xref: /illumos-gate/usr/src/uts/common/inet/udp/udp.c (revision f5fc5c043551e3eab595da7d8b5a53479207738f)
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 2008 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 
84 /*
85  * The ipsec_info.h header file is here since it has the definition for the
86  * M_CTL message types used by IP to convey information to the ULP. The
87  * ipsec_info.h needs the pfkeyv2.h, hence the latter's presence.
88  */
89 #include <net/pfkeyv2.h>
90 #include <inet/ipsec_info.h>
91 
92 #include <sys/tsol/label.h>
93 #include <sys/tsol/tnet.h>
94 #include <rpc/pmap_prot.h>
95 
96 /*
97  * Synchronization notes:
98  *
99  * UDP is MT and uses the usual kernel synchronization primitives. There are 2
100  * locks, the fanout lock (uf_lock) and the udp endpoint lock udp_rwlock.
101  * We also use conn_lock when updating things that affect the IP classifier
102  * lookup.
103  * The lock order is udp_rwlock -> uf_lock and is udp_rwlock -> conn_lock.
104  *
105  * The fanout lock uf_lock:
106  * When a UDP endpoint is bound to a local port, it is inserted into
107  * a bind hash list.  The list consists of an array of udp_fanout_t buckets.
108  * The size of the array is controlled by the udp_bind_fanout_size variable.
109  * This variable can be changed in /etc/system if the default value is
110  * not large enough.  Each bind hash bucket is protected by a per bucket
111  * lock.  It protects the udp_bind_hash and udp_ptpbhn fields in the udp_t
112  * structure and a few other fields in the udp_t. A UDP endpoint is removed
113  * from the bind hash list only when it is being unbound or being closed.
114  * The per bucket lock also protects a UDP endpoint's state changes.
115  *
116  * The udp_rwlock:
117  * This protects most of the other fields in the udp_t. The exact list of
118  * fields which are protected by each of the above locks is documented in
119  * the udp_t structure definition.
120  *
121  * Plumbing notes:
122  * UDP is always a device driver. For compatibility with mibopen() code
123  * it is possible to I_PUSH "udp", but that results in pushing a passthrough
124  * dummy module.
125  *
126  * The above implies that we don't support any intermediate module to
127  * reside in between /dev/ip and udp -- in fact, we never supported such
128  * scenario in the past as the inter-layer communication semantics have
129  * always been private.
130  */
131 
132 /* For /etc/system control */
133 uint_t udp_bind_fanout_size = UDP_BIND_FANOUT_SIZE;
134 
135 #define	NDD_TOO_QUICK_MSG \
136 	"ndd get info rate too high for non-privileged users, try again " \
137 	"later.\n"
138 #define	NDD_OUT_OF_BUF_MSG	"<< Out of buffer >>\n"
139 
140 /* Option processing attrs */
141 typedef struct udpattrs_s {
142 	union {
143 		ip6_pkt_t	*udpattr_ipp6;	/* For V6 */
144 		ip4_pkt_t 	*udpattr_ipp4;	/* For V4 */
145 	} udpattr_ippu;
146 #define	udpattr_ipp6 udpattr_ippu.udpattr_ipp6
147 #define	udpattr_ipp4 udpattr_ippu.udpattr_ipp4
148 	mblk_t		*udpattr_mb;
149 	boolean_t	udpattr_credset;
150 } udpattrs_t;
151 
152 static void	udp_addr_req(queue_t *q, mblk_t *mp);
153 static void	udp_tpi_bind(queue_t *q, mblk_t *mp);
154 static void	udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp);
155 static void	udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock);
156 static int	udp_build_hdrs(udp_t *udp);
157 static void	udp_capability_req(queue_t *q, mblk_t *mp);
158 static int	udp_tpi_close(queue_t *q, int flags);
159 static void	udp_tpi_connect(queue_t *q, mblk_t *mp);
160 static void	udp_tpi_disconnect(queue_t *q, mblk_t *mp);
161 static void	udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
162 		    int sys_error);
163 static void	udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive,
164 		    t_scalar_t tlierr, int unixerr);
165 static int	udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp,
166 		    cred_t *cr);
167 static int	udp_extra_priv_ports_add(queue_t *q, mblk_t *mp,
168 		    char *value, caddr_t cp, cred_t *cr);
169 static int	udp_extra_priv_ports_del(queue_t *q, mblk_t *mp,
170 		    char *value, caddr_t cp, cred_t *cr);
171 static void	udp_icmp_error(conn_t *, mblk_t *);
172 static void	udp_icmp_error_ipv6(conn_t *, mblk_t *);
173 static void	udp_info_req(queue_t *q, mblk_t *mp);
174 static void	udp_input(void *, mblk_t *, void *);
175 static mblk_t	*udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim,
176 		    t_scalar_t addr_length);
177 static void	udp_lrput(queue_t *, mblk_t *);
178 static void	udp_lwput(queue_t *, mblk_t *);
179 static int	udp_open(queue_t *q, dev_t *devp, int flag, int sflag,
180 		    cred_t *credp, boolean_t isv6);
181 static int	udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
182 		    cred_t *credp);
183 static int	udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
184 		    cred_t *credp);
185 static  int	udp_unitdata_opt_process(queue_t *q, mblk_t *mp,
186 		    int *errorp, udpattrs_t *udpattrs);
187 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
188 static int	udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
189 static boolean_t udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt);
190 static int	udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
191 		    cred_t *cr);
192 static void	udp_report_item(mblk_t *mp, udp_t *udp);
193 static int	udp_rinfop(queue_t *q, infod_t *dp);
194 static int	udp_rrw(queue_t *q, struiod_t *dp);
195 static int	udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
196 		    cred_t *cr);
197 static void	udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp,
198 		    ipha_t *ipha);
199 static void	udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr,
200 		    t_scalar_t destlen, t_scalar_t err);
201 static void	udp_tpi_unbind(queue_t *q, mblk_t *mp);
202 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port,
203     boolean_t random);
204 static mblk_t	*udp_output_v4(conn_t *, mblk_t *, ipaddr_t, uint16_t, uint_t,
205 		    int *, boolean_t, struct nmsghdr *, cred_t *, pid_t);
206 static mblk_t	*udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6,
207 		    int *error, struct nmsghdr *msg, cred_t *cr, pid_t pid);
208 static void	udp_wput_other(queue_t *q, mblk_t *mp);
209 static void	udp_wput_iocdata(queue_t *q, mblk_t *mp);
210 static void	udp_wput_fallback(queue_t *q, mblk_t *mp);
211 static size_t	udp_set_rcv_hiwat(udp_t *udp, size_t size);
212 
213 static void	*udp_stack_init(netstackid_t stackid, netstack_t *ns);
214 static void	udp_stack_fini(netstackid_t stackid, void *arg);
215 
216 static void	*udp_kstat_init(netstackid_t stackid);
217 static void	udp_kstat_fini(netstackid_t stackid, kstat_t *ksp);
218 static void	*udp_kstat2_init(netstackid_t, udp_stat_t *);
219 static void	udp_kstat2_fini(netstackid_t, kstat_t *);
220 static int	udp_kstat_update(kstat_t *kp, int rw);
221 
222 static void	udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp,
223 		    uint_t pkt_len);
224 static void	udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing);
225 static void	udp_xmit(queue_t *, mblk_t *, ire_t *ire, conn_t *, zoneid_t);
226 
227 static int	udp_send_connected(conn_t *, mblk_t *, struct nmsghdr *,
228 		    cred_t *, pid_t);
229 
230 /* Common routine for TPI and socket module */
231 static conn_t	*udp_do_open(cred_t *, boolean_t, int);
232 static void	udp_do_close(conn_t *);
233 static int	udp_do_bind(conn_t *, struct sockaddr *, socklen_t, cred_t *,
234     boolean_t);
235 static int	udp_do_unbind(conn_t *);
236 static int	udp_do_getsockname(udp_t *, struct sockaddr *, uint_t *);
237 static int	udp_do_getpeername(udp_t *, struct sockaddr *, uint_t *);
238 
239 int		udp_getsockname(sock_lower_handle_t,
240     struct sockaddr *, socklen_t *, cred_t *);
241 int		udp_getpeername(sock_lower_handle_t,
242     struct sockaddr *, socklen_t *, cred_t *);
243 static int	udp_do_connect(conn_t *, const struct sockaddr *, socklen_t);
244 static int	udp_post_ip_bind_connect(udp_t *, mblk_t *, int);
245 
246 #define	UDP_RECV_HIWATER	(56 * 1024)
247 #define	UDP_RECV_LOWATER	128
248 #define	UDP_XMIT_HIWATER	(56 * 1024)
249 #define	UDP_XMIT_LOWATER	1024
250 
251 static struct module_info udp_mod_info =  {
252 	UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER
253 };
254 
255 /*
256  * Entry points for UDP as a device.
257  * We have separate open functions for the /dev/udp and /dev/udp6 devices.
258  */
259 static struct qinit udp_rinitv4 = {
260 	NULL, NULL, udp_openv4, udp_tpi_close, NULL,
261 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
262 };
263 
264 static struct qinit udp_rinitv6 = {
265 	NULL, NULL, udp_openv6, udp_tpi_close, NULL,
266 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
267 };
268 
269 static struct qinit udp_winit = {
270 	(pfi_t)udp_wput, (pfi_t)ip_wsrv, NULL, NULL, NULL,
271 	&udp_mod_info, NULL, NULL, NULL, STRUIOT_NONE
272 };
273 
274 /* UDP entry point during fallback */
275 struct qinit udp_fallback_sock_winit = {
276 	(pfi_t)udp_wput_fallback, NULL, NULL, NULL, NULL, &udp_mod_info
277 };
278 
279 /*
280  * UDP needs to handle I_LINK and I_PLINK since ifconfig
281  * likes to use it as a place to hang the various streams.
282  */
283 static struct qinit udp_lrinit = {
284 	(pfi_t)udp_lrput, NULL, udp_openv4, udp_tpi_close, NULL,
285 	&udp_mod_info
286 };
287 
288 static struct qinit udp_lwinit = {
289 	(pfi_t)udp_lwput, NULL, udp_openv4, udp_tpi_close, NULL,
290 	&udp_mod_info
291 };
292 
293 /* For AF_INET aka /dev/udp */
294 struct streamtab udpinfov4 = {
295 	&udp_rinitv4, &udp_winit, &udp_lrinit, &udp_lwinit
296 };
297 
298 /* For AF_INET6 aka /dev/udp6 */
299 struct streamtab udpinfov6 = {
300 	&udp_rinitv6, &udp_winit, &udp_lrinit, &udp_lwinit
301 };
302 
303 static	sin_t	sin_null;	/* Zero address for quick clears */
304 static	sin6_t	sin6_null;	/* Zero address for quick clears */
305 
306 #define	UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH)
307 
308 /* Default structure copied into T_INFO_ACK messages */
309 static struct T_info_ack udp_g_t_info_ack_ipv4 = {
310 	T_INFO_ACK,
311 	UDP_MAXPACKET_IPV4,	/* TSDU_size. Excl. headers */
312 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
313 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
314 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
315 	sizeof (sin_t),	/* ADDR_size. */
316 	0,		/* OPT_size - not initialized here */
317 	UDP_MAXPACKET_IPV4,	/* TIDU_size.  Excl. headers */
318 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
319 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
320 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
321 };
322 
323 #define	UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN)
324 
325 static	struct T_info_ack udp_g_t_info_ack_ipv6 = {
326 	T_INFO_ACK,
327 	UDP_MAXPACKET_IPV6,	/* TSDU_size.  Excl. headers */
328 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
329 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
330 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
331 	sizeof (sin6_t), /* ADDR_size. */
332 	0,		/* OPT_size - not initialized here */
333 	UDP_MAXPACKET_IPV6,	/* TIDU_size. Excl. headers */
334 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
335 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
336 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
337 };
338 
339 /* largest UDP port number */
340 #define	UDP_MAX_PORT	65535
341 
342 /*
343  * Table of ND variables supported by udp.  These are loaded into us_nd
344  * in udp_open.
345  * All of these are alterable, within the min/max values given, at run time.
346  */
347 /* BEGIN CSTYLED */
348 udpparam_t udp_param_arr[] = {
349  /*min		max		value		name */
350  { 0L,		256,		32,		"udp_wroff_extra" },
351  { 1L,		255,		255,		"udp_ipv4_ttl" },
352  { 0,		IPV6_MAX_HOPS,	IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"},
353  { 1024,	(32 * 1024),	1024,		"udp_smallest_nonpriv_port" },
354  { 0,		1,		1,		"udp_do_checksum" },
355  { 1024,	UDP_MAX_PORT,	(32 * 1024),	"udp_smallest_anon_port" },
356  { 1024,	UDP_MAX_PORT,	UDP_MAX_PORT,	"udp_largest_anon_port" },
357  { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER,	"udp_xmit_hiwat"},
358  { 0,		     (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"},
359  { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER,	"udp_recv_hiwat"},
360  { 65536,	(1<<30),	2*1024*1024,	"udp_max_buf"},
361  { 100,		60000,		1000,		"udp_ndd_get_info_interval"},
362 };
363 /* END CSTYLED */
364 
365 /* Setable in /etc/system */
366 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */
367 uint32_t udp_random_anon_port = 1;
368 
369 /*
370  * Hook functions to enable cluster networking.
371  * On non-clustered systems these vectors must always be NULL
372  */
373 
374 void (*cl_inet_bind)(uchar_t protocol, sa_family_t addr_family,
375     uint8_t *laddrp, in_port_t lport) = NULL;
376 void (*cl_inet_unbind)(uint8_t protocol, sa_family_t addr_family,
377     uint8_t *laddrp, in_port_t lport) = NULL;
378 
379 typedef union T_primitives *t_primp_t;
380 
381 /*
382  * Return the next anonymous port in the privileged port range for
383  * bind checking.
384  *
385  * Trusted Extension (TX) notes: TX allows administrator to mark or
386  * reserve ports as Multilevel ports (MLP). MLP has special function
387  * on TX systems. Once a port is made MLP, it's not available as
388  * ordinary port. This creates "holes" in the port name space. It
389  * may be necessary to skip the "holes" find a suitable anon port.
390  */
391 static in_port_t
392 udp_get_next_priv_port(udp_t *udp)
393 {
394 	static in_port_t next_priv_port = IPPORT_RESERVED - 1;
395 	in_port_t nextport;
396 	boolean_t restart = B_FALSE;
397 	udp_stack_t *us = udp->udp_us;
398 
399 retry:
400 	if (next_priv_port < us->us_min_anonpriv_port ||
401 	    next_priv_port >= IPPORT_RESERVED) {
402 		next_priv_port = IPPORT_RESERVED - 1;
403 		if (restart)
404 			return (0);
405 		restart = B_TRUE;
406 	}
407 
408 	if (is_system_labeled() &&
409 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
410 	    next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) {
411 		next_priv_port = nextport;
412 		goto retry;
413 	}
414 
415 	return (next_priv_port--);
416 }
417 
418 /* UDP bind hash report triggered via the Named Dispatch mechanism. */
419 /* ARGSUSED */
420 static int
421 udp_bind_hash_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
422 {
423 	udp_fanout_t	*udpf;
424 	int		i;
425 	zoneid_t	zoneid;
426 	conn_t		*connp;
427 	udp_t		*udp;
428 	udp_stack_t	*us;
429 
430 	connp = Q_TO_CONN(q);
431 	udp = connp->conn_udp;
432 	us = udp->udp_us;
433 
434 	/* Refer to comments in udp_status_report(). */
435 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
436 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
437 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
438 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
439 			return (0);
440 		}
441 	}
442 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
443 		/* The following may work even if we cannot get a large buf. */
444 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
445 		return (0);
446 	}
447 
448 	(void) mi_mpprintf(mp,
449 	    "UDP     " MI_COL_HDRPAD_STR
450 	/*   12345678[89ABCDEF] */
451 	    " zone lport src addr        dest addr       port  state");
452 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
453 
454 	zoneid = connp->conn_zoneid;
455 
456 	for (i = 0; i < us->us_bind_fanout_size; i++) {
457 		udpf = &us->us_bind_fanout[i];
458 		mutex_enter(&udpf->uf_lock);
459 
460 		/* Print the hash index. */
461 		udp = udpf->uf_udp;
462 		if (zoneid != GLOBAL_ZONEID) {
463 			/* skip to first entry in this zone; might be none */
464 			while (udp != NULL &&
465 			    udp->udp_connp->conn_zoneid != zoneid)
466 				udp = udp->udp_bind_hash;
467 		}
468 		if (udp != NULL) {
469 			uint_t print_len, buf_len;
470 
471 			buf_len = mp->b_cont->b_datap->db_lim -
472 			    mp->b_cont->b_wptr;
473 			print_len = snprintf((char *)mp->b_cont->b_wptr,
474 			    buf_len, "%d\n", i);
475 			if (print_len < buf_len) {
476 				mp->b_cont->b_wptr += print_len;
477 			} else {
478 				mp->b_cont->b_wptr += buf_len;
479 			}
480 			for (; udp != NULL; udp = udp->udp_bind_hash) {
481 				if (zoneid == GLOBAL_ZONEID ||
482 				    zoneid == udp->udp_connp->conn_zoneid)
483 					udp_report_item(mp->b_cont, udp);
484 			}
485 		}
486 		mutex_exit(&udpf->uf_lock);
487 	}
488 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
489 	return (0);
490 }
491 
492 /*
493  * Hash list removal routine for udp_t structures.
494  */
495 static void
496 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock)
497 {
498 	udp_t	*udpnext;
499 	kmutex_t *lockp;
500 	udp_stack_t *us = udp->udp_us;
501 
502 	if (udp->udp_ptpbhn == NULL)
503 		return;
504 
505 	/*
506 	 * Extract the lock pointer in case there are concurrent
507 	 * hash_remove's for this instance.
508 	 */
509 	ASSERT(udp->udp_port != 0);
510 	if (!caller_holds_lock) {
511 		lockp = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
512 		    us->us_bind_fanout_size)].uf_lock;
513 		ASSERT(lockp != NULL);
514 		mutex_enter(lockp);
515 	}
516 	if (udp->udp_ptpbhn != NULL) {
517 		udpnext = udp->udp_bind_hash;
518 		if (udpnext != NULL) {
519 			udpnext->udp_ptpbhn = udp->udp_ptpbhn;
520 			udp->udp_bind_hash = NULL;
521 		}
522 		*udp->udp_ptpbhn = udpnext;
523 		udp->udp_ptpbhn = NULL;
524 	}
525 	if (!caller_holds_lock) {
526 		mutex_exit(lockp);
527 	}
528 }
529 
530 static void
531 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp)
532 {
533 	udp_t	**udpp;
534 	udp_t	*udpnext;
535 
536 	ASSERT(MUTEX_HELD(&uf->uf_lock));
537 	ASSERT(udp->udp_ptpbhn == NULL);
538 	udpp = &uf->uf_udp;
539 	udpnext = udpp[0];
540 	if (udpnext != NULL) {
541 		/*
542 		 * If the new udp bound to the INADDR_ANY address
543 		 * and the first one in the list is not bound to
544 		 * INADDR_ANY we skip all entries until we find the
545 		 * first one bound to INADDR_ANY.
546 		 * This makes sure that applications binding to a
547 		 * specific address get preference over those binding to
548 		 * INADDR_ANY.
549 		 */
550 		if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) &&
551 		    !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) {
552 			while ((udpnext = udpp[0]) != NULL &&
553 			    !V6_OR_V4_INADDR_ANY(
554 			    udpnext->udp_bound_v6src)) {
555 				udpp = &(udpnext->udp_bind_hash);
556 			}
557 			if (udpnext != NULL)
558 				udpnext->udp_ptpbhn = &udp->udp_bind_hash;
559 		} else {
560 			udpnext->udp_ptpbhn = &udp->udp_bind_hash;
561 		}
562 	}
563 	udp->udp_bind_hash = udpnext;
564 	udp->udp_ptpbhn = udpp;
565 	udpp[0] = udp;
566 }
567 
568 /*
569  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
570  * passed to udp_wput.
571  * It associates a port number and local address with the stream.
572  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP
573  * protocol type (IPPROTO_UDP) placed in the message following the address.
574  * A T_BIND_ACK message is passed upstream when ip acknowledges the request.
575  * (Called as writer.)
576  *
577  * Note that UDP over IPv4 and IPv6 sockets can use the same port number
578  * without setting SO_REUSEADDR. This is needed so that they
579  * can be viewed as two independent transport protocols.
580  * However, anonymouns ports are allocated from the same range to avoid
581  * duplicating the us->us_next_port_to_try.
582  */
583 static void
584 udp_tpi_bind(queue_t *q, mblk_t *mp)
585 {
586 	sin_t		*sin;
587 	sin6_t		*sin6;
588 	mblk_t		*mp1;
589 	struct T_bind_req *tbr;
590 	conn_t		*connp;
591 	udp_t		*udp;
592 	int		error;
593 	struct sockaddr	*sa;
594 
595 	connp = Q_TO_CONN(q);
596 	udp = connp->conn_udp;
597 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
598 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
599 		    "udp_bind: bad req, len %u",
600 		    (uint_t)(mp->b_wptr - mp->b_rptr));
601 		udp_err_ack(q, mp, TPROTO, 0);
602 		return;
603 	}
604 	if (udp->udp_state != TS_UNBND) {
605 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
606 		    "udp_bind: bad state, %u", udp->udp_state);
607 		udp_err_ack(q, mp, TOUTSTATE, 0);
608 		return;
609 	}
610 	/*
611 	 * Reallocate the message to make sure we have enough room for an
612 	 * address and the protocol type.
613 	 */
614 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
615 	if (!mp1) {
616 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
617 		return;
618 	}
619 
620 	mp = mp1;
621 
622 	/* Reset the message type in preparation for shipping it back. */
623 	DB_TYPE(mp) = M_PCPROTO;
624 
625 	tbr = (struct T_bind_req *)mp->b_rptr;
626 	switch (tbr->ADDR_length) {
627 	case 0:			/* Request for a generic port */
628 		tbr->ADDR_offset = sizeof (struct T_bind_req);
629 		if (udp->udp_family == AF_INET) {
630 			tbr->ADDR_length = sizeof (sin_t);
631 			sin = (sin_t *)&tbr[1];
632 			*sin = sin_null;
633 			sin->sin_family = AF_INET;
634 			mp->b_wptr = (uchar_t *)&sin[1];
635 			sa = (struct sockaddr *)sin;
636 		} else {
637 			ASSERT(udp->udp_family == AF_INET6);
638 			tbr->ADDR_length = sizeof (sin6_t);
639 			sin6 = (sin6_t *)&tbr[1];
640 			*sin6 = sin6_null;
641 			sin6->sin6_family = AF_INET6;
642 			mp->b_wptr = (uchar_t *)&sin6[1];
643 			sa = (struct sockaddr *)sin6;
644 		}
645 		break;
646 
647 	case sizeof (sin_t):	/* Complete IPv4 address */
648 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
649 		    sizeof (sin_t));
650 		if (sa == NULL || !OK_32PTR((char *)sa)) {
651 			udp_err_ack(q, mp, TSYSERR, EINVAL);
652 			return;
653 		}
654 		if (udp->udp_family != AF_INET ||
655 		    sa->sa_family != AF_INET) {
656 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
657 			return;
658 		}
659 		break;
660 
661 	case sizeof (sin6_t):	/* complete IPv6 address */
662 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
663 		    sizeof (sin6_t));
664 		if (sa == NULL || !OK_32PTR((char *)sa)) {
665 			udp_err_ack(q, mp, TSYSERR, EINVAL);
666 			return;
667 		}
668 		if (udp->udp_family != AF_INET6 ||
669 		    sa->sa_family != AF_INET6) {
670 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
671 			return;
672 		}
673 		break;
674 
675 	default:		/* Invalid request */
676 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
677 		    "udp_bind: bad ADDR_length length %u", tbr->ADDR_length);
678 		udp_err_ack(q, mp, TBADADDR, 0);
679 		return;
680 	}
681 
682 
683 	cred_t *cr = DB_CREDDEF(mp, connp->conn_cred);
684 	error = udp_do_bind(connp, sa, tbr->ADDR_length, cr,
685 	    tbr->PRIM_type != O_T_BIND_REQ);
686 
687 	if (error != 0) {
688 		if (error > 0) {
689 			udp_err_ack(q, mp, TSYSERR, error);
690 		} else {
691 			udp_err_ack(q, mp, -error, 0);
692 		}
693 	} else {
694 		tbr->PRIM_type = T_BIND_ACK;
695 		qreply(q, mp);
696 	}
697 }
698 
699 /*
700  * This routine handles each T_CONN_REQ message passed to udp.  It
701  * associates a default destination address with the stream.
702  *
703  * This routine sends down a T_BIND_REQ to IP with the following mblks:
704  *	T_BIND_REQ	- specifying local and remote address/port
705  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
706  *	T_OK_ACK	- for the T_CONN_REQ
707  *	T_CONN_CON	- to keep the TPI user happy
708  *
709  * The connect completes in udp_do_connect.
710  * When a T_BIND_ACK is received information is extracted from the IRE
711  * and the two appended messages are sent to the TPI user.
712  * Should udp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
713  * convert it to an error ack for the appropriate primitive.
714  */
715 static void
716 udp_tpi_connect(queue_t *q, mblk_t *mp)
717 {
718 	mblk_t	*mp1;
719 	udp_t	*udp;
720 	conn_t	*connp = Q_TO_CONN(q);
721 	int	error;
722 	socklen_t	len;
723 	struct sockaddr		*sa;
724 	struct T_conn_req	*tcr;
725 
726 	udp = connp->conn_udp;
727 	tcr = (struct T_conn_req *)mp->b_rptr;
728 
729 	/* A bit of sanity checking */
730 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
731 		udp_err_ack(q, mp, TPROTO, 0);
732 		return;
733 	}
734 
735 	if (tcr->OPT_length != 0) {
736 		udp_err_ack(q, mp, TBADOPT, 0);
737 		return;
738 	}
739 
740 	/*
741 	 * Determine packet type based on type of address passed in
742 	 * the request should contain an IPv4 or IPv6 address.
743 	 * Make sure that address family matches the type of
744 	 * family of the the address passed down
745 	 */
746 	len = tcr->DEST_length;
747 	switch (tcr->DEST_length) {
748 	default:
749 		udp_err_ack(q, mp, TBADADDR, 0);
750 		return;
751 
752 	case sizeof (sin_t):
753 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
754 		    sizeof (sin_t));
755 		break;
756 
757 	case sizeof (sin6_t):
758 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
759 		    sizeof (sin6_t));
760 		break;
761 	}
762 
763 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
764 	if (error != 0) {
765 		udp_err_ack(q, mp, TSYSERR, error);
766 		return;
767 	}
768 
769 	/*
770 	 * We have to send a connection confirmation to
771 	 * keep TLI happy.
772 	 */
773 	if (udp->udp_family == AF_INET) {
774 		mp1 = mi_tpi_conn_con(NULL, (char *)sa,
775 		    sizeof (sin_t), NULL, 0);
776 	} else {
777 		mp1 = mi_tpi_conn_con(NULL, (char *)sa,
778 		    sizeof (sin6_t), NULL, 0);
779 	}
780 	if (mp1 == NULL) {
781 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
782 		return;
783 	}
784 
785 	/*
786 	 * ok_ack for T_CONN_REQ
787 	 */
788 	mp = mi_tpi_ok_ack_alloc(mp);
789 	if (mp == NULL) {
790 		/* Unable to reuse the T_CONN_REQ for the ack. */
791 		freemsg(mp1);
792 		udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
793 		return;
794 	}
795 
796 	error = udp_do_connect(connp, sa, len);
797 	if (error != 0) {
798 		freeb(mp1);
799 		if (error < 0)
800 			udp_err_ack(q, mp, -error, 0);
801 		else
802 			udp_err_ack(q, mp, TSYSERR, error);
803 	} else {
804 		putnext(connp->conn_rq, mp);
805 		putnext(connp->conn_rq, mp1);
806 	}
807 }
808 
809 static int
810 udp_tpi_close(queue_t *q, int flags)
811 {
812 	conn_t	*connp;
813 
814 	if (flags & SO_FALLBACK) {
815 		/*
816 		 * stream is being closed while in fallback
817 		 * simply free the resources that were allocated
818 		 */
819 		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
820 		qprocsoff(q);
821 		goto done;
822 	}
823 
824 	connp = Q_TO_CONN(q);
825 	udp_do_close(connp);
826 done:
827 	q->q_ptr = WR(q)->q_ptr = NULL;
828 	return (0);
829 }
830 
831 /*
832  * Called in the close path to quiesce the conn
833  */
834 void
835 udp_quiesce_conn(conn_t *connp)
836 {
837 	udp_t	*udp = connp->conn_udp;
838 
839 	if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) {
840 		/*
841 		 * Running in cluster mode - register unbind information
842 		 */
843 		if (udp->udp_ipversion == IPV4_VERSION) {
844 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
845 			    (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))),
846 			    (in_port_t)udp->udp_port);
847 		} else {
848 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
849 			    (uint8_t *)(&(udp->udp_v6src)),
850 			    (in_port_t)udp->udp_port);
851 		}
852 	}
853 
854 	udp_bind_hash_remove(udp, B_FALSE);
855 
856 }
857 
858 void
859 udp_close_free(conn_t *connp)
860 {
861 	udp_t *udp = connp->conn_udp;
862 
863 	/* If there are any options associated with the stream, free them. */
864 	if (udp->udp_ip_snd_options != NULL) {
865 		mi_free((char *)udp->udp_ip_snd_options);
866 		udp->udp_ip_snd_options = NULL;
867 		udp->udp_ip_snd_options_len = 0;
868 	}
869 
870 	if (udp->udp_ip_rcv_options != NULL) {
871 		mi_free((char *)udp->udp_ip_rcv_options);
872 		udp->udp_ip_rcv_options = NULL;
873 		udp->udp_ip_rcv_options_len = 0;
874 	}
875 
876 	/* Free memory associated with sticky options */
877 	if (udp->udp_sticky_hdrs_len != 0) {
878 		kmem_free(udp->udp_sticky_hdrs,
879 		    udp->udp_sticky_hdrs_len);
880 		udp->udp_sticky_hdrs = NULL;
881 		udp->udp_sticky_hdrs_len = 0;
882 	}
883 
884 	ip6_pkt_free(&udp->udp_sticky_ipp);
885 
886 	/*
887 	 * Clear any fields which the kmem_cache constructor clears.
888 	 * Only udp_connp needs to be preserved.
889 	 * TBD: We should make this more efficient to avoid clearing
890 	 * everything.
891 	 */
892 	ASSERT(udp->udp_connp == connp);
893 	bzero(udp, sizeof (udp_t));
894 	udp->udp_connp = connp;
895 }
896 
897 static int
898 udp_do_disconnect(conn_t *connp)
899 {
900 	udp_t	*udp;
901 	mblk_t	*ire_mp;
902 	udp_fanout_t *udpf;
903 	udp_stack_t *us;
904 	int	error;
905 
906 	udp = connp->conn_udp;
907 	us = udp->udp_us;
908 	rw_enter(&udp->udp_rwlock, RW_WRITER);
909 	if (udp->udp_state != TS_DATA_XFER || udp->udp_pending_op != -1) {
910 		rw_exit(&udp->udp_rwlock);
911 		return (-TOUTSTATE);
912 	}
913 	udp->udp_pending_op = T_DISCON_REQ;
914 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
915 	    us->us_bind_fanout_size)];
916 	mutex_enter(&udpf->uf_lock);
917 	udp->udp_v6src = udp->udp_bound_v6src;
918 	udp->udp_state = TS_IDLE;
919 	mutex_exit(&udpf->uf_lock);
920 
921 	if (udp->udp_family == AF_INET6) {
922 		/* Rebuild the header template */
923 		error = udp_build_hdrs(udp);
924 		if (error != 0) {
925 			udp->udp_pending_op = -1;
926 			rw_exit(&udp->udp_rwlock);
927 			return (error);
928 		}
929 	}
930 
931 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
932 	if (ire_mp == NULL) {
933 		mutex_enter(&udpf->uf_lock);
934 		udp->udp_pending_op = -1;
935 		mutex_exit(&udpf->uf_lock);
936 		rw_exit(&udp->udp_rwlock);
937 		return (ENOMEM);
938 	}
939 
940 	rw_exit(&udp->udp_rwlock);
941 
942 	if (udp->udp_family == AF_INET6) {
943 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, IPPROTO_UDP,
944 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
945 	} else {
946 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, IPPROTO_UDP,
947 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port, B_TRUE);
948 	}
949 
950 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
951 }
952 
953 
954 static void
955 udp_tpi_disconnect(queue_t *q, mblk_t *mp)
956 {
957 	conn_t	*connp = Q_TO_CONN(q);
958 	int	error;
959 
960 	/*
961 	 * Allocate the largest primitive we need to send back
962 	 * T_error_ack is > than T_ok_ack
963 	 */
964 	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
965 	if (mp == NULL) {
966 		/* Unable to reuse the T_DISCON_REQ for the ack. */
967 		udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
968 		return;
969 	}
970 
971 	error = udp_do_disconnect(connp);
972 
973 	if (error != 0) {
974 		if (error < 0) {
975 			udp_err_ack(q, mp, -error, 0);
976 		} else {
977 			udp_err_ack(q, mp, TSYSERR, error);
978 		}
979 	} else {
980 		mp = mi_tpi_ok_ack_alloc(mp);
981 		ASSERT(mp != NULL);
982 		qreply(q, mp);
983 	}
984 }
985 
986 int
987 udp_disconnect(conn_t *connp)
988 {
989 	int error;
990 	udp_t *udp = connp->conn_udp;
991 
992 	udp->udp_dgram_errind = B_FALSE;
993 
994 	error = udp_do_disconnect(connp);
995 
996 	if (error < 0)
997 		error = proto_tlitosyserr(-error);
998 
999 	return (error);
1000 }
1001 
1002 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
1003 static void
1004 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
1005 {
1006 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1007 		qreply(q, mp);
1008 }
1009 
1010 /* Shorthand to generate and send TPI error acks to our client */
1011 static void
1012 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error,
1013     int sys_error)
1014 {
1015 	struct T_error_ack	*teackp;
1016 
1017 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1018 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1019 		teackp = (struct T_error_ack *)mp->b_rptr;
1020 		teackp->ERROR_prim = primitive;
1021 		teackp->TLI_error = t_error;
1022 		teackp->UNIX_error = sys_error;
1023 		qreply(q, mp);
1024 	}
1025 }
1026 
1027 /*ARGSUSED*/
1028 static int
1029 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
1030 {
1031 	int i;
1032 	udp_t		*udp = Q_TO_UDP(q);
1033 	udp_stack_t *us = udp->udp_us;
1034 
1035 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1036 		if (us->us_epriv_ports[i] != 0)
1037 			(void) mi_mpprintf(mp, "%d ", us->us_epriv_ports[i]);
1038 	}
1039 	return (0);
1040 }
1041 
1042 /* ARGSUSED */
1043 static int
1044 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1045     cred_t *cr)
1046 {
1047 	long	new_value;
1048 	int	i;
1049 	udp_t		*udp = Q_TO_UDP(q);
1050 	udp_stack_t *us = udp->udp_us;
1051 
1052 	/*
1053 	 * Fail the request if the new value does not lie within the
1054 	 * port number limits.
1055 	 */
1056 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1057 	    new_value <= 0 || new_value >= 65536) {
1058 		return (EINVAL);
1059 	}
1060 
1061 	/* Check if the value is already in the list */
1062 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1063 		if (new_value == us->us_epriv_ports[i]) {
1064 			return (EEXIST);
1065 		}
1066 	}
1067 	/* Find an empty slot */
1068 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1069 		if (us->us_epriv_ports[i] == 0)
1070 			break;
1071 	}
1072 	if (i == us->us_num_epriv_ports) {
1073 		return (EOVERFLOW);
1074 	}
1075 
1076 	/* Set the new value */
1077 	us->us_epriv_ports[i] = (in_port_t)new_value;
1078 	return (0);
1079 }
1080 
1081 /* ARGSUSED */
1082 static int
1083 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1084     cred_t *cr)
1085 {
1086 	long	new_value;
1087 	int	i;
1088 	udp_t		*udp = Q_TO_UDP(q);
1089 	udp_stack_t *us = udp->udp_us;
1090 
1091 	/*
1092 	 * Fail the request if the new value does not lie within the
1093 	 * port number limits.
1094 	 */
1095 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1096 	    new_value <= 0 || new_value >= 65536) {
1097 		return (EINVAL);
1098 	}
1099 
1100 	/* Check that the value is already in the list */
1101 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1102 		if (us->us_epriv_ports[i] == new_value)
1103 			break;
1104 	}
1105 	if (i == us->us_num_epriv_ports) {
1106 		return (ESRCH);
1107 	}
1108 
1109 	/* Clear the value */
1110 	us->us_epriv_ports[i] = 0;
1111 	return (0);
1112 }
1113 
1114 /* At minimum we need 4 bytes of UDP header */
1115 #define	ICMP_MIN_UDP_HDR	4
1116 
1117 /*
1118  * udp_icmp_error is called by udp_input to process ICMP msgs. passed up by IP.
1119  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1120  * Assumes that IP has pulled up everything up to and including the ICMP header.
1121  */
1122 static void
1123 udp_icmp_error(conn_t *connp, mblk_t *mp)
1124 			    {
1125 	icmph_t *icmph;
1126 	ipha_t	*ipha;
1127 	int	iph_hdr_length;
1128 	udpha_t	*udpha;
1129 	sin_t	sin;
1130 	sin6_t	sin6;
1131 	mblk_t	*mp1;
1132 	int	error = 0;
1133 	udp_t	*udp = connp->conn_udp;
1134 
1135 	mp1 = NULL;
1136 	ipha = (ipha_t *)mp->b_rptr;
1137 
1138 	ASSERT(OK_32PTR(mp->b_rptr));
1139 
1140 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1141 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1142 		udp_icmp_error_ipv6(connp, mp);
1143 		return;
1144 	}
1145 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
1146 
1147 	/* Skip past the outer IP and ICMP headers */
1148 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1149 	icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length];
1150 	ipha = (ipha_t *)&icmph[1];
1151 
1152 	/* Skip past the inner IP and find the ULP header */
1153 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1154 	udpha = (udpha_t *)((char *)ipha + iph_hdr_length);
1155 
1156 	switch (icmph->icmph_type) {
1157 	case ICMP_DEST_UNREACHABLE:
1158 		switch (icmph->icmph_code) {
1159 		case ICMP_FRAGMENTATION_NEEDED:
1160 			/*
1161 			 * IP has already adjusted the path MTU.
1162 			 */
1163 			break;
1164 		case ICMP_PORT_UNREACHABLE:
1165 		case ICMP_PROTOCOL_UNREACHABLE:
1166 			error = ECONNREFUSED;
1167 			break;
1168 		default:
1169 			/* Transient errors */
1170 			break;
1171 		}
1172 		break;
1173 	default:
1174 		/* Transient errors */
1175 		break;
1176 	}
1177 	if (error == 0) {
1178 		freemsg(mp);
1179 		return;
1180 	}
1181 
1182 	/*
1183 	 * Deliver T_UDERROR_IND when the application has asked for it.
1184 	 * The socket layer enables this automatically when connected.
1185 	 */
1186 	if (!udp->udp_dgram_errind) {
1187 		freemsg(mp);
1188 		return;
1189 	}
1190 
1191 
1192 	switch (udp->udp_family) {
1193 	case AF_INET:
1194 		sin = sin_null;
1195 		sin.sin_family = AF_INET;
1196 		sin.sin_addr.s_addr = ipha->ipha_dst;
1197 		sin.sin_port = udpha->uha_dst_port;
1198 		if (IPCL_IS_NONSTR(connp)) {
1199 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1200 			if (udp->udp_state == TS_DATA_XFER) {
1201 				if (sin.sin_port == udp->udp_dstport &&
1202 				    sin.sin_addr.s_addr ==
1203 				    V4_PART_OF_V6(udp->udp_v6dst)) {
1204 
1205 					rw_exit(&udp->udp_rwlock);
1206 					(*connp->conn_upcalls->su_set_error)
1207 					    (connp->conn_upper_handle, error);
1208 					goto done;
1209 				}
1210 			} else {
1211 				udp->udp_delayed_error = error;
1212 				*((sin_t *)&udp->udp_delayed_addr) = sin;
1213 			}
1214 			rw_exit(&udp->udp_rwlock);
1215 		} else {
1216 			mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t),
1217 			    NULL, 0, error);
1218 		}
1219 		break;
1220 	case AF_INET6:
1221 		sin6 = sin6_null;
1222 		sin6.sin6_family = AF_INET6;
1223 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
1224 		sin6.sin6_port = udpha->uha_dst_port;
1225 		if (IPCL_IS_NONSTR(connp)) {
1226 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1227 			if (udp->udp_state == TS_DATA_XFER) {
1228 				if (sin6.sin6_port == udp->udp_dstport &&
1229 				    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1230 				    &udp->udp_v6dst)) {
1231 					rw_exit(&udp->udp_rwlock);
1232 					(*connp->conn_upcalls->su_set_error)
1233 					    (connp->conn_upper_handle, error);
1234 					goto done;
1235 				}
1236 			} else {
1237 				udp->udp_delayed_error = error;
1238 				*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1239 			}
1240 			rw_exit(&udp->udp_rwlock);
1241 		} else {
1242 
1243 			mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1244 			    NULL, 0, error);
1245 		}
1246 		break;
1247 	}
1248 	if (mp1 != NULL)
1249 		putnext(connp->conn_rq, mp1);
1250 done:
1251 	freemsg(mp);
1252 }
1253 
1254 /*
1255  * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6.
1256  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1257  * Assumes that IP has pulled up all the extension headers as well as the
1258  * ICMPv6 header.
1259  */
1260 static void
1261 udp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
1262 {
1263 	icmp6_t		*icmp6;
1264 	ip6_t		*ip6h, *outer_ip6h;
1265 	uint16_t	iph_hdr_length;
1266 	uint8_t		*nexthdrp;
1267 	udpha_t		*udpha;
1268 	sin6_t		sin6;
1269 	mblk_t		*mp1;
1270 	int		error = 0;
1271 	udp_t		*udp = connp->conn_udp;
1272 	udp_stack_t	*us = udp->udp_us;
1273 
1274 	outer_ip6h = (ip6_t *)mp->b_rptr;
1275 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1276 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1277 	else
1278 		iph_hdr_length = IPV6_HDR_LEN;
1279 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1280 	ip6h = (ip6_t *)&icmp6[1];
1281 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1282 		freemsg(mp);
1283 		return;
1284 	}
1285 	udpha = (udpha_t *)((char *)ip6h + iph_hdr_length);
1286 
1287 	switch (icmp6->icmp6_type) {
1288 	case ICMP6_DST_UNREACH:
1289 		switch (icmp6->icmp6_code) {
1290 		case ICMP6_DST_UNREACH_NOPORT:
1291 			error = ECONNREFUSED;
1292 			break;
1293 		case ICMP6_DST_UNREACH_ADMIN:
1294 		case ICMP6_DST_UNREACH_NOROUTE:
1295 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1296 		case ICMP6_DST_UNREACH_ADDR:
1297 			/* Transient errors */
1298 			break;
1299 		default:
1300 			break;
1301 		}
1302 		break;
1303 	case ICMP6_PACKET_TOO_BIG: {
1304 		struct T_unitdata_ind	*tudi;
1305 		struct T_opthdr		*toh;
1306 		size_t			udi_size;
1307 		mblk_t			*newmp;
1308 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1309 		    sizeof (struct ip6_mtuinfo);
1310 		sin6_t			*sin6;
1311 		struct ip6_mtuinfo	*mtuinfo;
1312 
1313 		/*
1314 		 * If the application has requested to receive path mtu
1315 		 * information, send up an empty message containing an
1316 		 * IPV6_PATHMTU ancillary data item.
