xref: /titanic_51/usr/src/uts/common/inet/udp/udp.c (revision 53a7b6b6763f5865522a76e5e887390a8f4777d7)
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/cmn_err.h>
44 #include <sys/kmem.h>
45 #include <sys/policy.h>
46 #include <sys/ucred.h>
47 #include <sys/zone.h>
48 
49 #include <sys/socket.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/mi.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_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 void	udp_bind_result(conn_t *, mblk_t *);
157 static void	udp_bind_ack(conn_t *, mblk_t *mp);
158 static void	udp_bind_error(conn_t *, mblk_t *mp);
159 static int	udp_build_hdrs(udp_t *udp);
160 static void	udp_capability_req(queue_t *q, mblk_t *mp);
161 static int	udp_close(queue_t *q);
162 static void	udp_connect(queue_t *q, mblk_t *mp);
163 static void	udp_disconnect(queue_t *q, mblk_t *mp);
164 static void	udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
165 		    int sys_error);
166 static void	udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive,
167 		    t_scalar_t tlierr, int unixerr);
168 static int	udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp,
169 		    cred_t *cr);
170 static int	udp_extra_priv_ports_add(queue_t *q, mblk_t *mp,
171 		    char *value, caddr_t cp, cred_t *cr);
172 static int	udp_extra_priv_ports_del(queue_t *q, mblk_t *mp,
173 		    char *value, caddr_t cp, cred_t *cr);
174 static void	udp_icmp_error(queue_t *q, mblk_t *mp);
175 static void	udp_icmp_error_ipv6(queue_t *q, mblk_t *mp);
176 static void	udp_info_req(queue_t *q, mblk_t *mp);
177 static void	udp_input(void *, mblk_t *, void *);
178 static mblk_t	*udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim,
179 		    t_scalar_t addr_length);
180 static void	udp_lrput(queue_t *, mblk_t *);
181 static void	udp_lwput(queue_t *, mblk_t *);
182 static int	udp_open(queue_t *q, dev_t *devp, int flag, int sflag,
183 		    cred_t *credp, boolean_t isv6);
184 static int	udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
185 		    cred_t *credp);
186 static int	udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
187 		    cred_t *credp);
188 static  int	udp_unitdata_opt_process(queue_t *q, mblk_t *mp,
189 		    int *errorp, udpattrs_t *udpattrs);
190 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
191 static int	udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
192 static boolean_t udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt);
193 static int	udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
194 		    cred_t *cr);
195 static void	udp_report_item(mblk_t *mp, udp_t *udp);
196 static int	udp_rinfop(queue_t *q, infod_t *dp);
197 static int	udp_rrw(queue_t *q, struiod_t *dp);
198 static int	udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
199 		    cred_t *cr);
200 static void	udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp,
201 		    ipha_t *ipha);
202 static void	udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr,
203 		    t_scalar_t destlen, t_scalar_t err);
204 static void	udp_unbind(queue_t *q, mblk_t *mp);
205 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port,
206     boolean_t random);
207 static mblk_t	*udp_output_v4(conn_t *, mblk_t *, ipaddr_t, uint16_t, uint_t,
208     int *, boolean_t);
209 static mblk_t	*udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6,
210 		    int *error);
211 static void	udp_wput_other(queue_t *q, mblk_t *mp);
212 static void	udp_wput_iocdata(queue_t *q, mblk_t *mp);
213 static size_t	udp_set_rcv_hiwat(udp_t *udp, size_t size);
214 
215 static void	*udp_stack_init(netstackid_t stackid, netstack_t *ns);
216 static void	udp_stack_fini(netstackid_t stackid, void *arg);
217 
218 static void	*udp_kstat_init(netstackid_t stackid);
219 static void	udp_kstat_fini(netstackid_t stackid, kstat_t *ksp);
220 static void	*udp_kstat2_init(netstackid_t, udp_stat_t *);
221 static void	udp_kstat2_fini(netstackid_t, kstat_t *);
222 static int	udp_kstat_update(kstat_t *kp, int rw);
223 
224 static void	udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp,
225 		    uint_t pkt_len);
226 static void	udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing);
227 static void	udp_xmit(queue_t *, mblk_t *, ire_t *ire, conn_t *, zoneid_t);
228 
229 #define	UDP_RECV_HIWATER	(56 * 1024)
230 #define	UDP_RECV_LOWATER	128
231 #define	UDP_XMIT_HIWATER	(56 * 1024)
232 #define	UDP_XMIT_LOWATER	1024
233 
234 static struct module_info udp_mod_info =  {
235 	UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER
236 };
237 
238 /*
239  * Entry points for UDP as a device.
240  * We have separate open functions for the /dev/udp and /dev/udp6 devices.
241  */
242 static struct qinit udp_rinitv4 = {
243 	NULL, NULL, udp_openv4, udp_close, NULL,
244 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
245 };
246 
247 static struct qinit udp_rinitv6 = {
248 	NULL, NULL, udp_openv6, udp_close, NULL,
249 	&udp_mod_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
250 };
251 
252 static struct qinit udp_winit = {
253 	(pfi_t)udp_wput, (pfi_t)ip_wsrv, NULL, NULL, NULL,
254 	&udp_mod_info, NULL, NULL, NULL, STRUIOT_NONE
255 };
256 
257 /*
258  * UDP needs to handle I_LINK and I_PLINK since ifconfig
259  * likes to use it as a place to hang the various streams.
260  */
261 static struct qinit udp_lrinit = {
262 	(pfi_t)udp_lrput, NULL, udp_openv4, udp_close, NULL,
263 	&udp_mod_info
264 };
265 
266 static struct qinit udp_lwinit = {
267 	(pfi_t)udp_lwput, NULL, udp_openv4, udp_close, NULL,
268 	&udp_mod_info
269 };
270 
271 /* For AF_INET aka /dev/udp */
272 struct streamtab udpinfov4 = {
273 	&udp_rinitv4, &udp_winit, &udp_lrinit, &udp_lwinit
274 };
275 
276 /* For AF_INET6 aka /dev/udp6 */
277 struct streamtab udpinfov6 = {
278 	&udp_rinitv6, &udp_winit, &udp_lrinit, &udp_lwinit
279 };
280 
281 static	sin_t	sin_null;	/* Zero address for quick clears */
282 static	sin6_t	sin6_null;	/* Zero address for quick clears */
283 
284 #define	UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH)
285 
286 /* Default structure copied into T_INFO_ACK messages */
287 static struct T_info_ack udp_g_t_info_ack_ipv4 = {
288 	T_INFO_ACK,
289 	UDP_MAXPACKET_IPV4,	/* TSDU_size. Excl. headers */
290 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
291 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
292 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
293 	sizeof (sin_t),	/* ADDR_size. */
294 	0,		/* OPT_size - not initialized here */
295 	UDP_MAXPACKET_IPV4,	/* TIDU_size.  Excl. headers */
296 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
297 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
298 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
299 };
300 
301 #define	UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN)
302 
303 static	struct T_info_ack udp_g_t_info_ack_ipv6 = {
304 	T_INFO_ACK,
305 	UDP_MAXPACKET_IPV6,	/* TSDU_size.  Excl. headers */
306 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
307 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
308 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
309 	sizeof (sin6_t), /* ADDR_size. */
310 	0,		/* OPT_size - not initialized here */
311 	UDP_MAXPACKET_IPV6,	/* TIDU_size. Excl. headers */
312 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
313 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
314 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
315 };
316 
317 /* largest UDP port number */
318 #define	UDP_MAX_PORT	65535
319 
320 /*
321  * Table of ND variables supported by udp.  These are loaded into us_nd
322  * in udp_open.
323  * All of these are alterable, within the min/max values given, at run time.
324  */
325 /* BEGIN CSTYLED */
326 udpparam_t udp_param_arr[] = {
327  /*min		max		value		name */
328  { 0L,		256,		32,		"udp_wroff_extra" },
329  { 1L,		255,		255,		"udp_ipv4_ttl" },
330  { 0,		IPV6_MAX_HOPS,	IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"},
331  { 1024,	(32 * 1024),	1024,		"udp_smallest_nonpriv_port" },
332  { 0,		1,		1,		"udp_do_checksum" },
333  { 1024,	UDP_MAX_PORT,	(32 * 1024),	"udp_smallest_anon_port" },
334  { 1024,	UDP_MAX_PORT,	UDP_MAX_PORT,	"udp_largest_anon_port" },
335  { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER,	"udp_xmit_hiwat"},
336  { 0,		     (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"},
337  { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER,	"udp_recv_hiwat"},
338  { 65536,	(1<<30),	2*1024*1024,	"udp_max_buf"},
339  { 100,		60000,		1000,		"udp_ndd_get_info_interval"},
340 };
341 /* END CSTYLED */
342 
343 /* Setable in /etc/system */
344 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */
345 uint32_t udp_random_anon_port = 1;
346 
347 /*
348  * Hook functions to enable cluster networking.
349  * On non-clustered systems these vectors must always be NULL
350  */
351 
352 void (*cl_inet_bind)(uchar_t protocol, sa_family_t addr_family,
353     uint8_t *laddrp, in_port_t lport) = NULL;
354 void (*cl_inet_unbind)(uint8_t protocol, sa_family_t addr_family,
355     uint8_t *laddrp, in_port_t lport) = NULL;
356 
357 typedef union T_primitives *t_primp_t;
358 
359 /*
360  * Return the next anonymous port in the privileged port range for
361  * bind checking.
362  *
363  * Trusted Extension (TX) notes: TX allows administrator to mark or
364  * reserve ports as Multilevel ports (MLP). MLP has special function
365  * on TX systems. Once a port is made MLP, it's not available as
366  * ordinary port. This creates "holes" in the port name space. It
367  * may be necessary to skip the "holes" find a suitable anon port.
368  */
369 static in_port_t
370 udp_get_next_priv_port(udp_t *udp)
371 {
372 	static in_port_t next_priv_port = IPPORT_RESERVED - 1;
373 	in_port_t nextport;
374 	boolean_t restart = B_FALSE;
375 	udp_stack_t *us = udp->udp_us;
376 
377 retry:
378 	if (next_priv_port < us->us_min_anonpriv_port ||
379 	    next_priv_port >= IPPORT_RESERVED) {
380 		next_priv_port = IPPORT_RESERVED - 1;
381 		if (restart)
382 			return (0);
383 		restart = B_TRUE;
384 	}
385 
386 	if (is_system_labeled() &&
387 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
388 	    next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) {
389 		next_priv_port = nextport;
390 		goto retry;
391 	}
392 
393 	return (next_priv_port--);
394 }
395 
396 /* UDP bind hash report triggered via the Named Dispatch mechanism. */
397 /* ARGSUSED */
398 static int
399 udp_bind_hash_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
400 {
401 	udp_fanout_t	*udpf;
402 	int		i;
403 	zoneid_t	zoneid;
404 	conn_t		*connp;
405 	udp_t		*udp;
406 	udp_stack_t	*us;
407 
408 	connp = Q_TO_CONN(q);
409 	udp = connp->conn_udp;
410 	us = udp->udp_us;
411 
412 	/* Refer to comments in udp_status_report(). */
413 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
414 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
415 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
416 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
417 			return (0);
418 		}
419 	}
420 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
421 		/* The following may work even if we cannot get a large buf. */
422 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
423 		return (0);
424 	}
425 
426 	(void) mi_mpprintf(mp,
427 	    "UDP     " MI_COL_HDRPAD_STR
428 	/*   12345678[89ABCDEF] */
429 	    " zone lport src addr        dest addr       port  state");
430 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
431 
432 	zoneid = connp->conn_zoneid;
433 
434 	for (i = 0; i < us->us_bind_fanout_size; i++) {
435 		udpf = &us->us_bind_fanout[i];
436 		mutex_enter(&udpf->uf_lock);
437 
438 		/* Print the hash index. */
439 		udp = udpf->uf_udp;
440 		if (zoneid != GLOBAL_ZONEID) {
441 			/* skip to first entry in this zone; might be none */
442 			while (udp != NULL &&
443 			    udp->udp_connp->conn_zoneid != zoneid)
444 				udp = udp->udp_bind_hash;
445 		}
446 		if (udp != NULL) {
447 			uint_t print_len, buf_len;
448 
449 			buf_len = mp->b_cont->b_datap->db_lim -
450 			    mp->b_cont->b_wptr;
451 			print_len = snprintf((char *)mp->b_cont->b_wptr,
452 			    buf_len, "%d\n", i);
453 			if (print_len < buf_len) {
454 				mp->b_cont->b_wptr += print_len;
455 			} else {
456 				mp->b_cont->b_wptr += buf_len;
457 			}
458 			for (; udp != NULL; udp = udp->udp_bind_hash) {
459 				if (zoneid == GLOBAL_ZONEID ||
460 				    zoneid == udp->udp_connp->conn_zoneid)
461 					udp_report_item(mp->b_cont, udp);
462 			}
463 		}
464 		mutex_exit(&udpf->uf_lock);
465 	}
466 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
467 	return (0);
468 }
469 
470 /*
471  * Hash list removal routine for udp_t structures.
472  */
473 static void
474 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock)
475 {
476 	udp_t	*udpnext;
477 	kmutex_t *lockp;
478 	udp_stack_t *us = udp->udp_us;
479 
480 	if (udp->udp_ptpbhn == NULL)
481 		return;
482 
483 	/*
484 	 * Extract the lock pointer in case there are concurrent
485 	 * hash_remove's for this instance.
486 	 */
487 	ASSERT(udp->udp_port != 0);
488 	if (!caller_holds_lock) {
489 		lockp = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
490 		    us->us_bind_fanout_size)].uf_lock;
491 		ASSERT(lockp != NULL);
492 		mutex_enter(lockp);
493 	}
494 	if (udp->udp_ptpbhn != NULL) {
495 		udpnext = udp->udp_bind_hash;
496 		if (udpnext != NULL) {
497 			udpnext->udp_ptpbhn = udp->udp_ptpbhn;
498 			udp->udp_bind_hash = NULL;
499 		}
500 		*udp->udp_ptpbhn = udpnext;
501 		udp->udp_ptpbhn = NULL;
502 	}
503 	if (!caller_holds_lock) {
504 		mutex_exit(lockp);
505 	}
506 }
507 
508 static void
509 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp)
510 {
511 	udp_t	**udpp;
512 	udp_t	*udpnext;
513 
514 	ASSERT(MUTEX_HELD(&uf->uf_lock));
515 	ASSERT(udp->udp_ptpbhn == NULL);
516 	udpp = &uf->uf_udp;
517 	udpnext = udpp[0];
518 	if (udpnext != NULL) {
519 		/*
520 		 * If the new udp bound to the INADDR_ANY address
521 		 * and the first one in the list is not bound to
522 		 * INADDR_ANY we skip all entries until we find the
523 		 * first one bound to INADDR_ANY.
524 		 * This makes sure that applications binding to a
525 		 * specific address get preference over those binding to
526 		 * INADDR_ANY.
527 		 */
528 		if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) &&
529 		    !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) {
530 			while ((udpnext = udpp[0]) != NULL &&
531 			    !V6_OR_V4_INADDR_ANY(
532 			    udpnext->udp_bound_v6src)) {
533 				udpp = &(udpnext->udp_bind_hash);
534 			}
535 			if (udpnext != NULL)
536 				udpnext->udp_ptpbhn = &udp->udp_bind_hash;
537 		} else {
538 			udpnext->udp_ptpbhn = &udp->udp_bind_hash;
539 		}
540 	}
541 	udp->udp_bind_hash = udpnext;
542 	udp->udp_ptpbhn = udpp;
543 	udpp[0] = udp;
544 }
545 
546 /*
547  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
548  * passed to udp_wput.
549  * It associates a port number and local address with the stream.
550  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP
551  * protocol type (IPPROTO_UDP) placed in the message following the address.
552  * A T_BIND_ACK message is passed upstream when ip acknowledges the request.
553  * (Called as writer.)
554  *
555  * Note that UDP over IPv4 and IPv6 sockets can use the same port number
556  * without setting SO_REUSEADDR. This is needed so that they
557  * can be viewed as two independent transport protocols.
558  * However, anonymouns ports are allocated from the same range to avoid
559  * duplicating the us->us_next_port_to_try.
560  */
561 static void
562 udp_bind(queue_t *q, mblk_t *mp)
563 {
564 	sin_t		*sin;
565 	sin6_t		*sin6;
566 	mblk_t		*mp1;
567 	in_port_t	port;		/* Host byte order */
568 	in_port_t	requested_port;	/* Host byte order */
569 	struct T_bind_req *tbr;
570 	int		count;
571 	in6_addr_t	v6src;
572 	boolean_t	bind_to_req_port_only;
573 	int		loopmax;
574 	udp_fanout_t	*udpf;
575 	in_port_t	lport;		/* Network byte order */
576 	zoneid_t	zoneid;
577 	conn_t		*connp;
578 	udp_t		*udp;
579 	boolean_t	is_inaddr_any;
580 	mlp_type_t	addrtype, mlptype;
581 	udp_stack_t	*us;
582 
583 	connp = Q_TO_CONN(q);
584 	udp = connp->conn_udp;
585 	us = udp->udp_us;
586 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
587 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
588 		    "udp_bind: bad req, len %u",
589 		    (uint_t)(mp->b_wptr - mp->b_rptr));
590 		udp_err_ack(q, mp, TPROTO, 0);
591 		return;
592 	}
593 	if (udp->udp_state != TS_UNBND) {
594 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
595 		    "udp_bind: bad state, %u", udp->udp_state);
596 		udp_err_ack(q, mp, TOUTSTATE, 0);
597 		return;
598 	}
599 	/*
600 	 * Reallocate the message to make sure we have enough room for an
601 	 * address and the protocol type.
602 	 */
603 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
604 	if (!mp1) {
605 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
606 		return;
607 	}
608 
609 	mp = mp1;
610 	tbr = (struct T_bind_req *)mp->b_rptr;
611 	switch (tbr->ADDR_length) {
612 	case 0:			/* Request for a generic port */
613 		tbr->ADDR_offset = sizeof (struct T_bind_req);
614 		if (udp->udp_family == AF_INET) {
615 			tbr->ADDR_length = sizeof (sin_t);
616 			sin = (sin_t *)&tbr[1];
617 			*sin = sin_null;
618 			sin->sin_family = AF_INET;
619 			mp->b_wptr = (uchar_t *)&sin[1];
620 		} else {
621 			ASSERT(udp->udp_family == AF_INET6);
622 			tbr->ADDR_length = sizeof (sin6_t);
623 			sin6 = (sin6_t *)&tbr[1];
624 			*sin6 = sin6_null;
625 			sin6->sin6_family = AF_INET6;
626 			mp->b_wptr = (uchar_t *)&sin6[1];
627 		}
628 		port = 0;
629 		break;
630 
631 	case sizeof (sin_t):	/* Complete IPv4 address */
632 		sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset,
633 		    sizeof (sin_t));
634 		if (sin == NULL || !OK_32PTR((char *)sin)) {
635 			udp_err_ack(q, mp, TSYSERR, EINVAL);
636 			return;
637 		}
638 		if (udp->udp_family != AF_INET ||
639 		    sin->sin_family != AF_INET) {
640 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
641 			return;
642 		}
643 		port = ntohs(sin->sin_port);
644 		break;
645 
646 	case sizeof (sin6_t):	/* complete IPv6 address */
647 		sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset,
648 		    sizeof (sin6_t));
649 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
650 			udp_err_ack(q, mp, TSYSERR, EINVAL);
651 			return;
652 		}
653 		if (udp->udp_family != AF_INET6 ||
654 		    sin6->sin6_family != AF_INET6) {
655 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
656 			return;
657 		}
658 		port = ntohs(sin6->sin6_port);
659 		break;
660 
661 	default:		/* Invalid request */
662 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
663 		    "udp_bind: bad ADDR_length length %u", tbr->ADDR_length);
664 		udp_err_ack(q, mp, TBADADDR, 0);
665 		return;
666 	}
667 
668 	requested_port = port;
669 
670 	if (requested_port == 0 || tbr->PRIM_type == O_T_BIND_REQ)
671 		bind_to_req_port_only = B_FALSE;
672 	else			/* T_BIND_REQ and requested_port != 0 */
673 		bind_to_req_port_only = B_TRUE;
674 
675 	if (requested_port == 0) {
676 		/*
677 		 * If the application passed in zero for the port number, it
678 		 * doesn't care which port number we bind to. Get one in the
679 		 * valid range.
