen;
3483 	ip6_hbh_t *dstopt = NULL;
3484 	const ip6_hbh_t *srcopt = ipp->ipp_hopopts;
3485 	size_t tlen, olen, plen;
3486 	boolean_t deleting;
3487 	const struct ip6_opt *sopt, *lastpad;
3488 	struct ip6_opt *dopt;
3489 
3490 	if ((toh = (struct T_opthdr *)dbuf) != NULL) {
3491 		toh->level = IPPROTO_IPV6;
3492 		toh->name = IPV6_HOPOPTS;
3493 		toh->status = 0;
3494 		dstopt = (ip6_hbh_t *)(toh + 1);
3495 	}
3496 
3497 	/*
3498 	 * If labeling is enabled, then skip the label option
3499 	 * but get other options if there are any.
3500 	 */
3501 	if (is_system_labeled()) {
3502 		dopt = NULL;
3503 		if (dstopt != NULL) {
3504 			/* will fill in ip6h_len later */
3505 			dstopt->ip6h_nxt = srcopt->ip6h_nxt;
3506 			dopt = (struct ip6_opt *)(dstopt + 1);
3507 		}
3508 		sopt = (const struct ip6_opt *)(srcopt + 1);
3509 		hol -= sizeof (*srcopt);
3510 		tlen = sizeof (*dstopt);
3511 		lastpad = NULL;
3512 		deleting = B_FALSE;
3513 		/*
3514 		 * This loop finds the first (lastpad pointer) of any number of
3515 		 * pads that preceeds the security option, then treats the
3516 		 * security option as though it were a pad, and then finds the
3517 		 * next non-pad option (or end of list).
3518 		 *
3519 		 * It then treats the entire block as one big pad.  To preserve
3520 		 * alignment of any options that follow, or just the end of the
3521 		 * list, it computes a minimal new padding size that keeps the
3522 		 * same alignment for the next option.
3523 		 *
3524 		 * If it encounters just a sequence of pads with no security
3525 		 * option, those are copied as-is rather than collapsed.
3526 		 *
3527 		 * Note that to handle the end of list case, the code makes one
3528 		 * loop with 'hol' set to zero.
3529 		 */
3530 		for (;;) {
3531 			if (hol > 0) {
3532 				if (sopt->ip6o_type == IP6OPT_PAD1) {
3533 					if (lastpad == NULL)
3534 						lastpad = sopt;
3535 					sopt = (const struct ip6_opt *)
3536 					    &sopt->ip6o_len;
3537 					hol--;
3538 					continue;
3539 				}
3540 				olen = sopt->ip6o_len + sizeof (*sopt);
3541 				if (olen > hol)
3542 					olen = hol;
3543 				if (sopt->ip6o_type == IP6OPT_PADN ||
3544 				    sopt->ip6o_type == ip6opt_ls) {
3545 					if (sopt->ip6o_type == ip6opt_ls)
3546 						deleting = B_TRUE;
3547 					if (lastpad == NULL)
3548 						lastpad = sopt;
3549 					sopt = (const struct ip6_opt *)
3550 					    ((const char *)sopt + olen);
3551 					hol -= olen;
3552 					continue;
3553 				}
3554 			} else {
3555 				/* if nothing was copied at all, then delete */
3556 				if (tlen == sizeof (*dstopt))
3557 					return (0);
3558 				/* last pass; pick up any trailing padding */
3559 				olen = 0;
3560 			}
3561 			if (deleting) {
3562 				/*
3563 				 * compute aligning effect of deleted material
3564 				 * to reproduce with pad.
3565 				 */
3566 				plen = ((const char *)sopt -
3567 				    (const char *)lastpad) & 7;
3568 				tlen += plen;
3569 				if (dopt != NULL) {
3570 					if (plen == 1) {
3571 						dopt->ip6o_type = IP6OPT_PAD1;
3572 					} else if (plen > 1) {
3573 						plen -= sizeof (*dopt);
3574 						dopt->ip6o_type = IP6OPT_PADN;
3575 						dopt->ip6o_len = plen;
3576 						if (plen > 0)
3577 							bzero(dopt + 1, plen);
3578 					}
3579 					dopt = (struct ip6_opt *)
3580 					    ((char *)dopt + plen);
3581 				}
3582 				deleting = B_FALSE;
3583 				lastpad = NULL;
3584 			}
3585 			/* if there's uncopied padding, then copy that now */
3586 			if (lastpad != NULL) {
3587 				olen += (const char *)sopt -
3588 				    (const char *)lastpad;
3589 				sopt = lastpad;
3590 				lastpad = NULL;
3591 			}
3592 			if (dopt != NULL && olen > 0) {
3593 				bcopy(sopt, dopt, olen);
3594 				dopt = (struct ip6_opt *)((char *)dopt + olen);
3595 			}
3596 			if (hol == 0)
3597 				break;
3598 			tlen += olen;
3599 			sopt = (const struct ip6_opt *)
3600 			    ((const char *)sopt + olen);
3601 			hol -= olen;
3602 		}
3603 		/* go back and patch up the length value, rounded upward */
3604 		if (dstopt != NULL)
3605 			dstopt->ip6h_len = (tlen - 1) >> 3;
3606 	} else {
3607 		tlen = hol;
3608 		if (dstopt != NULL)
3609 			bcopy(srcopt, dstopt, hol);
3610 	}
3611 
3612 	tlen += sizeof (*toh);
3613 	if (toh != NULL)
3614 		toh->len = tlen;
3615 
3616 	return (tlen);
3617 }
3618 
3619 /*
3620  * Update udp_rcv_opt_len from the packet.
3621  * Called when options received, and when no options received but
3622  * udp_ip_recv_opt_len has previously recorded options.
3623  */
3624 static void
3625 udp_save_ip_rcv_opt(udp_t *udp, void *opt, int opt_len)
3626 {
3627 	/* Save the options if any */
3628 	if (opt_len > 0) {
3629 		if (opt_len > udp->udp_ip_rcv_options_len) {
3630 			/* Need to allocate larger buffer */
3631 			if (udp->udp_ip_rcv_options_len != 0)
3632 				mi_free((char *)udp->udp_ip_rcv_options);
3633 			udp->udp_ip_rcv_options_len = 0;
3634 			udp->udp_ip_rcv_options =
3635 			    (uchar_t *)mi_alloc(opt_len, BPRI_HI);
3636 			if (udp->udp_ip_rcv_options != NULL)
3637 				udp->udp_ip_rcv_options_len = opt_len;
3638 		}
3639 		if (udp->udp_ip_rcv_options_len != 0) {
3640 			bcopy(opt, udp->udp_ip_rcv_options, opt_len);
3641 			/* Adjust length if we are resusing the space */
3642 			udp->udp_ip_rcv_options_len = opt_len;
3643 		}
3644 	} else if (udp->udp_ip_rcv_options_len != 0) {
3645 		/* Clear out previously recorded options */
3646 		mi_free((char *)udp->udp_ip_rcv_options);
3647 		udp->udp_ip_rcv_options = NULL;
3648 		udp->udp_ip_rcv_options_len = 0;
3649 	}
3650 }
3651 
3652 static void
3653 udp_queue_fallback(udp_t *udp, mblk_t *mp)
3654 {
3655 	ASSERT(MUTEX_HELD(&udp->udp_recv_lock));
3656 	if (IPCL_IS_NONSTR(udp->udp_connp)) {
3657 		/*
3658 		 * fallback has started but messages have not been moved yet
3659 		 */
3660 		if (udp->udp_fallback_queue_head == NULL) {
3661 			ASSERT(udp->udp_fallback_queue_tail == NULL);
3662 			udp->udp_fallback_queue_head = mp;
3663 			udp->udp_fallback_queue_tail = mp;
3664 		} else {
3665 			ASSERT(udp->udp_fallback_queue_tail != NULL);
3666 			udp->udp_fallback_queue_tail->b_next = mp;
3667 			udp->udp_fallback_queue_tail = mp;
3668 		}
3669 		mutex_exit(&udp->udp_recv_lock);
3670 	} else {
3671 		/*
3672 		 * no more fallbacks possible, ok to drop lock.
3673 		 */
3674 		mutex_exit(&udp->udp_recv_lock);
3675 		putnext(udp->udp_connp->conn_rq, mp);
3676 	}
3677 }
3678 
3679 /* ARGSUSED2 */
3680 static void
3681 udp_input(void *arg1, mblk_t *mp, void *arg2)
3682 {
3683 	conn_t *connp = (conn_t *)arg1;
3684 	struct T_unitdata_ind	*tudi;
3685 	uchar_t			*rptr;		/* Pointer to IP header */
3686 	int			hdr_length;	/* Length of IP+UDP headers */
3687 	int			opt_len;
3688 	int			udi_size;	/* Size of T_unitdata_ind */
3689 	int			mp_len;
3690 	udp_t			*udp;
3691 	udpha_t			*udpha;
3692 	int			ipversion;
3693 	ip6_pkt_t		ipp;
3694 	ip6_t			*ip6h;
3695 	ip6i_t			*ip6i;
3696 	mblk_t			*mp1;
3697 	mblk_t			*options_mp = NULL;
3698 	ip_pktinfo_t		*pinfo = NULL;
3699 	cred_t			*cr = NULL;
3700 	pid_t			cpid;
3701 	uint32_t		udp_ip_rcv_options_len;
3702 	udp_bits_t		udp_bits;
3703 	cred_t			*rcr = connp->conn_cred;
3704 	udp_stack_t *us;
3705 
3706 	ASSERT(connp->conn_flags & IPCL_UDPCONN);
3707 
3708 	udp = connp->conn_udp;
3709 	us = udp->udp_us;
3710 	rptr = mp->b_rptr;
3711 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3712 	ASSERT(OK_32PTR(rptr));
3713 
3714 	/*
3715 	 * IP should have prepended the options data in an M_CTL
3716 	 * Check M_CTL "type" to make sure are not here bcos of
3717 	 * a valid ICMP message
3718 	 */
3719 	if (DB_TYPE(mp) == M_CTL) {
3720 		if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3721 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3722 		    IN_PKTINFO) {
3723 			/*
3724 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3725 			 * has been prepended to the packet by IP. We need to
3726 			 * extract the mblk and adjust the rptr
3727 			 */
3728 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3729 			options_mp = mp;
3730 			mp = mp->b_cont;
3731 			rptr = mp->b_rptr;
3732 			UDP_STAT(us, udp_in_pktinfo);
3733 		} else {
3734 			/*
3735 			 * ICMP messages.
3736 			 */
3737 			udp_icmp_error(connp, mp);
3738 			return;
3739 		}
3740 	}
3741 
3742 	mp_len = msgdsize(mp);
3743 	/*
3744 	 * This is the inbound data path.
3745 	 * First, we check to make sure the IP version number is correct,
3746 	 * and then pull the IP and UDP headers into the first mblk.
3747 	 */
3748 
3749 	/* Initialize regardless if ipversion is IPv4 or IPv6 */
3750 	ipp.ipp_fields = 0;
3751 
3752 	ipversion = IPH_HDR_VERSION(rptr);
3753 
3754 	rw_enter(&udp->udp_rwlock, RW_READER);
3755 	udp_ip_rcv_options_len = udp->udp_ip_rcv_options_len;
3756 	udp_bits = udp->udp_bits;
3757 	rw_exit(&udp->udp_rwlock);
3758 
3759 	switch (ipversion) {
3760 	case IPV4_VERSION:
3761 		ASSERT(MBLKL(mp) >= sizeof (ipha_t));
3762 		ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP);
3763 		hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
3764 		opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE);
3765 		if ((opt_len > 0 || udp_ip_rcv_options_len > 0) &&
3766 		    udp->udp_family == AF_INET) {
3767 			/*
3768 			 * Record/update udp_ip_rcv_options with the lock
3769 			 * held. Not needed for AF_INET6 sockets
3770 			 * since they don't support a getsockopt of IP_OPTIONS.
3771 			 */
3772 			rw_enter(&udp->udp_rwlock, RW_WRITER);
3773 			udp_save_ip_rcv_opt(udp, rptr + IP_SIMPLE_HDR_LENGTH,
3774 			    opt_len);
3775 			rw_exit(&udp->udp_rwlock);
3776 		}
3777 		/* Handle IPV6_RECVPKTINFO even for IPv4 packet. */
3778 		if ((udp->udp_family == AF_INET6) && (pinfo != NULL) &&
3779 		    udp->udp_ip_recvpktinfo) {
3780 			if (pinfo->ip_pkt_flags & IPF_RECVIF) {
3781 				ipp.ipp_fields |= IPPF_IFINDEX;
3782 				ipp.ipp_ifindex = pinfo->ip_pkt_ifindex;
3783 			}
3784 		}
3785 		break;
3786 	case IPV6_VERSION:
3787 		/*
3788 		 * IPv6 packets can only be received by applications
3789 		 * that are prepared to receive IPv6 addresses.
3790 		 * The IP fanout must ensure this.
3791 		 */
3792 		ASSERT(udp->udp_family == AF_INET6);
3793 
3794 		ip6h = (ip6_t *)rptr;
3795 		ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr);
3796 
3797 		if (ip6h->ip6_nxt != IPPROTO_UDP) {
3798 			uint8_t nexthdrp;
3799 			/* Look for ifindex information */
3800 			if (ip6h->ip6_nxt == IPPROTO_RAW) {
3801 				ip6i = (ip6i_t *)ip6h;
3802 				if ((uchar_t *)&ip6i[1] > mp->b_wptr)
3803 					goto tossit;
3804 
3805 				if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3806 					ASSERT(ip6i->ip6i_ifindex != 0);
3807 					ipp.ipp_fields |= IPPF_IFINDEX;
3808 					ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3809 				}
3810 				rptr = (uchar_t *)&ip6i[1];
3811 				mp->b_rptr = rptr;
3812 				if (rptr == mp->b_wptr) {
3813 					mp1 = mp->b_cont;
3814 					freeb(mp);
3815 					mp = mp1;
3816 					rptr = mp->b_rptr;
3817 				}
3818 				if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE))
3819 					goto tossit;
3820 				ip6h = (ip6_t *)rptr;
3821 				mp_len = msgdsize(mp);
3822 			}
3823 			/*
3824 			 * Find any potentially interesting extension headers
3825 			 * as well as the length of the IPv6 + extension
3826 			 * headers.
3827 			 */
3828 			hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) +
3829 			    UDPH_SIZE;
3830 			ASSERT(nexthdrp == IPPROTO_UDP);
3831 		} else {
3832 			hdr_length = IPV6_HDR_LEN + UDPH_SIZE;
3833 			ip6i = NULL;
3834 		}
3835 		break;
3836 	default:
3837 		ASSERT(0);
3838 	}
3839 
3840 	/*
3841 	 * IP inspected the UDP header thus all of it must be in the mblk.
3842 	 * UDP length check is performed for IPv6 packets and IPv4 packets
3843 	 * to check if the size of the packet as specified
3844 	 * by the header is the same as the physical size of the packet.
3845 	 * FIXME? Didn't IP already check this?
3846 	 */
3847 	udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
3848 	if ((MBLKL(mp) < hdr_length) ||
3849 	    (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) {
3850 		goto tossit;
3851 	}
3852 
3853 
3854 	/* Walk past the headers unless IP_RECVHDR was set. */
3855 	if (!udp_bits.udpb_rcvhdr) {
3856 		mp->b_rptr = rptr + hdr_length;
3857 		mp_len -= hdr_length;
3858 	}
3859 
3860 	/*
3861 	 * This is the inbound data path.  Packets are passed upstream as
3862 	 * T_UNITDATA_IND messages with full IP headers still attached.
3863 	 */
3864 	if (udp->udp_family == AF_INET) {
3865 		sin_t *sin;
3866 
3867 		ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
3868 
3869 		/*
3870 		 * Normally only send up the source address.
3871 		 * If IP_RECVDSTADDR is set we include the destination IP
3872 		 * address as an option. With IP_RECVOPTS we include all
3873 		 * the IP options.
3874 		 */
3875 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3876 		if (udp_bits.udpb_recvdstaddr) {
3877 			udi_size += sizeof (struct T_opthdr) +
3878 			    sizeof (struct in_addr);
3879 			UDP_STAT(us, udp_in_recvdstaddr);
3880 		}
3881 
3882 		if (udp_bits.udpb_ip_recvpktinfo && (pinfo != NULL) &&
3883 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3884 			udi_size += sizeof (struct T_opthdr) +
3885 			    sizeof (struct in_pktinfo);
3886 			UDP_STAT(us, udp_ip_rcvpktinfo);
3887 		}
3888 
3889 		if ((udp_bits.udpb_recvopts) && opt_len > 0) {
3890 			udi_size += sizeof (struct T_opthdr) + opt_len;
3891 			UDP_STAT(us, udp_in_recvopts);
3892 		}
3893 
3894 		/*
3895 		 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
3896 		 * space accordingly
3897 		 */
3898 		if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
3899 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3900 			udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
3901 			UDP_STAT(us, udp_in_recvif);
3902 		}
3903 
3904 		if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
3905 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
3906 			udi_size += sizeof (struct T_opthdr) +
3907 			    sizeof (struct sockaddr_dl);
3908 			UDP_STAT(us, udp_in_recvslla);
3909 		}
3910 
3911 		if ((udp_bits.udpb_recvucred) &&
3912 		    (cr = DB_CRED(mp)) != NULL) {
3913 			udi_size += sizeof (struct T_opthdr) + ucredsize;
3914 			cpid = DB_CPID(mp);
3915 			UDP_STAT(us, udp_in_recvucred);
3916 		}
3917 
3918 		/*
3919 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3920 		 * buffer. Since gethrestime() expects a pointer aligned
3921 		 * argument, we allocate space necessary for extra
3922 		 * alignment (even though it might not be used).
3923 		 */
3924 		if (udp_bits.udpb_timestamp) {
3925 			udi_size += sizeof (struct T_opthdr) +
3926 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3927 			UDP_STAT(us, udp_in_timestamp);
3928 		}
3929 
3930 		/*
3931 		 * If IP_RECVTTL is set allocate the appropriate sized buffer
3932 		 */
3933 		if (udp_bits.udpb_recvttl) {
3934 			udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
3935 			UDP_STAT(us, udp_in_recvttl);
3936 		}
3937 
3938 		/* Allocate a message block for the T_UNITDATA_IND structure. */
3939 		mp1 = allocb(udi_size, BPRI_MED);
3940 		if (mp1 == NULL) {
3941 			freemsg(mp);
3942 			if (options_mp != NULL)
3943 				freeb(options_mp);
3944 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
3945 			return;
3946 		}
3947 		mp1->b_cont = mp;
3948 		mp = mp1;
3949 		mp->b_datap->db_type = M_PROTO;
3950 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3951 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3952 		tudi->PRIM_type = T_UNITDATA_IND;
3953 		tudi->SRC_length = sizeof (sin_t);
3954 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3955 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
3956 		    sizeof (sin_t);
3957 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3958 		tudi->OPT_length = udi_size;
3959 		sin = (sin_t *)&tudi[1];
3960 		sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
3961 		sin->sin_port =	udpha->uha_src_port;
3962 		sin->sin_family = udp->udp_family;
3963 		*(uint32_t *)&sin->sin_zero[0] = 0;
3964 		*(uint32_t *)&sin->sin_zero[4] = 0;
3965 
3966 		/*
3967 		 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
3968 		 * IP_RECVTTL has been set.
3969 		 */
3970 		if (udi_size != 0) {
3971 			/*
3972 			 * Copy in destination address before options to avoid
3973 			 * any padding issues.
3974 			 */
3975 			char *dstopt;
3976 
3977 			dstopt = (char *)&sin[1];
3978 			if (udp_bits.udpb_recvdstaddr) {
3979 				struct T_opthdr *toh;
3980 				ipaddr_t *dstptr;
3981 
3982 				toh = (struct T_opthdr *)dstopt;
3983 				toh->level = IPPROTO_IP;
3984 				toh->name = IP_RECVDSTADDR;
3985 				toh->len = sizeof (struct T_opthdr) +
3986 				    sizeof (ipaddr_t);
3987 				toh->status = 0;
3988 				dstopt += sizeof (struct T_opthdr);
3989 				dstptr = (ipaddr_t *)dstopt;
3990 				*dstptr = ((ipha_t *)rptr)->ipha_dst;
3991 				dstopt += sizeof (ipaddr_t);
3992 				udi_size -= toh->len;
3993 			}
3994 
3995 			if (udp_bits.udpb_recvopts && opt_len > 0) {
3996 				struct T_opthdr *toh;
3997 
3998 				toh = (struct T_opthdr *)dstopt;
3999 				toh->level = IPPROTO_IP;
4000 				toh->name = IP_RECVOPTS;
4001 				toh->len = sizeof (struct T_opthdr) + opt_len;
4002 				toh->status = 0;
4003 				dstopt += sizeof (struct T_opthdr);
4004 				bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt,
4005 				    opt_len);
4006 				dstopt += opt_len;
4007 				udi_size -= toh->len;
4008 			}
4009 
4010 			if ((udp_bits.udpb_ip_recvpktinfo) && (pinfo != NULL) &&
4011 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
4012 				struct T_opthdr *toh;
4013 				struct in_pktinfo *pktinfop;
4014 
4015 				toh = (struct T_opthdr *)dstopt;
4016 				toh->level = IPPROTO_IP;
4017 				toh->name = IP_PKTINFO;
4018 				toh->len = sizeof (struct T_opthdr) +
4019 				    sizeof (*pktinfop);
4020 				toh->status = 0;
4021 				dstopt += sizeof (struct T_opthdr);
4022 				pktinfop = (struct in_pktinfo *)dstopt;
4023 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
4024 				pktinfop->ipi_spec_dst =
4025 				    pinfo->ip_pkt_match_addr;
4026 				pktinfop->ipi_addr.s_addr =
4027 				    ((ipha_t *)rptr)->ipha_dst;
4028 
4029 				dstopt += sizeof (struct in_pktinfo);
4030 				udi_size -= toh->len;
4031 			}
4032 
4033 			if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
4034 			    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
4035 
4036 				struct T_opthdr *toh;
4037 				struct sockaddr_dl	*dstptr;
4038 
4039 				toh = (struct T_opthdr *)dstopt;
4040 				toh->level = IPPROTO_IP;
4041 				toh->name = IP_RECVSLLA;
4042 				toh->len = sizeof (struct T_opthdr) +
4043 				    sizeof (struct sockaddr_dl);
4044 				toh->status = 0;
4045 				dstopt += sizeof (struct T_opthdr);
4046 				dstptr = (struct sockaddr_dl *)dstopt;
4047 				bcopy(&pinfo->ip_pkt_slla, dstptr,
4048 				    sizeof (struct sockaddr_dl));
4049 				dstopt += sizeof (struct sockaddr_dl);
4050 				udi_size -= toh->len;
4051 			}
4052 
4053 			if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
4054 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4055 
4056 				struct T_opthdr *toh;
4057 				uint_t		*dstptr;
4058 
4059 				toh = (struct T_opthdr *)dstopt;
4060 				toh->level = IPPROTO_IP;
4061 				toh->name = IP_RECVIF;
4062 				toh->len = sizeof (struct T_opthdr) +
4063 				    sizeof (uint_t);
4064 				toh->status = 0;
4065 				dstopt += sizeof (struct T_opthdr);
4066 				dstptr = (uint_t *)dstopt;
4067 				*dstptr = pinfo->ip_pkt_ifindex;
4068 				dstopt += sizeof (uint_t);
4069 				udi_size -= toh->len;
4070 			}
4071 
4072 			if (cr != NULL) {
4073 				struct T_opthdr *toh;
4074 
4075 				toh = (struct T_opthdr *)dstopt;
4076 				toh->level = SOL_SOCKET;
4077 				toh->name = SCM_UCRED;
4078 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4079 				toh->status = 0;
4080 				dstopt += sizeof (struct T_opthdr);
4081 				(void) cred2ucred(cr, cpid, dstopt, rcr);
4082 				dstopt += ucredsize;
4083 				udi_size -= toh->len;
4084 			}
4085 
4086 			if (udp_bits.udpb_timestamp) {
4087 				struct	T_opthdr *toh;
4088 
4089 				toh = (struct T_opthdr *)dstopt;
4090 				toh->level = SOL_SOCKET;
4091 				toh->name = SCM_TIMESTAMP;
4092 				toh->len = sizeof (struct T_opthdr) +
4093 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4094 				toh->status = 0;
4095 				dstopt += sizeof (struct T_opthdr);
4096 				/* Align for gethrestime() */
4097 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
4098 				    sizeof (intptr_t));
4099 				gethrestime((timestruc_t *)dstopt);
4100 				dstopt = (char *)toh + toh->len;
4101 				udi_size -= toh->len;
4102 			}
4103 
4104 			/*
4105 			 * CAUTION:
4106 			 * Due to aligment issues
4107 			 * Processing of IP_RECVTTL option
4108 			 * should always be the last. Adding
4109 			 * any option processing after this will
4110 			 * cause alignment panic.
4111 			 */
4112 			if (udp_bits.udpb_recvttl) {
4113 				struct	T_opthdr *toh;
4114 				uint8_t	*dstptr;
4115 
4116 				toh = (struct T_opthdr *)dstopt;
4117 				toh->level = IPPROTO_IP;
4118 				toh->name = IP_RECVTTL;
4119 				toh->len = sizeof (struct T_opthdr) +
4120 				    sizeof (uint8_t);
4121 				toh->status = 0;
4122 				dstopt += sizeof (struct T_opthdr);
4123 				dstptr = (uint8_t *)dstopt;
4124 				*dstptr = ((ipha_t *)rptr)->ipha_ttl;
4125 				dstopt += sizeof (uint8_t);
4126 				udi_size -= toh->len;
4127 			}
4128 
4129 			/* Consumed all of allocated space */
4130 			ASSERT(udi_size == 0);
4131 		}
4132 	} else {
4133 		sin6_t *sin6;
4134 
4135 		/*
4136 		 * Handle both IPv4 and IPv6 packets for IPv6 sockets.
4137 		 *
4138 		 * Normally we only send up the address. If receiving of any
4139 		 * optional receive side information is enabled, we also send
4140 		 * that up as options.
4141 		 */
4142 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
4143 
4144 		if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
4145 		    IPPF_RTHDR|IPPF_IFINDEX)) {
4146 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4147 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4148 				size_t hlen;
4149 
4150 				UDP_STAT(us, udp_in_recvhopopts);
4151 				hlen = copy_hop_opts(&ipp, NULL);
4152 				if (hlen == 0)
4153 					ipp.ipp_fields &= ~IPPF_HOPOPTS;
4154 				udi_size += hlen;
4155 			}
4156 			if (((udp_bits.udpb_ipv6_recvdstopts) ||
4157 			    udp_bits.udpb_old_ipv6_recvdstopts) &&
4158 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4159 				udi_size += sizeof (struct T_opthdr) +
4160 				    ipp.ipp_dstoptslen;
4161 				UDP_STAT(us, udp_in_recvdstopts);
4162 			}
4163 			if ((((udp_bits.udpb_ipv6_recvdstopts) &&
4164 			    udp_bits.udpb_ipv6_recvrthdr &&
4165 			    (ipp.ipp_fields & IPPF_RTHDR)) ||
4166 			    (udp_bits.udpb_ipv6_recvrthdrdstopts)) &&
4167 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4168 				udi_size += sizeof (struct T_opthdr) +
4169 				    ipp.ipp_rtdstoptslen;
4170 				UDP_STAT(us, udp_in_recvrtdstopts);
4171 			}
4172 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4173 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4174 				udi_size += sizeof (struct T_opthdr) +
4175 				    ipp.ipp_rthdrlen;
4176 				UDP_STAT(us, udp_in_recvrthdr);
4177 			}
4178 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4179 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4180 				udi_size += sizeof (struct T_opthdr) +
4181 				    sizeof (struct in6_pktinfo);
4182 				UDP_STAT(us, udp_in_recvpktinfo);
4183 			}
4184 
4185 		}
4186 		if ((udp_bits.udpb_recvucred) &&
4187 		    (cr = DB_CRED(mp)) != NULL) {
4188 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4189 			cpid = DB_CPID(mp);
4190 			UDP_STAT(us, udp_in_recvucred);
4191 		}
4192 
4193 		/*
4194 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4195 		 * buffer. Since gethrestime() expects a pointer aligned
4196 		 * argument, we allocate space necessary for extra
4197 		 * alignment (even though it might not be used).