1317 		 */
1318 		if (!udp->udp_ipv6_recvpathmtu)
1319 			break;
1320 
1321 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1322 		    opt_length;
1323 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1324 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
1325 			break;
1326 		}
1327 
1328 		/*
1329 		 * newmp->b_cont is left to NULL on purpose.  This is an
1330 		 * empty message containing only ancillary data.
1331 		 */
1332 		newmp->b_datap->db_type = M_PROTO;
1333 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1334 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1335 		tudi->PRIM_type = T_UNITDATA_IND;
1336 		tudi->SRC_length = sizeof (sin6_t);
1337 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1338 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1339 		tudi->OPT_length = opt_length;
1340 
1341 		sin6 = (sin6_t *)&tudi[1];
1342 		bzero(sin6, sizeof (sin6_t));
1343 		sin6->sin6_family = AF_INET6;
1344 		sin6->sin6_addr = udp->udp_v6dst;
1345 
1346 		toh = (struct T_opthdr *)&sin6[1];
1347 		toh->level = IPPROTO_IPV6;
1348 		toh->name = IPV6_PATHMTU;
1349 		toh->len = opt_length;
1350 		toh->status = 0;
1351 
1352 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1353 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1354 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1355 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1356 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1357 		/*
1358 		 * We've consumed everything we need from the original
1359 		 * message.  Free it, then send our empty message.
1360 		 */
1361 		freemsg(mp);
1362 		if (!IPCL_IS_NONSTR(connp)) {
1363 			putnext(connp->conn_rq, newmp);
1364 		} else {
1365 			(*connp->conn_upcalls->su_recv)
1366 			    (connp->conn_upper_handle, newmp, 0, 0, &error,
1367 			    NULL);
1368 		}
1369 		return;
1370 	}
1371 	case ICMP6_TIME_EXCEEDED:
1372 		/* Transient errors */
1373 		break;
1374 	case ICMP6_PARAM_PROB:
1375 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1376 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1377 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1378 		    (uchar_t *)nexthdrp) {
1379 			error = ECONNREFUSED;
1380 			break;
1381 		}
1382 		break;
1383 	}
1384 	if (error == 0) {
1385 		freemsg(mp);
1386 		return;
1387 	}
1388 
1389 	/*
1390 	 * Deliver T_UDERROR_IND when the application has asked for it.
1391 	 * The socket layer enables this automatically when connected.
1392 	 */
1393 	if (!udp->udp_dgram_errind) {
1394 		freemsg(mp);
1395 		return;
1396 	}
1397 
1398 	sin6 = sin6_null;
1399 	sin6.sin6_family = AF_INET6;
1400 	sin6.sin6_addr = ip6h->ip6_dst;
1401 	sin6.sin6_port = udpha->uha_dst_port;
1402 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1403 
1404 	if (IPCL_IS_NONSTR(connp)) {
1405 		rw_enter(&udp->udp_rwlock, RW_WRITER);
1406 		if (udp->udp_state == TS_DATA_XFER) {
1407 			if (sin6.sin6_port == udp->udp_dstport &&
1408 			    IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1409 			    &udp->udp_v6dst)) {
1410 				rw_exit(&udp->udp_rwlock);
1411 				(*connp->conn_upcalls->su_set_error)
1412 				    (connp->conn_upper_handle, error);
1413 				goto done;
1414 			}
1415 		} else {
1416 			udp->udp_delayed_error = error;
1417 			*((sin6_t *)&udp->udp_delayed_addr) = sin6;
1418 		}
1419 		rw_exit(&udp->udp_rwlock);
1420 	} else {
1421 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1422 		    NULL, 0, error);
1423 		if (mp1 != NULL)
1424 			putnext(connp->conn_rq, mp1);
1425 	}
1426 
1427 done:
1428 	freemsg(mp);
1429 }
1430 
1431 /*
1432  * This routine responds to T_ADDR_REQ messages.  It is called by udp_wput.
1433  * The local address is filled in if endpoint is bound. The remote address
1434  * is filled in if remote address has been precified ("connected endpoint")
1435  * (The concept of connected CLTS sockets is alien to published TPI
1436  *  but we support it anyway).
1437  */
1438 static void
1439 udp_addr_req(queue_t *q, mblk_t *mp)
1440 {
1441 	sin_t	*sin;
1442 	sin6_t	*sin6;
1443 	mblk_t	*ackmp;
1444 	struct T_addr_ack *taa;
1445 	udp_t	*udp = Q_TO_UDP(q);
1446 
1447 	/* Make it large enough for worst case */
1448 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1449 	    2 * sizeof (sin6_t), 1);
1450 	if (ackmp == NULL) {
1451 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1452 		return;
1453 	}
1454 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1455 
1456 	bzero(taa, sizeof (struct T_addr_ack));
1457 	ackmp->b_wptr = (uchar_t *)&taa[1];
1458 
1459 	taa->PRIM_type = T_ADDR_ACK;
1460 	ackmp->b_datap->db_type = M_PCPROTO;
1461 	rw_enter(&udp->udp_rwlock, RW_READER);
1462 	/*
1463 	 * Note: Following code assumes 32 bit alignment of basic
1464 	 * data structures like sin_t and struct T_addr_ack.
1465 	 */
1466 	if (udp->udp_state != TS_UNBND) {
1467 		/*
1468 		 * Fill in local address first
1469 		 */
1470 		taa->LOCADDR_offset = sizeof (*taa);
1471 		if (udp->udp_family == AF_INET) {
1472 			taa->LOCADDR_length = sizeof (sin_t);
1473 			sin = (sin_t *)&taa[1];
1474 			/* Fill zeroes and then initialize non-zero fields */
1475 			*sin = sin_null;
1476 			sin->sin_family = AF_INET;
1477 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
1478 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1479 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src,
1480 				    sin->sin_addr.s_addr);
1481 			} else {
1482 				/*
1483 				 * INADDR_ANY
1484 				 * udp_v6src is not set, we might be bound to
1485 				 * broadcast/multicast. Use udp_bound_v6src as
1486 				 * local address instead (that could
1487 				 * also still be INADDR_ANY)
1488 				 */
1489 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src,
1490 				    sin->sin_addr.s_addr);
1491 			}
1492 			sin->sin_port = udp->udp_port;
1493 			ackmp->b_wptr = (uchar_t *)&sin[1];
1494 			if (udp->udp_state == TS_DATA_XFER) {
1495 				/*
1496 				 * connected, fill remote address too
1497 				 */
1498 				taa->REMADDR_length = sizeof (sin_t);
1499 				/* assumed 32-bit alignment */
1500 				taa->REMADDR_offset = taa->LOCADDR_offset +
1501 				    taa->LOCADDR_length;
1502 
1503 				sin = (sin_t *)(ackmp->b_rptr +
1504 				    taa->REMADDR_offset);
1505 				/* initialize */
1506 				*sin = sin_null;
1507 				sin->sin_family = AF_INET;
1508 				sin->sin_addr.s_addr =
1509 				    V4_PART_OF_V6(udp->udp_v6dst);
1510 				sin->sin_port = udp->udp_dstport;
1511 				ackmp->b_wptr = (uchar_t *)&sin[1];
1512 			}
1513 		} else {
1514 			taa->LOCADDR_length = sizeof (sin6_t);
1515 			sin6 = (sin6_t *)&taa[1];
1516 			/* Fill zeroes and then initialize non-zero fields */
1517 			*sin6 = sin6_null;
1518 			sin6->sin6_family = AF_INET6;
1519 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
1520 				sin6->sin6_addr = udp->udp_v6src;
1521 			} else {
1522 				/*
1523 				 * UNSPECIFIED
1524 				 * udp_v6src is not set, we might be bound to
1525 				 * broadcast/multicast. Use udp_bound_v6src as
1526 				 * local address instead (that could
1527 				 * also still be UNSPECIFIED)
1528 				 */
1529 				sin6->sin6_addr =
1530 				    udp->udp_bound_v6src;
1531 			}
1532 			sin6->sin6_port = udp->udp_port;
1533 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1534 			if (udp->udp_state == TS_DATA_XFER) {
1535 				/*
1536 				 * connected, fill remote address too
1537 				 */
1538 				taa->REMADDR_length = sizeof (sin6_t);
1539 				/* assumed 32-bit alignment */
1540 				taa->REMADDR_offset = taa->LOCADDR_offset +
1541 				    taa->LOCADDR_length;
1542 
1543 				sin6 = (sin6_t *)(ackmp->b_rptr +
1544 				    taa->REMADDR_offset);
1545 				/* initialize */
1546 				*sin6 = sin6_null;
1547 				sin6->sin6_family = AF_INET6;
1548 				sin6->sin6_addr = udp->udp_v6dst;
1549 				sin6->sin6_port =  udp->udp_dstport;
1550 				ackmp->b_wptr = (uchar_t *)&sin6[1];
1551 			}
1552 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1553 		}
1554 	}
1555 	rw_exit(&udp->udp_rwlock);
1556 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1557 	qreply(q, ackmp);
1558 }
1559 
1560 static void
1561 udp_copy_info(struct T_info_ack *tap, udp_t *udp)
1562 {
1563 	if (udp->udp_family == AF_INET) {
1564 		*tap = udp_g_t_info_ack_ipv4;
1565 	} else {
1566 		*tap = udp_g_t_info_ack_ipv6;
1567 	}
1568 	tap->CURRENT_state = udp->udp_state;
1569 	tap->OPT_size = udp_max_optsize;
1570 }
1571 
1572 static void
1573 udp_do_capability_ack(udp_t *udp, struct T_capability_ack *tcap,
1574     t_uscalar_t cap_bits1)
1575 {
1576 	tcap->CAP_bits1 = 0;
1577 
1578 	if (cap_bits1 & TC1_INFO) {
1579 		udp_copy_info(&tcap->INFO_ack, udp);
1580 		tcap->CAP_bits1 |= TC1_INFO;
1581 	}
1582 }
1583 
1584 /*
1585  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1586  * udp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1587  * udp_g_t_info_ack.  The current state of the stream is copied from
1588  * udp_state.
1589  */
1590 static void
1591 udp_capability_req(queue_t *q, mblk_t *mp)
1592 {
1593 	t_uscalar_t		cap_bits1;
1594 	struct T_capability_ack	*tcap;
1595 	udp_t	*udp = Q_TO_UDP(q);
1596 
1597 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1598 
1599 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1600 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1601 	if (!mp)
1602 		return;
1603 
1604 	tcap = (struct T_capability_ack *)mp->b_rptr;
1605 	udp_do_capability_ack(udp, tcap, cap_bits1);
1606 
1607 	qreply(q, mp);
1608 }
1609 
1610 /*
1611  * This routine responds to T_INFO_REQ messages.  It is called by udp_wput.
1612  * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack.
1613  * The current state of the stream is copied from udp_state.
1614  */
1615 static void
1616 udp_info_req(queue_t *q, mblk_t *mp)
1617 {
1618 	udp_t *udp = Q_TO_UDP(q);
1619 
1620 	/* Create a T_INFO_ACK message. */
1621 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1622 	    T_INFO_ACK);
1623 	if (!mp)
1624 		return;
1625 	udp_copy_info((struct T_info_ack *)mp->b_rptr, udp);
1626 	qreply(q, mp);
1627 }
1628 
1629 /*
1630  * IP recognizes seven kinds of bind requests:
1631  *
1632  * - A zero-length address binds only to the protocol number.
1633  *
1634  * - A 4-byte address is treated as a request to
1635  * validate that the address is a valid local IPv4
1636  * address, appropriate for an application to bind to.
1637  * IP does the verification, but does not make any note
1638  * of the address at this time.
1639  *
1640  * - A 16-byte address contains is treated as a request
1641  * to validate a local IPv6 address, as the 4-byte
1642  * address case above.
1643  *
1644  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
1645  * use it for the inbound fanout of packets.
1646  *
1647  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
1648  * use it for the inbound fanout of packets.
1649  *
1650  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
1651  * information consisting of local and remote addresses
1652  * and ports.  In this case, the addresses are both
1653  * validated as appropriate for this operation, and, if
1654  * so, the information is retained for use in the
1655  * inbound fanout.
1656  *
1657  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
1658  * fanout information, like the 12-byte case above.
1659  *
1660  * IP will also fill in the IRE request mblk with information
1661  * regarding our peer.  In all cases, we notify IP of our protocol
1662  * type by appending a single protocol byte to the bind request.
1663  */
1664 static mblk_t *
1665 udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, t_scalar_t addr_length)
1666 {
1667 	char	*cp;
1668 	mblk_t	*mp;
1669 	struct T_bind_req *tbr;
1670 	ipa_conn_t	*ac;
1671 	ipa6_conn_t	*ac6;
1672 	sin_t		*sin;
1673 	sin6_t		*sin6;
1674 
1675 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
1676 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
1677 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
1678 	if (!mp)
1679 		return (mp);
1680 	mp->b_datap->db_type = M_PROTO;
1681 	tbr = (struct T_bind_req *)mp->b_rptr;
1682 	tbr->PRIM_type = bind_prim;
1683 	tbr->ADDR_offset = sizeof (*tbr);
1684 	tbr->CONIND_number = 0;
1685 	tbr->ADDR_length = addr_length;
1686 	cp = (char *)&tbr[1];
1687 	switch (addr_length) {
1688 	case sizeof (ipa_conn_t):
1689 		ASSERT(udp->udp_family == AF_INET);
1690 		/* Append a request for an IRE */
1691 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1692 		if (!mp->b_cont) {
1693 			freemsg(mp);
1694 			return (NULL);
1695 		}
1696 		mp->b_cont->b_wptr += sizeof (ire_t);
1697 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1698 
1699 		/* cp known to be 32 bit aligned */
1700 		ac = (ipa_conn_t *)cp;
1701 		ac->ac_laddr = V4_PART_OF_V6(udp->udp_v6src);
1702 		ac->ac_faddr = V4_PART_OF_V6(udp->udp_v6dst);
1703 		ac->ac_fport = udp->udp_dstport;
1704 		ac->ac_lport = udp->udp_port;
1705 		break;
1706 
1707 	case sizeof (ipa6_conn_t):
1708 		ASSERT(udp->udp_family == AF_INET6);
1709 		/* Append a request for an IRE */
1710 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1711 		if (!mp->b_cont) {
1712 			freemsg(mp);
1713 			return (NULL);
1714 		}
1715 		mp->b_cont->b_wptr += sizeof (ire_t);
1716 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1717 
1718 		/* cp known to be 32 bit aligned */
1719 		ac6 = (ipa6_conn_t *)cp;
1720 		ac6->ac6_laddr = udp->udp_v6src;
1721 		ac6->ac6_faddr = udp->udp_v6dst;
1722 		ac6->ac6_fport = udp->udp_dstport;
1723 		ac6->ac6_lport = udp->udp_port;
1724 		break;
1725 
1726 	case sizeof (sin_t):
1727 		ASSERT(udp->udp_family == AF_INET);
1728 		/* Append a request for an IRE */
1729 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1730 		if (!mp->b_cont) {
1731 			freemsg(mp);
1732 			return (NULL);
1733 		}
1734 		mp->b_cont->b_wptr += sizeof (ire_t);
1735 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1736 
1737 		sin = (sin_t *)cp;
1738 		*sin = sin_null;
1739 		sin->sin_family = AF_INET;
1740 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_bound_v6src);
1741 		sin->sin_port = udp->udp_port;
1742 		break;
1743 
1744 	case sizeof (sin6_t):
1745 		ASSERT(udp->udp_family == AF_INET6);
1746 		/* Append a request for an IRE */
1747 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1748 		if (!mp->b_cont) {
1749 			freemsg(mp);
1750 			return (NULL);
1751 		}
1752 		mp->b_cont->b_wptr += sizeof (ire_t);
1753 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1754 
1755 		sin6 = (sin6_t *)cp;
1756 		*sin6 = sin6_null;
1757 		sin6->sin6_family = AF_INET6;
1758 		sin6->sin6_addr = udp->udp_bound_v6src;
1759 		sin6->sin6_port = udp->udp_port;
1760 		break;
1761 	}
1762 	/* Add protocol number to end */
1763 	cp[addr_length] = (char)IPPROTO_UDP;
1764 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
1765 	return (mp);
1766 }
1767 
1768 /* For /dev/udp aka AF_INET open */
1769 static int
1770 udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1771 {
1772 	return (udp_open(q, devp, flag, sflag, credp, B_FALSE));
1773 }
1774 
1775 /* For /dev/udp6 aka AF_INET6 open */
1776 static int
1777 udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1778 {
1779 	return (udp_open(q, devp, flag, sflag, credp, B_TRUE));
1780 }
1781 
1782 /*
1783  * This is the open routine for udp.  It allocates a udp_t structure for
1784  * the stream and, on the first open of the module, creates an ND table.
1785  */
1786 /*ARGSUSED2*/
1787 static int
1788 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1789     boolean_t isv6)
1790 {
1791 	int		error;
1792 	udp_t		*udp;
1793 	conn_t		*connp;
1794 	dev_t		conn_dev;
1795 	udp_stack_t	*us;
1796 	vmem_t		*minor_arena;
1797 
1798 	TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q);
1799 
1800 	/* If the stream is already open, return immediately. */
1801 	if (q->q_ptr != NULL)
1802 		return (0);
1803 
1804 	if (sflag == MODOPEN)
1805 		return (EINVAL);
1806 
1807 	if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
1808 	    ((conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
1809 		minor_arena = ip_minor_arena_la;
1810 	} else {
1811 		/*
1812 		 * Either minor numbers in the large arena were exhausted
1813 		 * or a non socket application is doing the open.
1814 		 * Try to allocate from the small arena.
1815 		 */
1816 		if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0)
1817 			return (EBUSY);
1818 
1819 		minor_arena = ip_minor_arena_sa;
1820 	}
1821 
1822 	if (flag & SO_FALLBACK) {
1823 		/*
1824 		 * Non streams socket needs a stream to fallback to
1825 		 */
1826 		RD(q)->q_ptr = (void *)conn_dev;
1827 		WR(q)->q_qinfo = &udp_fallback_sock_winit;
1828 		WR(q)->q_ptr = (void *)minor_arena;
1829 		qprocson(q);
1830 		return (0);
1831 	}
1832 
1833 	connp = udp_do_open(credp, isv6, KM_SLEEP);
1834 	if (connp == NULL) {
1835 		inet_minor_free(minor_arena, conn_dev);
1836 		return (ENOMEM);
1837 	}
1838 	udp = connp->conn_udp;
1839 	us = udp->udp_us;
1840 
1841 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1842 	connp->conn_dev = conn_dev;
1843 	connp->conn_minor_arena = minor_arena;
1844 
1845 	/*
1846 	 * Initialize the udp_t structure for this stream.
1847 	 */
1848 	q->q_ptr = connp;
1849 	WR(q)->q_ptr = connp;
1850 	connp->conn_rq = q;
1851 	connp->conn_wq = WR(q);
1852 
1853 	rw_enter(&udp->udp_rwlock, RW_WRITER);
1854 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
1855 	ASSERT(connp->conn_udp == udp);
1856 	ASSERT(udp->udp_connp == connp);
1857 
1858 	if (flag & SO_SOCKSTR) {
1859 		connp->conn_flags |= IPCL_SOCKET;
1860 		udp->udp_issocket = B_TRUE;
1861 		udp->udp_direct_sockfs = B_TRUE;
1862 	}
1863 
1864 	q->q_hiwat = us->us_recv_hiwat;
1865 	WR(q)->q_hiwat = us->us_xmit_hiwat;
1866 	WR(q)->q_lowat = us->us_xmit_lowat;
1867 
1868 	qprocson(q);
1869 
1870 	if (udp->udp_family == AF_INET6) {
1871 		/* Build initial header template for transmit */
1872 		if ((error = udp_build_hdrs(udp)) != 0) {
1873 			rw_exit(&udp->udp_rwlock);
1874 			qprocsoff(q);
1875 			inet_minor_free(minor_arena, conn_dev);
1876 			ipcl_conn_destroy(connp);
1877 			return (error);
1878 		}
1879 	}
1880 	rw_exit(&udp->udp_rwlock);
1881 
1882 	/* Set the Stream head write offset and high watermark. */
1883 	(void) proto_set_tx_wroff(q, connp,
1884 	    udp->udp_max_hdr_len + us->us_wroff_extra);
1885 	/* XXX udp_set_rcv_hiwat() doesn't hold the lock, is it a bug??? */
1886 	(void) proto_set_rx_hiwat(q, connp, udp_set_rcv_hiwat(udp, q->q_hiwat));
1887 
1888 	mutex_enter(&connp->conn_lock);
1889 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1890 	mutex_exit(&connp->conn_lock);
1891 	return (0);
1892 }
1893 
1894 /*
1895  * Which UDP options OK to set through T_UNITDATA_REQ...
1896  */
1897 /* ARGSUSED */
1898 static boolean_t
1899 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1900 {
1901 	return (B_TRUE);
1902 }
1903 
1904 /*
1905  * This routine gets default values of certain options whose default
1906  * values are maintained by protcol specific code
1907  */
1908 /* ARGSUSED */
1909 int
1910 udp_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
1911 {
1912 	udp_t		*udp = Q_TO_UDP(q);
1913 	udp_stack_t *us = udp->udp_us;
1914 	int *i1 = (int *)ptr;
1915 
1916 	switch (level) {
1917 	case IPPROTO_IP:
1918 		switch (name) {
1919 		case IP_MULTICAST_TTL:
1920 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1921 			return (sizeof (uchar_t));
1922 		case IP_MULTICAST_LOOP:
1923 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1924 			return (sizeof (uchar_t));
1925 		}
1926 		break;
1927 	case IPPROTO_IPV6:
1928 		switch (name) {
1929 		case IPV6_MULTICAST_HOPS:
1930 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1931 			return (sizeof (int));
1932 		case IPV6_MULTICAST_LOOP:
1933 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1934 			return (sizeof (int));
1935 		case IPV6_UNICAST_HOPS:
1936 			*i1 = us->us_ipv6_hoplimit;
1937 			return (sizeof (int));
1938 		}
1939 		break;
1940 	}
1941 	return (-1);
1942 }
1943 
1944 /*
1945  * This routine retrieves the current status of socket options.
1946  * It returns the size of the option retrieved.
1947  */
1948 static int
1949 udp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
1950 {
1951 	udp_t		*udp = connp->conn_udp;
1952 	udp_stack_t	*us = udp->udp_us;
1953 	int		*i1 = (int *)ptr;
1954 	ip6_pkt_t 	*ipp = &udp->udp_sticky_ipp;
1955 	int		len;
1956 
1957 	ASSERT(RW_READ_HELD(&udp->udp_rwlock));
1958 	switch (level) {
1959 	case SOL_SOCKET:
1960 		switch (name) {
1961 		case SO_DEBUG:
1962 			*i1 = udp->udp_debug;
1963 			break;	/* goto sizeof (int) option return */
1964 		case SO_REUSEADDR:
1965 			*i1 = udp->udp_reuseaddr;
1966 			break;	/* goto sizeof (int) option return */
1967 		case SO_TYPE:
1968 			*i1 = SOCK_DGRAM;
1969 			break;	/* goto sizeof (int) option return */
1970 
1971 		/*
1972 		 * The following three items are available here,
1973 		 * but are only meaningful to IP.
1974 		 */
1975 		case SO_DONTROUTE:
1976 			*i1 = udp->udp_dontroute;
1977 			break;	/* goto sizeof (int) option return */
1978 		case SO_USELOOPBACK:
1979 			*i1 = udp->udp_useloopback;
1980 			break;	/* goto sizeof (int) option return */
1981 		case SO_BROADCAST:
1982 			*i1 = udp->udp_broadcast;
1983 			break;	/* goto sizeof (int) option return */
1984 
1985 		case SO_SNDBUF:
1986 			*i1 = udp->udp_xmit_hiwat;
1987 			break;	/* goto sizeof (int) option return */
1988 		case SO_RCVBUF:
1989 			*i1 = udp->udp_rcv_disply_hiwat;
1990 			break;	/* goto sizeof (int) option return */
1991 		case SO_DGRAM_ERRIND:
1992 			*i1 = udp->udp_dgram_errind;
1993 			break;	/* goto sizeof (int) option return */
1994 		case SO_RECVUCRED:
1995 			*i1 = udp->udp_recvucred;
1996 			break;	/* goto sizeof (int) option return */
1997 		case SO_TIMESTAMP:
1998 			*i1 = udp->udp_timestamp;
1999 			break;	/* goto sizeof (int) option return */
2000 		case SO_ANON_MLP:
2001 			*i1 = connp->conn_anon_mlp;
2002 			break;	/* goto sizeof (int) option return */
2003 		case SO_MAC_EXEMPT:
2004 			*i1 = connp->conn_mac_exempt;
2005 			break;	/* goto sizeof (int) option return */
2006 		case SO_ALLZONES:
2007 			*i1 = connp->conn_allzones;
2008 			break;	/* goto sizeof (int) option return */
2009 		case SO_EXCLBIND:
2010 			*i1 = udp->udp_exclbind ? SO_EXCLBIND : 0;
2011 			break;
2012 		case SO_PROTOTYPE:
2013 			*i1 = IPPROTO_UDP;
2014 			break;
2015 		case SO_DOMAIN:
2016 			*i1 = udp->udp_family;
2017 			break;
2018 		default:
2019 			return (-1);
2020 		}
2021 		break;
2022 	case IPPROTO_IP:
2023 		if (udp->udp_family != AF_INET)
2024 			return (-1);
2025 		switch (name) {
2026 		case IP_OPTIONS:
2027 		case T_IP_OPTIONS:
2028 			len = udp->udp_ip_rcv_options_len - udp->udp_label_len;
2029 			if (len > 0) {
2030 				bcopy(udp->udp_ip_rcv_options +
2031 				    udp->udp_label_len, ptr, len);
2032 			}
2033 			return (len);
2034 		case IP_TOS:
2035 		case T_IP_TOS:
2036 			*i1 = (int)udp->udp_type_of_service;
2037 			break;	/* goto sizeof (int) option return */
2038 		case IP_TTL:
2039 			*i1 = (int)udp->udp_ttl;
2040 			break;	/* goto sizeof (int) option return */
2041 		case IP_DHCPINIT_IF:
2042 			return (-EINVAL);
2043 		case IP_NEXTHOP:
2044 		case IP_RECVPKTINFO:
2045 			/*
2046 			 * This also handles IP_PKTINFO.
2047 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2048 			 * Differentiation is based on the size of the argument
2049 			 * passed in.
2050 			 * This option is handled in IP which will return an
2051 			 * error for IP_PKTINFO as it's not supported as a
2052 			 * sticky option.
2053 			 */
2054 			return (-EINVAL);
2055 		case IP_MULTICAST_IF:
2056 			/* 0 address if not set */
2057 			*(ipaddr_t *)ptr = udp->udp_multicast_if_addr;
2058 			return (sizeof (ipaddr_t));
2059 		case IP_MULTICAST_TTL:
2060 			*(uchar_t *)ptr = udp->udp_multicast_ttl;
2061 			return (sizeof (uchar_t));
2062 		case IP_MULTICAST_LOOP:
2063 			*ptr = connp->conn_multicast_loop;
2064 			return (sizeof (uint8_t));
2065 		case IP_RECVOPTS:
2066 			*i1 = udp->udp_recvopts;
2067 			break;	/* goto sizeof (int) option return */
2068 		case IP_RECVDSTADDR:
2069 			*i1 = udp->udp_recvdstaddr;
2070 			break;	/* goto sizeof (int) option return */
2071 		case IP_RECVIF:
2072 			*i1 = udp->udp_recvif;
2073 			break;	/* goto sizeof (int) option return */
2074 		case IP_RECVSLLA:
2075 			*i1 = udp->udp_recvslla;
2076 			break;	/* goto sizeof (int) option return */
2077 		case IP_RECVTTL:
2078 			*i1 = udp->udp_recvttl;
2079 			break;	/* goto sizeof (int) option return */
2080 		case IP_ADD_MEMBERSHIP:
2081 		case IP_DROP_MEMBERSHIP:
2082 		case IP_BLOCK_SOURCE:
2083 		case IP_UNBLOCK_SOURCE:
2084 		case IP_ADD_SOURCE_MEMBERSHIP:
2085 		case IP_DROP_SOURCE_MEMBERSHIP:
2086 		case MCAST_JOIN_GROUP:
2087 		case MCAST_LEAVE_GROUP:
2088 		case MCAST_BLOCK_SOURCE:
2089 		case MCAST_UNBLOCK_SOURCE:
2090 		case MCAST_JOIN_SOURCE_GROUP:
2091 		case MCAST_LEAVE_SOURCE_GROUP:
2092 		case IP_DONTFAILOVER_IF:
2093 			/* cannot "get" the value for these */
2094 			return (-1);
2095 		case IP_BOUND_IF:
2096 			/* Zero if not set */
2097 			*i1 = udp->udp_bound_if;
2098 			break;	/* goto sizeof (int) option return */
2099 		case IP_UNSPEC_SRC:
2100 			*i1 = udp->udp_unspec_source;
2101 			break;	/* goto sizeof (int) option return */
2102 		case IP_BROADCAST_TTL:
2103 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
2104 			return (sizeof (uchar_t));
2105 		default:
2106 			return (-1);
2107 		}
2108 		break;
2109 	case IPPROTO_IPV6:
2110 		if (udp->udp_family != AF_INET6)
2111 			return (-1);
2112 		switch (name) {
2113 		case IPV6_UNICAST_HOPS:
2114 			*i1 = (unsigned int)udp->udp_ttl;
2115 			break;	/* goto sizeof (int) option return */
2116 		case IPV6_MULTICAST_IF:
2117 			/* 0 index if not set */
2118 			*i1 = udp->udp_multicast_if_index;
2119 			break;	/* goto sizeof (int) option return */
2120 		case IPV6_MULTICAST_HOPS:
2121 			*i1 = udp->udp_multicast_ttl;
2122 			break;	/* goto sizeof (int) option return */
2123 		case IPV6_MULTICAST_LOOP:
2124 			*i1 = connp->conn_multicast_loop;
2125 			break;	/* goto sizeof (int) option return */
2126 		case IPV6_JOIN_GROUP:
2127 		case IPV6_LEAVE_GROUP:
2128 		case MCAST_JOIN_GROUP:
2129 		case MCAST_LEAVE_GROUP:
2130 		case MCAST_BLOCK_SOURCE:
2131 		case MCAST_UNBLOCK_SOURCE:
2132 		case MCAST_JOIN_SOURCE_GROUP:
2133 		case MCAST_LEAVE_SOURCE_GROUP:
2134 			/* cannot "get" the value for these */
2135 			return (-1);
2136 		case IPV6_BOUND_IF:
2137 			/* Zero if not set */
2138 			*i1 = udp->udp_bound_if;
2139 			break;	/* goto sizeof (int) option return */
2140 		case IPV6_UNSPEC_SRC:
2141 			*i1 = udp->udp_unspec_source;
2142 			break;	/* goto sizeof (int) option return */
2143 		case IPV6_RECVPKTINFO:
2144 			*i1 = udp->udp_ip_recvpktinfo;
2145 			break;	/* goto sizeof (int) option return */
2146 		case IPV6_RECVTCLASS:
2147 			*i1 = udp->udp_ipv6_recvtclass;
2148 			break;	/* goto sizeof (int) option return */
2149 		case IPV6_RECVPATHMTU:
2150 			*i1 = udp->udp_ipv6_recvpathmtu;
2151 			break;	/* goto sizeof (int) option return */
2152 		case IPV6_RECVHOPLIMIT:
2153 			*i1 = udp->udp_ipv6_recvhoplimit;
2154 			break;	/* goto sizeof (int) option return */
2155 		case IPV6_RECVHOPOPTS:
2156 			*i1 = udp->udp_ipv6_recvhopopts;
2157 			break;	/* goto sizeof (int) option return */
2158 		case IPV6_RECVDSTOPTS:
2159 			*i1 = udp->udp_ipv6_recvdstopts;
2160 			break;	/* goto sizeof (int) option return */
2161 		case _OLD_IPV6_RECVDSTOPTS:
2162 			*i1 = udp->udp_old_ipv6_recvdstopts;
2163 			break;	/* goto sizeof (int) option return */
2164 		case IPV6_RECVRTHDRDSTOPTS:
2165 			*i1 = udp->udp_ipv6_recvrthdrdstopts;
2166 			break;	/* goto sizeof (int) option return */
2167 		case IPV6_RECVRTHDR:
2168 			*i1 = udp->udp_ipv6_recvrthdr;
2169 			break;	/* goto sizeof (int) option return */
2170 		case IPV6_PKTINFO: {
2171 			/* XXX assumes that caller has room for max size! */
2172 			struct in6_pktinfo *pkti;
2173 
2174 			pkti = (struct in6_pktinfo *)ptr;
2175 			if (ipp->ipp_fields & IPPF_IFINDEX)
2176 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
2177 			else
2178 				pkti->ipi6_ifindex = 0;
2179 			if (ipp->ipp_fields & IPPF_ADDR)
2180 				pkti->ipi6_addr = ipp->ipp_addr;
2181 			else
2182 				pkti->ipi6_addr = ipv6_all_zeros;
2183 			return (sizeof (struct in6_pktinfo));
2184 		}
2185 		case IPV6_TCLASS:
2186 			if (ipp->ipp_fields & IPPF_TCLASS)
2187 				*i1 = ipp->ipp_tclass;
2188 			else
2189 				*i1 = IPV6_FLOW_TCLASS(
2190 				    IPV6_DEFAULT_VERS_AND_FLOW);
2191 			break;	/* goto sizeof (int) option return */
2192 		case IPV6_NEXTHOP: {
2193 			sin6_t *sin6 = (sin6_t *)ptr;
2194 
2195 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2196 				return (0);
2197 			*sin6 = sin6_null;
2198 			sin6->sin6_family = AF_INET6;
2199 			sin6->sin6_addr = ipp->ipp_nexthop;
2200 			return (sizeof (sin6_t));
2201 		}
2202 		case IPV6_HOPOPTS:
2203 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2204 				return (0);
2205 			if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6)
2206 				return (0);
2207 			/*
2208 			 * The cipso/label option is added by kernel.
2209 			 * User is not usually aware of this option.
2210 			 * We copy out the hbh opt after the label option.
2211 			 */
2212 			bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6,
2213 			    ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2214 			if (udp->udp_label_len_v6 > 0) {
2215 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2216 				ptr[1] = (ipp->ipp_hopoptslen -
2217 				    udp->udp_label_len_v6 + 7) / 8 - 1;
2218 			}
2219 			return (ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2220 		case IPV6_RTHDRDSTOPTS:
2221 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2222 				return (0);
2223 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2224 			return (ipp->ipp_rtdstoptslen);
2225 		case IPV6_RTHDR:
2226 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2227 				return (0);
2228 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2229 			return (ipp->ipp_rthdrlen);
2230 		case IPV6_DSTOPTS:
2231 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
2232 				return (0);
2233 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2234 			return (ipp->ipp_dstoptslen);
2235 		case IPV6_PATHMTU:
2236 			return (ip_fill_mtuinfo(&udp->udp_v6dst,
2237 			    udp->udp_dstport, (struct ip6_mtuinfo *)ptr,
2238 			    us->us_netstack));
2239 		default:
2240 			return (-1);
2241 		}
2242 		break;
2243 	case IPPROTO_UDP:
2244 		switch (name) {
2245 		case UDP_ANONPRIVBIND:
2246 			*i1 = udp->udp_anon_priv_bind;
2247 			break;
2248 		case UDP_EXCLBIND:
2249 			*i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0;
2250 			break;
2251 		case UDP_RCVHDR:
2252 			*i1 = udp->udp_rcvhdr ? 1 : 0;
2253 			break;
2254 		case UDP_NAT_T_ENDPOINT:
2255 			*i1 = udp->udp_nat_t_endpoint;
2256 			break;
2257 		default:
2258 			return (-1);
2259 		}
2260 		break;
2261 	default:
2262 		return (-1);
2263 	}
2264 	return (sizeof (int));
2265 }
2266 
2267 int
2268 udp_tpi_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2269 {
2270 	udp_t   *udp;
2271 	int	err;
2272 
2273 	udp = Q_TO_UDP(q);
2274 
2275 	rw_enter(&udp->udp_rwlock, RW_READER);
2276 	err = udp_opt_get(Q_TO_CONN(q), level, name, ptr);
2277 	rw_exit(&udp->udp_rwlock);
2278 	return (err);
2279 }
2280 
2281 /*
2282  * This routine sets socket options.
2283  */
2284 /* ARGSUSED */
2285 static int
2286 udp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
2287     uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
2288     void *thisdg_attrs, boolean_t checkonly)
2289 {
2290 	udpattrs_t *attrs = thisdg_attrs;
2291 	int	*i1 = (int *)invalp;
2292 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2293 	udp_t	*udp = connp->conn_udp;
2294 	udp_stack_t	*us = udp->udp_us;
2295 	int	error;
2296 	uint_t	newlen;
2297 	size_t	sth_wroff;
2298 
2299 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
2300 	/*
2301 	 * For fixed length options, no sanity check
2302 	 * of passed in length is done. It is assumed *_optcom_req()
2303 	 * routines do the right thing.
2304 	 */
2305 	switch (level) {
2306 	case SOL_SOCKET:
2307 		switch (name) {
2308 		case SO_REUSEADDR:
2309 			if (!checkonly) {
2310 				udp->udp_reuseaddr = onoff;
2311 				PASS_OPT_TO_IP(connp);
2312 			}
2313 			break;
2314 		case SO_DEBUG:
2315 			if (!checkonly)
2316 				udp->udp_debug = onoff;
2317 			break;
2318 		/*
2319 		 * The following three items are available here,
2320 		 * but are only meaningful to IP.
2321 		 */
2322 		case SO_DONTROUTE:
2323 			if (!checkonly) {
2324 				udp->udp_dontroute = onoff;
2325 				PASS_OPT_TO_IP(connp);
2326 			}
2327 			break;
2328 		case SO_USELOOPBACK:
2329 			if (!checkonly) {
2330 				udp->udp_useloopback = onoff;
2331 				PASS_OPT_TO_IP(connp);
2332 			}
2333 			break;
2334 		case SO_BROADCAST:
2335 			if (!checkonly) {
2336 				udp->udp_broadcast = onoff;
2337 				PASS_OPT_TO_IP(connp);
2338 			}
2339 			break;
2340 
2341 		case SO_SNDBUF:
2342 			if (*i1 > us->us_max_buf) {
2343 				*outlenp = 0;
2344 				return (ENOBUFS);
2345 			}
2346 			if (!checkonly) {
2347 				udp->udp_xmit_hiwat = *i1;
2348 				connp->conn_wq->q_hiwat = *i1;
2349 			}
2350 			break;
2351 		case SO_RCVBUF:
2352 			if (*i1 > us->us_max_buf) {
2353 				*outlenp = 0;
2354 				return (ENOBUFS);
2355 			}
2356 			if (!checkonly) {
2357 				int size;
2358 
2359 				udp->udp_rcv_disply_hiwat = *i1;
2360 				size = udp_set_rcv_hiwat(udp, *i1);
2361 				rw_exit(&udp->udp_rwlock);
2362 				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
2363 				    size);
2364 				rw_enter(&udp->udp_rwlock, RW_WRITER);
2365 			}
2366 			break;
2367 		case SO_DGRAM_ERRIND:
2368 			if (!checkonly)
2369 				udp->udp_dgram_errind = onoff;
2370 			break;
2371 		case SO_RECVUCRED:
2372 			if (!checkonly)
2373 				udp->udp_recvucred = onoff;
2374 			break;
2375 		case SO_ALLZONES:
2376 			/*
2377 			 * "soft" error (negative)
2378 			 * option not handled at this level
2379 			 * Do not modify *outlenp.
2380 			 */
2381 			return (-EINVAL);
2382 		case SO_TIMESTAMP:
2383 			if (!checkonly)
2384 				udp->udp_timestamp = onoff;
2385 			break;
2386 		case SO_ANON_MLP:
2387 			if (!checkonly) {
2388 				connp->conn_anon_mlp = onoff;
2389 				PASS_OPT_TO_IP(connp);
2390 			}
2391 			break;
2392 		case SO_MAC_EXEMPT:
2393 			if (secpolicy_net_mac_aware(cr) != 0 ||
2394 			    udp->udp_state != TS_UNBND)
2395 				return (EACCES);
2396 			if (!checkonly) {
2397 				connp->conn_mac_exempt = onoff;
2398 				PASS_OPT_TO_IP(connp);
2399 			}
2400 			break;
2401 		case SCM_UCRED: {
2402 			struct ucred_s *ucr;
2403 			cred_t *cr, *newcr;
2404 			ts_label_t *tsl;
2405 
2406 			/*
2407 			 * Only sockets that have proper privileges and are
2408 			 * bound to MLPs will have any other value here, so
2409 			 * this implicitly tests for privilege to set label.
2410 			 */
2411 			if (connp->conn_mlp_type == mlptSingle)
2412 				break;
2413 			ucr = (struct ucred_s *)invalp;
2414 			if (inlen != ucredsize ||
2415 			    ucr->uc_labeloff < sizeof (*ucr) ||
2416 			    ucr->uc_labeloff + sizeof (bslabel_t) > inlen)
2417 				return (EINVAL);
2418 			if (!checkonly) {
2419 				mblk_t *mb;
2420 
2421 				if (attrs == NULL ||
2422 				    (mb = attrs->udpattr_mb) == NULL)
2423 					return (EINVAL);
2424 				if ((cr = DB_CRED(mb)) == NULL)
2425 					cr = udp->udp_connp->conn_cred;
2426 				ASSERT(cr != NULL);
2427 				if ((tsl = crgetlabel(cr)) == NULL)
2428 					return (EINVAL);
2429 				newcr = copycred_from_bslabel(cr, UCLABEL(ucr),
2430 				    tsl->tsl_doi, KM_NOSLEEP);
2431 				if (newcr == NULL)
2432 					return (ENOSR);
2433 				mblk_setcred(mb, newcr);
2434 				attrs->udpattr_credset = B_TRUE;
2435 				crfree(newcr);
2436 			}
2437 			break;
2438 		}
2439 		case SO_EXCLBIND:
2440 			if (!checkonly)
2441 				udp->udp_exclbind = onoff;
2442 			break;
2443 		default:
2444 			*outlenp = 0;
2445 			return (EINVAL);
2446 		}
2447 		break;
2448 	case IPPROTO_IP:
2449 		if (udp->udp_family != AF_INET) {
2450 			*outlenp = 0;
2451 			return (ENOPROTOOPT);
2452 		}
2453 		switch (name) {
2454 		case IP_OPTIONS:
2455 		case T_IP_OPTIONS:
2456 			/* Save options for use by IP. */
2457 			newlen = inlen + udp->udp_label_len;
2458 			if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
2459 				*outlenp = 0;
2460 				return (EINVAL);
2461 			}
2462 			if (checkonly)
2463 				break;
2464 
2465 			/*
2466 			 * Update the stored options taking into account
2467 			 * any CIPSO option which we should not overwrite.