680 		 */
681 		if (udp->udp_anon_priv_bind) {
682 			port = udp_get_next_priv_port(udp);
683 		} else {
684 			port = udp_update_next_port(udp,
685 			    us->us_next_port_to_try, B_TRUE);
686 		}
687 	} else {
688 		/*
689 		 * If the port is in the well-known privileged range,
690 		 * make sure the caller was privileged.
691 		 */
692 		int i;
693 		boolean_t priv = B_FALSE;
694 
695 		if (port < us->us_smallest_nonpriv_port) {
696 			priv = B_TRUE;
697 		} else {
698 			for (i = 0; i < us->us_num_epriv_ports; i++) {
699 				if (port == us->us_epriv_ports[i]) {
700 					priv = B_TRUE;
701 					break;
702 				}
703 			}
704 		}
705 
706 		if (priv) {
707 			cred_t *cr = DB_CREDDEF(mp, connp->conn_cred);
708 
709 			if (secpolicy_net_privaddr(cr, port,
710 			    IPPROTO_UDP) != 0) {
711 				udp_err_ack(q, mp, TACCES, 0);
712 				return;
713 			}
714 		}
715 	}
716 
717 	if (port == 0) {
718 		udp_err_ack(q, mp, TNOADDR, 0);
719 		return;
720 	}
721 
722 	/*
723 	 * The state must be TS_UNBND. TPI mandates that users must send
724 	 * TPI primitives only 1 at a time and wait for the response before
725 	 * sending the next primitive.
726 	 */
727 	rw_enter(&udp->udp_rwlock, RW_WRITER);
728 	if (udp->udp_state != TS_UNBND || udp->udp_pending_op != -1) {
729 		rw_exit(&udp->udp_rwlock);
730 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
731 		    "udp_bind: bad state, %u", udp->udp_state);
732 		udp_err_ack(q, mp, TOUTSTATE, 0);
733 		return;
734 	}
735 	udp->udp_pending_op = tbr->PRIM_type;
736 	/*
737 	 * Copy the source address into our udp structure. This address
738 	 * may still be zero; if so, IP will fill in the correct address
739 	 * each time an outbound packet is passed to it. Since the udp is
740 	 * not yet in the bind hash list, we don't grab the uf_lock to
741 	 * change udp_ipversion
742 	 */
743 	if (udp->udp_family == AF_INET) {
744 		ASSERT(sin != NULL);
745 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
746 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
747 		    udp->udp_ip_snd_options_len;
748 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src);
749 	} else {
750 		ASSERT(sin6 != NULL);
751 		v6src = sin6->sin6_addr;
752 		if (IN6_IS_ADDR_V4MAPPED(&v6src)) {
753 			/*
754 			 * no need to hold the uf_lock to set the udp_ipversion
755 			 * since we are not yet in the fanout list
756 			 */
757 			udp->udp_ipversion = IPV4_VERSION;
758 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
759 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
760 		} else {
761 			udp->udp_ipversion = IPV6_VERSION;
762 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
763 		}
764 	}
765 
766 	/*
767 	 * If udp_reuseaddr is not set, then we have to make sure that
768 	 * the IP address and port number the application requested
769 	 * (or we selected for the application) is not being used by
770 	 * another stream.  If another stream is already using the
771 	 * requested IP address and port, the behavior depends on
772 	 * "bind_to_req_port_only". If set the bind fails; otherwise we
773 	 * search for any an unused port to bind to the the stream.
774 	 *
775 	 * As per the BSD semantics, as modified by the Deering multicast
776 	 * changes, if udp_reuseaddr is set, then we allow multiple binds
777 	 * to the same port independent of the local IP address.
778 	 *
779 	 * This is slightly different than in SunOS 4.X which did not
780 	 * support IP multicast. Note that the change implemented by the
781 	 * Deering multicast code effects all binds - not only binding
782 	 * to IP multicast addresses.
783 	 *
784 	 * Note that when binding to port zero we ignore SO_REUSEADDR in
785 	 * order to guarantee a unique port.
786 	 */
787 
788 	count = 0;
789 	if (udp->udp_anon_priv_bind) {
790 		/*
791 		 * loopmax = (IPPORT_RESERVED-1) -
792 		 *    us->us_min_anonpriv_port + 1
793 		 */
794 		loopmax = IPPORT_RESERVED - us->us_min_anonpriv_port;
795 	} else {
796 		loopmax = us->us_largest_anon_port -
797 		    us->us_smallest_anon_port + 1;
798 	}
799 
800 	is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src);
801 	zoneid = connp->conn_zoneid;
802 
803 	for (;;) {
804 		udp_t		*udp1;
805 		boolean_t	found_exclbind = B_FALSE;
806 
807 		/*
808 		 * Walk through the list of udp streams bound to
809 		 * requested port with the same IP address.
810 		 */
811 		lport = htons(port);
812 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(lport,
813 		    us->us_bind_fanout_size)];
814 		mutex_enter(&udpf->uf_lock);
815 		for (udp1 = udpf->uf_udp; udp1 != NULL;
816 		    udp1 = udp1->udp_bind_hash) {
817 			if (lport != udp1->udp_port)
818 				continue;
819 
820 			/*
821 			 * On a labeled system, we must treat bindings to ports
822 			 * on shared IP addresses by sockets with MAC exemption
823 			 * privilege as being in all zones, as there's
824 			 * otherwise no way to identify the right receiver.
825 			 */
826 			if (!(IPCL_ZONE_MATCH(udp1->udp_connp, zoneid) ||
827 			    IPCL_ZONE_MATCH(connp,
828 			    udp1->udp_connp->conn_zoneid)) &&
829 			    !connp->conn_mac_exempt && \
830 			    !udp1->udp_connp->conn_mac_exempt)
831 				continue;
832 
833 			/*
834 			 * If UDP_EXCLBIND is set for either the bound or
835 			 * binding endpoint, the semantics of bind
836 			 * is changed according to the following chart.
837 			 *
838 			 * spec = specified address (v4 or v6)
839 			 * unspec = unspecified address (v4 or v6)
840 			 * A = specified addresses are different for endpoints
841 			 *
842 			 * bound	bind to		allowed?
843 			 * -------------------------------------
844 			 * unspec	unspec		no
845 			 * unspec	spec		no
846 			 * spec		unspec		no
847 			 * spec		spec		yes if A
848 			 *
849 			 * For labeled systems, SO_MAC_EXEMPT behaves the same
850 			 * as UDP_EXCLBIND, except that zoneid is ignored.
851 			 */
852 			if (udp1->udp_exclbind || udp->udp_exclbind ||
853 			    udp1->udp_connp->conn_mac_exempt ||
854 			    connp->conn_mac_exempt) {
855 				if (V6_OR_V4_INADDR_ANY(
856 				    udp1->udp_bound_v6src) ||
857 				    is_inaddr_any ||
858 				    IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
859 				    &v6src)) {
860 					found_exclbind = B_TRUE;
861 					break;
862 				}
863 				continue;
864 			}
865 
866 			/*
867 			 * Check ipversion to allow IPv4 and IPv6 sockets to
868 			 * have disjoint port number spaces.
869 			 */
870 			if (udp->udp_ipversion != udp1->udp_ipversion) {
871 
872 				/*
873 				 * On the first time through the loop, if the
874 				 * the user intentionally specified a
875 				 * particular port number, then ignore any
876 				 * bindings of the other protocol that may
877 				 * conflict. This allows the user to bind IPv6
878 				 * alone and get both v4 and v6, or bind both
879 				 * both and get each seperately. On subsequent
880 				 * times through the loop, we're checking a
881 				 * port that we chose (not the user) and thus
882 				 * we do not allow casual duplicate bindings.
883 				 */
884 				if (count == 0 && requested_port != 0)
885 					continue;
886 			}
887 
888 			/*
889 			 * No difference depending on SO_REUSEADDR.
890 			 *
891 			 * If existing port is bound to a
892 			 * non-wildcard IP address and
893 			 * the requesting stream is bound to
894 			 * a distinct different IP addresses
895 			 * (non-wildcard, also), keep going.
896 			 */
897 			if (!is_inaddr_any &&
898 			    !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) &&
899 			    !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
900 			    &v6src)) {
901 				continue;
902 			}
903 			break;
904 		}
905 
906 		if (!found_exclbind &&
907 		    (udp->udp_reuseaddr && requested_port != 0)) {
908 			break;
909 		}
910 
911 		if (udp1 == NULL) {
912 			/*
913 			 * No other stream has this IP address
914 			 * and port number. We can use it.
915 			 */
916 			break;
917 		}
918 		mutex_exit(&udpf->uf_lock);
919 		if (bind_to_req_port_only) {
920 			/*
921 			 * We get here only when requested port
922 			 * is bound (and only first  of the for()
923 			 * loop iteration).
924 			 *
925 			 * The semantics of this bind request
926 			 * require it to fail so we return from
927 			 * the routine (and exit the loop).
928 			 *
929 			 */
930 			udp->udp_pending_op = -1;
931 			rw_exit(&udp->udp_rwlock);
932 			udp_err_ack(q, mp, TADDRBUSY, 0);
933 			return;
934 		}
935 
936 		if (udp->udp_anon_priv_bind) {
937 			port = udp_get_next_priv_port(udp);
938 		} else {
939 			if ((count == 0) && (requested_port != 0)) {
940 				/*
941 				 * If the application wants us to find
942 				 * a port, get one to start with. Set
943 				 * requested_port to 0, so that we will
944 				 * update us->us_next_port_to_try below.
945 				 */
946 				port = udp_update_next_port(udp,
947 				    us->us_next_port_to_try, B_TRUE);
948 				requested_port = 0;
949 			} else {
950 				port = udp_update_next_port(udp, port + 1,
951 				    B_FALSE);
952 			}
953 		}
954 
955 		if (port == 0 || ++count >= loopmax) {
956 			/*
957 			 * We've tried every possible port number and
958 			 * there are none available, so send an error
959 			 * to the user.
960 			 */
961 			udp->udp_pending_op = -1;
962 			rw_exit(&udp->udp_rwlock);
963 			udp_err_ack(q, mp, TNOADDR, 0);
964 			return;
965 		}
966 	}
967 
968 	/*
969 	 * Copy the source address into our udp structure.  This address
970 	 * may still be zero; if so, ip will fill in the correct address
971 	 * each time an outbound packet is passed to it.
972 	 * If we are binding to a broadcast or multicast address then
973 	 * udp_bind_ack will clear the source address when it receives
974 	 * the T_BIND_ACK.
975 	 */
976 	udp->udp_v6src = udp->udp_bound_v6src = v6src;
977 	udp->udp_port = lport;
978 	/*
979 	 * Now reset the the next anonymous port if the application requested
980 	 * an anonymous port, or we handed out the next anonymous port.
981 	 */
982 	if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) {
983 		us->us_next_port_to_try = port + 1;
984 	}
985 
986 	/* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */
987 	if (udp->udp_family == AF_INET) {
988 		sin->sin_port = udp->udp_port;
989 	} else {
990 		int error;
991 
992 		sin6->sin6_port = udp->udp_port;
993 		/* Rebuild the header template */
994 		error = udp_build_hdrs(udp);
995 		if (error != 0) {
996 			udp->udp_pending_op = -1;
997 			rw_exit(&udp->udp_rwlock);
998 			mutex_exit(&udpf->uf_lock);
999 			udp_err_ack(q, mp, TSYSERR, error);
1000 			return;
1001 		}
1002 	}
1003 	udp->udp_state = TS_IDLE;
1004 	udp_bind_hash_insert(udpf, udp);
1005 	mutex_exit(&udpf->uf_lock);
1006 	rw_exit(&udp->udp_rwlock);
1007 
1008 	if (cl_inet_bind) {
1009 		/*
1010 		 * Running in cluster mode - register bind information
1011 		 */
1012 		if (udp->udp_ipversion == IPV4_VERSION) {
1013 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET,
1014 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
1015 			    (in_port_t)udp->udp_port);
1016 		} else {
1017 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET6,
1018 			    (uint8_t *)&(udp->udp_v6src),
1019 			    (in_port_t)udp->udp_port);
1020 		}
1021 
1022 	}
1023 
1024 	connp->conn_anon_port = (is_system_labeled() && requested_port == 0);
1025 	if (is_system_labeled() && (!connp->conn_anon_port ||
1026 	    connp->conn_anon_mlp)) {
1027 		uint16_t mlpport;
1028 		cred_t *cr = connp->conn_cred;
1029 		zone_t *zone;
1030 
1031 		zone = crgetzone(cr);
1032 		connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth :
1033 		    mlptSingle;
1034 		addrtype = tsol_mlp_addr_type(zone->zone_id, IPV6_VERSION,
1035 		    &v6src, us->us_netstack->netstack_ip);
1036 		if (addrtype == mlptSingle) {
1037 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1038 			udp->udp_pending_op = -1;
1039 			rw_exit(&udp->udp_rwlock);
1040 			udp_err_ack(q, mp, TNOADDR, 0);
1041 			connp->conn_anon_port = B_FALSE;
1042 			connp->conn_mlp_type = mlptSingle;
1043 			return;
1044 		}
1045 		mlpport = connp->conn_anon_port ? PMAPPORT : port;
1046 		mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport,
1047 		    addrtype);
1048 		if (mlptype != mlptSingle &&
1049 		    (connp->conn_mlp_type == mlptSingle ||
1050 		    secpolicy_net_bindmlp(cr) != 0)) {
1051 			if (udp->udp_debug) {
1052 				(void) strlog(UDP_MOD_ID, 0, 1,
1053 				    SL_ERROR|SL_TRACE,
1054 				    "udp_bind: no priv for multilevel port %d",
1055 				    mlpport);
1056 			}
1057 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1058 			udp->udp_pending_op = -1;
1059 			rw_exit(&udp->udp_rwlock);
1060 			udp_err_ack(q, mp, TACCES, 0);
1061 			connp->conn_anon_port = B_FALSE;
1062 			connp->conn_mlp_type = mlptSingle;
1063 			return;
1064 		}
1065 
1066 		/*
1067 		 * If we're specifically binding a shared IP address and the
1068 		 * port is MLP on shared addresses, then check to see if this
1069 		 * zone actually owns the MLP.  Reject if not.
1070 		 */
1071 		if (mlptype == mlptShared && addrtype == mlptShared) {
1072 			/*
1073 			 * No need to handle exclusive-stack zones since
1074 			 * ALL_ZONES only applies to the shared stack.
1075 			 */
1076 			zoneid_t mlpzone;
1077 
1078 			mlpzone = tsol_mlp_findzone(IPPROTO_UDP,
1079 			    htons(mlpport));
1080 			if (connp->conn_zoneid != mlpzone) {
1081 				if (udp->udp_debug) {
1082 					(void) strlog(UDP_MOD_ID, 0, 1,
1083 					    SL_ERROR|SL_TRACE,
1084 					    "udp_bind: attempt to bind port "
1085 					    "%d on shared addr in zone %d "
1086 					    "(should be %d)",
1087 					    mlpport, connp->conn_zoneid,
1088 					    mlpzone);
1089 				}
1090 				rw_enter(&udp->udp_rwlock, RW_WRITER);
1091 				udp->udp_pending_op = -1;
1092 				rw_exit(&udp->udp_rwlock);
1093 				udp_err_ack(q, mp, TACCES, 0);
1094 				connp->conn_anon_port = B_FALSE;
1095 				connp->conn_mlp_type = mlptSingle;
1096 				return;
1097 			}
1098 		}
1099 		if (connp->conn_anon_port) {
1100 			int error;
1101 
1102 			error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp,
1103 			    port, B_TRUE);
1104 			if (error != 0) {
1105 				if (udp->udp_debug) {
1106 					(void) strlog(UDP_MOD_ID, 0, 1,
1107 					    SL_ERROR|SL_TRACE,
1108 					    "udp_bind: cannot establish anon "
1109 					    "MLP for port %d", port);
1110 				}
1111 				rw_enter(&udp->udp_rwlock, RW_WRITER);
1112 				udp->udp_pending_op = -1;
1113 				rw_exit(&udp->udp_rwlock);
1114 				udp_err_ack(q, mp, TACCES, 0);
1115 				connp->conn_anon_port = B_FALSE;
1116 				connp->conn_mlp_type = mlptSingle;
1117 				return;
1118 			}
1119 		}
1120 		connp->conn_mlp_type = mlptype;
1121 	}
1122 
1123 	/* Pass the protocol number in the message following the address. */
1124 	*mp->b_wptr++ = IPPROTO_UDP;
1125 	if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
1126 		/*
1127 		 * Append a request for an IRE if udp_v6src not
1128 		 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address).
1129 		 */
1130 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1131 		if (!mp->b_cont) {
1132 			rw_enter(&udp->udp_rwlock, RW_WRITER);
1133 			udp->udp_pending_op = -1;
1134 			rw_exit(&udp->udp_rwlock);
1135 			udp_err_ack(q, mp, TSYSERR, ENOMEM);
1136 			return;
1137 		}
1138 		mp->b_cont->b_wptr += sizeof (ire_t);
1139 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1140 	}
1141 	if (udp->udp_family == AF_INET6)
1142 		mp = ip_bind_v6(q, mp, connp, NULL);
1143 	else
1144 		mp = ip_bind_v4(q, mp, connp);
1145 
1146 	/* The above return NULL if the bind needs to be deferred */
1147 	if (mp != NULL)
1148 		udp_bind_result(connp, mp);
1149 	else
1150 		CONN_INC_REF(connp);
1151 }
1152 
1153 /*
1154  * This is called from ip_wput_nondata to handle the results of a
1155  * deferred UDP bind. It is called once the bind has been completed.
1156  */
1157 void
1158 udp_resume_bind(conn_t *connp, mblk_t *mp)
1159 {
1160 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
1161 
1162 	udp_bind_result(connp, mp);
1163 
1164 	CONN_OPER_PENDING_DONE(connp);
1165 }
1166 
1167 /*
1168  * This routine handles each T_CONN_REQ message passed to udp.  It
1169  * associates a default destination address with the stream.
1170  *
1171  * This routine sends down a T_BIND_REQ to IP with the following mblks:
1172  *	T_BIND_REQ	- specifying local and remote address/port
1173  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
1174  *	T_OK_ACK	- for the T_CONN_REQ
1175  *	T_CONN_CON	- to keep the TPI user happy
1176  *
1177  * The connect completes in udp_bind_result.
1178  * When a T_BIND_ACK is received information is extracted from the IRE
1179  * and the two appended messages are sent to the TPI user.
1180  * Should udp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
1181  * convert it to an error ack for the appropriate primitive.
1182  */
1183 static void
1184 udp_connect(queue_t *q, mblk_t *mp)
1185 {
1186 	sin6_t	*sin6;
1187 	sin_t	*sin;
1188 	struct T_conn_req	*tcr;
1189 	in6_addr_t v6dst;
1190 	ipaddr_t v4dst;
1191 	uint16_t dstport;
1192 	uint32_t flowinfo;
1193 	mblk_t	*mp1, *mp2;
1194 	udp_fanout_t	*udpf;
1195 	udp_t	*udp, *udp1;
1196 	ushort_t	ipversion;
1197 	udp_stack_t	*us;
1198 	conn_t		*connp = Q_TO_CONN(q);
1199 
1200 	udp = connp->conn_udp;
1201 	tcr = (struct T_conn_req *)mp->b_rptr;
1202 	us = udp->udp_us;
1203 
1204 	/* A bit of sanity checking */
1205 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
1206 		udp_err_ack(q, mp, TPROTO, 0);
1207 		return;
1208 	}
1209 
1210 	if (tcr->OPT_length != 0) {
1211 		udp_err_ack(q, mp, TBADOPT, 0);
1212 		return;
1213 	}
1214 
1215 	/*
1216 	 * Determine packet type based on type of address passed in
1217 	 * the request should contain an IPv4 or IPv6 address.