4198 		 */
4199 		if (udp_bits.udpb_timestamp) {
4200 			udi_size += sizeof (struct T_opthdr) +
4201 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4202 			UDP_STAT(us, udp_in_timestamp);
4203 		}
4204 
4205 		if (udp_bits.udpb_ipv6_recvhoplimit) {
4206 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4207 			UDP_STAT(us, udp_in_recvhoplimit);
4208 		}
4209 
4210 		if (udp_bits.udpb_ipv6_recvtclass) {
4211 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4212 			UDP_STAT(us, udp_in_recvtclass);
4213 		}
4214 
4215 		mp1 = allocb(udi_size, BPRI_MED);
4216 		if (mp1 == NULL) {
4217 			freemsg(mp);
4218 			if (options_mp != NULL)
4219 				freeb(options_mp);
4220 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
4221 			return;
4222 		}
4223 		mp1->b_cont = mp;
4224 		mp = mp1;
4225 		mp->b_datap->db_type = M_PROTO;
4226 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4227 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4228 		tudi->PRIM_type = T_UNITDATA_IND;
4229 		tudi->SRC_length = sizeof (sin6_t);
4230 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4231 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4232 		    sizeof (sin6_t);
4233 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
4234 		tudi->OPT_length = udi_size;
4235 		sin6 = (sin6_t *)&tudi[1];
4236 		if (ipversion == IPV4_VERSION) {
4237 			in6_addr_t v6dst;
4238 
4239 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src,
4240 			    &sin6->sin6_addr);
4241 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst,
4242 			    &v6dst);
4243 			sin6->sin6_flowinfo = 0;
4244 			sin6->sin6_scope_id = 0;
4245 			sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst,
4246 			    connp->conn_zoneid, us->us_netstack);
4247 		} else {
4248 			sin6->sin6_addr = ip6h->ip6_src;
4249 			/* No sin6_flowinfo per API */
4250 			sin6->sin6_flowinfo = 0;
4251 			/* For link-scope source pass up scope id */
4252 			if ((ipp.ipp_fields & IPPF_IFINDEX) &&
4253 			    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
4254 				sin6->sin6_scope_id = ipp.ipp_ifindex;
4255 			else
4256 				sin6->sin6_scope_id = 0;
4257 			sin6->__sin6_src_id = ip_srcid_find_addr(
4258 			    &ip6h->ip6_dst, connp->conn_zoneid,
4259 			    us->us_netstack);
4260 		}
4261 		sin6->sin6_port = udpha->uha_src_port;
4262 		sin6->sin6_family = udp->udp_family;
4263 
4264 		if (udi_size != 0) {
4265 			uchar_t *dstopt;
4266 
4267 			dstopt = (uchar_t *)&sin6[1];
4268 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4269 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4270 				struct T_opthdr *toh;
4271 				struct in6_pktinfo *pkti;
4272 
4273 				toh = (struct T_opthdr *)dstopt;
4274 				toh->level = IPPROTO_IPV6;
4275 				toh->name = IPV6_PKTINFO;
4276 				toh->len = sizeof (struct T_opthdr) +
4277 				    sizeof (*pkti);
4278 				toh->status = 0;
4279 				dstopt += sizeof (struct T_opthdr);
4280 				pkti = (struct in6_pktinfo *)dstopt;
4281 				if (ipversion == IPV6_VERSION)
4282 					pkti->ipi6_addr = ip6h->ip6_dst;
4283 				else
4284 					IN6_IPADDR_TO_V4MAPPED(
4285 					    ((ipha_t *)rptr)->ipha_dst,
4286 					    &pkti->ipi6_addr);
4287 				pkti->ipi6_ifindex = ipp.ipp_ifindex;
4288 				dstopt += sizeof (*pkti);
4289 				udi_size -= toh->len;
4290 			}
4291 			if (udp_bits.udpb_ipv6_recvhoplimit) {
4292 				struct T_opthdr *toh;
4293 
4294 				toh = (struct T_opthdr *)dstopt;
4295 				toh->level = IPPROTO_IPV6;
4296 				toh->name = IPV6_HOPLIMIT;
4297 				toh->len = sizeof (struct T_opthdr) +
4298 				    sizeof (uint_t);
4299 				toh->status = 0;
4300 				dstopt += sizeof (struct T_opthdr);
4301 				if (ipversion == IPV6_VERSION)
4302 					*(uint_t *)dstopt = ip6h->ip6_hops;
4303 				else
4304 					*(uint_t *)dstopt =
4305 					    ((ipha_t *)rptr)->ipha_ttl;
4306 				dstopt += sizeof (uint_t);
4307 				udi_size -= toh->len;
4308 			}
4309 			if (udp_bits.udpb_ipv6_recvtclass) {
4310 				struct T_opthdr *toh;
4311 
4312 				toh = (struct T_opthdr *)dstopt;
4313 				toh->level = IPPROTO_IPV6;
4314 				toh->name = IPV6_TCLASS;
4315 				toh->len = sizeof (struct T_opthdr) +
4316 				    sizeof (uint_t);
4317 				toh->status = 0;
4318 				dstopt += sizeof (struct T_opthdr);
4319 				if (ipversion == IPV6_VERSION) {
4320 					*(uint_t *)dstopt =
4321 					    IPV6_FLOW_TCLASS(ip6h->ip6_flow);
4322 				} else {
4323 					ipha_t *ipha = (ipha_t *)rptr;
4324 					*(uint_t *)dstopt =
4325 					    ipha->ipha_type_of_service;
4326 				}
4327 				dstopt += sizeof (uint_t);
4328 				udi_size -= toh->len;
4329 			}
4330 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4331 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4332 				size_t hlen;
4333 
4334 				hlen = copy_hop_opts(&ipp, dstopt);
4335 				dstopt += hlen;
4336 				udi_size -= hlen;
4337 			}
4338 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4339 			    (udp_bits.udpb_ipv6_recvrthdr) &&
4340 			    (ipp.ipp_fields & IPPF_RTHDR) &&
4341 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4342 				struct T_opthdr *toh;
4343 
4344 				toh = (struct T_opthdr *)dstopt;
4345 				toh->level = IPPROTO_IPV6;
4346 				toh->name = IPV6_DSTOPTS;
4347 				toh->len = sizeof (struct T_opthdr) +
4348 				    ipp.ipp_rtdstoptslen;
4349 				toh->status = 0;
4350 				dstopt += sizeof (struct T_opthdr);
4351 				bcopy(ipp.ipp_rtdstopts, dstopt,
4352 				    ipp.ipp_rtdstoptslen);
4353 				dstopt += ipp.ipp_rtdstoptslen;
4354 				udi_size -= toh->len;
4355 			}
4356 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4357 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4358 				struct T_opthdr *toh;
4359 
4360 				toh = (struct T_opthdr *)dstopt;
4361 				toh->level = IPPROTO_IPV6;
4362 				toh->name = IPV6_RTHDR;
4363 				toh->len = sizeof (struct T_opthdr) +
4364 				    ipp.ipp_rthdrlen;
4365 				toh->status = 0;
4366 				dstopt += sizeof (struct T_opthdr);
4367 				bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
4368 				dstopt += ipp.ipp_rthdrlen;
4369 				udi_size -= toh->len;
4370 			}
4371 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4372 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4373 				struct T_opthdr *toh;
4374 
4375 				toh = (struct T_opthdr *)dstopt;
4376 				toh->level = IPPROTO_IPV6;
4377 				toh->name = IPV6_DSTOPTS;
4378 				toh->len = sizeof (struct T_opthdr) +
4379 				    ipp.ipp_dstoptslen;
4380 				toh->status = 0;
4381 				dstopt += sizeof (struct T_opthdr);
4382 				bcopy(ipp.ipp_dstopts, dstopt,
4383 				    ipp.ipp_dstoptslen);
4384 				dstopt += ipp.ipp_dstoptslen;
4385 				udi_size -= toh->len;
4386 			}
4387 			if (cr != NULL) {
4388 				struct T_opthdr *toh;
4389 
4390 				toh = (struct T_opthdr *)dstopt;
4391 				toh->level = SOL_SOCKET;
4392 				toh->name = SCM_UCRED;
4393 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4394 				toh->status = 0;
4395 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
4396 				dstopt += toh->len;
4397 				udi_size -= toh->len;
4398 			}
4399 			if (udp_bits.udpb_timestamp) {
4400 				struct	T_opthdr *toh;
4401 
4402 				toh = (struct T_opthdr *)dstopt;
4403 				toh->level = SOL_SOCKET;
4404 				toh->name = SCM_TIMESTAMP;
4405 				toh->len = sizeof (struct T_opthdr) +
4406 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4407 				toh->status = 0;
4408 				dstopt += sizeof (struct T_opthdr);
4409 				/* Align for gethrestime() */
4410 				dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
4411 				    sizeof (intptr_t));
4412 				gethrestime((timestruc_t *)dstopt);
4413 				dstopt = (uchar_t *)toh + toh->len;
4414 				udi_size -= toh->len;
4415 			}
4416 
4417 			/* Consumed all of allocated space */
4418 			ASSERT(udi_size == 0);
4419 		}
4420 #undef	sin6
4421 		/* No IP_RECVDSTADDR for IPv6. */
4422 	}
4423 
4424 	BUMP_MIB(&us->us_udp_mib, udpHCInDatagrams);
4425 	if (options_mp != NULL)
4426 		freeb(options_mp);
4427 
4428 	if (IPCL_IS_NONSTR(connp)) {
4429 		int error;
4430 
4431 		if ((*connp->conn_upcalls->su_recv)
4432 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
4433 		    NULL) < 0) {
4434 			mutex_enter(&udp->udp_recv_lock);
4435 			if (error == ENOSPC) {
4436 				/*
4437 				 * let's confirm while holding the lock
4438 				 */
4439 				if ((*connp->conn_upcalls->su_recv)
4440 				    (connp->conn_upper_handle, NULL, 0, 0,
4441 				    &error, NULL) < 0) {
4442 					if (error == ENOSPC) {
4443 						connp->conn_flow_cntrld =
4444 						    B_TRUE;
4445 					} else {
4446 						ASSERT(error == EOPNOTSUPP);
4447 					}
4448 				}
4449 				mutex_exit(&udp->udp_recv_lock);
4450 			} else {
4451 				ASSERT(error == EOPNOTSUPP);
4452 				udp_queue_fallback(udp, mp);
4453 			}
4454 		}
4455 	} else {
4456 		putnext(connp->conn_rq, mp);
4457 	}
4458 	ASSERT(MUTEX_NOT_HELD(&udp->udp_recv_lock));
4459 	return;
4460 
4461 tossit:
4462 	freemsg(mp);
4463 	if (options_mp != NULL)
4464 		freeb(options_mp);
4465 	BUMP_MIB(&us->us_udp_mib, udpInErrors);
4466 }
4467 
4468 /*
4469  * return SNMP stuff in buffer in mpdata. We don't hold any lock and report
4470  * information that can be changing beneath us.
4471  */
4472 mblk_t *
4473 udp_snmp_get(queue_t *q, mblk_t *mpctl)
4474 {
4475 	mblk_t			*mpdata;
4476 	mblk_t			*mp_conn_ctl;
4477 	mblk_t			*mp_attr_ctl;
4478 	mblk_t			*mp6_conn_ctl;
4479 	mblk_t			*mp6_attr_ctl;
4480 	mblk_t			*mp_conn_tail;
4481 	mblk_t			*mp_attr_tail;
4482 	mblk_t			*mp6_conn_tail;
4483 	mblk_t			*mp6_attr_tail;
4484 	struct opthdr		*optp;
4485 	mib2_udpEntry_t		ude;
4486 	mib2_udp6Entry_t	ude6;
4487 	mib2_transportMLPEntry_t mlp;
4488 	int			state;
4489 	zoneid_t		zoneid;
4490 	int			i;
4491 	connf_t			*connfp;
4492 	conn_t			*connp = Q_TO_CONN(q);
4493 	int			v4_conn_idx;
4494 	int			v6_conn_idx;
4495 	boolean_t		needattr;
4496 	udp_t			*udp;
4497 	ip_stack_t		*ipst = connp->conn_netstack->netstack_ip;
4498 	udp_stack_t		*us = connp->conn_netstack->netstack_udp;
4499 	mblk_t			*mp2ctl;
4500 
4501 	/*
4502 	 * make a copy of the original message
4503 	 */
4504 	mp2ctl = copymsg(mpctl);
4505 
4506 	mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL;
4507 	if (mpctl == NULL ||
4508 	    (mpdata = mpctl->b_cont) == NULL ||
4509 	    (mp_conn_ctl = copymsg(mpctl)) == NULL ||
4510 	    (mp_attr_ctl = copymsg(mpctl)) == NULL ||
4511 	    (mp6_conn_ctl = copymsg(mpctl)) == NULL ||
4512 	    (mp6_attr_ctl = copymsg(mpctl)) == NULL) {
4513 		freemsg(mp_conn_ctl);
4514 		freemsg(mp_attr_ctl);
4515 		freemsg(mp6_conn_ctl);
4516 		freemsg(mpctl);
4517 		freemsg(mp2ctl);
4518 		return (0);
4519 	}
4520 
4521 	zoneid = connp->conn_zoneid;
4522 
4523 	/* fixed length structure for IPv4 and IPv6 counters */
4524 	SET_MIB(us->us_udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t));
4525 	SET_MIB(us->us_udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t));
4526 	/* synchronize 64- and 32-bit counters */
4527 	SYNC32_MIB(&us->us_udp_mib, udpInDatagrams, udpHCInDatagrams);
4528 	SYNC32_MIB(&us->us_udp_mib, udpOutDatagrams, udpHCOutDatagrams);
4529 
4530 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4531 	optp->level = MIB2_UDP;
4532 	optp->name = 0;
4533 	(void) snmp_append_data(mpdata, (char *)&us->us_udp_mib,
4534 	    sizeof (us->us_udp_mib));
4535 	optp->len = msgdsize(mpdata);
4536 	qreply(q, mpctl);
4537 
4538 	mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL;
4539 	v4_conn_idx = v6_conn_idx = 0;
4540 
4541 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4542 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4543 		connp = NULL;
4544 
4545 		while ((connp = ipcl_get_next_conn(connfp, connp,
4546 		    IPCL_UDPCONN))) {
4547 			udp = connp->conn_udp;
4548 			if (zoneid != connp->conn_zoneid)
4549 				continue;
4550 
4551 			/*
4552 			 * Note that the port numbers are sent in
4553 			 * host byte order
4554 			 */
4555 
4556 			if (udp->udp_state == TS_UNBND)
4557 				state = MIB2_UDP_unbound;
4558 			else if (udp->udp_state == TS_IDLE)
4559 				state = MIB2_UDP_idle;
4560 			else if (udp->udp_state == TS_DATA_XFER)
4561 				state = MIB2_UDP_connected;
4562 			else
4563 				state = MIB2_UDP_unknown;
4564 
4565 			needattr = B_FALSE;
4566 			bzero(&mlp, sizeof (mlp));
4567 			if (connp->conn_mlp_type != mlptSingle) {
4568 				if (connp->conn_mlp_type == mlptShared ||
4569 				    connp->conn_mlp_type == mlptBoth)
4570 					mlp.tme_flags |= MIB2_TMEF_SHARED;
4571 				if (connp->conn_mlp_type == mlptPrivate ||
4572 				    connp->conn_mlp_type == mlptBoth)
4573 					mlp.tme_flags |= MIB2_TMEF_PRIVATE;
4574 				needattr = B_TRUE;
4575 			}
4576 
4577 			/*
4578 			 * Create an IPv4 table entry for IPv4 entries and also
4579 			 * any IPv6 entries which are bound to in6addr_any
4580 			 * (i.e. anything a IPv4 peer could connect/send to).
4581 			 */
4582 			if (udp->udp_ipversion == IPV4_VERSION ||
4583 			    (udp->udp_state <= TS_IDLE &&
4584 			    IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) {
4585 				ude.udpEntryInfo.ue_state = state;
4586 				/*
4587 				 * If in6addr_any this will set it to
4588 				 * INADDR_ANY
4589 				 */
4590 				ude.udpLocalAddress =
4591 				    V4_PART_OF_V6(udp->udp_v6src);
4592 				ude.udpLocalPort = ntohs(udp->udp_port);
4593 				if (udp->udp_state == TS_DATA_XFER) {
4594 					/*
4595 					 * Can potentially get here for
4596 					 * v6 socket if another process
4597 					 * (say, ping) has just done a
4598 					 * sendto(), changing the state
4599 					 * from the TS_IDLE above to
4600 					 * TS_DATA_XFER by the time we hit
4601 					 * this part of the code.
4602 					 */
4603 					ude.udpEntryInfo.ue_RemoteAddress =
4604 					    V4_PART_OF_V6(udp->udp_v6dst);
4605 					ude.udpEntryInfo.ue_RemotePort =
4606 					    ntohs(udp->udp_dstport);
4607 				} else {
4608 					ude.udpEntryInfo.ue_RemoteAddress = 0;
4609 					ude.udpEntryInfo.ue_RemotePort = 0;
4610 				}
4611 
4612 				/*
4613 				 * We make the assumption that all udp_t
4614 				 * structs will be created within an address
4615 				 * region no larger than 32-bits.
4616 				 */
4617 				ude.udpInstance = (uint32_t)(uintptr_t)udp;
4618 				ude.udpCreationProcess =
4619 				    (udp->udp_open_pid < 0) ?
4620 				    MIB2_UNKNOWN_PROCESS :
4621 				    udp->udp_open_pid;
4622 				ude.udpCreationTime = udp->udp_open_time;
4623 
4624 				(void) snmp_append_data2(mp_conn_ctl->b_cont,
4625 				    &mp_conn_tail, (char *)&ude, sizeof (ude));
4626 				mlp.tme_connidx = v4_conn_idx++;
4627 				if (needattr)
4628 					(void) snmp_append_data2(
4629 					    mp_attr_ctl->b_cont, &mp_attr_tail,
4630 					    (char *)&mlp, sizeof (mlp));
4631 			}
4632 			if (udp->udp_ipversion == IPV6_VERSION) {
4633 				ude6.udp6EntryInfo.ue_state  = state;
4634 				ude6.udp6LocalAddress = udp->udp_v6src;
4635 				ude6.udp6LocalPort = ntohs(udp->udp_port);
4636 				ude6.udp6IfIndex = udp->udp_bound_if;
4637 				if (udp->udp_state == TS_DATA_XFER) {
4638 					ude6.udp6EntryInfo.ue_RemoteAddress =
4639 					    udp->udp_v6dst;
4640 					ude6.udp6EntryInfo.ue_RemotePort =
4641 					    ntohs(udp->udp_dstport);
4642 				} else {
4643 					ude6.udp6EntryInfo.ue_RemoteAddress =
4644 					    sin6_null.sin6_addr;
4645 					ude6.udp6EntryInfo.ue_RemotePort = 0;
4646 				}
4647 				/*
4648 				 * We make the assumption that all udp_t
4649 				 * structs will be created within an address
4650 				 * region no larger than 32-bits.
4651 				 */
4652 				ude6.udp6Instance = (uint32_t)(uintptr_t)udp;
4653 				ude6.udp6CreationProcess =
4654 				    (udp->udp_open_pid < 0) ?
4655 				    MIB2_UNKNOWN_PROCESS :
4656 				    udp->udp_open_pid;
4657 				ude6.udp6CreationTime = udp->udp_open_time;
4658 
4659 				(void) snmp_append_data2(mp6_conn_ctl->b_cont,
4660 				    &mp6_conn_tail, (char *)&ude6,
4661 				    sizeof (ude6));
4662 				mlp.tme_connidx = v6_conn_idx++;
4663 				if (needattr)
4664 					(void) snmp_append_data2(
4665 					    mp6_attr_ctl->b_cont,
4666 					    &mp6_attr_tail, (char *)&mlp,
4667 					    sizeof (mlp));
4668 			}
4669 		}
4670 	}
4671 
4672 	/* IPv4 UDP endpoints */
4673 	optp = (struct opthdr *)&mp_conn_ctl->b_rptr[
4674 	    sizeof (struct T_optmgmt_ack)];
4675 	optp->level = MIB2_UDP;
4676 	optp->name = MIB2_UDP_ENTRY;
4677 	optp->len = msgdsize(mp_conn_ctl->b_cont);
4678 	qreply(q, mp_conn_ctl);
4679 
4680 	/* table of MLP attributes... */
4681 	optp = (struct opthdr *)&mp_attr_ctl->b_rptr[
4682 	    sizeof (struct T_optmgmt_ack)];
4683 	optp->level = MIB2_UDP;
4684 	optp->name = EXPER_XPORT_MLP;
4685 	optp->len = msgdsize(mp_attr_ctl->b_cont);
4686 	if (optp->len == 0)
4687 		freemsg(mp_attr_ctl);
4688 	else
4689 		qreply(q, mp_attr_ctl);
4690 
4691 	/* IPv6 UDP endpoints */
4692 	optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[
4693 	    sizeof (struct T_optmgmt_ack)];
4694 	optp->level = MIB2_UDP6;
4695 	optp->name = MIB2_UDP6_ENTRY;
4696 	optp->len = msgdsize(mp6_conn_ctl->b_cont);
4697 	qreply(q, mp6_conn_ctl);
4698 
4699 	/* table of MLP attributes... */
4700 	optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[
4701 	    sizeof (struct T_optmgmt_ack)];
4702 	optp->level = MIB2_UDP6;
4703 	optp->name = EXPER_XPORT_MLP;
4704 	optp->len = msgdsize(mp6_attr_ctl->b_cont);
4705 	if (optp->len == 0)
4706 		freemsg(mp6_attr_ctl);
4707 	else
4708 		qreply(q, mp6_attr_ctl);
4709 
4710 	return (mp2ctl);
4711 }
4712 
4713 /*
4714  * Return 0 if invalid set request, 1 otherwise, including non-udp requests.
4715  * NOTE: Per MIB-II, UDP has no writable data.
4716  * TODO:  If this ever actually tries to set anything, it needs to be
4717  * to do the appropriate locking.
4718  */
4719 /* ARGSUSED */
4720 int
4721 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4722     uchar_t *ptr, int len)
4723 {
4724 	switch (level) {
4725 	case MIB2_UDP:
4726 		return (0);
4727 	default:
4728 		return (1);
4729 	}
4730 }
4731 
4732 static void
4733 udp_report_item(mblk_t *mp, udp_t *udp)
4734 {
4735 	char *state;
4736 	char addrbuf1[INET6_ADDRSTRLEN];
4737 	char addrbuf2[INET6_ADDRSTRLEN];
4738 	uint_t print_len, buf_len;
4739 
4740 	buf_len = mp->b_datap->db_lim - mp->b_wptr;
4741 	ASSERT(buf_len >= 0);
4742 	if (buf_len == 0)
4743 		return;
4744 
4745 	if (udp->udp_state == TS_UNBND)
4746 		state = "UNBOUND";
4747 	else if (udp->udp_state == TS_IDLE)
4748 		state = "IDLE";
4749 	else if (udp->udp_state == TS_DATA_XFER)
4750 		state = "CONNECTED";
4751 	else
4752 		state = "UnkState";
4753 	print_len = snprintf((char *)mp->b_wptr, buf_len,
4754 	    MI_COL_PTRFMT_STR "%4d %5u %s %s %5u %s\n",
4755 	    (void *)udp, udp->udp_connp->conn_zoneid, ntohs(udp->udp_port),
4756 	    inet_ntop(AF_INET6, &udp->udp_v6src, addrbuf1, sizeof (addrbuf1)),
4757 	    inet_ntop(AF_INET6, &udp->udp_v6dst, addrbuf2, sizeof (addrbuf2)),
4758 	    ntohs(udp->udp_dstport), state);
4759 	if (print_len < buf_len) {
4760 		mp->b_wptr += print_len;
4761 	} else {
4762 		mp->b_wptr += buf_len;
4763 	}
4764 }
4765 
4766 /* Report for ndd "udp_status" */
4767 /* ARGSUSED */
4768 static int
4769 udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4770 {
4771 	zoneid_t zoneid;
4772 	connf_t	*connfp;
4773 	conn_t	*connp = Q_TO_CONN(q);
4774 	udp_t	*udp = connp->conn_udp;
4775 	int	i;
4776 	udp_stack_t *us = udp->udp_us;
4777 	ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
4778 
4779 	/*
4780 	 * Because of the ndd constraint, at most we can have 64K buffer
4781 	 * to put in all UDP info.  So to be more efficient, just
4782 	 * allocate a 64K buffer here, assuming we need that large buffer.
4783 	 * This may be a problem as any user can read udp_status.  Therefore
4784 	 * we limit the rate of doing this using us_ndd_get_info_interval.
4785 	 * This should be OK as normal users should not do this too often.
4786 	 */
4787 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
4788 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
4789 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
4790 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
4791 			return (0);
4792 		}
4793 	}
4794 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
4795 		/* The following may work even if we cannot get a large buf. */
4796 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
4797 		return (0);
4798 	}
4799 	(void) mi_mpprintf(mp,
4800 	    "UDP     " MI_COL_HDRPAD_STR
4801 	/*   12345678[89ABCDEF] */
4802 	    " zone lport src addr        dest addr       port  state");
4803 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
4804 
4805 	zoneid = connp->conn_zoneid;
4806 
4807 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4808 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4809 		connp = NULL;
4810 
4811 		while ((connp = ipcl_get_next_conn(connfp, connp,
4812 		    IPCL_UDPCONN))) {
4813 			udp = connp->conn_udp;
4814 			if (zoneid != GLOBAL_ZONEID &&
4815 			    zoneid != connp->conn_zoneid)
4816 				continue;
4817 
4818 			udp_report_item(mp->b_cont, udp);
4819 		}
4820 	}
4821 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
4822 	return (0);
4823 }
4824 
4825 /*
4826  * This routine creates a T_UDERROR_IND message and passes it upstream.
4827  * The address and options are copied from the T_UNITDATA_REQ message
4828  * passed in mp.  This message is freed.
4829  */
4830 static void
4831 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen,
4832     t_scalar_t err)
4833 {
4834 	struct T_unitdata_req *tudr;
4835 	mblk_t	*mp1;
4836 	uchar_t	*optaddr;
4837 	t_scalar_t optlen;
4838 
4839 	if (DB_TYPE(mp) == M_DATA) {
4840 		ASSERT(destaddr != NULL && destlen != 0);
4841 		optaddr = NULL;
4842 		optlen = 0;
4843 	} else {
4844 		if ((mp->b_wptr < mp->b_rptr) ||
4845 		    (MBLKL(mp)) < sizeof (struct T_unitdata_req)) {
4846 			goto done;
4847 		}
4848 		tudr = (struct T_unitdata_req *)mp->b_rptr;
4849 		destaddr = mp->b_rptr + tudr->DEST_offset;
4850 		if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr ||
4851 		    destaddr + tudr->DEST_length < mp->b_rptr ||
4852 		    destaddr + tudr->DEST_length > mp->b_wptr) {
4853 			goto done;
4854 		}
4855 		optaddr = mp->b_rptr + tudr->OPT_offset;
4856 		if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr ||
4857 		    optaddr + tudr->OPT_length < mp->b_rptr ||
4858 		    optaddr + tudr->OPT_length > mp->b_wptr) {
4859 			goto done;
4860 		}
4861 		destlen = tudr->DEST_length;
4862 		optlen = tudr->OPT_length;
4863 	}
4864 
4865 	mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen,
4866 	    (char *)optaddr, optlen, err);
4867 	if (mp1 != NULL)
4868 		qreply(q, mp1);
4869 
4870 done:
4871 	freemsg(mp);
4872 }
4873 
4874 /*
4875  * This routine removes a port number association from a stream.  It
4876  * is called by udp_wput to handle T_UNBIND_REQ messages.
4877  */
4878 static void
4879 udp_tpi_unbind(queue_t *q, mblk_t *mp)
4880 {
4881 	conn_t	*connp = Q_TO_CONN(q);
4882 	int	error;
4883 
4884 	error = udp_do_unbind(connp);
4885 	if (error) {
4886 		if (error < 0)
4887 			udp_err_ack(q, mp, -error, 0);
4888 		else
4889 			udp_err_ack(q, mp, TSYSERR, error);
4890 		return;
4891 	}
4892 
4893 	mp = mi_tpi_ok_ack_alloc(mp);
4894 	ASSERT(mp != NULL);
4895 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4896 	qreply(q, mp);
4897 }
4898 
4899 /*
4900  * Don't let port fall into the privileged range.
4901  * Since the extra privileged ports can be arbitrary we also
4902  * ensure that we exclude those from consideration.
4903  * us->us_epriv_ports is not sorted thus we loop over it until
4904  * there are no changes.