2468 			 */
2469 			if (!tsol_option_set(&udp->udp_ip_snd_options,
2470 			    &udp->udp_ip_snd_options_len,
2471 			    udp->udp_label_len, invalp, inlen)) {
2472 				*outlenp = 0;
2473 				return (ENOMEM);
2474 			}
2475 
2476 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2477 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
2478 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
2479 			rw_exit(&udp->udp_rwlock);
2480 			(void) proto_set_tx_wroff(connp->conn_rq, connp,
2481 			    sth_wroff);
2482 			rw_enter(&udp->udp_rwlock, RW_WRITER);
2483 			break;
2484 
2485 		case IP_TTL:
2486 			if (!checkonly) {
2487 				udp->udp_ttl = (uchar_t)*i1;
2488 			}
2489 			break;
2490 		case IP_TOS:
2491 		case T_IP_TOS:
2492 			if (!checkonly) {
2493 				udp->udp_type_of_service = (uchar_t)*i1;
2494 			}
2495 			break;
2496 		case IP_MULTICAST_IF: {
2497 			/*
2498 			 * TODO should check OPTMGMT reply and undo this if
2499 			 * there is an error.
2500 			 */
2501 			struct in_addr *inap = (struct in_addr *)invalp;
2502 			if (!checkonly) {
2503 				udp->udp_multicast_if_addr =
2504 				    inap->s_addr;
2505 				PASS_OPT_TO_IP(connp);
2506 			}
2507 			break;
2508 		}
2509 		case IP_MULTICAST_TTL:
2510 			if (!checkonly)
2511 				udp->udp_multicast_ttl = *invalp;
2512 			break;
2513 		case IP_MULTICAST_LOOP:
2514 			if (!checkonly) {
2515 				connp->conn_multicast_loop = *invalp;
2516 				PASS_OPT_TO_IP(connp);
2517 			}
2518 			break;
2519 		case IP_RECVOPTS:
2520 			if (!checkonly)
2521 				udp->udp_recvopts = onoff;
2522 			break;
2523 		case IP_RECVDSTADDR:
2524 			if (!checkonly)
2525 				udp->udp_recvdstaddr = onoff;
2526 			break;
2527 		case IP_RECVIF:
2528 			if (!checkonly) {
2529 				udp->udp_recvif = onoff;
2530 				PASS_OPT_TO_IP(connp);
2531 			}
2532 			break;
2533 		case IP_RECVSLLA:
2534 			if (!checkonly) {
2535 				udp->udp_recvslla = onoff;
2536 				PASS_OPT_TO_IP(connp);
2537 			}
2538 			break;
2539 		case IP_RECVTTL:
2540 			if (!checkonly)
2541 				udp->udp_recvttl = onoff;
2542 			break;
2543 		case IP_PKTINFO: {
2544 			/*
2545 			 * This also handles IP_RECVPKTINFO.
2546 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
2547 			 * Differentiation is based on the size of the
2548 			 * argument passed in.
2549 			 */
2550 			struct in_pktinfo *pktinfop;
2551 			ip4_pkt_t *attr_pktinfop;
2552 
2553 			if (checkonly)
2554 				break;
2555 
2556 			if (inlen == sizeof (int)) {
2557 				/*
2558 				 * This is IP_RECVPKTINFO option.
2559 				 * Keep a local copy of whether this option is
2560 				 * set or not and pass it down to IP for
2561 				 * processing.
2562 				 */
2563 
2564 				udp->udp_ip_recvpktinfo = onoff;
2565 				return (-EINVAL);
2566 			}
2567 
2568 			if (attrs == NULL ||
2569 			    (attr_pktinfop = attrs->udpattr_ipp4) == NULL) {
2570 				/*
2571 				 * sticky option or no buffer to return
2572 				 * the results.
2573 				 */
2574 				return (EINVAL);
2575 			}
2576 
2577 			if (inlen != sizeof (struct in_pktinfo))
2578 				return (EINVAL);
2579 
2580 			pktinfop = (struct in_pktinfo *)invalp;
2581 
2582 			/*
2583 			 * At least one of the values should be specified
2584 			 */
2585 			if (pktinfop->ipi_ifindex == 0 &&
2586 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2587 				return (EINVAL);
2588 			}
2589 
2590 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2591 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2592 
2593 			break;
2594 		}
2595 		case IP_ADD_MEMBERSHIP:
2596 		case IP_DROP_MEMBERSHIP:
2597 		case IP_BLOCK_SOURCE:
2598 		case IP_UNBLOCK_SOURCE:
2599 		case IP_ADD_SOURCE_MEMBERSHIP:
2600 		case IP_DROP_SOURCE_MEMBERSHIP:
2601 		case MCAST_JOIN_GROUP:
2602 		case MCAST_LEAVE_GROUP:
2603 		case MCAST_BLOCK_SOURCE:
2604 		case MCAST_UNBLOCK_SOURCE:
2605 		case MCAST_JOIN_SOURCE_GROUP:
2606 		case MCAST_LEAVE_SOURCE_GROUP:
2607 		case IP_SEC_OPT:
2608 		case IP_NEXTHOP:
2609 		case IP_DHCPINIT_IF:
2610 			/*
2611 			 * "soft" error (negative)
2612 			 * option not handled at this level
2613 			 * Do not modify *outlenp.
2614 			 */
2615 			return (-EINVAL);
2616 		case IP_BOUND_IF:
2617 			if (!checkonly) {
2618 				udp->udp_bound_if = *i1;
2619 				PASS_OPT_TO_IP(connp);
2620 			}
2621 			break;
2622 		case IP_UNSPEC_SRC:
2623 			if (!checkonly) {
2624 				udp->udp_unspec_source = onoff;
2625 				PASS_OPT_TO_IP(connp);
2626 			}
2627 			break;
2628 		case IP_BROADCAST_TTL:
2629 			if (!checkonly)
2630 				connp->conn_broadcast_ttl = *invalp;
2631 			break;
2632 		default:
2633 			*outlenp = 0;
2634 			return (EINVAL);
2635 		}
2636 		break;
2637 	case IPPROTO_IPV6: {
2638 		ip6_pkt_t		*ipp;
2639 		boolean_t		sticky;
2640 
2641 		if (udp->udp_family != AF_INET6) {
2642 			*outlenp = 0;
2643 			return (ENOPROTOOPT);
2644 		}
2645 		/*
2646 		 * Deal with both sticky options and ancillary data
2647 		 */
2648 		sticky = B_FALSE;
2649 		if (attrs == NULL || (ipp = attrs->udpattr_ipp6) ==
2650 		    NULL) {
2651 			/* sticky options, or none */
2652 			ipp = &udp->udp_sticky_ipp;
2653 			sticky = B_TRUE;
2654 		}
2655 
2656 		switch (name) {
2657 		case IPV6_MULTICAST_IF:
2658 			if (!checkonly) {
2659 				udp->udp_multicast_if_index = *i1;
2660 				PASS_OPT_TO_IP(connp);
2661 			}
2662 			break;
2663 		case IPV6_UNICAST_HOPS:
2664 			/* -1 means use default */
2665 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2666 				*outlenp = 0;
2667 				return (EINVAL);
2668 			}
2669 			if (!checkonly) {
2670 				if (*i1 == -1) {
2671 					udp->udp_ttl = ipp->ipp_unicast_hops =
2672 					    us->us_ipv6_hoplimit;
2673 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2674 					/* Pass modified value to IP. */
2675 					*i1 = udp->udp_ttl;
2676 				} else {
2677 					udp->udp_ttl = ipp->ipp_unicast_hops =
2678 					    (uint8_t)*i1;
2679 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2680 				}
2681 				/* Rebuild the header template */
2682 				error = udp_build_hdrs(udp);
2683 				if (error != 0) {
2684 					*outlenp = 0;
2685 					return (error);
2686 				}
2687 			}
2688 			break;
2689 		case IPV6_MULTICAST_HOPS:
2690 			/* -1 means use default */
2691 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2692 				*outlenp = 0;
2693 				return (EINVAL);
2694 			}
2695 			if (!checkonly) {
2696 				if (*i1 == -1) {
2697 					udp->udp_multicast_ttl =
2698 					    ipp->ipp_multicast_hops =
2699 					    IP_DEFAULT_MULTICAST_TTL;
2700 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2701 					/* Pass modified value to IP. */
2702 					*i1 = udp->udp_multicast_ttl;
2703 				} else {
2704 					udp->udp_multicast_ttl =
2705 					    ipp->ipp_multicast_hops =
2706 					    (uint8_t)*i1;
2707 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2708 				}
2709 			}
2710 			break;
2711 		case IPV6_MULTICAST_LOOP:
2712 			if (*i1 != 0 && *i1 != 1) {
2713 				*outlenp = 0;
2714 				return (EINVAL);
2715 			}
2716 			if (!checkonly) {
2717 				connp->conn_multicast_loop = *i1;
2718 				PASS_OPT_TO_IP(connp);
2719 			}
2720 			break;
2721 		case IPV6_JOIN_GROUP:
2722 		case IPV6_LEAVE_GROUP:
2723 		case MCAST_JOIN_GROUP:
2724 		case MCAST_LEAVE_GROUP:
2725 		case MCAST_BLOCK_SOURCE:
2726 		case MCAST_UNBLOCK_SOURCE:
2727 		case MCAST_JOIN_SOURCE_GROUP:
2728 		case MCAST_LEAVE_SOURCE_GROUP:
2729 			/*
2730 			 * "soft" error (negative)
2731 			 * option not handled at this level
2732 			 * Note: Do not modify *outlenp
2733 			 */
2734 			return (-EINVAL);
2735 		case IPV6_BOUND_IF:
2736 			if (!checkonly) {
2737 				udp->udp_bound_if = *i1;
2738 				PASS_OPT_TO_IP(connp);
2739 			}
2740 			break;
2741 		case IPV6_UNSPEC_SRC:
2742 			if (!checkonly) {
2743 				udp->udp_unspec_source = onoff;
2744 				PASS_OPT_TO_IP(connp);
2745 			}
2746 			break;
2747 		/*
2748 		 * Set boolean switches for ancillary data delivery
2749 		 */
2750 		case IPV6_RECVPKTINFO:
2751 			if (!checkonly) {
2752 				udp->udp_ip_recvpktinfo = onoff;
2753 				PASS_OPT_TO_IP(connp);
2754 			}
2755 			break;
2756 		case IPV6_RECVTCLASS:
2757 			if (!checkonly) {
2758 				udp->udp_ipv6_recvtclass = onoff;
2759 				PASS_OPT_TO_IP(connp);
2760 			}
2761 			break;
2762 		case IPV6_RECVPATHMTU:
2763 			if (!checkonly) {
2764 				udp->udp_ipv6_recvpathmtu = onoff;
2765 				PASS_OPT_TO_IP(connp);
2766 			}
2767 			break;
2768 		case IPV6_RECVHOPLIMIT:
2769 			if (!checkonly) {
2770 				udp->udp_ipv6_recvhoplimit = onoff;
2771 				PASS_OPT_TO_IP(connp);
2772 			}
2773 			break;
2774 		case IPV6_RECVHOPOPTS:
2775 			if (!checkonly) {
2776 				udp->udp_ipv6_recvhopopts = onoff;
2777 				PASS_OPT_TO_IP(connp);
2778 			}
2779 			break;
2780 		case IPV6_RECVDSTOPTS:
2781 			if (!checkonly) {
2782 				udp->udp_ipv6_recvdstopts = onoff;
2783 				PASS_OPT_TO_IP(connp);
2784 			}
2785 			break;
2786 		case _OLD_IPV6_RECVDSTOPTS:
2787 			if (!checkonly)
2788 				udp->udp_old_ipv6_recvdstopts = onoff;
2789 			break;
2790 		case IPV6_RECVRTHDRDSTOPTS:
2791 			if (!checkonly) {
2792 				udp->udp_ipv6_recvrthdrdstopts = onoff;
2793 				PASS_OPT_TO_IP(connp);
2794 			}
2795 			break;
2796 		case IPV6_RECVRTHDR:
2797 			if (!checkonly) {
2798 				udp->udp_ipv6_recvrthdr = onoff;
2799 				PASS_OPT_TO_IP(connp);
2800 			}
2801 			break;
2802 		/*
2803 		 * Set sticky options or ancillary data.
2804 		 * If sticky options, (re)build any extension headers
2805 		 * that might be needed as a result.
2806 		 */
2807 		case IPV6_PKTINFO:
2808 			/*
2809 			 * The source address and ifindex are verified
2810 			 * in ip_opt_set(). For ancillary data the
2811 			 * source address is checked in ip_wput_v6.
2812 			 */
2813 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2814 				return (EINVAL);
2815 			if (checkonly)
2816 				break;
2817 
2818 			if (inlen == 0) {
2819 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2820 				ipp->ipp_sticky_ignored |=
2821 				    (IPPF_IFINDEX|IPPF_ADDR);
2822 			} else {
2823 				struct in6_pktinfo *pkti;
2824 
2825 				pkti = (struct in6_pktinfo *)invalp;
2826 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2827 				ipp->ipp_addr = pkti->ipi6_addr;
2828 				if (ipp->ipp_ifindex != 0)
2829 					ipp->ipp_fields |= IPPF_IFINDEX;
2830 				else
2831 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2832 				if (!IN6_IS_ADDR_UNSPECIFIED(
2833 				    &ipp->ipp_addr))
2834 					ipp->ipp_fields |= IPPF_ADDR;
2835 				else
2836 					ipp->ipp_fields &= ~IPPF_ADDR;
2837 			}
2838 			if (sticky) {
2839 				error = udp_build_hdrs(udp);
2840 				if (error != 0)
2841 					return (error);
2842 				PASS_OPT_TO_IP(connp);
2843 			}
2844 			break;
2845 		case IPV6_HOPLIMIT:
2846 			if (sticky)
2847 				return (EINVAL);
2848 			if (inlen != 0 && inlen != sizeof (int))
2849 				return (EINVAL);
2850 			if (checkonly)
2851 				break;
2852 
2853 			if (inlen == 0) {
2854 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2855 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2856 			} else {
2857 				if (*i1 > 255 || *i1 < -1)
2858 					return (EINVAL);
2859 				if (*i1 == -1)
2860 					ipp->ipp_hoplimit =
2861 					    us->us_ipv6_hoplimit;
2862 				else
2863 					ipp->ipp_hoplimit = *i1;
2864 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2865 			}
2866 			break;
2867 		case IPV6_TCLASS:
2868 			if (inlen != 0 && inlen != sizeof (int))
2869 				return (EINVAL);
2870 			if (checkonly)
2871 				break;
2872 
2873 			if (inlen == 0) {
2874 				ipp->ipp_fields &= ~IPPF_TCLASS;
2875 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2876 			} else {
2877 				if (*i1 > 255 || *i1 < -1)
2878 					return (EINVAL);
2879 				if (*i1 == -1)
2880 					ipp->ipp_tclass = 0;
2881 				else
2882 					ipp->ipp_tclass = *i1;
2883 				ipp->ipp_fields |= IPPF_TCLASS;
2884 			}
2885 			if (sticky) {
2886 				error = udp_build_hdrs(udp);
2887 				if (error != 0)
2888 					return (error);
2889 			}
2890 			break;
2891 		case IPV6_NEXTHOP:
2892 			/*
2893 			 * IP will verify that the nexthop is reachable
2894 			 * and fail for sticky options.
2895 			 */
2896 			if (inlen != 0 && inlen != sizeof (sin6_t))
2897 				return (EINVAL);
2898 			if (checkonly)
2899 				break;
2900 
2901 			if (inlen == 0) {
2902 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2903 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2904 			} else {
2905 				sin6_t *sin6 = (sin6_t *)invalp;
2906 
2907 				if (sin6->sin6_family != AF_INET6) {
2908 					return (EAFNOSUPPORT);
2909 				}
2910 				if (IN6_IS_ADDR_V4MAPPED(
2911 				    &sin6->sin6_addr))
2912 					return (EADDRNOTAVAIL);
2913 				ipp->ipp_nexthop = sin6->sin6_addr;
2914 				if (!IN6_IS_ADDR_UNSPECIFIED(
2915 				    &ipp->ipp_nexthop))
2916 					ipp->ipp_fields |= IPPF_NEXTHOP;
2917 				else
2918 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2919 			}
2920 			if (sticky) {
2921 				error = udp_build_hdrs(udp);
2922 				if (error != 0)
2923 					return (error);
2924 				PASS_OPT_TO_IP(connp);
2925 			}
2926 			break;
2927 		case IPV6_HOPOPTS: {
2928 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2929 			/*
2930 			 * Sanity checks - minimum size, size a multiple of
2931 			 * eight bytes, and matching size passed in.
2932 			 */
2933 			if (inlen != 0 &&
2934 			    inlen != (8 * (hopts->ip6h_len + 1)))
2935 				return (EINVAL);
2936 
2937 			if (checkonly)
2938 				break;
2939 
2940 			error = optcom_pkt_set(invalp, inlen, sticky,
2941 			    (uchar_t **)&ipp->ipp_hopopts,
2942 			    &ipp->ipp_hopoptslen,
2943 			    sticky ? udp->udp_label_len_v6 : 0);
2944 			if (error != 0)
2945 				return (error);
2946 			if (ipp->ipp_hopoptslen == 0) {
2947 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2948 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2949 			} else {
2950 				ipp->ipp_fields |= IPPF_HOPOPTS;
2951 			}
2952 			if (sticky) {
2953 				error = udp_build_hdrs(udp);
2954 				if (error != 0)
2955 					return (error);
2956 			}
2957 			break;
2958 		}
2959 		case IPV6_RTHDRDSTOPTS: {
2960 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2961 
2962 			/*
2963 			 * Sanity checks - minimum size, size a multiple of
2964 			 * eight bytes, and matching size passed in.
2965 			 */
2966 			if (inlen != 0 &&
2967 			    inlen != (8 * (dopts->ip6d_len + 1)))
2968 				return (EINVAL);
2969 
2970 			if (checkonly)
2971 				break;
2972 
2973 			if (inlen == 0) {
2974 				if (sticky &&
2975 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2976 					kmem_free(ipp->ipp_rtdstopts,
2977 					    ipp->ipp_rtdstoptslen);
2978 					ipp->ipp_rtdstopts = NULL;
2979 					ipp->ipp_rtdstoptslen = 0;
2980 				}
2981 
2982 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2983 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2984 			} else {
2985 				error = optcom_pkt_set(invalp, inlen, sticky,
2986 				    (uchar_t **)&ipp->ipp_rtdstopts,
2987 				    &ipp->ipp_rtdstoptslen, 0);
2988 				if (error != 0)
2989 					return (error);
2990 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2991 			}
2992 			if (sticky) {
2993 				error = udp_build_hdrs(udp);
2994 				if (error != 0)
2995 					return (error);
2996 			}
2997 			break;
2998 		}
2999 		case IPV6_DSTOPTS: {
3000 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
3001 
3002 			/*
3003 			 * Sanity checks - minimum size, size a multiple of
3004 			 * eight bytes, and matching size passed in.
3005 			 */
3006 			if (inlen != 0 &&
3007 			    inlen != (8 * (dopts->ip6d_len + 1)))
3008 				return (EINVAL);
3009 
3010 			if (checkonly)
3011 				break;
3012 
3013 			if (inlen == 0) {
3014 				if (sticky &&
3015 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
3016 					kmem_free(ipp->ipp_dstopts,
3017 					    ipp->ipp_dstoptslen);
3018 					ipp->ipp_dstopts = NULL;
3019 					ipp->ipp_dstoptslen = 0;
3020 				}
3021 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
3022 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
3023 			} else {
3024 				error = optcom_pkt_set(invalp, inlen, sticky,
3025 				    (uchar_t **)&ipp->ipp_dstopts,
3026 				    &ipp->ipp_dstoptslen, 0);
3027 				if (error != 0)
3028 					return (error);
3029 				ipp->ipp_fields |= IPPF_DSTOPTS;
3030 			}
3031 			if (sticky) {
3032 				error = udp_build_hdrs(udp);
3033 				if (error != 0)
3034 					return (error);
3035 			}
3036 			break;
3037 		}
3038 		case IPV6_RTHDR: {
3039 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
3040 
3041 			/*
3042 			 * Sanity checks - minimum size, size a multiple of
3043 			 * eight bytes, and matching size passed in.
3044 			 */
3045 			if (inlen != 0 &&
3046 			    inlen != (8 * (rt->ip6r_len + 1)))
3047 				return (EINVAL);
3048 
3049 			if (checkonly)
3050 				break;
3051 
3052 			if (inlen == 0) {
3053 				if (sticky &&
3054 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
3055 					kmem_free(ipp->ipp_rthdr,
3056 					    ipp->ipp_rthdrlen);
3057 					ipp->ipp_rthdr = NULL;
3058 					ipp->ipp_rthdrlen = 0;
3059 				}
3060 				ipp->ipp_fields &= ~IPPF_RTHDR;
3061 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
3062 			} else {
3063 				error = optcom_pkt_set(invalp, inlen, sticky,
3064 				    (uchar_t **)&ipp->ipp_rthdr,
3065 				    &ipp->ipp_rthdrlen, 0);
3066 				if (error != 0)
3067 					return (error);
3068 				ipp->ipp_fields |= IPPF_RTHDR;
3069 			}
3070 			if (sticky) {
3071 				error = udp_build_hdrs(udp);
3072 				if (error != 0)
3073 					return (error);
3074 			}
3075 			break;
3076 		}
3077 
3078 		case IPV6_DONTFRAG:
3079 			if (checkonly)
3080 				break;
3081 
3082 			if (onoff) {
3083 				ipp->ipp_fields |= IPPF_DONTFRAG;
3084 			} else {
3085 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
3086 			}
3087 			break;
3088 
3089 		case IPV6_USE_MIN_MTU:
3090 			if (inlen != sizeof (int))
3091 				return (EINVAL);
3092 
3093 			if (*i1 < -1 || *i1 > 1)
3094 				return (EINVAL);
3095 
3096 			if (checkonly)
3097 				break;
3098 
3099 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3100 			ipp->ipp_use_min_mtu = *i1;
3101 			break;
3102 
3103 		case IPV6_BOUND_PIF:
3104 		case IPV6_SEC_OPT:
3105 		case IPV6_DONTFAILOVER_IF:
3106 		case IPV6_SRC_PREFERENCES:
3107 		case IPV6_V6ONLY:
3108 			/* Handled at the IP level */
3109 			return (-EINVAL);
3110 		default:
3111 			*outlenp = 0;
3112 			return (EINVAL);
3113 		}
3114 		break;
3115 		}		/* end IPPROTO_IPV6 */
3116 	case IPPROTO_UDP:
3117 		switch (name) {
3118 		case UDP_ANONPRIVBIND:
3119 			if ((error = secpolicy_net_privaddr(cr, 0,
3120 			    IPPROTO_UDP)) != 0) {
3121 				*outlenp = 0;
3122 				return (error);
3123 			}
3124 			if (!checkonly) {
3125 				udp->udp_anon_priv_bind = onoff;
3126 			}
3127 			break;
3128 		case UDP_EXCLBIND:
3129 			if (!checkonly)
3130 				udp->udp_exclbind = onoff;
3131 			break;
3132 		case UDP_RCVHDR:
3133 			if (!checkonly)
3134 				udp->udp_rcvhdr = onoff;
3135 			break;
3136 		case UDP_NAT_T_ENDPOINT:
3137 			if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
3138 				*outlenp = 0;
3139 				return (error);
3140 			}
3141 
3142 			/*
3143 			 * Use udp_family instead so we can avoid ambiguitites
3144 			 * with AF_INET6 sockets that may switch from IPv4
3145 			 * to IPv6.
3146 			 */
3147 			if (udp->udp_family != AF_INET) {
3148 				*outlenp = 0;
3149 				return (EAFNOSUPPORT);
3150 			}
3151 
3152 			if (!checkonly) {
3153 				int size;
3154 
3155 				udp->udp_nat_t_endpoint = onoff;
3156 
3157 				udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3158 				    UDPH_SIZE + udp->udp_ip_snd_options_len;
3159 
3160 				/* Also, adjust wroff */
3161 				if (onoff) {
3162 					udp->udp_max_hdr_len +=
3163 					    sizeof (uint32_t);
3164 				}
3165 				size = udp->udp_max_hdr_len +
3166 				    us->us_wroff_extra;
3167 				(void) proto_set_tx_wroff(connp->conn_rq, connp,
3168 				    size);
3169 			}
3170 			break;
3171 		default:
3172 			*outlenp = 0;
3173 			return (EINVAL);
3174 		}
3175 		break;
3176 	default:
3177 		*outlenp = 0;
3178 		return (EINVAL);
3179 	}
3180 	/*
3181 	 * Common case of OK return with outval same as inval.
3182 	 */
3183 	if (invalp != outvalp) {
3184 		/* don't trust bcopy for identical src/dst */
3185 		(void) bcopy(invalp, outvalp, inlen);
3186 	}
3187 	*outlenp = inlen;
3188 	return (0);
3189 }
3190 
3191 int
3192 udp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3193     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3194     void *thisdg_attrs, cred_t *cr)
3195 {
3196 	int		error;
3197 	boolean_t	checkonly;
3198 
3199 	error = 0;
3200 	switch (optset_context) {
3201 	case SETFN_OPTCOM_CHECKONLY:
3202 		checkonly = B_TRUE;
3203 		/*
3204 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3205 		 * inlen != 0 implies value supplied and
3206 		 * 	we have to "pretend" to set it.
3207 		 * inlen == 0 implies that there is no
3208 		 * 	value part in T_CHECK request and just validation
3209 		 * done elsewhere should be enough, we just return here.
3210 		 */
3211 		if (inlen == 0) {
3212 			*outlenp = 0;
3213 			goto done;
3214 		}
3215 		break;
3216 	case SETFN_OPTCOM_NEGOTIATE:
3217 		checkonly = B_FALSE;
3218 		break;
3219 	case SETFN_UD_NEGOTIATE:
3220 	case SETFN_CONN_NEGOTIATE:
3221 		checkonly = B_FALSE;
3222 		/*
3223 		 * Negotiating local and "association-related" options
3224 		 * through T_UNITDATA_REQ.
3225 		 *
3226 		 * Following routine can filter out ones we do not
3227 		 * want to be "set" this way.
3228 		 */
3229 		if (!udp_opt_allow_udr_set(level, name)) {
3230 			*outlenp = 0;
3231 			error = EINVAL;
3232 			goto done;
3233 		}
3234 		break;
3235 	default:
3236 		/*
3237 		 * We should never get here
3238 		 */
3239 		*outlenp = 0;
3240 		error = EINVAL;
3241 		goto done;
3242 	}
3243 
3244 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3245 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3246 
3247 	error = udp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3248 	    outvalp, cr, thisdg_attrs, checkonly);
3249 done:
3250 	return (error);
3251 }
3252 
3253 /* ARGSUSED */
3254 int
3255 udp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3256     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3257     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3258 {
3259 	conn_t  *connp =  Q_TO_CONN(q);
3260 	int error;
3261 	udp_t	*udp = connp->conn_udp;
3262 
3263 	rw_enter(&udp->udp_rwlock, RW_WRITER);
3264 	error = udp_opt_set(connp, optset_context, level, name, inlen, invalp,
3265 	    outlenp, outvalp, thisdg_attrs, cr);
3266 	rw_exit(&udp->udp_rwlock);
3267 	return (error);
3268 }
3269 
3270 /*
3271  * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl.
3272  * The headers include ip6i_t (if needed), ip6_t, any sticky extension
3273  * headers, and the udp header.
3274  * Returns failure if can't allocate memory.
3275  */
3276 static int
3277 udp_build_hdrs(udp_t *udp)
3278 {
3279 	udp_stack_t *us = udp->udp_us;
3280 	uchar_t	*hdrs;
3281 	uint_t	hdrs_len;
3282 	ip6_t	*ip6h;
3283 	ip6i_t	*ip6i;
3284 	udpha_t	*udpha;
3285 	ip6_pkt_t *ipp = &udp->udp_sticky_ipp;
3286 	size_t	sth_wroff;
3287 	conn_t	*connp = udp->udp_connp;
3288 
3289 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
3290 	ASSERT(connp != NULL);
3291 
3292 	hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE;
3293 	ASSERT(hdrs_len != 0);
3294 	if (hdrs_len != udp->udp_sticky_hdrs_len) {
3295 		/* Need to reallocate */
3296 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3297 		if (hdrs == NULL)
3298 			return (ENOMEM);
3299 
3300 		if (udp->udp_sticky_hdrs_len != 0) {
3301 			kmem_free(udp->udp_sticky_hdrs,
3302 			    udp->udp_sticky_hdrs_len);
3303 		}
3304 		udp->udp_sticky_hdrs = hdrs;
3305 		udp->udp_sticky_hdrs_len = hdrs_len;
3306 	}
3307 	ip_build_hdrs_v6(udp->udp_sticky_hdrs,
3308 	    udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP);
3309 
3310 	/* Set header fields not in ipp */
3311 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3312 		ip6i = (ip6i_t *)udp->udp_sticky_hdrs;
3313 		ip6h = (ip6_t *)&ip6i[1];
3314 	} else {
3315 		ip6h = (ip6_t *)udp->udp_sticky_hdrs;
3316 	}
3317 
3318 	if (!(ipp->ipp_fields & IPPF_ADDR))
3319 		ip6h->ip6_src = udp->udp_v6src;
3320 
3321 	udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE);
3322 	udpha->uha_src_port = udp->udp_port;
3323 
3324 	/* Try to get everything in a single mblk */
3325 	if (hdrs_len > udp->udp_max_hdr_len) {
3326 		udp->udp_max_hdr_len = hdrs_len;
3327 		sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
3328 		rw_exit(&udp->udp_rwlock);
3329 		(void) proto_set_tx_wroff(udp->udp_connp->conn_rq,
3330 		    udp->udp_connp, sth_wroff);
3331 		rw_enter(&udp->udp_rwlock, RW_WRITER);
3332 	}
3333 	return (0);
3334 }
3335 
3336 /*
3337  * This routine retrieves the value of an ND variable in a udpparam_t
3338  * structure.  It is called through nd_getset when a user reads the
3339  * variable.
3340  */
3341 /* ARGSUSED */
3342 static int
3343 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3344 {
3345 	udpparam_t *udppa = (udpparam_t *)cp;
3346 
3347 	(void) mi_mpprintf(mp, "%d", udppa->udp_param_value);
3348 	return (0);
3349 }
3350 
3351 /*
3352  * Walk through the param array specified registering each element with the
3353  * named dispatch (ND) handler.
3354  */
3355 static boolean_t
3356 udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt)
3357 {
3358 	for (; cnt-- > 0; udppa++) {
3359 		if (udppa->udp_param_name && udppa->udp_param_name[0]) {
3360 			if (!nd_load(ndp, udppa->udp_param_name,
3361 			    udp_param_get, udp_param_set,
3362 			    (caddr_t)udppa)) {
3363 				nd_free(ndp);
3364 				return (B_FALSE);
3365 			}
3366 		}
3367 	}
3368 	if (!nd_load(ndp, "udp_extra_priv_ports",
3369 	    udp_extra_priv_ports_get, NULL, NULL)) {
3370 		nd_free(ndp);
3371 		return (B_FALSE);
3372 	}
3373 	if (!nd_load(ndp, "udp_extra_priv_ports_add",
3374 	    NULL, udp_extra_priv_ports_add, NULL)) {
3375 		nd_free(ndp);
3376 		return (B_FALSE);
3377 	}
3378 	if (!nd_load(ndp, "udp_extra_priv_ports_del",
3379 	    NULL, udp_extra_priv_ports_del, NULL)) {
3380 		nd_free(ndp);
3381 		return (B_FALSE);
3382 	}
3383 	if (!nd_load(ndp, "udp_status", udp_status_report, NULL,
3384 	    NULL)) {
3385 		nd_free(ndp);
3386 		return (B_FALSE);
3387 	}
3388 	if (!nd_load(ndp, "udp_bind_hash", udp_bind_hash_report, NULL,
3389 	    NULL)) {
3390 		nd_free(ndp);
3391 		return (B_FALSE);
3392 	}
3393 	return (B_TRUE);
3394 }
3395 
3396 /* This routine sets an ND variable in a udpparam_t structure. */
3397 /* ARGSUSED */
3398 static int
3399 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3400 {
3401 	long		new_value;
3402 	udpparam_t	*udppa = (udpparam_t *)cp;
3403 
3404 	/*
3405 	 * Fail the request if the new value does not lie within the
3406 	 * required bounds.
3407 	 */
3408 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3409 	    new_value < udppa->udp_param_min ||
3410 	    new_value > udppa->udp_param_max) {
3411 		return (EINVAL);
3412 	}
3413 
3414 	/* Set the new value */
3415 	udppa->udp_param_value = new_value;
3416 	return (0);
3417 }
3418 
3419 /*
3420  * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with
3421  * T_opthdr) and return the number of bytes copied.  'dbuf' may be NULL to
3422  * just count the length needed for allocation.  If 'dbuf' is non-NULL,
3423  * then it's assumed to be allocated to be large enough.
3424  *
3425  * Returns zero if trimming of the security option causes all options to go
3426  * away.
3427  */
3428 static size_t
3429 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf)
3430 {
3431 	struct T_opthdr *toh;
3432 	size_t hol = ipp->ipp_hopoptslen;
3433 	ip6_hbh_t *dstopt = NULL;
3434 	const ip6_hbh_t *srcopt = ipp->ipp_hopopts;
3435 	size_t tlen, olen, plen;
3436 	boolean_t deleting;
3437 	const struct ip6_opt *sopt, *lastpad;
3438 	struct ip6_opt *dopt;
3439 
3440 	if ((toh = (struct T_opthdr *)dbuf) != NULL) {
3441 		toh->level = IPPROTO_IPV6;
3442 		toh->name = IPV6_HOPOPTS;
3443 		toh->status = 0;
3444 		dstopt = (ip6_hbh_t *)(toh + 1);
3445 	}
3446 
3447 	/*
3448 	 * If labeling is enabled, then skip the label option
3449 	 * but get other options if there are any.
3450 	 */
3451 	if (is_system_labeled()) {
3452 		dopt = NULL;
3453 		if (dstopt != NULL) {
3454 			/* will fill in ip6h_len later */
3455 			dstopt->ip6h_nxt = srcopt->ip6h_nxt;
3456 			dopt = (struct ip6_opt *)(dstopt + 1);
3457 		}
3458 		sopt = (const struct ip6_opt *)(srcopt + 1);
3459 		hol -= sizeof (*srcopt);
3460 		tlen = sizeof (*dstopt);
3461 		lastpad = NULL;
3462 		deleting = B_FALSE;
3463 		/*
3464 		 * This loop finds the first (lastpad pointer) of any number of
3465 		 * pads that preceeds the security option, then treats the
3466 		 * security option as though it were a pad, and then finds the
3467 		 * next non-pad option (or end of list).
3468 		 *
3469 		 * It then treats the entire block as one big pad.  To preserve
3470 		 * alignment of any options that follow, or just the end of the
3471 		 * list, it computes a minimal new padding size that keeps the
3472 		 * same alignment for the next option.
3473 		 *
3474 		 * If it encounters just a sequence of pads with no security
3475 		 * option, those are copied as-is rather than collapsed.
3476 		 *
3477 		 * Note that to handle the end of list case, the code makes one
3478 		 * loop with 'hol' set to zero.
3479 		 */
3480 		for (;;) {
3481 			if (hol > 0) {
3482 				if (sopt->ip6o_type == IP6OPT_PAD1) {
3483 					if (lastpad == NULL)
3484 						lastpad = sopt;
3485 					sopt = (const struct ip6_opt *)
3486 					    &sopt->ip6o_len;
3487 					hol--;
3488 					continue;
3489 				}
3490 				olen = sopt->ip6o_len + sizeof (*sopt);
3491 				if (olen > hol)
3492 					olen = hol;
3493 				if (sopt->ip6o_type == IP6OPT_PADN ||
3494 				    sopt->ip6o_type == ip6opt_ls) {
3495 					if (sopt->ip6o_type == ip6opt_ls)
3496 						deleting = B_TRUE;
3497 					if (lastpad == NULL)
3498 						lastpad = sopt;
3499 					sopt = (const struct ip6_opt *)
3500 					    ((const char *)sopt + olen);
3501 					hol -= olen;
3502 					continue;
3503 				}
3504 			} else {
3505 				/* if nothing was copied at all, then delete */
3506 				if (tlen == sizeof (*dstopt))
3507 					return (0);
3508 				/* last pass; pick up any trailing padding */
3509 				olen = 0;
3510 			}
3511 			if (deleting) {
3512 				/*
3513 				 * compute aligning effect of deleted material
3514 				 * to reproduce with pad.
3515 				 */
3516 				plen = ((const char *)sopt -
3517 				    (const char *)lastpad) & 7;
3518 				tlen += plen;
3519 				if (dopt != NULL) {
3520 					if (plen == 1) {
3521 						dopt->ip6o_type = IP6OPT_PAD1;
3522 					} else if (plen > 1) {
3523 						plen -= sizeof (*dopt);
3524 						dopt->ip6o_type = IP6OPT_PADN;
3525 						dopt->ip6o_len = plen;
3526 						if (plen > 0)
3527 							bzero(dopt + 1, plen);
3528 					}
3529 					dopt = (struct ip6_opt *)
3530 					    ((char *)dopt + plen);
3531 				}
3532 				deleting = B_FALSE;
3533 				lastpad = NULL;
3534 			}
3535 			/* if there's uncopied padding, then copy that now */
3536 			if (lastpad != NULL) {
3537 				olen += (const char *)sopt -
3538 				    (const char *)lastpad;
3539 				sopt = lastpad;
3540 				lastpad = NULL;
3541 			}
3542 			if (dopt != NULL && olen > 0) {
3543 				bcopy(sopt, dopt, olen);
3544 				dopt = (struct ip6_opt *)((char *)dopt + olen);
3545 			}
3546 			if (hol == 0)
3547 				break;
3548 			tlen += olen;
3549 			sopt = (const struct ip6_opt *)
3550 			    ((const char *)sopt + olen);
3551 			hol -= olen;
3552 		}
3553 		/* go back and patch up the length value, rounded upward */
3554 		if (dstopt != NULL)
3555 			dstopt->ip6h_len = (tlen - 1) >> 3;
3556 	} else {
3557 		tlen = hol;
3558 		if (dstopt != NULL)
3559 			bcopy(srcopt, dstopt, hol);
3560 	}
3561 
3562 	tlen += sizeof (*toh);
3563 	if (toh != NULL)
3564 		toh->len = tlen;
3565 
3566 	return (tlen);
3567 }
3568 
3569 /*
3570  * Update udp_rcv_opt_len from the packet.
3571  * Called when options received, and when no options received but
3572  * udp_ip_recv_opt_len has previously recorded options.
3573  */
3574 static void
3575 udp_save_ip_rcv_opt(udp_t *udp, void *opt, int opt_len)
3576 {
3577 	/* Save the options if any */
3578 	if (opt_len > 0) {
3579 		if (opt_len > udp->udp_ip_rcv_options_len) {
3580 			/* Need to allocate larger buffer */
3581 			if (udp->udp_ip_rcv_options_len != 0)
3582 				mi_free((char *)udp->udp_ip_rcv_options);
3583 			udp->udp_ip_rcv_options_len = 0;
3584 			udp->udp_ip_rcv_options =
3585 			    (uchar_t *)mi_alloc(opt_len, BPRI_HI);
3586 			if (udp->udp_ip_rcv_options != NULL)
3587 				udp->udp_ip_rcv_options_len = opt_len;
3588 		}
3589 		if (udp->udp_ip_rcv_options_len != 0) {
3590 			bcopy(opt, udp->udp_ip_rcv_options, opt_len);
3591 			/* Adjust length if we are resusing the space */
3592 			udp->udp_ip_rcv_options_len = opt_len;
3593 		}
3594 	} else if (udp->udp_ip_rcv_options_len != 0) {
3595 		/* Clear out previously recorded options */
3596 		mi_free((char *)udp->udp_ip_rcv_options);
3597 		udp->udp_ip_rcv_options = NULL;
3598 		udp->udp_ip_rcv_options_len = 0;
3599 	}
3600 }
3601 
3602 static void
3603 udp_queue_fallback(udp_t *udp, mblk_t *mp)
3604 {
3605 	ASSERT(MUTEX_HELD(&udp->udp_recv_lock));
3606 	if (IPCL_IS_NONSTR(udp->udp_connp)) {
3607 		/*
3608 		 * fallback has started but messages have not been moved yet
3609 		 */
3610 		if (udp->udp_fallback_queue_head == NULL) {
3611 			ASSERT(udp->udp_fallback_queue_tail == NULL);
3612 			udp->udp_fallback_queue_head = mp;
3613 			udp->udp_fallback_queue_tail = mp;
3614 		} else {
3615 			ASSERT(udp->udp_fallback_queue_tail != NULL);
3616 			udp->udp_fallback_queue_tail->b_next = mp;
3617 			udp->udp_fallback_queue_tail = mp;
3618 		}
3619 		mutex_exit(&udp->udp_recv_lock);
3620 	} else {
3621 		/*
3622 		 * no more fallbacks possible, ok to drop lock.
3623 		 */
3624 		mutex_exit(&udp->udp_recv_lock);
3625 		putnext(udp->udp_connp->conn_rq, mp);
3626 	}
3627 }
3628 
3629 /* ARGSUSED2 */
3630 static void
3631 udp_input(void *arg1, mblk_t *mp, void *arg2)
3632 {
3633 	conn_t *connp = (conn_t *)arg1;
3634 	struct T_unitdata_ind	*tudi;
3635 	uchar_t			*rptr;		/* Pointer to IP header */
3636 	int			hdr_length;	/* Length of IP+UDP headers */
3637 	int			opt_len;
3638 	int			udi_size;	/* Size of T_unitdata_ind */
3639 	int			mp_len;
3640 	udp_t			*udp;
3641 	udpha_t			*udpha;
3642 	int			ipversion;
3643 	ip6_pkt_t		ipp;
3644 	ip6_t			*ip6h;
3645 	ip6i_t			*ip6i;
3646 	mblk_t			*mp1;
3647 	mblk_t			*options_mp = NULL;
3648 	ip_pktinfo_t		*pinfo = NULL;
3649 	cred_t			*cr = NULL;
3650 	pid_t			cpid;
3651 	uint32_t		udp_ip_rcv_options_len;
3652 	udp_bits_t		udp_bits;
3653 	cred_t			*rcr = connp->conn_cred;
3654 	udp_stack_t *us;
3655 
3656 	ASSERT(connp->conn_flags & IPCL_UDPCONN);
3657 
3658 	udp = connp->conn_udp;
3659 	us = udp->udp_us;
3660 	rptr = mp->b_rptr;
3661 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3662 	ASSERT(OK_32PTR(rptr));
3663 
3664 	/*
3665 	 * IP should have prepended the options data in an M_CTL
3666 	 * Check M_CTL "type" to make sure are not here bcos of
3667 	 * a valid ICMP message
3668 	 */
3669 	if (DB_TYPE(mp) == M_CTL) {
3670 		if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3671 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3672 		    IN_PKTINFO) {
3673 			/*
3674 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3675 			 * has been prepended to the packet by IP. We need to
3676 			 * extract the mblk and adjust the rptr
3677 			 */
3678 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3679 			options_mp = mp;
3680 			mp = mp->b_cont;
3681 			rptr = mp->b_rptr;
3682 			UDP_STAT(us, udp_in_pktinfo);
3683 		} else {
3684 			/*
3685 			 * ICMP messages.