1218 	 * Make sure that address family matches the type of
1219 	 * family of the the address passed down
1220 	 */
1221 	switch (tcr->DEST_length) {
1222 	default:
1223 		udp_err_ack(q, mp, TBADADDR, 0);
1224 		return;
1225 
1226 	case sizeof (sin_t):
1227 		sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset,
1228 		    sizeof (sin_t));
1229 		if (sin == NULL || !OK_32PTR((char *)sin)) {
1230 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1231 			return;
1232 		}
1233 		if (udp->udp_family != AF_INET ||
1234 		    sin->sin_family != AF_INET) {
1235 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1236 			return;
1237 		}
1238 		v4dst = sin->sin_addr.s_addr;
1239 		dstport = sin->sin_port;
1240 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
1241 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
1242 		ipversion = IPV4_VERSION;
1243 		break;
1244 
1245 	case sizeof (sin6_t):
1246 		sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset,
1247 		    sizeof (sin6_t));
1248 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
1249 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1250 			return;
1251 		}
1252 		if (udp->udp_family != AF_INET6 ||
1253 		    sin6->sin6_family != AF_INET6) {
1254 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1255 			return;
1256 		}
1257 		v6dst = sin6->sin6_addr;
1258 		dstport = sin6->sin6_port;
1259 		if (IN6_IS_ADDR_V4MAPPED(&v6dst)) {
1260 			IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst);
1261 			ipversion = IPV4_VERSION;
1262 			flowinfo = 0;
1263 		} else {
1264 			ipversion = IPV6_VERSION;
1265 			flowinfo = sin6->sin6_flowinfo;
1266 		}
1267 		break;
1268 	}
1269 	if (dstport == 0) {
1270 		udp_err_ack(q, mp, TBADADDR, 0);
1271 		return;
1272 	}
1273 
1274 	rw_enter(&udp->udp_rwlock, RW_WRITER);
1275 
1276 	/*
1277 	 * This UDP must have bound to a port already before doing a connect.
1278 	 * TPI mandates that users must send TPI primitives only 1 at a time
1279 	 * and wait for the response before sending the next primitive.
1280 	 */
1281 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
1282 		rw_exit(&udp->udp_rwlock);
1283 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1284 		    "udp_connect: bad state, %u", udp->udp_state);
1285 		udp_err_ack(q, mp, TOUTSTATE, 0);
1286 		return;
1287 	}
1288 	udp->udp_pending_op = T_CONN_REQ;
1289 	ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL);
1290 
1291 	if (ipversion == IPV4_VERSION) {
1292 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
1293 		    udp->udp_ip_snd_options_len;
1294 	} else {
1295 		udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
1296 	}
1297 
1298 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
1299 	    us->us_bind_fanout_size)];
1300 
1301 	mutex_enter(&udpf->uf_lock);
1302 	if (udp->udp_state == TS_DATA_XFER) {
1303 		/* Already connected - clear out state */
1304 		udp->udp_v6src = udp->udp_bound_v6src;
1305 		udp->udp_state = TS_IDLE;
1306 	}
1307 
1308 	/*
1309 	 * Create a default IP header with no IP options.
1310 	 */
1311 	udp->udp_dstport = dstport;
1312 	udp->udp_ipversion = ipversion;
1313 	if (ipversion == IPV4_VERSION) {
1314 		/*
1315 		 * Interpret a zero destination to mean loopback.
1316 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
1317 		 * generate the T_CONN_CON.
1318 		 */
1319 		if (v4dst == INADDR_ANY) {
1320 			v4dst = htonl(INADDR_LOOPBACK);
1321 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
1322 			if (udp->udp_family == AF_INET) {
1323 				sin->sin_addr.s_addr = v4dst;
1324 			} else {
1325 				sin6->sin6_addr = v6dst;
1326 			}
1327 		}
1328 		udp->udp_v6dst = v6dst;
1329 		udp->udp_flowinfo = 0;
1330 
1331 		/*
1332 		 * If the destination address is multicast and
1333 		 * an outgoing multicast interface has been set,
1334 		 * use the address of that interface as our
1335 		 * source address if no source address has been set.
1336 		 */
1337 		if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY &&
1338 		    CLASSD(v4dst) &&
1339 		    udp->udp_multicast_if_addr != INADDR_ANY) {
1340 			IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr,
1341 			    &udp->udp_v6src);
1342 		}
1343 	} else {
1344 		ASSERT(udp->udp_ipversion == IPV6_VERSION);
1345 		/*
1346 		 * Interpret a zero destination to mean loopback.
1347 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
1348 		 * generate the T_CONN_CON.
1349 		 */
1350 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
1351 			v6dst = ipv6_loopback;
1352 			sin6->sin6_addr = v6dst;
1353 		}
1354 		udp->udp_v6dst = v6dst;
1355 		udp->udp_flowinfo = flowinfo;
1356 		/*
1357 		 * If the destination address is multicast and
1358 		 * an outgoing multicast interface has been set,
1359 		 * then the ip bind logic will pick the correct source
1360 		 * address (i.e. matching the outgoing multicast interface).
1361 		 */
1362 	}
1363 
1364 	/*
1365 	 * Verify that the src/port/dst/port is unique for all
1366 	 * connections in TS_DATA_XFER
1367 	 */
1368 	for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) {
1369 		if (udp1->udp_state != TS_DATA_XFER)
1370 			continue;
1371 		if (udp->udp_port != udp1->udp_port ||
1372 		    udp->udp_ipversion != udp1->udp_ipversion ||
1373 		    dstport != udp1->udp_dstport ||
1374 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) ||
1375 		    !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst) ||
1376 		    !(IPCL_ZONE_MATCH(udp->udp_connp,
1377 		    udp1->udp_connp->conn_zoneid) ||
1378 		    IPCL_ZONE_MATCH(udp1->udp_connp,
1379 		    udp->udp_connp->conn_zoneid)))
1380 			continue;
1381 		mutex_exit(&udpf->uf_lock);
1382 		udp->udp_pending_op = -1;
1383 		rw_exit(&udp->udp_rwlock);
1384 		udp_err_ack(q, mp, TBADADDR, 0);
1385 		return;
1386 	}
1387 	udp->udp_state = TS_DATA_XFER;
1388 	mutex_exit(&udpf->uf_lock);
1389 
1390 	/*
1391 	 * Send down bind to IP to verify that there is a route
1392 	 * and to determine the source address.
1393 	 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput.
1394 	 */
1395 	if (udp->udp_family == AF_INET)
1396 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa_conn_t));
1397 	else
1398 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa6_conn_t));
1399 	if (mp1 == NULL) {
1400 bind_failed:
1401 		mutex_enter(&udpf->uf_lock);
1402 		udp->udp_state = TS_IDLE;
1403 		udp->udp_pending_op = -1;
1404 		mutex_exit(&udpf->uf_lock);
1405 		rw_exit(&udp->udp_rwlock);
1406 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1407 		return;
1408 	}
1409 
1410 	rw_exit(&udp->udp_rwlock);
1411 	/*
1412 	 * We also have to send a connection confirmation to
1413 	 * keep TLI happy. Prepare it for udp_bind_result.
1414 	 */
1415 	if (udp->udp_family == AF_INET)
1416 		mp2 = mi_tpi_conn_con(NULL, (char *)sin,
1417 		    sizeof (*sin), NULL, 0);
1418 	else
1419 		mp2 = mi_tpi_conn_con(NULL, (char *)sin6,
1420 		    sizeof (*sin6), NULL, 0);
1421 	if (mp2 == NULL) {
1422 		freemsg(mp1);
1423 		rw_enter(&udp->udp_rwlock, RW_WRITER);
1424 		goto bind_failed;
1425 	}
1426 
1427 	mp = mi_tpi_ok_ack_alloc(mp);
1428 	if (mp == NULL) {
1429 		/* Unable to reuse the T_CONN_REQ for the ack. */
1430 		freemsg(mp2);
1431 		rw_enter(&udp->udp_rwlock, RW_WRITER);
1432 		mutex_enter(&udpf->uf_lock);
1433 		udp->udp_state = TS_IDLE;
1434 		udp->udp_pending_op = -1;
1435 		mutex_exit(&udpf->uf_lock);
1436 		rw_exit(&udp->udp_rwlock);
1437 		udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
1438 		return;
1439 	}
1440 
1441 	/* Hang onto the T_OK_ACK and T_CONN_CON for later. */
1442 	linkb(mp1, mp);
1443 	linkb(mp1, mp2);
1444 
1445 	mblk_setcred(mp1, connp->conn_cred);
1446 	if (udp->udp_family == AF_INET)
1447 		mp1 = ip_bind_v4(q, mp1, connp);
1448 	else
1449 		mp1 = ip_bind_v6(q, mp1, connp, NULL);
1450 
1451 	/* The above return NULL if the bind needs to be deferred */
1452 	if (mp1 != NULL)
1453 		udp_bind_result(connp, mp1);
1454 	else
1455 		CONN_INC_REF(connp);
1456 }
1457 
1458 static int
1459 udp_close(queue_t *q)
1460 {
1461 	conn_t	*connp = (conn_t *)q->q_ptr;
1462 	udp_t	*udp;
1463 
1464 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
1465 	udp = connp->conn_udp;
1466 
1467 	udp_quiesce_conn(connp);
1468 	ip_quiesce_conn(connp);
1469 	/*
1470 	 * Disable read-side synchronous stream
1471 	 * interface and drain any queued data.
1472 	 */
1473 	udp_rcv_drain(q, udp, B_TRUE);
1474 	ASSERT(!udp->udp_direct_sockfs);
1475 
1476 	qprocsoff(q);
1477 
1478 	ASSERT(udp->udp_rcv_cnt == 0);
1479 	ASSERT(udp->udp_rcv_msgcnt == 0);
1480 	ASSERT(udp->udp_rcv_list_head == NULL);
1481 	ASSERT(udp->udp_rcv_list_tail == NULL);
1482 
1483 	udp_close_free(connp);
1484 
1485 	/*
1486 	 * Now we are truly single threaded on this stream, and can
1487 	 * delete the things hanging off the connp, and finally the connp.
1488 	 * We removed this connp from the fanout list, it cannot be
1489 	 * accessed thru the fanouts, and we already waited for the
1490 	 * conn_ref to drop to 0. We are already in close, so
1491 	 * there cannot be any other thread from the top. qprocsoff
1492 	 * has completed, and service has completed or won't run in
1493 	 * future.
1494 	 */
1495 	ASSERT(connp->conn_ref == 1);
1496 	inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
1497 	connp->conn_ref--;
1498 	ipcl_conn_destroy(connp);
1499 
1500 	q->q_ptr = WR(q)->q_ptr = NULL;
1501 	return (0);
1502 }
1503 
1504 /*
1505  * Called in the close path to quiesce the conn
1506  */
1507 void
1508 udp_quiesce_conn(conn_t *connp)
1509 {
1510 	udp_t	*udp = connp->conn_udp;
1511 
1512 	if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) {
1513 		/*
1514 		 * Running in cluster mode - register unbind information
1515 		 */
1516 		if (udp->udp_ipversion == IPV4_VERSION) {
1517 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
1518 			    (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))),
1519 			    (in_port_t)udp->udp_port);
1520 		} else {
1521 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
1522 			    (uint8_t *)(&(udp->udp_v6src)),
1523 			    (in_port_t)udp->udp_port);
1524 		}
1525 	}
1526 
1527 	udp_bind_hash_remove(udp, B_FALSE);
1528 
1529 }
1530 
1531 void
1532 udp_close_free(conn_t *connp)
1533 {
1534 	udp_t *udp = connp->conn_udp;
1535 
1536 	/* If there are any options associated with the stream, free them. */
1537 	if (udp->udp_ip_snd_options != NULL) {
1538 		mi_free((char *)udp->udp_ip_snd_options);
1539 		udp->udp_ip_snd_options = NULL;
1540 		udp->udp_ip_snd_options_len = 0;
1541 	}
1542 
1543 	if (udp->udp_ip_rcv_options != NULL) {
1544 		mi_free((char *)udp->udp_ip_rcv_options);
1545 		udp->udp_ip_rcv_options = NULL;
1546 		udp->udp_ip_rcv_options_len = 0;
1547 	}
1548 
1549 	/* Free memory associated with sticky options */
1550 	if (udp->udp_sticky_hdrs_len != 0) {
1551 		kmem_free(udp->udp_sticky_hdrs,
1552 		    udp->udp_sticky_hdrs_len);
1553 		udp->udp_sticky_hdrs = NULL;
1554 		udp->udp_sticky_hdrs_len = 0;
1555 	}
1556 
1557 	ip6_pkt_free(&udp->udp_sticky_ipp);
1558 
1559 	/*
1560 	 * Clear any fields which the kmem_cache constructor clears.
1561 	 * Only udp_connp needs to be preserved.
1562 	 * TBD: We should make this more efficient to avoid clearing
1563 	 * everything.
1564 	 */
1565 	ASSERT(udp->udp_connp == connp);
1566 	bzero(udp, sizeof (udp_t));
1567 	udp->udp_connp = connp;
1568 }
1569 
1570 /*
1571  * This routine handles each T_DISCON_REQ message passed to udp
1572  * as an indicating that UDP is no longer connected. This results
1573  * in sending a T_BIND_REQ to IP to restore the binding to just
1574  * the local address/port.
1575  *
1576  * This routine sends down a T_BIND_REQ to IP with the following mblks:
1577  *	T_BIND_REQ	- specifying just the local address/port
1578  *	T_OK_ACK	- for the T_DISCON_REQ
1579  *
1580  * The disconnect completes in udp_bind_result.
1581  * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user.
1582  * Should udp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
1583  * convert it to an error ack for the appropriate primitive.
1584  */
1585 static void
1586 udp_disconnect(queue_t *q, mblk_t *mp)
1587 {
1588 	udp_t	*udp;
1589 	mblk_t	*mp1;
1590 	udp_fanout_t *udpf;
1591 	udp_stack_t *us;
1592 	conn_t	*connp = Q_TO_CONN(q);
1593 
1594 	udp = connp->conn_udp;
1595 	us = udp->udp_us;
1596 	rw_enter(&udp->udp_rwlock, RW_WRITER);
1597 	if (udp->udp_state != TS_DATA_XFER || udp->udp_pending_op != -1) {
1598 		rw_exit(&udp->udp_rwlock);
1599 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1600 		    "udp_disconnect: bad state, %u", udp->udp_state);
1601 		udp_err_ack(q, mp, TOUTSTATE, 0);
1602 		return;
1603 	}
1604 	udp->udp_pending_op = T_DISCON_REQ;
1605 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
1606 	    us->us_bind_fanout_size)];
1607 	mutex_enter(&udpf->uf_lock);
1608 	udp->udp_v6src = udp->udp_bound_v6src;
1609 	udp->udp_state = TS_IDLE;
1610 	mutex_exit(&udpf->uf_lock);
1611 
1612 	/*
1613 	 * Send down bind to IP to remove the full binding and revert
1614 	 * to the local address binding.
1615 	 */
1616 	if (udp->udp_family == AF_INET)
1617 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin_t));
1618 	else
1619 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin6_t));
1620 	if (mp1 == NULL) {
1621 		udp->udp_pending_op = -1;
1622 		rw_exit(&udp->udp_rwlock);
1623 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1624 		return;
1625 	}
1626 	mp = mi_tpi_ok_ack_alloc(mp);
1627 	if (mp == NULL) {
1628 		/* Unable to reuse the T_DISCON_REQ for the ack. */
1629 		udp->udp_pending_op = -1;
1630 		rw_exit(&udp->udp_rwlock);
1631 		udp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM);
1632 		return;
1633 	}
1634 
1635 	if (udp->udp_family == AF_INET6) {
1636 		int error;
1637 
1638 		/* Rebuild the header template */
1639 		error = udp_build_hdrs(udp);
1640 		if (error != 0) {
1641 			udp->udp_pending_op = -1;
1642 			rw_exit(&udp->udp_rwlock);
1643 			udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error);
1644 			freemsg(mp1);
1645 			return;
1646 		}
1647 	}
1648 
1649 	rw_exit(&udp->udp_rwlock);
1650 	/* Append the T_OK_ACK to the T_BIND_REQ for udp_bind_ack */
1651 	linkb(mp1, mp);
1652 
1653 	if (udp->udp_family == AF_INET6)
1654 		mp1 = ip_bind_v6(q, mp1, connp, NULL);
1655 	else
1656 		mp1 = ip_bind_v4(q, mp1, connp);
1657 
1658 	/* The above return NULL if the bind needs to be deferred */
1659 	if (mp1 != NULL)
1660 		udp_bind_result(connp, mp1);
1661 	else
1662 		CONN_INC_REF(connp);
1663 }
1664 
1665 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
1666 static void
1667 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
1668 {
1669 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1670 		qreply(q, mp);
1671 }
1672 
1673 /* Shorthand to generate and send TPI error acks to our client */
1674 static void
1675 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error,
1676     int sys_error)
1677 {
1678 	struct T_error_ack	*teackp;
1679 
1680 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1681 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1682 		teackp = (struct T_error_ack *)mp->b_rptr;
1683 		teackp->ERROR_prim = primitive;
1684 		teackp->TLI_error = t_error;
1685 		teackp->UNIX_error = sys_error;
1686 		qreply(q, mp);
1687 	}
1688 }
1689 
1690 /*ARGSUSED*/
1691 static int
1692 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
1693 {
1694 	int i;
1695 	udp_t		*udp = Q_TO_UDP(q);
1696 	udp_stack_t *us = udp->udp_us;
1697 
1698 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1699 		if (us->us_epriv_ports[i] != 0)
1700 			(void) mi_mpprintf(mp, "%d ", us->us_epriv_ports[i]);
1701 	}
1702 	return (0);
1703 }
1704 
1705 /* ARGSUSED */
1706 static int
1707 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1708     cred_t *cr)
1709 {
1710 	long	new_value;
1711 	int	i;
1712 	udp_t		*udp = Q_TO_UDP(q);
1713 	udp_stack_t *us = udp->udp_us;
1714 
1715 	/*
1716 	 * Fail the request if the new value does not lie within the
1717 	 * port number limits.
1718 	 */
1719 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1720 	    new_value <= 0 || new_value >= 65536) {
1721 		return (EINVAL);
1722 	}
1723 
1724 	/* Check if the value is already in the list */
1725 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1726 		if (new_value == us->us_epriv_ports[i]) {
1727 			return (EEXIST);
1728 		}
1729 	}
1730 	/* Find an empty slot */
1731 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1732 		if (us->us_epriv_ports[i] == 0)
1733 			break;
1734 	}
1735 	if (i == us->us_num_epriv_ports) {
1736 		return (EOVERFLOW);
1737 	}
1738 
1739 	/* Set the new value */
1740 	us->us_epriv_ports[i] = (in_port_t)new_value;
1741 	return (0);
1742 }
1743 
1744 /* ARGSUSED */
1745 static int
1746 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
1747     cred_t *cr)
1748 {
1749 	long	new_value;
1750 	int	i;
1751 	udp_t		*udp = Q_TO_UDP(q);
1752 	udp_stack_t *us = udp->udp_us;
1753 
1754 	/*
1755 	 * Fail the request if the new value does not lie within the
1756 	 * port number limits.
1757 	 */
1758 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
1759 	    new_value <= 0 || new_value >= 65536) {
1760 		return (EINVAL);
1761 	}
1762 
1763 	/* Check that the value is already in the list */
1764 	for (i = 0; i < us->us_num_epriv_ports; i++) {
1765 		if (us->us_epriv_ports[i] == new_value)
1766 			break;
1767 	}
1768 	if (i == us->us_num_epriv_ports) {
1769 		return (ESRCH);
1770 	}
1771 
1772 	/* Clear the value */
1773 	us->us_epriv_ports[i] = 0;
1774 	return (0);
1775 }
1776 
1777 /* At minimum we need 4 bytes of UDP header */
1778 #define	ICMP_MIN_UDP_HDR	4
1779 
1780 /*
1781  * udp_icmp_error is called by udp_input to process ICMP msgs. passed up by IP.
1782  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1783  * Assumes that IP has pulled up everything up to and including the ICMP header.
1784  */
1785 static void
1786 udp_icmp_error(queue_t *q, mblk_t *mp)
1787 {
1788 	icmph_t *icmph;
1789 	ipha_t	*ipha;
1790 	int	iph_hdr_length;
1791 	udpha_t	*udpha;
1792 	sin_t	sin;
1793 	sin6_t	sin6;
1794 	mblk_t	*mp1;
1795 	int	error = 0;
1796 	udp_t	*udp = Q_TO_UDP(q);
1797 
1798 	ipha = (ipha_t *)mp->b_rptr;
1799 
1800 	ASSERT(OK_32PTR(mp->b_rptr));
1801 
1802 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1803 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1804 		udp_icmp_error_ipv6(q, mp);
1805 		return;
1806 	}
1807 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
1808 
1809 	/* Skip past the outer IP and ICMP headers */
1810 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1811 	icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length];
1812 	ipha = (ipha_t *)&icmph[1];
1813 
1814 	/* Skip past the inner IP and find the ULP header */
1815 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1816 	udpha = (udpha_t *)((char *)ipha + iph_hdr_length);
1817 
1818 	switch (icmph->icmph_type) {
1819 	case ICMP_DEST_UNREACHABLE:
1820 		switch (icmph->icmph_code) {
1821 		case ICMP_FRAGMENTATION_NEEDED:
1822 			/*
1823 			 * IP has already adjusted the path MTU.