4905  */
4906 static in_port_t
4907 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random)
4908 {
4909 	int i;
4910 	in_port_t nextport;
4911 	boolean_t restart = B_FALSE;
4912 	udp_stack_t *us = udp->udp_us;
4913 
4914 	if (random && udp_random_anon_port != 0) {
4915 		(void) random_get_pseudo_bytes((uint8_t *)&port,
4916 		    sizeof (in_port_t));
4917 		/*
4918 		 * Unless changed by a sys admin, the smallest anon port
4919 		 * is 32768 and the largest anon port is 65535.  It is
4920 		 * very likely (50%) for the random port to be smaller
4921 		 * than the smallest anon port.  When that happens,
4922 		 * add port % (anon port range) to the smallest anon
4923 		 * port to get the random port.  It should fall into the
4924 		 * valid anon port range.
4925 		 */
4926 		if (port < us->us_smallest_anon_port) {
4927 			port = us->us_smallest_anon_port +
4928 			    port % (us->us_largest_anon_port -
4929 			    us->us_smallest_anon_port);
4930 		}
4931 	}
4932 
4933 retry:
4934 	if (port < us->us_smallest_anon_port)
4935 		port = us->us_smallest_anon_port;
4936 
4937 	if (port > us->us_largest_anon_port) {
4938 		port = us->us_smallest_anon_port;
4939 		if (restart)
4940 			return (0);
4941 		restart = B_TRUE;
4942 	}
4943 
4944 	if (port < us->us_smallest_nonpriv_port)
4945 		port = us->us_smallest_nonpriv_port;
4946 
4947 	for (i = 0; i < us->us_num_epriv_ports; i++) {
4948 		if (port == us->us_epriv_ports[i]) {
4949 			port++;
4950 			/*
4951 			 * Make sure that the port is in the
4952 			 * valid range.
4953 			 */
4954 			goto retry;
4955 		}
4956 	}
4957 
4958 	if (is_system_labeled() &&
4959 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
4960 	    port, IPPROTO_UDP, B_TRUE)) != 0) {
4961 		port = nextport;
4962 		goto retry;
4963 	}
4964 
4965 	return (port);
4966 }
4967 
4968 static int
4969 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst,
4970     boolean_t *update_lastdst)
4971 {
4972 	int err;
4973 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4974 	udp_t *udp = Q_TO_UDP(wq);
4975 	udp_stack_t	*us = udp->udp_us;
4976 
4977 	err = tsol_compute_label(DB_CREDDEF(mp, udp->udp_connp->conn_cred), dst,
4978 	    opt_storage, udp->udp_connp->conn_mac_exempt,
4979 	    us->us_netstack->netstack_ip);
4980 	if (err == 0) {
4981 		err = tsol_update_options(&udp->udp_ip_snd_options,
4982 		    &udp->udp_ip_snd_options_len, &udp->udp_label_len,
4983 		    opt_storage);
4984 	}
4985 	if (err != 0) {
4986 		DTRACE_PROBE4(
4987 		    tx__ip__log__info__updatelabel__udp,
4988 		    char *, "queue(1) failed to update options(2) on mp(3)",
4989 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4990 	} else {
4991 		*update_lastdst = B_TRUE;
4992 	}
4993 	return (err);
4994 }
4995 
4996 static mblk_t *
4997 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port,
4998     uint_t srcid, int *error, boolean_t insert_spi, struct nmsghdr *msg,
4999     cred_t *cr, pid_t pid)
5000 {
5001 	udp_t		*udp = connp->conn_udp;
5002 	mblk_t		*mp1 = mp;
5003 	mblk_t		*mp2;
5004 	ipha_t		*ipha;
5005 	int		ip_hdr_length;
5006 	uint32_t 	ip_len;
5007 	udpha_t		*udpha;
5008 	boolean_t 	lock_held = B_FALSE;
5009 	in_port_t	uha_src_port;
5010 	udpattrs_t	attrs;
5011 	uchar_t		ip_snd_opt[IP_MAX_OPT_LENGTH];
5012 	uint32_t	ip_snd_opt_len = 0;
5013 	ip4_pkt_t  	pktinfo;
5014 	ip4_pkt_t  	*pktinfop = &pktinfo;
5015 	ip_opt_info_t	optinfo;
5016 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5017 	udp_stack_t	*us = udp->udp_us;
5018 	ipsec_stack_t	*ipss = ipst->ips_netstack->netstack_ipsec;
5019 	queue_t		*q = connp->conn_wq;
5020 	ire_t		*ire;
5021 	in6_addr_t	v6dst;
5022 	boolean_t	update_lastdst = B_FALSE;
5023 
5024 	*error = 0;
5025 	pktinfop->ip4_ill_index = 0;
5026 	pktinfop->ip4_addr = INADDR_ANY;
5027 	optinfo.ip_opt_flags = 0;
5028 	optinfo.ip_opt_ill_index = 0;
5029 
5030 	if (v4dst == INADDR_ANY)
5031 		v4dst = htonl(INADDR_LOOPBACK);
5032 
5033 	/*
5034 	 * If options passed in, feed it for verification and handling
5035 	 */
5036 	attrs.udpattr_credset = B_FALSE;
5037 	if (IPCL_IS_NONSTR(connp)) {
5038 		if (msg->msg_controllen != 0) {
5039 			attrs.udpattr_ipp4 = pktinfop;
5040 			attrs.udpattr_mb = mp;
5041 
5042 			rw_enter(&udp->udp_rwlock, RW_WRITER);
5043 			*error = process_auxiliary_options(connp,
5044 			    msg->msg_control, msg->msg_controllen,
5045 			    &attrs, &udp_opt_obj, udp_opt_set);
5046 			rw_exit(&udp->udp_rwlock);
5047 			if (*error)
5048 				goto done;
5049 		}
5050 	} else {
5051 		if (DB_TYPE(mp) != M_DATA) {
5052 			mp1 = mp->b_cont;
5053 			if (((struct T_unitdata_req *)
5054 			    mp->b_rptr)->OPT_length != 0) {
5055 				attrs.udpattr_ipp4 = pktinfop;
5056 				attrs.udpattr_mb = mp;
5057 				if (udp_unitdata_opt_process(q, mp, error,
5058 				    &attrs) < 0)
5059 					goto done;
5060 				/*
5061 				 * Note: success in processing options.
5062 				 * mp option buffer represented by
5063 				 * OPT_length/offset now potentially modified
5064 				 * and contain option setting results
5065 				 */
5066 				ASSERT(*error == 0);
5067 			}
5068 		}
5069 	}
5070 
5071 	/* mp1 points to the M_DATA mblk carrying the packet */
5072 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
5073 
5074 	/*
5075 	 * Determine whether we need to mark the mblk with the user's
5076 	 * credentials.
5077 	 */
5078 	ire = connp->conn_ire_cache;
5079 	if (is_system_labeled() || CLASSD(v4dst) || (ire == NULL) ||
5080 	    (ire->ire_addr != v4dst) ||
5081 	    (ire->ire_type & (IRE_BROADCAST | IRE_LOCAL | IRE_LOOPBACK))) {
5082 		if (cr != NULL && DB_CRED(mp) == NULL)
5083 			msg_setcredpid(mp, cr, pid);
5084 	}
5085 
5086 	rw_enter(&udp->udp_rwlock, RW_READER);
5087 	lock_held = B_TRUE;
5088 
5089 	/*
5090 	 * Cluster and TSOL note:
5091 	 *    udp.udp_v6lastdst		is shared by Cluster and TSOL
5092 	 *    udp.udp_lastdstport	is used by Cluster
5093 	 *
5094 	 * Both Cluster and TSOL need to update the dest addr and/or port.
5095 	 * Updating is done after both Cluster and TSOL checks, protected
5096 	 * by conn_lock.
5097 	 */
5098 	mutex_enter(&connp->conn_lock);
5099 
5100 	if (cl_inet_connect2 != NULL &&
5101 	    (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
5102 	    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
5103 	    udp->udp_lastdstport != port)) {
5104 		mutex_exit(&connp->conn_lock);
5105 		*error = 0;
5106 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
5107 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &v6dst, port, *error);
5108 		if (*error != 0) {
5109 			*error = EHOSTUNREACH;
5110 			goto done;
5111 		}
5112 		update_lastdst = B_TRUE;
5113 		mutex_enter(&connp->conn_lock);
5114 	}
5115 
5116 	/*
5117 	 * Check if our saved options are valid; update if not.
5118 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
5119 	 * to different destination may require different labels,
5120 	 * or worse, UDP packets to same IP address may require
5121 	 * different labels due to use of shared all-zones address.
5122 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
5123 	 * and ip_snd_options_len are consistent for the current
5124 	 * destination and are updated atomically.
5125 	 */
5126 	if (is_system_labeled()) {
5127 		/* Using UDP MLP requires SCM_UCRED from user */
5128 		if (connp->conn_mlp_type != mlptSingle &&
5129 		    !attrs.udpattr_credset) {
5130 			mutex_exit(&connp->conn_lock);
5131 			DTRACE_PROBE4(
5132 			    tx__ip__log__info__output__udp,
5133 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
5134 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
5135 			*error = ECONNREFUSED;
5136 			goto done;
5137 		}
5138 		/*
5139 		 * update label option for this UDP socket if
5140 		 * - the destination has changed, or
5141 		 * - the UDP socket is MLP
5142 		 */
5143 		if ((!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
5144 		    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
5145 		    connp->conn_mlp_type != mlptSingle) &&
5146 		    (*error = udp_update_label(q, mp, v4dst, &update_lastdst))
5147 		    != 0) {
5148 			mutex_exit(&connp->conn_lock);
5149 			goto done;
5150 		}
5151 	}
5152 	if (update_lastdst) {
5153 		IN6_IPADDR_TO_V4MAPPED(v4dst, &udp->udp_v6lastdst);
5154 		udp->udp_lastdstport = port;
5155 	}
5156 	if (udp->udp_ip_snd_options_len > 0) {
5157 		ip_snd_opt_len = udp->udp_ip_snd_options_len;
5158 		bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
5159 	}
5160 	mutex_exit(&connp->conn_lock);
5161 
5162 	/* Add an IP header */
5163 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len +
5164 	    (insert_spi ? sizeof (uint32_t) : 0);
5165 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
5166 	if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) ||
5167 	    !OK_32PTR(ipha)) {
5168 		mp2 = allocb(ip_hdr_length + us->us_wroff_extra, BPRI_LO);
5169 		if (mp2 == NULL) {
5170 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5171 			    "udp_wput_end: q %p (%S)", q, "allocbfail2");
5172 			*error = ENOMEM;
5173 			goto done;
5174 		}
5175 		mp2->b_wptr = DB_LIM(mp2);
5176 		mp2->b_cont = mp1;
5177 		mp1 = mp2;
5178 		if (DB_TYPE(mp) != M_DATA)
5179 			mp->b_cont = mp1;
5180 		else
5181 			mp = mp1;
5182 
5183 		ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length);
5184 	}
5185 	ip_hdr_length -= (UDPH_SIZE + (insert_spi ? sizeof (uint32_t) : 0));
5186 #ifdef	_BIG_ENDIAN
5187 	/* Set version, header length, and tos */
5188 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5189 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
5190 	    udp->udp_type_of_service);
5191 	/* Set ttl and protocol */
5192 	*(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP;
5193 #else
5194 	/* Set version, header length, and tos */
5195 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5196 	    ((udp->udp_type_of_service << 8) |
5197 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
5198 	/* Set ttl and protocol */
5199 	*(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl;
5200 #endif
5201 	if (pktinfop->ip4_addr != INADDR_ANY) {
5202 		ipha->ipha_src = pktinfop->ip4_addr;
5203 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
5204 	} else {
5205 		/*
5206 		 * Copy our address into the packet.  If this is zero,
5207 		 * first look at __sin6_src_id for a hint. If we leave the
5208 		 * source as INADDR_ANY then ip will fill in the real source
5209 		 * address.
5210 		 */
5211 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src);
5212 		if (srcid != 0 && ipha->ipha_src == INADDR_ANY) {
5213 			in6_addr_t v6src;
5214 
5215 			ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid,
5216 			    us->us_netstack);
5217 			IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src);
5218 		}
5219 	}
5220 	uha_src_port = udp->udp_port;
5221 	if (ip_hdr_length == IP_SIMPLE_HDR_LENGTH) {
5222 		rw_exit(&udp->udp_rwlock);
5223 		lock_held = B_FALSE;
5224 	}
5225 
5226 	if (pktinfop->ip4_ill_index != 0) {
5227 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
5228 	}
5229 
5230 	ipha->ipha_fragment_offset_and_flags = 0;
5231 	ipha->ipha_ident = 0;
5232 
5233 	mp1->b_rptr = (uchar_t *)ipha;
5234 
5235 	ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <=
5236 	    (uintptr_t)UINT_MAX);
5237 
5238 	/* Determine length of packet */
5239 	ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha);
5240 	if ((mp2 = mp1->b_cont) != NULL) {
5241 		do {
5242 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
5243 			ip_len += (uint32_t)MBLKL(mp2);
5244 		} while ((mp2 = mp2->b_cont) != NULL);
5245 	}
5246 	/*
5247 	 * If the size of the packet is greater than the maximum allowed by
5248 	 * ip, return an error. Passing this down could cause panics because
5249 	 * the size will have wrapped and be inconsistent with the msg size.
5250 	 */
5251 	if (ip_len > IP_MAXPACKET) {
5252 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5253 		    "udp_wput_end: q %p (%S)", q, "IP length exceeded");
5254 		*error = EMSGSIZE;
5255 		goto done;
5256 	}
5257 	ipha->ipha_length = htons((uint16_t)ip_len);
5258 	ip_len -= ip_hdr_length;
5259 	ip_len = htons((uint16_t)ip_len);
5260 	udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length);
5261 
5262 	/* Insert all-0s SPI now. */
5263 	if (insert_spi)
5264 		*((uint32_t *)(udpha + 1)) = 0;
5265 
5266 	/*
5267 	 * Copy in the destination address
5268 	 */
5269 	ipha->ipha_dst = v4dst;
5270 
5271 	/*
5272 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
5273 	 */
5274 	if (CLASSD(v4dst))
5275 		ipha->ipha_ttl = udp->udp_multicast_ttl;
5276 
5277 	udpha->uha_dst_port = port;
5278 	udpha->uha_src_port = uha_src_port;
5279 
5280 	if (ip_snd_opt_len > 0) {
5281 		uint32_t	cksum;
5282 
5283 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
5284 		lock_held = B_FALSE;
5285 		rw_exit(&udp->udp_rwlock);
5286 		/*
5287 		 * Massage source route putting first source route in ipha_dst.
5288 		 * Ignore the destination in T_unitdata_req.
5289 		 * Create a checksum adjustment for a source route, if any.
5290 		 */
5291 		cksum = ip_massage_options(ipha, us->us_netstack);
5292 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5293 		cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) +
5294 		    (ipha->ipha_dst & 0xFFFF);
5295 		if ((int)cksum < 0)
5296 			cksum--;
5297 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5298 		/*
5299 		 * IP does the checksum if uha_checksum is non-zero,
5300 		 * We make it easy for IP to include our pseudo header
5301 		 * by putting our length in uha_checksum.
5302 		 */
5303 		cksum += ip_len;
5304 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5305 		/* There might be a carry. */
5306 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5307 #ifdef _LITTLE_ENDIAN
5308 		if (us->us_do_checksum)
5309 			ip_len = (cksum << 16) | ip_len;
5310 #else
5311 		if (us->us_do_checksum)
5312 			ip_len = (ip_len << 16) | cksum;
5313 		else
5314 			ip_len <<= 16;
5315 #endif
5316 	} else {
5317 		/*
5318 		 * IP does the checksum if uha_checksum is non-zero,
5319 		 * We make it easy for IP to include our pseudo header
5320 		 * by putting our length in uha_checksum.
5321 		 */
5322 		if (us->us_do_checksum)
5323 			ip_len |= (ip_len << 16);
5324 #ifndef _LITTLE_ENDIAN
5325 		else
5326 			ip_len <<= 16;
5327 #endif
5328 	}
5329 	ASSERT(!lock_held);
5330 	/* Set UDP length and checksum */
5331 	*((uint32_t *)&udpha->uha_length) = ip_len;
5332 	if (DB_CRED(mp) != NULL)
5333 		mblk_setcred(mp1, DB_CRED(mp));
5334 
5335 	if (DB_TYPE(mp) != M_DATA) {
5336 		ASSERT(mp != mp1);
5337 		freeb(mp);
5338 	}
5339 
5340 	/* mp has been consumed and we'll return success */
5341 	ASSERT(*error == 0);
5342 	mp = NULL;
5343 
5344 	/* We're done.  Pass the packet to ip. */
5345 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
5346 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5347 	    "udp_wput_end: q %p (%S)", q, "end");
5348 
5349 	if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 ||
5350 	    CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) ||
5351 	    connp->conn_dontroute ||
5352 	    connp->conn_outgoing_ill != NULL || optinfo.ip_opt_flags != 0 ||
5353 	    optinfo.ip_opt_ill_index != 0 ||
5354 	    ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION ||
5355 	    IPP_ENABLED(IPP_LOCAL_OUT, ipst) ||
5356 	    ipst->ips_ip_g_mrouter != NULL) {
5357 		UDP_STAT(us, udp_ip_send);
5358 		ip_output_options(connp, mp1, connp->conn_wq, IP_WPUT,
5359 		    &optinfo);
5360 	} else {
5361 		udp_send_data(udp, connp->conn_wq, mp1, ipha);
5362 	}
5363 
5364 done:
5365 	if (lock_held)
5366 		rw_exit(&udp->udp_rwlock);
5367 	if (*error != 0) {
5368 		ASSERT(mp != NULL);
5369 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5370 	}
5371 	return (mp);
5372 }
5373 
5374 static void
5375 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha)
5376 {
5377 	conn_t	*connp = udp->udp_connp;
5378 	ipaddr_t src, dst;
5379 	ire_t	*ire;
5380 	ipif_t	*ipif = NULL;
5381 	mblk_t	*ire_fp_mp;
5382 	boolean_t retry_caching;
5383 	udp_stack_t *us = udp->udp_us;
5384 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5385 
5386 	dst = ipha->ipha_dst;
5387 	src = ipha->ipha_src;
5388 	ASSERT(ipha->ipha_ident == 0);
5389 
5390 	if (CLASSD(dst)) {
5391 		int err;
5392 
5393 		ipif = conn_get_held_ipif(connp,
5394 		    &connp->conn_multicast_ipif, &err);
5395 
5396 		if (ipif == NULL || ipif->ipif_isv6 ||
5397 		    (ipif->ipif_ill->ill_phyint->phyint_flags &
5398 		    PHYI_LOOPBACK)) {
5399 			if (ipif != NULL)
5400 				ipif_refrele(ipif);
5401 			UDP_STAT(us, udp_ip_send);
5402 			ip_output(connp, mp, q, IP_WPUT);
5403 			return;
5404 		}
5405 	}
5406 
5407 	retry_caching = B_FALSE;
5408 	mutex_enter(&connp->conn_lock);
5409 	ire = connp->conn_ire_cache;
5410 	ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT));
5411 
5412 	if (ire == NULL || ire->ire_addr != dst ||
5413 	    (ire->ire_marks & IRE_MARK_CONDEMNED)) {
5414 		retry_caching = B_TRUE;
5415 	} else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) {
5416 		ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr;
5417 
5418 		ASSERT(ipif != NULL);
5419 		if (!IS_ON_SAME_LAN(stq_ill, ipif->ipif_ill))
5420 			retry_caching = B_TRUE;
5421 	}
5422 
5423 	if (!retry_caching) {
5424 		ASSERT(ire != NULL);
5425 		IRE_REFHOLD(ire);
5426 		mutex_exit(&connp->conn_lock);
5427 	} else {
5428 		boolean_t cached = B_FALSE;
5429 
5430 		connp->conn_ire_cache = NULL;
5431 		mutex_exit(&connp->conn_lock);
5432 
5433 		/* Release the old ire */
5434 		if (ire != NULL) {
5435 			IRE_REFRELE_NOTR(ire);
5436 			ire = NULL;
5437 		}
5438 
5439 		if (CLASSD(dst)) {
5440 			ASSERT(ipif != NULL);
5441 			ire = ire_ctable_lookup(dst, 0, 0, ipif,
5442 			    connp->conn_zoneid, MBLK_GETLABEL(mp),
5443 			    MATCH_IRE_ILL, ipst);
5444 		} else {
5445 			ASSERT(ipif == NULL);
5446 			ire = ire_cache_lookup(dst, connp->conn_zoneid,
5447 			    MBLK_GETLABEL(mp), ipst);
5448 		}
5449 
5450 		if (ire == NULL) {
5451 			if (ipif != NULL)
5452 				ipif_refrele(ipif);
5453 			UDP_STAT(us, udp_ire_null);
5454 			ip_output(connp, mp, q, IP_WPUT);
5455 			return;
5456 		}
5457 		IRE_REFHOLD_NOTR(ire);
5458 
5459 		mutex_enter(&connp->conn_lock);
5460 		if (CONN_CACHE_IRE(connp) && connp->conn_ire_cache == NULL &&
5461 		    !(ire->ire_marks & IRE_MARK_CONDEMNED)) {
5462 			irb_t		*irb = ire->ire_bucket;
5463 
5464 			/*
5465 			 * IRE's created for non-connection oriented transports
5466 			 * are normally initialized with IRE_MARK_TEMPORARY set
5467 			 * in the ire_marks. These IRE's are preferentially
5468 			 * reaped when the hash chain length in the cache
5469 			 * bucket exceeds the maximum value specified in
5470 			 * ip[6]_ire_max_bucket_cnt. This can severely affect
5471 			 * UDP performance if IRE cache entries that we need
5472 			 * to reuse are continually removed. To remedy this,
5473 			 * when we cache the IRE in the conn_t, we remove the
5474 			 * IRE_MARK_TEMPORARY bit from the ire_marks if it was
5475 			 * set.
5476 			 */
5477 			if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5478 				rw_enter(&irb->irb_lock, RW_WRITER);
5479 				if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5480 					ire->ire_marks &= ~IRE_MARK_TEMPORARY;
5481 					irb->irb_tmp_ire_cnt--;
5482 				}
5483 				rw_exit(&irb->irb_lock);
5484 			}
5485 			connp->conn_ire_cache = ire;
5486 			cached = B_TRUE;
5487 		}
5488 		mutex_exit(&connp->conn_lock);
5489 
5490 		/*
5491 		 * We can continue to use the ire but since it was not
5492 		 * cached, we should drop the extra reference.
5493 		 */
5494 		if (!cached)
5495 			IRE_REFRELE_NOTR(ire);
5496 	}
5497 	ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION);
5498 	ASSERT(!CLASSD(dst) || ipif != NULL);
5499 
5500 	/*
5501 	 * Check if we can take the fast-path.
5502 	 * Note that "incomplete" ire's (where the link-layer for next hop
5503 	 * is not resolved, or where the fast-path header in nce_fp_mp is not
5504 	 * available yet) are sent down the legacy (slow) path
5505 	 */
5506 	if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) ||
5507 	    (ire->ire_flags & RTF_MULTIRT) || (ire->ire_stq == NULL) ||
5508 	    (ire->ire_max_frag < ntohs(ipha->ipha_length)) ||
5509 	    ((ire->ire_nce == NULL) ||
5510 	    ((ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL)) ||
5511 	    connp->conn_nexthop_set || (MBLKL(ire_fp_mp) > MBLKHEAD(mp))) {
5512 		if (ipif != NULL)
5513 			ipif_refrele(ipif);
5514 		UDP_STAT(us, udp_ip_ire_send);
5515 		IRE_REFRELE(ire);
5516 		ip_output(connp, mp, q, IP_WPUT);
5517 		return;
5518 	}
5519 
5520 	if (src == INADDR_ANY && !connp->conn_unspec_src) {
5521 		if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC))
5522 			ipha->ipha_src = ipif->ipif_src_addr;
5523 		else
5524 			ipha->ipha_src = ire->ire_src_addr;
5525 	}
5526 
5527 	if (ipif != NULL)
5528 		ipif_refrele(ipif);
5529 
5530 	udp_xmit(connp->conn_wq, mp, ire, connp, connp->conn_zoneid);
5531 }
5532 
5533 static void
5534 udp_xmit(queue_t *q, mblk_t *mp, ire_t *ire, conn_t *connp, zoneid_t zoneid)
5535 {
5536 	ipaddr_t src, dst;
5537 	ill_t	*ill;
5538 	mblk_t	*ire_fp_mp;
5539 	uint_t	ire_fp_mp_len;
5540 	uint16_t *up;
5541 	uint32_t cksum, hcksum_txflags;
5542 	queue_t	*dev_q;
5543 	udp_t	*udp = connp->conn_udp;
5544 	ipha_t	*ipha = (ipha_t *)mp->b_rptr;
5545 	udp_stack_t	*us = udp->udp_us;
5546 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5547 	boolean_t	ll_multicast = B_FALSE;
5548 
5549 	dev_q = ire->ire_stq->q_next;
5550 	ASSERT(dev_q != NULL);
5551 
5552 	ill = ire_to_ill(ire);
5553 	ASSERT(ill != NULL);
5554 
5555 	/* is queue flow controlled? */
5556 	if (q->q_first != NULL || connp->conn_draining ||
5557 	    DEV_Q_FLOW_BLOCKED(dev_q)) {
5558 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
5559 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
5560 
5561 		if (ipst->ips_ip_output_queue)
5562 			(void) putq(connp->conn_wq, mp);
5563 		else
5564 			freemsg(mp);
5565 		ire_refrele(ire);
5566 		return;
5567 	}
5568 
5569 	ire_fp_mp = ire->ire_nce->nce_fp_mp;
5570 	ire_fp_mp_len = MBLKL(ire_fp_mp);
5571 	ASSERT(MBLKHEAD(mp) >= ire_fp_mp_len);
5572 
5573 	dst = ipha->ipha_dst;
5574 	src = ipha->ipha_src;
5575 
5576 
5577 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
5578 
5579 	ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1);
5580 #ifndef _BIG_ENDIAN
5581 	ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8);
5582 #endif
5583 
5584 	if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
5585 		ASSERT(ill->ill_hcksum_capab != NULL);
5586 		hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags;
5587 	} else {
5588 		hcksum_txflags = 0;
5589 	}
5590 
5591 	/* pseudo-header checksum (do it in parts for IP header checksum) */
5592 	cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF);
5593 
5594 	ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION);
5595 	up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH);
5596 	if (*up != 0) {
5597 		IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags,
5598 		    mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH,
5599 		    ntohs(ipha->ipha_length), cksum);
5600 
5601 		/* Software checksum? */
5602 		if (DB_CKSUMFLAGS(mp) == 0) {
5603 			UDP_STAT(us, udp_out_sw_cksum);
5604 			UDP_STAT_UPDATE(us, udp_out_sw_cksum_bytes,
5605 			    ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH);
5606 		}
5607 	}
5608 
5609 	if (!CLASSD(dst)) {
5610 		ipha->ipha_fragment_offset_and_flags |=
5611 		    (uint32_t)htons(ire->ire_frag_flag);
5612 	}
5613 
5614 	/* Calculate IP header checksum if hardware isn't capable */
5615 	if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) {
5616 		IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0],
5617 		    ((uint16_t *)ipha)[4]);
5618 	}
5619 
5620 	if (CLASSD(dst)) {
5621 		if (ilm_lookup_ill(ill, dst, ALL_ZONES) != NULL) {
5622 			ip_multicast_loopback(q, ill, mp,
5623 			    connp->conn_multicast_loop ? 0 :
5624 			    IP_FF_NO_MCAST_LOOP, zoneid);
5625 		}
5626 
5627 		/* If multicast TTL is 0 then we are done */
5628 		if (ipha->ipha_ttl == 0) {
5629 			freemsg(mp);
5630 			ire_refrele(ire);
5631 			return;
5632 		}
5633 		ll_multicast = B_TRUE;
5634 	}
5635 
5636 	ASSERT(DB_TYPE(ire_fp_mp) == M_DATA);
5637 	mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len;
5638 	bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len);
5639 
5640 	UPDATE_OB_PKT_COUNT(ire);
5641 	ire->ire_last_used_time = lbolt;
5642 
5643 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
5644 	UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
5645 	    ntohs(ipha->ipha_length));
5646 
5647 	DTRACE_PROBE4(ip4__physical__out__start,
5648 	    ill_t *, NULL, ill_t *, ill, ipha_t *, ipha, mblk_t *, mp);
5649 	FW_HOOKS(ipst->ips_ip4_physical_out_event,
5650 	    ipst->ips_ipv4firewall_physical_out, NULL, ill, ipha, mp, mp,
5651 	    ll_multicast, ipst);
5652 	DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);
5653 	if (ipst->ips_ipobs_enabled && mp != NULL) {
5654 		zoneid_t szone;
5655 
5656 		szone = ip_get_zoneid_v4(ipha->ipha_src, mp,
5657 		    ipst, ALL_ZONES);
5658 		ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone,
5659 		    ALL_ZONES, ill, IPV4_VERSION, ire_fp_mp_len, ipst);
5660 	}
5661 
5662 	if (mp != NULL) {
5663 		DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
5664 		    void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
5665 		    ipha_t *, ipha, ip6_t *, NULL, int, 0);
5666 
5667 		if (ILL_DIRECT_CAPABLE(ill)) {
5668 			ill_dld_direct_t *idd = &ill->ill_dld_capab->idc_direct;
5669 
5670 			(void) idd->idd_tx_df(idd->idd_tx_dh, mp,
5671 			    (uintptr_t)connp, 0);
5672 		} else {
5673 			putnext(ire->ire_stq, mp);
5674 		}
5675 	}
5676 	IRE_REFRELE(ire);
5677 }
5678 
5679 static boolean_t
5680 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst,
5681     boolean_t *update_lastdst)
5682 {
5683 	udp_t *udp = Q_TO_UDP(wq);
5684 	int err;
5685 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
5686 	udp_stack_t		*us = udp->udp_us;
5687 
5688 	err = tsol_compute_label_v6(DB_CREDDEF(mp, udp->udp_connp->conn_cred),
5689 	    dst, opt_storage, udp->udp_connp->conn_mac_exempt,
5690 	    us->us_netstack->netstack_ip);
5691 	if (err == 0) {
5692 		err = tsol_update_sticky(&udp->udp_sticky_ipp,
5693 		    &udp->udp_label_len_v6, opt_storage);
5694 	}
5695 	if (err != 0) {
5696 		DTRACE_PROBE4(
5697 		    tx__ip__log__drop__updatelabel__udp6,
5698 		    char *, "queue(1) failed to update options(2) on mp(3)",
5699 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
5700 	} else {
5701 		*update_lastdst = B_TRUE;
5702 	}
5703 	return (err);
5704 }
5705 
5706 static int
5707 udp_send_connected(conn_t *connp, mblk_t *mp, struct nmsghdr *msg, cred_t *cr,
5708     pid_t pid)
5709 {
5710 	udp_t		*udp = connp->conn_udp;
5711 	udp_stack_t	*us = udp->udp_us;
5712 	ipaddr_t	v4dst;
5713 	in_port_t	dstport;
5714 	boolean_t	mapped_addr;
5715 	struct sockaddr_storage ss;
5716 	sin_t		*sin;
5717 	sin6_t		*sin6;
5718 	struct sockaddr	*addr;
5719 	socklen_t	addrlen;
5720 	int		error;
5721 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5722 
5723 	/* M_DATA for connected socket */
5724 
5725 	ASSERT(udp->udp_issocket || IPCL_IS_NONSTR(connp));
5726 	UDP_DBGSTAT(us, udp_data_conn);
5727 
5728 	mutex_enter(&connp->conn_lock);
5729 	if (udp->udp_state != TS_DATA_XFER) {
5730 		mutex_exit(&connp->conn_lock);
5731 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5732 		UDP_STAT(us, udp_out_err_notconn);
5733 		freemsg(mp);
5734 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5735 		    "udp_wput_end: connp %p (%S)", connp,
5736 		    "not-connected; address required");
5737 		return (EDESTADDRREQ);
5738 	}
5739 
5740 	mapped_addr = IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst);
5741 	if (mapped_addr)
5742 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst);
5743 
5744 	/* Initialize addr and addrlen as if they're passed in */
5745 	if (udp->udp_family == AF_INET) {
5746 		sin = (sin_t *)&ss;
5747 		sin->sin_family = AF_INET;
5748 		dstport = sin->sin_port = udp->udp_dstport;
5749 		ASSERT(mapped_addr);
5750 		sin->sin_addr.s_addr = v4dst;
5751 		addr = (struct sockaddr *)sin;
5752 		addrlen = sizeof (*sin);
5753 	} else {
5754 		sin6 = (sin6_t *)&ss;
5755 		sin6->sin6_family = AF_INET6;
5756 		dstport = sin6->sin6_port = udp->udp_dstport;
5757 		sin6->sin6_flowinfo = udp->udp_flowinfo;
5758 		sin6->sin6_addr = udp->udp_v6dst;
5759 		sin6->sin6_scope_id = 0;
5760 		sin6->__sin6_src_id = 0;
5761 		addr = (struct sockaddr *)sin6;
5762 		addrlen = sizeof (*sin6);
5763 	}
5764 	mutex_exit(&connp->conn_lock);
5765 
5766 	if (mapped_addr) {
5767 		/*
5768 		 * Handle both AF_INET and AF_INET6; the latter
5769 		 * for IPV4 mapped destination addresses.  Note
5770 		 * here that both addr and addrlen point to the
5771 		 * corresponding struct depending on the address
5772 		 * family of the socket.