3686 			 */
3687 			udp_icmp_error(connp, mp);
3688 			return;
3689 		}
3690 	}
3691 
3692 	mp_len = msgdsize(mp);
3693 	/*
3694 	 * This is the inbound data path.
3695 	 * First, we check to make sure the IP version number is correct,
3696 	 * and then pull the IP and UDP headers into the first mblk.
3697 	 */
3698 
3699 	/* Initialize regardless if ipversion is IPv4 or IPv6 */
3700 	ipp.ipp_fields = 0;
3701 
3702 	ipversion = IPH_HDR_VERSION(rptr);
3703 
3704 	rw_enter(&udp->udp_rwlock, RW_READER);
3705 	udp_ip_rcv_options_len = udp->udp_ip_rcv_options_len;
3706 	udp_bits = udp->udp_bits;
3707 	rw_exit(&udp->udp_rwlock);
3708 
3709 	switch (ipversion) {
3710 	case IPV4_VERSION:
3711 		ASSERT(MBLKL(mp) >= sizeof (ipha_t));
3712 		ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP);
3713 		hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
3714 		opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE);
3715 		if ((opt_len > 0 || udp_ip_rcv_options_len > 0) &&
3716 		    udp->udp_family == AF_INET) {
3717 			/*
3718 			 * Record/update udp_ip_rcv_options with the lock
3719 			 * held. Not needed for AF_INET6 sockets
3720 			 * since they don't support a getsockopt of IP_OPTIONS.
3721 			 */
3722 			rw_enter(&udp->udp_rwlock, RW_WRITER);
3723 			udp_save_ip_rcv_opt(udp, rptr + IP_SIMPLE_HDR_LENGTH,
3724 			    opt_len);
3725 			rw_exit(&udp->udp_rwlock);
3726 		}
3727 		/* Handle IPV6_RECVPKTINFO even for IPv4 packet. */
3728 		if ((udp->udp_family == AF_INET6) && (pinfo != NULL) &&
3729 		    udp->udp_ip_recvpktinfo) {
3730 			if (pinfo->ip_pkt_flags & IPF_RECVIF) {
3731 				ipp.ipp_fields |= IPPF_IFINDEX;
3732 				ipp.ipp_ifindex = pinfo->ip_pkt_ifindex;
3733 			}
3734 		}
3735 		break;
3736 	case IPV6_VERSION:
3737 		/*
3738 		 * IPv6 packets can only be received by applications
3739 		 * that are prepared to receive IPv6 addresses.
3740 		 * The IP fanout must ensure this.
3741 		 */
3742 		ASSERT(udp->udp_family == AF_INET6);
3743 
3744 		ip6h = (ip6_t *)rptr;
3745 		ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr);
3746 
3747 		if (ip6h->ip6_nxt != IPPROTO_UDP) {
3748 			uint8_t nexthdrp;
3749 			/* Look for ifindex information */
3750 			if (ip6h->ip6_nxt == IPPROTO_RAW) {
3751 				ip6i = (ip6i_t *)ip6h;
3752 				if ((uchar_t *)&ip6i[1] > mp->b_wptr)
3753 					goto tossit;
3754 
3755 				if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3756 					ASSERT(ip6i->ip6i_ifindex != 0);
3757 					ipp.ipp_fields |= IPPF_IFINDEX;
3758 					ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3759 				}
3760 				rptr = (uchar_t *)&ip6i[1];
3761 				mp->b_rptr = rptr;
3762 				if (rptr == mp->b_wptr) {
3763 					mp1 = mp->b_cont;
3764 					freeb(mp);
3765 					mp = mp1;
3766 					rptr = mp->b_rptr;
3767 				}
3768 				if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE))
3769 					goto tossit;
3770 				ip6h = (ip6_t *)rptr;
3771 				mp_len = msgdsize(mp);
3772 			}
3773 			/*
3774 			 * Find any potentially interesting extension headers
3775 			 * as well as the length of the IPv6 + extension
3776 			 * headers.
3777 			 */
3778 			hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) +
3779 			    UDPH_SIZE;
3780 			ASSERT(nexthdrp == IPPROTO_UDP);
3781 		} else {
3782 			hdr_length = IPV6_HDR_LEN + UDPH_SIZE;
3783 			ip6i = NULL;
3784 		}
3785 		break;
3786 	default:
3787 		ASSERT(0);
3788 	}
3789 
3790 	/*
3791 	 * IP inspected the UDP header thus all of it must be in the mblk.
3792 	 * UDP length check is performed for IPv6 packets and IPv4 packets
3793 	 * to check if the size of the packet as specified
3794 	 * by the header is the same as the physical size of the packet.
3795 	 * FIXME? Didn't IP already check this?
3796 	 */
3797 	udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
3798 	if ((MBLKL(mp) < hdr_length) ||
3799 	    (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) {
3800 		goto tossit;
3801 	}
3802 
3803 
3804 	/* Walk past the headers unless IP_RECVHDR was set. */
3805 	if (!udp_bits.udpb_rcvhdr) {
3806 		mp->b_rptr = rptr + hdr_length;
3807 		mp_len -= hdr_length;
3808 	}
3809 
3810 	/*
3811 	 * This is the inbound data path.  Packets are passed upstream as
3812 	 * T_UNITDATA_IND messages with full IP headers still attached.
3813 	 */
3814 	if (udp->udp_family == AF_INET) {
3815 		sin_t *sin;
3816 
3817 		ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
3818 
3819 		/*
3820 		 * Normally only send up the source address.
3821 		 * If IP_RECVDSTADDR is set we include the destination IP
3822 		 * address as an option. With IP_RECVOPTS we include all
3823 		 * the IP options.
3824 		 */
3825 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3826 		if (udp_bits.udpb_recvdstaddr) {
3827 			udi_size += sizeof (struct T_opthdr) +
3828 			    sizeof (struct in_addr);
3829 			UDP_STAT(us, udp_in_recvdstaddr);
3830 		}
3831 
3832 		if (udp_bits.udpb_ip_recvpktinfo && (pinfo != NULL) &&
3833 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3834 			udi_size += sizeof (struct T_opthdr) +
3835 			    sizeof (struct in_pktinfo);
3836 			UDP_STAT(us, udp_ip_rcvpktinfo);
3837 		}
3838 
3839 		if ((udp_bits.udpb_recvopts) && opt_len > 0) {
3840 			udi_size += sizeof (struct T_opthdr) + opt_len;
3841 			UDP_STAT(us, udp_in_recvopts);
3842 		}
3843 
3844 		/*
3845 		 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
3846 		 * space accordingly
3847 		 */
3848 		if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
3849 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3850 			udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
3851 			UDP_STAT(us, udp_in_recvif);
3852 		}
3853 
3854 		if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
3855 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
3856 			udi_size += sizeof (struct T_opthdr) +
3857 			    sizeof (struct sockaddr_dl);
3858 			UDP_STAT(us, udp_in_recvslla);
3859 		}
3860 
3861 		if ((udp_bits.udpb_recvucred) &&
3862 		    (cr = DB_CRED(mp)) != NULL) {
3863 			udi_size += sizeof (struct T_opthdr) + ucredsize;
3864 			cpid = DB_CPID(mp);
3865 			UDP_STAT(us, udp_in_recvucred);
3866 		}
3867 
3868 		/*
3869 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3870 		 * buffer. Since gethrestime() expects a pointer aligned
3871 		 * argument, we allocate space necessary for extra
3872 		 * alignment (even though it might not be used).
3873 		 */
3874 		if (udp_bits.udpb_timestamp) {
3875 			udi_size += sizeof (struct T_opthdr) +
3876 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3877 			UDP_STAT(us, udp_in_timestamp);
3878 		}
3879 
3880 		/*
3881 		 * If IP_RECVTTL is set allocate the appropriate sized buffer
3882 		 */
3883 		if (udp_bits.udpb_recvttl) {
3884 			udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
3885 			UDP_STAT(us, udp_in_recvttl);
3886 		}
3887 
3888 		/* Allocate a message block for the T_UNITDATA_IND structure. */
3889 		mp1 = allocb(udi_size, BPRI_MED);
3890 		if (mp1 == NULL) {
3891 			freemsg(mp);
3892 			if (options_mp != NULL)
3893 				freeb(options_mp);
3894 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
3895 			return;
3896 		}
3897 		mp1->b_cont = mp;
3898 		mp = mp1;
3899 		mp->b_datap->db_type = M_PROTO;
3900 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3901 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3902 		tudi->PRIM_type = T_UNITDATA_IND;
3903 		tudi->SRC_length = sizeof (sin_t);
3904 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3905 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
3906 		    sizeof (sin_t);
3907 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3908 		tudi->OPT_length = udi_size;
3909 		sin = (sin_t *)&tudi[1];
3910 		sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
3911 		sin->sin_port =	udpha->uha_src_port;
3912 		sin->sin_family = udp->udp_family;
3913 		*(uint32_t *)&sin->sin_zero[0] = 0;
3914 		*(uint32_t *)&sin->sin_zero[4] = 0;
3915 
3916 		/*
3917 		 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
3918 		 * IP_RECVTTL has been set.
3919 		 */
3920 		if (udi_size != 0) {
3921 			/*
3922 			 * Copy in destination address before options to avoid
3923 			 * any padding issues.
3924 			 */
3925 			char *dstopt;
3926 
3927 			dstopt = (char *)&sin[1];
3928 			if (udp_bits.udpb_recvdstaddr) {
3929 				struct T_opthdr *toh;
3930 				ipaddr_t *dstptr;
3931 
3932 				toh = (struct T_opthdr *)dstopt;
3933 				toh->level = IPPROTO_IP;
3934 				toh->name = IP_RECVDSTADDR;
3935 				toh->len = sizeof (struct T_opthdr) +
3936 				    sizeof (ipaddr_t);
3937 				toh->status = 0;
3938 				dstopt += sizeof (struct T_opthdr);
3939 				dstptr = (ipaddr_t *)dstopt;
3940 				*dstptr = ((ipha_t *)rptr)->ipha_dst;
3941 				dstopt += sizeof (ipaddr_t);
3942 				udi_size -= toh->len;
3943 			}
3944 
3945 			if (udp_bits.udpb_recvopts && opt_len > 0) {
3946 				struct T_opthdr *toh;
3947 
3948 				toh = (struct T_opthdr *)dstopt;
3949 				toh->level = IPPROTO_IP;
3950 				toh->name = IP_RECVOPTS;
3951 				toh->len = sizeof (struct T_opthdr) + opt_len;
3952 				toh->status = 0;
3953 				dstopt += sizeof (struct T_opthdr);
3954 				bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt,
3955 				    opt_len);
3956 				dstopt += opt_len;
3957 				udi_size -= toh->len;
3958 			}
3959 
3960 			if ((udp_bits.udpb_ip_recvpktinfo) && (pinfo != NULL) &&
3961 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3962 				struct T_opthdr *toh;
3963 				struct in_pktinfo *pktinfop;
3964 
3965 				toh = (struct T_opthdr *)dstopt;
3966 				toh->level = IPPROTO_IP;
3967 				toh->name = IP_PKTINFO;
3968 				toh->len = sizeof (struct T_opthdr) +
3969 				    sizeof (*pktinfop);
3970 				toh->status = 0;
3971 				dstopt += sizeof (struct T_opthdr);
3972 				pktinfop = (struct in_pktinfo *)dstopt;
3973 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3974 				pktinfop->ipi_spec_dst =
3975 				    pinfo->ip_pkt_match_addr;
3976 				pktinfop->ipi_addr.s_addr =
3977 				    ((ipha_t *)rptr)->ipha_dst;
3978 
3979 				dstopt += sizeof (struct in_pktinfo);
3980 				udi_size -= toh->len;
3981 			}
3982 
3983 			if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
3984 			    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
3985 
3986 				struct T_opthdr *toh;
3987 				struct sockaddr_dl	*dstptr;
3988 
3989 				toh = (struct T_opthdr *)dstopt;
3990 				toh->level = IPPROTO_IP;
3991 				toh->name = IP_RECVSLLA;
3992 				toh->len = sizeof (struct T_opthdr) +
3993 				    sizeof (struct sockaddr_dl);
3994 				toh->status = 0;
3995 				dstopt += sizeof (struct T_opthdr);
3996 				dstptr = (struct sockaddr_dl *)dstopt;
3997 				bcopy(&pinfo->ip_pkt_slla, dstptr,
3998 				    sizeof (struct sockaddr_dl));
3999 				dstopt += sizeof (struct sockaddr_dl);
4000 				udi_size -= toh->len;
4001 			}
4002 
4003 			if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
4004 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4005 
4006 				struct T_opthdr *toh;
4007 				uint_t		*dstptr;
4008 
4009 				toh = (struct T_opthdr *)dstopt;
4010 				toh->level = IPPROTO_IP;
4011 				toh->name = IP_RECVIF;
4012 				toh->len = sizeof (struct T_opthdr) +
4013 				    sizeof (uint_t);
4014 				toh->status = 0;
4015 				dstopt += sizeof (struct T_opthdr);
4016 				dstptr = (uint_t *)dstopt;
4017 				*dstptr = pinfo->ip_pkt_ifindex;
4018 				dstopt += sizeof (uint_t);
4019 				udi_size -= toh->len;
4020 			}
4021 
4022 			if (cr != NULL) {
4023 				struct T_opthdr *toh;
4024 
4025 				toh = (struct T_opthdr *)dstopt;
4026 				toh->level = SOL_SOCKET;
4027 				toh->name = SCM_UCRED;
4028 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4029 				toh->status = 0;
4030 				dstopt += sizeof (struct T_opthdr);
4031 				(void) cred2ucred(cr, cpid, dstopt, rcr);
4032 				dstopt += ucredsize;
4033 				udi_size -= toh->len;
4034 			}
4035 
4036 			if (udp_bits.udpb_timestamp) {
4037 				struct	T_opthdr *toh;
4038 
4039 				toh = (struct T_opthdr *)dstopt;
4040 				toh->level = SOL_SOCKET;
4041 				toh->name = SCM_TIMESTAMP;
4042 				toh->len = sizeof (struct T_opthdr) +
4043 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4044 				toh->status = 0;
4045 				dstopt += sizeof (struct T_opthdr);
4046 				/* Align for gethrestime() */
4047 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
4048 				    sizeof (intptr_t));
4049 				gethrestime((timestruc_t *)dstopt);
4050 				dstopt = (char *)toh + toh->len;
4051 				udi_size -= toh->len;
4052 			}
4053 
4054 			/*
4055 			 * CAUTION:
4056 			 * Due to aligment issues
4057 			 * Processing of IP_RECVTTL option
4058 			 * should always be the last. Adding
4059 			 * any option processing after this will
4060 			 * cause alignment panic.
4061 			 */
4062 			if (udp_bits.udpb_recvttl) {
4063 				struct	T_opthdr *toh;
4064 				uint8_t	*dstptr;
4065 
4066 				toh = (struct T_opthdr *)dstopt;
4067 				toh->level = IPPROTO_IP;
4068 				toh->name = IP_RECVTTL;
4069 				toh->len = sizeof (struct T_opthdr) +
4070 				    sizeof (uint8_t);
4071 				toh->status = 0;
4072 				dstopt += sizeof (struct T_opthdr);
4073 				dstptr = (uint8_t *)dstopt;
4074 				*dstptr = ((ipha_t *)rptr)->ipha_ttl;
4075 				dstopt += sizeof (uint8_t);
4076 				udi_size -= toh->len;
4077 			}
4078 
4079 			/* Consumed all of allocated space */
4080 			ASSERT(udi_size == 0);
4081 		}
4082 	} else {
4083 		sin6_t *sin6;
4084 
4085 		/*
4086 		 * Handle both IPv4 and IPv6 packets for IPv6 sockets.
4087 		 *
4088 		 * Normally we only send up the address. If receiving of any
4089 		 * optional receive side information is enabled, we also send
4090 		 * that up as options.
4091 		 */
4092 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
4093 
4094 		if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
4095 		    IPPF_RTHDR|IPPF_IFINDEX)) {
4096 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4097 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4098 				size_t hlen;
4099 
4100 				UDP_STAT(us, udp_in_recvhopopts);
4101 				hlen = copy_hop_opts(&ipp, NULL);
4102 				if (hlen == 0)
4103 					ipp.ipp_fields &= ~IPPF_HOPOPTS;
4104 				udi_size += hlen;
4105 			}
4106 			if (((udp_bits.udpb_ipv6_recvdstopts) ||
4107 			    udp_bits.udpb_old_ipv6_recvdstopts) &&
4108 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4109 				udi_size += sizeof (struct T_opthdr) +
4110 				    ipp.ipp_dstoptslen;
4111 				UDP_STAT(us, udp_in_recvdstopts);
4112 			}
4113 			if ((((udp_bits.udpb_ipv6_recvdstopts) &&
4114 			    udp_bits.udpb_ipv6_recvrthdr &&
4115 			    (ipp.ipp_fields & IPPF_RTHDR)) ||
4116 			    (udp_bits.udpb_ipv6_recvrthdrdstopts)) &&
4117 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4118 				udi_size += sizeof (struct T_opthdr) +
4119 				    ipp.ipp_rtdstoptslen;
4120 				UDP_STAT(us, udp_in_recvrtdstopts);
4121 			}
4122 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4123 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4124 				udi_size += sizeof (struct T_opthdr) +
4125 				    ipp.ipp_rthdrlen;
4126 				UDP_STAT(us, udp_in_recvrthdr);
4127 			}
4128 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4129 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4130 				udi_size += sizeof (struct T_opthdr) +
4131 				    sizeof (struct in6_pktinfo);
4132 				UDP_STAT(us, udp_in_recvpktinfo);
4133 			}
4134 
4135 		}
4136 		if ((udp_bits.udpb_recvucred) &&
4137 		    (cr = DB_CRED(mp)) != NULL) {
4138 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4139 			cpid = DB_CPID(mp);
4140 			UDP_STAT(us, udp_in_recvucred);
4141 		}
4142 
4143 		/*
4144 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4145 		 * buffer. Since gethrestime() expects a pointer aligned
4146 		 * argument, we allocate space necessary for extra
4147 		 * alignment (even though it might not be used).
4148 		 */
4149 		if (udp_bits.udpb_timestamp) {
4150 			udi_size += sizeof (struct T_opthdr) +
4151 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4152 			UDP_STAT(us, udp_in_timestamp);
4153 		}
4154 
4155 		if (udp_bits.udpb_ipv6_recvhoplimit) {
4156 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4157 			UDP_STAT(us, udp_in_recvhoplimit);
4158 		}
4159 
4160 		if (udp_bits.udpb_ipv6_recvtclass) {
4161 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4162 			UDP_STAT(us, udp_in_recvtclass);
4163 		}
4164 
4165 		mp1 = allocb(udi_size, BPRI_MED);
4166 		if (mp1 == NULL) {
4167 			freemsg(mp);
4168 			if (options_mp != NULL)
4169 				freeb(options_mp);
4170 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
4171 			return;
4172 		}
4173 		mp1->b_cont = mp;
4174 		mp = mp1;
4175 		mp->b_datap->db_type = M_PROTO;
4176 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4177 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4178 		tudi->PRIM_type = T_UNITDATA_IND;
4179 		tudi->SRC_length = sizeof (sin6_t);
4180 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4181 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4182 		    sizeof (sin6_t);
4183 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
4184 		tudi->OPT_length = udi_size;
4185 		sin6 = (sin6_t *)&tudi[1];
4186 		if (ipversion == IPV4_VERSION) {
4187 			in6_addr_t v6dst;
4188 
4189 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src,
4190 			    &sin6->sin6_addr);
4191 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst,
4192 			    &v6dst);
4193 			sin6->sin6_flowinfo = 0;
4194 			sin6->sin6_scope_id = 0;
4195 			sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst,
4196 			    connp->conn_zoneid, us->us_netstack);
4197 		} else {
4198 			sin6->sin6_addr = ip6h->ip6_src;
4199 			/* No sin6_flowinfo per API */
4200 			sin6->sin6_flowinfo = 0;
4201 			/* For link-scope source pass up scope id */
4202 			if ((ipp.ipp_fields & IPPF_IFINDEX) &&
4203 			    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
4204 				sin6->sin6_scope_id = ipp.ipp_ifindex;
4205 			else
4206 				sin6->sin6_scope_id = 0;
4207 			sin6->__sin6_src_id = ip_srcid_find_addr(
4208 			    &ip6h->ip6_dst, connp->conn_zoneid,
4209 			    us->us_netstack);
4210 		}
4211 		sin6->sin6_port = udpha->uha_src_port;
4212 		sin6->sin6_family = udp->udp_family;
4213 
4214 		if (udi_size != 0) {
4215 			uchar_t *dstopt;
4216 
4217 			dstopt = (uchar_t *)&sin6[1];
4218 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4219 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4220 				struct T_opthdr *toh;
4221 				struct in6_pktinfo *pkti;
4222 
4223 				toh = (struct T_opthdr *)dstopt;
4224 				toh->level = IPPROTO_IPV6;
4225 				toh->name = IPV6_PKTINFO;
4226 				toh->len = sizeof (struct T_opthdr) +
4227 				    sizeof (*pkti);
4228 				toh->status = 0;
4229 				dstopt += sizeof (struct T_opthdr);
4230 				pkti = (struct in6_pktinfo *)dstopt;
4231 				if (ipversion == IPV6_VERSION)
4232 					pkti->ipi6_addr = ip6h->ip6_dst;
4233 				else
4234 					IN6_IPADDR_TO_V4MAPPED(
4235 					    ((ipha_t *)rptr)->ipha_dst,
4236 					    &pkti->ipi6_addr);
4237 				pkti->ipi6_ifindex = ipp.ipp_ifindex;
4238 				dstopt += sizeof (*pkti);
4239 				udi_size -= toh->len;
4240 			}
4241 			if (udp_bits.udpb_ipv6_recvhoplimit) {
4242 				struct T_opthdr *toh;
4243 
4244 				toh = (struct T_opthdr *)dstopt;
4245 				toh->level = IPPROTO_IPV6;
4246 				toh->name = IPV6_HOPLIMIT;
4247 				toh->len = sizeof (struct T_opthdr) +
4248 				    sizeof (uint_t);
4249 				toh->status = 0;
4250 				dstopt += sizeof (struct T_opthdr);
4251 				if (ipversion == IPV6_VERSION)
4252 					*(uint_t *)dstopt = ip6h->ip6_hops;
4253 				else
4254 					*(uint_t *)dstopt =
4255 					    ((ipha_t *)rptr)->ipha_ttl;
4256 				dstopt += sizeof (uint_t);
4257 				udi_size -= toh->len;
4258 			}
4259 			if (udp_bits.udpb_ipv6_recvtclass) {
4260 				struct T_opthdr *toh;
4261 
4262 				toh = (struct T_opthdr *)dstopt;
4263 				toh->level = IPPROTO_IPV6;
4264 				toh->name = IPV6_TCLASS;
4265 				toh->len = sizeof (struct T_opthdr) +
4266 				    sizeof (uint_t);
4267 				toh->status = 0;
4268 				dstopt += sizeof (struct T_opthdr);
4269 				if (ipversion == IPV6_VERSION) {
4270 					*(uint_t *)dstopt =
4271 					    IPV6_FLOW_TCLASS(ip6h->ip6_flow);
4272 				} else {
4273 					ipha_t *ipha = (ipha_t *)rptr;
4274 					*(uint_t *)dstopt =
4275 					    ipha->ipha_type_of_service;
4276 				}
4277 				dstopt += sizeof (uint_t);
4278 				udi_size -= toh->len;
4279 			}
4280 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4281 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4282 				size_t hlen;
4283 
4284 				hlen = copy_hop_opts(&ipp, dstopt);
4285 				dstopt += hlen;
4286 				udi_size -= hlen;
4287 			}
4288 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4289 			    (udp_bits.udpb_ipv6_recvrthdr) &&
4290 			    (ipp.ipp_fields & IPPF_RTHDR) &&
4291 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4292 				struct T_opthdr *toh;
4293 
4294 				toh = (struct T_opthdr *)dstopt;
4295 				toh->level = IPPROTO_IPV6;
4296 				toh->name = IPV6_DSTOPTS;
4297 				toh->len = sizeof (struct T_opthdr) +
4298 				    ipp.ipp_rtdstoptslen;
4299 				toh->status = 0;
4300 				dstopt += sizeof (struct T_opthdr);
4301 				bcopy(ipp.ipp_rtdstopts, dstopt,
4302 				    ipp.ipp_rtdstoptslen);
4303 				dstopt += ipp.ipp_rtdstoptslen;
4304 				udi_size -= toh->len;
4305 			}
4306 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4307 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4308 				struct T_opthdr *toh;
4309 
4310 				toh = (struct T_opthdr *)dstopt;
4311 				toh->level = IPPROTO_IPV6;
4312 				toh->name = IPV6_RTHDR;
4313 				toh->len = sizeof (struct T_opthdr) +
4314 				    ipp.ipp_rthdrlen;
4315 				toh->status = 0;
4316 				dstopt += sizeof (struct T_opthdr);
4317 				bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
4318 				dstopt += ipp.ipp_rthdrlen;
4319 				udi_size -= toh->len;
4320 			}
4321 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4322 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4323 				struct T_opthdr *toh;
4324 
4325 				toh = (struct T_opthdr *)dstopt;
4326 				toh->level = IPPROTO_IPV6;
4327 				toh->name = IPV6_DSTOPTS;
4328 				toh->len = sizeof (struct T_opthdr) +
4329 				    ipp.ipp_dstoptslen;
4330 				toh->status = 0;
4331 				dstopt += sizeof (struct T_opthdr);
4332 				bcopy(ipp.ipp_dstopts, dstopt,
4333 				    ipp.ipp_dstoptslen);
4334 				dstopt += ipp.ipp_dstoptslen;
4335 				udi_size -= toh->len;
4336 			}
4337 			if (cr != NULL) {
4338 				struct T_opthdr *toh;
4339 
4340 				toh = (struct T_opthdr *)dstopt;
4341 				toh->level = SOL_SOCKET;
4342 				toh->name = SCM_UCRED;
4343 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4344 				toh->status = 0;
4345 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
4346 				dstopt += toh->len;
4347 				udi_size -= toh->len;
4348 			}
4349 			if (udp_bits.udpb_timestamp) {
4350 				struct	T_opthdr *toh;
4351 
4352 				toh = (struct T_opthdr *)dstopt;
4353 				toh->level = SOL_SOCKET;
4354 				toh->name = SCM_TIMESTAMP;
4355 				toh->len = sizeof (struct T_opthdr) +
4356 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4357 				toh->status = 0;
4358 				dstopt += sizeof (struct T_opthdr);
4359 				/* Align for gethrestime() */
4360 				dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
4361 				    sizeof (intptr_t));
4362 				gethrestime((timestruc_t *)dstopt);
4363 				dstopt = (uchar_t *)toh + toh->len;
4364 				udi_size -= toh->len;
4365 			}
4366 
4367 			/* Consumed all of allocated space */
4368 			ASSERT(udi_size == 0);
4369 		}
4370 #undef	sin6
4371 		/* No IP_RECVDSTADDR for IPv6. */
4372 	}
4373 
4374 	BUMP_MIB(&us->us_udp_mib, udpHCInDatagrams);
4375 	if (options_mp != NULL)
4376 		freeb(options_mp);
4377 
4378 	if (IPCL_IS_NONSTR(connp)) {
4379 		int error;
4380 
4381 		if ((*connp->conn_upcalls->su_recv)
4382 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
4383 		    NULL) < 0) {
4384 			mutex_enter(&udp->udp_recv_lock);
4385 			if (error == ENOSPC) {
4386 				/*
4387 				 * let's confirm while holding the lock
4388 				 */
4389 				if ((*connp->conn_upcalls->su_recv)
4390 				    (connp->conn_upper_handle, NULL, 0, 0,
4391 				    &error, NULL) < 0) {
4392 					if (error == ENOSPC) {
4393 						connp->conn_flow_cntrld =
4394 						    B_TRUE;
4395 					} else {
4396 						ASSERT(error == EOPNOTSUPP);
4397 					}
4398 				}
4399 				mutex_exit(&udp->udp_recv_lock);
4400 			} else {
4401 				ASSERT(error == EOPNOTSUPP);
4402 				udp_queue_fallback(udp, mp);
4403 			}
4404 		}
4405 	} else {
4406 		putnext(connp->conn_rq, mp);
4407 	}
4408 	ASSERT(MUTEX_NOT_HELD(&udp->udp_recv_lock));
4409 	return;
4410 
4411 tossit:
4412 	freemsg(mp);
4413 	if (options_mp != NULL)
4414 		freeb(options_mp);
4415 	BUMP_MIB(&us->us_udp_mib, udpInErrors);
4416 }
4417 
4418 /*
4419  * return SNMP stuff in buffer in mpdata. We don't hold any lock and report
4420  * information that can be changing beneath us.
4421  */
4422 mblk_t *
4423 udp_snmp_get(queue_t *q, mblk_t *mpctl)
4424 {
4425 	mblk_t			*mpdata;
4426 	mblk_t			*mp_conn_ctl;
4427 	mblk_t			*mp_attr_ctl;
4428 	mblk_t			*mp6_conn_ctl;
4429 	mblk_t			*mp6_attr_ctl;
4430 	mblk_t			*mp_conn_tail;
4431 	mblk_t			*mp_attr_tail;
4432 	mblk_t			*mp6_conn_tail;
4433 	mblk_t			*mp6_attr_tail;
4434 	struct opthdr		*optp;
4435 	mib2_udpEntry_t		ude;
4436 	mib2_udp6Entry_t	ude6;
4437 	mib2_transportMLPEntry_t mlp;
4438 	int			state;
4439 	zoneid_t		zoneid;
4440 	int			i;
4441 	connf_t			*connfp;
4442 	conn_t			*connp = Q_TO_CONN(q);
4443 	int			v4_conn_idx;
4444 	int			v6_conn_idx;
4445 	boolean_t		needattr;
4446 	udp_t			*udp;
4447 	ip_stack_t		*ipst = connp->conn_netstack->netstack_ip;
4448 	udp_stack_t		*us = connp->conn_netstack->netstack_udp;
4449 	mblk_t			*mp2ctl;
4450 
4451 	/*
4452 	 * make a copy of the original message
4453 	 */
4454 	mp2ctl = copymsg(mpctl);
4455 
4456 	mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL;
4457 	if (mpctl == NULL ||
4458 	    (mpdata = mpctl->b_cont) == NULL ||
4459 	    (mp_conn_ctl = copymsg(mpctl)) == NULL ||
4460 	    (mp_attr_ctl = copymsg(mpctl)) == NULL ||
4461 	    (mp6_conn_ctl = copymsg(mpctl)) == NULL ||
4462 	    (mp6_attr_ctl = copymsg(mpctl)) == NULL) {
4463 		freemsg(mp_conn_ctl);
4464 		freemsg(mp_attr_ctl);
4465 		freemsg(mp6_conn_ctl);
4466 		freemsg(mpctl);
4467 		freemsg(mp2ctl);
4468 		return (0);
4469 	}
4470 
4471 	zoneid = connp->conn_zoneid;
4472 
4473 	/* fixed length structure for IPv4 and IPv6 counters */
4474 	SET_MIB(us->us_udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t));
4475 	SET_MIB(us->us_udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t));
4476 	/* synchronize 64- and 32-bit counters */
4477 	SYNC32_MIB(&us->us_udp_mib, udpInDatagrams, udpHCInDatagrams);
4478 	SYNC32_MIB(&us->us_udp_mib, udpOutDatagrams, udpHCOutDatagrams);
4479 
4480 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4481 	optp->level = MIB2_UDP;
4482 	optp->name = 0;
4483 	(void) snmp_append_data(mpdata, (char *)&us->us_udp_mib,
4484 	    sizeof (us->us_udp_mib));
4485 	optp->len = msgdsize(mpdata);
4486 	qreply(q, mpctl);
4487 
4488 	mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL;
4489 	v4_conn_idx = v6_conn_idx = 0;
4490 
4491 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4492 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4493 		connp = NULL;
4494 
4495 		while ((connp = ipcl_get_next_conn(connfp, connp,
4496 		    IPCL_UDPCONN))) {
4497 			udp = connp->conn_udp;
4498 			if (zoneid != connp->conn_zoneid)
4499 				continue;
4500 
4501 			/*
4502 			 * Note that the port numbers are sent in
4503 			 * host byte order
4504 			 */
4505 
4506 			if (udp->udp_state == TS_UNBND)
4507 				state = MIB2_UDP_unbound;
4508 			else if (udp->udp_state == TS_IDLE)
4509 				state = MIB2_UDP_idle;
4510 			else if (udp->udp_state == TS_DATA_XFER)
4511 				state = MIB2_UDP_connected;
4512 			else
4513 				state = MIB2_UDP_unknown;
4514 
4515 			needattr = B_FALSE;
4516 			bzero(&mlp, sizeof (mlp));
4517 			if (connp->conn_mlp_type != mlptSingle) {
4518 				if (connp->conn_mlp_type == mlptShared ||
4519 				    connp->conn_mlp_type == mlptBoth)
4520 					mlp.tme_flags |= MIB2_TMEF_SHARED;
4521 				if (connp->conn_mlp_type == mlptPrivate ||
4522 				    connp->conn_mlp_type == mlptBoth)
4523 					mlp.tme_flags |= MIB2_TMEF_PRIVATE;
4524 				needattr = B_TRUE;
4525 			}
4526 
4527 			/*
4528 			 * Create an IPv4 table entry for IPv4 entries and also
4529 			 * any IPv6 entries which are bound to in6addr_any
4530 			 * (i.e. anything a IPv4 peer could connect/send to).
4531 			 */
4532 			if (udp->udp_ipversion == IPV4_VERSION ||
4533 			    (udp->udp_state <= TS_IDLE &&
4534 			    IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) {
4535 				ude.udpEntryInfo.ue_state = state;
4536 				/*
4537 				 * If in6addr_any this will set it to
4538 				 * INADDR_ANY
4539 				 */
4540 				ude.udpLocalAddress =
4541 				    V4_PART_OF_V6(udp->udp_v6src);
4542 				ude.udpLocalPort = ntohs(udp->udp_port);
4543 				if (udp->udp_state == TS_DATA_XFER) {
4544 					/*
4545 					 * Can potentially get here for
4546 					 * v6 socket if another process
4547 					 * (say, ping) has just done a
4548 					 * sendto(), changing the state
4549 					 * from the TS_IDLE above to
4550 					 * TS_DATA_XFER by the time we hit
4551 					 * this part of the code.
4552 					 */
4553 					ude.udpEntryInfo.ue_RemoteAddress =
4554 					    V4_PART_OF_V6(udp->udp_v6dst);
4555 					ude.udpEntryInfo.ue_RemotePort =
4556 					    ntohs(udp->udp_dstport);
4557 				} else {
4558 					ude.udpEntryInfo.ue_RemoteAddress = 0;
4559 					ude.udpEntryInfo.ue_RemotePort = 0;
4560 				}
4561 
4562 				/*
4563 				 * We make the assumption that all udp_t
4564 				 * structs will be created within an address
4565 				 * region no larger than 32-bits.
4566 				 */
4567 				ude.udpInstance = (uint32_t)(uintptr_t)udp;
4568 				ude.udpCreationProcess =
4569 				    (udp->udp_open_pid < 0) ?
4570 				    MIB2_UNKNOWN_PROCESS :
4571 				    udp->udp_open_pid;
4572 				ude.udpCreationTime = udp->udp_open_time;
4573 
4574 				(void) snmp_append_data2(mp_conn_ctl->b_cont,
4575 				    &mp_conn_tail, (char *)&ude, sizeof (ude));
4576 				mlp.tme_connidx = v4_conn_idx++;
4577 				if (needattr)
4578 					(void) snmp_append_data2(
4579 					    mp_attr_ctl->b_cont, &mp_attr_tail,
4580 					    (char *)&mlp, sizeof (mlp));
4581 			}
4582 			if (udp->udp_ipversion == IPV6_VERSION) {
4583 				ude6.udp6EntryInfo.ue_state  = state;
4584 				ude6.udp6LocalAddress = udp->udp_v6src;
4585 				ude6.udp6LocalPort = ntohs(udp->udp_port);
4586 				ude6.udp6IfIndex = udp->udp_bound_if;
4587 				if (udp->udp_state == TS_DATA_XFER) {
4588 					ude6.udp6EntryInfo.ue_RemoteAddress =
4589 					    udp->udp_v6dst;
4590 					ude6.udp6EntryInfo.ue_RemotePort =
4591 					    ntohs(udp->udp_dstport);
4592 				} else {
4593 					ude6.udp6EntryInfo.ue_RemoteAddress =
4594 					    sin6_null.sin6_addr;
4595 					ude6.udp6EntryInfo.ue_RemotePort = 0;
4596 				}
4597 				/*
4598 				 * We make the assumption that all udp_t
4599 				 * structs will be created within an address
4600 				 * region no larger than 32-bits.
4601 				 */
4602 				ude6.udp6Instance = (uint32_t)(uintptr_t)udp;
4603 				ude6.udp6CreationProcess =
4604 				    (udp->udp_open_pid < 0) ?
4605 				    MIB2_UNKNOWN_PROCESS :
4606 				    udp->udp_open_pid;
4607 				ude6.udp6CreationTime = udp->udp_open_time;
4608 
4609 				(void) snmp_append_data2(mp6_conn_ctl->b_cont,
4610 				    &mp6_conn_tail, (char *)&ude6,
4611 				    sizeof (ude6));
4612 				mlp.tme_connidx = v6_conn_idx++;
4613 				if (needattr)
4614 					(void) snmp_append_data2(
4615 					    mp6_attr_ctl->b_cont,
4616 					    &mp6_attr_tail, (char *)&mlp,
4617 					    sizeof (mlp));
4618 			}
4619 		}
4620 	}
4621 
4622 	/* IPv4 UDP endpoints */
4623 	optp = (struct opthdr *)&mp_conn_ctl->b_rptr[
4624 	    sizeof (struct T_optmgmt_ack)];
4625 	optp->level = MIB2_UDP;
4626 	optp->name = MIB2_UDP_ENTRY;
4627 	optp->len = msgdsize(mp_conn_ctl->b_cont);
4628 	qreply(q, mp_conn_ctl);
4629 
4630 	/* table of MLP attributes... */
4631 	optp = (struct opthdr *)&mp_attr_ctl->b_rptr[
4632 	    sizeof (struct T_optmgmt_ack)];
4633 	optp->level = MIB2_UDP;
4634 	optp->name = EXPER_XPORT_MLP;
4635 	optp->len = msgdsize(mp_attr_ctl->b_cont);
4636 	if (optp->len == 0)
4637 		freemsg(mp_attr_ctl);
4638 	else
4639 		qreply(q, mp_attr_ctl);
4640 
4641 	/* IPv6 UDP endpoints */
4642 	optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[
4643 	    sizeof (struct T_optmgmt_ack)];
4644 	optp->level = MIB2_UDP6;
4645 	optp->name = MIB2_UDP6_ENTRY;
4646 	optp->len = msgdsize(mp6_conn_ctl->b_cont);
4647 	qreply(q, mp6_conn_ctl);
4648 
4649 	/* table of MLP attributes... */
4650 	optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[
4651 	    sizeof (struct T_optmgmt_ack)];
4652 	optp->level = MIB2_UDP6;
4653 	optp->name = EXPER_XPORT_MLP;
4654 	optp->len = msgdsize(mp6_attr_ctl->b_cont);
4655 	if (optp->len == 0)
4656 		freemsg(mp6_attr_ctl);
4657 	else
4658 		qreply(q, mp6_attr_ctl);
4659 
4660 	return (mp2ctl);
4661 }
4662 
4663 /*
4664  * Return 0 if invalid set request, 1 otherwise, including non-udp requests.
4665  * NOTE: Per MIB-II, UDP has no writable data.
4666  * TODO:  If this ever actually tries to set anything, it needs to be
4667  * to do the appropriate locking.
4668  */
4669 /* ARGSUSED */
4670 int
4671 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4672     uchar_t *ptr, int len)
4673 {
4674 	switch (level) {
4675 	case MIB2_UDP:
4676 		return (0);
4677 	default:
4678 		return (1);
4679 	}
4680 }
4681 
4682 static void
4683 udp_report_item(mblk_t *mp, udp_t *udp)
4684 {
4685 	char *state;
4686 	char addrbuf1[INET6_ADDRSTRLEN];
4687 	char addrbuf2[INET6_ADDRSTRLEN];
4688 	uint_t print_len, buf_len;
4689 
4690 	buf_len = mp->b_datap->db_lim - mp->b_wptr;
4691 	ASSERT(buf_len >= 0);
4692 	if (buf_len == 0)
4693 		return;
4694 
4695 	if (udp->udp_state == TS_UNBND)
4696 		state = "UNBOUND";
4697 	else if (udp->udp_state == TS_IDLE)
4698 		state = "IDLE";
4699 	else if (udp->udp_state == TS_DATA_XFER)
4700 		state = "CONNECTED";
4701 	else
4702 		state = "UnkState";
4703 	print_len = snprintf((char *)mp->b_wptr, buf_len,
4704 	    MI_COL_PTRFMT_STR "%4d %5u %s %s %5u %s\n",
4705 	    (void *)udp, udp->udp_connp->conn_zoneid, ntohs(udp->udp_port),
4706 	    inet_ntop(AF_INET6, &udp->udp_v6src, addrbuf1, sizeof (addrbuf1)),
4707 	    inet_ntop(AF_INET6, &udp->udp_v6dst, addrbuf2, sizeof (addrbuf2)),
4708 	    ntohs(udp->udp_dstport), state);
4709 	if (print_len < buf_len) {
4710 		mp->b_wptr += print_len;
4711 	} else {
4712 		mp->b_wptr += buf_len;
4713 	}
4714 }
4715 
4716 /* Report for ndd "udp_status" */
4717 /* ARGSUSED */
4718 static int
4719 udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4720 {
4721 	zoneid_t zoneid;
4722 	connf_t	*connfp;
4723 	conn_t	*connp = Q_TO_CONN(q);
4724 	udp_t	*udp = connp->conn_udp;
4725 	int	i;
4726 	udp_stack_t *us = udp->udp_us;
4727 	ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
4728 
4729 	/*
4730 	 * Because of the ndd constraint, at most we can have 64K buffer
4731 	 * to put in all UDP info.  So to be more efficient, just
4732 	 * allocate a 64K buffer here, assuming we need that large buffer.
4733 	 * This may be a problem as any user can read udp_status.  Therefore
4734 	 * we limit the rate of doing this using us_ndd_get_info_interval.
4735 	 * This should be OK as normal users should not do this too often.