1824 			 */
1825 			break;
1826 		case ICMP_PORT_UNREACHABLE:
1827 		case ICMP_PROTOCOL_UNREACHABLE:
1828 			error = ECONNREFUSED;
1829 			break;
1830 		default:
1831 			/* Transient errors */
1832 			break;
1833 		}
1834 		break;
1835 	default:
1836 		/* Transient errors */
1837 		break;
1838 	}
1839 	if (error == 0) {
1840 		freemsg(mp);
1841 		return;
1842 	}
1843 
1844 	/*
1845 	 * Deliver T_UDERROR_IND when the application has asked for it.
1846 	 * The socket layer enables this automatically when connected.
1847 	 */
1848 	if (!udp->udp_dgram_errind) {
1849 		freemsg(mp);
1850 		return;
1851 	}
1852 
1853 	switch (udp->udp_family) {
1854 	case AF_INET:
1855 		sin = sin_null;
1856 		sin.sin_family = AF_INET;
1857 		sin.sin_addr.s_addr = ipha->ipha_dst;
1858 		sin.sin_port = udpha->uha_dst_port;
1859 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0,
1860 		    error);
1861 		break;
1862 	case AF_INET6:
1863 		sin6 = sin6_null;
1864 		sin6.sin6_family = AF_INET6;
1865 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
1866 		sin6.sin6_port = udpha->uha_dst_port;
1867 
1868 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1869 		    NULL, 0, error);
1870 		break;
1871 	}
1872 	if (mp1)
1873 		putnext(q, mp1);
1874 	freemsg(mp);
1875 }
1876 
1877 /*
1878  * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6.
1879  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
1880  * Assumes that IP has pulled up all the extension headers as well as the
1881  * ICMPv6 header.
1882  */
1883 static void
1884 udp_icmp_error_ipv6(queue_t *q, mblk_t *mp)
1885 {
1886 	icmp6_t		*icmp6;
1887 	ip6_t		*ip6h, *outer_ip6h;
1888 	uint16_t	iph_hdr_length;
1889 	uint8_t		*nexthdrp;
1890 	udpha_t		*udpha;
1891 	sin6_t		sin6;
1892 	mblk_t		*mp1;
1893 	int		error = 0;
1894 	udp_t		*udp = Q_TO_UDP(q);
1895 	udp_stack_t	*us = udp->udp_us;
1896 
1897 	outer_ip6h = (ip6_t *)mp->b_rptr;
1898 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1899 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1900 	else
1901 		iph_hdr_length = IPV6_HDR_LEN;
1902 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1903 	ip6h = (ip6_t *)&icmp6[1];
1904 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1905 		freemsg(mp);
1906 		return;
1907 	}
1908 	udpha = (udpha_t *)((char *)ip6h + iph_hdr_length);
1909 
1910 	switch (icmp6->icmp6_type) {
1911 	case ICMP6_DST_UNREACH:
1912 		switch (icmp6->icmp6_code) {
1913 		case ICMP6_DST_UNREACH_NOPORT:
1914 			error = ECONNREFUSED;
1915 			break;
1916 		case ICMP6_DST_UNREACH_ADMIN:
1917 		case ICMP6_DST_UNREACH_NOROUTE:
1918 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1919 		case ICMP6_DST_UNREACH_ADDR:
1920 			/* Transient errors */
1921 			break;
1922 		default:
1923 			break;
1924 		}
1925 		break;
1926 	case ICMP6_PACKET_TOO_BIG: {
1927 		struct T_unitdata_ind	*tudi;
1928 		struct T_opthdr		*toh;
1929 		size_t			udi_size;
1930 		mblk_t			*newmp;
1931 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1932 		    sizeof (struct ip6_mtuinfo);
1933 		sin6_t			*sin6;
1934 		struct ip6_mtuinfo	*mtuinfo;
1935 
1936 		/*
1937 		 * If the application has requested to receive path mtu
1938 		 * information, send up an empty message containing an
1939 		 * IPV6_PATHMTU ancillary data item.
1940 		 */
1941 		if (!udp->udp_ipv6_recvpathmtu)
1942 			break;
1943 
1944 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1945 		    opt_length;
1946 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1947 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
1948 			break;
1949 		}
1950 
1951 		/*
1952 		 * newmp->b_cont is left to NULL on purpose.  This is an
1953 		 * empty message containing only ancillary data.
1954 		 */
1955 		newmp->b_datap->db_type = M_PROTO;
1956 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1957 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1958 		tudi->PRIM_type = T_UNITDATA_IND;
1959 		tudi->SRC_length = sizeof (sin6_t);
1960 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1961 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1962 		tudi->OPT_length = opt_length;
1963 
1964 		sin6 = (sin6_t *)&tudi[1];
1965 		bzero(sin6, sizeof (sin6_t));
1966 		sin6->sin6_family = AF_INET6;
1967 		sin6->sin6_addr = udp->udp_v6dst;
1968 
1969 		toh = (struct T_opthdr *)&sin6[1];
1970 		toh->level = IPPROTO_IPV6;
1971 		toh->name = IPV6_PATHMTU;
1972 		toh->len = opt_length;
1973 		toh->status = 0;
1974 
1975 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1976 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1977 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1978 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1979 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1980 		/*
1981 		 * We've consumed everything we need from the original
1982 		 * message.  Free it, then send our empty message.
1983 		 */
1984 		freemsg(mp);
1985 		putnext(q, newmp);
1986 		return;
1987 	}
1988 	case ICMP6_TIME_EXCEEDED:
1989 		/* Transient errors */
1990 		break;
1991 	case ICMP6_PARAM_PROB:
1992 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1993 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1994 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1995 		    (uchar_t *)nexthdrp) {
1996 			error = ECONNREFUSED;
1997 			break;
1998 		}
1999 		break;
2000 	}
2001 	if (error == 0) {
2002 		freemsg(mp);
2003 		return;
2004 	}
2005 
2006 	/*
2007 	 * Deliver T_UDERROR_IND when the application has asked for it.
2008 	 * The socket layer enables this automatically when connected.
2009 	 */
2010 	if (!udp->udp_dgram_errind) {
2011 		freemsg(mp);
2012 		return;
2013 	}
2014 
2015 	sin6 = sin6_null;
2016 	sin6.sin6_family = AF_INET6;
2017 	sin6.sin6_addr = ip6h->ip6_dst;
2018 	sin6.sin6_port = udpha->uha_dst_port;
2019 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
2020 
2021 	mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0,
2022 	    error);
2023 	if (mp1)
2024 		putnext(q, mp1);
2025 	freemsg(mp);
2026 }
2027 
2028 /*
2029  * This routine responds to T_ADDR_REQ messages.  It is called by udp_wput.
2030  * The local address is filled in if endpoint is bound. The remote address
2031  * is filled in if remote address has been precified ("connected endpoint")
2032  * (The concept of connected CLTS sockets is alien to published TPI
2033  *  but we support it anyway).
2034  */
2035 static void
2036 udp_addr_req(queue_t *q, mblk_t *mp)
2037 {
2038 	sin_t	*sin;
2039 	sin6_t	*sin6;
2040 	mblk_t	*ackmp;
2041 	struct T_addr_ack *taa;
2042 	udp_t	*udp = Q_TO_UDP(q);
2043 
2044 	/* Make it large enough for worst case */
2045 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
2046 	    2 * sizeof (sin6_t), 1);
2047 	if (ackmp == NULL) {
2048 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
2049 		return;
2050 	}
2051 	taa = (struct T_addr_ack *)ackmp->b_rptr;
2052 
2053 	bzero(taa, sizeof (struct T_addr_ack));
2054 	ackmp->b_wptr = (uchar_t *)&taa[1];
2055 
2056 	taa->PRIM_type = T_ADDR_ACK;
2057 	ackmp->b_datap->db_type = M_PCPROTO;
2058 	rw_enter(&udp->udp_rwlock, RW_READER);
2059 	/*
2060 	 * Note: Following code assumes 32 bit alignment of basic
2061 	 * data structures like sin_t and struct T_addr_ack.
2062 	 */
2063 	if (udp->udp_state != TS_UNBND) {
2064 		/*
2065 		 * Fill in local address first
2066 		 */
2067 		taa->LOCADDR_offset = sizeof (*taa);
2068 		if (udp->udp_family == AF_INET) {
2069 			taa->LOCADDR_length = sizeof (sin_t);
2070 			sin = (sin_t *)&taa[1];
2071 			/* Fill zeroes and then initialize non-zero fields */
2072 			*sin = sin_null;
2073 			sin->sin_family = AF_INET;
2074 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
2075 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
2076 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src,
2077 				    sin->sin_addr.s_addr);
2078 			} else {
2079 				/*
2080 				 * INADDR_ANY
2081 				 * udp_v6src is not set, we might be bound to
2082 				 * broadcast/multicast. Use udp_bound_v6src as
2083 				 * local address instead (that could
2084 				 * also still be INADDR_ANY)
2085 				 */
2086 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src,
2087 				    sin->sin_addr.s_addr);
2088 			}
2089 			sin->sin_port = udp->udp_port;
2090 			ackmp->b_wptr = (uchar_t *)&sin[1];
2091 			if (udp->udp_state == TS_DATA_XFER) {
2092 				/*
2093 				 * connected, fill remote address too
2094 				 */
2095 				taa->REMADDR_length = sizeof (sin_t);
2096 				/* assumed 32-bit alignment */
2097 				taa->REMADDR_offset = taa->LOCADDR_offset +
2098 				    taa->LOCADDR_length;
2099 
2100 				sin = (sin_t *)(ackmp->b_rptr +
2101 				    taa->REMADDR_offset);
2102 				/* initialize */
2103 				*sin = sin_null;
2104 				sin->sin_family = AF_INET;
2105 				sin->sin_addr.s_addr =
2106 				    V4_PART_OF_V6(udp->udp_v6dst);
2107 				sin->sin_port = udp->udp_dstport;
2108 				ackmp->b_wptr = (uchar_t *)&sin[1];
2109 			}
2110 		} else {
2111 			taa->LOCADDR_length = sizeof (sin6_t);
2112 			sin6 = (sin6_t *)&taa[1];
2113 			/* Fill zeroes and then initialize non-zero fields */
2114 			*sin6 = sin6_null;
2115 			sin6->sin6_family = AF_INET6;
2116 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
2117 				sin6->sin6_addr = udp->udp_v6src;
2118 			} else {
2119 				/*
2120 				 * UNSPECIFIED
2121 				 * udp_v6src is not set, we might be bound to
2122 				 * broadcast/multicast. Use udp_bound_v6src as
2123 				 * local address instead (that could
2124 				 * also still be UNSPECIFIED)
2125 				 */
2126 				sin6->sin6_addr =
2127 				    udp->udp_bound_v6src;
2128 			}
2129 			sin6->sin6_port = udp->udp_port;
2130 			ackmp->b_wptr = (uchar_t *)&sin6[1];
2131 			if (udp->udp_state == TS_DATA_XFER) {
2132 				/*
2133 				 * connected, fill remote address too
2134 				 */
2135 				taa->REMADDR_length = sizeof (sin6_t);
2136 				/* assumed 32-bit alignment */
2137 				taa->REMADDR_offset = taa->LOCADDR_offset +
2138 				    taa->LOCADDR_length;
2139 
2140 				sin6 = (sin6_t *)(ackmp->b_rptr +
2141 				    taa->REMADDR_offset);
2142 				/* initialize */
2143 				*sin6 = sin6_null;
2144 				sin6->sin6_family = AF_INET6;
2145 				sin6->sin6_addr = udp->udp_v6dst;
2146 				sin6->sin6_port =  udp->udp_dstport;
2147 				ackmp->b_wptr = (uchar_t *)&sin6[1];
2148 			}
2149 			ackmp->b_wptr = (uchar_t *)&sin6[1];
2150 		}
2151 	}
2152 	rw_exit(&udp->udp_rwlock);
2153 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
2154 	qreply(q, ackmp);
2155 }
2156 
2157 static void
2158 udp_copy_info(struct T_info_ack *tap, udp_t *udp)
2159 {
2160 	if (udp->udp_family == AF_INET) {
2161 		*tap = udp_g_t_info_ack_ipv4;
2162 	} else {
2163 		*tap = udp_g_t_info_ack_ipv6;
2164 	}
2165 	tap->CURRENT_state = udp->udp_state;
2166 	tap->OPT_size = udp_max_optsize;
2167 }
2168 
2169 /*
2170  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
2171  * udp_wput.  Much of the T_CAPABILITY_ACK information is copied from
2172  * udp_g_t_info_ack.  The current state of the stream is copied from
2173  * udp_state.
2174  */
2175 static void
2176 udp_capability_req(queue_t *q, mblk_t *mp)
2177 {
2178 	t_uscalar_t		cap_bits1;
2179 	struct T_capability_ack	*tcap;
2180 	udp_t	*udp = Q_TO_UDP(q);
2181 
2182 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
2183 
2184 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
2185 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
2186 	if (!mp)
2187 		return;
2188 
2189 	tcap = (struct T_capability_ack *)mp->b_rptr;
2190 	tcap->CAP_bits1 = 0;
2191 
2192 	if (cap_bits1 & TC1_INFO) {
2193 		udp_copy_info(&tcap->INFO_ack, udp);
2194 		tcap->CAP_bits1 |= TC1_INFO;
2195 	}
2196 
2197 	qreply(q, mp);
2198 }
2199 
2200 /*
2201  * This routine responds to T_INFO_REQ messages.  It is called by udp_wput.
2202  * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack.
2203  * The current state of the stream is copied from udp_state.
2204  */
2205 static void
2206 udp_info_req(queue_t *q, mblk_t *mp)
2207 {
2208 	udp_t *udp = Q_TO_UDP(q);
2209 
2210 	/* Create a T_INFO_ACK message. */
2211 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
2212 	    T_INFO_ACK);
2213 	if (!mp)
2214 		return;
2215 	udp_copy_info((struct T_info_ack *)mp->b_rptr, udp);
2216 	qreply(q, mp);
2217 }
2218 
2219 /*
2220  * IP recognizes seven kinds of bind requests:
2221  *
2222  * - A zero-length address binds only to the protocol number.
2223  *
2224  * - A 4-byte address is treated as a request to
2225  * validate that the address is a valid local IPv4
2226  * address, appropriate for an application to bind to.
2227  * IP does the verification, but does not make any note
2228  * of the address at this time.
2229  *
2230  * - A 16-byte address contains is treated as a request
2231  * to validate a local IPv6 address, as the 4-byte
2232  * address case above.
2233  *
2234  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
2235  * use it for the inbound fanout of packets.
2236  *
2237  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
2238  * use it for the inbound fanout of packets.
2239  *
2240  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
2241  * information consisting of local and remote addresses
2242  * and ports.  In this case, the addresses are both
2243  * validated as appropriate for this operation, and, if
2244  * so, the information is retained for use in the
2245  * inbound fanout.
2246  *
2247  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
2248  * fanout information, like the 12-byte case above.
2249  *
2250  * IP will also fill in the IRE request mblk with information
2251  * regarding our peer.  In all cases, we notify IP of our protocol
2252  * type by appending a single protocol byte to the bind request.
2253  */
2254 static mblk_t *
2255 udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, t_scalar_t addr_length)
2256 {
2257 	char	*cp;
2258 	mblk_t	*mp;
2259 	struct T_bind_req *tbr;
2260 	ipa_conn_t	*ac;
2261 	ipa6_conn_t	*ac6;
2262 	sin_t		*sin;
2263 	sin6_t		*sin6;
2264 
2265 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
2266 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
2267 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
2268 	if (!mp)
2269 		return (mp);
2270 	mp->b_datap->db_type = M_PROTO;
2271 	tbr = (struct T_bind_req *)mp->b_rptr;
2272 	tbr->PRIM_type = bind_prim;
2273 	tbr->ADDR_offset = sizeof (*tbr);
2274 	tbr->CONIND_number = 0;
2275 	tbr->ADDR_length = addr_length;
2276 	cp = (char *)&tbr[1];
2277 	switch (addr_length) {
2278 	case sizeof (ipa_conn_t):
2279 		ASSERT(udp->udp_family == AF_INET);
2280 		/* Append a request for an IRE */
2281 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2282 		if (!mp->b_cont) {
2283 			freemsg(mp);
2284 			return (NULL);
2285 		}
2286 		mp->b_cont->b_wptr += sizeof (ire_t);
2287 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2288 
2289 		/* cp known to be 32 bit aligned */
2290 		ac = (ipa_conn_t *)cp;
2291 		ac->ac_laddr = V4_PART_OF_V6(udp->udp_v6src);
2292 		ac->ac_faddr = V4_PART_OF_V6(udp->udp_v6dst);
2293 		ac->ac_fport = udp->udp_dstport;
2294 		ac->ac_lport = udp->udp_port;
2295 		break;
2296 
2297 	case sizeof (ipa6_conn_t):
2298 		ASSERT(udp->udp_family == AF_INET6);
2299 		/* Append a request for an IRE */
2300 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2301 		if (!mp->b_cont) {
2302 			freemsg(mp);
2303 			return (NULL);
2304 		}
2305 		mp->b_cont->b_wptr += sizeof (ire_t);
2306 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2307 
2308 		/* cp known to be 32 bit aligned */
2309 		ac6 = (ipa6_conn_t *)cp;
2310 		ac6->ac6_laddr = udp->udp_v6src;
2311 		ac6->ac6_faddr = udp->udp_v6dst;
2312 		ac6->ac6_fport = udp->udp_dstport;
2313 		ac6->ac6_lport = udp->udp_port;
2314 		break;
2315 
2316 	case sizeof (sin_t):
2317 		ASSERT(udp->udp_family == AF_INET);
2318 		/* Append a request for an IRE */
2319 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2320 		if (!mp->b_cont) {
2321 			freemsg(mp);
2322 			return (NULL);
2323 		}
2324 		mp->b_cont->b_wptr += sizeof (ire_t);
2325 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2326 
2327 		sin = (sin_t *)cp;
2328 		*sin = sin_null;
2329 		sin->sin_family = AF_INET;
2330 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_bound_v6src);
2331 		sin->sin_port = udp->udp_port;
2332 		break;
2333 
2334 	case sizeof (sin6_t):
2335 		ASSERT(udp->udp_family == AF_INET6);
2336 		/* Append a request for an IRE */
2337 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2338 		if (!mp->b_cont) {
2339 			freemsg(mp);
2340 			return (NULL);
2341 		}
2342 		mp->b_cont->b_wptr += sizeof (ire_t);
2343 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2344 
2345 		sin6 = (sin6_t *)cp;
2346 		*sin6 = sin6_null;
2347 		sin6->sin6_family = AF_INET6;
2348 		sin6->sin6_addr = udp->udp_bound_v6src;
2349 		sin6->sin6_port = udp->udp_port;
2350 		break;
2351 	}
2352 	/* Add protocol number to end */
2353 	cp[addr_length] = (char)IPPROTO_UDP;
2354 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
2355 	return (mp);
2356 }
2357 
2358 /* For /dev/udp aka AF_INET open */
2359 static int
2360 udp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
2361 {
2362 	return (udp_open(q, devp, flag, sflag, credp, B_FALSE));
2363 }
2364 
2365 /* For /dev/udp6 aka AF_INET6 open */
2366 static int
2367 udp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
2368 {
2369 	return (udp_open(q, devp, flag, sflag, credp, B_TRUE));
2370 }
2371 
2372 /*
2373  * This is the open routine for udp.  It allocates a udp_t structure for
2374  * the stream and, on the first open of the module, creates an ND table.