5773 		 */
5774 		mp = udp_output_v4(connp, mp, v4dst, dstport, 0, &error,
5775 		    insert_spi, msg, cr, pid);
5776 	} else {
5777 		mp = udp_output_v6(connp, mp, sin6, &error, msg, cr, pid);
5778 	}
5779 	if (error == 0) {
5780 		ASSERT(mp == NULL);
5781 		return (0);
5782 	}
5783 
5784 	UDP_STAT(us, udp_out_err_output);
5785 	ASSERT(mp != NULL);
5786 	if (IPCL_IS_NONSTR(connp)) {
5787 		freemsg(mp);
5788 		return (error);
5789 	} else {
5790 		/* mp is freed by the following routine */
5791 		udp_ud_err(connp->conn_wq, mp, (uchar_t *)addr,
5792 		    (t_scalar_t)addrlen, (t_scalar_t)error);
5793 		return (0);
5794 	}
5795 }
5796 
5797 /* ARGSUSED */
5798 static int
5799 udp_send_not_connected(conn_t *connp,  mblk_t *mp, struct sockaddr *addr,
5800     socklen_t addrlen, struct nmsghdr *msg, cred_t *cr, pid_t pid)
5801 {
5802 
5803 	udp_t		*udp = connp->conn_udp;
5804 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5805 	int		error = 0;
5806 	sin6_t		*sin6;
5807 	sin_t		*sin;
5808 	uint_t		srcid;
5809 	uint16_t	port;
5810 	ipaddr_t	v4dst;
5811 
5812 
5813 	ASSERT(addr != NULL);
5814 
5815 	switch (udp->udp_family) {
5816 	case AF_INET6:
5817 		sin6 = (sin6_t *)addr;
5818 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
5819 			/*
5820 			 * Destination is a non-IPv4-compatible IPv6 address.
5821 			 * Send out an IPv6 format packet.
5822 			 */
5823 			mp = udp_output_v6(connp, mp, sin6, &error, msg, cr,
5824 			    pid);
5825 			if (error != 0)
5826 				goto ud_error;
5827 
5828 			return (0);
5829 		}
5830 		/*
5831 		 * If the local address is not zero or a mapped address
5832 		 * return an error.  It would be possible to send an IPv4
5833 		 * packet but the response would never make it back to the
5834 		 * application since it is bound to a non-mapped address.
5835 		 */
5836 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) &&
5837 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
5838 			error = EADDRNOTAVAIL;
5839 			goto ud_error;
5840 		}
5841 		/* Send IPv4 packet without modifying udp_ipversion */
5842 		/* Extract port and ipaddr */
5843 		port = sin6->sin6_port;
5844 		IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst);
5845 		srcid = sin6->__sin6_src_id;
5846 		break;
5847 
5848 	case AF_INET:
5849 		sin = (sin_t *)addr;
5850 		/* Extract port and ipaddr */
5851 		port = sin->sin_port;
5852 		v4dst = sin->sin_addr.s_addr;
5853 		srcid = 0;
5854 		break;
5855 	}
5856 
5857 	mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error, insert_spi,
5858 	    msg, cr, pid);
5859 
5860 	if (error == 0) {
5861 		ASSERT(mp == NULL);
5862 		return (0);
5863 	}
5864 
5865 ud_error:
5866 	ASSERT(mp != NULL);
5867 
5868 	return (error);
5869 }
5870 
5871 /*
5872  * This routine handles all messages passed downstream.  It either
5873  * consumes the message or passes it downstream; it never queues a
5874  * a message.
5875  *
5876  * Also entry point for sockfs when udp is in "direct sockfs" mode.  This mode
5877  * is valid when we are directly beneath the stream head, and thus sockfs
5878  * is able to bypass STREAMS and directly call us, passing along the sockaddr
5879  * structure without the cumbersome T_UNITDATA_REQ interface for the case of
5880  * connected endpoints.
5881  */
5882 void
5883 udp_wput(queue_t *q, mblk_t *mp)
5884 {
5885 	conn_t		*connp = Q_TO_CONN(q);
5886 	udp_t		*udp = connp->conn_udp;
5887 	int		error = 0;
5888 	struct sockaddr	*addr;
5889 	socklen_t	addrlen;
5890 	udp_stack_t	*us = udp->udp_us;
5891 
5892 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START,
5893 	    "udp_wput_start: queue %p mp %p", q, mp);
5894 
5895 	/*
5896 	 * We directly handle several cases here: T_UNITDATA_REQ message
5897 	 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for connected
5898 	 * socket.
5899 	 */
5900 	switch (DB_TYPE(mp)) {
5901 	case M_DATA:
5902 		/*
5903 		 * Quick check for error cases. Checks will be done again
5904 		 * under the lock later on
5905 		 */
5906 		if (!udp->udp_direct_sockfs || udp->udp_state != TS_DATA_XFER) {
5907 			/* Not connected; address is required */
5908 			BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5909 			UDP_STAT(us, udp_out_err_notconn);
5910 			freemsg(mp);
5911 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5912 			    "udp_wput_end: connp %p (%S)", connp,
5913 			    "not-connected; address required");
5914 			return;
5915 		}
5916 		(void) udp_send_connected(connp, mp, NULL, NULL, -1);
5917 		return;
5918 
5919 	case M_PROTO:
5920 	case M_PCPROTO: {
5921 		struct T_unitdata_req *tudr;
5922 
5923 		ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX);
5924 		tudr = (struct T_unitdata_req *)mp->b_rptr;
5925 
5926 		/* Handle valid T_UNITDATA_REQ here */
5927 		if (MBLKL(mp) >= sizeof (*tudr) &&
5928 		    ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) {
5929 			if (mp->b_cont == NULL) {
5930 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5931 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5932 				error = EPROTO;
5933 				goto ud_error;
5934 			}
5935 
5936 			if (!MBLKIN(mp, 0, tudr->DEST_offset +
5937 			    tudr->DEST_length)) {
5938 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5939 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5940 				error = EADDRNOTAVAIL;
5941 				goto ud_error;
5942 			}
5943 			/*
5944 			 * If a port has not been bound to the stream, fail.
5945 			 * This is not a problem when sockfs is directly
5946 			 * above us, because it will ensure that the socket
5947 			 * is first bound before allowing data to be sent.
5948 			 */
5949 			if (udp->udp_state == TS_UNBND) {
5950 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5951 				    "udp_wput_end: q %p (%S)", q, "outstate");
5952 				error = EPROTO;
5953 				goto ud_error;
5954 			}
5955 			addr = (struct sockaddr *)
5956 			    &mp->b_rptr[tudr->DEST_offset];
5957 			addrlen = tudr->DEST_length;
5958 			if (tudr->OPT_length != 0)
5959 				UDP_STAT(us, udp_out_opt);
5960 			break;
5961 		}
5962 		/* FALLTHRU */
5963 	}
5964 	default:
5965 		udp_wput_other(q, mp);
5966 		return;
5967 	}
5968 	ASSERT(addr != NULL);
5969 
5970 	error = udp_send_not_connected(connp,  mp, addr, addrlen, NULL, NULL,
5971 	    -1);
5972 	if (error != 0) {
5973 ud_error:
5974 		UDP_STAT(us, udp_out_err_output);
5975 		ASSERT(mp != NULL);
5976 		/* mp is freed by the following routine */
5977 		udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
5978 		    (t_scalar_t)error);
5979 	}
5980 }
5981 
5982 /* ARGSUSED */
5983 static void
5984 udp_wput_fallback(queue_t *wq, mblk_t *mp)
5985 {
5986 #ifdef DEBUG
5987 	cmn_err(CE_CONT, "udp_wput_fallback: Message in fallback \n");
5988 #endif
5989 	freemsg(mp);
5990 }
5991 
5992 
5993 /*
5994  * udp_output_v6():
5995  * Assumes that udp_wput did some sanity checking on the destination
5996  * address.
5997  */
5998 static mblk_t *
5999 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error,
6000     struct nmsghdr *msg, cred_t *cr, pid_t pid)
6001 {
6002 	ip6_t		*ip6h;
6003 	ip6i_t		*ip6i;	/* mp1->b_rptr even if no ip6i_t */
6004 	mblk_t		*mp1 = mp;
6005 	mblk_t		*mp2;
6006 	int		udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
6007 	size_t		ip_len;
6008 	udpha_t		*udph;
6009 	udp_t		*udp = connp->conn_udp;
6010 	udp_stack_t	*us = udp->udp_us;
6011 	queue_t		*q = connp->conn_wq;
6012 	ip6_pkt_t	ipp_s;	/* For ancillary data options */
6013 	ip6_pkt_t	*ipp = &ipp_s;
6014 	ip6_pkt_t	*tipp;	/* temporary ipp */
6015 	uint32_t	csum = 0;
6016 	uint_t		ignore = 0;
6017 	uint_t		option_exists = 0, is_sticky = 0;
6018 	uint8_t		*cp;
6019 	uint8_t		*nxthdr_ptr;
6020 	in6_addr_t	ip6_dst;
6021 	in_port_t	port;
6022 	udpattrs_t	attrs;
6023 	boolean_t	opt_present;
6024 	ip6_hbh_t	*hopoptsptr = NULL;
6025 	uint_t		hopoptslen = 0;
6026 	boolean_t	is_ancillary = B_FALSE;
6027 	size_t		sth_wroff = 0;
6028 	ire_t		*ire;
6029 	boolean_t	update_lastdst = B_FALSE;
6030 
6031 	*error = 0;
6032 
6033 	/*
6034 	 * If the local address is a mapped address return
6035 	 * an error.
6036 	 * It would be possible to send an IPv6 packet but the
6037 	 * response would never make it back to the application
6038 	 * since it is bound to a mapped address.
6039 	 */
6040 	if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) {
6041 		*error = EADDRNOTAVAIL;
6042 		goto done;
6043 	}
6044 
6045 	ipp->ipp_fields = 0;
6046 	ipp->ipp_sticky_ignored = 0;
6047 
6048 	/*
6049 	 * If TPI options passed in, feed it for verification and handling
6050 	 */
6051 	attrs.udpattr_credset = B_FALSE;
6052 	opt_present = B_FALSE;
6053 	if (IPCL_IS_NONSTR(connp)) {
6054 		if (msg->msg_controllen != 0) {
6055 			attrs.udpattr_ipp6 = ipp;
6056 			attrs.udpattr_mb = mp;
6057 
6058 			rw_enter(&udp->udp_rwlock, RW_WRITER);
6059 			*error = process_auxiliary_options(connp,
6060 			    msg->msg_control, msg->msg_controllen,
6061 			    &attrs, &udp_opt_obj, udp_opt_set);
6062 			rw_exit(&udp->udp_rwlock);
6063 			if (*error)
6064 				goto done;
6065 			ASSERT(*error == 0);
6066 			opt_present = B_TRUE;
6067 		}
6068 	} else {
6069 		if (DB_TYPE(mp) != M_DATA) {
6070 			mp1 = mp->b_cont;
6071 			if (((struct T_unitdata_req *)
6072 			    mp->b_rptr)->OPT_length != 0) {
6073 				attrs.udpattr_ipp6 = ipp;
6074 				attrs.udpattr_mb = mp;
6075 				if (udp_unitdata_opt_process(q, mp, error,
6076 				    &attrs) < 0) {
6077 					goto done;
6078 				}
6079 				ASSERT(*error == 0);
6080 				opt_present = B_TRUE;
6081 			}
6082 		}
6083 	}
6084 
6085 	/*
6086 	 * Determine whether we need to mark the mblk with the user's
6087 	 * credentials.
6088 	 */
6089 	ire = connp->conn_ire_cache;
6090 	if (is_system_labeled() || IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) ||
6091 	    (ire == NULL) ||
6092 	    (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, &sin6->sin6_addr)) ||
6093 	    (ire->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) {
6094 		if (cr != NULL && DB_CRED(mp) == NULL)
6095 			msg_setcredpid(mp, cr, pid);
6096 	}
6097 
6098 	rw_enter(&udp->udp_rwlock, RW_READER);
6099 	ignore = ipp->ipp_sticky_ignored;
6100 
6101 	/* mp1 points to the M_DATA mblk carrying the packet */
6102 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
6103 
6104 	if (sin6->sin6_scope_id != 0 &&
6105 	    IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
6106 		/*
6107 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
6108 		 * option nor ancillary data.  It needs to be
6109 		 * explicitly set in options_exists.
6110 		 */
6111 		option_exists |= IPPF_SCOPE_ID;
6112 	}
6113 
6114 	/*
6115 	 * Compute the destination address
6116 	 */
6117 	ip6_dst = sin6->sin6_addr;
6118 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
6119 		ip6_dst = ipv6_loopback;
6120 
6121 	port = sin6->sin6_port;
6122 
6123 	/*
6124 	 * Cluster and TSOL notes, Cluster check:
6125 	 * see comments in udp_output_v4().
6126 	 */
6127 	mutex_enter(&connp->conn_lock);
6128 
6129 	if (cl_inet_connect2 != NULL &&
6130 	    (!IN6_ARE_ADDR_EQUAL(&ip6_dst, &udp->udp_v6lastdst) ||
6131 	    port != udp->udp_lastdstport)) {
6132 		mutex_exit(&connp->conn_lock);
6133 		*error = 0;
6134 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &ip6_dst, port, *error);
6135 		if (*error != 0) {
6136 			*error = EHOSTUNREACH;
6137 			rw_exit(&udp->udp_rwlock);
6138 			goto done;
6139 		}
6140 		update_lastdst = B_TRUE;
6141 		mutex_enter(&connp->conn_lock);
6142 	}
6143 
6144 	/*
6145 	 * If we're not going to the same destination as last time, then
6146 	 * recompute the label required.  This is done in a separate routine to
6147 	 * avoid blowing up our stack here.
6148 	 *
6149 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
6150 	 * to different destination may require different labels,
6151 	 * or worse, UDP packets to same IP address may require
6152 	 * different labels due to use of shared all-zones address.
6153 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
6154 	 * and sticky ipp_hopoptslen are consistent for the current
6155 	 * destination and are updated atomically.
6156 	 */
6157 	if (is_system_labeled()) {
6158 		/* Using UDP MLP requires SCM_UCRED from user */
6159 		if (connp->conn_mlp_type != mlptSingle &&
6160 		    !attrs.udpattr_credset) {
6161 			DTRACE_PROBE4(
6162 			    tx__ip__log__info__output__udp6,
6163 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
6164 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
6165 			*error = ECONNREFUSED;
6166 			rw_exit(&udp->udp_rwlock);
6167 			mutex_exit(&connp->conn_lock);
6168 			goto done;
6169 		}
6170 		/*
6171 		 * update label option for this UDP socket if
6172 		 * - the destination has changed, or
6173 		 * - the UDP socket is MLP
6174 		 */
6175 		if ((opt_present ||
6176 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst) ||
6177 		    connp->conn_mlp_type != mlptSingle) &&
6178 		    (*error = udp_update_label_v6(q, mp, &ip6_dst,
6179 		    &update_lastdst)) != 0) {
6180 			rw_exit(&udp->udp_rwlock);
6181 			mutex_exit(&connp->conn_lock);
6182 			goto done;
6183 		}
6184 	}
6185 
6186 	if (update_lastdst) {
6187 		udp->udp_v6lastdst = ip6_dst;
6188 		udp->udp_lastdstport = port;
6189 	}
6190 
6191 	/*
6192 	 * If there's a security label here, then we ignore any options the
6193 	 * user may try to set.  We keep the peer's label as a hidden sticky
6194 	 * option. We make a private copy of this label before releasing the
6195 	 * lock so that label is kept consistent with the destination addr.
6196 	 */
6197 	if (udp->udp_label_len_v6 > 0) {
6198 		ignore &= ~IPPF_HOPOPTS;
6199 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
6200 	}
6201 
6202 	if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) {
6203 		/* No sticky options nor ancillary data. */
6204 		mutex_exit(&connp->conn_lock);
6205 		goto no_options;
6206 	}
6207 
6208 	/*
6209 	 * Go through the options figuring out where each is going to
6210 	 * come from and build two masks.  The first mask indicates if
6211 	 * the option exists at all.  The second mask indicates if the
6212 	 * option is sticky or ancillary.
6213 	 */
6214 	if (!(ignore & IPPF_HOPOPTS)) {
6215 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
6216 			option_exists |= IPPF_HOPOPTS;
6217 			udp_ip_hdr_len += ipp->ipp_hopoptslen;
6218 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
6219 			option_exists |= IPPF_HOPOPTS;
6220 			is_sticky |= IPPF_HOPOPTS;
6221 			ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0);
6222 			hopoptsptr = kmem_alloc(
6223 			    udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
6224 			if (hopoptsptr == NULL) {
6225 				*error = ENOMEM;
6226 				mutex_exit(&connp->conn_lock);
6227 				goto done;
6228 			}
6229 			hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen;
6230 			bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr,
6231 			    hopoptslen);
6232 			udp_ip_hdr_len += hopoptslen;
6233 		}
6234 	}
6235 	mutex_exit(&connp->conn_lock);
6236 
6237 	if (!(ignore & IPPF_RTHDR)) {
6238 		if (ipp->ipp_fields & IPPF_RTHDR) {
6239 			option_exists |= IPPF_RTHDR;
6240 			udp_ip_hdr_len += ipp->ipp_rthdrlen;
6241 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
6242 			option_exists |= IPPF_RTHDR;
6243 			is_sticky |= IPPF_RTHDR;
6244 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen;
6245 		}
6246 	}
6247 
6248 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
6249 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
6250 			option_exists |= IPPF_RTDSTOPTS;
6251 			udp_ip_hdr_len += ipp->ipp_rtdstoptslen;
6252 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
6253 			option_exists |= IPPF_RTDSTOPTS;
6254 			is_sticky |= IPPF_RTDSTOPTS;
6255 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen;
6256 		}
6257 	}
6258 
6259 	if (!(ignore & IPPF_DSTOPTS)) {
6260 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
6261 			option_exists |= IPPF_DSTOPTS;
6262 			udp_ip_hdr_len += ipp->ipp_dstoptslen;
6263 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
6264 			option_exists |= IPPF_DSTOPTS;
6265 			is_sticky |= IPPF_DSTOPTS;
6266 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen;
6267 		}
6268 	}
6269 
6270 	if (!(ignore & IPPF_IFINDEX)) {
6271 		if (ipp->ipp_fields & IPPF_IFINDEX) {
6272 			option_exists |= IPPF_IFINDEX;
6273 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
6274 			option_exists |= IPPF_IFINDEX;
6275 			is_sticky |= IPPF_IFINDEX;
6276 		}
6277 	}
6278 
6279 	if (!(ignore & IPPF_ADDR)) {
6280 		if (ipp->ipp_fields & IPPF_ADDR) {
6281 			option_exists |= IPPF_ADDR;
6282 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) {
6283 			option_exists |= IPPF_ADDR;
6284 			is_sticky |= IPPF_ADDR;
6285 		}
6286 	}
6287 
6288 	if (!(ignore & IPPF_DONTFRAG)) {
6289 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
6290 			option_exists |= IPPF_DONTFRAG;
6291 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
6292 			option_exists |= IPPF_DONTFRAG;
6293 			is_sticky |= IPPF_DONTFRAG;
6294 		}
6295 	}
6296 
6297 	if (!(ignore & IPPF_USE_MIN_MTU)) {
6298 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
6299 			option_exists |= IPPF_USE_MIN_MTU;
6300 		} else if (udp->udp_sticky_ipp.ipp_fields &
6301 		    IPPF_USE_MIN_MTU) {
6302 			option_exists |= IPPF_USE_MIN_MTU;
6303 			is_sticky |= IPPF_USE_MIN_MTU;
6304 		}
6305 	}
6306 
6307 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
6308 		option_exists |= IPPF_HOPLIMIT;
6309 	/* IPV6_HOPLIMIT can never be sticky */
6310 	ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
6311 
6312 	if (!(ignore & IPPF_UNICAST_HOPS) &&
6313 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
6314 		option_exists |= IPPF_UNICAST_HOPS;
6315 		is_sticky |= IPPF_UNICAST_HOPS;
6316 	}
6317 
6318 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
6319 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
6320 		option_exists |= IPPF_MULTICAST_HOPS;
6321 		is_sticky |= IPPF_MULTICAST_HOPS;
6322 	}
6323 
6324 	if (!(ignore & IPPF_TCLASS)) {
6325 		if (ipp->ipp_fields & IPPF_TCLASS) {
6326 			option_exists |= IPPF_TCLASS;
6327 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
6328 			option_exists |= IPPF_TCLASS;
6329 			is_sticky |= IPPF_TCLASS;
6330 		}
6331 	}
6332 
6333 	if (!(ignore & IPPF_NEXTHOP) &&
6334 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_NEXTHOP)) {
6335 		option_exists |= IPPF_NEXTHOP;
6336 		is_sticky |= IPPF_NEXTHOP;
6337 	}
6338 
6339 no_options:
6340 
6341 	/*
6342 	 * If any options carried in the ip6i_t were specified, we
6343 	 * need to account for the ip6i_t in the data we'll be sending
6344 	 * down.
6345 	 */
6346 	if (option_exists & IPPF_HAS_IP6I)
6347 		udp_ip_hdr_len += sizeof (ip6i_t);
6348 
6349 	/* check/fix buffer config, setup pointers into it */
6350 	ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len];
6351 	if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) ||
6352 	    !OK_32PTR(ip6h)) {
6353 
6354 		/* Try to get everything in a single mblk next time */
6355 		if (udp_ip_hdr_len > udp->udp_max_hdr_len) {
6356 			udp->udp_max_hdr_len = udp_ip_hdr_len;
6357 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
6358 		}
6359 
6360 		mp2 = allocb(udp_ip_hdr_len + us->us_wroff_extra, BPRI_LO);
6361 		if (mp2 == NULL) {
6362 			*error = ENOMEM;
6363 			rw_exit(&udp->udp_rwlock);
6364 			goto done;
6365 		}
6366 		mp2->b_wptr = DB_LIM(mp2);
6367 		mp2->b_cont = mp1;
6368 		mp1 = mp2;
6369 		if (DB_TYPE(mp) != M_DATA)
6370 			mp->b_cont = mp1;
6371 		else
6372 			mp = mp1;
6373 
6374 		ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len);
6375 	}
6376 	mp1->b_rptr = (unsigned char *)ip6h;
6377 	ip6i = (ip6i_t *)ip6h;
6378 
6379 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp)
6380 	if (option_exists & IPPF_HAS_IP6I) {
6381 		ip6h = (ip6_t *)&ip6i[1];
6382 		ip6i->ip6i_flags = 0;
6383 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6384 
6385 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
6386 		if (option_exists & IPPF_SCOPE_ID) {
6387 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6388 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
6389 		} else if (option_exists & IPPF_IFINDEX) {
6390 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
6391 			ASSERT(tipp->ipp_ifindex != 0);
6392 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6393 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
6394 		}
6395 
6396 		if (option_exists & IPPF_ADDR) {
6397 			/*
6398 			 * Enable per-packet source address verification if
6399 			 * IPV6_PKTINFO specified the source address.
6400 			 * ip6_src is set in the transport's _wput function.
6401 			 */
6402 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
6403 		}
6404 
6405 		if (option_exists & IPPF_DONTFRAG) {
6406 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
6407 		}
6408 
6409 		if (option_exists & IPPF_USE_MIN_MTU) {
6410 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
6411 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
6412 		}
6413 
6414 		if (option_exists & IPPF_NEXTHOP) {
6415 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
6416 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
6417 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
6418 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
6419 		}
6420 
6421 		/*
6422 		 * tell IP this is an ip6i_t private header
6423 		 */
6424 		ip6i->ip6i_nxt = IPPROTO_RAW;
6425 	}
6426 
6427 	/* Initialize IPv6 header */
6428 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6429 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
6430 
6431 	/* Set the hoplimit of the outgoing packet. */
6432 	if (option_exists & IPPF_HOPLIMIT) {
6433 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
6434 		ip6h->ip6_hops = ipp->ipp_hoplimit;
6435 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6436 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
6437 		ip6h->ip6_hops = udp->udp_multicast_ttl;
6438 		if (option_exists & IPPF_MULTICAST_HOPS)
6439 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6440 	} else {
6441 		ip6h->ip6_hops = udp->udp_ttl;
6442 		if (option_exists & IPPF_UNICAST_HOPS)
6443 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6444 	}
6445 
6446 	if (option_exists & IPPF_ADDR) {
6447 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
6448 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
6449 		ip6h->ip6_src = tipp->ipp_addr;
6450 	} else {
6451 		/*
6452 		 * The source address was not set using IPV6_PKTINFO.