4736 	 */
4737 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
4738 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
4739 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
4740 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
4741 			return (0);
4742 		}
4743 	}
4744 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
4745 		/* The following may work even if we cannot get a large buf. */
4746 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
4747 		return (0);
4748 	}
4749 	(void) mi_mpprintf(mp,
4750 	    "UDP     " MI_COL_HDRPAD_STR
4751 	/*   12345678[89ABCDEF] */
4752 	    " zone lport src addr        dest addr       port  state");
4753 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
4754 
4755 	zoneid = connp->conn_zoneid;
4756 
4757 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4758 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4759 		connp = NULL;
4760 
4761 		while ((connp = ipcl_get_next_conn(connfp, connp,
4762 		    IPCL_UDPCONN))) {
4763 			udp = connp->conn_udp;
4764 			if (zoneid != GLOBAL_ZONEID &&
4765 			    zoneid != connp->conn_zoneid)
4766 				continue;
4767 
4768 			udp_report_item(mp->b_cont, udp);
4769 		}
4770 	}
4771 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
4772 	return (0);
4773 }
4774 
4775 /*
4776  * This routine creates a T_UDERROR_IND message and passes it upstream.
4777  * The address and options are copied from the T_UNITDATA_REQ message
4778  * passed in mp.  This message is freed.
4779  */
4780 static void
4781 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen,
4782     t_scalar_t err)
4783 {
4784 	struct T_unitdata_req *tudr;
4785 	mblk_t	*mp1;
4786 	uchar_t	*optaddr;
4787 	t_scalar_t optlen;
4788 
4789 	if (DB_TYPE(mp) == M_DATA) {
4790 		ASSERT(destaddr != NULL && destlen != 0);
4791 		optaddr = NULL;
4792 		optlen = 0;
4793 	} else {
4794 		if ((mp->b_wptr < mp->b_rptr) ||
4795 		    (MBLKL(mp)) < sizeof (struct T_unitdata_req)) {
4796 			goto done;
4797 		}
4798 		tudr = (struct T_unitdata_req *)mp->b_rptr;
4799 		destaddr = mp->b_rptr + tudr->DEST_offset;
4800 		if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr ||
4801 		    destaddr + tudr->DEST_length < mp->b_rptr ||
4802 		    destaddr + tudr->DEST_length > mp->b_wptr) {
4803 			goto done;
4804 		}
4805 		optaddr = mp->b_rptr + tudr->OPT_offset;
4806 		if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr ||
4807 		    optaddr + tudr->OPT_length < mp->b_rptr ||
4808 		    optaddr + tudr->OPT_length > mp->b_wptr) {
4809 			goto done;
4810 		}
4811 		destlen = tudr->DEST_length;
4812 		optlen = tudr->OPT_length;
4813 	}
4814 
4815 	mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen,
4816 	    (char *)optaddr, optlen, err);
4817 	if (mp1 != NULL)
4818 		qreply(q, mp1);
4819 
4820 done:
4821 	freemsg(mp);
4822 }
4823 
4824 /*
4825  * This routine removes a port number association from a stream.  It
4826  * is called by udp_wput to handle T_UNBIND_REQ messages.
4827  */
4828 static void
4829 udp_tpi_unbind(queue_t *q, mblk_t *mp)
4830 {
4831 	conn_t	*connp = Q_TO_CONN(q);
4832 	int	error;
4833 
4834 	error = udp_do_unbind(connp);
4835 	if (error) {
4836 		if (error < 0)
4837 			udp_err_ack(q, mp, -error, 0);
4838 		else
4839 			udp_err_ack(q, mp, TSYSERR, error);
4840 		return;
4841 	}
4842 
4843 	mp = mi_tpi_ok_ack_alloc(mp);
4844 	ASSERT(mp != NULL);
4845 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4846 	qreply(q, mp);
4847 }
4848 
4849 /*
4850  * Don't let port fall into the privileged range.
4851  * Since the extra privileged ports can be arbitrary we also
4852  * ensure that we exclude those from consideration.
4853  * us->us_epriv_ports is not sorted thus we loop over it until
4854  * there are no changes.
4855  */
4856 static in_port_t
4857 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random)
4858 {
4859 	int i;
4860 	in_port_t nextport;
4861 	boolean_t restart = B_FALSE;
4862 	udp_stack_t *us = udp->udp_us;
4863 
4864 	if (random && udp_random_anon_port != 0) {
4865 		(void) random_get_pseudo_bytes((uint8_t *)&port,
4866 		    sizeof (in_port_t));
4867 		/*
4868 		 * Unless changed by a sys admin, the smallest anon port
4869 		 * is 32768 and the largest anon port is 65535.  It is
4870 		 * very likely (50%) for the random port to be smaller
4871 		 * than the smallest anon port.  When that happens,
4872 		 * add port % (anon port range) to the smallest anon
4873 		 * port to get the random port.  It should fall into the
4874 		 * valid anon port range.
4875 		 */
4876 		if (port < us->us_smallest_anon_port) {
4877 			port = us->us_smallest_anon_port +
4878 			    port % (us->us_largest_anon_port -
4879 			    us->us_smallest_anon_port);
4880 		}
4881 	}
4882 
4883 retry:
4884 	if (port < us->us_smallest_anon_port)
4885 		port = us->us_smallest_anon_port;
4886 
4887 	if (port > us->us_largest_anon_port) {
4888 		port = us->us_smallest_anon_port;
4889 		if (restart)
4890 			return (0);
4891 		restart = B_TRUE;
4892 	}
4893 
4894 	if (port < us->us_smallest_nonpriv_port)
4895 		port = us->us_smallest_nonpriv_port;
4896 
4897 	for (i = 0; i < us->us_num_epriv_ports; i++) {
4898 		if (port == us->us_epriv_ports[i]) {
4899 			port++;
4900 			/*
4901 			 * Make sure that the port is in the
4902 			 * valid range.
4903 			 */
4904 			goto retry;
4905 		}
4906 	}
4907 
4908 	if (is_system_labeled() &&
4909 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
4910 	    port, IPPROTO_UDP, B_TRUE)) != 0) {
4911 		port = nextport;
4912 		goto retry;
4913 	}
4914 
4915 	return (port);
4916 }
4917 
4918 static int
4919 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst)
4920 {
4921 	int err;
4922 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4923 	udp_t *udp = Q_TO_UDP(wq);
4924 	udp_stack_t	*us = udp->udp_us;
4925 
4926 	err = tsol_compute_label(DB_CREDDEF(mp, udp->udp_connp->conn_cred), dst,
4927 	    opt_storage, udp->udp_connp->conn_mac_exempt,
4928 	    us->us_netstack->netstack_ip);
4929 	if (err == 0) {
4930 		err = tsol_update_options(&udp->udp_ip_snd_options,
4931 		    &udp->udp_ip_snd_options_len, &udp->udp_label_len,
4932 		    opt_storage);
4933 	}
4934 	if (err != 0) {
4935 		DTRACE_PROBE4(
4936 		    tx__ip__log__info__updatelabel__udp,
4937 		    char *, "queue(1) failed to update options(2) on mp(3)",
4938 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4939 	} else {
4940 		IN6_IPADDR_TO_V4MAPPED(dst, &udp->udp_v6lastdst);
4941 	}
4942 	return (err);
4943 }
4944 
4945 static mblk_t *
4946 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port,
4947     uint_t srcid, int *error, boolean_t insert_spi, struct nmsghdr *msg,
4948     cred_t *cr, pid_t pid)
4949 {
4950 	udp_t		*udp = connp->conn_udp;
4951 	mblk_t		*mp1 = mp;
4952 	mblk_t		*mp2;
4953 	ipha_t		*ipha;
4954 	int		ip_hdr_length;
4955 	uint32_t 	ip_len;
4956 	udpha_t		*udpha;
4957 	boolean_t 	lock_held = B_FALSE;
4958 	in_port_t	uha_src_port;
4959 	udpattrs_t	attrs;
4960 	uchar_t		ip_snd_opt[IP_MAX_OPT_LENGTH];
4961 	uint32_t	ip_snd_opt_len = 0;
4962 	ip4_pkt_t  	pktinfo;
4963 	ip4_pkt_t  	*pktinfop = &pktinfo;
4964 	ip_opt_info_t	optinfo;
4965 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
4966 	udp_stack_t	*us = udp->udp_us;
4967 	ipsec_stack_t	*ipss = ipst->ips_netstack->netstack_ipsec;
4968 	queue_t		*q = connp->conn_wq;
4969 	ire_t		*ire;
4970 
4971 
4972 	*error = 0;
4973 	pktinfop->ip4_ill_index = 0;
4974 	pktinfop->ip4_addr = INADDR_ANY;
4975 	optinfo.ip_opt_flags = 0;
4976 	optinfo.ip_opt_ill_index = 0;
4977 
4978 	if (v4dst == INADDR_ANY)
4979 		v4dst = htonl(INADDR_LOOPBACK);
4980 
4981 	/*
4982 	 * If options passed in, feed it for verification and handling
4983 	 */
4984 	attrs.udpattr_credset = B_FALSE;
4985 	if (IPCL_IS_NONSTR(connp)) {
4986 		if (msg->msg_controllen != 0) {
4987 			attrs.udpattr_ipp4 = pktinfop;
4988 			attrs.udpattr_mb = mp;
4989 
4990 			rw_enter(&udp->udp_rwlock, RW_WRITER);
4991 			*error = process_auxiliary_options(connp,
4992 			    msg->msg_control, msg->msg_controllen,
4993 			    &attrs, &udp_opt_obj, udp_opt_set);
4994 			rw_exit(&udp->udp_rwlock);
4995 			if (*error)
4996 				goto done;
4997 		}
4998 	} else {
4999 		if (DB_TYPE(mp) != M_DATA) {
5000 			mp1 = mp->b_cont;
5001 			if (((struct T_unitdata_req *)
5002 			    mp->b_rptr)->OPT_length != 0) {
5003 				attrs.udpattr_ipp4 = pktinfop;
5004 				attrs.udpattr_mb = mp;
5005 				if (udp_unitdata_opt_process(q, mp, error,
5006 				    &attrs) < 0)
5007 					goto done;
5008 				/*
5009 				 * Note: success in processing options.
5010 				 * mp option buffer represented by
5011 				 * OPT_length/offset now potentially modified
5012 				 * and contain option setting results
5013 				 */
5014 				ASSERT(*error == 0);
5015 			}
5016 		}
5017 	}
5018 
5019 	/* mp1 points to the M_DATA mblk carrying the packet */
5020 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
5021 
5022 	/*
5023 	 * Determine whether we need to mark the mblk with the user's
5024 	 * credentials.
5025 	 */
5026 	ire = connp->conn_ire_cache;
5027 	if (is_system_labeled() || CLASSD(v4dst) || (ire == NULL) ||
5028 	    (ire->ire_addr != v4dst) ||
5029 	    (ire->ire_type & (IRE_BROADCAST | IRE_LOCAL | IRE_LOOPBACK))) {
5030 		if (cr != NULL && DB_CRED(mp) == NULL)
5031 			msg_setcredpid(mp, cr, pid);
5032 	}
5033 
5034 	rw_enter(&udp->udp_rwlock, RW_READER);
5035 	lock_held = B_TRUE;
5036 	/*
5037 	 * Check if our saved options are valid; update if not.
5038 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
5039 	 * to different destination may require different labels,
5040 	 * or worse, UDP packets to same IP address may require
5041 	 * different labels due to use of shared all-zones address.
5042 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
5043 	 * and ip_snd_options_len are consistent for the current
5044 	 * destination and are updated atomically.
5045 	 */
5046 	mutex_enter(&connp->conn_lock);
5047 	if (is_system_labeled()) {
5048 		/* Using UDP MLP requires SCM_UCRED from user */
5049 		if (connp->conn_mlp_type != mlptSingle &&
5050 		    !attrs.udpattr_credset) {
5051 			mutex_exit(&connp->conn_lock);
5052 			DTRACE_PROBE4(
5053 			    tx__ip__log__info__output__udp,
5054 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
5055 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
5056 			*error = ECONNREFUSED;
5057 			goto done;
5058 		}
5059 		/*
5060 		 * update label option for this UDP socket if
5061 		 * - the destination has changed, or
5062 		 * - the UDP socket is MLP
5063 		 */
5064 		if ((!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
5065 		    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
5066 		    connp->conn_mlp_type != mlptSingle) &&
5067 		    (*error = udp_update_label(q, mp, v4dst)) != 0) {
5068 			mutex_exit(&connp->conn_lock);
5069 			goto done;
5070 		}
5071 	}
5072 	if (udp->udp_ip_snd_options_len > 0) {
5073 		ip_snd_opt_len = udp->udp_ip_snd_options_len;
5074 		bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
5075 	}
5076 	mutex_exit(&connp->conn_lock);
5077 
5078 	/* Add an IP header */
5079 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len +
5080 	    (insert_spi ? sizeof (uint32_t) : 0);
5081 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
5082 	if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) ||
5083 	    !OK_32PTR(ipha)) {
5084 		mp2 = allocb(ip_hdr_length + us->us_wroff_extra, BPRI_LO);
5085 		if (mp2 == NULL) {
5086 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5087 			    "udp_wput_end: q %p (%S)", q, "allocbfail2");
5088 			*error = ENOMEM;
5089 			goto done;
5090 		}
5091 		mp2->b_wptr = DB_LIM(mp2);
5092 		mp2->b_cont = mp1;
5093 		mp1 = mp2;
5094 		if (DB_TYPE(mp) != M_DATA)
5095 			mp->b_cont = mp1;
5096 		else
5097 			mp = mp1;
5098 
5099 		ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length);
5100 	}
5101 	ip_hdr_length -= (UDPH_SIZE + (insert_spi ? sizeof (uint32_t) : 0));
5102 #ifdef	_BIG_ENDIAN
5103 	/* Set version, header length, and tos */
5104 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5105 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
5106 	    udp->udp_type_of_service);
5107 	/* Set ttl and protocol */
5108 	*(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP;
5109 #else
5110 	/* Set version, header length, and tos */
5111 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5112 	    ((udp->udp_type_of_service << 8) |
5113 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
5114 	/* Set ttl and protocol */
5115 	*(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl;
5116 #endif
5117 	if (pktinfop->ip4_addr != INADDR_ANY) {
5118 		ipha->ipha_src = pktinfop->ip4_addr;
5119 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
5120 	} else {
5121 		/*
5122 		 * Copy our address into the packet.  If this is zero,
5123 		 * first look at __sin6_src_id for a hint. If we leave the
5124 		 * source as INADDR_ANY then ip will fill in the real source
5125 		 * address.
5126 		 */
5127 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src);
5128 		if (srcid != 0 && ipha->ipha_src == INADDR_ANY) {
5129 			in6_addr_t v6src;
5130 
5131 			ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid,
5132 			    us->us_netstack);
5133 			IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src);
5134 		}
5135 	}
5136 	uha_src_port = udp->udp_port;
5137 	if (ip_hdr_length == IP_SIMPLE_HDR_LENGTH) {
5138 		rw_exit(&udp->udp_rwlock);
5139 		lock_held = B_FALSE;
5140 	}
5141 
5142 	if (pktinfop->ip4_ill_index != 0) {
5143 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
5144 	}
5145 
5146 	ipha->ipha_fragment_offset_and_flags = 0;
5147 	ipha->ipha_ident = 0;
5148 
5149 	mp1->b_rptr = (uchar_t *)ipha;
5150 
5151 	ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <=
5152 	    (uintptr_t)UINT_MAX);
5153 
5154 	/* Determine length of packet */
5155 	ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha);
5156 	if ((mp2 = mp1->b_cont) != NULL) {
5157 		do {
5158 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
5159 			ip_len += (uint32_t)MBLKL(mp2);
5160 		} while ((mp2 = mp2->b_cont) != NULL);
5161 	}
5162 	/*
5163 	 * If the size of the packet is greater than the maximum allowed by
5164 	 * ip, return an error. Passing this down could cause panics because
5165 	 * the size will have wrapped and be inconsistent with the msg size.
5166 	 */
5167 	if (ip_len > IP_MAXPACKET) {
5168 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5169 		    "udp_wput_end: q %p (%S)", q, "IP length exceeded");
5170 		*error = EMSGSIZE;
5171 		goto done;
5172 	}
5173 	ipha->ipha_length = htons((uint16_t)ip_len);
5174 	ip_len -= ip_hdr_length;
5175 	ip_len = htons((uint16_t)ip_len);
5176 	udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length);
5177 
5178 	/* Insert all-0s SPI now. */
5179 	if (insert_spi)
5180 		*((uint32_t *)(udpha + 1)) = 0;
5181 
5182 	/*
5183 	 * Copy in the destination address
5184 	 */
5185 	ipha->ipha_dst = v4dst;
5186 
5187 	/*
5188 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
5189 	 */
5190 	if (CLASSD(v4dst))
5191 		ipha->ipha_ttl = udp->udp_multicast_ttl;
5192 
5193 	udpha->uha_dst_port = port;
5194 	udpha->uha_src_port = uha_src_port;
5195 
5196 	if (ip_snd_opt_len > 0) {
5197 		uint32_t	cksum;
5198 
5199 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
5200 		lock_held = B_FALSE;
5201 		rw_exit(&udp->udp_rwlock);
5202 		/*
5203 		 * Massage source route putting first source route in ipha_dst.
5204 		 * Ignore the destination in T_unitdata_req.
5205 		 * Create a checksum adjustment for a source route, if any.
5206 		 */
5207 		cksum = ip_massage_options(ipha, us->us_netstack);
5208 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5209 		cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) +
5210 		    (ipha->ipha_dst & 0xFFFF);
5211 		if ((int)cksum < 0)
5212 			cksum--;
5213 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5214 		/*
5215 		 * IP does the checksum if uha_checksum is non-zero,
5216 		 * We make it easy for IP to include our pseudo header
5217 		 * by putting our length in uha_checksum.
5218 		 */
5219 		cksum += ip_len;
5220 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5221 		/* There might be a carry. */
5222 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5223 #ifdef _LITTLE_ENDIAN
5224 		if (us->us_do_checksum)
5225 			ip_len = (cksum << 16) | ip_len;
5226 #else
5227 		if (us->us_do_checksum)
5228 			ip_len = (ip_len << 16) | cksum;
5229 		else
5230 			ip_len <<= 16;
5231 #endif
5232 	} else {
5233 		/*
5234 		 * IP does the checksum if uha_checksum is non-zero,
5235 		 * We make it easy for IP to include our pseudo header
5236 		 * by putting our length in uha_checksum.
5237 		 */
5238 		if (us->us_do_checksum)
5239 			ip_len |= (ip_len << 16);
5240 #ifndef _LITTLE_ENDIAN
5241 		else
5242 			ip_len <<= 16;
5243 #endif
5244 	}
5245 	ASSERT(!lock_held);
5246 	/* Set UDP length and checksum */
5247 	*((uint32_t *)&udpha->uha_length) = ip_len;
5248 	if (DB_CRED(mp) != NULL)
5249 		mblk_setcred(mp1, DB_CRED(mp));
5250 
5251 	if (DB_TYPE(mp) != M_DATA) {
5252 		ASSERT(mp != mp1);
5253 		freeb(mp);
5254 	}
5255 
5256 	/* mp has been consumed and we'll return success */
5257 	ASSERT(*error == 0);
5258 	mp = NULL;
5259 
5260 	/* We're done.  Pass the packet to ip. */
5261 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
5262 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5263 	    "udp_wput_end: q %p (%S)", q, "end");
5264 
5265 	if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 ||
5266 	    CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) ||
5267 	    connp->conn_dontroute ||
5268 	    connp->conn_nofailover_ill != NULL ||
5269 	    connp->conn_outgoing_ill != NULL || optinfo.ip_opt_flags != 0 ||
5270 	    optinfo.ip_opt_ill_index != 0 ||
5271 	    ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION ||
5272 	    IPP_ENABLED(IPP_LOCAL_OUT, ipst) ||
5273 	    ipst->ips_ip_g_mrouter != NULL) {
5274 		UDP_STAT(us, udp_ip_send);
5275 		ip_output_options(connp, mp1, connp->conn_wq, IP_WPUT,
5276 		    &optinfo);
5277 	} else {
5278 		udp_send_data(udp, connp->conn_wq, mp1, ipha);
5279 	}
5280 
5281 done:
5282 	if (lock_held)
5283 		rw_exit(&udp->udp_rwlock);
5284 	if (*error != 0) {
5285 		ASSERT(mp != NULL);
5286 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5287 	}
5288 	return (mp);
5289 }
5290 
5291 static void
5292 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha)
5293 {
5294 	conn_t	*connp = udp->udp_connp;
5295 	ipaddr_t src, dst;
5296 	ire_t	*ire;
5297 	ipif_t	*ipif = NULL;
5298 	mblk_t	*ire_fp_mp;
5299 	boolean_t retry_caching;
5300 	udp_stack_t *us = udp->udp_us;
5301 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5302 
5303 	dst = ipha->ipha_dst;
5304 	src = ipha->ipha_src;
5305 	ASSERT(ipha->ipha_ident == 0);
5306 
5307 	if (CLASSD(dst)) {
5308 		int err;
5309 
5310 		ipif = conn_get_held_ipif(connp,
5311 		    &connp->conn_multicast_ipif, &err);
5312 
5313 		if (ipif == NULL || ipif->ipif_isv6 ||
5314 		    (ipif->ipif_ill->ill_phyint->phyint_flags &
5315 		    PHYI_LOOPBACK)) {
5316 			if (ipif != NULL)
5317 				ipif_refrele(ipif);
5318 			UDP_STAT(us, udp_ip_send);
5319 			ip_output(connp, mp, q, IP_WPUT);
5320 			return;
5321 		}
5322 	}
5323 
5324 	retry_caching = B_FALSE;
5325 	mutex_enter(&connp->conn_lock);
5326 	ire = connp->conn_ire_cache;
5327 	ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT));
5328 
5329 	if (ire == NULL || ire->ire_addr != dst ||
5330 	    (ire->ire_marks & IRE_MARK_CONDEMNED)) {
5331 		retry_caching = B_TRUE;
5332 	} else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) {
5333 		ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr;
5334 
5335 		ASSERT(ipif != NULL);
5336 		if (stq_ill != ipif->ipif_ill && (stq_ill->ill_group == NULL ||
5337 		    stq_ill->ill_group != ipif->ipif_ill->ill_group))
5338 			retry_caching = B_TRUE;
5339 	}
5340 
5341 	if (!retry_caching) {
5342 		ASSERT(ire != NULL);
5343 		IRE_REFHOLD(ire);
5344 		mutex_exit(&connp->conn_lock);
5345 	} else {
5346 		boolean_t cached = B_FALSE;
5347 
5348 		connp->conn_ire_cache = NULL;
5349 		mutex_exit(&connp->conn_lock);
5350 
5351 		/* Release the old ire */
5352 		if (ire != NULL) {
5353 			IRE_REFRELE_NOTR(ire);
5354 			ire = NULL;
5355 		}
5356 
5357 		if (CLASSD(dst)) {
5358 			ASSERT(ipif != NULL);
5359 			ire = ire_ctable_lookup(dst, 0, 0, ipif,
5360 			    connp->conn_zoneid, MBLK_GETLABEL(mp),
5361 			    MATCH_IRE_ILL_GROUP, ipst);
5362 		} else {
5363 			ASSERT(ipif == NULL);
5364 			ire = ire_cache_lookup(dst, connp->conn_zoneid,
5365 			    MBLK_GETLABEL(mp), ipst);
5366 		}
5367 
5368 		if (ire == NULL) {
5369 			if (ipif != NULL)
5370 				ipif_refrele(ipif);
5371 			UDP_STAT(us, udp_ire_null);
5372 			ip_output(connp, mp, q, IP_WPUT);
5373 			return;
5374 		}
5375 		IRE_REFHOLD_NOTR(ire);
5376 
5377 		mutex_enter(&connp->conn_lock);
5378 		if (CONN_CACHE_IRE(connp) && connp->conn_ire_cache == NULL &&
5379 		    !(ire->ire_marks & IRE_MARK_CONDEMNED)) {
5380 			irb_t		*irb = ire->ire_bucket;
5381 
5382 			/*
5383 			 * IRE's created for non-connection oriented transports
5384 			 * are normally initialized with IRE_MARK_TEMPORARY set
5385 			 * in the ire_marks. These IRE's are preferentially
5386 			 * reaped when the hash chain length in the cache
5387 			 * bucket exceeds the maximum value specified in
5388 			 * ip[6]_ire_max_bucket_cnt. This can severely affect
5389 			 * UDP performance if IRE cache entries that we need
5390 			 * to reuse are continually removed. To remedy this,
5391 			 * when we cache the IRE in the conn_t, we remove the
5392 			 * IRE_MARK_TEMPORARY bit from the ire_marks if it was
5393 			 * set.
5394 			 */
5395 			if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5396 				rw_enter(&irb->irb_lock, RW_WRITER);
5397 				if (ire->ire_marks & IRE_MARK_TEMPORARY) {
5398 					ire->ire_marks &= ~IRE_MARK_TEMPORARY;
5399 					irb->irb_tmp_ire_cnt--;
5400 				}
5401 				rw_exit(&irb->irb_lock);
5402 			}
5403 			connp->conn_ire_cache = ire;
5404 			cached = B_TRUE;
5405 		}
5406 		mutex_exit(&connp->conn_lock);
5407 
5408 		/*
5409 		 * We can continue to use the ire but since it was not
5410 		 * cached, we should drop the extra reference.
5411 		 */
5412 		if (!cached)
5413 			IRE_REFRELE_NOTR(ire);
5414 	}
5415 	ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION);
5416 	ASSERT(!CLASSD(dst) || ipif != NULL);
5417 
5418 	/*
5419 	 * Check if we can take the fast-path.
5420 	 * Note that "incomplete" ire's (where the link-layer for next hop
5421 	 * is not resolved, or where the fast-path header in nce_fp_mp is not
5422 	 * available yet) are sent down the legacy (slow) path
5423 	 */
5424 	if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) ||
5425 	    (ire->ire_flags & RTF_MULTIRT) || (ire->ire_stq == NULL) ||
5426 	    (ire->ire_max_frag < ntohs(ipha->ipha_length)) ||
5427 	    ((ire->ire_nce == NULL) ||
5428 	    ((ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL)) ||
5429 	    connp->conn_nexthop_set || (MBLKL(ire_fp_mp) > MBLKHEAD(mp))) {
5430 		if (ipif != NULL)
5431 			ipif_refrele(ipif);
5432 		UDP_STAT(us, udp_ip_ire_send);
5433 		IRE_REFRELE(ire);
5434 		ip_output(connp, mp, q, IP_WPUT);
5435 		return;
5436 	}
5437 
5438 	if (src == INADDR_ANY && !connp->conn_unspec_src) {
5439 		if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC))
5440 			ipha->ipha_src = ipif->ipif_src_addr;
5441 		else
5442 			ipha->ipha_src = ire->ire_src_addr;
5443 	}
5444 
5445 	if (ipif != NULL)
5446 		ipif_refrele(ipif);
5447 
5448 	udp_xmit(connp->conn_wq, mp, ire, connp, connp->conn_zoneid);
5449 }
5450 
5451 static void
5452 udp_xmit(queue_t *q, mblk_t *mp, ire_t *ire, conn_t *connp, zoneid_t zoneid)
5453 {
5454 	ipaddr_t src, dst;
5455 	ill_t	*ill;
5456 	mblk_t	*ire_fp_mp;
5457 	uint_t	ire_fp_mp_len;
5458 	uint16_t *up;
5459 	uint32_t cksum, hcksum_txflags;
5460 	queue_t	*dev_q;
5461 	udp_t	*udp = connp->conn_udp;
5462 	ipha_t	*ipha = (ipha_t *)mp->b_rptr;
5463 	udp_stack_t	*us = udp->udp_us;
5464 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5465 	boolean_t	ll_multicast = B_FALSE;
5466 
5467 	dev_q = ire->ire_stq->q_next;
5468 	ASSERT(dev_q != NULL);
5469 
5470 	ill = ire_to_ill(ire);
5471 	ASSERT(ill != NULL);
5472 
5473 	/* is queue flow controlled? */
5474 	if (q->q_first != NULL || connp->conn_draining ||
5475 	    DEV_Q_FLOW_BLOCKED(dev_q)) {
5476 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
5477 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
5478 
5479 		if (ipst->ips_ip_output_queue)
5480 			(void) putq(connp->conn_wq, mp);
5481 		else
5482 			freemsg(mp);
5483 		ire_refrele(ire);
5484 		return;
5485 	}
5486 
5487 	ire_fp_mp = ire->ire_nce->nce_fp_mp;
5488 	ire_fp_mp_len = MBLKL(ire_fp_mp);
5489 	ASSERT(MBLKHEAD(mp) >= ire_fp_mp_len);
5490 
5491 	dst = ipha->ipha_dst;
5492 	src = ipha->ipha_src;
5493 
5494 
5495 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
5496 
5497 	ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1);
5498 #ifndef _BIG_ENDIAN
5499 	ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8);
5500 #endif
5501 
5502 	if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
5503 		ASSERT(ill->ill_hcksum_capab != NULL);
5504 		hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags;
5505 	} else {
5506 		hcksum_txflags = 0;
5507 	}
5508 
5509 	/* pseudo-header checksum (do it in parts for IP header checksum) */
5510 	cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF);
5511 
5512 	ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION);
5513 	up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH);
5514 	if (*up != 0) {
5515 		IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags,
5516 		    mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH,
5517 		    ntohs(ipha->ipha_length), cksum);
5518 
5519 		/* Software checksum? */
5520 		if (DB_CKSUMFLAGS(mp) == 0) {
5521 			UDP_STAT(us, udp_out_sw_cksum);
5522 			UDP_STAT_UPDATE(us, udp_out_sw_cksum_bytes,
5523 			    ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH);
5524 		}
5525 	}
5526 
5527 	if (!CLASSD(dst)) {
5528 		ipha->ipha_fragment_offset_and_flags |=
5529 		    (uint32_t)htons(ire->ire_frag_flag);
5530 	}
5531 
5532 	/* Calculate IP header checksum if hardware isn't capable */
5533 	if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) {
5534 		IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0],
5535 		    ((uint16_t *)ipha)[4]);
5536 	}
5537 
5538 	if (CLASSD(dst)) {
5539 		boolean_t ilm_exists;
5540 
5541 		ILM_WALKER_HOLD(ill);
5542 		ilm_exists = (ilm_lookup_ill(ill, dst, ALL_ZONES) != NULL);
5543 		ILM_WALKER_RELE(ill);
5544 		if (ilm_exists) {
5545 			ip_multicast_loopback(q, ill, mp,
5546 			    connp->conn_multicast_loop ? 0 :
5547 			    IP_FF_NO_MCAST_LOOP, zoneid);
5548 		}
5549 
5550 		/* If multicast TTL is 0 then we are done */
5551 		if (ipha->ipha_ttl == 0) {
5552 			freemsg(mp);
5553 			ire_refrele(ire);
5554 			return;
5555 		}
5556 		ll_multicast = B_TRUE;
5557 	}
5558 
5559 	ASSERT(DB_TYPE(ire_fp_mp) == M_DATA);
5560 	mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len;
5561 	bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len);
5562 
5563 	UPDATE_OB_PKT_COUNT(ire);
5564 	ire->ire_last_used_time = lbolt;
5565 
5566 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
5567 	UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
5568 	    ntohs(ipha->ipha_length));
5569 
5570 	DTRACE_PROBE4(ip4__physical__out__start,
5571 	    ill_t *, NULL, ill_t *, ill, ipha_t *, ipha, mblk_t *, mp);
5572 	FW_HOOKS(ipst->ips_ip4_physical_out_event,
5573 	    ipst->ips_ipv4firewall_physical_out, NULL, ill, ipha, mp, mp,
5574 	    ll_multicast, ipst);
5575 	DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);
5576 	if (ipst->ips_ipobs_enabled && mp != NULL) {
5577 		zoneid_t szone;
5578 
5579 		szone = ip_get_zoneid_v4(ipha->ipha_src, mp,
5580 		    ipst, ALL_ZONES);
5581 		ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone,
5582 		    ALL_ZONES, ill, IPV4_VERSION, ire_fp_mp_len, ipst);
5583 	}
5584 
5585 	if (mp != NULL) {
5586 		DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
5587 		    void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
5588 		    ipha_t *, ipha, ip6_t *, NULL, int, 0);
5589 
5590 		if (ILL_DIRECT_CAPABLE(ill)) {
5591 			ill_dld_direct_t *idd = &ill->ill_dld_capab->idc_direct;
5592 
5593 			(void) idd->idd_tx_df(idd->idd_tx_dh, mp,
5594 			    (uintptr_t)connp, 0);
5595 		} else {
5596 			putnext(ire->ire_stq, mp);
5597 		}
5598 	}
5599 	IRE_REFRELE(ire);
5600 }
5601 
5602 static boolean_t
5603 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst)
5604 {
5605 	udp_t *udp = Q_TO_UDP(wq);
5606 	int err;
5607 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
5608 	udp_stack_t		*us = udp->udp_us;
5609 
5610 	err = tsol_compute_label_v6(DB_CREDDEF(mp, udp->udp_connp->conn_cred),
5611 	    dst, opt_storage, udp->udp_connp->conn_mac_exempt,
5612 	    us->us_netstack->netstack_ip);
5613 	if (err == 0) {
5614 		err = tsol_update_sticky(&udp->udp_sticky_ipp,
5615 		    &udp->udp_label_len_v6, opt_storage);
5616 	}
5617 	if (err != 0) {
5618 		DTRACE_PROBE4(
5619 		    tx__ip__log__drop__updatelabel__udp6,
5620 		    char *, "queue(1) failed to update options(2) on mp(3)",
5621 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
5622 	} else {
5623 		udp->udp_v6lastdst = *dst;
5624 	}
5625 	return (err);
5626 }
5627 
5628 static int
5629 udp_send_connected(conn_t *connp, mblk_t *mp, struct nmsghdr *msg, cred_t *cr,
5630     pid_t pid)
5631 {
5632 	udp_t		*udp = connp->conn_udp;
5633 	udp_stack_t	*us = udp->udp_us;
5634 	ipaddr_t	v4dst;
5635 	in_port_t	dstport;
5636 	boolean_t	mapped_addr;
5637 	struct sockaddr_storage ss;
5638 	sin_t		*sin;
5639 	sin6_t		*sin6;
5640 	struct sockaddr	*addr;
5641 	socklen_t	addrlen;
5642 	int		error;
5643 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5644 
5645 	/* M_DATA for connected socket */
5646 
5647 	ASSERT(udp->udp_issocket || IPCL_IS_NONSTR(connp));
5648 	UDP_DBGSTAT(us, udp_data_conn);
5649 
5650 	mutex_enter(&connp->conn_lock);
5651 	if (udp->udp_state != TS_DATA_XFER) {
5652 		mutex_exit(&connp->conn_lock);
5653 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5654 		UDP_STAT(us, udp_out_err_notconn);
5655 		freemsg(mp);
5656 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5657 		    "udp_wput_end: connp %p (%S)", connp,
5658 		    "not-connected; address required");
5659 		return (EDESTADDRREQ);
5660 	}
5661 
5662 	mapped_addr = IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst);
5663 	if (mapped_addr)
5664 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst);
5665 
5666 	/* Initialize addr and addrlen as if they're passed in */
5667 	if (udp->udp_family == AF_INET) {
5668 		sin = (sin_t *)&ss;
5669 		sin->sin_family = AF_INET;
5670 		dstport = sin->sin_port = udp->udp_dstport;
5671 		ASSERT(mapped_addr);
5672 		sin->sin_addr.s_addr = v4dst;
5673 		addr = (struct sockaddr *)sin;
5674 		addrlen = sizeof (*sin);
5675 	} else {
5676 		sin6 = (sin6_t *)&ss;
5677 		sin6->sin6_family = AF_INET6;
5678 		dstport = sin6->sin6_port = udp->udp_dstport;
5679 		sin6->sin6_flowinfo = udp->udp_flowinfo;
5680 		sin6->sin6_addr = udp->udp_v6dst;
5681 		sin6->sin6_scope_id = 0;
5682 		sin6->__sin6_src_id = 0;
5683 		addr = (struct sockaddr *)sin6;
5684 		addrlen = sizeof (*sin6);
5685 	}
5686 	mutex_exit(&connp->conn_lock);
5687 
5688 	if (mapped_addr) {
5689 		/*
5690 		 * Handle both AF_INET and AF_INET6; the latter
5691 		 * for IPV4 mapped destination addresses.  Note
5692 		 * here that both addr and addrlen point to the
5693 		 * corresponding struct depending on the address
5694 		 * family of the socket.
5695 		 */
5696 		mp = udp_output_v4(connp, mp, v4dst, dstport, 0, &error,
5697 		    insert_spi, msg, cr, pid);
5698 	} else {
5699 		mp = udp_output_v6(connp, mp, sin6, &error, msg, cr, pid);
5700 	}
5701 	if (error == 0) {
5702 		ASSERT(mp == NULL);
5703 		return (0);
5704 	}
5705 
5706 	UDP_STAT(us, udp_out_err_output);
5707 	ASSERT(mp != NULL);
5708 	if (IPCL_IS_NONSTR(connp)) {
5709 		freemsg(mp);
5710 		return (error);
5711 	} else {
5712 		/* mp is freed by the following routine */
5713 		udp_ud_err(connp->conn_wq, mp, (uchar_t *)addr,
5714 		    (t_scalar_t)addrlen, (t_scalar_t)error);
5715 		return (0);
5716 	}
5717 }
5718 
5719 /* ARGSUSED */
5720 static int
5721 udp_send_not_connected(conn_t *connp,  mblk_t *mp, struct sockaddr *addr,
5722     socklen_t addrlen, struct nmsghdr *msg, cred_t *cr, pid_t pid)
5723 {
5724 
5725 	udp_t		*udp = connp->conn_udp;
5726 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
5727 	int		error = 0;
5728 	sin6_t		*sin6;
5729 	sin_t		*sin;
5730 	uint_t		srcid;
5731 	uint16_t	port;
5732 	ipaddr_t	v4dst;
5733 
5734 
5735 	ASSERT(addr != NULL);
5736 
5737 	switch (udp->udp_family) {
5738 	case AF_INET6:
5739 		sin6 = (sin6_t *)addr;
5740 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
5741 			/*
5742 			 * Destination is a non-IPv4-compatible IPv6 address.
5743 			 * Send out an IPv6 format packet.
5744 			 */
5745 			mp = udp_output_v6(connp, mp, sin6, &error, msg, cr,
5746 			    pid);
5747 			if (error != 0)
5748 				goto ud_error;
5749 
5750 			return (0);
5751 		}
5752 		/*
5753 		 * If the local address is not zero or a mapped address
5754 		 * return an error.  It would be possible to send an IPv4
5755 		 * packet but the response would never make it back to the
5756 		 * application since it is bound to a non-mapped address.
5757 		 */
5758 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) &&
5759 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
5760 			error = EADDRNOTAVAIL;
5761 			goto ud_error;
5762 		}
5763 		/* Send IPv4 packet without modifying udp_ipversion */
5764 		/* Extract port and ipaddr */
5765 		port = sin6->sin6_port;
5766 		IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst);
5767 		srcid = sin6->__sin6_src_id;
5768 		break;
5769 
5770 	case AF_INET:
5771 		sin = (sin_t *)addr;
5772 		/* Extract port and ipaddr */
5773 		port = sin->sin_port;
5774 		v4dst = sin->sin_addr.s_addr;
5775 		srcid = 0;
5776 		break;
5777 	}
5778 
5779 	mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error, insert_spi,
5780 	    msg, cr, pid);
5781 
5782 	if (error == 0) {
5783 		ASSERT(mp == NULL);
5784 		return (0);
5785 	}
5786 
5787 ud_error:
5788 	ASSERT(mp != NULL);
5789 
5790 	return (error);
5791 }
5792 
5793 /*
5794  * This routine handles all messages passed downstream.  It either
5795  * consumes the message or passes it downstream; it never queues a
5796  * a message.
5797  *
5798  * Also entry point for sockfs when udp is in "direct sockfs" mode.  This mode
5799  * is valid when we are directly beneath the stream head, and thus sockfs
5800  * is able to bypass STREAMS and directly call us, passing along the sockaddr
5801  * structure without the cumbersome T_UNITDATA_REQ interface for the case of
5802  * connected endpoints.
5803  */
5804 void
5805 udp_wput(queue_t *q, mblk_t *mp)
5806 {
5807 	conn_t		*connp = Q_TO_CONN(q);
5808 	udp_t		*udp = connp->conn_udp;
5809 	int		error = 0;
5810 	struct sockaddr	*addr;
5811 	socklen_t	addrlen;
5812 	udp_stack_t	*us = udp->udp_us;
5813 
5814 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START,
5815 	    "udp_wput_start: queue %p mp %p", q, mp);
5816 
5817 	/*
5818 	 * We directly handle several cases here: T_UNITDATA_REQ message
5819 	 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for connected
5820 	 * socket.
5821 	 */
5822 	switch (DB_TYPE(mp)) {
5823 	case M_DATA:
5824 		/*
5825 		 * Quick check for error cases. Checks will be done again
5826 		 * under the lock later on
5827 		 */
5828 		if (!udp->udp_direct_sockfs || udp->udp_state != TS_DATA_XFER) {
5829 			/* Not connected; address is required */
5830 			BUMP_MIB(&us->us_udp_mib, udpOutErrors);
5831 			UDP_STAT(us, udp_out_err_notconn);
5832 			freemsg(mp);
5833 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5834 			    "udp_wput_end: connp %p (%S)", connp,
5835 			    "not-connected; address required");
5836 			return;
5837 		}
5838 		(void) udp_send_connected(connp, mp, NULL, NULL, -1);
5839 		return;
5840 
5841 	case M_PROTO:
5842 	case M_PCPROTO: {
5843 		struct T_unitdata_req *tudr;
5844 
5845 		ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX);
5846 		tudr = (struct T_unitdata_req *)mp->b_rptr;
5847 
5848 		/* Handle valid T_UNITDATA_REQ here */
5849 		if (MBLKL(mp) >= sizeof (*tudr) &&
5850 		    ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) {
5851 			if (mp->b_cont == NULL) {
5852 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5853 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5854 				error = EPROTO;
5855 				goto ud_error;
5856 			}
5857 
5858 			if (!MBLKIN(mp, 0, tudr->DEST_offset +
5859 			    tudr->DEST_length)) {
5860 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5861 				    "udp_wput_end: q %p (%S)", q, "badaddr");
5862 				error = EADDRNOTAVAIL;
5863 				goto ud_error;
5864 			}
5865 			/*
5866 			 * If a port has not been bound to the stream, fail.
5867 			 * This is not a problem when sockfs is directly
5868 			 * above us, because it will ensure that the socket
5869 			 * is first bound before allowing data to be sent.