2375  */
2376 /*ARGSUSED2*/
2377 static int
2378 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
2379     boolean_t isv6)
2380 {
2381 	int		err;
2382 	udp_t		*udp;
2383 	conn_t		*connp;
2384 	dev_t		conn_dev;
2385 	zoneid_t	zoneid;
2386 	netstack_t	*ns;
2387 	udp_stack_t	*us;
2388 	vmem_t		*minor_arena;
2389 
2390 	TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q);
2391 
2392 	/* If the stream is already open, return immediately. */
2393 	if (q->q_ptr != NULL)
2394 		return (0);
2395 
2396 	if (sflag == MODOPEN)
2397 		return (EINVAL);
2398 
2399 	ns = netstack_find_by_cred(credp);
2400 	ASSERT(ns != NULL);
2401 	us = ns->netstack_udp;
2402 	ASSERT(us != NULL);
2403 
2404 	/*
2405 	 * For exclusive stacks we set the zoneid to zero
2406 	 * to make UDP operate as if in the global zone.
2407 	 */
2408 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
2409 		zoneid = GLOBAL_ZONEID;
2410 	else
2411 		zoneid = crgetzoneid(credp);
2412 
2413 	if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
2414 	    ((conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
2415 		minor_arena = ip_minor_arena_la;
2416 	} else {
2417 		/*
2418 		 * Either minor numbers in the large arena were exhausted
2419 		 * or a non socket application is doing the open.
2420 		 * Try to allocate from the small arena.
2421 		 */
2422 		if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0) {
2423 			netstack_rele(ns);
2424 			return (EBUSY);
2425 		}
2426 		minor_arena = ip_minor_arena_sa;
2427 	}
2428 
2429 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
2430 
2431 	connp = ipcl_conn_create(IPCL_UDPCONN, KM_SLEEP, ns);
2432 	connp->conn_dev = conn_dev;
2433 	connp->conn_minor_arena = minor_arena;
2434 	udp = connp->conn_udp;
2435 
2436 	/*
2437 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
2438 	 * done by netstack_find_by_cred()
2439 	 */
2440 	netstack_rele(ns);
2441 
2442 	/*
2443 	 * Initialize the udp_t structure for this stream.
2444 	 */
2445 	q->q_ptr = connp;
2446 	WR(q)->q_ptr = connp;
2447 	connp->conn_rq = q;
2448 	connp->conn_wq = WR(q);
2449 
2450 	rw_enter(&udp->udp_rwlock, RW_WRITER);
2451 	ASSERT(connp->conn_ulp == IPPROTO_UDP);
2452 	ASSERT(connp->conn_udp == udp);
2453 	ASSERT(udp->udp_connp == connp);
2454 
2455 	/* Set the initial state of the stream and the privilege status. */
2456 	udp->udp_state = TS_UNBND;
2457 	if (isv6) {
2458 		udp->udp_family = AF_INET6;
2459 		udp->udp_ipversion = IPV6_VERSION;
2460 		udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
2461 		udp->udp_ttl = us->us_ipv6_hoplimit;
2462 		connp->conn_af_isv6 = B_TRUE;
2463 		connp->conn_flags |= IPCL_ISV6;
2464 	} else {
2465 		udp->udp_family = AF_INET;
2466 		udp->udp_ipversion = IPV4_VERSION;
2467 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE;
2468 		udp->udp_ttl = us->us_ipv4_ttl;
2469 		connp->conn_af_isv6 = B_FALSE;
2470 		connp->conn_flags &= ~IPCL_ISV6;
2471 	}
2472 
2473 	udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2474 	udp->udp_pending_op = -1;
2475 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
2476 	connp->conn_zoneid = zoneid;
2477 
2478 	udp->udp_open_time = lbolt64;
2479 	udp->udp_open_pid = curproc->p_pid;
2480 
2481 	/*
2482 	 * If the caller has the process-wide flag set, then default to MAC
2483 	 * exempt mode.  This allows read-down to unlabeled hosts.
2484 	 */
2485 	if (getpflags(NET_MAC_AWARE, credp) != 0)
2486 		connp->conn_mac_exempt = B_TRUE;
2487 
2488 	if (flag & SO_SOCKSTR) {
2489 		connp->conn_flags |= IPCL_SOCKET;
2490 		udp->udp_issocket = B_TRUE;
2491 		udp->udp_direct_sockfs = B_TRUE;
2492 	}
2493 
2494 	connp->conn_ulp_labeled = is_system_labeled();
2495 
2496 	udp->udp_us = us;
2497 
2498 	q->q_hiwat = us->us_recv_hiwat;
2499 	WR(q)->q_hiwat = us->us_xmit_hiwat;
2500 	WR(q)->q_lowat = us->us_xmit_lowat;
2501 
2502 	connp->conn_recv = udp_input;
2503 	crhold(credp);
2504 	connp->conn_cred = credp;
2505 
2506 	mutex_enter(&connp->conn_lock);
2507 	connp->conn_state_flags &= ~CONN_INCIPIENT;
2508 	mutex_exit(&connp->conn_lock);
2509 
2510 	qprocson(q);
2511 
2512 	if (udp->udp_family == AF_INET6) {
2513 		/* Build initial header template for transmit */
2514 		if ((err = udp_build_hdrs(udp)) != 0) {
2515 			rw_exit(&udp->udp_rwlock);
2516 			qprocsoff(q);
2517 			ipcl_conn_destroy(connp);
2518 			return (err);
2519 		}
2520 	}
2521 	rw_exit(&udp->udp_rwlock);
2522 
2523 	/* Set the Stream head write offset and high watermark. */
2524 	(void) mi_set_sth_wroff(q,
2525 	    udp->udp_max_hdr_len + us->us_wroff_extra);
2526 	(void) mi_set_sth_hiwat(q, udp_set_rcv_hiwat(udp, q->q_hiwat));
2527 
2528 	return (0);
2529 }
2530 
2531 /*
2532  * Which UDP options OK to set through T_UNITDATA_REQ...
2533  */
2534 /* ARGSUSED */
2535 static boolean_t
2536 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
2537 {
2538 	return (B_TRUE);
2539 }
2540 
2541 /*
2542  * This routine gets default values of certain options whose default
2543  * values are maintained by protcol specific code
2544  */
2545 /* ARGSUSED */
2546 int
2547 udp_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2548 {
2549 	udp_t		*udp = Q_TO_UDP(q);
2550 	udp_stack_t *us = udp->udp_us;
2551 	int *i1 = (int *)ptr;
2552 
2553 	switch (level) {
2554 	case IPPROTO_IP:
2555 		switch (name) {
2556 		case IP_MULTICAST_TTL:
2557 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
2558 			return (sizeof (uchar_t));
2559 		case IP_MULTICAST_LOOP:
2560 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
2561 			return (sizeof (uchar_t));
2562 		}
2563 		break;
2564 	case IPPROTO_IPV6:
2565 		switch (name) {
2566 		case IPV6_MULTICAST_HOPS:
2567 			*i1 = IP_DEFAULT_MULTICAST_TTL;
2568 			return (sizeof (int));
2569 		case IPV6_MULTICAST_LOOP:
2570 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
2571 			return (sizeof (int));
2572 		case IPV6_UNICAST_HOPS:
2573 			*i1 = us->us_ipv6_hoplimit;
2574 			return (sizeof (int));
2575 		}
2576 		break;
2577 	}
2578 	return (-1);
2579 }
2580 
2581 /*
2582  * This routine retrieves the current status of socket options.
2583  * It returns the size of the option retrieved.
2584  */
2585 int
2586 udp_opt_get_locked(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2587 {
2588 	int	*i1 = (int *)ptr;
2589 	conn_t	*connp;
2590 	udp_t	*udp;
2591 	ip6_pkt_t *ipp;
2592 	int	len;
2593 	udp_stack_t	*us;
2594 
2595 	connp = Q_TO_CONN(q);
2596 	udp = connp->conn_udp;
2597 	ipp = &udp->udp_sticky_ipp;
2598 	us = udp->udp_us;
2599 
2600 	switch (level) {
2601 	case SOL_SOCKET:
2602 		switch (name) {
2603 		case SO_DEBUG:
2604 			*i1 = udp->udp_debug;
2605 			break;	/* goto sizeof (int) option return */
2606 		case SO_REUSEADDR:
2607 			*i1 = udp->udp_reuseaddr;
2608 			break;	/* goto sizeof (int) option return */
2609 		case SO_TYPE:
2610 			*i1 = SOCK_DGRAM;
2611 			break;	/* goto sizeof (int) option return */
2612 
2613 		/*
2614 		 * The following three items are available here,
2615 		 * but are only meaningful to IP.
2616 		 */
2617 		case SO_DONTROUTE:
2618 			*i1 = udp->udp_dontroute;
2619 			break;	/* goto sizeof (int) option return */
2620 		case SO_USELOOPBACK:
2621 			*i1 = udp->udp_useloopback;
2622 			break;	/* goto sizeof (int) option return */
2623 		case SO_BROADCAST:
2624 			*i1 = udp->udp_broadcast;
2625 			break;	/* goto sizeof (int) option return */
2626 
2627 		case SO_SNDBUF:
2628 			*i1 = q->q_hiwat;
2629 			break;	/* goto sizeof (int) option return */
2630 		case SO_RCVBUF:
2631 			*i1 = RD(q)->q_hiwat;
2632 			break;	/* goto sizeof (int) option return */
2633 		case SO_DGRAM_ERRIND:
2634 			*i1 = udp->udp_dgram_errind;
2635 			break;	/* goto sizeof (int) option return */
2636 		case SO_RECVUCRED:
2637 			*i1 = udp->udp_recvucred;
2638 			break;	/* goto sizeof (int) option return */
2639 		case SO_TIMESTAMP:
2640 			*i1 = udp->udp_timestamp;
2641 			break;	/* goto sizeof (int) option return */
2642 		case SO_ANON_MLP:
2643 			*i1 = connp->conn_anon_mlp;
2644 			break;	/* goto sizeof (int) option return */
2645 		case SO_MAC_EXEMPT:
2646 			*i1 = connp->conn_mac_exempt;
2647 			break;	/* goto sizeof (int) option return */
2648 		case SO_ALLZONES:
2649 			*i1 = connp->conn_allzones;
2650 			break;	/* goto sizeof (int) option return */
2651 		case SO_EXCLBIND:
2652 			*i1 = udp->udp_exclbind ? SO_EXCLBIND : 0;
2653 			break;
2654 		case SO_PROTOTYPE:
2655 			*i1 = IPPROTO_UDP;
2656 			break;
2657 		case SO_DOMAIN:
2658 			*i1 = udp->udp_family;
2659 			break;
2660 		default:
2661 			return (-1);
2662 		}
2663 		break;
2664 	case IPPROTO_IP:
2665 		if (udp->udp_family != AF_INET)
2666 			return (-1);
2667 		switch (name) {
2668 		case IP_OPTIONS:
2669 		case T_IP_OPTIONS:
2670 			len = udp->udp_ip_rcv_options_len - udp->udp_label_len;
2671 			if (len > 0) {
2672 				bcopy(udp->udp_ip_rcv_options +
2673 				    udp->udp_label_len, ptr, len);
2674 			}
2675 			return (len);
2676 		case IP_TOS:
2677 		case T_IP_TOS:
2678 			*i1 = (int)udp->udp_type_of_service;
2679 			break;	/* goto sizeof (int) option return */
2680 		case IP_TTL:
2681 			*i1 = (int)udp->udp_ttl;
2682 			break;	/* goto sizeof (int) option return */
2683 		case IP_DHCPINIT_IF:
2684 			return (-EINVAL);
2685 		case IP_NEXTHOP:
2686 		case IP_RECVPKTINFO:
2687 			/*
2688 			 * This also handles IP_PKTINFO.
2689 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2690 			 * Differentiation is based on the size of the argument
2691 			 * passed in.
2692 			 * This option is handled in IP which will return an
2693 			 * error for IP_PKTINFO as it's not supported as a
2694 			 * sticky option.
2695 			 */
2696 			return (-EINVAL);
2697 		case IP_MULTICAST_IF:
2698 			/* 0 address if not set */
2699 			*(ipaddr_t *)ptr = udp->udp_multicast_if_addr;
2700 			return (sizeof (ipaddr_t));
2701 		case IP_MULTICAST_TTL:
2702 			*(uchar_t *)ptr = udp->udp_multicast_ttl;
2703 			return (sizeof (uchar_t));
2704 		case IP_MULTICAST_LOOP:
2705 			*ptr = connp->conn_multicast_loop;
2706 			return (sizeof (uint8_t));
2707 		case IP_RECVOPTS:
2708 			*i1 = udp->udp_recvopts;
2709 			break;	/* goto sizeof (int) option return */
2710 		case IP_RECVDSTADDR:
2711 			*i1 = udp->udp_recvdstaddr;
2712 			break;	/* goto sizeof (int) option return */
2713 		case IP_RECVIF:
2714 			*i1 = udp->udp_recvif;
2715 			break;	/* goto sizeof (int) option return */
2716 		case IP_RECVSLLA:
2717 			*i1 = udp->udp_recvslla;
2718 			break;	/* goto sizeof (int) option return */
2719 		case IP_RECVTTL:
2720 			*i1 = udp->udp_recvttl;
2721 			break;	/* goto sizeof (int) option return */
2722 		case IP_ADD_MEMBERSHIP:
2723 		case IP_DROP_MEMBERSHIP:
2724 		case IP_BLOCK_SOURCE:
2725 		case IP_UNBLOCK_SOURCE:
2726 		case IP_ADD_SOURCE_MEMBERSHIP:
2727 		case IP_DROP_SOURCE_MEMBERSHIP:
2728 		case MCAST_JOIN_GROUP:
2729 		case MCAST_LEAVE_GROUP:
2730 		case MCAST_BLOCK_SOURCE:
2731 		case MCAST_UNBLOCK_SOURCE:
2732 		case MCAST_JOIN_SOURCE_GROUP:
2733 		case MCAST_LEAVE_SOURCE_GROUP:
2734 		case IP_DONTFAILOVER_IF:
2735 			/* cannot "get" the value for these */
2736 			return (-1);
2737 		case IP_BOUND_IF:
2738 			/* Zero if not set */
2739 			*i1 = udp->udp_bound_if;
2740 			break;	/* goto sizeof (int) option return */
2741 		case IP_UNSPEC_SRC:
2742 			*i1 = udp->udp_unspec_source;
2743 			break;	/* goto sizeof (int) option return */
2744 		case IP_BROADCAST_TTL:
2745 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
2746 			return (sizeof (uchar_t));
2747 		default:
2748 			return (-1);
2749 		}
2750 		break;
2751 	case IPPROTO_IPV6:
2752 		if (udp->udp_family != AF_INET6)
2753 			return (-1);
2754 		switch (name) {
2755 		case IPV6_UNICAST_HOPS:
2756 			*i1 = (unsigned int)udp->udp_ttl;
2757 			break;	/* goto sizeof (int) option return */
2758 		case IPV6_MULTICAST_IF:
2759 			/* 0 index if not set */
2760 			*i1 = udp->udp_multicast_if_index;
2761 			break;	/* goto sizeof (int) option return */
2762 		case IPV6_MULTICAST_HOPS:
2763 			*i1 = udp->udp_multicast_ttl;
2764 			break;	/* goto sizeof (int) option return */
2765 		case IPV6_MULTICAST_LOOP:
2766 			*i1 = connp->conn_multicast_loop;
2767 			break;	/* goto sizeof (int) option return */
2768 		case IPV6_JOIN_GROUP:
2769 		case IPV6_LEAVE_GROUP:
2770 		case MCAST_JOIN_GROUP:
2771 		case MCAST_LEAVE_GROUP:
2772 		case MCAST_BLOCK_SOURCE:
2773 		case MCAST_UNBLOCK_SOURCE:
2774 		case MCAST_JOIN_SOURCE_GROUP:
2775 		case MCAST_LEAVE_SOURCE_GROUP:
2776 			/* cannot "get" the value for these */
2777 			return (-1);
2778 		case IPV6_BOUND_IF:
2779 			/* Zero if not set */
2780 			*i1 = udp->udp_bound_if;
2781 			break;	/* goto sizeof (int) option return */
2782 		case IPV6_UNSPEC_SRC:
2783 			*i1 = udp->udp_unspec_source;
2784 			break;	/* goto sizeof (int) option return */
2785 		case IPV6_RECVPKTINFO:
2786 			*i1 = udp->udp_ip_recvpktinfo;
2787 			break;	/* goto sizeof (int) option return */
2788 		case IPV6_RECVTCLASS:
2789 			*i1 = udp->udp_ipv6_recvtclass;
2790 			break;	/* goto sizeof (int) option return */
2791 		case IPV6_RECVPATHMTU:
2792 			*i1 = udp->udp_ipv6_recvpathmtu;
2793 			break;	/* goto sizeof (int) option return */
2794 		case IPV6_RECVHOPLIMIT:
2795 			*i1 = udp->udp_ipv6_recvhoplimit;
2796 			break;	/* goto sizeof (int) option return */
2797 		case IPV6_RECVHOPOPTS:
2798 			*i1 = udp->udp_ipv6_recvhopopts;
2799 			break;	/* goto sizeof (int) option return */
2800 		case IPV6_RECVDSTOPTS:
2801 			*i1 = udp->udp_ipv6_recvdstopts;
2802 			break;	/* goto sizeof (int) option return */
2803 		case _OLD_IPV6_RECVDSTOPTS:
2804 			*i1 = udp->udp_old_ipv6_recvdstopts;
2805 			break;	/* goto sizeof (int) option return */
2806 		case IPV6_RECVRTHDRDSTOPTS:
2807 			*i1 = udp->udp_ipv6_recvrthdrdstopts;
2808 			break;	/* goto sizeof (int) option return */
2809 		case IPV6_RECVRTHDR:
2810 			*i1 = udp->udp_ipv6_recvrthdr;
2811 			break;	/* goto sizeof (int) option return */
2812 		case IPV6_PKTINFO: {
2813 			/* XXX assumes that caller has room for max size! */
2814 			struct in6_pktinfo *pkti;
2815 
2816 			pkti = (struct in6_pktinfo *)ptr;
2817 			if (ipp->ipp_fields & IPPF_IFINDEX)
2818 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
2819 			else
2820 				pkti->ipi6_ifindex = 0;
2821 			if (ipp->ipp_fields & IPPF_ADDR)
2822 				pkti->ipi6_addr = ipp->ipp_addr;
2823 			else
2824 				pkti->ipi6_addr = ipv6_all_zeros;
2825 			return (sizeof (struct in6_pktinfo));
2826 		}
2827 		case IPV6_TCLASS:
2828 			if (ipp->ipp_fields & IPPF_TCLASS)
2829 				*i1 = ipp->ipp_tclass;
2830 			else
2831 				*i1 = IPV6_FLOW_TCLASS(
2832 				    IPV6_DEFAULT_VERS_AND_FLOW);
2833 			break;	/* goto sizeof (int) option return */
2834 		case IPV6_NEXTHOP: {
2835 			sin6_t *sin6 = (sin6_t *)ptr;
2836 
2837 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2838 				return (0);
2839 			*sin6 = sin6_null;
2840 			sin6->sin6_family = AF_INET6;
2841 			sin6->sin6_addr = ipp->ipp_nexthop;
2842 			return (sizeof (sin6_t));
2843 		}
2844 		case IPV6_HOPOPTS:
2845 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2846 				return (0);
2847 			if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6)
2848 				return (0);
2849 			/*
2850 			 * The cipso/label option is added by kernel.
2851 			 * User is not usually aware of this option.
2852 			 * We copy out the hbh opt after the label option.
2853 			 */
2854 			bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6,
2855 			    ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2856 			if (udp->udp_label_len_v6 > 0) {
2857 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2858 				ptr[1] = (ipp->ipp_hopoptslen -
2859 				    udp->udp_label_len_v6 + 7) / 8 - 1;
2860 			}
2861 			return (ipp->ipp_hopoptslen - udp->udp_label_len_v6);
2862 		case IPV6_RTHDRDSTOPTS:
2863 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2864 				return (0);
2865 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2866 			return (ipp->ipp_rtdstoptslen);
2867 		case IPV6_RTHDR:
2868 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2869 				return (0);
2870 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2871 			return (ipp->ipp_rthdrlen);
2872 		case IPV6_DSTOPTS:
2873 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
2874 				return (0);
2875 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2876 			return (ipp->ipp_dstoptslen);
2877 		case IPV6_PATHMTU:
2878 			return (ip_fill_mtuinfo(&udp->udp_v6dst,
2879 			    udp->udp_dstport, (struct ip6_mtuinfo *)ptr,
2880 			    us->us_netstack));
2881 		default:
2882 			return (-1);
2883 		}
2884 		break;
2885 	case IPPROTO_UDP:
2886 		switch (name) {
2887 		case UDP_ANONPRIVBIND:
2888 			*i1 = udp->udp_anon_priv_bind;
2889 			break;
2890 		case UDP_EXCLBIND:
2891 			*i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0;
2892 			break;
2893 		case UDP_RCVHDR:
2894 			*i1 = udp->udp_rcvhdr ? 1 : 0;
2895 			break;
2896 		case UDP_NAT_T_ENDPOINT:
2897 			*i1 = udp->udp_nat_t_endpoint;
2898 			break;
2899 		default:
2900 			return (-1);
2901 		}
2902 		break;
2903 	default:
2904 		return (-1);
2905 	}
2906 	return (sizeof (int));
2907 }
2908 
2909 int
2910 udp_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
2911 {
2912 	udp_t	*udp;
2913 	int	err;
2914 
2915 	udp = Q_TO_UDP(q);
2916 
2917 	rw_enter(&udp->udp_rwlock, RW_READER);
2918 	err = udp_opt_get_locked(q, level, name, ptr);
2919 	rw_exit(&udp->udp_rwlock);
2920 	return (err);
2921 }
2922 
2923 /*
2924  * This routine sets socket options.