6453 		 * First look at the bound source.
6454 		 * If unspecified fallback to __sin6_src_id.
6455 		 */
6456 		ip6h->ip6_src = udp->udp_v6src;
6457 		if (sin6->__sin6_src_id != 0 &&
6458 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
6459 			ip_srcid_find_id(sin6->__sin6_src_id,
6460 			    &ip6h->ip6_src, connp->conn_zoneid,
6461 			    us->us_netstack);
6462 		}
6463 	}
6464 
6465 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
6466 	cp = (uint8_t *)&ip6h[1];
6467 
6468 	/*
6469 	 * Here's where we have to start stringing together
6470 	 * any extension headers in the right order:
6471 	 * Hop-by-hop, destination, routing, and final destination opts.
6472 	 */
6473 	if (option_exists & IPPF_HOPOPTS) {
6474 		/* Hop-by-hop options */
6475 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
6476 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
6477 		if (hopoptslen == 0) {
6478 			hopoptsptr = tipp->ipp_hopopts;
6479 			hopoptslen = tipp->ipp_hopoptslen;
6480 			is_ancillary = B_TRUE;
6481 		}
6482 
6483 		*nxthdr_ptr = IPPROTO_HOPOPTS;
6484 		nxthdr_ptr = &hbh->ip6h_nxt;
6485 
6486 		bcopy(hopoptsptr, cp, hopoptslen);
6487 		cp += hopoptslen;
6488 
6489 		if (hopoptsptr != NULL && !is_ancillary) {
6490 			kmem_free(hopoptsptr, hopoptslen);
6491 			hopoptsptr = NULL;
6492 			hopoptslen = 0;
6493 		}
6494 	}
6495 	/*
6496 	 * En-route destination options
6497 	 * Only do them if there's a routing header as well
6498 	 */
6499 	if (option_exists & IPPF_RTDSTOPTS) {
6500 		ip6_dest_t *dst = (ip6_dest_t *)cp;
6501 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
6502 
6503 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6504 		nxthdr_ptr = &dst->ip6d_nxt;
6505 
6506 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
6507 		cp += tipp->ipp_rtdstoptslen;
6508 	}
6509 	/*
6510 	 * Routing header next
6511 	 */
6512 	if (option_exists & IPPF_RTHDR) {
6513 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
6514 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
6515 
6516 		*nxthdr_ptr = IPPROTO_ROUTING;
6517 		nxthdr_ptr = &rt->ip6r_nxt;
6518 
6519 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
6520 		cp += tipp->ipp_rthdrlen;
6521 	}
6522 	/*
6523 	 * Do ultimate destination options
6524 	 */
6525 	if (option_exists & IPPF_DSTOPTS) {
6526 		ip6_dest_t *dest = (ip6_dest_t *)cp;
6527 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
6528 
6529 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6530 		nxthdr_ptr = &dest->ip6d_nxt;
6531 
6532 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
6533 		cp += tipp->ipp_dstoptslen;
6534 	}
6535 	/*
6536 	 * Now set the last header pointer to the proto passed in
6537 	 */
6538 	ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE));
6539 	*nxthdr_ptr = IPPROTO_UDP;
6540 
6541 	/* Update UDP header */
6542 	udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE);
6543 	udph->uha_dst_port = sin6->sin6_port;
6544 	udph->uha_src_port = udp->udp_port;
6545 
6546 	/*
6547 	 * Copy in the destination address
6548 	 */
6549 	ip6h->ip6_dst = ip6_dst;
6550 
6551 	ip6h->ip6_vcf =
6552 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
6553 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
6554 
6555 	if (option_exists & IPPF_TCLASS) {
6556 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
6557 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
6558 		    tipp->ipp_tclass);
6559 	}
6560 	rw_exit(&udp->udp_rwlock);
6561 
6562 	if (option_exists & IPPF_RTHDR) {
6563 		ip6_rthdr_t	*rth;
6564 
6565 		/*
6566 		 * Perform any processing needed for source routing.
6567 		 * We know that all extension headers will be in the same mblk
6568 		 * as the IPv6 header.
6569 		 */
6570 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
6571 		if (rth != NULL && rth->ip6r_segleft != 0) {
6572 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
6573 				/*
6574 				 * Drop packet - only support Type 0 routing.
6575 				 * Notify the application as well.
6576 				 */
6577 				*error = EPROTO;
6578 				goto done;
6579 			}
6580 
6581 			/*
6582 			 * rth->ip6r_len is twice the number of
6583 			 * addresses in the header. Thus it must be even.
6584 			 */
6585 			if (rth->ip6r_len & 0x1) {
6586 				*error = EPROTO;
6587 				goto done;
6588 			}
6589 			/*
6590 			 * Shuffle the routing header and ip6_dst
6591 			 * addresses, and get the checksum difference
6592 			 * between the first hop (in ip6_dst) and
6593 			 * the destination (in the last routing hdr entry).
6594 			 */
6595 			csum = ip_massage_options_v6(ip6h, rth,
6596 			    us->us_netstack);
6597 			/*
6598 			 * Verify that the first hop isn't a mapped address.
6599 			 * Routers along the path need to do this verification
6600 			 * for subsequent hops.
6601 			 */
6602 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
6603 				*error = EADDRNOTAVAIL;
6604 				goto done;
6605 			}
6606 
6607 			cp += (rth->ip6r_len + 1)*8;
6608 		}
6609 	}
6610 
6611 	/* count up length of UDP packet */
6612 	ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN;
6613 	if ((mp2 = mp1->b_cont) != NULL) {
6614 		do {
6615 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
6616 			ip_len += (uint32_t)MBLKL(mp2);
6617 		} while ((mp2 = mp2->b_cont) != NULL);
6618 	}
6619 
6620 	/*
6621 	 * If the size of the packet is greater than the maximum allowed by
6622 	 * ip, return an error. Passing this down could cause panics because
6623 	 * the size will have wrapped and be inconsistent with the msg size.
6624 	 */
6625 	if (ip_len > IP_MAXPACKET) {
6626 		*error = EMSGSIZE;
6627 		goto done;
6628 	}
6629 
6630 	/* Store the UDP length. Subtract length of extension hdrs */
6631 	udph->uha_length = htons(ip_len + IPV6_HDR_LEN -
6632 	    (int)((uchar_t *)udph - (uchar_t *)ip6h));
6633 
6634 	/*
6635 	 * We make it easy for IP to include our pseudo header
6636 	 * by putting our length in uh_checksum, modified (if
6637 	 * we have a routing header) by the checksum difference
6638 	 * between the ultimate destination and first hop addresses.
6639 	 * Note: UDP over IPv6 must always checksum the packet.
6640 	 */
6641 	csum += udph->uha_length;
6642 	csum = (csum & 0xFFFF) + (csum >> 16);
6643 	udph->uha_checksum = (uint16_t)csum;
6644 
6645 #ifdef _LITTLE_ENDIAN
6646 	ip_len = htons(ip_len);
6647 #endif
6648 	ip6h->ip6_plen = ip_len;
6649 	if (DB_CRED(mp) != NULL)
6650 		mblk_setcred(mp1, DB_CRED(mp));
6651 
6652 	if (DB_TYPE(mp) != M_DATA) {
6653 		ASSERT(mp != mp1);
6654 		freeb(mp);
6655 	}
6656 
6657 	/* mp has been consumed and we'll return success */
6658 	ASSERT(*error == 0);
6659 	mp = NULL;
6660 
6661 	/* We're done. Pass the packet to IP */
6662 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
6663 	ip_output_v6(connp, mp1, q, IP_WPUT);
6664 
6665 done:
6666 	if (sth_wroff != 0) {
6667 		(void) proto_set_tx_wroff(RD(q), connp,
6668 		    udp->udp_max_hdr_len + us->us_wroff_extra);
6669 	}
6670 	if (hopoptsptr != NULL && !is_ancillary) {
6671 		kmem_free(hopoptsptr, hopoptslen);
6672 		hopoptsptr = NULL;
6673 	}
6674 	if (*error != 0) {
6675 		ASSERT(mp != NULL);
6676 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6677 	}
6678 	return (mp);
6679 }
6680 
6681 
6682 static int
6683 i_udp_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6684 {
6685 	sin_t *sin = (sin_t *)sa;
6686 	sin6_t *sin6 = (sin6_t *)sa;
6687 
6688 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6689 
6690 	if (udp->udp_state != TS_DATA_XFER)
6691 		return (ENOTCONN);
6692 
6693 	switch (udp->udp_family) {
6694 	case AF_INET:
6695 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6696 
6697 		if (*salenp < sizeof (sin_t))
6698 			return (EINVAL);
6699 
6700 		*salenp = sizeof (sin_t);
6701 		*sin = sin_null;
6702 		sin->sin_family = AF_INET;
6703 		sin->sin_port = udp->udp_dstport;
6704 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
6705 		break;
6706 
6707 	case AF_INET6:
6708 		if (*salenp < sizeof (sin6_t))
6709 			return (EINVAL);
6710 
6711 		*salenp = sizeof (sin6_t);
6712 		*sin6 = sin6_null;
6713 		sin6->sin6_family = AF_INET6;
6714 		sin6->sin6_port = udp->udp_dstport;
6715 		sin6->sin6_addr = udp->udp_v6dst;
6716 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6717 		break;
6718 	}
6719 
6720 	return (0);
6721 }
6722 
6723 static int
6724 udp_getmyname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6725 {
6726 	sin_t *sin = (sin_t *)sa;
6727 	sin6_t *sin6 = (sin6_t *)sa;
6728 
6729 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6730 
6731 	switch (udp->udp_family) {
6732 	case AF_INET:
6733 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6734 
6735 		if (*salenp < sizeof (sin_t))
6736 			return (EINVAL);
6737 
6738 		*salenp = sizeof (sin_t);
6739 		*sin = sin_null;
6740 		sin->sin_family = AF_INET;
6741 		sin->sin_port = udp->udp_port;
6742 
6743 		/*
6744 		 * If udp_v6src is unspecified, we might be bound to broadcast
6745 		 * / multicast.  Use udp_bound_v6src as local address instead
6746 		 * (that could also still be unspecified).
6747 		 */
6748 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
6749 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
6750 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
6751 		} else {
6752 			sin->sin_addr.s_addr =
6753 			    V4_PART_OF_V6(udp->udp_bound_v6src);
6754 		}
6755 		break;
6756 
6757 	case AF_INET6:
6758 		if (*salenp < sizeof (sin6_t))
6759 			return (EINVAL);
6760 
6761 		*salenp = sizeof (sin6_t);
6762 		*sin6 = sin6_null;
6763 		sin6->sin6_family = AF_INET6;
6764 		sin6->sin6_port = udp->udp_port;
6765 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6766 
6767 		/*
6768 		 * If udp_v6src is unspecified, we might be bound to broadcast
6769 		 * / multicast.  Use udp_bound_v6src as local address instead
6770 		 * (that could also still be unspecified).
6771 		 */
6772 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))
6773 			sin6->sin6_addr = udp->udp_v6src;
6774 		else
6775 			sin6->sin6_addr = udp->udp_bound_v6src;
6776 		break;
6777 	}
6778 
6779 	return (0);
6780 }
6781 
6782 /*
6783  * Handle special out-of-band ioctl requests (see PSARC/2008/265).
6784  */
6785 static void
6786 udp_wput_cmdblk(queue_t *q, mblk_t *mp)
6787 {
6788 	void	*data;
6789 	mblk_t	*datamp = mp->b_cont;
6790 	udp_t	*udp = Q_TO_UDP(q);
6791 	cmdblk_t *cmdp = (cmdblk_t *)mp->b_rptr;
6792 
6793 	if (datamp == NULL || MBLKL(datamp) < cmdp->cb_len) {
6794 		cmdp->cb_error = EPROTO;
6795 		qreply(q, mp);
6796 		return;
6797 	}
6798 	data = datamp->b_rptr;
6799 
6800 	rw_enter(&udp->udp_rwlock, RW_READER);
6801 	switch (cmdp->cb_cmd) {
6802 	case TI_GETPEERNAME:
6803 		cmdp->cb_error = i_udp_getpeername(udp, data, &cmdp->cb_len);
6804 		break;
6805 	case TI_GETMYNAME:
6806 		cmdp->cb_error = udp_getmyname(udp, data, &cmdp->cb_len);
6807 		break;
6808 	default:
6809 		cmdp->cb_error = EINVAL;
6810 		break;
6811 	}
6812 	rw_exit(&udp->udp_rwlock);
6813 
6814 	qreply(q, mp);
6815 }
6816 
6817 static void
6818 udp_disable_direct_sockfs(udp_t *udp)
6819 {
6820 	udp->udp_issocket = B_FALSE;
6821 	if (udp->udp_direct_sockfs) {
6822 		/*
6823 		 * Disable read-side synchronous stream interface and
6824 		 * drain any queued data.
6825 		 */
6826 		udp_rcv_drain(udp->udp_connp->conn_rq, udp, B_FALSE);
6827 		ASSERT(!udp->udp_direct_sockfs);
6828 		UDP_STAT(udp->udp_us, udp_sock_fallback);
6829 	}
6830 }
6831 
6832 static void
6833 udp_wput_other(queue_t *q, mblk_t *mp)
6834 {
6835 	uchar_t	*rptr = mp->b_rptr;
6836 	struct datab *db;
6837 	struct iocblk *iocp;
6838 	cred_t	*cr;
6839 	conn_t	*connp = Q_TO_CONN(q);
6840 	udp_t	*udp = connp->conn_udp;
6841 	udp_stack_t *us;
6842 
6843 	TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START,
6844 	    "udp_wput_other_start: q %p", q);
6845 
6846 	us = udp->udp_us;
6847 	db = mp->b_datap;
6848 
6849 	cr = DB_CREDDEF(mp, connp->conn_cred);
6850 
6851 	switch (db->db_type) {
6852 	case M_CMD:
6853 		udp_wput_cmdblk(q, mp);
6854 		return;
6855 
6856 	case M_PROTO:
6857 	case M_PCPROTO:
6858 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
6859 			freemsg(mp);
6860 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6861 			    "udp_wput_other_end: q %p (%S)", q, "protoshort");
6862 			return;
6863 		}
6864 		switch (((t_primp_t)rptr)->type) {
6865 		case T_ADDR_REQ:
6866 			udp_addr_req(q, mp);
6867 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6868 			    "udp_wput_other_end: q %p (%S)", q, "addrreq");
6869 			return;
6870 		case O_T_BIND_REQ:
6871 		case T_BIND_REQ:
6872 			udp_tpi_bind(q, mp);
6873 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6874 			    "udp_wput_other_end: q %p (%S)", q, "bindreq");
6875 			return;
6876 		case T_CONN_REQ:
6877 			udp_tpi_connect(q, mp);
6878 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6879 			    "udp_wput_other_end: q %p (%S)", q, "connreq");
6880 			return;
6881 		case T_CAPABILITY_REQ:
6882 			udp_capability_req(q, mp);
6883 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6884 			    "udp_wput_other_end: q %p (%S)", q, "capabreq");
6885 			return;
6886 		case T_INFO_REQ:
6887 			udp_info_req(q, mp);
6888 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6889 			    "udp_wput_other_end: q %p (%S)", q, "inforeq");
6890 			return;
6891 		case T_UNITDATA_REQ:
6892 			/*
6893 			 * If a T_UNITDATA_REQ gets here, the address must
6894 			 * be bad.  Valid T_UNITDATA_REQs are handled
6895 			 * in udp_wput.
6896 			 */
6897 			udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL);
6898 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6899 			    "udp_wput_other_end: q %p (%S)", q, "unitdatareq");
6900 			return;
6901 		case T_UNBIND_REQ:
6902 			udp_tpi_unbind(q, mp);
6903 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6904 			    "udp_wput_other_end: q %p (%S)", q, "unbindreq");
6905 			return;
6906 		case T_SVR4_OPTMGMT_REQ:
6907 			if (!snmpcom_req(q, mp, udp_snmp_set, ip_snmp_get,
6908 			    cr)) {
6909 				(void) svr4_optcom_req(q,
6910 				    mp, cr, &udp_opt_obj, B_TRUE);
6911 			}
6912 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6913 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6914 			return;
6915 
6916 		case T_OPTMGMT_REQ:
6917 			(void) tpi_optcom_req(q, mp, cr, &udp_opt_obj, B_TRUE);
6918 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6919 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6920 			return;
6921 
6922 		case T_DISCON_REQ:
6923 			udp_tpi_disconnect(q, mp);
6924 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6925 			    "udp_wput_other_end: q %p (%S)", q, "disconreq");
6926 			return;
6927 
6928 		/* The following TPI message is not supported by udp. */
6929 		case O_T_CONN_RES:
6930 		case T_CONN_RES:
6931 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6932 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6933 			    "udp_wput_other_end: q %p (%S)", q,
6934 			    "connres/disconreq");
6935 			return;
6936 
6937 		/* The following 3 TPI messages are illegal for udp. */
6938 		case T_DATA_REQ:
6939 		case T_EXDATA_REQ:
6940 		case T_ORDREL_REQ:
6941 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6942 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6943 			    "udp_wput_other_end: q %p (%S)", q,
6944 			    "data/exdata/ordrel");
6945 			return;
6946 		default:
6947 			break;
6948 		}
6949 		break;
6950 	case M_FLUSH:
6951 		if (*rptr & FLUSHW)
6952 			flushq(q, FLUSHDATA);
6953 		break;
6954 	case M_IOCTL:
6955 		iocp = (struct iocblk *)mp->b_rptr;
6956 		switch (iocp->ioc_cmd) {
6957 		case TI_GETPEERNAME:
6958 			if (udp->udp_state != TS_DATA_XFER) {
6959 				/*
6960 				 * If a default destination address has not
6961 				 * been associated with the stream, then we
6962 				 * don't know the peer's name.
6963 				 */
6964 				iocp->ioc_error = ENOTCONN;
6965 				iocp->ioc_count = 0;
6966 				mp->b_datap->db_type = M_IOCACK;
6967 				qreply(q, mp);
6968 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6969 				    "udp_wput_other_end: q %p (%S)", q,
6970 				    "getpeername");
6971 				return;
6972 			}
6973 			/* FALLTHRU */
6974 		case TI_GETMYNAME: {
6975 			/*
6976 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
6977 			 * need to copyin the user's strbuf structure.
6978 			 * Processing will continue in the M_IOCDATA case
6979 			 * below.
6980 			 */
6981 			mi_copyin(q, mp, NULL,
6982 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
6983 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6984 			    "udp_wput_other_end: q %p (%S)", q, "getmyname");
6985 			return;
6986 			}
6987 		case ND_SET:
6988 			/* nd_getset performs the necessary checking */
6989 		case ND_GET:
6990 			if (nd_getset(q, us->us_nd, mp)) {
6991 				qreply(q, mp);
6992 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6993 				    "udp_wput_other_end: q %p (%S)", q, "get");
6994 				return;
6995 			}
6996 			break;
6997 		case _SIOCSOCKFALLBACK:
6998 			/*
6999 			 * Either sockmod is about to be popped and the
7000 			 * socket would now be treated as a plain stream,
7001 			 * or a module is about to be pushed so we could
7002 			 * no longer use read-side synchronous stream.
7003 			 * Drain any queued data and disable direct sockfs
7004 			 * interface from now on.
7005 			 */
7006 			if (!udp->udp_issocket) {
7007 				DB_TYPE(mp) = M_IOCNAK;
7008 				iocp->ioc_error = EINVAL;
7009 			} else {
7010 				udp_disable_direct_sockfs(udp);
7011 
7012 				DB_TYPE(mp) = M_IOCACK;
7013 				iocp->ioc_error = 0;
7014 			}
7015 			iocp->ioc_count = 0;
7016 			iocp->ioc_rval = 0;
7017 			qreply(q, mp);
7018 			return;
7019 		default:
7020 			break;
7021 		}
7022 		break;
7023 	case M_IOCDATA:
7024 		udp_wput_iocdata(q, mp);
7025 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7026 		    "udp_wput_other_end: q %p (%S)", q, "iocdata");
7027 		return;
7028 	default:
7029 		/* Unrecognized messages are passed through without change. */
7030 		break;
7031 	}
7032 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7033 	    "udp_wput_other_end: q %p (%S)", q, "end");
7034 	ip_output(connp, mp, q, IP_WPUT);
7035 }
7036 
7037 /*
7038  * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA
7039  * messages.
7040  */
7041 static void
7042 udp_wput_iocdata(queue_t *q, mblk_t *mp)
7043 {
7044 	mblk_t		*mp1;
7045 	struct	iocblk *iocp = (struct iocblk *)mp->b_rptr;
7046 	STRUCT_HANDLE(strbuf, sb);
7047 	udp_t		*udp = Q_TO_UDP(q);
7048 	int		error;
7049 	uint_t		addrlen;
7050 
7051 	/* Make sure it is one of ours. */
7052 	switch (iocp->ioc_cmd) {
7053 	case TI_GETMYNAME:
7054 	case TI_GETPEERNAME:
7055 		break;
7056 	default:
7057 		ip_output(udp->udp_connp, mp, q, IP_WPUT);
7058 		return;
7059 	}
7060 
7061 	switch (mi_copy_state(q, mp, &mp1)) {
7062 	case -1:
7063 		return;
7064 	case MI_COPY_CASE(MI_COPY_IN, 1):
7065 		break;
7066 	case MI_COPY_CASE(MI_COPY_OUT, 1):
7067 		/*
7068 		 * The address has been copied out, so now
7069 		 * copyout the strbuf.
7070 		 */
7071 		mi_copyout(q, mp);
7072 		return;
7073 	case MI_COPY_CASE(MI_COPY_OUT, 2):
7074 		/*
7075 		 * The address and strbuf have been copied out.
7076 		 * We're done, so just acknowledge the original
7077 		 * M_IOCTL.
7078 		 */
7079 		mi_copy_done(q, mp, 0);
7080 		return;
7081 	default:
7082 		/*
7083 		 * Something strange has happened, so acknowledge
7084 		 * the original M_IOCTL with an EPROTO error.
7085 		 */
7086 		mi_copy_done(q, mp, EPROTO);
7087 		return;
7088 	}
7089 
7090 	/*
7091 	 * Now we have the strbuf structure for TI_GETMYNAME
7092 	 * and TI_GETPEERNAME.  Next we copyout the requested
7093 	 * address and then we'll copyout the strbuf.
7094 	 */
7095 	STRUCT_SET_HANDLE(sb, iocp->ioc_flag, (void *)mp1->b_rptr);
7096 	addrlen = udp->udp_family == AF_INET ? sizeof (sin_t) : sizeof (sin6_t);
7097 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
7098 		mi_copy_done(q, mp, EINVAL);
7099 		return;
7100 	}
7101 
7102 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
7103 
7104 	if (mp1 == NULL)
7105 		return;
7106 
7107 	rw_enter(&udp->udp_rwlock, RW_READER);
7108 	switch (iocp->ioc_cmd) {
7109 	case TI_GETMYNAME:
7110 		error = udp_do_getsockname(udp, (void *)mp1->b_rptr, &addrlen);
7111 		break;
7112 	case TI_GETPEERNAME:
7113 		error = udp_do_getpeername(udp, (void *)mp1->b_rptr, &addrlen);
7114 		break;
7115 	}
7116 	rw_exit(&udp->udp_rwlock);
7117 
7118 	if (error != 0) {
7119 		mi_copy_done(q, mp, error);
7120 	} else {
7121 		mp1->b_wptr += addrlen;
7122 		STRUCT_FSET(sb, len, addrlen);
7123 
7124 		/* Copy out the address */
7125 		mi_copyout(q, mp);
7126 	}
7127 }
7128 
7129 static int
7130 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
7131     udpattrs_t *udpattrs)
7132 {
7133 	struct T_unitdata_req *udreqp;
7134 	int is_absreq_failure;
7135 	cred_t *cr;
7136 	conn_t	*connp = Q_TO_CONN(q);
7137 
7138 	ASSERT(((t_primp_t)mp->b_rptr)->type);
7139 
7140 	cr = DB_CREDDEF(mp, connp->conn_cred);
7141 
7142 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
7143 
7144 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
7145 	    udreqp->OPT_offset, cr, &udp_opt_obj,
7146 	    udpattrs, &is_absreq_failure);
7147 
7148 	if (*errorp != 0) {
7149 		/*
7150 		 * Note: No special action needed in this
7151 		 * module for "is_absreq_failure"
7152 		 */
7153 		return (-1);		/* failure */
7154 	}
7155 	ASSERT(is_absreq_failure == 0);
7156 	return (0);	/* success */
7157 }
7158 
7159 void
7160 udp_ddi_g_init(void)
7161 {
7162 	udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr,
7163 	    udp_opt_obj.odb_opt_arr_cnt);
7164 
7165 	/*
7166 	 * We want to be informed each time a stack is created or
7167 	 * destroyed in the kernel, so we can maintain the
7168 	 * set of udp_stack_t's.
7169 	 */
7170 	netstack_register(NS_UDP, udp_stack_init, NULL, udp_stack_fini);
7171 }
7172 
7173 void
7174 udp_ddi_g_destroy(void)
7175 {
7176 	netstack_unregister(NS_UDP);
7177 }
7178 
7179 #define	INET_NAME	"ip"
7180 
7181 /*
7182  * Initialize the UDP stack instance.
7183  */
7184 static void *
7185 udp_stack_init(netstackid_t stackid, netstack_t *ns)
7186 {
7187 	udp_stack_t	*us;
7188 	udpparam_t	*pa;
7189 	int		i;
7190 	int		error = 0;
7191 	major_t		major;
7192 
7193 	us = (udp_stack_t *)kmem_zalloc(sizeof (*us), KM_SLEEP);
7194 	us->us_netstack = ns;
7195 
7196 	us->us_num_epriv_ports = UDP_NUM_EPRIV_PORTS;
7197 	us->us_epriv_ports[0] = 2049;
7198 	us->us_epriv_ports[1] = 4045;
7199 
7200 	/*
7201 	 * The smallest anonymous port in the priviledged port range which UDP
7202 	 * looks for free port.  Use in the option UDP_ANONPRIVBIND.
7203 	 */
7204 	us->us_min_anonpriv_port = 512;
7205 
7206 	us->us_bind_fanout_size = udp_bind_fanout_size;
7207 
7208 	/* Roundup variable that might have been modified in /etc/system */
7209 	if (us->us_bind_fanout_size & (us->us_bind_fanout_size - 1)) {
7210 		/* Not a power of two. Round up to nearest power of two */
7211 		for (i = 0; i < 31; i++) {
7212 			if (us->us_bind_fanout_size < (1 << i))
7213 				break;
7214 		}
7215 		us->us_bind_fanout_size = 1 << i;
7216 	}
7217 	us->us_bind_fanout = kmem_zalloc(us->us_bind_fanout_size *
7218 	    sizeof (udp_fanout_t), KM_SLEEP);
7219 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7220 		mutex_init(&us->us_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT,
7221 		    NULL);
7222 	}
7223 
7224 	pa = (udpparam_t *)kmem_alloc(sizeof (udp_param_arr), KM_SLEEP);
7225 
7226 	us->us_param_arr = pa;
7227 	bcopy(udp_param_arr, us->us_param_arr, sizeof (udp_param_arr));
7228 
7229 	(void) udp_param_register(&us->us_nd,
7230 	    us->us_param_arr, A_CNT(udp_param_arr));
7231 
7232 	us->us_kstat = udp_kstat2_init(stackid, &us->us_statistics);
7233 	us->us_mibkp = udp_kstat_init(stackid);
7234 
7235 	major = mod_name_to_major(INET_NAME);
7236 	error = ldi_ident_from_major(major, &us->us_ldi_ident);
7237 	ASSERT(error == 0);
7238 	return (us);
7239 }
7240 
7241 /*
7242  * Free the UDP stack instance.