5870 			 */
5871 			if (udp->udp_state == TS_UNBND) {
5872 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5873 				    "udp_wput_end: q %p (%S)", q, "outstate");
5874 				error = EPROTO;
5875 				goto ud_error;
5876 			}
5877 			addr = (struct sockaddr *)
5878 			    &mp->b_rptr[tudr->DEST_offset];
5879 			addrlen = tudr->DEST_length;
5880 			if (tudr->OPT_length != 0)
5881 				UDP_STAT(us, udp_out_opt);
5882 			break;
5883 		}
5884 		/* FALLTHRU */
5885 	}
5886 	default:
5887 		udp_wput_other(q, mp);
5888 		return;
5889 	}
5890 	ASSERT(addr != NULL);
5891 
5892 	error = udp_send_not_connected(connp,  mp, addr, addrlen, NULL, NULL,
5893 	    -1);
5894 	if (error != 0) {
5895 ud_error:
5896 		UDP_STAT(us, udp_out_err_output);
5897 		ASSERT(mp != NULL);
5898 		/* mp is freed by the following routine */
5899 		udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
5900 		    (t_scalar_t)error);
5901 	}
5902 }
5903 
5904 /* ARGSUSED */
5905 static void
5906 udp_wput_fallback(queue_t *wq, mblk_t *mp)
5907 {
5908 #ifdef DEBUG
5909 	cmn_err(CE_CONT, "udp_wput_fallback: Message in fallback \n");
5910 #endif
5911 	freemsg(mp);
5912 }
5913 
5914 
5915 /*
5916  * udp_output_v6():
5917  * Assumes that udp_wput did some sanity checking on the destination
5918  * address.
5919  */
5920 static mblk_t *
5921 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error,
5922     struct nmsghdr *msg, cred_t *cr, pid_t pid)
5923 {
5924 	ip6_t		*ip6h;
5925 	ip6i_t		*ip6i;	/* mp1->b_rptr even if no ip6i_t */
5926 	mblk_t		*mp1 = mp;
5927 	mblk_t		*mp2;
5928 	int		udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
5929 	size_t		ip_len;
5930 	udpha_t		*udph;
5931 	udp_t		*udp = connp->conn_udp;
5932 	udp_stack_t	*us = udp->udp_us;
5933 	queue_t		*q = connp->conn_wq;
5934 	ip6_pkt_t	ipp_s;	/* For ancillary data options */
5935 	ip6_pkt_t	*ipp = &ipp_s;
5936 	ip6_pkt_t	*tipp;	/* temporary ipp */
5937 	uint32_t	csum = 0;
5938 	uint_t		ignore = 0;
5939 	uint_t		option_exists = 0, is_sticky = 0;
5940 	uint8_t		*cp;
5941 	uint8_t		*nxthdr_ptr;
5942 	in6_addr_t	ip6_dst;
5943 	udpattrs_t	attrs;
5944 	boolean_t	opt_present;
5945 	ip6_hbh_t	*hopoptsptr = NULL;
5946 	uint_t		hopoptslen = 0;
5947 	boolean_t	is_ancillary = B_FALSE;
5948 	size_t		sth_wroff = 0;
5949 	ire_t		*ire;
5950 
5951 	*error = 0;
5952 
5953 	/*
5954 	 * If the local address is a mapped address return
5955 	 * an error.
5956 	 * It would be possible to send an IPv6 packet but the
5957 	 * response would never make it back to the application
5958 	 * since it is bound to a mapped address.
5959 	 */
5960 	if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) {
5961 		*error = EADDRNOTAVAIL;
5962 		goto done;
5963 	}
5964 
5965 	ipp->ipp_fields = 0;
5966 	ipp->ipp_sticky_ignored = 0;
5967 
5968 	/*
5969 	 * If TPI options passed in, feed it for verification and handling
5970 	 */
5971 	attrs.udpattr_credset = B_FALSE;
5972 	opt_present = B_FALSE;
5973 	if (IPCL_IS_NONSTR(connp)) {
5974 		if (msg->msg_controllen != 0) {
5975 			attrs.udpattr_ipp6 = ipp;
5976 			attrs.udpattr_mb = mp;
5977 
5978 			rw_enter(&udp->udp_rwlock, RW_WRITER);
5979 			*error = process_auxiliary_options(connp,
5980 			    msg->msg_control, msg->msg_controllen,
5981 			    &attrs, &udp_opt_obj, udp_opt_set);
5982 			rw_exit(&udp->udp_rwlock);
5983 			if (*error)
5984 				goto done;
5985 			ASSERT(*error == 0);
5986 			opt_present = B_TRUE;
5987 		}
5988 	} else {
5989 		if (DB_TYPE(mp) != M_DATA) {
5990 			mp1 = mp->b_cont;
5991 			if (((struct T_unitdata_req *)
5992 			    mp->b_rptr)->OPT_length != 0) {
5993 				attrs.udpattr_ipp6 = ipp;
5994 				attrs.udpattr_mb = mp;
5995 				if (udp_unitdata_opt_process(q, mp, error,
5996 				    &attrs) < 0) {
5997 					goto done;
5998 				}
5999 				ASSERT(*error == 0);
6000 				opt_present = B_TRUE;
6001 			}
6002 		}
6003 	}
6004 
6005 	/*
6006 	 * Determine whether we need to mark the mblk with the user's
6007 	 * credentials.
6008 	 */
6009 	ire = connp->conn_ire_cache;
6010 	if (is_system_labeled() || IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) ||
6011 	    (ire == NULL) ||
6012 	    (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, &sin6->sin6_addr)) ||
6013 	    (ire->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) {
6014 		if (cr != NULL && DB_CRED(mp) == NULL)
6015 			msg_setcredpid(mp, cr, pid);
6016 	}
6017 
6018 	rw_enter(&udp->udp_rwlock, RW_READER);
6019 	ignore = ipp->ipp_sticky_ignored;
6020 
6021 	/* mp1 points to the M_DATA mblk carrying the packet */
6022 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
6023 
6024 	if (sin6->sin6_scope_id != 0 &&
6025 	    IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
6026 		/*
6027 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
6028 		 * option nor ancillary data.  It needs to be
6029 		 * explicitly set in options_exists.
6030 		 */
6031 		option_exists |= IPPF_SCOPE_ID;
6032 	}
6033 
6034 	/*
6035 	 * Compute the destination address
6036 	 */
6037 	ip6_dst = sin6->sin6_addr;
6038 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
6039 		ip6_dst = ipv6_loopback;
6040 
6041 	/*
6042 	 * If we're not going to the same destination as last time, then
6043 	 * recompute the label required.  This is done in a separate routine to
6044 	 * avoid blowing up our stack here.
6045 	 *
6046 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
6047 	 * to different destination may require different labels,
6048 	 * or worse, UDP packets to same IP address may require
6049 	 * different labels due to use of shared all-zones address.
6050 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
6051 	 * and sticky ipp_hopoptslen are consistent for the current
6052 	 * destination and are updated atomically.
6053 	 */
6054 	mutex_enter(&connp->conn_lock);
6055 	if (is_system_labeled()) {
6056 		/* Using UDP MLP requires SCM_UCRED from user */
6057 		if (connp->conn_mlp_type != mlptSingle &&
6058 		    !attrs.udpattr_credset) {
6059 			DTRACE_PROBE4(
6060 			    tx__ip__log__info__output__udp6,
6061 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
6062 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
6063 			*error = ECONNREFUSED;
6064 			rw_exit(&udp->udp_rwlock);
6065 			mutex_exit(&connp->conn_lock);
6066 			goto done;
6067 		}
6068 		/*
6069 		 * update label option for this UDP socket if
6070 		 * - the destination has changed, or
6071 		 * - the UDP socket is MLP
6072 		 */
6073 		if ((opt_present ||
6074 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst) ||
6075 		    connp->conn_mlp_type != mlptSingle) &&
6076 		    (*error = udp_update_label_v6(q, mp, &ip6_dst)) != 0) {
6077 			rw_exit(&udp->udp_rwlock);
6078 			mutex_exit(&connp->conn_lock);
6079 			goto done;
6080 		}
6081 	}
6082 
6083 	/*
6084 	 * If there's a security label here, then we ignore any options the
6085 	 * user may try to set.  We keep the peer's label as a hidden sticky
6086 	 * option. We make a private copy of this label before releasing the
6087 	 * lock so that label is kept consistent with the destination addr.
6088 	 */
6089 	if (udp->udp_label_len_v6 > 0) {
6090 		ignore &= ~IPPF_HOPOPTS;
6091 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
6092 	}
6093 
6094 	if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) {
6095 		/* No sticky options nor ancillary data. */
6096 		mutex_exit(&connp->conn_lock);
6097 		goto no_options;
6098 	}
6099 
6100 	/*
6101 	 * Go through the options figuring out where each is going to
6102 	 * come from and build two masks.  The first mask indicates if
6103 	 * the option exists at all.  The second mask indicates if the
6104 	 * option is sticky or ancillary.
6105 	 */
6106 	if (!(ignore & IPPF_HOPOPTS)) {
6107 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
6108 			option_exists |= IPPF_HOPOPTS;
6109 			udp_ip_hdr_len += ipp->ipp_hopoptslen;
6110 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
6111 			option_exists |= IPPF_HOPOPTS;
6112 			is_sticky |= IPPF_HOPOPTS;
6113 			ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0);
6114 			hopoptsptr = kmem_alloc(
6115 			    udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
6116 			if (hopoptsptr == NULL) {
6117 				*error = ENOMEM;
6118 				mutex_exit(&connp->conn_lock);
6119 				goto done;
6120 			}
6121 			hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen;
6122 			bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr,
6123 			    hopoptslen);
6124 			udp_ip_hdr_len += hopoptslen;
6125 		}
6126 	}
6127 	mutex_exit(&connp->conn_lock);
6128 
6129 	if (!(ignore & IPPF_RTHDR)) {
6130 		if (ipp->ipp_fields & IPPF_RTHDR) {
6131 			option_exists |= IPPF_RTHDR;
6132 			udp_ip_hdr_len += ipp->ipp_rthdrlen;
6133 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
6134 			option_exists |= IPPF_RTHDR;
6135 			is_sticky |= IPPF_RTHDR;
6136 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen;
6137 		}
6138 	}
6139 
6140 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
6141 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
6142 			option_exists |= IPPF_RTDSTOPTS;
6143 			udp_ip_hdr_len += ipp->ipp_rtdstoptslen;
6144 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
6145 			option_exists |= IPPF_RTDSTOPTS;
6146 			is_sticky |= IPPF_RTDSTOPTS;
6147 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen;
6148 		}
6149 	}
6150 
6151 	if (!(ignore & IPPF_DSTOPTS)) {
6152 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
6153 			option_exists |= IPPF_DSTOPTS;
6154 			udp_ip_hdr_len += ipp->ipp_dstoptslen;
6155 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
6156 			option_exists |= IPPF_DSTOPTS;
6157 			is_sticky |= IPPF_DSTOPTS;
6158 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen;
6159 		}
6160 	}
6161 
6162 	if (!(ignore & IPPF_IFINDEX)) {
6163 		if (ipp->ipp_fields & IPPF_IFINDEX) {
6164 			option_exists |= IPPF_IFINDEX;
6165 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
6166 			option_exists |= IPPF_IFINDEX;
6167 			is_sticky |= IPPF_IFINDEX;
6168 		}
6169 	}
6170 
6171 	if (!(ignore & IPPF_ADDR)) {
6172 		if (ipp->ipp_fields & IPPF_ADDR) {
6173 			option_exists |= IPPF_ADDR;
6174 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) {
6175 			option_exists |= IPPF_ADDR;
6176 			is_sticky |= IPPF_ADDR;
6177 		}
6178 	}
6179 
6180 	if (!(ignore & IPPF_DONTFRAG)) {
6181 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
6182 			option_exists |= IPPF_DONTFRAG;
6183 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
6184 			option_exists |= IPPF_DONTFRAG;
6185 			is_sticky |= IPPF_DONTFRAG;
6186 		}
6187 	}
6188 
6189 	if (!(ignore & IPPF_USE_MIN_MTU)) {
6190 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
6191 			option_exists |= IPPF_USE_MIN_MTU;
6192 		} else if (udp->udp_sticky_ipp.ipp_fields &
6193 		    IPPF_USE_MIN_MTU) {
6194 			option_exists |= IPPF_USE_MIN_MTU;
6195 			is_sticky |= IPPF_USE_MIN_MTU;
6196 		}
6197 	}
6198 
6199 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
6200 		option_exists |= IPPF_HOPLIMIT;
6201 	/* IPV6_HOPLIMIT can never be sticky */
6202 	ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
6203 
6204 	if (!(ignore & IPPF_UNICAST_HOPS) &&
6205 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
6206 		option_exists |= IPPF_UNICAST_HOPS;
6207 		is_sticky |= IPPF_UNICAST_HOPS;
6208 	}
6209 
6210 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
6211 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
6212 		option_exists |= IPPF_MULTICAST_HOPS;
6213 		is_sticky |= IPPF_MULTICAST_HOPS;
6214 	}
6215 
6216 	if (!(ignore & IPPF_TCLASS)) {
6217 		if (ipp->ipp_fields & IPPF_TCLASS) {
6218 			option_exists |= IPPF_TCLASS;
6219 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
6220 			option_exists |= IPPF_TCLASS;
6221 			is_sticky |= IPPF_TCLASS;
6222 		}
6223 	}
6224 
6225 	if (!(ignore & IPPF_NEXTHOP) &&
6226 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_NEXTHOP)) {
6227 		option_exists |= IPPF_NEXTHOP;
6228 		is_sticky |= IPPF_NEXTHOP;
6229 	}
6230 
6231 no_options:
6232 
6233 	/*
6234 	 * If any options carried in the ip6i_t were specified, we
6235 	 * need to account for the ip6i_t in the data we'll be sending
6236 	 * down.
6237 	 */
6238 	if (option_exists & IPPF_HAS_IP6I)
6239 		udp_ip_hdr_len += sizeof (ip6i_t);
6240 
6241 	/* check/fix buffer config, setup pointers into it */
6242 	ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len];
6243 	if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) ||
6244 	    !OK_32PTR(ip6h)) {
6245 
6246 		/* Try to get everything in a single mblk next time */
6247 		if (udp_ip_hdr_len > udp->udp_max_hdr_len) {
6248 			udp->udp_max_hdr_len = udp_ip_hdr_len;
6249 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
6250 		}
6251 
6252 		mp2 = allocb(udp_ip_hdr_len + us->us_wroff_extra, BPRI_LO);
6253 		if (mp2 == NULL) {
6254 			*error = ENOMEM;
6255 			rw_exit(&udp->udp_rwlock);
6256 			goto done;
6257 		}
6258 		mp2->b_wptr = DB_LIM(mp2);
6259 		mp2->b_cont = mp1;
6260 		mp1 = mp2;
6261 		if (DB_TYPE(mp) != M_DATA)
6262 			mp->b_cont = mp1;
6263 		else
6264 			mp = mp1;
6265 
6266 		ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len);
6267 	}
6268 	mp1->b_rptr = (unsigned char *)ip6h;
6269 	ip6i = (ip6i_t *)ip6h;
6270 
6271 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp)
6272 	if (option_exists & IPPF_HAS_IP6I) {
6273 		ip6h = (ip6_t *)&ip6i[1];
6274 		ip6i->ip6i_flags = 0;
6275 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6276 
6277 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
6278 		if (option_exists & IPPF_SCOPE_ID) {
6279 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6280 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
6281 		} else if (option_exists & IPPF_IFINDEX) {
6282 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
6283 			ASSERT(tipp->ipp_ifindex != 0);
6284 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6285 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
6286 		}
6287 
6288 		if (option_exists & IPPF_ADDR) {
6289 			/*
6290 			 * Enable per-packet source address verification if
6291 			 * IPV6_PKTINFO specified the source address.
6292 			 * ip6_src is set in the transport's _wput function.
6293 			 */
6294 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
6295 		}
6296 
6297 		if (option_exists & IPPF_DONTFRAG) {
6298 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
6299 		}
6300 
6301 		if (option_exists & IPPF_USE_MIN_MTU) {
6302 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
6303 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
6304 		}
6305 
6306 		if (option_exists & IPPF_NEXTHOP) {
6307 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
6308 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
6309 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
6310 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
6311 		}
6312 
6313 		/*
6314 		 * tell IP this is an ip6i_t private header
6315 		 */
6316 		ip6i->ip6i_nxt = IPPROTO_RAW;
6317 	}
6318 
6319 	/* Initialize IPv6 header */
6320 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6321 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
6322 
6323 	/* Set the hoplimit of the outgoing packet. */
6324 	if (option_exists & IPPF_HOPLIMIT) {
6325 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
6326 		ip6h->ip6_hops = ipp->ipp_hoplimit;
6327 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6328 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
6329 		ip6h->ip6_hops = udp->udp_multicast_ttl;
6330 		if (option_exists & IPPF_MULTICAST_HOPS)
6331 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6332 	} else {
6333 		ip6h->ip6_hops = udp->udp_ttl;
6334 		if (option_exists & IPPF_UNICAST_HOPS)
6335 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
6336 	}
6337 
6338 	if (option_exists & IPPF_ADDR) {
6339 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
6340 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
6341 		ip6h->ip6_src = tipp->ipp_addr;
6342 	} else {
6343 		/*
6344 		 * The source address was not set using IPV6_PKTINFO.
6345 		 * First look at the bound source.
6346 		 * If unspecified fallback to __sin6_src_id.
6347 		 */
6348 		ip6h->ip6_src = udp->udp_v6src;
6349 		if (sin6->__sin6_src_id != 0 &&
6350 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
6351 			ip_srcid_find_id(sin6->__sin6_src_id,
6352 			    &ip6h->ip6_src, connp->conn_zoneid,
6353 			    us->us_netstack);
6354 		}
6355 	}
6356 
6357 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
6358 	cp = (uint8_t *)&ip6h[1];
6359 
6360 	/*
6361 	 * Here's where we have to start stringing together
6362 	 * any extension headers in the right order:
6363 	 * Hop-by-hop, destination, routing, and final destination opts.
6364 	 */
6365 	if (option_exists & IPPF_HOPOPTS) {
6366 		/* Hop-by-hop options */
6367 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
6368 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
6369 		if (hopoptslen == 0) {
6370 			hopoptsptr = tipp->ipp_hopopts;
6371 			hopoptslen = tipp->ipp_hopoptslen;
6372 			is_ancillary = B_TRUE;
6373 		}
6374 
6375 		*nxthdr_ptr = IPPROTO_HOPOPTS;
6376 		nxthdr_ptr = &hbh->ip6h_nxt;
6377 
6378 		bcopy(hopoptsptr, cp, hopoptslen);
6379 		cp += hopoptslen;
6380 
6381 		if (hopoptsptr != NULL && !is_ancillary) {
6382 			kmem_free(hopoptsptr, hopoptslen);
6383 			hopoptsptr = NULL;
6384 			hopoptslen = 0;
6385 		}
6386 	}
6387 	/*
6388 	 * En-route destination options
6389 	 * Only do them if there's a routing header as well
6390 	 */
6391 	if (option_exists & IPPF_RTDSTOPTS) {
6392 		ip6_dest_t *dst = (ip6_dest_t *)cp;
6393 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
6394 
6395 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6396 		nxthdr_ptr = &dst->ip6d_nxt;
6397 
6398 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
6399 		cp += tipp->ipp_rtdstoptslen;
6400 	}
6401 	/*
6402 	 * Routing header next
6403 	 */
6404 	if (option_exists & IPPF_RTHDR) {
6405 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
6406 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
6407 
6408 		*nxthdr_ptr = IPPROTO_ROUTING;
6409 		nxthdr_ptr = &rt->ip6r_nxt;
6410 
6411 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
6412 		cp += tipp->ipp_rthdrlen;
6413 	}
6414 	/*
6415 	 * Do ultimate destination options
6416 	 */
6417 	if (option_exists & IPPF_DSTOPTS) {
6418 		ip6_dest_t *dest = (ip6_dest_t *)cp;
6419 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
6420 
6421 		*nxthdr_ptr = IPPROTO_DSTOPTS;
6422 		nxthdr_ptr = &dest->ip6d_nxt;
6423 
6424 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
6425 		cp += tipp->ipp_dstoptslen;
6426 	}
6427 	/*
6428 	 * Now set the last header pointer to the proto passed in
6429 	 */
6430 	ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE));
6431 	*nxthdr_ptr = IPPROTO_UDP;
6432 
6433 	/* Update UDP header */
6434 	udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE);
6435 	udph->uha_dst_port = sin6->sin6_port;
6436 	udph->uha_src_port = udp->udp_port;
6437 
6438 	/*
6439 	 * Copy in the destination address
6440 	 */
6441 	ip6h->ip6_dst = ip6_dst;
6442 
6443 	ip6h->ip6_vcf =
6444 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
6445 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
6446 
6447 	if (option_exists & IPPF_TCLASS) {
6448 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
6449 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
6450 		    tipp->ipp_tclass);
6451 	}
6452 	rw_exit(&udp->udp_rwlock);
6453 
6454 	if (option_exists & IPPF_RTHDR) {
6455 		ip6_rthdr_t	*rth;
6456 
6457 		/*
6458 		 * Perform any processing needed for source routing.
6459 		 * We know that all extension headers will be in the same mblk
6460 		 * as the IPv6 header.
6461 		 */
6462 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
6463 		if (rth != NULL && rth->ip6r_segleft != 0) {
6464 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
6465 				/*
6466 				 * Drop packet - only support Type 0 routing.
6467 				 * Notify the application as well.
6468 				 */
6469 				*error = EPROTO;
6470 				goto done;
6471 			}
6472 
6473 			/*
6474 			 * rth->ip6r_len is twice the number of
6475 			 * addresses in the header. Thus it must be even.
6476 			 */
6477 			if (rth->ip6r_len & 0x1) {
6478 				*error = EPROTO;
6479 				goto done;
6480 			}
6481 			/*
6482 			 * Shuffle the routing header and ip6_dst
6483 			 * addresses, and get the checksum difference
6484 			 * between the first hop (in ip6_dst) and
6485 			 * the destination (in the last routing hdr entry).
6486 			 */
6487 			csum = ip_massage_options_v6(ip6h, rth,
6488 			    us->us_netstack);
6489 			/*
6490 			 * Verify that the first hop isn't a mapped address.
6491 			 * Routers along the path need to do this verification
6492 			 * for subsequent hops.
6493 			 */
6494 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
6495 				*error = EADDRNOTAVAIL;
6496 				goto done;
6497 			}
6498 
6499 			cp += (rth->ip6r_len + 1)*8;
6500 		}
6501 	}
6502 
6503 	/* count up length of UDP packet */
6504 	ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN;
6505 	if ((mp2 = mp1->b_cont) != NULL) {
6506 		do {
6507 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
6508 			ip_len += (uint32_t)MBLKL(mp2);
6509 		} while ((mp2 = mp2->b_cont) != NULL);
6510 	}
6511 
6512 	/*
6513 	 * If the size of the packet is greater than the maximum allowed by
6514 	 * ip, return an error. Passing this down could cause panics because
6515 	 * the size will have wrapped and be inconsistent with the msg size.
6516 	 */
6517 	if (ip_len > IP_MAXPACKET) {
6518 		*error = EMSGSIZE;
6519 		goto done;
6520 	}
6521 
6522 	/* Store the UDP length. Subtract length of extension hdrs */
6523 	udph->uha_length = htons(ip_len + IPV6_HDR_LEN -
6524 	    (int)((uchar_t *)udph - (uchar_t *)ip6h));
6525 
6526 	/*
6527 	 * We make it easy for IP to include our pseudo header
6528 	 * by putting our length in uh_checksum, modified (if
6529 	 * we have a routing header) by the checksum difference
6530 	 * between the ultimate destination and first hop addresses.
6531 	 * Note: UDP over IPv6 must always checksum the packet.
6532 	 */
6533 	csum += udph->uha_length;
6534 	csum = (csum & 0xFFFF) + (csum >> 16);
6535 	udph->uha_checksum = (uint16_t)csum;
6536 
6537 #ifdef _LITTLE_ENDIAN
6538 	ip_len = htons(ip_len);
6539 #endif
6540 	ip6h->ip6_plen = ip_len;
6541 	if (DB_CRED(mp) != NULL)
6542 		mblk_setcred(mp1, DB_CRED(mp));
6543 
6544 	if (DB_TYPE(mp) != M_DATA) {
6545 		ASSERT(mp != mp1);
6546 		freeb(mp);
6547 	}
6548 
6549 	/* mp has been consumed and we'll return success */
6550 	ASSERT(*error == 0);
6551 	mp = NULL;
6552 
6553 	/* We're done. Pass the packet to IP */
6554 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
6555 	ip_output_v6(connp, mp1, q, IP_WPUT);
6556 
6557 done:
6558 	if (sth_wroff != 0) {
6559 		(void) proto_set_tx_wroff(RD(q), connp,
6560 		    udp->udp_max_hdr_len + us->us_wroff_extra);
6561 	}
6562 	if (hopoptsptr != NULL && !is_ancillary) {
6563 		kmem_free(hopoptsptr, hopoptslen);
6564 		hopoptsptr = NULL;
6565 	}
6566 	if (*error != 0) {
6567 		ASSERT(mp != NULL);
6568 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6569 	}
6570 	return (mp);
6571 }
6572 
6573 
6574 static int
6575 i_udp_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6576 {
6577 	sin_t *sin = (sin_t *)sa;
6578 	sin6_t *sin6 = (sin6_t *)sa;
6579 
6580 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6581 
6582 	if (udp->udp_state != TS_DATA_XFER)
6583 		return (ENOTCONN);
6584 
6585 	switch (udp->udp_family) {
6586 	case AF_INET:
6587 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6588 
6589 		if (*salenp < sizeof (sin_t))
6590 			return (EINVAL);
6591 
6592 		*salenp = sizeof (sin_t);
6593 		*sin = sin_null;
6594 		sin->sin_family = AF_INET;
6595 		sin->sin_port = udp->udp_dstport;
6596 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
6597 		break;
6598 
6599 	case AF_INET6:
6600 		if (*salenp < sizeof (sin6_t))
6601 			return (EINVAL);
6602 
6603 		*salenp = sizeof (sin6_t);
6604 		*sin6 = sin6_null;
6605 		sin6->sin6_family = AF_INET6;
6606 		sin6->sin6_port = udp->udp_dstport;
6607 		sin6->sin6_addr = udp->udp_v6dst;
6608 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6609 		break;
6610 	}
6611 
6612 	return (0);
6613 }
6614 
6615 static int
6616 udp_getmyname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
6617 {
6618 	sin_t *sin = (sin_t *)sa;
6619 	sin6_t *sin6 = (sin6_t *)sa;
6620 
6621 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
6622 
6623 	switch (udp->udp_family) {
6624 	case AF_INET:
6625 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
6626 
6627 		if (*salenp < sizeof (sin_t))
6628 			return (EINVAL);
6629 
6630 		*salenp = sizeof (sin_t);
6631 		*sin = sin_null;
6632 		sin->sin_family = AF_INET;
6633 		sin->sin_port = udp->udp_port;
6634 
6635 		/*
6636 		 * If udp_v6src is unspecified, we might be bound to broadcast
6637 		 * / multicast.  Use udp_bound_v6src as local address instead
6638 		 * (that could also still be unspecified).
6639 		 */
6640 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
6641 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
6642 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
6643 		} else {
6644 			sin->sin_addr.s_addr =
6645 			    V4_PART_OF_V6(udp->udp_bound_v6src);
6646 		}
6647 		break;
6648 
6649 	case AF_INET6:
6650 		if (*salenp < sizeof (sin6_t))
6651 			return (EINVAL);
6652 
6653 		*salenp = sizeof (sin6_t);
6654 		*sin6 = sin6_null;
6655 		sin6->sin6_family = AF_INET6;
6656 		sin6->sin6_port = udp->udp_port;
6657 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6658 
6659 		/*
6660 		 * If udp_v6src is unspecified, we might be bound to broadcast
6661 		 * / multicast.  Use udp_bound_v6src as local address instead
6662 		 * (that could also still be unspecified).
6663 		 */
6664 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))
6665 			sin6->sin6_addr = udp->udp_v6src;
6666 		else
6667 			sin6->sin6_addr = udp->udp_bound_v6src;
6668 		break;
6669 	}
6670 
6671 	return (0);
6672 }
6673 
6674 /*
6675  * Handle special out-of-band ioctl requests (see PSARC/2008/265).
6676  */
6677 static void
6678 udp_wput_cmdblk(queue_t *q, mblk_t *mp)
6679 {
6680 	void	*data;
6681 	mblk_t	*datamp = mp->b_cont;
6682 	udp_t	*udp = Q_TO_UDP(q);
6683 	cmdblk_t *cmdp = (cmdblk_t *)mp->b_rptr;
6684 
6685 	if (datamp == NULL || MBLKL(datamp) < cmdp->cb_len) {
6686 		cmdp->cb_error = EPROTO;
6687 		qreply(q, mp);
6688 		return;
6689 	}
6690 	data = datamp->b_rptr;
6691 
6692 	rw_enter(&udp->udp_rwlock, RW_READER);
6693 	switch (cmdp->cb_cmd) {
6694 	case TI_GETPEERNAME:
6695 		cmdp->cb_error = i_udp_getpeername(udp, data, &cmdp->cb_len);
6696 		break;
6697 	case TI_GETMYNAME:
6698 		cmdp->cb_error = udp_getmyname(udp, data, &cmdp->cb_len);
6699 		break;
6700 	default:
6701 		cmdp->cb_error = EINVAL;
6702 		break;
6703 	}
6704 	rw_exit(&udp->udp_rwlock);
6705 
6706 	qreply(q, mp);
6707 }
6708 
6709 static void
6710 udp_disable_direct_sockfs(udp_t *udp)
6711 {
6712 	udp->udp_issocket = B_FALSE;
6713 	if (udp->udp_direct_sockfs) {
6714 		/*
6715 		 * Disable read-side synchronous stream interface and
6716 		 * drain any queued data.
6717 		 */
6718 		udp_rcv_drain(udp->udp_connp->conn_rq, udp, B_FALSE);
6719 		ASSERT(!udp->udp_direct_sockfs);
6720 		UDP_STAT(udp->udp_us, udp_sock_fallback);
6721 	}
6722 }
6723 
6724 static void
6725 udp_wput_other(queue_t *q, mblk_t *mp)
6726 {
6727 	uchar_t	*rptr = mp->b_rptr;
6728 	struct datab *db;
6729 	struct iocblk *iocp;
6730 	cred_t	*cr;
6731 	conn_t	*connp = Q_TO_CONN(q);
6732 	udp_t	*udp = connp->conn_udp;
6733 	udp_stack_t *us;
6734 
6735 	TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START,
6736 	    "udp_wput_other_start: q %p", q);
6737 
6738 	us = udp->udp_us;
6739 	db = mp->b_datap;
6740 
6741 	cr = DB_CREDDEF(mp, connp->conn_cred);
6742 
6743 	switch (db->db_type) {
6744 	case M_CMD:
6745 		udp_wput_cmdblk(q, mp);
6746 		return;
6747 
6748 	case M_PROTO:
6749 	case M_PCPROTO:
6750 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
6751 			freemsg(mp);
6752 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6753 			    "udp_wput_other_end: q %p (%S)", q, "protoshort");
6754 			return;
6755 		}
6756 		switch (((t_primp_t)rptr)->type) {
6757 		case T_ADDR_REQ:
6758 			udp_addr_req(q, mp);
6759 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6760 			    "udp_wput_other_end: q %p (%S)", q, "addrreq");
6761 			return;
6762 		case O_T_BIND_REQ:
6763 		case T_BIND_REQ:
6764 			udp_tpi_bind(q, mp);
6765 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6766 			    "udp_wput_other_end: q %p (%S)", q, "bindreq");
6767 			return;
6768 		case T_CONN_REQ:
6769 			udp_tpi_connect(q, mp);
6770 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6771 			    "udp_wput_other_end: q %p (%S)", q, "connreq");
6772 			return;
6773 		case T_CAPABILITY_REQ:
6774 			udp_capability_req(q, mp);
6775 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6776 			    "udp_wput_other_end: q %p (%S)", q, "capabreq");
6777 			return;
6778 		case T_INFO_REQ:
6779 			udp_info_req(q, mp);
6780 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6781 			    "udp_wput_other_end: q %p (%S)", q, "inforeq");
6782 			return;
6783 		case T_UNITDATA_REQ:
6784 			/*
6785 			 * If a T_UNITDATA_REQ gets here, the address must
6786 			 * be bad.  Valid T_UNITDATA_REQs are handled
6787 			 * in udp_wput.
6788 			 */
6789 			udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL);
6790 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6791 			    "udp_wput_other_end: q %p (%S)", q, "unitdatareq");
6792 			return;
6793 		case T_UNBIND_REQ:
6794 			udp_tpi_unbind(q, mp);
6795 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6796 			    "udp_wput_other_end: q %p (%S)", q, "unbindreq");
6797 			return;
6798 		case T_SVR4_OPTMGMT_REQ:
6799 			if (!snmpcom_req(q, mp, udp_snmp_set, ip_snmp_get,
6800 			    cr)) {
6801 				(void) svr4_optcom_req(q,
6802 				    mp, cr, &udp_opt_obj, B_TRUE);
6803 			}
6804 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6805 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6806 			return;
6807 
6808 		case T_OPTMGMT_REQ:
6809 			(void) tpi_optcom_req(q, mp, cr, &udp_opt_obj, B_TRUE);
6810 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6811 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
6812 			return;
6813 
6814 		case T_DISCON_REQ:
6815 			udp_tpi_disconnect(q, mp);
6816 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6817 			    "udp_wput_other_end: q %p (%S)", q, "disconreq");
6818 			return;
6819 
6820 		/* The following TPI message is not supported by udp. */
6821 		case O_T_CONN_RES:
6822 		case T_CONN_RES:
6823 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6824 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6825 			    "udp_wput_other_end: q %p (%S)", q,
6826 			    "connres/disconreq");
6827 			return;
6828 
6829 		/* The following 3 TPI messages are illegal for udp. */
6830 		case T_DATA_REQ:
6831 		case T_EXDATA_REQ:
6832 		case T_ORDREL_REQ:
6833 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
6834 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6835 			    "udp_wput_other_end: q %p (%S)", q,
6836 			    "data/exdata/ordrel");
6837 			return;
6838 		default:
6839 			break;
6840 		}
6841 		break;
6842 	case M_FLUSH:
6843 		if (*rptr & FLUSHW)
6844 			flushq(q, FLUSHDATA);
6845 		break;
6846 	case M_IOCTL:
6847 		iocp = (struct iocblk *)mp->b_rptr;
6848 		switch (iocp->ioc_cmd) {
6849 		case TI_GETPEERNAME:
6850 			if (udp->udp_state != TS_DATA_XFER) {
6851 				/*
6852 				 * If a default destination address has not
6853 				 * been associated with the stream, then we
6854 				 * don't know the peer's name.
6855 				 */
6856 				iocp->ioc_error = ENOTCONN;
6857 				iocp->ioc_count = 0;
6858 				mp->b_datap->db_type = M_IOCACK;
6859 				qreply(q, mp);
6860 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6861 				    "udp_wput_other_end: q %p (%S)", q,
6862 				    "getpeername");
6863 				return;
6864 			}
6865 			/* FALLTHRU */
6866 		case TI_GETMYNAME: {
6867 			/*
6868 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
6869 			 * need to copyin the user's strbuf structure.
6870 			 * Processing will continue in the M_IOCDATA case
6871 			 * below.
6872 			 */
6873 			mi_copyin(q, mp, NULL,
6874 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
6875 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6876 			    "udp_wput_other_end: q %p (%S)", q, "getmyname");
6877 			return;
6878 			}
6879 		case ND_SET:
6880 			/* nd_getset performs the necessary checking */
6881 		case ND_GET:
6882 			if (nd_getset(q, us->us_nd, mp)) {
6883 				qreply(q, mp);
6884 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6885 				    "udp_wput_other_end: q %p (%S)", q, "get");
6886 				return;
6887 			}
6888 			break;
6889 		case _SIOCSOCKFALLBACK:
6890 			/*
6891 			 * Either sockmod is about to be popped and the
6892 			 * socket would now be treated as a plain stream,
6893 			 * or a module is about to be pushed so we could
6894 			 * no longer use read-side synchronous stream.
6895 			 * Drain any queued data and disable direct sockfs
6896 			 * interface from now on.
6897 			 */
6898 			if (!udp->udp_issocket) {
6899 				DB_TYPE(mp) = M_IOCNAK;
6900 				iocp->ioc_error = EINVAL;
6901 			} else {
6902 				udp_disable_direct_sockfs(udp);
6903 
6904 				DB_TYPE(mp) = M_IOCACK;
6905 				iocp->ioc_error = 0;
6906 			}
6907 			iocp->ioc_count = 0;
6908 			iocp->ioc_rval = 0;
6909 			qreply(q, mp);
6910 			return;
6911 		default:
6912 			break;
6913 		}
6914 		break;
6915 	case M_IOCDATA:
6916 		udp_wput_iocdata(q, mp);
6917 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6918 		    "udp_wput_other_end: q %p (%S)", q, "iocdata");
6919 		return;
6920 	default:
6921 		/* Unrecognized messages are passed through without change. */
6922 		break;
6923 	}
6924 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
6925 	    "udp_wput_other_end: q %p (%S)", q, "end");
6926 	ip_output(connp, mp, q, IP_WPUT);
6927 }
6928 
6929 /*
6930  * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA
6931  * messages.
6932  */
6933 static void
6934 udp_wput_iocdata(queue_t *q, mblk_t *mp)
6935 {
6936 	mblk_t		*mp1;
6937 	struct	iocblk *iocp = (struct iocblk *)mp->b_rptr;
6938 	STRUCT_HANDLE(strbuf, sb);
6939 	udp_t		*udp = Q_TO_UDP(q);
6940 	int		error;
6941 	uint_t		addrlen;
6942 
6943 	/* Make sure it is one of ours. */
6944 	switch (iocp->ioc_cmd) {
6945 	case TI_GETMYNAME:
6946 	case TI_GETPEERNAME:
6947 		break;
6948 	default:
6949 		ip_output(udp->udp_connp, mp, q, IP_WPUT);
6950 		return;
6951 	}
6952 
6953 	switch (mi_copy_state(q, mp, &mp1)) {
6954 	case -1:
6955 		return;
6956 	case MI_COPY_CASE(MI_COPY_IN, 1):
6957 		break;
6958 	case MI_COPY_CASE(MI_COPY_OUT, 1):
6959 		/*
6960 		 * The address has been copied out, so now
6961 		 * copyout the strbuf.
6962 		 */
6963 		mi_copyout(q, mp);
6964 		return;
6965 	case MI_COPY_CASE(MI_COPY_OUT, 2):
6966 		/*
6967 		 * The address and strbuf have been copied out.
6968 		 * We're done, so just acknowledge the original
6969 		 * M_IOCTL.
6970 		 */
6971 		mi_copy_done(q, mp, 0);
6972 		return;
6973 	default:
6974 		/*
6975 		 * Something strange has happened, so acknowledge
6976 		 * the original M_IOCTL with an EPROTO error.
6977 		 */
6978 		mi_copy_done(q, mp, EPROTO);
6979 		return;
6980 	}
6981 
6982 	/*
6983 	 * Now we have the strbuf structure for TI_GETMYNAME
6984 	 * and TI_GETPEERNAME.  Next we copyout the requested
6985 	 * address and then we'll copyout the strbuf.
6986 	 */
6987 	STRUCT_SET_HANDLE(sb, iocp->ioc_flag, (void *)mp1->b_rptr);
6988 	addrlen = udp->udp_family == AF_INET ? sizeof (sin_t) : sizeof (sin6_t);
6989 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
6990 		mi_copy_done(q, mp, EINVAL);
6991 		return;
6992 	}
6993 
6994 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
6995 
6996 	if (mp1 == NULL)
6997 		return;
6998 
6999 	rw_enter(&udp->udp_rwlock, RW_READER);
7000 	switch (iocp->ioc_cmd) {
7001 	case TI_GETMYNAME:
7002 		error = udp_do_getsockname(udp, (void *)mp1->b_rptr, &addrlen);
7003 		break;
7004 	case TI_GETPEERNAME:
7005 		error = udp_do_getpeername(udp, (void *)mp1->b_rptr, &addrlen);
7006 		break;
7007 	}
7008 	rw_exit(&udp->udp_rwlock);
7009 
7010 	if (error != 0) {
7011 		mi_copy_done(q, mp, error);
7012 	} else {
7013 		mp1->b_wptr += addrlen;
7014 		STRUCT_FSET(sb, len, addrlen);
7015 
7016 		/* Copy out the address */
7017 		mi_copyout(q, mp);
7018 	}
7019 }
7020 
7021 static int
7022 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
7023     udpattrs_t *udpattrs)
7024 {
7025 	struct T_unitdata_req *udreqp;
7026 	int is_absreq_failure;
7027 	cred_t *cr;
7028 	conn_t	*connp = Q_TO_CONN(q);
7029 
7030 	ASSERT(((t_primp_t)mp->b_rptr)->type);
7031 
7032 	cr = DB_CREDDEF(mp, connp->conn_cred);
7033 
7034 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
7035 
7036 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
7037 	    udreqp->OPT_offset, cr, &udp_opt_obj,
7038 	    udpattrs, &is_absreq_failure);
7039 
7040 	if (*errorp != 0) {
7041 		/*
7042 		 * Note: No special action needed in this
7043 		 * module for "is_absreq_failure"
7044 		 */
7045 		return (-1);		/* failure */
7046 	}
7047 	ASSERT(is_absreq_failure == 0);
7048 	return (0);	/* success */
7049 }
7050 
7051 void
7052 udp_ddi_g_init(void)
7053 {
7054 	udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr,
7055 	    udp_opt_obj.odb_opt_arr_cnt);
7056 
7057 	/*
7058 	 * We want to be informed each time a stack is created or
7059 	 * destroyed in the kernel, so we can maintain the
7060 	 * set of udp_stack_t's.
7061 	 */
7062 	netstack_register(NS_UDP, udp_stack_init, NULL, udp_stack_fini);
7063 }
7064 
7065 void
7066 udp_ddi_g_destroy(void)
7067 {
7068 	netstack_unregister(NS_UDP);
7069 }
7070 
7071 #define	INET_NAME	"ip"
7072 
7073 /*
7074  * Initialize the UDP stack instance.
7075  */
7076 static void *
7077 udp_stack_init(netstackid_t stackid, netstack_t *ns)
7078 {
7079 	udp_stack_t	*us;
7080 	udpparam_t	*pa;
7081 	int		i;
7082 	int		error = 0;
7083 	major_t		major;
7084 
7085 	us = (udp_stack_t *)kmem_zalloc(sizeof (*us), KM_SLEEP);
7086 	us->us_netstack = ns;
7087 
7088 	us->us_num_epriv_ports = UDP_NUM_EPRIV_PORTS;
7089 	us->us_epriv_ports[0] = 2049;
7090 	us->us_epriv_ports[1] = 4045;
7091 
7092 	/*
7093 	 * The smallest anonymous port in the priviledged port range which UDP
7094 	 * looks for free port.  Use in the option UDP_ANONPRIVBIND.