2925  */
2926 /* ARGSUSED */
2927 int
2928 udp_opt_set_locked(queue_t *q, uint_t optset_context, int level,
2929     int name, uint_t inlen, uchar_t *invalp, uint_t *outlenp,
2930     uchar_t *outvalp, void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
2931 {
2932 	udpattrs_t *attrs = thisdg_attrs;
2933 	int	*i1 = (int *)invalp;
2934 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2935 	boolean_t checkonly;
2936 	int	error;
2937 	conn_t	*connp;
2938 	udp_t	*udp;
2939 	uint_t	newlen;
2940 	udp_stack_t *us;
2941 	size_t	sth_wroff;
2942 
2943 	connp = Q_TO_CONN(q);
2944 	udp = connp->conn_udp;
2945 	us = udp->udp_us;
2946 
2947 	switch (optset_context) {
2948 	case SETFN_OPTCOM_CHECKONLY:
2949 		checkonly = B_TRUE;
2950 		/*
2951 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
2952 		 * inlen != 0 implies value supplied and
2953 		 * 	we have to "pretend" to set it.
2954 		 * inlen == 0 implies that there is no
2955 		 * 	value part in T_CHECK request and just validation
2956 		 * done elsewhere should be enough, we just return here.
2957 		 */
2958 		if (inlen == 0) {
2959 			*outlenp = 0;
2960 			return (0);
2961 		}
2962 		break;
2963 	case SETFN_OPTCOM_NEGOTIATE:
2964 		checkonly = B_FALSE;
2965 		break;
2966 	case SETFN_UD_NEGOTIATE:
2967 	case SETFN_CONN_NEGOTIATE:
2968 		checkonly = B_FALSE;
2969 		/*
2970 		 * Negotiating local and "association-related" options
2971 		 * through T_UNITDATA_REQ.
2972 		 *
2973 		 * Following routine can filter out ones we do not
2974 		 * want to be "set" this way.
2975 		 */
2976 		if (!udp_opt_allow_udr_set(level, name)) {
2977 			*outlenp = 0;
2978 			return (EINVAL);
2979 		}
2980 		break;
2981 	default:
2982 		/*
2983 		 * We should never get here
2984 		 */
2985 		*outlenp = 0;
2986 		return (EINVAL);
2987 	}
2988 
2989 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
2990 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
2991 
2992 	/*
2993 	 * For fixed length options, no sanity check
2994 	 * of passed in length is done. It is assumed *_optcom_req()
2995 	 * routines do the right thing.
2996 	 */
2997 
2998 	switch (level) {
2999 	case SOL_SOCKET:
3000 		switch (name) {
3001 		case SO_REUSEADDR:
3002 			if (!checkonly)
3003 				udp->udp_reuseaddr = onoff;
3004 			break;
3005 		case SO_DEBUG:
3006 			if (!checkonly)
3007 				udp->udp_debug = onoff;
3008 			break;
3009 		/*
3010 		 * The following three items are available here,
3011 		 * but are only meaningful to IP.
3012 		 */
3013 		case SO_DONTROUTE:
3014 			if (!checkonly)
3015 				udp->udp_dontroute = onoff;
3016 			break;
3017 		case SO_USELOOPBACK:
3018 			if (!checkonly)
3019 				udp->udp_useloopback = onoff;
3020 			break;
3021 		case SO_BROADCAST:
3022 			if (!checkonly)
3023 				udp->udp_broadcast = onoff;
3024 			break;
3025 
3026 		case SO_SNDBUF:
3027 			if (*i1 > us->us_max_buf) {
3028 				*outlenp = 0;
3029 				return (ENOBUFS);
3030 			}
3031 			if (!checkonly) {
3032 				q->q_hiwat = *i1;
3033 			}
3034 			break;
3035 		case SO_RCVBUF:
3036 			if (*i1 > us->us_max_buf) {
3037 				*outlenp = 0;
3038 				return (ENOBUFS);
3039 			}
3040 			if (!checkonly) {
3041 				RD(q)->q_hiwat = *i1;
3042 				rw_exit(&udp->udp_rwlock);
3043 				(void) mi_set_sth_hiwat(RD(q),
3044 				    udp_set_rcv_hiwat(udp, *i1));
3045 				rw_enter(&udp->udp_rwlock, RW_WRITER);
3046 			}
3047 			break;
3048 		case SO_DGRAM_ERRIND:
3049 			if (!checkonly)
3050 				udp->udp_dgram_errind = onoff;
3051 			break;
3052 		case SO_RECVUCRED:
3053 			if (!checkonly)
3054 				udp->udp_recvucred = onoff;
3055 			break;
3056 		case SO_ALLZONES:
3057 			/*
3058 			 * "soft" error (negative)
3059 			 * option not handled at this level
3060 			 * Do not modify *outlenp.
3061 			 */
3062 			return (-EINVAL);
3063 		case SO_TIMESTAMP:
3064 			if (!checkonly)
3065 				udp->udp_timestamp = onoff;
3066 			break;
3067 		case SO_ANON_MLP:
3068 			/* Pass option along to IP level for handling */
3069 			return (-EINVAL);
3070 		case SO_MAC_EXEMPT:
3071 			/* Pass option along to IP level for handling */
3072 			return (-EINVAL);
3073 		case SCM_UCRED: {
3074 			struct ucred_s *ucr;
3075 			cred_t *cr, *newcr;
3076 			ts_label_t *tsl;
3077 
3078 			/*
3079 			 * Only sockets that have proper privileges and are
3080 			 * bound to MLPs will have any other value here, so
3081 			 * this implicitly tests for privilege to set label.
3082 			 */
3083 			if (connp->conn_mlp_type == mlptSingle)
3084 				break;
3085 			ucr = (struct ucred_s *)invalp;
3086 			if (inlen != ucredsize ||
3087 			    ucr->uc_labeloff < sizeof (*ucr) ||
3088 			    ucr->uc_labeloff + sizeof (bslabel_t) > inlen)
3089 				return (EINVAL);
3090 			if (!checkonly) {
3091 				mblk_t *mb;
3092 
3093 				if (attrs == NULL ||
3094 				    (mb = attrs->udpattr_mb) == NULL)
3095 					return (EINVAL);
3096 				if ((cr = DB_CRED(mb)) == NULL)
3097 					cr = udp->udp_connp->conn_cred;
3098 				ASSERT(cr != NULL);
3099 				if ((tsl = crgetlabel(cr)) == NULL)
3100 					return (EINVAL);
3101 				newcr = copycred_from_bslabel(cr, UCLABEL(ucr),
3102 				    tsl->tsl_doi, KM_NOSLEEP);
3103 				if (newcr == NULL)
3104 					return (ENOSR);
3105 				mblk_setcred(mb, newcr);
3106 				attrs->udpattr_credset = B_TRUE;
3107 				crfree(newcr);
3108 			}
3109 			break;
3110 		}
3111 		case SO_EXCLBIND:
3112 			if (!checkonly)
3113 				udp->udp_exclbind = onoff;
3114 			break;
3115 		default:
3116 			*outlenp = 0;
3117 			return (EINVAL);
3118 		}
3119 		break;
3120 	case IPPROTO_IP:
3121 		if (udp->udp_family != AF_INET) {
3122 			*outlenp = 0;
3123 			return (ENOPROTOOPT);
3124 		}
3125 		switch (name) {
3126 		case IP_OPTIONS:
3127 		case T_IP_OPTIONS:
3128 			/* Save options for use by IP. */
3129 			newlen = inlen + udp->udp_label_len;
3130 			if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
3131 				*outlenp = 0;
3132 				return (EINVAL);
3133 			}
3134 			if (checkonly)
3135 				break;
3136 
3137 			/*
3138 			 * Update the stored options taking into account
3139 			 * any CIPSO option which we should not overwrite.
3140 			 */
3141 			if (!tsol_option_set(&udp->udp_ip_snd_options,
3142 			    &udp->udp_ip_snd_options_len,
3143 			    udp->udp_label_len, invalp, inlen)) {
3144 				*outlenp = 0;
3145 				return (ENOMEM);
3146 			}
3147 
3148 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3149 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
3150 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
3151 			rw_exit(&udp->udp_rwlock);
3152 			(void) mi_set_sth_wroff(RD(q), sth_wroff);
3153 			rw_enter(&udp->udp_rwlock, RW_WRITER);
3154 			break;
3155 
3156 		case IP_TTL:
3157 			if (!checkonly) {
3158 				udp->udp_ttl = (uchar_t)*i1;
3159 			}
3160 			break;
3161 		case IP_TOS:
3162 		case T_IP_TOS:
3163 			if (!checkonly) {
3164 				udp->udp_type_of_service = (uchar_t)*i1;
3165 			}
3166 			break;
3167 		case IP_MULTICAST_IF: {
3168 			/*
3169 			 * TODO should check OPTMGMT reply and undo this if
3170 			 * there is an error.
3171 			 */
3172 			struct in_addr *inap = (struct in_addr *)invalp;
3173 			if (!checkonly) {
3174 				udp->udp_multicast_if_addr =
3175 				    inap->s_addr;
3176 			}
3177 			break;
3178 		}
3179 		case IP_MULTICAST_TTL:
3180 			if (!checkonly)
3181 				udp->udp_multicast_ttl = *invalp;
3182 			break;
3183 		case IP_MULTICAST_LOOP:
3184 			if (!checkonly)
3185 				connp->conn_multicast_loop = *invalp;
3186 			break;
3187 		case IP_RECVOPTS:
3188 			if (!checkonly)
3189 				udp->udp_recvopts = onoff;
3190 			break;
3191 		case IP_RECVDSTADDR:
3192 			if (!checkonly)
3193 				udp->udp_recvdstaddr = onoff;
3194 			break;
3195 		case IP_RECVIF:
3196 			if (!checkonly)
3197 				udp->udp_recvif = onoff;
3198 			break;
3199 		case IP_RECVSLLA:
3200 			if (!checkonly)
3201 				udp->udp_recvslla = onoff;
3202 			break;
3203 		case IP_RECVTTL:
3204 			if (!checkonly)
3205 				udp->udp_recvttl = onoff;
3206 			break;
3207 		case IP_PKTINFO: {
3208 			/*
3209 			 * This also handles IP_RECVPKTINFO.
3210 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
3211 			 * Differentiation is based on the size of the
3212 			 * argument passed in.
3213 			 */
3214 			struct in_pktinfo *pktinfop;
3215 			ip4_pkt_t *attr_pktinfop;
3216 
3217 			if (checkonly)
3218 				break;
3219 
3220 			if (inlen == sizeof (int)) {
3221 				/*
3222 				 * This is IP_RECVPKTINFO option.
3223 				 * Keep a local copy of whether this option is
3224 				 * set or not and pass it down to IP for
3225 				 * processing.
3226 				 */
3227 
3228 				udp->udp_ip_recvpktinfo = onoff;
3229 				return (-EINVAL);
3230 			}
3231 
3232 			if (attrs == NULL ||
3233 			    (attr_pktinfop = attrs->udpattr_ipp4) == NULL) {
3234 				/*
3235 				 * sticky option or no buffer to return
3236 				 * the results.
3237 				 */
3238 				return (EINVAL);
3239 			}
3240 
3241 			if (inlen != sizeof (struct in_pktinfo))
3242 				return (EINVAL);
3243 
3244 			pktinfop = (struct in_pktinfo *)invalp;
3245 
3246 			/*
3247 			 * At least one of the values should be specified
3248 			 */
3249 			if (pktinfop->ipi_ifindex == 0 &&
3250 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
3251 				return (EINVAL);
3252 			}
3253 
3254 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
3255 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
3256 
3257 			break;
3258 		}
3259 		case IP_ADD_MEMBERSHIP:
3260 		case IP_DROP_MEMBERSHIP:
3261 		case IP_BLOCK_SOURCE:
3262 		case IP_UNBLOCK_SOURCE:
3263 		case IP_ADD_SOURCE_MEMBERSHIP:
3264 		case IP_DROP_SOURCE_MEMBERSHIP:
3265 		case MCAST_JOIN_GROUP:
3266 		case MCAST_LEAVE_GROUP:
3267 		case MCAST_BLOCK_SOURCE:
3268 		case MCAST_UNBLOCK_SOURCE:
3269 		case MCAST_JOIN_SOURCE_GROUP:
3270 		case MCAST_LEAVE_SOURCE_GROUP:
3271 		case IP_SEC_OPT:
3272 		case IP_NEXTHOP:
3273 		case IP_DHCPINIT_IF:
3274 			/*
3275 			 * "soft" error (negative)
3276 			 * option not handled at this level
3277 			 * Do not modify *outlenp.
3278 			 */
3279 			return (-EINVAL);
3280 		case IP_BOUND_IF:
3281 			if (!checkonly)
3282 				udp->udp_bound_if = *i1;
3283 			break;
3284 		case IP_UNSPEC_SRC:
3285 			if (!checkonly)
3286 				udp->udp_unspec_source = onoff;
3287 			break;
3288 		case IP_BROADCAST_TTL:
3289 			if (!checkonly)
3290 				connp->conn_broadcast_ttl = *invalp;
3291 			break;
3292 		default:
3293 			*outlenp = 0;
3294 			return (EINVAL);
3295 		}
3296 		break;
3297 	case IPPROTO_IPV6: {
3298 		ip6_pkt_t		*ipp;
3299 		boolean_t		sticky;
3300 
3301 		if (udp->udp_family != AF_INET6) {
3302 			*outlenp = 0;
3303 			return (ENOPROTOOPT);
3304 		}
3305 		/*
3306 		 * Deal with both sticky options and ancillary data
3307 		 */
3308 		sticky = B_FALSE;
3309 		if (attrs == NULL || (ipp = attrs->udpattr_ipp6) ==
3310 		    NULL) {
3311 			/* sticky options, or none */
3312 			ipp = &udp->udp_sticky_ipp;
3313 			sticky = B_TRUE;
3314 		}
3315 
3316 		switch (name) {
3317 		case IPV6_MULTICAST_IF:
3318 			if (!checkonly)
3319 				udp->udp_multicast_if_index = *i1;
3320 			break;
3321 		case IPV6_UNICAST_HOPS:
3322 			/* -1 means use default */
3323 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
3324 				*outlenp = 0;
3325 				return (EINVAL);
3326 			}
3327 			if (!checkonly) {
3328 				if (*i1 == -1) {
3329 					udp->udp_ttl = ipp->ipp_unicast_hops =
3330 					    us->us_ipv6_hoplimit;
3331 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
3332 					/* Pass modified value to IP. */
3333 					*i1 = udp->udp_ttl;
3334 				} else {
3335 					udp->udp_ttl = ipp->ipp_unicast_hops =
3336 					    (uint8_t)*i1;
3337 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
3338 				}
3339 				/* Rebuild the header template */
3340 				error = udp_build_hdrs(udp);
3341 				if (error != 0) {
3342 					*outlenp = 0;
3343 					return (error);
3344 				}
3345 			}
3346 			break;
3347 		case IPV6_MULTICAST_HOPS:
3348 			/* -1 means use default */
3349 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
3350 				*outlenp = 0;
3351 				return (EINVAL);
3352 			}
3353 			if (!checkonly) {
3354 				if (*i1 == -1) {
3355 					udp->udp_multicast_ttl =
3356 					    ipp->ipp_multicast_hops =
3357 					    IP_DEFAULT_MULTICAST_TTL;
3358 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
3359 					/* Pass modified value to IP. */
3360 					*i1 = udp->udp_multicast_ttl;
3361 				} else {
3362 					udp->udp_multicast_ttl =
3363 					    ipp->ipp_multicast_hops =
3364 					    (uint8_t)*i1;
3365 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
3366 				}
3367 			}
3368 			break;
3369 		case IPV6_MULTICAST_LOOP:
3370 			if (*i1 != 0 && *i1 != 1) {
3371 				*outlenp = 0;
3372 				return (EINVAL);
3373 			}
3374 			if (!checkonly)
3375 				connp->conn_multicast_loop = *i1;
3376 			break;
3377 		case IPV6_JOIN_GROUP:
3378 		case IPV6_LEAVE_GROUP:
3379 		case MCAST_JOIN_GROUP:
3380 		case MCAST_LEAVE_GROUP:
3381 		case MCAST_BLOCK_SOURCE:
3382 		case MCAST_UNBLOCK_SOURCE:
3383 		case MCAST_JOIN_SOURCE_GROUP:
3384 		case MCAST_LEAVE_SOURCE_GROUP:
3385 			/*
3386 			 * "soft" error (negative)
3387 			 * option not handled at this level
3388 			 * Note: Do not modify *outlenp
3389 			 */
3390 			return (-EINVAL);
3391 		case IPV6_BOUND_IF:
3392 			if (!checkonly)
3393 				udp->udp_bound_if = *i1;
3394 			break;
3395 		case IPV6_UNSPEC_SRC:
3396 			if (!checkonly)
3397 				udp->udp_unspec_source = onoff;
3398 			break;
3399 		/*
3400 		 * Set boolean switches for ancillary data delivery
3401 		 */
3402 		case IPV6_RECVPKTINFO:
3403 			if (!checkonly)
3404 				udp->udp_ip_recvpktinfo = onoff;
3405 			break;
3406 		case IPV6_RECVTCLASS:
3407 			if (!checkonly) {
3408 				udp->udp_ipv6_recvtclass = onoff;
3409 			}
3410 			break;
3411 		case IPV6_RECVPATHMTU:
3412 			if (!checkonly) {
3413 				udp->udp_ipv6_recvpathmtu = onoff;
3414 			}
3415 			break;
3416 		case IPV6_RECVHOPLIMIT:
3417 			if (!checkonly)
3418 				udp->udp_ipv6_recvhoplimit = onoff;
3419 			break;
3420 		case IPV6_RECVHOPOPTS:
3421 			if (!checkonly)
3422 				udp->udp_ipv6_recvhopopts = onoff;
3423 			break;
3424 		case IPV6_RECVDSTOPTS:
3425 			if (!checkonly)
3426 				udp->udp_ipv6_recvdstopts = onoff;
3427 			break;
3428 		case _OLD_IPV6_RECVDSTOPTS:
3429 			if (!checkonly)
3430 				udp->udp_old_ipv6_recvdstopts = onoff;
3431 			break;
3432 		case IPV6_RECVRTHDRDSTOPTS:
3433 			if (!checkonly)
3434 				udp->udp_ipv6_recvrthdrdstopts = onoff;
3435 			break;
3436 		case IPV6_RECVRTHDR:
3437 			if (!checkonly)
3438 				udp->udp_ipv6_recvrthdr = onoff;
3439 			break;
3440 		/*
3441 		 * Set sticky options or ancillary data.
3442 		 * If sticky options, (re)build any extension headers
3443 		 * that might be needed as a result.
3444 		 */
3445 		case IPV6_PKTINFO:
3446 			/*
3447 			 * The source address and ifindex are verified
3448 			 * in ip_opt_set(). For ancillary data the
3449 			 * source address is checked in ip_wput_v6.