7243  */
7244 static void
7245 udp_stack_fini(netstackid_t stackid, void *arg)
7246 {
7247 	udp_stack_t *us = (udp_stack_t *)arg;
7248 	int i;
7249 
7250 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7251 		mutex_destroy(&us->us_bind_fanout[i].uf_lock);
7252 	}
7253 
7254 	kmem_free(us->us_bind_fanout, us->us_bind_fanout_size *
7255 	    sizeof (udp_fanout_t));
7256 
7257 	us->us_bind_fanout = NULL;
7258 
7259 	nd_free(&us->us_nd);
7260 	kmem_free(us->us_param_arr, sizeof (udp_param_arr));
7261 	us->us_param_arr = NULL;
7262 
7263 	udp_kstat_fini(stackid, us->us_mibkp);
7264 	us->us_mibkp = NULL;
7265 
7266 	udp_kstat2_fini(stackid, us->us_kstat);
7267 	us->us_kstat = NULL;
7268 	bzero(&us->us_statistics, sizeof (us->us_statistics));
7269 
7270 	ldi_ident_release(us->us_ldi_ident);
7271 	kmem_free(us, sizeof (*us));
7272 }
7273 
7274 static void *
7275 udp_kstat2_init(netstackid_t stackid, udp_stat_t *us_statisticsp)
7276 {
7277 	kstat_t *ksp;
7278 
7279 	udp_stat_t template = {
7280 		{ "udp_ip_send",		KSTAT_DATA_UINT64 },
7281 		{ "udp_ip_ire_send",		KSTAT_DATA_UINT64 },
7282 		{ "udp_ire_null",		KSTAT_DATA_UINT64 },
7283 		{ "udp_drain",			KSTAT_DATA_UINT64 },
7284 		{ "udp_sock_fallback",		KSTAT_DATA_UINT64 },
7285 		{ "udp_rrw_busy",		KSTAT_DATA_UINT64 },
7286 		{ "udp_rrw_msgcnt",		KSTAT_DATA_UINT64 },
7287 		{ "udp_out_sw_cksum",		KSTAT_DATA_UINT64 },
7288 		{ "udp_out_sw_cksum_bytes",	KSTAT_DATA_UINT64 },
7289 		{ "udp_out_opt",		KSTAT_DATA_UINT64 },
7290 		{ "udp_out_err_notconn",	KSTAT_DATA_UINT64 },
7291 		{ "udp_out_err_output",		KSTAT_DATA_UINT64 },
7292 		{ "udp_out_err_tudr",		KSTAT_DATA_UINT64 },
7293 		{ "udp_in_pktinfo",		KSTAT_DATA_UINT64 },
7294 		{ "udp_in_recvdstaddr",		KSTAT_DATA_UINT64 },
7295 		{ "udp_in_recvopts",		KSTAT_DATA_UINT64 },
7296 		{ "udp_in_recvif",		KSTAT_DATA_UINT64 },
7297 		{ "udp_in_recvslla",		KSTAT_DATA_UINT64 },
7298 		{ "udp_in_recvucred",		KSTAT_DATA_UINT64 },
7299 		{ "udp_in_recvttl",		KSTAT_DATA_UINT64 },
7300 		{ "udp_in_recvhopopts",		KSTAT_DATA_UINT64 },
7301 		{ "udp_in_recvhoplimit",	KSTAT_DATA_UINT64 },
7302 		{ "udp_in_recvdstopts",		KSTAT_DATA_UINT64 },
7303 		{ "udp_in_recvrtdstopts",	KSTAT_DATA_UINT64 },
7304 		{ "udp_in_recvrthdr",		KSTAT_DATA_UINT64 },
7305 		{ "udp_in_recvpktinfo",		KSTAT_DATA_UINT64 },
7306 		{ "udp_in_recvtclass",		KSTAT_DATA_UINT64 },
7307 		{ "udp_in_timestamp",		KSTAT_DATA_UINT64 },
7308 #ifdef DEBUG
7309 		{ "udp_data_conn",		KSTAT_DATA_UINT64 },
7310 		{ "udp_data_notconn",		KSTAT_DATA_UINT64 },
7311 #endif
7312 	};
7313 
7314 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, "udpstat", "net",
7315 	    KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
7316 	    KSTAT_FLAG_VIRTUAL, stackid);
7317 
7318 	if (ksp == NULL)
7319 		return (NULL);
7320 
7321 	bcopy(&template, us_statisticsp, sizeof (template));
7322 	ksp->ks_data = (void *)us_statisticsp;
7323 	ksp->ks_private = (void *)(uintptr_t)stackid;
7324 
7325 	kstat_install(ksp);
7326 	return (ksp);
7327 }
7328 
7329 static void
7330 udp_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
7331 {
7332 	if (ksp != NULL) {
7333 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7334 		kstat_delete_netstack(ksp, stackid);
7335 	}
7336 }
7337 
7338 static void *
7339 udp_kstat_init(netstackid_t stackid)
7340 {
7341 	kstat_t	*ksp;
7342 
7343 	udp_named_kstat_t template = {
7344 		{ "inDatagrams",	KSTAT_DATA_UINT64, 0 },
7345 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
7346 		{ "outDatagrams",	KSTAT_DATA_UINT64, 0 },
7347 		{ "entrySize",		KSTAT_DATA_INT32, 0 },
7348 		{ "entry6Size",		KSTAT_DATA_INT32, 0 },
7349 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
7350 	};
7351 
7352 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, UDP_MOD_NAME, "mib2",
7353 	    KSTAT_TYPE_NAMED,
7354 	    NUM_OF_FIELDS(udp_named_kstat_t), 0, stackid);
7355 
7356 	if (ksp == NULL || ksp->ks_data == NULL)
7357 		return (NULL);
7358 
7359 	template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t);
7360 	template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t);
7361 
7362 	bcopy(&template, ksp->ks_data, sizeof (template));
7363 	ksp->ks_update = udp_kstat_update;
7364 	ksp->ks_private = (void *)(uintptr_t)stackid;
7365 
7366 	kstat_install(ksp);
7367 	return (ksp);
7368 }
7369 
7370 static void
7371 udp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
7372 {
7373 	if (ksp != NULL) {
7374 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7375 		kstat_delete_netstack(ksp, stackid);
7376 	}
7377 }
7378 
7379 static int
7380 udp_kstat_update(kstat_t *kp, int rw)
7381 {
7382 	udp_named_kstat_t *udpkp;
7383 	netstackid_t	stackid = (netstackid_t)(uintptr_t)kp->ks_private;
7384 	netstack_t	*ns;
7385 	udp_stack_t	*us;
7386 
7387 	if ((kp == NULL) || (kp->ks_data == NULL))
7388 		return (EIO);
7389 
7390 	if (rw == KSTAT_WRITE)
7391 		return (EACCES);
7392 
7393 	ns = netstack_find_by_stackid(stackid);
7394 	if (ns == NULL)
7395 		return (-1);
7396 	us = ns->netstack_udp;
7397 	if (us == NULL) {
7398 		netstack_rele(ns);
7399 		return (-1);
7400 	}
7401 	udpkp = (udp_named_kstat_t *)kp->ks_data;
7402 
7403 	udpkp->inDatagrams.value.ui64 =	us->us_udp_mib.udpHCInDatagrams;
7404 	udpkp->inErrors.value.ui32 =	us->us_udp_mib.udpInErrors;
7405 	udpkp->outDatagrams.value.ui64 = us->us_udp_mib.udpHCOutDatagrams;
7406 	udpkp->outErrors.value.ui32 =	us->us_udp_mib.udpOutErrors;
7407 	netstack_rele(ns);
7408 	return (0);
7409 }
7410 
7411 /*
7412  * Read-side synchronous stream info entry point, called as a
7413  * result of handling certain STREAMS ioctl operations.
7414  */
7415 static int
7416 udp_rinfop(queue_t *q, infod_t *dp)
7417 {
7418 	mblk_t	*mp;
7419 	uint_t	cmd = dp->d_cmd;
7420 	int	res = 0;
7421 	int	error = 0;
7422 	udp_t	*udp = Q_TO_UDP(q);
7423 	struct stdata *stp = STREAM(q);
7424 
7425 	mutex_enter(&udp->udp_drain_lock);
7426 	/* If shutdown on read has happened, return nothing */
7427 	mutex_enter(&stp->sd_lock);
7428 	if (stp->sd_flag & STREOF) {
7429 		mutex_exit(&stp->sd_lock);
7430 		goto done;
7431 	}
7432 	mutex_exit(&stp->sd_lock);
7433 
7434 	if ((mp = udp->udp_rcv_list_head) == NULL)
7435 		goto done;
7436 
7437 	ASSERT(DB_TYPE(mp) != M_DATA && mp->b_cont != NULL);
7438 
7439 	if (cmd & INFOD_COUNT) {
7440 		/*
7441 		 * Return the number of messages.
7442 		 */
7443 		dp->d_count += udp->udp_rcv_msgcnt;
7444 		res |= INFOD_COUNT;
7445 	}
7446 	if (cmd & INFOD_BYTES) {
7447 		/*
7448 		 * Return size of all data messages.
7449 		 */
7450 		dp->d_bytes += udp->udp_rcv_cnt;
7451 		res |= INFOD_BYTES;
7452 	}
7453 	if (cmd & INFOD_FIRSTBYTES) {
7454 		/*
7455 		 * Return size of first data message.
7456 		 */
7457 		dp->d_bytes = msgdsize(mp);
7458 		res |= INFOD_FIRSTBYTES;
7459 		dp->d_cmd &= ~INFOD_FIRSTBYTES;
7460 	}
7461 	if (cmd & INFOD_COPYOUT) {
7462 		mblk_t *mp1 = mp->b_cont;
7463 		int n;
7464 		/*
7465 		 * Return data contents of first message.
7466 		 */
7467 		ASSERT(DB_TYPE(mp1) == M_DATA);
7468 		while (mp1 != NULL && dp->d_uiop->uio_resid > 0) {
7469 			n = MIN(dp->d_uiop->uio_resid, MBLKL(mp1));
7470 			if (n != 0 && (error = uiomove((char *)mp1->b_rptr, n,
7471 			    UIO_READ, dp->d_uiop)) != 0) {
7472 				goto done;
7473 			}
7474 			mp1 = mp1->b_cont;
7475 		}
7476 		res |= INFOD_COPYOUT;
7477 		dp->d_cmd &= ~INFOD_COPYOUT;
7478 	}
7479 done:
7480 	mutex_exit(&udp->udp_drain_lock);
7481 
7482 	dp->d_res |= res;
7483 
7484 	return (error);
7485 }
7486 
7487 /*
7488  * Read-side synchronous stream entry point.  This is called as a result
7489  * of recv/read operation done at sockfs, and is guaranteed to execute
7490  * outside of the interrupt thread context.  It returns a single datagram
7491  * (b_cont chain of T_UNITDATA_IND plus data) to the upper layer.
7492  */
7493 static int
7494 udp_rrw(queue_t *q, struiod_t *dp)
7495 {
7496 	mblk_t	*mp;
7497 	udp_t	*udp = Q_TO_UDP(q);
7498 	udp_stack_t *us = udp->udp_us;
7499 
7500 	/*
7501 	 * Dequeue datagram from the head of the list and return
7502 	 * it to caller; also ensure that RSLEEP sd_wakeq flag is
7503 	 * set/cleared depending on whether or not there's data
7504 	 * remaining in the list.
7505 	 */
7506 	mutex_enter(&udp->udp_drain_lock);
7507 	if (!udp->udp_direct_sockfs) {
7508 		mutex_exit(&udp->udp_drain_lock);
7509 		UDP_STAT(us, udp_rrw_busy);
7510 		return (EBUSY);
7511 	}
7512 	if ((mp = udp->udp_rcv_list_head) != NULL) {
7513 		uint_t size = msgdsize(mp);
7514 
7515 		/* Last datagram in the list? */
7516 		if ((udp->udp_rcv_list_head = mp->b_next) == NULL)
7517 			udp->udp_rcv_list_tail = NULL;
7518 		mp->b_next = NULL;
7519 
7520 		udp->udp_rcv_cnt -= size;
7521 		udp->udp_rcv_msgcnt--;
7522 		UDP_STAT(us, udp_rrw_msgcnt);
7523 
7524 		/* No longer flow-controlling? */
7525 		if (udp->udp_rcv_cnt < udp->udp_rcv_hiwat &&
7526 		    udp->udp_rcv_msgcnt < udp->udp_rcv_hiwat)
7527 			udp->udp_drain_qfull = B_FALSE;
7528 	}
7529 	if (udp->udp_rcv_list_head == NULL) {
7530 		/*
7531 		 * Either we just dequeued the last datagram or
7532 		 * we get here from sockfs and have nothing to
7533 		 * return; in this case clear RSLEEP.
7534 		 */
7535 		ASSERT(udp->udp_rcv_cnt == 0);
7536 		ASSERT(udp->udp_rcv_msgcnt == 0);
7537 		ASSERT(udp->udp_rcv_list_tail == NULL);
7538 		STR_WAKEUP_CLEAR(STREAM(q));
7539 	} else {
7540 		/*
7541 		 * More data follows; we need udp_rrw() to be
7542 		 * called in future to pick up the rest.
7543 		 */
7544 		STR_WAKEUP_SET(STREAM(q));
7545 	}
7546 	mutex_exit(&udp->udp_drain_lock);
7547 	dp->d_mp = mp;
7548 	return (0);
7549 }
7550 
7551 /*
7552  * Enqueue a completely-built T_UNITDATA_IND message into the receive
7553  * list; this is typically executed within the interrupt thread context
7554  * and so we do things as quickly as possible.
7555  */
7556 static void
7557 udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, uint_t pkt_len)
7558 {
7559 	ASSERT(q == RD(q));
7560 	ASSERT(pkt_len == msgdsize(mp));
7561 	ASSERT(mp->b_next == NULL && mp->b_cont != NULL);
7562 	ASSERT(DB_TYPE(mp) == M_PROTO && DB_TYPE(mp->b_cont) == M_DATA);
7563 	ASSERT(MBLKL(mp) >= sizeof (struct T_unitdata_ind));
7564 
7565 	mutex_enter(&udp->udp_drain_lock);
7566 	/*
7567 	 * Wake up and signal the receiving app; it is okay to do this
7568 	 * before enqueueing the mp because we are holding the drain lock.
7569 	 * One of the advantages of synchronous stream is the ability for
7570 	 * us to find out when the application performs a read on the
7571 	 * socket by way of udp_rrw() entry point being called.  We need
7572 	 * to generate SIGPOLL/SIGIO for each received data in the case
7573 	 * of asynchronous socket just as in the strrput() case.  However,
7574 	 * we only wake the application up when necessary, i.e. during the
7575 	 * first enqueue.  When udp_rrw() is called, we send up a single
7576 	 * datagram upstream and call STR_WAKEUP_SET() again when there
7577 	 * are still data remaining in our receive queue.
7578 	 */
7579 	STR_WAKEUP_SENDSIG(STREAM(q), udp->udp_rcv_list_head);
7580 	if (udp->udp_rcv_list_head == NULL)
7581 		udp->udp_rcv_list_head = mp;
7582 	else
7583 		udp->udp_rcv_list_tail->b_next = mp;
7584 	udp->udp_rcv_list_tail = mp;
7585 	udp->udp_rcv_cnt += pkt_len;
7586 	udp->udp_rcv_msgcnt++;
7587 
7588 	/* Need to flow-control? */
7589 	if (udp->udp_rcv_cnt >= udp->udp_rcv_hiwat ||
7590 	    udp->udp_rcv_msgcnt >= udp->udp_rcv_hiwat)
7591 		udp->udp_drain_qfull = B_TRUE;
7592 
7593 	mutex_exit(&udp->udp_drain_lock);
7594 }
7595 
7596 /*
7597  * Drain the contents of receive list to the module upstream; we do
7598  * this during close or when we fallback to the slow mode due to
7599  * sockmod being popped or a module being pushed on top of us.
7600  */
7601 static void
7602 udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing)
7603 {
7604 	mblk_t *mp;
7605 	udp_stack_t *us = udp->udp_us;
7606 
7607 	mutex_enter(&udp->udp_drain_lock);
7608 	/*
7609 	 * There is no race with a concurrent udp_input() sending
7610 	 * up packets using putnext() after we have cleared the
7611 	 * udp_direct_sockfs flag but before we have completed
7612 	 * sending up the packets in udp_rcv_list, since we are
7613 	 * either a writer or we have quiesced the conn.
7614 	 */
7615 	udp->udp_direct_sockfs = B_FALSE;
7616 	mutex_exit(&udp->udp_drain_lock);
7617 
7618 	if (udp->udp_rcv_list_head != NULL)
7619 		UDP_STAT(us, udp_drain);
7620 
7621 	/*
7622 	 * Send up everything via putnext(); note here that we
7623 	 * don't need the udp_drain_lock to protect us since
7624 	 * nothing can enter udp_rrw() and that we currently
7625 	 * have exclusive access to this udp.
7626 	 */
7627 	while ((mp = udp->udp_rcv_list_head) != NULL) {
7628 		udp->udp_rcv_list_head = mp->b_next;
7629 		mp->b_next = NULL;
7630 		udp->udp_rcv_cnt -= msgdsize(mp);
7631 		udp->udp_rcv_msgcnt--;
7632 		if (closing) {
7633 			freemsg(mp);
7634 		} else {
7635 			ASSERT(q == RD(q));
7636 			putnext(q, mp);
7637 		}
7638 	}
7639 	ASSERT(udp->udp_rcv_cnt == 0);
7640 	ASSERT(udp->udp_rcv_msgcnt == 0);
7641 	ASSERT(udp->udp_rcv_list_head == NULL);
7642 	udp->udp_rcv_list_tail = NULL;
7643 	udp->udp_drain_qfull = B_FALSE;
7644 }
7645 
7646 static size_t
7647 udp_set_rcv_hiwat(udp_t *udp, size_t size)
7648 {
7649 	udp_stack_t *us = udp->udp_us;
7650 
7651 	/* We add a bit of extra buffering */
7652 	size += size >> 1;
7653 	if (size > us->us_max_buf)
7654 		size = us->us_max_buf;
7655 
7656 	udp->udp_rcv_hiwat = size;
7657 	return (size);
7658 }
7659 
7660 /*
7661  * For the lower queue so that UDP can be a dummy mux.
7662  * Nobody should be sending
7663  * packets up this stream
7664  */
7665 static void
7666 udp_lrput(queue_t *q, mblk_t *mp)
7667 {
7668 	mblk_t *mp1;
7669 
7670 	switch (mp->b_datap->db_type) {
7671 	case M_FLUSH:
7672 		/* Turn around */
7673 		if (*mp->b_rptr & FLUSHW) {
7674 			*mp->b_rptr &= ~FLUSHR;
7675 			qreply(q, mp);
7676 			return;
7677 		}
7678 		break;
7679 	}
7680 	/* Could receive messages that passed through ar_rput */
7681 	for (mp1 = mp; mp1; mp1 = mp1->b_cont)
7682 		mp1->b_prev = mp1->b_next = NULL;
7683 	freemsg(mp);
7684 }
7685 
7686 /*
7687  * For the lower queue so that UDP can be a dummy mux.
7688  * Nobody should be sending packets down this stream.
7689  */
7690 /* ARGSUSED */
7691 void
7692 udp_lwput(queue_t *q, mblk_t *mp)
7693 {
7694 	freemsg(mp);
7695 }
7696 
7697 /*
7698  * Below routines for UDP socket module.
7699  */
7700 
7701 static conn_t *
7702 udp_do_open(cred_t *credp, boolean_t isv6, int flags)
7703 {
7704 	udp_t		*udp;
7705 	conn_t		*connp;
7706 	zoneid_t 	zoneid;
7707 	netstack_t 	*ns;
7708 	udp_stack_t 	*us;
7709 
7710 	ns = netstack_find_by_cred(credp);
7711 	ASSERT(ns != NULL);
7712 	us = ns->netstack_udp;
7713 	ASSERT(us != NULL);
7714 
7715 	/*
7716 	 * For exclusive stacks we set the zoneid to zero
7717 	 * to make UDP operate as if in the global zone.
7718 	 */
7719 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
7720 		zoneid = GLOBAL_ZONEID;
7721 	else
7722 		zoneid = crgetzoneid(credp);
7723 
7724 	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);
7725 
7726 	connp = ipcl_conn_create(IPCL_UDPCONN, flags, ns);
7727 	if (connp == NULL) {
7728 		netstack_rele(ns);
7729 		return (NULL);
7730 	}
7731 	udp = connp->conn_udp;
7732 
7733 	/*
7734 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
7735 	 * done by netstack_find_by_cred()
7736 	 */
7737 	netstack_rele(ns);
7738 
7739 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7740 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
7741 	ASSERT(connp->conn_udp == udp);
7742 	ASSERT(udp->udp_connp == connp);
7743 
7744 	/* Set the initial state of the stream and the privilege status. */
7745 	udp->udp_state = TS_UNBND;
7746 	if (isv6) {
7747 		udp->udp_family = AF_INET6;
7748 		udp->udp_ipversion = IPV6_VERSION;
7749 		udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
7750 		udp->udp_ttl = us->us_ipv6_hoplimit;
7751 		connp->conn_af_isv6 = B_TRUE;
7752 		connp->conn_flags |= IPCL_ISV6;
7753 	} else {
7754 		udp->udp_family = AF_INET;
7755 		udp->udp_ipversion = IPV4_VERSION;
7756 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE;
7757 		udp->udp_ttl = us->us_ipv4_ttl;
7758 		connp->conn_af_isv6 = B_FALSE;
7759 		connp->conn_flags &= ~IPCL_ISV6;
7760 	}
7761 
7762 	udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
7763 	udp->udp_pending_op = -1;
7764 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
7765 	connp->conn_zoneid = zoneid;
7766 
7767 	udp->udp_open_time = lbolt64;
7768 	udp->udp_open_pid = curproc->p_pid;
7769 
7770 	/*
7771 	 * If the caller has the process-wide flag set, then default to MAC
7772 	 * exempt mode.  This allows read-down to unlabeled hosts.
7773 	 */
7774 	if (getpflags(NET_MAC_AWARE, credp) != 0)
7775 		connp->conn_mac_exempt = B_TRUE;
7776 
7777 	connp->conn_ulp_labeled = is_system_labeled();
7778 
7779 	udp->udp_us = us;
7780 
7781 	connp->conn_recv = udp_input;
7782 	crhold(credp);
7783 	connp->conn_cred = credp;
7784 
7785 	*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
7786 
7787 	rw_exit(&udp->udp_rwlock);
7788 
7789 	return (connp);
7790 }
7791 
7792 /* ARGSUSED */
7793 sock_lower_handle_t
7794 udp_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
7795     uint_t *smodep, int *errorp, int flags, cred_t *credp)
7796 {
7797 	udp_t		*udp = NULL;
7798 	udp_stack_t	*us;
7799 	conn_t		*connp;
7800 	boolean_t	isv6;
7801 
7802 	if (type != SOCK_DGRAM || (family != AF_INET && family != AF_INET6) ||
7803 	    (proto != 0 && proto != IPPROTO_UDP)) {
7804 		*errorp = EPROTONOSUPPORT;
7805 		return (NULL);
7806 	}
7807 
7808 	if (family == AF_INET6)
7809 		isv6 = B_TRUE;
7810 	else
7811 		isv6 = B_FALSE;
7812 
7813 	connp = udp_do_open(credp, isv6, flags);
7814 	if (connp == NULL) {
7815 		*errorp = ENOMEM;
7816 		return (NULL);
7817 	}
7818 
7819 	udp = connp->conn_udp;
7820 	ASSERT(udp != NULL);
7821 	us = udp->udp_us;
7822 	ASSERT(us != NULL);
7823 
7824 	connp->conn_flags |= IPCL_NONSTR | IPCL_SOCKET;
7825 
7826 	/* Set flow control */
7827 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7828 	(void) udp_set_rcv_hiwat(udp, us->us_recv_hiwat);
7829 	udp->udp_rcv_disply_hiwat = us->us_recv_hiwat;
7830 	udp->udp_rcv_lowat = udp_mod_info.mi_lowat;
7831 	udp->udp_xmit_hiwat = us->us_xmit_hiwat;
7832 	udp->udp_xmit_lowat = us->us_xmit_lowat;
7833 
7834 	if (udp->udp_family == AF_INET6) {
7835 		/* Build initial header template for transmit */
7836 		if ((*errorp = udp_build_hdrs(udp)) != 0) {
7837 			rw_exit(&udp->udp_rwlock);
7838 			ipcl_conn_destroy(connp);
7839 			return (NULL);
7840 		}
7841 	}
7842 	rw_exit(&udp->udp_rwlock);
7843 
7844 	connp->conn_flow_cntrld = B_FALSE;
7845 
7846 	ASSERT(us->us_ldi_ident != NULL);
7847 
7848 	if ((*errorp = ip_create_helper_stream(connp, us->us_ldi_ident)) != 0) {
7849 		ip1dbg(("udp_create: create of IP helper stream failed\n"));
7850 		udp_do_close(connp);
7851 		return (NULL);
7852 	}
7853 
7854 	/* Set the send flow control */
7855 	connp->conn_wq->q_hiwat = us->us_xmit_hiwat;
7856 	connp->conn_wq->q_lowat = us->us_xmit_lowat;
7857 
7858 	mutex_enter(&connp->conn_lock);
7859 	connp->conn_state_flags &= ~CONN_INCIPIENT;
7860 	mutex_exit(&connp->conn_lock);
7861 
7862 	*errorp = 0;
7863 	*smodep = SM_ATOMIC;
7864 	*sock_downcalls = &sock_udp_downcalls;
7865 	return ((sock_lower_handle_t)connp);
7866 }
7867 
7868 /* ARGSUSED */
7869 void
7870 udp_activate(sock_lower_handle_t proto_handle, sock_upper_handle_t sock_handle,
7871     sock_upcalls_t *sock_upcalls, int flags, cred_t *cr)
7872 {
7873 	conn_t 		*connp = (conn_t *)proto_handle;
7874 	udp_t 		*udp = connp->conn_udp;
7875 	udp_stack_t	*us = udp->udp_us;
7876 	struct sock_proto_props sopp;
7877 
7878 	connp->conn_upcalls = sock_upcalls;
7879 	connp->conn_upper_handle = sock_handle;
7880 
7881 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT |
7882 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
7883 	sopp.sopp_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
7884 	sopp.sopp_maxblk = INFPSZ;
7885 	sopp.sopp_rxhiwat = udp->udp_rcv_hiwat;
7886 	sopp.sopp_maxaddrlen = sizeof (sin6_t);
7887 	sopp.sopp_maxpsz =
7888 	    (udp->udp_family == AF_INET) ? UDP_MAXPACKET_IPV4 :
7889 	    UDP_MAXPACKET_IPV6;
7890 	sopp.sopp_minpsz = (udp_mod_info.mi_minpsz == 1) ? 0 :
7891 	    udp_mod_info.mi_minpsz;
7892 
7893 	(*connp->conn_upcalls->su_set_proto_props)(connp->conn_upper_handle,
7894 	    &sopp);
7895 }
7896 
7897 static void
7898 udp_do_close(conn_t *connp)
7899 {
7900 	udp_t	*udp;
7901 
7902 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
7903 	udp = connp->conn_udp;
7904 
7905 	udp_quiesce_conn(connp);
7906 	ip_quiesce_conn(connp);
7907 
7908 	if (!IPCL_IS_NONSTR(connp)) {
7909 		/*
7910 		 * Disable read-side synchronous stream
7911 		 * interface and drain any queued data.
7912 		 */
7913 		ASSERT(connp->conn_wq != NULL);
7914 		udp_rcv_drain(connp->conn_wq, udp, B_TRUE);
7915 		ASSERT(!udp->udp_direct_sockfs);
7916 
7917 		ASSERT(connp->conn_rq != NULL);
7918 		qprocsoff(connp->conn_rq);
7919 	}
7920 
7921 	ASSERT(udp->udp_rcv_cnt == 0);
7922 	ASSERT(udp->udp_rcv_msgcnt == 0);
7923 	ASSERT(udp->udp_rcv_list_head == NULL);
7924 	ASSERT(udp->udp_rcv_list_tail == NULL);
7925 
7926 	udp_close_free(connp);
7927 
7928 	/*
7929 	 * Now we are truly single threaded on this stream, and can
7930 	 * delete the things hanging off the connp, and finally the connp.
7931 	 * We removed this connp from the fanout list, it cannot be
7932 	 * accessed thru the fanouts, and we already waited for the
7933 	 * conn_ref to drop to 0. We are already in close, so
7934 	 * there cannot be any other thread from the top. qprocsoff
7935 	 * has completed, and service has completed or won't run in
7936 	 * future.