7095 	 */
7096 	us->us_min_anonpriv_port = 512;
7097 
7098 	us->us_bind_fanout_size = udp_bind_fanout_size;
7099 
7100 	/* Roundup variable that might have been modified in /etc/system */
7101 	if (us->us_bind_fanout_size & (us->us_bind_fanout_size - 1)) {
7102 		/* Not a power of two. Round up to nearest power of two */
7103 		for (i = 0; i < 31; i++) {
7104 			if (us->us_bind_fanout_size < (1 << i))
7105 				break;
7106 		}
7107 		us->us_bind_fanout_size = 1 << i;
7108 	}
7109 	us->us_bind_fanout = kmem_zalloc(us->us_bind_fanout_size *
7110 	    sizeof (udp_fanout_t), KM_SLEEP);
7111 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7112 		mutex_init(&us->us_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT,
7113 		    NULL);
7114 	}
7115 
7116 	pa = (udpparam_t *)kmem_alloc(sizeof (udp_param_arr), KM_SLEEP);
7117 
7118 	us->us_param_arr = pa;
7119 	bcopy(udp_param_arr, us->us_param_arr, sizeof (udp_param_arr));
7120 
7121 	(void) udp_param_register(&us->us_nd,
7122 	    us->us_param_arr, A_CNT(udp_param_arr));
7123 
7124 	us->us_kstat = udp_kstat2_init(stackid, &us->us_statistics);
7125 	us->us_mibkp = udp_kstat_init(stackid);
7126 
7127 	major = mod_name_to_major(INET_NAME);
7128 	error = ldi_ident_from_major(major, &us->us_ldi_ident);
7129 	ASSERT(error == 0);
7130 	return (us);
7131 }
7132 
7133 /*
7134  * Free the UDP stack instance.
7135  */
7136 static void
7137 udp_stack_fini(netstackid_t stackid, void *arg)
7138 {
7139 	udp_stack_t *us = (udp_stack_t *)arg;
7140 	int i;
7141 
7142 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7143 		mutex_destroy(&us->us_bind_fanout[i].uf_lock);
7144 	}
7145 
7146 	kmem_free(us->us_bind_fanout, us->us_bind_fanout_size *
7147 	    sizeof (udp_fanout_t));
7148 
7149 	us->us_bind_fanout = NULL;
7150 
7151 	nd_free(&us->us_nd);
7152 	kmem_free(us->us_param_arr, sizeof (udp_param_arr));
7153 	us->us_param_arr = NULL;
7154 
7155 	udp_kstat_fini(stackid, us->us_mibkp);
7156 	us->us_mibkp = NULL;
7157 
7158 	udp_kstat2_fini(stackid, us->us_kstat);
7159 	us->us_kstat = NULL;
7160 	bzero(&us->us_statistics, sizeof (us->us_statistics));
7161 
7162 	ldi_ident_release(us->us_ldi_ident);
7163 	kmem_free(us, sizeof (*us));
7164 }
7165 
7166 static void *
7167 udp_kstat2_init(netstackid_t stackid, udp_stat_t *us_statisticsp)
7168 {
7169 	kstat_t *ksp;
7170 
7171 	udp_stat_t template = {
7172 		{ "udp_ip_send",		KSTAT_DATA_UINT64 },
7173 		{ "udp_ip_ire_send",		KSTAT_DATA_UINT64 },
7174 		{ "udp_ire_null",		KSTAT_DATA_UINT64 },
7175 		{ "udp_drain",			KSTAT_DATA_UINT64 },
7176 		{ "udp_sock_fallback",		KSTAT_DATA_UINT64 },
7177 		{ "udp_rrw_busy",		KSTAT_DATA_UINT64 },
7178 		{ "udp_rrw_msgcnt",		KSTAT_DATA_UINT64 },
7179 		{ "udp_out_sw_cksum",		KSTAT_DATA_UINT64 },
7180 		{ "udp_out_sw_cksum_bytes",	KSTAT_DATA_UINT64 },
7181 		{ "udp_out_opt",		KSTAT_DATA_UINT64 },
7182 		{ "udp_out_err_notconn",	KSTAT_DATA_UINT64 },
7183 		{ "udp_out_err_output",		KSTAT_DATA_UINT64 },
7184 		{ "udp_out_err_tudr",		KSTAT_DATA_UINT64 },
7185 		{ "udp_in_pktinfo",		KSTAT_DATA_UINT64 },
7186 		{ "udp_in_recvdstaddr",		KSTAT_DATA_UINT64 },
7187 		{ "udp_in_recvopts",		KSTAT_DATA_UINT64 },
7188 		{ "udp_in_recvif",		KSTAT_DATA_UINT64 },
7189 		{ "udp_in_recvslla",		KSTAT_DATA_UINT64 },
7190 		{ "udp_in_recvucred",		KSTAT_DATA_UINT64 },
7191 		{ "udp_in_recvttl",		KSTAT_DATA_UINT64 },
7192 		{ "udp_in_recvhopopts",		KSTAT_DATA_UINT64 },
7193 		{ "udp_in_recvhoplimit",	KSTAT_DATA_UINT64 },
7194 		{ "udp_in_recvdstopts",		KSTAT_DATA_UINT64 },
7195 		{ "udp_in_recvrtdstopts",	KSTAT_DATA_UINT64 },
7196 		{ "udp_in_recvrthdr",		KSTAT_DATA_UINT64 },
7197 		{ "udp_in_recvpktinfo",		KSTAT_DATA_UINT64 },
7198 		{ "udp_in_recvtclass",		KSTAT_DATA_UINT64 },
7199 		{ "udp_in_timestamp",		KSTAT_DATA_UINT64 },
7200 #ifdef DEBUG
7201 		{ "udp_data_conn",		KSTAT_DATA_UINT64 },
7202 		{ "udp_data_notconn",		KSTAT_DATA_UINT64 },
7203 #endif
7204 	};
7205 
7206 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, "udpstat", "net",
7207 	    KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
7208 	    KSTAT_FLAG_VIRTUAL, stackid);
7209 
7210 	if (ksp == NULL)
7211 		return (NULL);
7212 
7213 	bcopy(&template, us_statisticsp, sizeof (template));
7214 	ksp->ks_data = (void *)us_statisticsp;
7215 	ksp->ks_private = (void *)(uintptr_t)stackid;
7216 
7217 	kstat_install(ksp);
7218 	return (ksp);
7219 }
7220 
7221 static void
7222 udp_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
7223 {
7224 	if (ksp != NULL) {
7225 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7226 		kstat_delete_netstack(ksp, stackid);
7227 	}
7228 }
7229 
7230 static void *
7231 udp_kstat_init(netstackid_t stackid)
7232 {
7233 	kstat_t	*ksp;
7234 
7235 	udp_named_kstat_t template = {
7236 		{ "inDatagrams",	KSTAT_DATA_UINT64, 0 },
7237 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
7238 		{ "outDatagrams",	KSTAT_DATA_UINT64, 0 },
7239 		{ "entrySize",		KSTAT_DATA_INT32, 0 },
7240 		{ "entry6Size",		KSTAT_DATA_INT32, 0 },
7241 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
7242 	};
7243 
7244 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, UDP_MOD_NAME, "mib2",
7245 	    KSTAT_TYPE_NAMED,
7246 	    NUM_OF_FIELDS(udp_named_kstat_t), 0, stackid);
7247 
7248 	if (ksp == NULL || ksp->ks_data == NULL)
7249 		return (NULL);
7250 
7251 	template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t);
7252 	template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t);
7253 
7254 	bcopy(&template, ksp->ks_data, sizeof (template));
7255 	ksp->ks_update = udp_kstat_update;
7256 	ksp->ks_private = (void *)(uintptr_t)stackid;
7257 
7258 	kstat_install(ksp);
7259 	return (ksp);
7260 }
7261 
7262 static void
7263 udp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
7264 {
7265 	if (ksp != NULL) {
7266 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7267 		kstat_delete_netstack(ksp, stackid);
7268 	}
7269 }
7270 
7271 static int
7272 udp_kstat_update(kstat_t *kp, int rw)
7273 {
7274 	udp_named_kstat_t *udpkp;
7275 	netstackid_t	stackid = (netstackid_t)(uintptr_t)kp->ks_private;
7276 	netstack_t	*ns;
7277 	udp_stack_t	*us;
7278 
7279 	if ((kp == NULL) || (kp->ks_data == NULL))
7280 		return (EIO);
7281 
7282 	if (rw == KSTAT_WRITE)
7283 		return (EACCES);
7284 
7285 	ns = netstack_find_by_stackid(stackid);
7286 	if (ns == NULL)
7287 		return (-1);
7288 	us = ns->netstack_udp;
7289 	if (us == NULL) {
7290 		netstack_rele(ns);
7291 		return (-1);
7292 	}
7293 	udpkp = (udp_named_kstat_t *)kp->ks_data;
7294 
7295 	udpkp->inDatagrams.value.ui64 =	us->us_udp_mib.udpHCInDatagrams;
7296 	udpkp->inErrors.value.ui32 =	us->us_udp_mib.udpInErrors;
7297 	udpkp->outDatagrams.value.ui64 = us->us_udp_mib.udpHCOutDatagrams;
7298 	udpkp->outErrors.value.ui32 =	us->us_udp_mib.udpOutErrors;
7299 	netstack_rele(ns);
7300 	return (0);
7301 }
7302 
7303 /*
7304  * Read-side synchronous stream info entry point, called as a
7305  * result of handling certain STREAMS ioctl operations.
7306  */
7307 static int
7308 udp_rinfop(queue_t *q, infod_t *dp)
7309 {
7310 	mblk_t	*mp;
7311 	uint_t	cmd = dp->d_cmd;
7312 	int	res = 0;
7313 	int	error = 0;
7314 	udp_t	*udp = Q_TO_UDP(q);
7315 	struct stdata *stp = STREAM(q);
7316 
7317 	mutex_enter(&udp->udp_drain_lock);
7318 	/* If shutdown on read has happened, return nothing */
7319 	mutex_enter(&stp->sd_lock);
7320 	if (stp->sd_flag & STREOF) {
7321 		mutex_exit(&stp->sd_lock);
7322 		goto done;
7323 	}
7324 	mutex_exit(&stp->sd_lock);
7325 
7326 	if ((mp = udp->udp_rcv_list_head) == NULL)
7327 		goto done;
7328 
7329 	ASSERT(DB_TYPE(mp) != M_DATA && mp->b_cont != NULL);
7330 
7331 	if (cmd & INFOD_COUNT) {
7332 		/*
7333 		 * Return the number of messages.
7334 		 */
7335 		dp->d_count += udp->udp_rcv_msgcnt;
7336 		res |= INFOD_COUNT;
7337 	}
7338 	if (cmd & INFOD_BYTES) {
7339 		/*
7340 		 * Return size of all data messages.
7341 		 */
7342 		dp->d_bytes += udp->udp_rcv_cnt;
7343 		res |= INFOD_BYTES;
7344 	}
7345 	if (cmd & INFOD_FIRSTBYTES) {
7346 		/*
7347 		 * Return size of first data message.
7348 		 */
7349 		dp->d_bytes = msgdsize(mp);
7350 		res |= INFOD_FIRSTBYTES;
7351 		dp->d_cmd &= ~INFOD_FIRSTBYTES;
7352 	}
7353 	if (cmd & INFOD_COPYOUT) {
7354 		mblk_t *mp1 = mp->b_cont;
7355 		int n;
7356 		/*
7357 		 * Return data contents of first message.
7358 		 */
7359 		ASSERT(DB_TYPE(mp1) == M_DATA);
7360 		while (mp1 != NULL && dp->d_uiop->uio_resid > 0) {
7361 			n = MIN(dp->d_uiop->uio_resid, MBLKL(mp1));
7362 			if (n != 0 && (error = uiomove((char *)mp1->b_rptr, n,
7363 			    UIO_READ, dp->d_uiop)) != 0) {
7364 				goto done;
7365 			}
7366 			mp1 = mp1->b_cont;
7367 		}
7368 		res |= INFOD_COPYOUT;
7369 		dp->d_cmd &= ~INFOD_COPYOUT;
7370 	}
7371 done:
7372 	mutex_exit(&udp->udp_drain_lock);
7373 
7374 	dp->d_res |= res;
7375 
7376 	return (error);
7377 }
7378 
7379 /*
7380  * Read-side synchronous stream entry point.  This is called as a result
7381  * of recv/read operation done at sockfs, and is guaranteed to execute
7382  * outside of the interrupt thread context.  It returns a single datagram
7383  * (b_cont chain of T_UNITDATA_IND plus data) to the upper layer.
7384  */
7385 static int
7386 udp_rrw(queue_t *q, struiod_t *dp)
7387 {
7388 	mblk_t	*mp;
7389 	udp_t	*udp = Q_TO_UDP(q);
7390 	udp_stack_t *us = udp->udp_us;
7391 
7392 	/*
7393 	 * Dequeue datagram from the head of the list and return
7394 	 * it to caller; also ensure that RSLEEP sd_wakeq flag is
7395 	 * set/cleared depending on whether or not there's data
7396 	 * remaining in the list.
7397 	 */
7398 	mutex_enter(&udp->udp_drain_lock);
7399 	if (!udp->udp_direct_sockfs) {
7400 		mutex_exit(&udp->udp_drain_lock);
7401 		UDP_STAT(us, udp_rrw_busy);
7402 		return (EBUSY);
7403 	}
7404 	if ((mp = udp->udp_rcv_list_head) != NULL) {
7405 		uint_t size = msgdsize(mp);
7406 
7407 		/* Last datagram in the list? */
7408 		if ((udp->udp_rcv_list_head = mp->b_next) == NULL)
7409 			udp->udp_rcv_list_tail = NULL;
7410 		mp->b_next = NULL;
7411 
7412 		udp->udp_rcv_cnt -= size;
7413 		udp->udp_rcv_msgcnt--;
7414 		UDP_STAT(us, udp_rrw_msgcnt);
7415 
7416 		/* No longer flow-controlling? */
7417 		if (udp->udp_rcv_cnt < udp->udp_rcv_hiwat &&
7418 		    udp->udp_rcv_msgcnt < udp->udp_rcv_hiwat)
7419 			udp->udp_drain_qfull = B_FALSE;
7420 	}
7421 	if (udp->udp_rcv_list_head == NULL) {
7422 		/*
7423 		 * Either we just dequeued the last datagram or
7424 		 * we get here from sockfs and have nothing to
7425 		 * return; in this case clear RSLEEP.
7426 		 */
7427 		ASSERT(udp->udp_rcv_cnt == 0);
7428 		ASSERT(udp->udp_rcv_msgcnt == 0);
7429 		ASSERT(udp->udp_rcv_list_tail == NULL);
7430 		STR_WAKEUP_CLEAR(STREAM(q));
7431 	} else {
7432 		/*
7433 		 * More data follows; we need udp_rrw() to be
7434 		 * called in future to pick up the rest.
7435 		 */
7436 		STR_WAKEUP_SET(STREAM(q));
7437 	}
7438 	mutex_exit(&udp->udp_drain_lock);
7439 	dp->d_mp = mp;
7440 	return (0);
7441 }
7442 
7443 /*
7444  * Enqueue a completely-built T_UNITDATA_IND message into the receive
7445  * list; this is typically executed within the interrupt thread context
7446  * and so we do things as quickly as possible.
7447  */
7448 static void
7449 udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, uint_t pkt_len)
7450 {
7451 	ASSERT(q == RD(q));
7452 	ASSERT(pkt_len == msgdsize(mp));
7453 	ASSERT(mp->b_next == NULL && mp->b_cont != NULL);
7454 	ASSERT(DB_TYPE(mp) == M_PROTO && DB_TYPE(mp->b_cont) == M_DATA);
7455 	ASSERT(MBLKL(mp) >= sizeof (struct T_unitdata_ind));
7456 
7457 	mutex_enter(&udp->udp_drain_lock);
7458 	/*
7459 	 * Wake up and signal the receiving app; it is okay to do this
7460 	 * before enqueueing the mp because we are holding the drain lock.
7461 	 * One of the advantages of synchronous stream is the ability for
7462 	 * us to find out when the application performs a read on the
7463 	 * socket by way of udp_rrw() entry point being called.  We need
7464 	 * to generate SIGPOLL/SIGIO for each received data in the case
7465 	 * of asynchronous socket just as in the strrput() case.  However,
7466 	 * we only wake the application up when necessary, i.e. during the
7467 	 * first enqueue.  When udp_rrw() is called, we send up a single
7468 	 * datagram upstream and call STR_WAKEUP_SET() again when there
7469 	 * are still data remaining in our receive queue.
7470 	 */
7471 	STR_WAKEUP_SENDSIG(STREAM(q), udp->udp_rcv_list_head);
7472 	if (udp->udp_rcv_list_head == NULL)
7473 		udp->udp_rcv_list_head = mp;
7474 	else
7475 		udp->udp_rcv_list_tail->b_next = mp;
7476 	udp->udp_rcv_list_tail = mp;
7477 	udp->udp_rcv_cnt += pkt_len;
7478 	udp->udp_rcv_msgcnt++;
7479 
7480 	/* Need to flow-control? */
7481 	if (udp->udp_rcv_cnt >= udp->udp_rcv_hiwat ||
7482 	    udp->udp_rcv_msgcnt >= udp->udp_rcv_hiwat)
7483 		udp->udp_drain_qfull = B_TRUE;
7484 
7485 	mutex_exit(&udp->udp_drain_lock);
7486 }
7487 
7488 /*
7489  * Drain the contents of receive list to the module upstream; we do
7490  * this during close or when we fallback to the slow mode due to
7491  * sockmod being popped or a module being pushed on top of us.
7492  */
7493 static void
7494 udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing)
7495 {
7496 	mblk_t *mp;
7497 	udp_stack_t *us = udp->udp_us;
7498 
7499 	mutex_enter(&udp->udp_drain_lock);
7500 	/*
7501 	 * There is no race with a concurrent udp_input() sending
7502 	 * up packets using putnext() after we have cleared the
7503 	 * udp_direct_sockfs flag but before we have completed
7504 	 * sending up the packets in udp_rcv_list, since we are
7505 	 * either a writer or we have quiesced the conn.
7506 	 */
7507 	udp->udp_direct_sockfs = B_FALSE;
7508 	mutex_exit(&udp->udp_drain_lock);
7509 
7510 	if (udp->udp_rcv_list_head != NULL)
7511 		UDP_STAT(us, udp_drain);
7512 
7513 	/*
7514 	 * Send up everything via putnext(); note here that we
7515 	 * don't need the udp_drain_lock to protect us since
7516 	 * nothing can enter udp_rrw() and that we currently
7517 	 * have exclusive access to this udp.
7518 	 */
7519 	while ((mp = udp->udp_rcv_list_head) != NULL) {
7520 		udp->udp_rcv_list_head = mp->b_next;
7521 		mp->b_next = NULL;
7522 		udp->udp_rcv_cnt -= msgdsize(mp);
7523 		udp->udp_rcv_msgcnt--;
7524 		if (closing) {
7525 			freemsg(mp);
7526 		} else {
7527 			ASSERT(q == RD(q));
7528 			putnext(q, mp);
7529 		}
7530 	}
7531 	ASSERT(udp->udp_rcv_cnt == 0);
7532 	ASSERT(udp->udp_rcv_msgcnt == 0);
7533 	ASSERT(udp->udp_rcv_list_head == NULL);
7534 	udp->udp_rcv_list_tail = NULL;
7535 	udp->udp_drain_qfull = B_FALSE;
7536 }
7537 
7538 static size_t
7539 udp_set_rcv_hiwat(udp_t *udp, size_t size)
7540 {
7541 	udp_stack_t *us = udp->udp_us;
7542 
7543 	/* We add a bit of extra buffering */
7544 	size += size >> 1;
7545 	if (size > us->us_max_buf)
7546 		size = us->us_max_buf;
7547 
7548 	udp->udp_rcv_hiwat = size;
7549 	return (size);
7550 }
7551 
7552 /*
7553  * For the lower queue so that UDP can be a dummy mux.
7554  * Nobody should be sending
7555  * packets up this stream
7556  */
7557 static void
7558 udp_lrput(queue_t *q, mblk_t *mp)
7559 {
7560 	mblk_t *mp1;
7561 
7562 	switch (mp->b_datap->db_type) {
7563 	case M_FLUSH:
7564 		/* Turn around */
7565 		if (*mp->b_rptr & FLUSHW) {
7566 			*mp->b_rptr &= ~FLUSHR;
7567 			qreply(q, mp);
7568 			return;
7569 		}
7570 		break;
7571 	}
7572 	/* Could receive messages that passed through ar_rput */
7573 	for (mp1 = mp; mp1; mp1 = mp1->b_cont)
7574 		mp1->b_prev = mp1->b_next = NULL;
7575 	freemsg(mp);
7576 }
7577 
7578 /*
7579  * For the lower queue so that UDP can be a dummy mux.
7580  * Nobody should be sending packets down this stream.
7581  */
7582 /* ARGSUSED */
7583 void
7584 udp_lwput(queue_t *q, mblk_t *mp)
7585 {
7586 	freemsg(mp);
7587 }
7588 
7589 /*
7590  * Below routines for UDP socket module.
7591  */
7592 
7593 static conn_t *
7594 udp_do_open(cred_t *credp, boolean_t isv6, int flags)
7595 {
7596 	udp_t		*udp;
7597 	conn_t		*connp;
7598 	zoneid_t 	zoneid;
7599 	netstack_t 	*ns;
7600 	udp_stack_t 	*us;
7601 
7602 	ns = netstack_find_by_cred(credp);
7603 	ASSERT(ns != NULL);
7604 	us = ns->netstack_udp;
7605 	ASSERT(us != NULL);
7606 
7607 	/*
7608 	 * For exclusive stacks we set the zoneid to zero
7609 	 * to make UDP operate as if in the global zone.
7610 	 */
7611 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
7612 		zoneid = GLOBAL_ZONEID;
7613 	else
7614 		zoneid = crgetzoneid(credp);
7615 
7616 	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);
7617 
7618 	connp = ipcl_conn_create(IPCL_UDPCONN, flags, ns);
7619 	if (connp == NULL) {
7620 		netstack_rele(ns);
7621 		return (NULL);
7622 	}
7623 	udp = connp->conn_udp;
7624 
7625 	/*
7626 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
7627 	 * done by netstack_find_by_cred()
7628 	 */
7629 	netstack_rele(ns);
7630 
7631 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7632 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
7633 	ASSERT(connp->conn_udp == udp);
7634 	ASSERT(udp->udp_connp == connp);
7635 
7636 	/* Set the initial state of the stream and the privilege status. */
7637 	udp->udp_state = TS_UNBND;
7638 	if (isv6) {
7639 		udp->udp_family = AF_INET6;
7640 		udp->udp_ipversion = IPV6_VERSION;
7641 		udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
7642 		udp->udp_ttl = us->us_ipv6_hoplimit;
7643 		connp->conn_af_isv6 = B_TRUE;
7644 		connp->conn_flags |= IPCL_ISV6;
7645 	} else {
7646 		udp->udp_family = AF_INET;
7647 		udp->udp_ipversion = IPV4_VERSION;
7648 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE;
7649 		udp->udp_ttl = us->us_ipv4_ttl;
7650 		connp->conn_af_isv6 = B_FALSE;
7651 		connp->conn_flags &= ~IPCL_ISV6;
7652 	}
7653 
7654 	udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
7655 	udp->udp_pending_op = -1;
7656 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
7657 	connp->conn_zoneid = zoneid;
7658 
7659 	udp->udp_open_time = lbolt64;
7660 	udp->udp_open_pid = curproc->p_pid;
7661 
7662 	/*
7663 	 * If the caller has the process-wide flag set, then default to MAC
7664 	 * exempt mode.  This allows read-down to unlabeled hosts.
7665 	 */
7666 	if (getpflags(NET_MAC_AWARE, credp) != 0)
7667 		connp->conn_mac_exempt = B_TRUE;
7668 
7669 	connp->conn_ulp_labeled = is_system_labeled();
7670 
7671 	udp->udp_us = us;
7672 
7673 	connp->conn_recv = udp_input;
7674 	crhold(credp);
7675 	connp->conn_cred = credp;
7676 
7677 	*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
7678 
7679 	rw_exit(&udp->udp_rwlock);
7680 
7681 	return (connp);
7682 }
7683 
7684 /* ARGSUSED */
7685 sock_lower_handle_t
7686 udp_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
7687     uint_t *smodep, int *errorp, int flags, cred_t *credp)
7688 {
7689 	udp_t		*udp = NULL;
7690 	udp_stack_t	*us;
7691 	conn_t		*connp;
7692 	boolean_t	isv6;
7693 
7694 	if (type != SOCK_DGRAM || (family != AF_INET && family != AF_INET6) ||
7695 	    (proto != 0 && proto != IPPROTO_UDP)) {
7696 		*errorp = EPROTONOSUPPORT;
7697 		return (NULL);
7698 	}
7699 
7700 	if (family == AF_INET6)
7701 		isv6 = B_TRUE;
7702 	else
7703 		isv6 = B_FALSE;
7704 
7705 	connp = udp_do_open(credp, isv6, flags);
7706 	if (connp == NULL) {
7707 		*errorp = ENOMEM;
7708 		return (NULL);
7709 	}
7710 
7711 	udp = connp->conn_udp;
7712 	ASSERT(udp != NULL);
7713 	us = udp->udp_us;
7714 	ASSERT(us != NULL);
7715 
7716 	connp->conn_flags |= IPCL_NONSTR | IPCL_SOCKET;
7717 
7718 	/* Set flow control */
7719 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7720 	(void) udp_set_rcv_hiwat(udp, us->us_recv_hiwat);
7721 	udp->udp_rcv_disply_hiwat = us->us_recv_hiwat;
7722 	udp->udp_rcv_lowat = udp_mod_info.mi_lowat;
7723 	udp->udp_xmit_hiwat = us->us_xmit_hiwat;
7724 	udp->udp_xmit_lowat = us->us_xmit_lowat;
7725 
7726 	if (udp->udp_family == AF_INET6) {
7727 		/* Build initial header template for transmit */
7728 		if ((*errorp = udp_build_hdrs(udp)) != 0) {
7729 			rw_exit(&udp->udp_rwlock);
7730 			ipcl_conn_destroy(connp);
7731 			return (NULL);
7732 		}
7733 	}
7734 	rw_exit(&udp->udp_rwlock);
7735 
7736 	connp->conn_flow_cntrld = B_FALSE;
7737 
7738 	ASSERT(us->us_ldi_ident != NULL);
7739 
7740 	if ((*errorp = ip_create_helper_stream(connp, us->us_ldi_ident)) != 0) {
7741 		ip1dbg(("create of IP helper stream failed\n"));
7742 		udp_do_close(connp);
7743 		return (NULL);
7744 	}
7745 
7746 	/* Set the send flow control */
7747 	connp->conn_wq->q_hiwat = us->us_xmit_hiwat;
7748 	connp->conn_wq->q_lowat = us->us_xmit_lowat;
7749 
7750 	mutex_enter(&connp->conn_lock);
7751 	connp->conn_state_flags &= ~CONN_INCIPIENT;
7752 	mutex_exit(&connp->conn_lock);
7753 
7754 	*errorp = 0;
7755 	*smodep = SM_ATOMIC;
7756 	*sock_downcalls = &sock_udp_downcalls;
7757 	return ((sock_lower_handle_t)connp);
7758 }
7759 
7760 /* ARGSUSED */
7761 void
7762 udp_activate(sock_lower_handle_t proto_handle, sock_upper_handle_t sock_handle,
7763     sock_upcalls_t *sock_upcalls, int flags, cred_t *cr)
7764 {
7765 	conn_t 		*connp = (conn_t *)proto_handle;
7766 	udp_t 		*udp = connp->conn_udp;
7767 	udp_stack_t	*us = udp->udp_us;
7768 	struct sock_proto_props sopp;
7769 
7770 	connp->conn_upcalls = sock_upcalls;
7771 	connp->conn_upper_handle = sock_handle;
7772 
7773 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT |
7774 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
7775 	sopp.sopp_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
7776 	sopp.sopp_maxblk = INFPSZ;
7777 	sopp.sopp_rxhiwat = udp->udp_rcv_hiwat;
7778 	sopp.sopp_maxaddrlen = sizeof (sin6_t);
7779 	sopp.sopp_maxpsz =
7780 	    (udp->udp_family == AF_INET) ? UDP_MAXPACKET_IPV4 :
7781 	    UDP_MAXPACKET_IPV6;
7782 	sopp.sopp_minpsz = (udp_mod_info.mi_minpsz == 1) ? 0 :
7783 	    udp_mod_info.mi_minpsz;
7784 
7785 	(*connp->conn_upcalls->su_set_proto_props)(connp->conn_upper_handle,
7786 	    &sopp);
7787 }
7788 
7789 static void
7790 udp_do_close(conn_t *connp)
7791 {
7792 	udp_t	*udp;
7793 
7794 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
7795 	udp = connp->conn_udp;
7796 
7797 	udp_quiesce_conn(connp);
7798 	ip_quiesce_conn(connp);
7799 
7800 	if (!IPCL_IS_NONSTR(connp)) {
7801 		/*
7802 		 * Disable read-side synchronous stream
7803 		 * interface and drain any queued data.
7804 		 */
7805 		ASSERT(connp->conn_wq != NULL);
7806 		udp_rcv_drain(connp->conn_wq, udp, B_TRUE);
7807 		ASSERT(!udp->udp_direct_sockfs);
7808 
7809 		ASSERT(connp->conn_rq != NULL);
7810 		qprocsoff(connp->conn_rq);
7811 	}
7812 
7813 	ASSERT(udp->udp_rcv_cnt == 0);
7814 	ASSERT(udp->udp_rcv_msgcnt == 0);
7815 	ASSERT(udp->udp_rcv_list_head == NULL);
7816 	ASSERT(udp->udp_rcv_list_tail == NULL);
7817 
7818 	udp_close_free(connp);
7819 
7820 	/*
7821 	 * Now we are truly single threaded on this stream, and can
7822 	 * delete the things hanging off the connp, and finally the connp.
7823 	 * We removed this connp from the fanout list, it cannot be
7824 	 * accessed thru the fanouts, and we already waited for the
7825 	 * conn_ref to drop to 0. We are already in close, so
7826 	 * there cannot be any other thread from the top. qprocsoff
7827 	 * has completed, and service has completed or won't run in
7828 	 * future.
7829 	 */
7830 	ASSERT(connp->conn_ref == 1);
7831 	if (!IPCL_IS_NONSTR(connp)) {
7832 		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
7833 	} else {
7834 		ip_close_helper_stream(connp);
7835 	}
7836 
7837 	connp->conn_ref--;
7838 	ipcl_conn_destroy(connp);
7839 }
7840 
7841 /* ARGSUSED */
7842 int
7843 udp_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
7844 {
7845 	conn_t	*connp = (conn_t *)proto_handle;
7846 
7847 	udp_do_close(connp);
7848 	return (0);
7849 }
7850 
7851 static int
7852 udp_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len, cred_t *cr,
7853     boolean_t bind_to_req_port_only)
7854 {
7855 	sin_t		*sin;
7856 	sin6_t		*sin6;
7857 	sin6_t		sin6addr;
7858 	in_port_t	port;		/* Host byte order */
7859 	in_port_t	requested_port;	/* Host byte order */
7860 	int		count;
7861 	in6_addr_t	v6src;
7862 	int		loopmax;
7863 	udp_fanout_t	*udpf;
7864 	in_port_t	lport;		/* Network byte order */
7865 	zoneid_t	zoneid;
7866 	udp_t		*udp;
7867 	boolean_t	is_inaddr_any;
7868 	mlp_type_t	addrtype, mlptype;
7869 	udp_stack_t	*us;
7870 	int		error = 0;
7871 	mblk_t		*mp = NULL;
7872 
7873 	udp = connp->conn_udp;
7874 	us = udp->udp_us;
7875 
7876 	if (udp->udp_state != TS_UNBND) {
7877 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7878 		    "udp_bind: bad state, %u", udp->udp_state);
7879 		return (-TOUTSTATE);
7880 	}
7881 
7882 	switch (len) {
7883 	case 0:
7884 		if (udp->udp_family == AF_INET) {
7885 			sin = (sin_t *)&sin6addr;
7886 			*sin = sin_null;
7887 			sin->sin_family = AF_INET;
7888 			sin->sin_addr.s_addr = INADDR_ANY;
7889 			udp->udp_ipversion = IPV4_VERSION;
7890 		} else {
7891 			ASSERT(udp->udp_family == AF_INET6);
7892 			sin6 = (sin6_t *)&sin6addr;
7893 			*sin6 = sin6_null;
7894 			sin6->sin6_family = AF_INET6;
7895 			V6_SET_ZERO(sin6->sin6_addr);
7896 			udp->udp_ipversion = IPV6_VERSION;
7897 		}
7898 		port = 0;
7899 		break;
7900 
7901 	case sizeof (sin_t):	/* Complete IPv4 address */
7902 		sin = (sin_t *)sa;
7903 
7904 		if (sin == NULL || !OK_32PTR((char *)sin))
7905 			return (EINVAL);
7906 
7907 		if (udp->udp_family != AF_INET ||
7908 		    sin->sin_family != AF_INET) {
7909 			return (EAFNOSUPPORT);
7910 		}
7911 		port = ntohs(sin->sin_port);
7912 		break;
7913 
7914 	case sizeof (sin6_t):	/* complete IPv6 address */
7915 		sin6 = (sin6_t *)sa;
7916 
7917 		if (sin6 == NULL || !OK_32PTR((char *)sin6))
7918 			return (EINVAL);
7919 
7920 		if (udp->udp_family != AF_INET6 ||
7921 		    sin6->sin6_family != AF_INET6) {
7922 			return (EAFNOSUPPORT);
7923 		}
7924 		port = ntohs(sin6->sin6_port);
7925 		break;
7926 
7927 	default:		/* Invalid request */
7928 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7929 		    "udp_bind: bad ADDR_length length %u", len);
7930 		return (-TBADADDR);
7931 	}
7932 
7933 	requested_port = port;
7934 
7935 	if (requested_port == 0 || !bind_to_req_port_only)
7936 		bind_to_req_port_only = B_FALSE;
7937 	else		/* T_BIND_REQ and requested_port != 0 */
7938 		bind_to_req_port_only = B_TRUE;
7939 
7940 	if (requested_port == 0) {
7941 		/*
7942 		 * If the application passed in zero for the port number, it
7943 		 * doesn't care which port number we bind to. Get one in the
7944 		 * valid range.
7945 		 */
7946 		if (udp->udp_anon_priv_bind) {
7947 			port = udp_get_next_priv_port(udp);
7948 		} else {
7949 			port = udp_update_next_port(udp,
7950 			    us->us_next_port_to_try, B_TRUE);
7951 		}
7952 	} else {
7953 		/*
7954 		 * If the port is in the well-known privileged range,
7955 		 * make sure the caller was privileged.
7956 		 */
7957 		int i;
7958 		boolean_t priv = B_FALSE;
7959 
7960 		if (port < us->us_smallest_nonpriv_port) {
7961 			priv = B_TRUE;
7962 		} else {
7963 			for (i = 0; i < us->us_num_epriv_ports; i++) {
7964 				if (port == us->us_epriv_ports[i]) {
7965 					priv = B_TRUE;
7966 					break;
7967 				}
7968 			}
7969 		}
7970 
7971 		if (priv) {
7972 			if (secpolicy_net_privaddr(cr, port, IPPROTO_UDP) != 0)
7973 				return (-TACCES);
7974 		}
7975 	}
7976 
7977 	if (port == 0)
7978 		return (-TNOADDR);
7979 
7980 	/*
7981 	 * The state must be TS_UNBND. TPI mandates that users must send
7982 	 * TPI primitives only 1 at a time and wait for the response before
7983 	 * sending the next primitive.
7984 	 */
7985 	rw_enter(&udp->udp_rwlock, RW_WRITER);
7986 	if (udp->udp_state != TS_UNBND || udp->udp_pending_op != -1) {
7987 		rw_exit(&udp->udp_rwlock);
7988 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
7989 		    "udp_bind: bad state, %u", udp->udp_state);
7990 		return (-TOUTSTATE);
7991 	}
7992 	/* XXX how to remove the T_BIND_REQ? Should set it before calling */
7993 	udp->udp_pending_op = T_BIND_REQ;
7994 	/*
7995 	 * Copy the source address into our udp structure. This address
7996 	 * may still be zero; if so, IP will fill in the correct address
7997 	 * each time an outbound packet is passed to it. Since the udp is
7998 	 * not yet in the bind hash list, we don't grab the uf_lock to
7999 	 * change udp_ipversion
8000 	 */
8001 	if (udp->udp_family == AF_INET) {
8002 		ASSERT(sin != NULL);
8003 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8004 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
8005 		    udp->udp_ip_snd_options_len;
8006 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src);
8007 	} else {
8008 		ASSERT(sin6 != NULL);
8009 		v6src = sin6->sin6_addr;
8010 		if (IN6_IS_ADDR_V4MAPPED(&v6src)) {
8011 			/*
8012 			 * no need to hold the uf_lock to set the udp_ipversion
8013 			 * since we are not yet in the fanout list
8014 			 */
8015 			udp->udp_ipversion = IPV4_VERSION;
8016 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
8017 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
8018 		} else {
8019 			udp->udp_ipversion = IPV6_VERSION;
8020 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
8021 		}
8022 	}
8023 
8024 	/*
8025 	 * If udp_reuseaddr is not set, then we have to make sure that
8026 	 * the IP address and port number the application requested
8027 	 * (or we selected for the application) is not being used by
8028 	 * another stream.  If another stream is already using the
8029 	 * requested IP address and port, the behavior depends on
8030 	 * "bind_to_req_port_only". If set the bind fails; otherwise we
8031 	 * search for any an unused port to bind to the the stream.
8032 	 *
8033 	 * As per the BSD semantics, as modified by the Deering multicast
8034 	 * changes, if udp_reuseaddr is set, then we allow multiple binds
8035 	 * to the same port independent of the local IP address.
8036 	 *
8037 	 * This is slightly different than in SunOS 4.X which did not
8038 	 * support IP multicast. Note that the change implemented by the
8039 	 * Deering multicast code effects all binds - not only binding
8040 	 * to IP multicast addresses.
8041 	 *
8042 	 * Note that when binding to port zero we ignore SO_REUSEADDR in
8043 	 * order to guarantee a unique port.
8044 	 */
8045 
8046 	count = 0;
8047 	if (udp->udp_anon_priv_bind) {
8048 		/*
8049 		 * loopmax = (IPPORT_RESERVED-1) -
8050 		 *    us->us_min_anonpriv_port + 1
8051 		 */
8052 		loopmax = IPPORT_RESERVED - us->us_min_anonpriv_port;
8053 	} else {
8054 		loopmax = us->us_largest_anon_port -
8055 		    us->us_smallest_anon_port + 1;
8056 	}
8057 
8058 	is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src);
8059 	zoneid = connp->conn_zoneid;
8060 
8061 	for (;;) {
8062 		udp_t		*udp1;
8063 		boolean_t	found_exclbind = B_FALSE;
8064 
8065 		/*
8066 		 * Walk through the list of udp streams bound to
8067 		 * requested port with the same IP address.
8068 		 */
8069 		lport = htons(port);
8070 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(lport,
8071 		    us->us_bind_fanout_size)];
8072 		mutex_enter(&udpf->uf_lock);
8073 		for (udp1 = udpf->uf_udp; udp1 != NULL;
8074 		    udp1 = udp1->udp_bind_hash) {
8075 			if (lport != udp1->udp_port)
8076 				continue;
8077 
8078 			/*
8079 			 * On a labeled system, we must treat bindings to ports
8080 			 * on shared IP addresses by sockets with MAC exemption
8081 			 * privilege as being in all zones, as there's
8082 			 * otherwise no way to identify the right receiver.
8083 			 */
8084 			if (!(IPCL_ZONE_MATCH(udp1->udp_connp, zoneid) ||
8085 			    IPCL_ZONE_MATCH(connp,
8086 			    udp1->udp_connp->conn_zoneid)) &&
8087 			    !connp->conn_mac_exempt && \
8088 			    !udp1->udp_connp->conn_mac_exempt)
8089 				continue;
8090 
8091 			/*
8092 			 * If UDP_EXCLBIND is set for either the bound or
8093 			 * binding endpoint, the semantics of bind
8094 			 * is changed according to the following chart.
8095 			 *
8096 			 * spec = specified address (v4 or v6)
8097 			 * unspec = unspecified address (v4 or v6)
8098 			 * A = specified addresses are different for endpoints
8099 			 *
8100 			 * bound	bind to		allowed?
8101 			 * -------------------------------------
8102 			 * unspec	unspec		no
8103 			 * unspec	spec		no
8104 			 * spec		unspec		no
8105 			 * spec		spec		yes if A
8106 			 *
8107 			 * For labeled systems, SO_MAC_EXEMPT behaves the same
8108 			 * as UDP_EXCLBIND, except that zoneid is ignored.
8109 			 */
8110 			if (udp1->udp_exclbind || udp->udp_exclbind ||
8111 			    udp1->udp_connp->conn_mac_exempt ||
8112 			    connp->conn_mac_exempt) {
8113 				if (V6_OR_V4_INADDR_ANY(
8114 				    udp1->udp_bound_v6src) ||
8115 				    is_inaddr_any ||
8116 				    IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
8117 				    &v6src)) {
8118 					found_exclbind = B_TRUE;
8119 					break;
8120 				}
8121 				continue;
8122 			}
8123 
8124 			/*
8125 			 * Check ipversion to allow IPv4 and IPv6 sockets to
8126 			 * have disjoint port number spaces.
8127 			 */
8128 			if (udp->udp_ipversion != udp1->udp_ipversion) {
8129 
8130 				/*
8131 				 * On the first time through the loop, if the
8132 				 * the user intentionally specified a
8133 				 * particular port number, then ignore any
8134 				 * bindings of the other protocol that may
8135 				 * conflict. This allows the user to bind IPv6
8136 				 * alone and get both v4 and v6, or bind both
8137 				 * both and get each seperately. On subsequent
8138 				 * times through the loop, we're checking a
8139 				 * port that we chose (not the user) and thus
8140 				 * we do not allow casual duplicate bindings.
8141 				 */
8142 				if (count == 0 && requested_port != 0)
8143 					continue;
8144 			}
8145 
8146 			/*
8147 			 * No difference depending on SO_REUSEADDR.
8148 			 *
8149 			 * If existing port is bound to a
8150 			 * non-wildcard IP address and
8151 			 * the requesting stream is bound to
8152 			 * a distinct different IP addresses
8153 			 * (non-wildcard, also), keep going.
8154 			 */
8155 			if (!is_inaddr_any &&
8156 			    !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) &&
8157 			    !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
8158 			    &v6src)) {
8159 				continue;
8160 			}
8161 			break;
8162 		}
8163 
8164 		if (!found_exclbind &&
8165 		    (udp->udp_reuseaddr && requested_port != 0)) {
8166 			break;
8167 		}
8168 
8169 		if (udp1 == NULL) {
8170 			/*
8171 			 * No other stream has this IP address
8172 			 * and port number. We can use it.
8173 			 */
8174 			break;
8175 		}
8176 		mutex_exit(&udpf->uf_lock);
8177 		if (bind_to_req_port_only) {
8178 			/*
8179 			 * We get here only when requested port
8180 			 * is bound (and only first  of the for()
8181 			 * loop iteration).
8182 			 *
8183 			 * The semantics of this bind request
8184 			 * require it to fail so we return from
8185 			 * the routine (and exit the loop).
8186 			 *
8187 			 */
8188 			udp->udp_pending_op = -1;
8189 			rw_exit(&udp->udp_rwlock);
8190 			return (-TADDRBUSY);
8191 		}
8192 
8193 		if (udp->udp_anon_priv_bind) {
8194 			port = udp_get_next_priv_port(udp);
8195 		} else {
8196 			if ((count == 0) && (requested_port != 0)) {
8197 				/*
8198 				 * If the application wants us to find
8199 				 * a port, get one to start with. Set
8200 				 * requested_port to 0, so that we will
8201 				 * update us->us_next_port_to_try below.