3450 			 */
3451 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
3452 				return (EINVAL);
3453 			if (checkonly)
3454 				break;
3455 
3456 			if (inlen == 0) {
3457 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
3458 				ipp->ipp_sticky_ignored |=
3459 				    (IPPF_IFINDEX|IPPF_ADDR);
3460 			} else {
3461 				struct in6_pktinfo *pkti;
3462 
3463 				pkti = (struct in6_pktinfo *)invalp;
3464 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
3465 				ipp->ipp_addr = pkti->ipi6_addr;
3466 				if (ipp->ipp_ifindex != 0)
3467 					ipp->ipp_fields |= IPPF_IFINDEX;
3468 				else
3469 					ipp->ipp_fields &= ~IPPF_IFINDEX;
3470 				if (!IN6_IS_ADDR_UNSPECIFIED(
3471 				    &ipp->ipp_addr))
3472 					ipp->ipp_fields |= IPPF_ADDR;
3473 				else
3474 					ipp->ipp_fields &= ~IPPF_ADDR;
3475 			}
3476 			if (sticky) {
3477 				error = udp_build_hdrs(udp);
3478 				if (error != 0)
3479 					return (error);
3480 			}
3481 			break;
3482 		case IPV6_HOPLIMIT:
3483 			if (sticky)
3484 				return (EINVAL);
3485 			if (inlen != 0 && inlen != sizeof (int))
3486 				return (EINVAL);
3487 			if (checkonly)
3488 				break;
3489 
3490 			if (inlen == 0) {
3491 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
3492 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
3493 			} else {
3494 				if (*i1 > 255 || *i1 < -1)
3495 					return (EINVAL);
3496 				if (*i1 == -1)
3497 					ipp->ipp_hoplimit =
3498 					    us->us_ipv6_hoplimit;
3499 				else
3500 					ipp->ipp_hoplimit = *i1;
3501 				ipp->ipp_fields |= IPPF_HOPLIMIT;
3502 			}
3503 			break;
3504 		case IPV6_TCLASS:
3505 			if (inlen != 0 && inlen != sizeof (int))
3506 				return (EINVAL);
3507 			if (checkonly)
3508 				break;
3509 
3510 			if (inlen == 0) {
3511 				ipp->ipp_fields &= ~IPPF_TCLASS;
3512 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
3513 			} else {
3514 				if (*i1 > 255 || *i1 < -1)
3515 					return (EINVAL);
3516 				if (*i1 == -1)
3517 					ipp->ipp_tclass = 0;
3518 				else
3519 					ipp->ipp_tclass = *i1;
3520 				ipp->ipp_fields |= IPPF_TCLASS;
3521 			}
3522 			if (sticky) {
3523 				error = udp_build_hdrs(udp);
3524 				if (error != 0)
3525 					return (error);
3526 			}
3527 			break;
3528 		case IPV6_NEXTHOP:
3529 			/*
3530 			 * IP will verify that the nexthop is reachable
3531 			 * and fail for sticky options.
3532 			 */
3533 			if (inlen != 0 && inlen != sizeof (sin6_t))
3534 				return (EINVAL);
3535 			if (checkonly)
3536 				break;
3537 
3538 			if (inlen == 0) {
3539 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
3540 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
3541 			} else {
3542 				sin6_t *sin6 = (sin6_t *)invalp;
3543 
3544 				if (sin6->sin6_family != AF_INET6)
3545 					return (EAFNOSUPPORT);
3546 				if (IN6_IS_ADDR_V4MAPPED(
3547 				    &sin6->sin6_addr))
3548 					return (EADDRNOTAVAIL);
3549 				ipp->ipp_nexthop = sin6->sin6_addr;
3550 				if (!IN6_IS_ADDR_UNSPECIFIED(
3551 				    &ipp->ipp_nexthop))
3552 					ipp->ipp_fields |= IPPF_NEXTHOP;
3553 				else
3554 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
3555 			}
3556 			if (sticky) {
3557 				error = udp_build_hdrs(udp);
3558 				if (error != 0)
3559 					return (error);
3560 			}
3561 			break;
3562 		case IPV6_HOPOPTS: {
3563 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
3564 			/*
3565 			 * Sanity checks - minimum size, size a multiple of
3566 			 * eight bytes, and matching size passed in.
3567 			 */
3568 			if (inlen != 0 &&
3569 			    inlen != (8 * (hopts->ip6h_len + 1)))
3570 				return (EINVAL);
3571 
3572 			if (checkonly)
3573 				break;
3574 
3575 			error = optcom_pkt_set(invalp, inlen, sticky,
3576 			    (uchar_t **)&ipp->ipp_hopopts,
3577 			    &ipp->ipp_hopoptslen,
3578 			    sticky ? udp->udp_label_len_v6 : 0);
3579 			if (error != 0)
3580 				return (error);
3581 			if (ipp->ipp_hopoptslen == 0) {
3582 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
3583 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
3584 			} else {
3585 				ipp->ipp_fields |= IPPF_HOPOPTS;
3586 			}
3587 			if (sticky) {
3588 				error = udp_build_hdrs(udp);
3589 				if (error != 0)
3590 					return (error);
3591 			}
3592 			break;
3593 		}
3594 		case IPV6_RTHDRDSTOPTS: {
3595 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
3596 
3597 			/*
3598 			 * Sanity checks - minimum size, size a multiple of
3599 			 * eight bytes, and matching size passed in.
3600 			 */
3601 			if (inlen != 0 &&
3602 			    inlen != (8 * (dopts->ip6d_len + 1)))
3603 				return (EINVAL);
3604 
3605 			if (checkonly)
3606 				break;
3607 
3608 			if (inlen == 0) {
3609 				if (sticky &&
3610 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
3611 					kmem_free(ipp->ipp_rtdstopts,
3612 					    ipp->ipp_rtdstoptslen);
3613 					ipp->ipp_rtdstopts = NULL;
3614 					ipp->ipp_rtdstoptslen = 0;
3615 				}
3616 
3617 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
3618 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
3619 			} else {
3620 				error = optcom_pkt_set(invalp, inlen, sticky,
3621 				    (uchar_t **)&ipp->ipp_rtdstopts,
3622 				    &ipp->ipp_rtdstoptslen, 0);
3623 				if (error != 0)
3624 					return (error);
3625 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
3626 			}
3627 			if (sticky) {
3628 				error = udp_build_hdrs(udp);
3629 				if (error != 0)
3630 					return (error);
3631 			}
3632 			break;
3633 		}
3634 		case IPV6_DSTOPTS: {
3635 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
3636 
3637 			/*
3638 			 * Sanity checks - minimum size, size a multiple of
3639 			 * eight bytes, and matching size passed in.
3640 			 */
3641 			if (inlen != 0 &&
3642 			    inlen != (8 * (dopts->ip6d_len + 1)))
3643 				return (EINVAL);
3644 
3645 			if (checkonly)
3646 				break;
3647 
3648 			if (inlen == 0) {
3649 				if (sticky &&
3650 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
3651 					kmem_free(ipp->ipp_dstopts,
3652 					    ipp->ipp_dstoptslen);
3653 					ipp->ipp_dstopts = NULL;
3654 					ipp->ipp_dstoptslen = 0;
3655 				}
3656 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
3657 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
3658 			} else {
3659 				error = optcom_pkt_set(invalp, inlen, sticky,
3660 				    (uchar_t **)&ipp->ipp_dstopts,
3661 				    &ipp->ipp_dstoptslen, 0);
3662 				if (error != 0)
3663 					return (error);
3664 				ipp->ipp_fields |= IPPF_DSTOPTS;
3665 			}
3666 			if (sticky) {
3667 				error = udp_build_hdrs(udp);
3668 				if (error != 0)
3669 					return (error);
3670 			}
3671 			break;
3672 		}
3673 		case IPV6_RTHDR: {
3674 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
3675 
3676 			/*
3677 			 * Sanity checks - minimum size, size a multiple of
3678 			 * eight bytes, and matching size passed in.
3679 			 */
3680 			if (inlen != 0 &&
3681 			    inlen != (8 * (rt->ip6r_len + 1)))
3682 				return (EINVAL);
3683 
3684 			if (checkonly)
3685 				break;
3686 
3687 			if (inlen == 0) {
3688 				if (sticky &&
3689 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
3690 					kmem_free(ipp->ipp_rthdr,
3691 					    ipp->ipp_rthdrlen);
3692 					ipp->ipp_rthdr = NULL;
3693 					ipp->ipp_rthdrlen = 0;
3694 				}
3695 				ipp->ipp_fields &= ~IPPF_RTHDR;
3696 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
3697 			} else {
3698 				error = optcom_pkt_set(invalp, inlen, sticky,
3699 				    (uchar_t **)&ipp->ipp_rthdr,
3700 				    &ipp->ipp_rthdrlen, 0);
3701 				if (error != 0)
3702 					return (error);
3703 				ipp->ipp_fields |= IPPF_RTHDR;
3704 			}
3705 			if (sticky) {
3706 				error = udp_build_hdrs(udp);
3707 				if (error != 0)
3708 					return (error);
3709 			}
3710 			break;
3711 		}
3712 
3713 		case IPV6_DONTFRAG:
3714 			if (checkonly)
3715 				break;
3716 
3717 			if (onoff) {
3718 				ipp->ipp_fields |= IPPF_DONTFRAG;
3719 			} else {
3720 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
3721 			}
3722 			break;
3723 
3724 		case IPV6_USE_MIN_MTU:
3725 			if (inlen != sizeof (int))
3726 				return (EINVAL);
3727 
3728 			if (*i1 < -1 || *i1 > 1)
3729 				return (EINVAL);
3730 
3731 			if (checkonly)
3732 				break;
3733 
3734 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3735 			ipp->ipp_use_min_mtu = *i1;
3736 			break;
3737 
3738 		case IPV6_BOUND_PIF:
3739 		case IPV6_SEC_OPT:
3740 		case IPV6_DONTFAILOVER_IF:
3741 		case IPV6_SRC_PREFERENCES:
3742 		case IPV6_V6ONLY:
3743 			/* Handled at the IP level */
3744 			return (-EINVAL);
3745 		default:
3746 			*outlenp = 0;
3747 			return (EINVAL);
3748 		}
3749 		break;
3750 		}		/* end IPPROTO_IPV6 */
3751 	case IPPROTO_UDP:
3752 		switch (name) {
3753 		case UDP_ANONPRIVBIND:
3754 			if ((error = secpolicy_net_privaddr(cr, 0,
3755 			    IPPROTO_UDP)) != 0) {
3756 				*outlenp = 0;
3757 				return (error);
3758 			}
3759 			if (!checkonly) {
3760 				udp->udp_anon_priv_bind = onoff;
3761 			}
3762 			break;
3763 		case UDP_EXCLBIND:
3764 			if (!checkonly)
3765 				udp->udp_exclbind = onoff;
3766 			break;
3767 		case UDP_RCVHDR:
3768 			if (!checkonly)
3769 				udp->udp_rcvhdr = onoff;
3770 			break;
3771 		case UDP_NAT_T_ENDPOINT:
3772 			if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
3773 				*outlenp = 0;
3774 				return (error);
3775 			}
3776 
3777 			/*
3778 			 * Use udp_family instead so we can avoid ambiguitites
3779 			 * with AF_INET6 sockets that may switch from IPv4
3780 			 * to IPv6.
3781 			 */
3782 			if (udp->udp_family != AF_INET) {
3783 				*outlenp = 0;
3784 				return (EAFNOSUPPORT);
3785 			}
3786 
3787 			if (!checkonly) {
3788 				udp->udp_nat_t_endpoint = onoff;
3789 
3790 				udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3791 				    UDPH_SIZE + udp->udp_ip_snd_options_len;
3792 
3793 				/* Also, adjust wroff */
3794 				if (onoff) {
3795 					udp->udp_max_hdr_len +=
3796 					    sizeof (uint32_t);
3797 				}
3798 				(void) mi_set_sth_wroff(RD(q),
3799 				    udp->udp_max_hdr_len + us->us_wroff_extra);
3800 			}
3801 			break;
3802 		default:
3803 			*outlenp = 0;
3804 			return (EINVAL);
3805 		}
3806 		break;
3807 	default:
3808 		*outlenp = 0;
3809 		return (EINVAL);
3810 	}
3811 	/*
3812 	 * Common case of OK return with outval same as inval.
3813 	 */
3814 	if (invalp != outvalp) {
3815 		/* don't trust bcopy for identical src/dst */
3816 		(void) bcopy(invalp, outvalp, inlen);
3817 	}
3818 	*outlenp = inlen;
3819 	return (0);
3820 }
3821 
3822 int
3823 udp_opt_set(queue_t *q, uint_t optset_context, int level,
3824     int name, uint_t inlen, uchar_t *invalp, uint_t *outlenp,
3825     uchar_t *outvalp, void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3826 {
3827 	udp_t	*udp;
3828 	int	err;
3829 
3830 	udp = Q_TO_UDP(q);
3831 
3832 	rw_enter(&udp->udp_rwlock, RW_WRITER);
3833 	err = udp_opt_set_locked(q, optset_context, level, name, inlen, invalp,
3834 	    outlenp, outvalp, thisdg_attrs, cr, mblk);
3835 	rw_exit(&udp->udp_rwlock);
3836 	return (err);
3837 }
3838 
3839 /*
3840  * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl.
3841  * The headers include ip6i_t (if needed), ip6_t, any sticky extension
3842  * headers, and the udp header.
3843  * Returns failure if can't allocate memory.
3844  */
3845 static int
3846 udp_build_hdrs(udp_t *udp)
3847 {
3848 	udp_stack_t *us = udp->udp_us;
3849 	uchar_t	*hdrs;
3850 	uint_t	hdrs_len;
3851 	ip6_t	*ip6h;
3852 	ip6i_t	*ip6i;
3853 	udpha_t	*udpha;
3854 	ip6_pkt_t *ipp = &udp->udp_sticky_ipp;
3855 	size_t	sth_wroff;
3856 
3857 	ASSERT(RW_WRITE_HELD(&udp->udp_rwlock));
3858 	hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE;
3859 	ASSERT(hdrs_len != 0);
3860 	if (hdrs_len != udp->udp_sticky_hdrs_len) {
3861 		/* Need to reallocate */
3862 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3863 		if (hdrs == NULL)
3864 			return (ENOMEM);
3865 
3866 		if (udp->udp_sticky_hdrs_len != 0) {
3867 			kmem_free(udp->udp_sticky_hdrs,
3868 			    udp->udp_sticky_hdrs_len);
3869 		}
3870 		udp->udp_sticky_hdrs = hdrs;
3871 		udp->udp_sticky_hdrs_len = hdrs_len;
3872 	}
3873 	ip_build_hdrs_v6(udp->udp_sticky_hdrs,
3874 	    udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP);
3875 
3876 	/* Set header fields not in ipp */
3877 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3878 		ip6i = (ip6i_t *)udp->udp_sticky_hdrs;
3879 		ip6h = (ip6_t *)&ip6i[1];
3880 	} else {
3881 		ip6h = (ip6_t *)udp->udp_sticky_hdrs;
3882 	}
3883 
3884 	if (!(ipp->ipp_fields & IPPF_ADDR))
3885 		ip6h->ip6_src = udp->udp_v6src;
3886 
3887 	udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE);
3888 	udpha->uha_src_port = udp->udp_port;
3889 
3890 	/* Try to get everything in a single mblk */
3891 	if (hdrs_len > udp->udp_max_hdr_len) {
3892 		udp->udp_max_hdr_len = hdrs_len;
3893 		sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
3894 		rw_exit(&udp->udp_rwlock);
3895 		(void) mi_set_sth_wroff(udp->udp_connp->conn_rq, sth_wroff);
3896 		rw_enter(&udp->udp_rwlock, RW_WRITER);
3897 	}
3898 	return (0);
3899 }
3900 
3901 /*
3902  * This routine retrieves the value of an ND variable in a udpparam_t
3903  * structure.  It is called through nd_getset when a user reads the
3904  * variable.
3905  */
3906 /* ARGSUSED */
3907 static int
3908 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3909 {
3910 	udpparam_t *udppa = (udpparam_t *)cp;
3911 
3912 	(void) mi_mpprintf(mp, "%d", udppa->udp_param_value);
3913 	return (0);
3914 }
3915 
3916 /*
3917  * Walk through the param array specified registering each element with the
3918  * named dispatch (ND) handler.
3919  */
3920 static boolean_t
3921 udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt)
3922 {
3923 	for (; cnt-- > 0; udppa++) {
3924 		if (udppa->udp_param_name && udppa->udp_param_name[0]) {
3925 			if (!nd_load(ndp, udppa->udp_param_name,
3926 			    udp_param_get, udp_param_set,
3927 			    (caddr_t)udppa)) {
3928 				nd_free(ndp);
3929 				return (B_FALSE);
3930 			}
3931 		}
3932 	}
3933 	if (!nd_load(ndp, "udp_extra_priv_ports",
3934 	    udp_extra_priv_ports_get, NULL, NULL)) {
3935 		nd_free(ndp);
3936 		return (B_FALSE);
3937 	}
3938 	if (!nd_load(ndp, "udp_extra_priv_ports_add",
3939 	    NULL, udp_extra_priv_ports_add, NULL)) {
3940 		nd_free(ndp);
3941 		return (B_FALSE);
3942 	}
3943 	if (!nd_load(ndp, "udp_extra_priv_ports_del",
3944 	    NULL, udp_extra_priv_ports_del, NULL)) {
3945 		nd_free(ndp);
3946 		return (B_FALSE);
3947 	}
3948 	if (!nd_load(ndp, "udp_status", udp_status_report, NULL,
3949 	    NULL)) {
3950 		nd_free(ndp);
3951 		return (B_FALSE);
3952 	}
3953 	if (!nd_load(ndp, "udp_bind_hash", udp_bind_hash_report, NULL,
3954 	    NULL)) {
3955 		nd_free(ndp);
3956 		return (B_FALSE);
3957 	}
3958 	return (B_TRUE);
3959 }
3960 
3961 /* This routine sets an ND variable in a udpparam_t structure. */
3962 /* ARGSUSED */
3963 static int
3964 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3965 {
3966 	long		new_value;
3967 	udpparam_t	*udppa = (udpparam_t *)cp;
3968 
3969 	/*
3970 	 * Fail the request if the new value does not lie within the
3971 	 * required bounds.
3972 	 */
3973 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3974 	    new_value < udppa->udp_param_min ||
3975 	    new_value > udppa->udp_param_max) {
3976 		return (EINVAL);
3977 	}
3978 
3979 	/* Set the new value */
3980 	udppa->udp_param_value = new_value;
3981 	return (0);
3982 }
3983 
3984 /*
3985  * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with
3986  * T_opthdr) and return the number of bytes copied.  'dbuf' may be NULL to
3987  * just count the length needed for allocation.  If 'dbuf' is non-NULL,
3988  * then it's assumed to be allocated to be large enough.
3989  *
3990  * Returns zero if trimming of the security option causes all options to go
3991  * away.
3992  */
3993 static size_t
3994 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf)
3995 {
3996 	struct T_opthdr *toh;
3997 	size_t hol = ipp->ipp_hopoptslen;
3998 	ip6_hbh_t *dstopt = NULL;
3999 	const ip6_hbh_t *srcopt = ipp->ipp_hopopts;
4000 	size_t tlen, olen, plen;
4001 	boolean_t deleting;
4002 	const struct ip6_opt *sopt, *lastpad;
4003 	struct ip6_opt *dopt;
4004 
4005 	if ((toh = (struct T_opthdr *)dbuf) != NULL) {
4006 		toh->level = IPPROTO_IPV6;
4007 		toh->name = IPV6_HOPOPTS;
4008 		toh->status = 0;
4009 		dstopt = (ip6_hbh_t *)(toh + 1);
4010 	}
4011 
4012 	/*
4013 	 * If labeling is enabled, then skip the label option
4014 	 * but get other options if there are any.
4015 	 */
4016 	if (is_system_labeled()) {
4017 		dopt = NULL;
4018 		if (dstopt != NULL) {
4019 			/* will fill in ip6h_len later */
4020 			dstopt->ip6h_nxt = srcopt->ip6h_nxt;
4021 			dopt = (struct ip6_opt *)(dstopt + 1);
4022 		}
4023 		sopt = (const struct ip6_opt *)(srcopt + 1);
4024 		hol -= sizeof (*srcopt);
4025 		tlen = sizeof (*dstopt);
4026 		lastpad = NULL;
4027 		deleting = B_FALSE;
4028 		/*
4029 		 * This loop finds the first (lastpad pointer) of any number of
4030 		 * pads that preceeds the security option, then treats the
4031 		 * security option as though it were a pad, and then finds the
4032 		 * next non-pad option (or end of list).