7937 	 */
7938 	ASSERT(connp->conn_ref == 1);
7939 	if (!IPCL_IS_NONSTR(connp)) {
7940 		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
7941 	} else {
7942 		ip_free_helper_stream(connp);
7943 	}
7944 
7945 	connp->conn_ref--;
7946 	ipcl_conn_destroy(connp);
7947 }
7948 
7949 /* ARGSUSED */
7950 int
7951 udp_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
7952 {
7953 	conn_t	*connp = (conn_t *)proto_handle;
7954 
7955 	udp_do_close(connp);
7956 	return (0);
7957 }
7958 
7959 static int
7960 udp_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len, cred_t *cr,
7961     boolean_t bind_to_req_port_only)
7962 {
7963 	sin_t		*sin;
7964 	sin6_t		*sin6;
7965 	sin6_t		sin6addr;
7966 	in_port_t	port;		/* Host byte order */
7967 	in_port_t	requested_port;	/* Host byte order */
7968 	int		count;
7969 	in6_addr_t	v6src;
7970 	int		loopmax;
7971 	udp_fanout_t	*udpf;
7972 	in_port_t	lport;		/* Network byte order */
7973 	zoneid_t	zoneid;
7974 	udp_t		*udp;
7975 	boolean_t	is_inaddr_any;
7976 	mlp_type_t	addrtype, mlptype;
7977 	udp_stack_t	*us;
7978 	int		error = 0;
7979 	mblk_t		*mp = NULL;
7980 
7981 	udp = connp->conn_udp;
7982 	us = udp->udp_us;
7983 
7984 	if (udp->udp_state != TS_UNBND) {
7985 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7986 		    "udp_bind: bad state, %u", udp->udp_state);
7987 		return (-TOUTSTATE);
7988 	}
7989 
7990 	switch (len) {
7991 	case 0:
7992 		if (udp->udp_family == AF_INET) {
7993 			sin = (sin_t *)&sin6addr;
7994 			*sin = sin_null;
7995 			sin->sin_family = AF_INET;
7996 			sin->sin_addr.s_addr = INADDR_ANY;
7997 			udp->udp_ipversion = IPV4_VERSION;
7998 		} else {
7999 			ASSERT(udp->udp_family == AF_INET6);
8000 			sin6 = (sin6_t *)&sin6addr;
8001 			*sin6 = sin6_null;
8002 			sin6->sin6_family = AF_INET6;
8003 			V6_SET_ZERO(sin6->sin6_addr);
8004 			udp->udp_ipversion = IPV6_VERSION;
8005 		}
8006 		port = 0;
8007 		break;
8008 
8009 	case sizeof (sin_t):	/* Complete IPv4 address */
8010 		sin = (sin_t *)sa;
8011 
8012 		if (sin == NULL || !OK_32PTR((char *)sin))
8013 			return (EINVAL);
8014 
8015 		if (udp->udp_family != AF_INET ||
8016 		    sin->sin_family != AF_INET) {
8017 			return (EAFNOSUPPORT);
8018 		}
8019 		port = ntohs(sin->sin_port);
8020 		break;
8021 
8022 	case sizeof (sin6_t):	/* complete IPv6 address */
8023 		sin6 = (sin6_t *)sa;
8024 
8025 		if (sin6 == NULL || !OK_32PTR((char *)sin6))
8026 			return (EINVAL);
8027 
8028 		if (udp->udp_family != AF_INET6 ||
8029 		    sin6->sin6_family != AF_INET6) {
8030 			return (EAFNOSUPPORT);
8031 		}
8032 		port = ntohs(sin6->sin6_port);
8033 		break;
8034 
8035 	default:		/* Invalid request */
8036 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
8037 		    "udp_bind: bad ADDR_length length %u", len);
8038 		return (-TBADADDR);
8039 	}
8040 
8041 	requested_port = port;
8042 
8043 	if (requested_port == 0 || !bind_to_req_port_only)
8044 		bind_to_req_port_only = B_FALSE;
8045 	else		/* T_BIND_REQ and requested_port != 0 */
8046 		bind_to_req_port_only = B_TRUE;
8047 
8048 	if (requested_port == 0) {
8049 		/*
8050 		 * If the application passed in zero for the port number, it
8051 		 * doesn't care which port number we bind to. Get one in the
8052 		 * valid range.
8053 		 */
8054 		if (udp->udp_anon_priv_bind) {
8055 			port = udp_get_next_priv_port(udp);
8056 		} else {
8057 			port = udp_update_next_port(udp,
8058 			    us->us_next_port_to_try, B_TRUE);
8059 		}
8060 	} else {
8061 		/*
8062 		 * If the port is in the well-known privileged range,
8063 		 * make sure the caller was privileged.
8064 		 */
8065 		int i;
8066 		boolean_t priv = B_FALSE;
8067 
8068 		if (port < us->us_smallest_nonpriv_port) {
8069 			priv = B_TRUE;
8070 		} else {
8071 			for (i = 0; i < us->us_num_epriv_ports; i++) {
8072 				if (port == us->us_epriv_ports[i]) {
8073 					priv = B_TRUE;
8074 					break;
8075 				}
8076 			}
8077 		}
8078 
8079 		if (priv) {
8080 			if (secpolicy_net_privaddr(cr, port, IPPROTO_UDP) != 0)
8081 				return (-TACCES);
8082 		}
8083 	}
8084 
8085 	if (port == 0)
8086 		return (-TNOADDR);
8087 
8088 	/*
8089 	 * The state must be TS_UNBND. TPI mandates that users must send
8090 	 * TPI primitives only 1 at a time and wait for the response before
8091 	 * sending the next primitive.
8092 	 */
8093 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8094 	if (udp->udp_state != TS_UNBND || udp->udp_pending_op != -1) {
8095 		rw_exit(&udp->udp_rwlock);
8096 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
8097 		    "udp_bind: bad state, %u", udp->udp_state);
8098 		return (-TOUTSTATE);
8099 	}
8100 	/* XXX how to remove the T_BIND_REQ? Should set it before calling */
8101 	udp->udp_pending_op = T_BIND_REQ;
8102 	/*
8103 	 * Copy the source address into our udp structure. This address
8104 	 * may still be zero; if so, IP will fill in the correct address
8105 	 * each time an outbound packet is passed to it. Since the udp is
8106 	 * not yet in the bind hash list, we don't grab the uf_lock to
8107 	 * change udp_ipversion
8108 	 */
8109 	if (udp->udp_family == AF_INET) {
8110 		ASSERT(sin != NULL);
8111 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8112 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
8113 		    udp->udp_ip_snd_options_len;
8114 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src);
8115 	} else {
8116 		ASSERT(sin6 != NULL);
8117 		v6src = sin6->sin6_addr;
8118 		if (IN6_IS_ADDR_V4MAPPED(&v6src)) {
8119 			/*
8120 			 * no need to hold the uf_lock to set the udp_ipversion
8121 			 * since we are not yet in the fanout list
8122 			 */
8123 			udp->udp_ipversion = IPV4_VERSION;
8124 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
8125 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
8126 		} else {
8127 			udp->udp_ipversion = IPV6_VERSION;
8128 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
8129 		}
8130 	}
8131 
8132 	/*
8133 	 * If udp_reuseaddr is not set, then we have to make sure that
8134 	 * the IP address and port number the application requested
8135 	 * (or we selected for the application) is not being used by
8136 	 * another stream.  If another stream is already using the
8137 	 * requested IP address and port, the behavior depends on
8138 	 * "bind_to_req_port_only". If set the bind fails; otherwise we
8139 	 * search for any an unused port to bind to the the stream.
8140 	 *
8141 	 * As per the BSD semantics, as modified by the Deering multicast
8142 	 * changes, if udp_reuseaddr is set, then we allow multiple binds
8143 	 * to the same port independent of the local IP address.
8144 	 *
8145 	 * This is slightly different than in SunOS 4.X which did not
8146 	 * support IP multicast. Note that the change implemented by the
8147 	 * Deering multicast code effects all binds - not only binding
8148 	 * to IP multicast addresses.
8149 	 *
8150 	 * Note that when binding to port zero we ignore SO_REUSEADDR in
8151 	 * order to guarantee a unique port.
8152 	 */
8153 
8154 	count = 0;
8155 	if (udp->udp_anon_priv_bind) {
8156 		/*
8157 		 * loopmax = (IPPORT_RESERVED-1) -
8158 		 *    us->us_min_anonpriv_port + 1
8159 		 */
8160 		loopmax = IPPORT_RESERVED - us->us_min_anonpriv_port;
8161 	} else {
8162 		loopmax = us->us_largest_anon_port -
8163 		    us->us_smallest_anon_port + 1;
8164 	}
8165 
8166 	is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src);
8167 	zoneid = connp->conn_zoneid;
8168 
8169 	for (;;) {
8170 		udp_t		*udp1;
8171 		boolean_t	found_exclbind = B_FALSE;
8172 
8173 		/*
8174 		 * Walk through the list of udp streams bound to
8175 		 * requested port with the same IP address.
8176 		 */
8177 		lport = htons(port);
8178 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(lport,
8179 		    us->us_bind_fanout_size)];
8180 		mutex_enter(&udpf->uf_lock);
8181 		for (udp1 = udpf->uf_udp; udp1 != NULL;
8182 		    udp1 = udp1->udp_bind_hash) {
8183 			if (lport != udp1->udp_port)
8184 				continue;
8185 
8186 			/*
8187 			 * On a labeled system, we must treat bindings to ports
8188 			 * on shared IP addresses by sockets with MAC exemption
8189 			 * privilege as being in all zones, as there's
8190 			 * otherwise no way to identify the right receiver.
8191 			 */
8192 			if (!(IPCL_ZONE_MATCH(udp1->udp_connp, zoneid) ||
8193 			    IPCL_ZONE_MATCH(connp,
8194 			    udp1->udp_connp->conn_zoneid)) &&
8195 			    !connp->conn_mac_exempt && \
8196 			    !udp1->udp_connp->conn_mac_exempt)
8197 				continue;
8198 
8199 			/*
8200 			 * If UDP_EXCLBIND is set for either the bound or
8201 			 * binding endpoint, the semantics of bind
8202 			 * is changed according to the following chart.
8203 			 *
8204 			 * spec = specified address (v4 or v6)
8205 			 * unspec = unspecified address (v4 or v6)
8206 			 * A = specified addresses are different for endpoints
8207 			 *
8208 			 * bound	bind to		allowed?
8209 			 * -------------------------------------
8210 			 * unspec	unspec		no
8211 			 * unspec	spec		no
8212 			 * spec		unspec		no
8213 			 * spec		spec		yes if A
8214 			 *
8215 			 * For labeled systems, SO_MAC_EXEMPT behaves the same
8216 			 * as UDP_EXCLBIND, except that zoneid is ignored.
8217 			 */
8218 			if (udp1->udp_exclbind || udp->udp_exclbind ||
8219 			    udp1->udp_connp->conn_mac_exempt ||
8220 			    connp->conn_mac_exempt) {
8221 				if (V6_OR_V4_INADDR_ANY(
8222 				    udp1->udp_bound_v6src) ||
8223 				    is_inaddr_any ||
8224 				    IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
8225 				    &v6src)) {
8226 					found_exclbind = B_TRUE;
8227 					break;
8228 				}
8229 				continue;
8230 			}
8231 
8232 			/*
8233 			 * Check ipversion to allow IPv4 and IPv6 sockets to
8234 			 * have disjoint port number spaces.
8235 			 */
8236 			if (udp->udp_ipversion != udp1->udp_ipversion) {
8237 
8238 				/*
8239 				 * On the first time through the loop, if the
8240 				 * the user intentionally specified a
8241 				 * particular port number, then ignore any
8242 				 * bindings of the other protocol that may
8243 				 * conflict. This allows the user to bind IPv6
8244 				 * alone and get both v4 and v6, or bind both
8245 				 * both and get each seperately. On subsequent
8246 				 * times through the loop, we're checking a
8247 				 * port that we chose (not the user) and thus
8248 				 * we do not allow casual duplicate bindings.
8249 				 */
8250 				if (count == 0 && requested_port != 0)
8251 					continue;
8252 			}
8253 
8254 			/*
8255 			 * No difference depending on SO_REUSEADDR.
8256 			 *
8257 			 * If existing port is bound to a
8258 			 * non-wildcard IP address and
8259 			 * the requesting stream is bound to
8260 			 * a distinct different IP addresses
8261 			 * (non-wildcard, also), keep going.
8262 			 */
8263 			if (!is_inaddr_any &&
8264 			    !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) &&
8265 			    !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
8266 			    &v6src)) {
8267 				continue;
8268 			}
8269 			break;
8270 		}
8271 
8272 		if (!found_exclbind &&
8273 		    (udp->udp_reuseaddr && requested_port != 0)) {
8274 			break;
8275 		}
8276 
8277 		if (udp1 == NULL) {
8278 			/*
8279 			 * No other stream has this IP address
8280 			 * and port number. We can use it.
8281 			 */
8282 			break;
8283 		}
8284 		mutex_exit(&udpf->uf_lock);
8285 		if (bind_to_req_port_only) {
8286 			/*
8287 			 * We get here only when requested port
8288 			 * is bound (and only first  of the for()
8289 			 * loop iteration).
8290 			 *
8291 			 * The semantics of this bind request
8292 			 * require it to fail so we return from
8293 			 * the routine (and exit the loop).
8294 			 *
8295 			 */
8296 			udp->udp_pending_op = -1;
8297 			rw_exit(&udp->udp_rwlock);
8298 			return (-TADDRBUSY);
8299 		}
8300 
8301 		if (udp->udp_anon_priv_bind) {
8302 			port = udp_get_next_priv_port(udp);
8303 		} else {
8304 			if ((count == 0) && (requested_port != 0)) {
8305 				/*
8306 				 * If the application wants us to find
8307 				 * a port, get one to start with. Set
8308 				 * requested_port to 0, so that we will
8309 				 * update us->us_next_port_to_try below.
8310 				 */
8311 				port = udp_update_next_port(udp,
8312 				    us->us_next_port_to_try, B_TRUE);
8313 				requested_port = 0;
8314 			} else {
8315 				port = udp_update_next_port(udp, port + 1,
8316 				    B_FALSE);
8317 			}
8318 		}
8319 
8320 		if (port == 0 || ++count >= loopmax) {
8321 			/*
8322 			 * We've tried every possible port number and
8323 			 * there are none available, so send an error
8324 			 * to the user.
8325 			 */
8326 			udp->udp_pending_op = -1;
8327 			rw_exit(&udp->udp_rwlock);
8328 			return (-TNOADDR);
8329 		}
8330 	}
8331 
8332 	/*
8333 	 * Copy the source address into our udp structure.  This address
8334 	 * may still be zero; if so, ip will fill in the correct address
8335 	 * each time an outbound packet is passed to it.
8336 	 * If we are binding to a broadcast or multicast address then
8337 	 * udp_post_ip_bind_connect will clear the source address
8338 	 * when udp_do_bind success.
8339 	 */
8340 	udp->udp_v6src = udp->udp_bound_v6src = v6src;
8341 	udp->udp_port = lport;
8342 	/*
8343 	 * Now reset the the next anonymous port if the application requested
8344 	 * an anonymous port, or we handed out the next anonymous port.
8345 	 */
8346 	if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) {
8347 		us->us_next_port_to_try = port + 1;
8348 	}
8349 
8350 	/* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */
8351 	if (udp->udp_family == AF_INET) {
8352 		sin->sin_port = udp->udp_port;
8353 	} else {
8354 		sin6->sin6_port = udp->udp_port;
8355 		/* Rebuild the header template */
8356 		error = udp_build_hdrs(udp);
8357 		if (error != 0) {
8358 			udp->udp_pending_op = -1;
8359 			rw_exit(&udp->udp_rwlock);
8360 			mutex_exit(&udpf->uf_lock);
8361 			return (error);
8362 		}
8363 	}
8364 	udp->udp_state = TS_IDLE;
8365 	udp_bind_hash_insert(udpf, udp);
8366 	mutex_exit(&udpf->uf_lock);
8367 	rw_exit(&udp->udp_rwlock);
8368 
8369 	if (cl_inet_bind) {
8370 		/*
8371 		 * Running in cluster mode - register bind information
8372 		 */
8373 		if (udp->udp_ipversion == IPV4_VERSION) {
8374 			(*cl_inet_bind)(connp->conn_netstack->netstack_stackid,
8375 			    IPPROTO_UDP, AF_INET,
8376 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8377 			    (in_port_t)udp->udp_port, NULL);
8378 		} else {
8379 			(*cl_inet_bind)(connp->conn_netstack->netstack_stackid,
8380 			    IPPROTO_UDP, AF_INET6,
8381 			    (uint8_t *)&(udp->udp_v6src),
8382 			    (in_port_t)udp->udp_port, NULL);
8383 		}
8384 	}
8385 
8386 	connp->conn_anon_port = (is_system_labeled() && requested_port == 0);
8387 	if (is_system_labeled() && (!connp->conn_anon_port ||
8388 	    connp->conn_anon_mlp)) {
8389 		uint16_t mlpport;
8390 		cred_t *cr = connp->conn_cred;
8391 		zone_t *zone;
8392 
8393 		zone = crgetzone(cr);
8394 		connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth :
8395 		    mlptSingle;
8396 		addrtype = tsol_mlp_addr_type(zone->zone_id, IPV6_VERSION,
8397 		    &v6src, us->us_netstack->netstack_ip);
8398 		if (addrtype == mlptSingle) {
8399 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8400 			udp->udp_pending_op = -1;
8401 			rw_exit(&udp->udp_rwlock);
8402 			connp->conn_anon_port = B_FALSE;
8403 			connp->conn_mlp_type = mlptSingle;
8404 			return (-TNOADDR);
8405 		}
8406 		mlpport = connp->conn_anon_port ? PMAPPORT : port;
8407 		mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport,
8408 		    addrtype);
8409 		if (mlptype != mlptSingle &&
8410 		    (connp->conn_mlp_type == mlptSingle ||
8411 		    secpolicy_net_bindmlp(cr) != 0)) {
8412 			if (udp->udp_debug) {
8413 				(void) strlog(UDP_MOD_ID, 0, 1,
8414 				    SL_ERROR|SL_TRACE,
8415 				    "udp_bind: no priv for multilevel port %d",
8416 				    mlpport);
8417 			}
8418 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8419 			udp->udp_pending_op = -1;
8420 			rw_exit(&udp->udp_rwlock);
8421 			connp->conn_anon_port = B_FALSE;
8422 			connp->conn_mlp_type = mlptSingle;
8423 			return (-TACCES);
8424 		}
8425 
8426 		/*
8427 		 * If we're specifically binding a shared IP address and the
8428 		 * port is MLP on shared addresses, then check to see if this
8429 		 * zone actually owns the MLP.  Reject if not.
8430 		 */
8431 		if (mlptype == mlptShared && addrtype == mlptShared) {
8432 			/*
8433 			 * No need to handle exclusive-stack zones since
8434 			 * ALL_ZONES only applies to the shared stack.
8435 			 */
8436 			zoneid_t mlpzone;
8437 
8438 			mlpzone = tsol_mlp_findzone(IPPROTO_UDP,
8439 			    htons(mlpport));
8440 			if (connp->conn_zoneid != mlpzone) {
8441 				if (udp->udp_debug) {
8442 					(void) strlog(UDP_MOD_ID, 0, 1,
8443 					    SL_ERROR|SL_TRACE,
8444 					    "udp_bind: attempt to bind port "
8445 					    "%d on shared addr in zone %d "
8446 					    "(should be %d)",
8447 					    mlpport, connp->conn_zoneid,
8448 					    mlpzone);
8449 				}
8450 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8451 				udp->udp_pending_op = -1;
8452 				rw_exit(&udp->udp_rwlock);
8453 				connp->conn_anon_port = B_FALSE;
8454 				connp->conn_mlp_type = mlptSingle;
8455 				return (-TACCES);
8456 			}
8457 		}
8458 		if (connp->conn_anon_port) {
8459 			error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp,
8460 			    port, B_TRUE);
8461 			if (error != 0) {
8462 				if (udp->udp_debug) {
8463 					(void) strlog(UDP_MOD_ID, 0, 1,
8464 					    SL_ERROR|SL_TRACE,
8465 					    "udp_bind: cannot establish anon "
8466 					    "MLP for port %d", port);
8467 				}
8468 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8469 				udp->udp_pending_op = -1;
8470 				rw_exit(&udp->udp_rwlock);
8471 				connp->conn_anon_port = B_FALSE;
8472 				connp->conn_mlp_type = mlptSingle;
8473 				return (-TACCES);
8474 			}
8475 		}
8476 		connp->conn_mlp_type = mlptype;
8477 	}
8478 
8479 	if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8480 		/*
8481 		 * Append a request for an IRE if udp_v6src not
8482 		 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address).
8483 		 */
8484 		mp = allocb(sizeof (ire_t), BPRI_HI);
8485 		if (!mp) {
8486 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8487 			udp->udp_pending_op = -1;
8488 			rw_exit(&udp->udp_rwlock);
8489 			return (ENOMEM);
8490 		}
8491 		mp->b_wptr += sizeof (ire_t);
8492 		mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8493 	}
8494 	if (udp->udp_family == AF_INET6) {
8495 		ASSERT(udp->udp_connp->conn_af_isv6);
8496 		error = ip_proto_bind_laddr_v6(connp, &mp, IPPROTO_UDP,
8497 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
8498 	} else {
8499 		ASSERT(!udp->udp_connp->conn_af_isv6);
8500 		error = ip_proto_bind_laddr_v4(connp, &mp, IPPROTO_UDP,
8501 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port,
8502 		    B_TRUE);
8503 	}
8504 
8505 	(void) udp_post_ip_bind_connect(udp, mp, error);
8506 	return (error);
8507 }
8508 
8509 int
8510 udp_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
8511     socklen_t len, cred_t *cr)
8512 {
8513 	int		error;
8514 	conn_t		*connp;
8515 
8516 	connp = (conn_t *)proto_handle;
8517 
8518 	if (sa == NULL)
8519 		error = udp_do_unbind(connp);
8520 	else
8521 		error = udp_do_bind(connp, sa, len, cr, B_TRUE);
8522 
8523 	if (error < 0) {
8524 		if (error == -TOUTSTATE)
8525 			error = EINVAL;
8526 		else
8527 			error = proto_tlitosyserr(-error);
8528 	}
8529 
8530 	return (error);
8531 }
8532 
8533 static int
8534 udp_implicit_bind(conn_t *connp, cred_t *cr)
8535 {
8536 	int error;
8537 
8538 	error = udp_do_bind(connp, NULL, 0, cr, B_FALSE);
8539 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
8540 }
8541 
8542 /*
8543  * This routine removes a port number association from a stream. It
8544  * is called by udp_unbind and udp_tpi_unbind.
8545  */
8546 static int
8547 udp_do_unbind(conn_t *connp)
8548 {
8549 	udp_t 		*udp = connp->conn_udp;
8550 	udp_fanout_t	*udpf;
8551 	udp_stack_t	*us = udp->udp_us;
8552 
8553 	if (cl_inet_unbind != NULL) {
8554 		/*
8555 		 * Running in cluster mode - register unbind information
8556 		 */
8557 		if (udp->udp_ipversion == IPV4_VERSION) {
8558 			(*cl_inet_unbind)(
8559 			    connp->conn_netstack->netstack_stackid,
8560 			    IPPROTO_UDP, AF_INET,
8561 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8562 			    (in_port_t)udp->udp_port, NULL);
8563 		} else {
8564 			(*cl_inet_unbind)(
8565 			    connp->conn_netstack->netstack_stackid,
8566 			    IPPROTO_UDP, AF_INET6,
8567 			    (uint8_t *)&(udp->udp_v6src),
8568 			    (in_port_t)udp->udp_port, NULL);
8569 		}
8570 	}
8571 
8572 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8573 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8574 		rw_exit(&udp->udp_rwlock);
8575 		return (-TOUTSTATE);
8576 	}
8577 	udp->udp_pending_op = T_UNBIND_REQ;
8578 	rw_exit(&udp->udp_rwlock);
8579 
8580 	/*
8581 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
8582 	 * and therefore ip_unbind must never return NULL.
8583 	 */
8584 	ip_unbind(connp);
8585 
8586 	/*
8587 	 * Once we're unbound from IP, the pending operation may be cleared
8588 	 * here.
8589 	 */
8590 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8591 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8592 	    us->us_bind_fanout_size)];
8593 
8594 	mutex_enter(&udpf->uf_lock);
8595 	udp_bind_hash_remove(udp, B_TRUE);
8596 	V6_SET_ZERO(udp->udp_v6src);
8597 	V6_SET_ZERO(udp->udp_bound_v6src);
8598 	udp->udp_port = 0;
8599 	mutex_exit(&udpf->uf_lock);
8600 
8601 	udp->udp_pending_op = -1;
8602 	udp->udp_state = TS_UNBND;
8603 	if (udp->udp_family == AF_INET6)
8604 		(void) udp_build_hdrs(udp);
8605 	rw_exit(&udp->udp_rwlock);
8606 
8607 	return (0);
8608 }
8609 
8610 static int
8611 udp_post_ip_bind_connect(udp_t *udp, mblk_t *ire_mp, int error)
8612 {
8613 	ire_t		*ire;
8614 	udp_fanout_t	*udpf;
8615 	udp_stack_t	*us = udp->udp_us;
8616 
8617 	ASSERT(udp->udp_pending_op != -1);
8618 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8619 	if (error == 0) {
8620 		/* For udp_do_connect() success */
8621 		/* udp_do_bind() success will do nothing in here */
8622 		/*
8623 		 * If a broadcast/multicast address was bound, set
8624 		 * the source address to 0.
8625 		 * This ensures no datagrams with broadcast address
8626 		 * as source address are emitted (which would violate
8627 		 * RFC1122 - Hosts requirements)
8628 		 *
8629 		 * Note that when connecting the returned IRE is
8630 		 * for the destination address and we only perform
8631 		 * the broadcast check for the source address (it
8632 		 * is OK to connect to a broadcast/multicast address.)
8633 		 */
8634 		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
8635 			ire = (ire_t *)ire_mp->b_rptr;
8636 
8637 			/*
8638 			 * Note: we get IRE_BROADCAST for IPv6 to "mark" a
8639 			 * multicast local address.
8640 			 */
8641 			udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8642 			    us->us_bind_fanout_size)];
8643 			if (ire->ire_type == IRE_BROADCAST &&
8644 			    udp->udp_state != TS_DATA_XFER) {
8645 				ASSERT(udp->udp_pending_op == T_BIND_REQ ||
8646 				    udp->udp_pending_op == O_T_BIND_REQ);
8647 				/*
8648 				 * This was just a local bind to a broadcast
8649 				 * addr.
8650 				 */
8651 				mutex_enter(&udpf->uf_lock);
8652 				V6_SET_ZERO(udp->udp_v6src);
8653 				mutex_exit(&udpf->uf_lock);
8654 				if (udp->udp_family == AF_INET6)
8655 					(void) udp_build_hdrs(udp);
8656 			} else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8657 				if (udp->udp_family == AF_INET6)
8658 					(void) udp_build_hdrs(udp);
8659 			}
8660 		}
8661 	} else {
8662 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8663 		    us->us_bind_fanout_size)];
8664 		mutex_enter(&udpf->uf_lock);
8665 
8666 		if (udp->udp_state == TS_DATA_XFER) {
8667 			/* Connect failed */
8668 			/* Revert back to the bound source */
8669 			udp->udp_v6src = udp->udp_bound_v6src;
8670 			udp->udp_state = TS_IDLE;
8671 		} else {
8672 			/* For udp_do_bind() failed */
8673 			V6_SET_ZERO(udp->udp_v6src);
8674 			V6_SET_ZERO(udp->udp_bound_v6src);
8675 			udp->udp_state = TS_UNBND;
8676 			udp_bind_hash_remove(udp, B_TRUE);
8677 			udp->udp_port = 0;
8678 		}
8679 		mutex_exit(&udpf->uf_lock);
8680 		if (udp->udp_family == AF_INET6)
8681 			(void) udp_build_hdrs(udp);
8682 	}
8683 	udp->udp_pending_op = -1;
8684 	rw_exit(&udp->udp_rwlock);
8685 	if (ire_mp != NULL)
8686 		freeb(ire_mp);
8687 	return (error);
8688 }
8689 
8690 /*
8691  * It associates a default destination address with the stream.