8202 				 */
8203 				port = udp_update_next_port(udp,
8204 				    us->us_next_port_to_try, B_TRUE);
8205 				requested_port = 0;
8206 			} else {
8207 				port = udp_update_next_port(udp, port + 1,
8208 				    B_FALSE);
8209 			}
8210 		}
8211 
8212 		if (port == 0 || ++count >= loopmax) {
8213 			/*
8214 			 * We've tried every possible port number and
8215 			 * there are none available, so send an error
8216 			 * to the user.
8217 			 */
8218 			udp->udp_pending_op = -1;
8219 			rw_exit(&udp->udp_rwlock);
8220 			return (-TNOADDR);
8221 		}
8222 	}
8223 
8224 	/*
8225 	 * Copy the source address into our udp structure.  This address
8226 	 * may still be zero; if so, ip will fill in the correct address
8227 	 * each time an outbound packet is passed to it.
8228 	 * If we are binding to a broadcast or multicast address then
8229 	 * udp_post_ip_bind_connect will clear the source address
8230 	 * when udp_do_bind success.
8231 	 */
8232 	udp->udp_v6src = udp->udp_bound_v6src = v6src;
8233 	udp->udp_port = lport;
8234 	/*
8235 	 * Now reset the the next anonymous port if the application requested
8236 	 * an anonymous port, or we handed out the next anonymous port.
8237 	 */
8238 	if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) {
8239 		us->us_next_port_to_try = port + 1;
8240 	}
8241 
8242 	/* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */
8243 	if (udp->udp_family == AF_INET) {
8244 		sin->sin_port = udp->udp_port;
8245 	} else {
8246 		sin6->sin6_port = udp->udp_port;
8247 		/* Rebuild the header template */
8248 		error = udp_build_hdrs(udp);
8249 		if (error != 0) {
8250 			udp->udp_pending_op = -1;
8251 			rw_exit(&udp->udp_rwlock);
8252 			mutex_exit(&udpf->uf_lock);
8253 			return (error);
8254 		}
8255 	}
8256 	udp->udp_state = TS_IDLE;
8257 	udp_bind_hash_insert(udpf, udp);
8258 	mutex_exit(&udpf->uf_lock);
8259 	rw_exit(&udp->udp_rwlock);
8260 
8261 	if (cl_inet_bind) {
8262 		/*
8263 		 * Running in cluster mode - register bind information
8264 		 */
8265 		if (udp->udp_ipversion == IPV4_VERSION) {
8266 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET,
8267 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8268 			    (in_port_t)udp->udp_port);
8269 		} else {
8270 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET6,
8271 			    (uint8_t *)&(udp->udp_v6src),
8272 			    (in_port_t)udp->udp_port);
8273 		}
8274 
8275 	}
8276 
8277 	connp->conn_anon_port = (is_system_labeled() && requested_port == 0);
8278 	if (is_system_labeled() && (!connp->conn_anon_port ||
8279 	    connp->conn_anon_mlp)) {
8280 		uint16_t mlpport;
8281 		cred_t *cr = connp->conn_cred;
8282 		zone_t *zone;
8283 
8284 		zone = crgetzone(cr);
8285 		connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth :
8286 		    mlptSingle;
8287 		addrtype = tsol_mlp_addr_type(zone->zone_id, IPV6_VERSION,
8288 		    &v6src, us->us_netstack->netstack_ip);
8289 		if (addrtype == mlptSingle) {
8290 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8291 			udp->udp_pending_op = -1;
8292 			rw_exit(&udp->udp_rwlock);
8293 			connp->conn_anon_port = B_FALSE;
8294 			connp->conn_mlp_type = mlptSingle;
8295 			return (-TNOADDR);
8296 		}
8297 		mlpport = connp->conn_anon_port ? PMAPPORT : port;
8298 		mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport,
8299 		    addrtype);
8300 		if (mlptype != mlptSingle &&
8301 		    (connp->conn_mlp_type == mlptSingle ||
8302 		    secpolicy_net_bindmlp(cr) != 0)) {
8303 			if (udp->udp_debug) {
8304 				(void) strlog(UDP_MOD_ID, 0, 1,
8305 				    SL_ERROR|SL_TRACE,
8306 				    "udp_bind: no priv for multilevel port %d",
8307 				    mlpport);
8308 			}
8309 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8310 			udp->udp_pending_op = -1;
8311 			rw_exit(&udp->udp_rwlock);
8312 			connp->conn_anon_port = B_FALSE;
8313 			connp->conn_mlp_type = mlptSingle;
8314 			return (-TACCES);
8315 		}
8316 
8317 		/*
8318 		 * If we're specifically binding a shared IP address and the
8319 		 * port is MLP on shared addresses, then check to see if this
8320 		 * zone actually owns the MLP.  Reject if not.
8321 		 */
8322 		if (mlptype == mlptShared && addrtype == mlptShared) {
8323 			/*
8324 			 * No need to handle exclusive-stack zones since
8325 			 * ALL_ZONES only applies to the shared stack.
8326 			 */
8327 			zoneid_t mlpzone;
8328 
8329 			mlpzone = tsol_mlp_findzone(IPPROTO_UDP,
8330 			    htons(mlpport));
8331 			if (connp->conn_zoneid != mlpzone) {
8332 				if (udp->udp_debug) {
8333 					(void) strlog(UDP_MOD_ID, 0, 1,
8334 					    SL_ERROR|SL_TRACE,
8335 					    "udp_bind: attempt to bind port "
8336 					    "%d on shared addr in zone %d "
8337 					    "(should be %d)",
8338 					    mlpport, connp->conn_zoneid,
8339 					    mlpzone);
8340 				}
8341 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8342 				udp->udp_pending_op = -1;
8343 				rw_exit(&udp->udp_rwlock);
8344 				connp->conn_anon_port = B_FALSE;
8345 				connp->conn_mlp_type = mlptSingle;
8346 				return (-TACCES);
8347 			}
8348 		}
8349 		if (connp->conn_anon_port) {
8350 			error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp,
8351 			    port, B_TRUE);
8352 			if (error != 0) {
8353 				if (udp->udp_debug) {
8354 					(void) strlog(UDP_MOD_ID, 0, 1,
8355 					    SL_ERROR|SL_TRACE,
8356 					    "udp_bind: cannot establish anon "
8357 					    "MLP for port %d", port);
8358 				}
8359 				rw_enter(&udp->udp_rwlock, RW_WRITER);
8360 				udp->udp_pending_op = -1;
8361 				rw_exit(&udp->udp_rwlock);
8362 				connp->conn_anon_port = B_FALSE;
8363 				connp->conn_mlp_type = mlptSingle;
8364 				return (-TACCES);
8365 			}
8366 		}
8367 		connp->conn_mlp_type = mlptype;
8368 	}
8369 
8370 	if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8371 		/*
8372 		 * Append a request for an IRE if udp_v6src not
8373 		 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address).
8374 		 */
8375 		mp = allocb(sizeof (ire_t), BPRI_HI);
8376 		if (!mp) {
8377 			rw_enter(&udp->udp_rwlock, RW_WRITER);
8378 			udp->udp_pending_op = -1;
8379 			rw_exit(&udp->udp_rwlock);
8380 			return (ENOMEM);
8381 		}
8382 		mp->b_wptr += sizeof (ire_t);
8383 		mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8384 	}
8385 	if (udp->udp_family == AF_INET6) {
8386 		ASSERT(udp->udp_connp->conn_af_isv6);
8387 		error = ip_proto_bind_laddr_v6(connp, &mp, IPPROTO_UDP,
8388 		    &udp->udp_bound_v6src, udp->udp_port, B_TRUE);
8389 	} else {
8390 		ASSERT(!udp->udp_connp->conn_af_isv6);
8391 		error = ip_proto_bind_laddr_v4(connp, &mp, IPPROTO_UDP,
8392 		    V4_PART_OF_V6(udp->udp_bound_v6src), udp->udp_port,
8393 		    B_TRUE);
8394 	}
8395 
8396 	(void) udp_post_ip_bind_connect(udp, mp, error);
8397 	return (error);
8398 }
8399 
8400 int
8401 udp_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
8402     socklen_t len, cred_t *cr)
8403 {
8404 	int		error;
8405 	conn_t		*connp;
8406 
8407 	connp = (conn_t *)proto_handle;
8408 
8409 	if (sa == NULL)
8410 		error = udp_do_unbind(connp);
8411 	else
8412 		error = udp_do_bind(connp, sa, len, cr, B_TRUE);
8413 
8414 	if (error < 0) {
8415 		if (error == -TOUTSTATE)
8416 			error = EINVAL;
8417 		else
8418 			error = proto_tlitosyserr(-error);
8419 	}
8420 
8421 	return (error);
8422 }
8423 
8424 static int
8425 udp_implicit_bind(conn_t *connp, cred_t *cr)
8426 {
8427 	int error;
8428 
8429 	error = udp_do_bind(connp, NULL, 0, cr, B_FALSE);
8430 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
8431 }
8432 
8433 /*
8434  * This routine removes a port number association from a stream. It
8435  * is called by udp_unbind and udp_tpi_unbind.
8436  */
8437 static int
8438 udp_do_unbind(conn_t *connp)
8439 {
8440 	udp_t 		*udp = connp->conn_udp;
8441 	udp_fanout_t	*udpf;
8442 	udp_stack_t	*us = udp->udp_us;
8443 
8444 	if (cl_inet_unbind != NULL) {
8445 		/*
8446 		 * Running in cluster mode - register unbind information
8447 		 */
8448 		if (udp->udp_ipversion == IPV4_VERSION) {
8449 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
8450 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
8451 			    (in_port_t)udp->udp_port);
8452 		} else {
8453 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
8454 			    (uint8_t *)&(udp->udp_v6src),
8455 			    (in_port_t)udp->udp_port);
8456 		}
8457 	}
8458 
8459 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8460 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8461 		rw_exit(&udp->udp_rwlock);
8462 		return (-TOUTSTATE);
8463 	}
8464 	udp->udp_pending_op = T_UNBIND_REQ;
8465 	rw_exit(&udp->udp_rwlock);
8466 
8467 	/*
8468 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
8469 	 * and therefore ip_unbind must never return NULL.
8470 	 */
8471 	ip_unbind(connp);
8472 
8473 	/*
8474 	 * Once we're unbound from IP, the pending operation may be cleared
8475 	 * here.
8476 	 */
8477 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8478 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8479 	    us->us_bind_fanout_size)];
8480 
8481 	mutex_enter(&udpf->uf_lock);
8482 	udp_bind_hash_remove(udp, B_TRUE);
8483 	V6_SET_ZERO(udp->udp_v6src);
8484 	V6_SET_ZERO(udp->udp_bound_v6src);
8485 	udp->udp_port = 0;
8486 	mutex_exit(&udpf->uf_lock);
8487 
8488 	udp->udp_pending_op = -1;
8489 	udp->udp_state = TS_UNBND;
8490 	if (udp->udp_family == AF_INET6)
8491 		(void) udp_build_hdrs(udp);
8492 	rw_exit(&udp->udp_rwlock);
8493 
8494 	return (0);
8495 }
8496 
8497 static int
8498 udp_post_ip_bind_connect(udp_t *udp, mblk_t *ire_mp, int error)
8499 {
8500 	ire_t		*ire;
8501 	udp_fanout_t	*udpf;
8502 	udp_stack_t	*us = udp->udp_us;
8503 
8504 	ASSERT(udp->udp_pending_op != -1);
8505 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8506 	if (error == 0) {
8507 		/* For udp_do_connect() success */
8508 		/* udp_do_bind() success will do nothing in here */
8509 		/*
8510 		 * If a broadcast/multicast address was bound, set
8511 		 * the source address to 0.
8512 		 * This ensures no datagrams with broadcast address
8513 		 * as source address are emitted (which would violate
8514 		 * RFC1122 - Hosts requirements)
8515 		 *
8516 		 * Note that when connecting the returned IRE is
8517 		 * for the destination address and we only perform
8518 		 * the broadcast check for the source address (it
8519 		 * is OK to connect to a broadcast/multicast address.)
8520 		 */
8521 		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
8522 			ire = (ire_t *)ire_mp->b_rptr;
8523 
8524 			/*
8525 			 * Note: we get IRE_BROADCAST for IPv6 to "mark" a
8526 			 * multicast local address.
8527 			 */
8528 			udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8529 			    us->us_bind_fanout_size)];
8530 			if (ire->ire_type == IRE_BROADCAST &&
8531 			    udp->udp_state != TS_DATA_XFER) {
8532 				ASSERT(udp->udp_pending_op == T_BIND_REQ ||
8533 				    udp->udp_pending_op == O_T_BIND_REQ);
8534 				/*
8535 				 * This was just a local bind to a broadcast
8536 				 * addr.
8537 				 */
8538 				mutex_enter(&udpf->uf_lock);
8539 				V6_SET_ZERO(udp->udp_v6src);
8540 				mutex_exit(&udpf->uf_lock);
8541 				if (udp->udp_family == AF_INET6)
8542 					(void) udp_build_hdrs(udp);
8543 			} else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
8544 				if (udp->udp_family == AF_INET6)
8545 					(void) udp_build_hdrs(udp);
8546 			}
8547 		}
8548 	} else {
8549 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8550 		    us->us_bind_fanout_size)];
8551 		mutex_enter(&udpf->uf_lock);
8552 
8553 		if (udp->udp_state == TS_DATA_XFER) {
8554 			/* Connect failed */
8555 			/* Revert back to the bound source */
8556 			udp->udp_v6src = udp->udp_bound_v6src;
8557 			udp->udp_state = TS_IDLE;
8558 		} else {
8559 			/* For udp_do_bind() failed */
8560 			V6_SET_ZERO(udp->udp_v6src);
8561 			V6_SET_ZERO(udp->udp_bound_v6src);
8562 			udp->udp_state = TS_UNBND;
8563 			udp_bind_hash_remove(udp, B_TRUE);
8564 			udp->udp_port = 0;
8565 		}
8566 		mutex_exit(&udpf->uf_lock);
8567 		if (udp->udp_family == AF_INET6)
8568 			(void) udp_build_hdrs(udp);
8569 	}
8570 	udp->udp_pending_op = -1;
8571 	rw_exit(&udp->udp_rwlock);
8572 	if (ire_mp != NULL)
8573 		freeb(ire_mp);
8574 	return (error);
8575 }
8576 
8577 /*
8578  * It associates a default destination address with the stream.
8579  */
8580 static int
8581 udp_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len)
8582 {
8583 	sin6_t		*sin6;
8584 	sin_t		*sin;
8585 	in6_addr_t 	v6dst;
8586 	ipaddr_t 	v4dst;
8587 	uint16_t 	dstport;
8588 	uint32_t 	flowinfo;
8589 	mblk_t		*ire_mp;
8590 	udp_fanout_t	*udpf;
8591 	udp_t		*udp, *udp1;
8592 	ushort_t	ipversion;
8593 	udp_stack_t	*us;
8594 	int		error;
8595 
8596 	udp = connp->conn_udp;
8597 	us = udp->udp_us;
8598 
8599 	/*
8600 	 * Address has been verified by the caller
8601 	 */
8602 	switch (len) {
8603 	default:
8604 		/*
8605 		 * Should never happen
8606 		 */
8607 		return (EINVAL);
8608 
8609 	case sizeof (sin_t):
8610 		sin = (sin_t *)sa;
8611 		v4dst = sin->sin_addr.s_addr;
8612 		dstport = sin->sin_port;
8613 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8614 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
8615 		ipversion = IPV4_VERSION;
8616 		break;
8617 
8618 	case sizeof (sin6_t):
8619 		sin6 = (sin6_t *)sa;
8620 		v6dst = sin6->sin6_addr;
8621 		dstport = sin6->sin6_port;
8622 		if (IN6_IS_ADDR_V4MAPPED(&v6dst)) {
8623 			IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst);
8624 			ipversion = IPV4_VERSION;
8625 			flowinfo = 0;
8626 		} else {
8627 			ipversion = IPV6_VERSION;
8628 			flowinfo = sin6->sin6_flowinfo;
8629 		}
8630 		break;
8631 	}
8632 
8633 	if (dstport == 0)
8634 		return (-TBADADDR);
8635 
8636 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8637 
8638 	/*
8639 	 * This UDP must have bound to a port already before doing a connect.
8640 	 * TPI mandates that users must send TPI primitives only 1 at a time
8641 	 * and wait for the response before sending the next primitive.
8642 	 */
8643 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
8644 		rw_exit(&udp->udp_rwlock);
8645 		(void) strlog(UDP_MOD_ID, 0, 1, SL_ERROR|SL_TRACE,
8646 		    "udp_connect: bad state, %u", udp->udp_state);
8647 		return (-TOUTSTATE);
8648 	}
8649 	udp->udp_pending_op = T_CONN_REQ;
8650 	ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL);
8651 
8652 	if (ipversion == IPV4_VERSION) {
8653 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
8654 		    udp->udp_ip_snd_options_len;
8655 	} else {
8656 		udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
8657 	}
8658 
8659 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
8660 	    us->us_bind_fanout_size)];
8661 
8662 	mutex_enter(&udpf->uf_lock);
8663 	if (udp->udp_state == TS_DATA_XFER) {
8664 		/* Already connected - clear out state */
8665 		udp->udp_v6src = udp->udp_bound_v6src;
8666 		udp->udp_state = TS_IDLE;
8667 	}
8668 
8669 	/*
8670 	 * Create a default IP header with no IP options.
8671 	 */
8672 	udp->udp_dstport = dstport;
8673 	udp->udp_ipversion = ipversion;
8674 	if (ipversion == IPV4_VERSION) {
8675 		/*
8676 		 * Interpret a zero destination to mean loopback.
8677 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8678 		 * generate the T_CONN_CON.
8679 		 */
8680 		if (v4dst == INADDR_ANY) {
8681 			v4dst = htonl(INADDR_LOOPBACK);
8682 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
8683 			if (udp->udp_family == AF_INET) {
8684 				sin->sin_addr.s_addr = v4dst;
8685 			} else {
8686 				sin6->sin6_addr = v6dst;
8687 			}
8688 		}
8689 		udp->udp_v6dst = v6dst;
8690 		udp->udp_flowinfo = 0;
8691 
8692 		/*
8693 		 * If the destination address is multicast and
8694 		 * an outgoing multicast interface has been set,
8695 		 * use the address of that interface as our
8696 		 * source address if no source address has been set.
8697 		 */
8698 		if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY &&
8699 		    CLASSD(v4dst) &&
8700 		    udp->udp_multicast_if_addr != INADDR_ANY) {
8701 			IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr,
8702 			    &udp->udp_v6src);
8703 		}
8704 	} else {
8705 		ASSERT(udp->udp_ipversion == IPV6_VERSION);
8706 		/*
8707 		 * Interpret a zero destination to mean loopback.
8708 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
8709 		 * generate the T_CONN_CON.
8710 		 */
8711 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
8712 			v6dst = ipv6_loopback;
8713 			sin6->sin6_addr = v6dst;
8714 		}
8715 		udp->udp_v6dst = v6dst;
8716 		udp->udp_flowinfo = flowinfo;
8717 		/*
8718 		 * If the destination address is multicast and
8719 		 * an outgoing multicast interface has been set,
8720 		 * then the ip bind logic will pick the correct source
8721 		 * address (i.e. matching the outgoing multicast interface).
8722 		 */
8723 	}
8724 
8725 	/*
8726 	 * Verify that the src/port/dst/port is unique for all
8727 	 * connections in TS_DATA_XFER
8728 	 */
8729 	for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) {
8730 		if (udp1->udp_state != TS_DATA_XFER)
8731 			continue;
8732 		if (udp->udp_port != udp1->udp_port ||
8733 		    udp->udp_ipversion != udp1->udp_ipversion ||
8734 		    dstport != udp1->udp_dstport ||
8735 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) ||
8736 		    !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst) ||
8737 		    !(IPCL_ZONE_MATCH(udp->udp_connp,
8738 		    udp1->udp_connp->conn_zoneid) ||
8739 		    IPCL_ZONE_MATCH(udp1->udp_connp,
8740 		    udp->udp_connp->conn_zoneid)))
8741 			continue;
8742 		mutex_exit(&udpf->uf_lock);
8743 		udp->udp_pending_op = -1;
8744 		rw_exit(&udp->udp_rwlock);
8745 		return (-TBADADDR);
8746 	}
8747 	udp->udp_state = TS_DATA_XFER;
8748 	mutex_exit(&udpf->uf_lock);
8749 
8750 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
8751 	if (ire_mp == NULL) {
8752 		mutex_enter(&udpf->uf_lock);
8753 		udp->udp_state = TS_IDLE;
8754 		udp->udp_pending_op = -1;
8755 		mutex_exit(&udpf->uf_lock);
8756 		rw_exit(&udp->udp_rwlock);
8757 		return (ENOMEM);
8758 	}
8759 
8760 	rw_exit(&udp->udp_rwlock);
8761 
8762 	ire_mp->b_wptr += sizeof (ire_t);
8763 	ire_mp->b_datap->db_type = IRE_DB_REQ_TYPE;
8764 
8765 	if (udp->udp_family == AF_INET) {
8766 		error = ip_proto_bind_connected_v4(connp, &ire_mp, IPPROTO_UDP,
8767 		    &V4_PART_OF_V6(udp->udp_v6src), udp->udp_port,
8768 		    V4_PART_OF_V6(udp->udp_v6dst), udp->udp_dstport,
8769 		    B_TRUE, B_TRUE);
8770 	} else {
8771 		error = ip_proto_bind_connected_v6(connp, &ire_mp, IPPROTO_UDP,
8772 		    &udp->udp_v6src, udp->udp_port, &udp->udp_v6dst,
8773 		    &udp->udp_sticky_ipp, udp->udp_dstport, B_TRUE, B_TRUE);
8774 	}
8775 
8776 	return (udp_post_ip_bind_connect(udp, ire_mp, error));
8777 }
8778 
8779 /* ARGSUSED */
8780 static int
8781 udp_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
8782     socklen_t len, sock_connid_t *id, cred_t *cr)
8783 {
8784 	conn_t	*connp = (conn_t *)proto_handle;
8785 	udp_t	*udp = connp->conn_udp;
8786 	int	error;
8787 	boolean_t did_bind = B_FALSE;
8788 
8789 	if (sa == NULL) {
8790 		/*
8791 		 * Disconnect
8792 		 * Make sure we are connected
8793 		 */
8794 		if (udp->udp_state != TS_DATA_XFER)
8795 			return (EINVAL);
8796 
8797 		error = udp_disconnect(connp);
8798 		return (error);
8799 	}
8800 
8801 	error = proto_verify_ip_addr(udp->udp_family, sa, len);
8802 	if (error != 0)
8803 		goto done;
8804 
8805 	/* do an implicit bind if necessary */
8806 	if (udp->udp_state == TS_UNBND) {
8807 		error = udp_implicit_bind(connp, cr);
8808 		/*
8809 		 * We could be racing with an actual bind, in which case
8810 		 * we would see EPROTO. We cross our fingers and try
8811 		 * to connect.
8812 		 */
8813 		if (!(error == 0 || error == EPROTO))
8814 			goto done;
8815 		did_bind = B_TRUE;
8816 	}
8817 	/*
8818 	 * set SO_DGRAM_ERRIND
8819 	 */
8820 	udp->udp_dgram_errind = B_TRUE;
8821 
8822 	error = udp_do_connect(connp, sa, len);
8823 
8824 	if (error != 0 && did_bind) {
8825 		int unbind_err;
8826 
8827 		unbind_err = udp_do_unbind(connp);
8828 		ASSERT(unbind_err == 0);
8829 	}
8830 
8831 	if (error == 0) {
8832 		*id = 0;
8833 		(*connp->conn_upcalls->su_connected)
8834 		    (connp->conn_upper_handle, 0, NULL, -1);
8835 	} else if (error < 0) {
8836 		error = proto_tlitosyserr(-error);
8837 	}
8838 
8839 done:
8840 	if (error != 0 && udp->udp_state == TS_DATA_XFER) {
8841 		/*
8842 		 * No need to hold locks to set state
8843 		 * after connect failure socket state is undefined
8844 		 * We set the state only to imitate old sockfs behavior
8845 		 */
8846 		udp->udp_state = TS_IDLE;
8847 	}
8848 	return (error);
8849 }
8850 
8851 /* ARGSUSED */
8852 int
8853 udp_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
8854     cred_t *cr)
8855 {
8856 	conn_t		*connp = (conn_t *)proto_handle;
8857 	udp_t		*udp = connp->conn_udp;
8858 	udp_stack_t	*us = udp->udp_us;
8859 	int		error = 0;
8860 
8861 	ASSERT(DB_TYPE(mp) == M_DATA);
8862 
8863 	/*
8864 	 * If the socket is connected and no change in destination
8865 	 */
8866 	if (msg->msg_namelen == 0) {
8867 		error = udp_send_connected(connp, mp, msg, cr, curproc->p_pid);
8868 		if (error == EDESTADDRREQ)
8869 			return (error);
8870 		else
8871 			return (udp->udp_dgram_errind ? error : 0);
8872 	}
8873 
8874 	/*
8875 	 * Do an implicit bind if necessary.
8876 	 */
8877 	if (udp->udp_state == TS_UNBND) {
8878 		error = udp_implicit_bind(connp, cr);
8879 		/*
8880 		 * We could be racing with an actual bind, in which case
8881 		 * we would see EPROTO. We cross our fingers and try
8882 		 * to send.
8883 		 */
8884 		if (!(error == 0 || error == EPROTO)) {
8885 			freemsg(mp);
8886 			return (error);
8887 		}
8888 	}
8889 
8890 	rw_enter(&udp->udp_rwlock, RW_WRITER);
8891 
8892 	if (msg->msg_name != NULL && udp->udp_state == TS_DATA_XFER) {
8893 		rw_exit(&udp->udp_rwlock);
8894 		freemsg(mp);
8895 		return (EISCONN);
8896 	}
8897 
8898 
8899 	if (udp->udp_delayed_error != 0) {
8900 		boolean_t	match;
8901 
8902 		error = udp->udp_delayed_error;
8903 		match = B_FALSE;
8904 		udp->udp_delayed_error = 0;
8905 		switch (udp->udp_family) {
8906 		case AF_INET: {
8907 			/* Compare just IP address and port */
8908 			sin_t *sin1 = (sin_t *)msg->msg_name;
8909 			sin_t *sin2 = (sin_t *)&udp->udp_delayed_addr;
8910 
8911 			if (msg->msg_namelen == sizeof (sin_t) &&
8912 			    sin1->sin_port == sin2->sin_port &&
8913 			    sin1->sin_addr.s_addr == sin2->sin_addr.s_addr)
8914 				match = B_TRUE;
8915 
8916 			break;
8917 		}
8918 		case AF_INET6: {
8919 			sin6_t	*sin1 = (sin6_t *)msg->msg_name;
8920 			sin6_t	*sin2 = (sin6_t *)&udp->udp_delayed_addr;
8921 
8922 			if (msg->msg_namelen == sizeof (sin6_t) &&
8923 			    sin1->sin6_port == sin2->sin6_port &&
8924 			    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
8925 			    &sin2->sin6_addr))
8926 				match = B_TRUE;
8927 			break;
8928 		}
8929 		default:
8930 			ASSERT(0);
8931 		}
8932 
8933 		*((sin6_t *)&udp->udp_delayed_addr) = sin6_null;
8934 
8935 		if (match) {
8936 			rw_exit(&udp->udp_rwlock);
8937 			freemsg(mp);
8938 			return (error);
8939 		}
8940 	}
8941 
8942 	error = proto_verify_ip_addr(udp->udp_family,
8943 	    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
8944 	rw_exit(&udp->udp_rwlock);
8945 
8946 	if (error != 0) {
8947 		freemsg(mp);
8948 		return (error);
8949 	}
8950 
8951 	error = udp_send_not_connected(connp, mp,
8952 	    (struct sockaddr  *)msg->msg_name, msg->msg_namelen, msg, cr,
8953 	    curproc->p_pid);
8954 	if (error != 0) {
8955 		UDP_STAT(us, udp_out_err_output);
8956 		freemsg(mp);
8957 	}
8958 	return (udp->udp_dgram_errind ? error : 0);
8959 }
8960 
8961 void
8962 udp_fallback(sock_lower_handle_t proto_handle, queue_t *q,
8963     boolean_t direct_sockfs, so_proto_quiesced_cb_t quiesced_cb)
8964 {
8965 	conn_t 	*connp = (conn_t *)proto_handle;
8966 	udp_t	*udp;
8967 	struct T_capability_ack tca;
8968 	struct sockaddr_in6 laddr, faddr;
8969 	socklen_t laddrlen, faddrlen;
8970 	short opts;
8971 	struct stroptions *stropt;
8972 	mblk_t *stropt_mp;
8973 	int error;
8974 
8975 	udp = connp->conn_udp;
8976 
8977 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
8978 
8979 	/*
8980 	 * setup the fallback stream that was allocated
8981 	 */
8982 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
8983 	connp->conn_minor_arena = WR(q)->q_ptr;
8984 
8985 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
8986 
8987 	WR(q)->q_qinfo = &udp_winit;
8988 
8989 	connp->conn_rq = RD(q);
8990 	connp->conn_wq = WR(q);
8991 
8992 	/* Notify stream head about options before sending up data */
8993 	stropt_mp->b_datap->db_type = M_SETOPTS;
8994 	stropt_mp->b_wptr += sizeof (*stropt);
8995 	stropt = (struct stroptions *)stropt_mp->b_rptr;
8996 	stropt->so_flags = SO_WROFF | SO_HIWAT;
8997 	stropt->so_wroff =
8998 	    (ushort_t)(udp->udp_max_hdr_len + udp->udp_us->us_wroff_extra);
8999 	stropt->so_hiwat = udp->udp_rcv_disply_hiwat;
9000 	putnext(RD(q), stropt_mp);
9001 
9002 	/*
9003 	 * Free the helper stream
9004 	 */
9005 	ip_close_helper_stream(connp);
9006 
9007 	if (!direct_sockfs)
9008 		udp_disable_direct_sockfs(udp);
9009 
9010 	/*
9011 	 * Collect the information needed to sync with the sonode
9012 	 */
9013 	udp_do_capability_ack(udp, &tca, TC1_INFO);
9014 
9015 	laddrlen = faddrlen = sizeof (sin6_t);
9016 	(void) udp_getsockname((sock_lower_handle_t)connp,
9017 	    (struct sockaddr *)&laddr, &laddrlen, NULL);
9018 	error = udp_getpeername((sock_lower_handle_t)connp,
9019 	    (struct sockaddr *)&faddr, &faddrlen, NULL);
9020 	if (error != 0)
9021 		faddrlen = 0;
9022 
9023 	opts = 0;
9024 	if (udp->udp_dgram_errind)
9025 		opts |= SO_DGRAM_ERRIND;
9026 	if (udp->udp_dontroute)
9027 		opts |= SO_DONTROUTE;
9028 
9029 	/*
9030 	 * Once we grab the drain lock, no data will be send up
9031 	 * to the socket. So we notify the socket that the endpoint
9032 	 * is quiescent and it's therefore safe move data from
9033 	 * the socket to the stream head.
9034 	 */
9035 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
9036 	    (struct sockaddr *)&laddr, laddrlen,
9037 	    (struct sockaddr *)&faddr, faddrlen, opts);
9038 
9039 	/*
9040 	 * push up any packets that were queued in udp_t
9041 	 */
9042 
9043 	mutex_enter(&udp->udp_recv_lock);
9044 	while (udp->udp_fallback_queue_head != NULL) {
9045 		mblk_t *mp;
9046 		mp = udp->udp_fallback_queue_head;
9047 		udp->udp_fallback_queue_head = mp->b_next;
9048 		mutex_exit(&udp->udp_recv_lock);
9049 		mp->b_next = NULL;
9050 		putnext(RD(q), mp);
9051 		mutex_enter(&udp->udp_recv_lock);
9052 	}
9053 	udp->udp_fallback_queue_tail = udp->udp_fallback_queue_head;
9054 	/*
9055 	 * No longer a streams less socket
9056 	 */
9057 	connp->conn_flags &= ~IPCL_NONSTR;
9058 	mutex_exit(&udp->udp_recv_lock);
9059 
9060 	ASSERT(connp->conn_ref >= 1);
9061 }
9062 
9063 static int
9064 udp_do_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
9065 {
9066 	sin_t	*sin = (sin_t *)sa;
9067 	sin6_t	*sin6 = (sin6_t *)sa;
9068 
9069 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
9070 	ASSERT(udp != NULL);
9071 
9072 	if (udp->udp_state != TS_DATA_XFER)
9073 		return (ENOTCONN);
9074 
9075 	switch (udp->udp_family) {
9076 	case AF_INET:
9077 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
9078 
9079 		if (*salenp < sizeof (sin_t))
9080 			return (EINVAL);
9081 
9082 		*salenp = sizeof (sin_t);
9083 		*sin = sin_null;
9084 		sin->sin_family = AF_INET;
9085 		sin->sin_port = udp->udp_dstport;
9086 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
9087 		break;
9088 	case AF_INET6:
9089 		if (*salenp < sizeof (sin6_t))
9090 			return (EINVAL);
9091 
9092 		*salenp = sizeof (sin6_t);
9093 		*sin6 = sin6_null;
9094 		sin6->sin6_family = AF_INET6;
9095 		sin6->sin6_port = udp->udp_dstport;
9096 		sin6->sin6_addr = udp->udp_v6dst;
9097 		sin6->sin6_flowinfo = udp->udp_flowinfo;
9098 		break;
9099 	}
9100 
9101 	return (0);
9102 }
9103 
9104 /* ARGSUSED */
9105 int
9106 udp_getpeername(sock_lower_handle_t  proto_handle, struct sockaddr *sa,
9107     socklen_t *salenp, cred_t *cr)
9108 {
9109 	conn_t	*connp = (conn_t *)proto_handle;
9110 	udp_t	*udp = connp->conn_udp;
9111 	int error;
9112 
9113 	ASSERT(udp != NULL);
9114 
9115 	rw_enter(&udp->udp_rwlock, RW_READER);
9116 
9117 	error = udp_do_getpeername(udp, sa, salenp);
9118 
9119 	rw_exit(&udp->udp_rwlock);
9120 
9121 	return (error);
9122 }
9123 
9124 static int
9125 udp_do_getsockname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
9126 {
9127 	sin_t	*sin = (sin_t *)sa;
9128 	sin6_t	*sin6 = (sin6_t *)sa;
9129 
9130 	ASSERT(udp != NULL);
9131 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
9132 
9133 	switch (udp->udp_family) {
9134 	case AF_INET:
9135 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
9136 
9137 		if (*salenp < sizeof (sin_t))
9138 			return (EINVAL);
9139 
9140 		*salenp = sizeof (sin_t);
9141 		*sin = sin_null;
9142 		sin->sin_family = AF_INET;
9143 		if (udp->udp_state == TS_UNBND) {
9144 			break;
9145 		}
9146 		sin->sin_port = udp->udp_port;
9147 
9148 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
9149 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9150 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
9151 		} else {
9152 			/*
9153 			 * INADDR_ANY
9154 			 * udp_v6src is not set, we might be bound to
9155 			 * broadcast/multicast. Use udp_bound_v6src as
9156 			 * local address instead (that could
9157 			 * also still be INADDR_ANY)
9158 			 */
9159 			sin->sin_addr.s_addr =
9160 			    V4_PART_OF_V6(udp->udp_bound_v6src);
9161 		}
9162 		break;
9163 
9164 	case AF_INET6:
9165 		if (*salenp < sizeof (sin6_t))
9166 			return (EINVAL);
9167 
9168 		*salenp = sizeof (sin6_t);
9169 		*sin6 = sin6_null;
9170 		sin6->sin6_family = AF_INET6;
9171 		if (udp->udp_state == TS_UNBND) {
9172 			break;
9173 		}
9174 		sin6->sin6_port = udp->udp_port;
9175 
9176 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
9177 			sin6->sin6_addr = udp->udp_v6src;
9178 		} else {
9179 			/*
9180 			 * UNSPECIFIED
9181 			 * udp_v6src is not set, we might be bound to
9182 			 * broadcast/multicast. Use udp_bound_v6src as
9183 			 * local address instead (that could
9184 			 * also still be UNSPECIFIED)
9185 			 */
9186 			sin6->sin6_addr = udp->udp_bound_v6src;
9187 		}
9188 	}
9189 	return (0);
9190 }
9191 
9192 /* ARGSUSED */
9193 int
9194 udp_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
9195     socklen_t *salenp, cred_t *cr)
9196 {
9197 	conn_t	*connp = (conn_t *)proto_handle;
9198 	udp_t	*udp = connp->conn_udp;
9199 	int error;
9200 
9201 	ASSERT(udp != NULL);
9202 	rw_enter(&udp->udp_rwlock, RW_READER);
9203 
9204 	error = udp_do_getsockname(udp, sa, salenp);
9205 
9206 	rw_exit(&udp->udp_rwlock);
9207 
9208 	return (error);
9209 }
9210 
9211 int
9212 udp_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9213     void *optvalp, socklen_t *optlen, cred_t *cr)
9214 {
9215 	conn_t		*connp = (conn_t *)proto_handle;
9216 	udp_t		*udp = connp->conn_udp;
9217 	int		error;
9218 	t_uscalar_t	max_optbuf_len;
9219 	void		*optvalp_buf;
9220 	int		len;
9221 
9222 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
9223 	    udp_opt_obj.odb_opt_des_arr,
9224 	    udp_opt_obj.odb_opt_arr_cnt,
9225 	    udp_opt_obj.odb_topmost_tpiprovider,
9226 	    B_FALSE, B_TRUE, cr);
9227 	if (error != 0) {
9228 		if (error < 0)
9229 			error = proto_tlitosyserr(-error);
9230 		return (error);
9231 	}
9232 
9233 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
9234 	rw_enter(&udp->udp_rwlock, RW_READER);
9235 	len = udp_opt_get(connp, level, option_name, optvalp_buf);
9236 	rw_exit(&udp->udp_rwlock);
9237 
9238 	if (len < 0) {
9239 		/*
9240 		 * Pass on to IP
9241 		 */
9242 		kmem_free(optvalp_buf, max_optbuf_len);
9243 		return (ip_get_options(connp, level, option_name,
9244 		    optvalp, optlen, cr));
9245 	} else {
9246 		/*
9247 		 * update optlen and copy option value
9248 		 */
9249 		t_uscalar_t size = MIN(len, *optlen);
9250 		bcopy(optvalp_buf, optvalp, size);
9251 		bcopy(&size, optlen, sizeof (size));
9252 
9253 		kmem_free(optvalp_buf, max_optbuf_len);
9254 		return (0);
9255 	}
9256 }
9257 
9258 int
9259 udp_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
9260     const void *optvalp, socklen_t optlen, cred_t *cr)
9261 {
9262 	conn_t		*connp = (conn_t *)proto_handle;
9263 	udp_t		*udp = connp->conn_udp;
9264 	int		error;
9265 
9266 	error = proto_opt_check(level, option_name, optlen, NULL,
9267 	    udp_opt_obj.odb_opt_des_arr,
9268 	    udp_opt_obj.odb_opt_arr_cnt,
9269 	    udp_opt_obj.odb_topmost_tpiprovider,
9270 	    B_TRUE, B_FALSE, cr);
9271 
9272 	if (error != 0) {
9273 		if (error < 0)
9274 			error = proto_tlitosyserr(-error);
9275 		return (error);
9276 	}
9277 
9278 	rw_enter(&udp->udp_rwlock, RW_WRITER);
9279 	error = udp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level, option_name,
9280 	    optlen, (uchar_t *)optvalp, (uint_t *)&optlen, (uchar_t *)optvalp,
9281 	    NULL, cr);
9282 	rw_exit(&udp->udp_rwlock);
9283 
9284 	if (error < 0) {
9285 		/*
9286 		 * Pass on to ip
9287 		 */
9288 		error = ip_set_options(connp, level, option_name, optvalp,
9289 		    optlen, cr);
9290 	}
9291 
9292 	return (error);
9293 }
9294 
9295 void
9296 udp_clr_flowctrl(sock_lower_handle_t proto_handle)
9297 {
9298 	conn_t	*connp = (conn_t *)proto_handle;
9299 	udp_t	*udp = connp->conn_udp;
9300 
9301 	mutex_enter(&udp->udp_recv_lock);
9302 	connp->conn_flow_cntrld = B_FALSE;
9303 	mutex_exit(&udp->udp_recv_lock);
9304 }
9305 
9306 /* ARGSUSED */
9307 int
9308 udp_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
9309 {
9310 	conn_t	*connp = (conn_t *)proto_handle;
9311 
9312 	/* shut down the send side */
9313 	if (how != SHUT_RD)
9314 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9315 		    SOCK_OPCTL_SHUT_SEND, 0);
9316 	/* shut down the recv side */
9317 	if (how != SHUT_WR)
9318 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
9319 		    SOCK_OPCTL_SHUT_RECV, 0);
9320 	return (0);
9321 }
9322 
9323 int
9324 udp_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
9325     int mode, int32_t *rvalp, cred_t *cr)
9326 {
9327 	conn_t  	*connp = (conn_t *)proto_handle;
9328 	int		error;
9329 
9330 	switch (cmd) {
9331 		case ND_SET:
9332 		case ND_GET:
9333 		case _SIOCSOCKFALLBACK:
9334 		case TI_GETPEERNAME:
9335 		case TI_GETMYNAME:
9336 			ip1dbg(("udp_ioctl: cmd 0x%x on non streams socket",
9337 			    cmd));
9338 			error = EINVAL;
9339 			break;
9340 		default:
9341 			/*
9342 			 * Pass on to IP using helper stream
9343 			 */
9344 			error = ldi_ioctl(
9345 			    connp->conn_helper_info->ip_helper_stream_handle,
9346 			    cmd, arg, mode, cr, rvalp);
9347 			break;
9348 	}
9349 	return (error);
9350 }
9351 
9352 /* ARGSUSED */
9353 int
9354 udp_accept(sock_lower_handle_t lproto_handle,
9355     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
9356     cred_t *cr)
9357 {
9358 	return (EOPNOTSUPP);
9359 }
9360 
9361 /* ARGSUSED */
9362 int
9363 udp_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
9364 {
9365 	return (EOPNOTSUPP);
9366 }
9367 
9368 sock_downcalls_t sock_udp_downcalls = {
9369 	udp_activate,		/* sd_activate */
9370 	udp_accept,		/* sd_accept */
9371 	udp_bind,		/* sd_bind */
9372 	udp_listen,		/* sd_listen */
9373 	udp_connect,		/* sd_connect */
9374 	udp_getpeername,	/* sd_getpeername */
9375 	udp_getsockname,	/* sd_getsockname */
9376 	udp_getsockopt,		/* sd_getsockopt */
9377 	udp_setsockopt,		/* sd_setsockopt */
9378 	udp_send,		/* sd_send */
9379 	NULL,			/* sd_send_uio */
9380 	NULL,			/* sd_recv_uio */
9381 	NULL,			/* sd_poll */
9382 	udp_shutdown,		/* sd_shutdown */
9383 	udp_clr_flowctrl,	/* sd_setflowctrl */
9384 	udp_ioctl,		/* sd_ioctl */
9385 	udp_close		/* sd_close */
9386 };
9387