8692  */
8693 static int
8694 udp_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len)
8695 {
8696 	sin6_t		*sin6;
8697 	sin_t		*sin;
8698 	in6_addr_t 	v6dst;
8699 	ipaddr_t 	v4dst;
8700 	uint16_t 	dstport;
8701 	uint32_t 	flowinfo;
8702 	mblk_t		*ire_mp;
8703 	udp_fanout_t	*udpf;
8704 	udp_t		*udp, *udp1;
8705 	ushort_t	ipversion;
8706 	udp_stack_t	*us;
8707 	int		error;
8708 
8709 	udp = connp->conn_udp;
8710 	us = udp->udp_us;
8711 
8712 	/*
8713 	 * Address has been verified by the caller
8714 	 */
8715 	switch (len) {
8716 	default:
8717 		/*
8718 		 * Should never happen
8719 		 */
8720 		return (EINVAL);
8721 
8722 	case sizeof (sin_t):
8723 		sin = (sin_t *)sa;
8724 		v4dst = sin->sin_addr.s_addr;
8725 		dstport = sin->sin_port;
8726 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8727 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8728 		ipversion = IPV4_VERSION;
8729 		break;
8730 
8731 	case sizeof (sin6_t):
8732 		sin6 = (sin6_t *)sa;
8733 		v6dst = sin6->sin6_addr;
8734 		dstport = sin6->sin6_port;
8735 		if (IN6_IS_ADDR_V4MAPPED(&v6dst)) {
8736 			IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst);
8737 			ipversion = IPV4_VERSION;
8738 			flowinfo = 0;
8739 		} else {
8740 			ipversion = IPV6_VERSION;
8741 			flowinfo = sin6->sin6_flowinfo;
8742 		}
8743 		break;
8744 	}
8745 
8746 	if (dstport == 0)
8747 		return (-TBADADDR);
8748 
8749 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8750 
8751 	/*
8752 	 * This UDP must have bound to a port already before doing a connect.
8753 	 * TPI mandates that users must send TPI primitives only 1 at a time
8754 	 * and wait for the response before sending the next primitive.
8755 	 */
8756 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8757 		rw_exit(&udp->udp_rwlock);
8758 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
8759 		    "udp_connect: bad state, %u", udp->udp_state);
8760 		return (-TOUTSTATE);
8761 	}
8762 	udp->udp_pending_op = T_CONN_REQ;
8763 	ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL);
8764 
8765 	if (ipversion == IPV4_VERSION) {
8766 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
8767 		    udp->udp_ip_snd_options_len;
8768 	} else {
8769 		udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
8770 	}
8771 
8772 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8773 	    us->us_bind_fanout_size)];
8774 
8775 	mutex_enter(&udpf->uf_lock);
8776 	if (udp->udp_state == TS_DATA_XFER) {
8777 		/* Already connected - clear out state */
8778 		udp->udp_v6src = udp->udp_bound_v6src;
8779 		udp->udp_state = TS_IDLE;
8780 	}
8781 
8782 	/*
8783 	 * Create a default IP header with no IP options.
8784 	 */
8785 	udp->udp_dstport = dstport;
8786 	udp->udp_ipversion = ipversion;
8787 	if (ipversion == IPV4_VERSION) {
8788 		/*
8789 		 * Interpret a zero destination to mean loopback.
8790 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8791 		 * generate the T_CONN_CON.
8792 		 */
8793 		if (v4dst == INADDR_ANY) {
8794 			v4dst = htonl(INADDR_LOOPBACK);
8795 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8796 			if (udp->udp_family == AF_INET) {
8797 				sin->sin_addr.s_addr = v4dst;
8798 			} else {
8799 				sin6->sin6_addr = v6dst;
8800 			}
8801 		}
8802 		udp->udp_v6dst = v6dst;
8803 		udp->udp_flowinfo = 0;
8804 
8805 		/*
8806 		 * If the destination address is multicast and
8807 		 * an outgoing multicast interface has been set,
8808 		 * use the address of that interface as our
8809 		 * source address if no source address has been set.
8810 		 */
8811 		if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY &&
8812 		    CLASSD(v4dst) &&
8813 		    udp->udp_multicast_if_addr != INADDR_ANY) {
8814 			IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr,
8815 			    &udp->udp_v6src);
8816 		}
8817 	} else {
8818 		ASSERT(udp->udp_ipversion == IPV6_VERSION);
8819 		/*
8820 		 * Interpret a zero destination to mean loopback.
8821 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8822 		 * generate the T_CONN_CON.
8823 		 */
8824 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
8825 			v6dst = ipv6_loopback;
8826 			sin6->sin6_addr = v6dst;
8827 		}
8828 		udp->udp_v6dst = v6dst;
8829 		udp->udp_flowinfo = flowinfo;
8830 		/*
8831 		 * If the destination address is multicast and
8832 		 * an outgoing multicast interface has been set,
8833 		 * then the ip bind logic will pick the correct source
8834 		 * address (i.e. matching the outgoing multicast interface).
8835 		 */
8836 	}
8837 
8838 	/*
8839 	 * Verify that the src/port/dst/port is unique for all
8840 	 * connections in TS_DATA_XFER
8841 	 */
8842 	for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) {
8843 		if (udp1->udp_state != TS_DATA_XFER)
8844 			continue;
8845 		if (udp->udp_port != udp1->udp_port ||
8846 		    udp->udp_ipversion != udp1->udp_ipversion ||
8847 		    dstport != udp1->udp_dstport ||
8848 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) ||
8849 		    !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst) ||
8850 		    !(IPCL_ZONE_MATCH(udp->udp_connp,
8851 		    udp1->udp_connp->conn_zoneid) ||
8852 		    IPCL_ZONE_MATCH(udp1->udp_connp,
8853 		    udp->udp_connp->conn_zoneid)))
8854 			continue;
8855 		mutex_exit(&udpf->uf_lock);
8856 		udp->udp_pending_op = -1;
8857 		rw_exit(&udp->udp_rwlock);
8858 		return (-TBADADDR);
8859 	}
8860 
8861 	if (cl_inet_connect2 != NULL) {
8862 		CL_INET_UDP_CONNECT(connp, udp, B_TRUE, &v6dst, dstport, error);
8863 		if (error != 0) {
8864 			mutex_exit(&udpf->uf_lock);
8865 			udp->udp_pending_op = -1;
8866 			rw_exit(&udp->udp_rwlock);
8867 			return (-TBADADDR);
8868 		}
8869 	}
8870 
8871 	udp->udp_state = TS_DATA_XFER;
8872 	mutex_exit(&udpf->uf_lock);
8873 
8874 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
8875 	if (ire_mp == NULL) {
8876 		mutex_enter(&udpf->uf_lock);
8877 		udp->udp_state = TS_IDLE;
8878 		udp->udp_pending_op = -1;
8879 		mutex_exit(&udpf->uf_lock);
8880 		rw_exit(&udp->udp_rwlock);
8881 		return (ENOMEM);
8882 	}
8883 
8884 	rw_exit(&udp->udp_rwlock);
8885 
8886 	ire_mp->b_wptr += sizeof (ire_t);
8887 	ire_mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8888 
8889 	if (udp->udp_family == AF_INET) {
8890 		error = ip_proto_bind_connected_v4(connp, &ire_mp, IPPROTO_UDP,
8891 		    &V4_PART_OF_V6(udp->udp_v6src), udp->udp_port,
8892 		    V4_PART_OF_V6(udp->udp_v6dst), udp->udp_dstport,
8893 		    B_TRUE, B_TRUE);
8894 	} else {
8895 		error = ip_proto_bind_connected_v6(connp, &ire_mp, IPPROTO_UDP,
8896 		    &udp->udp_v6src, udp->udp_port, &udp->udp_v6dst,
8897 		    &udp->udp_sticky_ipp, udp->udp_dstport, B_TRUE, B_TRUE);
8898 	}
8899 
8900 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
8901 }
8902 
8903 /* ARGSUSED */
8904 static int
8905 udp_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
8906     socklen_t len, sock_connid_t *id, cred_t *cr)
8907 {
8908 	conn_t	*connp = (conn_t *)proto_handle;
8909 	udp_t	*udp = connp->conn_udp;
8910 	int	error;
8911 	boolean_t did_bind = B_FALSE;
8912 
8913 	if (sa == NULL) {
8914 		/*
8915 		 * Disconnect
8916 		 * Make sure we are connected
8917 		 */
8918 		if (udp->udp_state != TS_DATA_XFER)
8919 			return (EINVAL);
8920 
8921 		error = udp_disconnect(connp);
8922 		return (error);
8923 	}
8924 
8925 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
8926 	if (error != 0)
8927 		goto done;
8928 
8929 	/* do an implicit bind if necessary */
8930 	if (udp->udp_state == TS_UNBND) {
8931 		error = udp_implicit_bind(connp, cr);
8932 		/*
8933 		 * We could be racing with an actual bind, in which case
8934 		 * we would see EPROTO. We cross our fingers and try
8935 		 * to connect.
8936 		 */
8937 		if (!(error == 0 || error == EPROTO))
8938 			goto done;
8939 		did_bind = B_TRUE;
8940 	}
8941 	/*
8942 	 * set SO_DGRAM_ERRIND
8943 	 */
8944 	udp->udp_dgram_errind = B_TRUE;
8945 
8946 	error = udp_do_connect(connp, sa, len);
8947 
8948 	if (error != 0 && did_bind) {
8949 		int unbind_err;
8950 
8951 		unbind_err = udp_do_unbind(connp);
8952 		ASSERT(unbind_err == 0);
8953 	}
8954 
8955 	if (error == 0) {
8956 		*id = 0;
8957 		(*connp->conn_upcalls->su_connected)
8958 		    (connp->conn_upper_handle, 0, NULL, -1);
8959 	} else if (error < 0) {
8960 		error = proto_tlitosyserr(-error);
8961 	}
8962 
8963 done:
8964 	if (error != 0 && udp->udp_state == TS_DATA_XFER) {
8965 		/*
8966 		 * No need to hold locks to set state
8967 		 * after connect failure socket state is undefined
8968 		 * We set the state only to imitate old sockfs behavior
8969 		 */
8970 		udp->udp_state = TS_IDLE;
8971 	}
8972 	return (error);
8973 }
8974 
8975 /* ARGSUSED */
8976 int
8977 udp_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
8978     cred_t *cr)
8979 {
8980 	conn_t		*connp = (conn_t *)proto_handle;
8981 	udp_t		*udp = connp->conn_udp;
8982 	udp_stack_t	*us = udp->udp_us;
8983 	int		error = 0;
8984 
8985 	ASSERT(DB_TYPE(mp) == M_DATA);
8986 
8987 	/*
8988 	 * If the socket is connected and no change in destination
8989 	 */
8990 	if (msg->msg_namelen == 0) {
8991 		error = udp_send_connected(connp, mp, msg, cr, curproc->p_pid);
8992 		if (error == EDESTADDRREQ)
8993 			return (error);
8994 		else
8995 			return (udp->udp_dgram_errind ? error : 0);
8996 	}
8997 
8998 	/*
8999 	 * Do an implicit bind if necessary.
9000 	 */
9001 	if (udp->udp_state == TS_UNBND) {
9002 		error = udp_implicit_bind(connp, cr);
9003 		/*
9004 		 * We could be racing with an actual bind, in which case
9005 		 * we would see EPROTO. We cross our fingers and try
9006 		 * to send.
9007 		 */
9008 		if (!(error == 0 || error == EPROTO)) {
9009 			freemsg(mp);
9010 			return (error);
9011 		}
9012 	}
9013 
9014 	rw_enter(&udp->udp_rwlock, RW_WRITER);
9015 
9016 	if (msg->msg_name != NULL && udp->udp_state == TS_DATA_XFER) {
9017 		rw_exit(&udp->udp_rwlock);
9018 		freemsg(mp);
9019 		return (EISCONN);
9020 	}
9021 
9022 
9023 	if (udp->udp_delayed_error != 0) {
9024 		boolean_t	match;
9025 
9026 		error = udp->udp_delayed_error;
9027 		match = B_FALSE;
9028 		udp->udp_delayed_error = 0;
9029 		switch (udp->udp_family) {
9030 		case AF_INET: {
9031 			/* Compare just IP address and port */
9032 			sin_t *sin1 = (sin_t *)msg->msg_name;
9033 			sin_t *sin2 = (sin_t *)&udp->udp_delayed_addr;
9034 
9035 			if (msg->msg_namelen == sizeof (sin_t) &&
9036 			    sin1->sin_port == sin2->sin_port &&
9037 			    sin1->sin_addr.s_addr == sin2->sin_addr.s_addr)
9038 				match = B_TRUE;
9039 
9040 			break;
9041 		}
9042 		case AF_INET6: {
9043 			sin6_t	*sin1 = (sin6_t *)msg->msg_name;
9044 			sin6_t	*sin2 = (sin6_t *)&udp->udp_delayed_addr;
9045 
9046 			if (msg->msg_namelen == sizeof (sin6_t) &&
9047 			    sin1->sin6_port == sin2->sin6_port &&
9048 			    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
9049 			    &sin2->sin6_addr))
9050 				match = B_TRUE;
9051 			break;
9052 		}
9053 		default:
9054 			ASSERT(0);
9055 		}
9056 
9057 		*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
9058 
9059 		if (match) {
9060 			rw_exit(&udp->udp_rwlock);
9061 			freemsg(mp);
9062 			return (error);
9063 		}
9064 	}
9065 
9066 	error = proto_verify_ip_addr(udp->udp_family,
9067 	    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
9068 	rw_exit(&udp->udp_rwlock);
9069 
9070 	if (error != 0) {
9071 		freemsg(mp);
9072 		return (error);
9073 	}
9074 
9075 	error = udp_send_not_connected(connp, mp,
9076 	    (struct sockaddr  *)msg->msg_name, msg->msg_namelen, msg, cr,
9077 	    curproc->p_pid);
9078 	if (error != 0) {
9079 		UDP_STAT(us, udp_out_err_output);
9080 		freemsg(mp);
9081 	}
9082 	return (udp->udp_dgram_errind ? error : 0);
9083 }
9084 
9085 void
9086 udp_fallback(sock_lower_handle_t proto_handle, queue_t *q,
9087     boolean_t direct_sockfs, so_proto_quiesced_cb_t quiesced_cb)
9088 {
9089 	conn_t 	*connp = (conn_t *)proto_handle;
9090 	udp_t	*udp;
9091 	struct T_capability_ack tca;
9092 	struct sockaddr_in6 laddr, faddr;
9093 	socklen_t laddrlen, faddrlen;
9094 	short opts;
9095 	struct stroptions *stropt;
9096 	mblk_t *stropt_mp;
9097 	int error;
9098 
9099 	udp = connp->conn_udp;
9100 
9101 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
9102 
9103 	/*
9104 	 * setup the fallback stream that was allocated
9105 	 */
9106 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
9107 	connp->conn_minor_arena = WR(q)->q_ptr;
9108 
9109 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
9110 
9111 	WR(q)->q_qinfo = &udp_winit;
9112 
9113 	connp->conn_rq = RD(q);
9114 	connp->conn_wq = WR(q);
9115 
9116 	/* Notify stream head about options before sending up data */
9117 	stropt_mp->b_datap->db_type = M_SETOPTS;
9118 	stropt_mp->b_wptr += sizeof (*stropt);
9119 	stropt = (struct stroptions *)stropt_mp->b_rptr;
9120 	stropt->so_flags = SO_WROFF | SO_HIWAT;
9121 	stropt->so_wroff =
9122 	    (ushort_t)(udp->udp_max_hdr_len + udp->udp_us->us_wroff_extra);
9123 	stropt->so_hiwat = udp->udp_rcv_disply_hiwat;
9124 	putnext(RD(q), stropt_mp);
9125 
9126 	/*
9127 	 * Free the helper stream
9128 	 */
9129 	ip_free_helper_stream(connp);
9130 
9131 	if (!direct_sockfs)
9132 		udp_disable_direct_sockfs(udp);
9133 
9134 	/*
9135 	 * Collect the information needed to sync with the sonode
9136 	 */
9137 	udp_do_capability_ack(udp, &tca, TC1_INFO);
9138 
9139 	laddrlen = faddrlen = sizeof (sin6_t);
9140 	(void) udp_getsockname((sock_lower_handle_t)connp,
9141 	    (struct sockaddr *)&laddr, &laddrlen, NULL);
9142 	error = udp_getpeername((sock_lower_handle_t)connp,
9143 	    (struct sockaddr *)&faddr, &faddrlen, NULL);
9144 	if (error != 0)
9145 		faddrlen = 0;
9146 
9147 	opts = 0;
9148 	if (udp->udp_dgram_errind)
9149 		opts |= SO_DGRAM_ERRIND;
9150 	if (udp->udp_dontroute)
9151 		opts |= SO_DONTROUTE;
9152 
9153 	/*
9154 	 * Once we grab the drain lock, no data will be send up
9155 	 * to the socket. So we notify the socket that the endpoint
9156 	 * is quiescent and it's therefore safe move data from
9157 	 * the socket to the stream head.
9158 	 */
9159 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
9160 	    (struct sockaddr *)&laddr, laddrlen,
9161 	    (struct sockaddr *)&faddr, faddrlen, opts);
9162 
9163 	/*
9164 	 * push up any packets that were queued in udp_t
9165 	 */
9166 
9167 	mutex_enter(&udp->udp_recv_lock);
9168 	while (udp->udp_fallback_queue_head != NULL) {
9169 		mblk_t *mp;
9170 		mp = udp->udp_fallback_queue_head;
9171 		udp->udp_fallback_queue_head = mp->b_next;
9172 		mutex_exit(&udp->udp_recv_lock);
9173 		mp->b_next = NULL;
9174 		putnext(RD(q), mp);
9175 		mutex_enter(&udp->udp_recv_lock);
9176 	}
9177 	udp->udp_fallback_queue_tail = udp->udp_fallback_queue_head;
9178 	/*
9179 	 * No longer a streams less socket
9180 	 */
9181 	connp->conn_flags &= ~IPCL_NONSTR;
9182 	mutex_exit(&udp->udp_recv_lock);
9183 
9184 	ASSERT(connp->conn_ref >= 1);
9185 }
9186 
9187 static int
9188 udp_do_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
9189 {
9190 	sin_t	*sin = (sin_t *)sa;
9191 	sin6_t	*sin6 = (sin6_t *)sa;
9192 
9193 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
9194 	ASSERT(udp != NULL);
9195 
9196 	if (udp->udp_state != TS_DATA_XFER)
9197 		return (ENOTCONN);
9198 
9199 	switch (udp->udp_family) {
9200 	case AF_INET:
9201 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
9202 
9203 		if (*salenp < sizeof (sin_t))
9204 			return (EINVAL);
9205 
9206 		*salenp = sizeof (sin_t);
9207 		*sin = sin_null;
9208 		sin->sin_family = AF_INET;
9209 		sin->sin_port = udp->udp_dstport;
9210 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
9211 		break;
9212 	case AF_INET6:
9213 		if (*salenp < sizeof (sin6_t))
9214 			return (EINVAL);
9215 
9216 		*salenp = sizeof (sin6_t);
9217 		*sin6 = sin6_null;
9218 		sin6->sin6_family = AF_INET6;
9219 		sin6->sin6_port = udp->udp_dstport;
9220 		sin6->sin6_addr = udp->udp_v6dst;
9221 		sin6->sin6_flowinfo = udp->udp_flowinfo;
9222 		break;
9223 	}
9224 
9225 	return (0);
9226 }
9227 
9228 /* ARGSUSED */
9229 int
9230 udp_getpeername(sock_lower_handle_t  proto_handle, struct sockaddr *sa,
9231     socklen_t *salenp, cred_t *cr)
9232 {
9233 	conn_t	*connp = (conn_t *)proto_handle;
9234 	udp_t	*udp = connp->conn_udp;
9235 	int error;
9236 
9237 	ASSERT(udp != NULL);
9238 
9239 	rw_enter(&udp->udp_rwlock, RW_READER);
9240 
9241 	error = udp_do_getpeername(udp, sa, salenp);
9242 
9243 	rw_exit(&udp->udp_rwlock);
9244 
9245 	return (error);
9246 }
9247 
9248 static int
9249 udp_do_getsockname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
9250 {
9251 	sin_t	*sin = (sin_t *)sa;
9252 	sin6_t	*sin6 = (sin6_t *)sa;
9253 
9254 	ASSERT(udp != NULL);
9255 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
9256 
9257 	switch (udp->udp_family) {
9258 	case AF_INET:
9259 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
9260 
9261 		if (*salenp < sizeof (sin_t))
9262 			return (EINVAL);
9263 
9264 		*salenp = sizeof (sin_t);
9265 		*sin = sin_null;
9266 		sin->sin_family = AF_INET;
9267 		if (udp->udp_state == TS_UNBND) {
9268 			break;
9269 		}
9270 		sin->sin_port = udp->udp_port;
9271 
9272 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
9273 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9274 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
9275 		} else {
9276 			/*
9277 			 * INADDR_ANY
9278 			 * udp_v6src is not set, we might be bound to
9279 			 * broadcast/multicast. Use udp_bound_v6src as
9280 			 * local address instead (that could
9281 			 * also still be INADDR_ANY)
9282 			 */
9283 			sin->sin_addr.s_addr =
9284 			    V4_PART_OF_V6(udp->udp_bound_v6src);
9285 		}
9286 		break;
9287 
9288 	case AF_INET6:
9289 		if (*salenp < sizeof (sin6_t))
9290 			return (EINVAL);
9291 
9292 		*salenp = sizeof (sin6_t);
9293 		*sin6 = sin6_null;
9294 		sin6->sin6_family = AF_INET6;
9295 		if (udp->udp_state == TS_UNBND) {
9296 			break;
9297 		}
9298 		sin6->sin6_port = udp->udp_port;
9299 
9300 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9301 			sin6->sin6_addr = udp->udp_v6src;
9302 		} else {
9303 			/*
9304 			 * UNSPECIFIED
9305 			 * udp_v6src is not set, we might be bound to
9306 			 * broadcast/multicast. Use udp_bound_v6src as
9307 			 * local address instead (that could
9308 			 * also still be UNSPECIFIED)
9309 			 */
9310 			sin6->sin6_addr = udp->udp_bound_v6src;
9311 		}
9312 	}
9313 	return (0);
9314 }
9315 
9316 /* ARGSUSED */
9317 int
9318 udp_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
9319     socklen_t *salenp, cred_t *cr)
9320 {
9321 	conn_t	*connp = (conn_t *)proto_handle;
9322 	udp_t	*udp = connp->conn_udp;
9323 	int error;
9324 
9325 	ASSERT(udp != NULL);
9326 	rw_enter(&udp->udp_rwlock, RW_READER);
9327 
9328 	error = udp_do_getsockname(udp, sa, salenp);
9329 
9330 	rw_exit(&udp->udp_rwlock);
9331 
9332 	return (error);
9333 }
9334 
9335 int
9336 udp_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9337     void *optvalp, socklen_t *optlen, cred_t *cr)
9338 {
9339 	conn_t		*connp = (conn_t *)proto_handle;
9340 	udp_t		*udp = connp->conn_udp;
9341 	int		error;
9342 	t_uscalar_t	max_optbuf_len;
9343 	void		*optvalp_buf;
9344 	int		len;
9345 
9346 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
9347 	    udp_opt_obj.odb_opt_des_arr,
9348 	    udp_opt_obj.odb_opt_arr_cnt,
9349 	    udp_opt_obj.odb_topmost_tpiprovider,
9350 	    B_FALSE, B_TRUE, cr);
9351 	if (error != 0) {
9352 		if (error < 0)
9353 			error = proto_tlitosyserr(-error);
9354 		return (error);
9355 	}
9356 
9357 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
9358 	rw_enter(&udp->udp_rwlock, RW_READER);
9359 	len = udp_opt_get(connp, level, option_name, optvalp_buf);
9360 	rw_exit(&udp->udp_rwlock);
9361 
9362 	if (len < 0) {
9363 		/*
9364 		 * Pass on to IP
9365 		 */
9366 		kmem_free(optvalp_buf, max_optbuf_len);
9367 		return (ip_get_options(connp, level, option_name,
9368 		    optvalp, optlen, cr));
9369 	} else {
9370 		/*
9371 		 * update optlen and copy option value
9372 		 */
9373 		t_uscalar_t size = MIN(len, *optlen);
9374 		bcopy(optvalp_buf, optvalp, size);
9375 		bcopy(&size, optlen, sizeof (size));
9376 
9377 		kmem_free(optvalp_buf, max_optbuf_len);
9378 		return (0);
9379 	}
9380 }
9381 
9382 int
9383 udp_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9384     const void *optvalp, socklen_t optlen, cred_t *cr)
9385 {
9386 	conn_t		*connp = (conn_t *)proto_handle;
9387 	udp_t		*udp = connp->conn_udp;
9388 	int		error;
9389 
9390 	error = proto_opt_check(level, option_name, optlen, NULL,
9391 	    udp_opt_obj.odb_opt_des_arr,
9392 	    udp_opt_obj.odb_opt_arr_cnt,
9393 	    udp_opt_obj.odb_topmost_tpiprovider,
9394 	    B_TRUE, B_FALSE, cr);
9395 
9396 	if (error != 0) {
9397 		if (error < 0)
9398 			error = proto_tlitosyserr(-error);
9399 		return (error);
9400 	}
9401 
9402 	rw_enter(&udp->udp_rwlock, RW_WRITER);
9403 	error = udp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level, option_name,
9404 	    optlen, (uchar_t *)optvalp, (uint_t *)&optlen, (uchar_t *)optvalp,
9405 	    NULL, cr);
9406 	rw_exit(&udp->udp_rwlock);
9407 
9408 	if (error < 0) {
9409 		/*
9410 		 * Pass on to ip
9411 		 */
9412 		error = ip_set_options(connp, level, option_name, optvalp,
9413 		    optlen, cr);
9414 	}
9415 
9416 	return (error);
9417 }
9418 
9419 void
9420 udp_clr_flowctrl(sock_lower_handle_t proto_handle)
9421 {
9422 	conn_t	*connp = (conn_t *)proto_handle;
9423 	udp_t	*udp = connp->conn_udp;
9424 
9425 	mutex_enter(&udp->udp_recv_lock);
9426 	connp->conn_flow_cntrld = B_FALSE;
9427 	mutex_exit(&udp->udp_recv_lock);
9428 }
9429 
9430 /* ARGSUSED */
9431 int
9432 udp_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
9433 {
9434 	conn_t	*connp = (conn_t *)proto_handle;
9435 
9436 	/* shut down the send side */
9437 	if (how != SHUT_RD)
9438 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9439 		    SOCK_OPCTL_SHUT_SEND, 0);
9440 	/* shut down the recv side */
9441 	if (how != SHUT_WR)
9442 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9443 		    SOCK_OPCTL_SHUT_RECV, 0);
9444 	return (0);
9445 }
9446 
9447 int
9448 udp_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
9449     int mode, int32_t *rvalp, cred_t *cr)
9450 {
9451 	conn_t  	*connp = (conn_t *)proto_handle;
9452 	int		error;
9453 
9454 	switch (cmd) {
9455 		case ND_SET:
9456 		case ND_GET:
9457 		case _SIOCSOCKFALLBACK:
9458 		case TI_GETPEERNAME:
9459 		case TI_GETMYNAME:
9460 			ip1dbg(("udp_ioctl: cmd 0x%x on non streams socket",
9461 			    cmd));
9462 			error = EINVAL;
9463 			break;
9464 		default:
9465 			/*
9466 			 * Pass on to IP using helper stream
9467 			 */
9468 			error = ldi_ioctl(connp->conn_helper_info->iphs_handle,
9469 			    cmd, arg, mode, cr, rvalp);
9470 			break;
9471 	}
9472 	return (error);
9473 }
9474 
9475 /* ARGSUSED */
9476 int
9477 udp_accept(sock_lower_handle_t lproto_handle,
9478     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
9479     cred_t *cr)
9480 {
9481 	return (EOPNOTSUPP);
9482 }
9483 
9484 /* ARGSUSED */
9485 int
9486 udp_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
9487 {
9488 	return (EOPNOTSUPP);
9489 }
9490 
9491 sock_downcalls_t sock_udp_downcalls = {
9492 	udp_activate,		/* sd_activate */
9493 	udp_accept,		/* sd_accept */
9494 	udp_bind,		/* sd_bind */
9495 	udp_listen,		/* sd_listen */
9496 	udp_connect,		/* sd_connect */
9497 	udp_getpeername,	/* sd_getpeername */
9498 	udp_getsockname,	/* sd_getsockname */
9499 	udp_getsockopt,		/* sd_getsockopt */
9500 	udp_setsockopt,		/* sd_setsockopt */
9501 	udp_send,		/* sd_send */
9502 	NULL,			/* sd_send_uio */
9503 	NULL,			/* sd_recv_uio */
9504 	NULL,			/* sd_poll */
9505 	udp_shutdown,		/* sd_shutdown */
9506 	udp_clr_flowctrl,	/* sd_setflowctrl */
9507 	udp_ioctl,		/* sd_ioctl */
9508 	udp_close		/* sd_close */
9509 };
9510