4033 		 *
4034 		 * It then treats the entire block as one big pad.  To preserve
4035 		 * alignment of any options that follow, or just the end of the
4036 		 * list, it computes a minimal new padding size that keeps the
4037 		 * same alignment for the next option.
4038 		 *
4039 		 * If it encounters just a sequence of pads with no security
4040 		 * option, those are copied as-is rather than collapsed.
4041 		 *
4042 		 * Note that to handle the end of list case, the code makes one
4043 		 * loop with 'hol' set to zero.
4044 		 */
4045 		for (;;) {
4046 			if (hol > 0) {
4047 				if (sopt->ip6o_type == IP6OPT_PAD1) {
4048 					if (lastpad == NULL)
4049 						lastpad = sopt;
4050 					sopt = (const struct ip6_opt *)
4051 					    &sopt->ip6o_len;
4052 					hol--;
4053 					continue;
4054 				}
4055 				olen = sopt->ip6o_len + sizeof (*sopt);
4056 				if (olen > hol)
4057 					olen = hol;
4058 				if (sopt->ip6o_type == IP6OPT_PADN ||
4059 				    sopt->ip6o_type == ip6opt_ls) {
4060 					if (sopt->ip6o_type == ip6opt_ls)
4061 						deleting = B_TRUE;
4062 					if (lastpad == NULL)
4063 						lastpad = sopt;
4064 					sopt = (const struct ip6_opt *)
4065 					    ((const char *)sopt + olen);
4066 					hol -= olen;
4067 					continue;
4068 				}
4069 			} else {
4070 				/* if nothing was copied at all, then delete */
4071 				if (tlen == sizeof (*dstopt))
4072 					return (0);
4073 				/* last pass; pick up any trailing padding */
4074 				olen = 0;
4075 			}
4076 			if (deleting) {
4077 				/*
4078 				 * compute aligning effect of deleted material
4079 				 * to reproduce with pad.
4080 				 */
4081 				plen = ((const char *)sopt -
4082 				    (const char *)lastpad) & 7;
4083 				tlen += plen;
4084 				if (dopt != NULL) {
4085 					if (plen == 1) {
4086 						dopt->ip6o_type = IP6OPT_PAD1;
4087 					} else if (plen > 1) {
4088 						plen -= sizeof (*dopt);
4089 						dopt->ip6o_type = IP6OPT_PADN;
4090 						dopt->ip6o_len = plen;
4091 						if (plen > 0)
4092 							bzero(dopt + 1, plen);
4093 					}
4094 					dopt = (struct ip6_opt *)
4095 					    ((char *)dopt + plen);
4096 				}
4097 				deleting = B_FALSE;
4098 				lastpad = NULL;
4099 			}
4100 			/* if there's uncopied padding, then copy that now */
4101 			if (lastpad != NULL) {
4102 				olen += (const char *)sopt -
4103 				    (const char *)lastpad;
4104 				sopt = lastpad;
4105 				lastpad = NULL;
4106 			}
4107 			if (dopt != NULL && olen > 0) {
4108 				bcopy(sopt, dopt, olen);
4109 				dopt = (struct ip6_opt *)((char *)dopt + olen);
4110 			}
4111 			if (hol == 0)
4112 				break;
4113 			tlen += olen;
4114 			sopt = (const struct ip6_opt *)
4115 			    ((const char *)sopt + olen);
4116 			hol -= olen;
4117 		}
4118 		/* go back and patch up the length value, rounded upward */
4119 		if (dstopt != NULL)
4120 			dstopt->ip6h_len = (tlen - 1) >> 3;
4121 	} else {
4122 		tlen = hol;
4123 		if (dstopt != NULL)
4124 			bcopy(srcopt, dstopt, hol);
4125 	}
4126 
4127 	tlen += sizeof (*toh);
4128 	if (toh != NULL)
4129 		toh->len = tlen;
4130 
4131 	return (tlen);
4132 }
4133 
4134 /*
4135  * Update udp_rcv_opt_len from the packet.
4136  * Called when options received, and when no options received but
4137  * udp_ip_recv_opt_len has previously recorded options.
4138  */
4139 static void
4140 udp_save_ip_rcv_opt(udp_t *udp, void *opt, int opt_len)
4141 {
4142 	/* Save the options if any */
4143 	if (opt_len > 0) {
4144 		if (opt_len > udp->udp_ip_rcv_options_len) {
4145 			/* Need to allocate larger buffer */
4146 			if (udp->udp_ip_rcv_options_len != 0)
4147 				mi_free((char *)udp->udp_ip_rcv_options);
4148 			udp->udp_ip_rcv_options_len = 0;
4149 			udp->udp_ip_rcv_options =
4150 			    (uchar_t *)mi_alloc(opt_len, BPRI_HI);
4151 			if (udp->udp_ip_rcv_options != NULL)
4152 				udp->udp_ip_rcv_options_len = opt_len;
4153 		}
4154 		if (udp->udp_ip_rcv_options_len != 0) {
4155 			bcopy(opt, udp->udp_ip_rcv_options, opt_len);
4156 			/* Adjust length if we are resusing the space */
4157 			udp->udp_ip_rcv_options_len = opt_len;
4158 		}
4159 	} else if (udp->udp_ip_rcv_options_len != 0) {
4160 		/* Clear out previously recorded options */
4161 		mi_free((char *)udp->udp_ip_rcv_options);
4162 		udp->udp_ip_rcv_options = NULL;
4163 		udp->udp_ip_rcv_options_len = 0;
4164 	}
4165 }
4166 
4167 /* ARGSUSED2 */
4168 static void
4169 udp_input(void *arg1, mblk_t *mp, void *arg2)
4170 {
4171 	conn_t *connp = (conn_t *)arg1;
4172 	struct T_unitdata_ind	*tudi;
4173 	uchar_t			*rptr;		/* Pointer to IP header */
4174 	int			hdr_length;	/* Length of IP+UDP headers */
4175 	int			opt_len;
4176 	int			udi_size;	/* Size of T_unitdata_ind */
4177 	int			mp_len;
4178 	udp_t			*udp;
4179 	udpha_t			*udpha;
4180 	int			ipversion;
4181 	ip6_pkt_t		ipp;
4182 	ip6_t			*ip6h;
4183 	ip6i_t			*ip6i;
4184 	mblk_t			*mp1;
4185 	mblk_t			*options_mp = NULL;
4186 	ip_pktinfo_t		*pinfo = NULL;
4187 	cred_t			*cr = NULL;
4188 	pid_t			cpid;
4189 	uint32_t		udp_ip_rcv_options_len;
4190 	udp_bits_t		udp_bits;
4191 	cred_t			*rcr = connp->conn_cred;
4192 	udp_stack_t *us;
4193 
4194 	ASSERT(connp->conn_flags & IPCL_UDPCONN);
4195 
4196 	udp = connp->conn_udp;
4197 	us = udp->udp_us;
4198 	rptr = mp->b_rptr;
4199 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
4200 	ASSERT(OK_32PTR(rptr));
4201 
4202 	/*
4203 	 * IP should have prepended the options data in an M_CTL
4204 	 * Check M_CTL "type" to make sure are not here bcos of
4205 	 * a valid ICMP message
4206 	 */
4207 	if (DB_TYPE(mp) == M_CTL) {
4208 		if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
4209 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
4210 		    IN_PKTINFO) {
4211 			/*
4212 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
4213 			 * has been prepended to the packet by IP. We need to
4214 			 * extract the mblk and adjust the rptr
4215 			 */
4216 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
4217 			options_mp = mp;
4218 			mp = mp->b_cont;
4219 			rptr = mp->b_rptr;
4220 			UDP_STAT(us, udp_in_pktinfo);
4221 		} else {
4222 			/*
4223 			 * ICMP messages.
4224 			 */
4225 			udp_icmp_error(connp->conn_rq, mp);
4226 			return;
4227 		}
4228 	}
4229 
4230 	mp_len = msgdsize(mp);
4231 	/*
4232 	 * This is the inbound data path.
4233 	 * First, we check to make sure the IP version number is correct,
4234 	 * and then pull the IP and UDP headers into the first mblk.
4235 	 */
4236 
4237 	/* Initialize regardless if ipversion is IPv4 or IPv6 */
4238 	ipp.ipp_fields = 0;
4239 
4240 	ipversion = IPH_HDR_VERSION(rptr);
4241 
4242 	rw_enter(&udp->udp_rwlock, RW_READER);
4243 	udp_ip_rcv_options_len = udp->udp_ip_rcv_options_len;
4244 	udp_bits = udp->udp_bits;
4245 	rw_exit(&udp->udp_rwlock);
4246 
4247 	switch (ipversion) {
4248 	case IPV4_VERSION:
4249 		ASSERT(MBLKL(mp) >= sizeof (ipha_t));
4250 		ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP);
4251 		hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
4252 		opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE);
4253 		if ((opt_len > 0 || udp_ip_rcv_options_len > 0) &&
4254 		    udp->udp_family == AF_INET) {
4255 			/*
4256 			 * Record/update udp_ip_rcv_options with the lock
4257 			 * held. Not needed for AF_INET6 sockets
4258 			 * since they don't support a getsockopt of IP_OPTIONS.
4259 			 */
4260 			rw_enter(&udp->udp_rwlock, RW_WRITER);
4261 			udp_save_ip_rcv_opt(udp, rptr + IP_SIMPLE_HDR_LENGTH,
4262 			    opt_len);
4263 			rw_exit(&udp->udp_rwlock);
4264 		}
4265 		/* Handle IPV6_RECVPKTINFO even for IPv4 packet. */
4266 		if ((udp->udp_family == AF_INET6) && (pinfo != NULL) &&
4267 		    udp->udp_ip_recvpktinfo) {
4268 			if (pinfo->ip_pkt_flags & IPF_RECVIF) {
4269 				ipp.ipp_fields |= IPPF_IFINDEX;
4270 				ipp.ipp_ifindex = pinfo->ip_pkt_ifindex;
4271 			}
4272 		}
4273 		break;
4274 	case IPV6_VERSION:
4275 		/*
4276 		 * IPv6 packets can only be received by applications
4277 		 * that are prepared to receive IPv6 addresses.
4278 		 * The IP fanout must ensure this.
4279 		 */
4280 		ASSERT(udp->udp_family == AF_INET6);
4281 
4282 		ip6h = (ip6_t *)rptr;
4283 		ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr);
4284 
4285 		if (ip6h->ip6_nxt != IPPROTO_UDP) {
4286 			uint8_t nexthdrp;
4287 			/* Look for ifindex information */
4288 			if (ip6h->ip6_nxt == IPPROTO_RAW) {
4289 				ip6i = (ip6i_t *)ip6h;
4290 				if ((uchar_t *)&ip6i[1] > mp->b_wptr)
4291 					goto tossit;
4292 
4293 				if (ip6i->ip6i_flags & IP6I_IFINDEX) {
4294 					ASSERT(ip6i->ip6i_ifindex != 0);
4295 					ipp.ipp_fields |= IPPF_IFINDEX;
4296 					ipp.ipp_ifindex = ip6i->ip6i_ifindex;
4297 				}
4298 				rptr = (uchar_t *)&ip6i[1];
4299 				mp->b_rptr = rptr;
4300 				if (rptr == mp->b_wptr) {
4301 					mp1 = mp->b_cont;
4302 					freeb(mp);
4303 					mp = mp1;
4304 					rptr = mp->b_rptr;
4305 				}
4306 				if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE))
4307 					goto tossit;
4308 				ip6h = (ip6_t *)rptr;
4309 				mp_len = msgdsize(mp);
4310 			}
4311 			/*
4312 			 * Find any potentially interesting extension headers
4313 			 * as well as the length of the IPv6 + extension
4314 			 * headers.
4315 			 */
4316 			hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) +
4317 			    UDPH_SIZE;
4318 			ASSERT(nexthdrp == IPPROTO_UDP);
4319 		} else {
4320 			hdr_length = IPV6_HDR_LEN + UDPH_SIZE;
4321 			ip6i = NULL;
4322 		}
4323 		break;
4324 	default:
4325 		ASSERT(0);
4326 	}
4327 
4328 	/*
4329 	 * IP inspected the UDP header thus all of it must be in the mblk.
4330 	 * UDP length check is performed for IPv6 packets and IPv4 packets
4331 	 * to check if the size of the packet as specified
4332 	 * by the header is the same as the physical size of the packet.
4333 	 * FIXME? Didn't IP already check this?
4334 	 */
4335 	udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
4336 	if ((MBLKL(mp) < hdr_length) ||
4337 	    (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) {
4338 		goto tossit;
4339 	}
4340 
4341 
4342 	/* Walk past the headers unless IP_RECVHDR was set. */
4343 	if (!udp_bits.udpb_rcvhdr) {
4344 		mp->b_rptr = rptr + hdr_length;
4345 		mp_len -= hdr_length;
4346 	}
4347 
4348 	/*
4349 	 * This is the inbound data path.  Packets are passed upstream as
4350 	 * T_UNITDATA_IND messages with full IP headers still attached.
4351 	 */
4352 	if (udp->udp_family == AF_INET) {
4353 		sin_t *sin;
4354 
4355 		ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
4356 
4357 		/*
4358 		 * Normally only send up the source address.
4359 		 * If IP_RECVDSTADDR is set we include the destination IP
4360 		 * address as an option. With IP_RECVOPTS we include all
4361 		 * the IP options.
4362 		 */
4363 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
4364 		if (udp_bits.udpb_recvdstaddr) {
4365 			udi_size += sizeof (struct T_opthdr) +
4366 			    sizeof (struct in_addr);
4367 			UDP_STAT(us, udp_in_recvdstaddr);
4368 		}
4369 
4370 		if (udp_bits.udpb_ip_recvpktinfo && (pinfo != NULL) &&
4371 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
4372 			udi_size += sizeof (struct T_opthdr) +
4373 			    sizeof (struct in_pktinfo);
4374 			UDP_STAT(us, udp_ip_rcvpktinfo);
4375 		}
4376 
4377 		if ((udp_bits.udpb_recvopts) && opt_len > 0) {
4378 			udi_size += sizeof (struct T_opthdr) + opt_len;
4379 			UDP_STAT(us, udp_in_recvopts);
4380 		}
4381 
4382 		/*
4383 		 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
4384 		 * space accordingly
4385 		 */
4386 		if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
4387 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4388 			udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
4389 			UDP_STAT(us, udp_in_recvif);
4390 		}
4391 
4392 		if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
4393 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
4394 			udi_size += sizeof (struct T_opthdr) +
4395 			    sizeof (struct sockaddr_dl);
4396 			UDP_STAT(us, udp_in_recvslla);
4397 		}
4398 
4399 		if ((udp_bits.udpb_recvucred) &&
4400 		    (cr = DB_CRED(mp)) != NULL) {
4401 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4402 			cpid = DB_CPID(mp);
4403 			UDP_STAT(us, udp_in_recvucred);
4404 		}
4405 
4406 		/* XXX FIXME: apply to AF_INET6 as well */
4407 		/*
4408 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4409 		 * buffer. Since gethrestime() expects a pointer aligned
4410 		 * argument, we allocate space necessary for extra
4411 		 * alignment (even though it might not be used).
4412 		 */
4413 		if (udp_bits.udpb_timestamp) {
4414 			udi_size += sizeof (struct T_opthdr) +
4415 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4416 			UDP_STAT(us, udp_in_timestamp);
4417 		}
4418 
4419 		/*
4420 		 * If IP_RECVTTL is set allocate the appropriate sized buffer
4421 		 */
4422 		if (udp_bits.udpb_recvttl) {
4423 			udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
4424 			UDP_STAT(us, udp_in_recvttl);
4425 		}
4426 
4427 		/* Allocate a message block for the T_UNITDATA_IND structure. */
4428 		mp1 = allocb(udi_size, BPRI_MED);
4429 		if (mp1 == NULL) {
4430 			freemsg(mp);
4431 			if (options_mp != NULL)
4432 				freeb(options_mp);
4433 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
4434 			return;
4435 		}
4436 		mp1->b_cont = mp;
4437 		mp = mp1;
4438 		mp->b_datap->db_type = M_PROTO;
4439 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4440 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4441 		tudi->PRIM_type = T_UNITDATA_IND;
4442 		tudi->SRC_length = sizeof (sin_t);
4443 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4444 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4445 		    sizeof (sin_t);
4446 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
4447 		tudi->OPT_length = udi_size;
4448 		sin = (sin_t *)&tudi[1];
4449 		sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
4450 		sin->sin_port =	udpha->uha_src_port;
4451 		sin->sin_family = udp->udp_family;
4452 		*(uint32_t *)&sin->sin_zero[0] = 0;
4453 		*(uint32_t *)&sin->sin_zero[4] = 0;
4454 
4455 		/*
4456 		 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
4457 		 * IP_RECVTTL has been set.
4458 		 */
4459 		if (udi_size != 0) {
4460 			/*
4461 			 * Copy in destination address before options to avoid
4462 			 * any padding issues.
4463 			 */
4464 			char *dstopt;
4465 
4466 			dstopt = (char *)&sin[1];
4467 			if (udp_bits.udpb_recvdstaddr) {
4468 				struct T_opthdr *toh;
4469 				ipaddr_t *dstptr;
4470 
4471 				toh = (struct T_opthdr *)dstopt;
4472 				toh->level = IPPROTO_IP;
4473 				toh->name = IP_RECVDSTADDR;
4474 				toh->len = sizeof (struct T_opthdr) +
4475 				    sizeof (ipaddr_t);
4476 				toh->status = 0;
4477 				dstopt += sizeof (struct T_opthdr);
4478 				dstptr = (ipaddr_t *)dstopt;
4479 				*dstptr = ((ipha_t *)rptr)->ipha_dst;
4480 				dstopt += sizeof (ipaddr_t);
4481 				udi_size -= toh->len;
4482 			}
4483 
4484 			if (udp_bits.udpb_recvopts && opt_len > 0) {
4485 				struct T_opthdr *toh;
4486 
4487 				toh = (struct T_opthdr *)dstopt;
4488 				toh->level = IPPROTO_IP;
4489 				toh->name = IP_RECVOPTS;
4490 				toh->len = sizeof (struct T_opthdr) + opt_len;
4491 				toh->status = 0;
4492 				dstopt += sizeof (struct T_opthdr);
4493 				bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt,
4494 				    opt_len);
4495 				dstopt += opt_len;
4496 				udi_size -= toh->len;
4497 			}
4498 
4499 			if ((udp_bits.udpb_ip_recvpktinfo) && (pinfo != NULL) &&
4500 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
4501 				struct T_opthdr *toh;
4502 				struct in_pktinfo *pktinfop;
4503 
4504 				toh = (struct T_opthdr *)dstopt;
4505 				toh->level = IPPROTO_IP;
4506 				toh->name = IP_PKTINFO;
4507 				toh->len = sizeof (struct T_opthdr) +
4508 				    sizeof (*pktinfop);
4509 				toh->status = 0;
4510 				dstopt += sizeof (struct T_opthdr);
4511 				pktinfop = (struct in_pktinfo *)dstopt;
4512 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
4513 				pktinfop->ipi_spec_dst =
4514 				    pinfo->ip_pkt_match_addr;
4515 				pktinfop->ipi_addr.s_addr =
4516 				    ((ipha_t *)rptr)->ipha_dst;
4517 
4518 				dstopt += sizeof (struct in_pktinfo);
4519 				udi_size -= toh->len;
4520 			}
4521 
4522 			if ((udp_bits.udpb_recvslla) && (pinfo != NULL) &&
4523 			    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
4524 
4525 				struct T_opthdr *toh;
4526 				struct sockaddr_dl	*dstptr;
4527 
4528 				toh = (struct T_opthdr *)dstopt;
4529 				toh->level = IPPROTO_IP;
4530 				toh->name = IP_RECVSLLA;
4531 				toh->len = sizeof (struct T_opthdr) +
4532 				    sizeof (struct sockaddr_dl);
4533 				toh->status = 0;
4534 				dstopt += sizeof (struct T_opthdr);
4535 				dstptr = (struct sockaddr_dl *)dstopt;
4536 				bcopy(&pinfo->ip_pkt_slla, dstptr,
4537 				    sizeof (struct sockaddr_dl));
4538 				dstopt += sizeof (struct sockaddr_dl);
4539 				udi_size -= toh->len;
4540 			}
4541 
4542 			if ((udp_bits.udpb_recvif) && (pinfo != NULL) &&
4543 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4544 
4545 				struct T_opthdr *toh;
4546 				uint_t		*dstptr;
4547 
4548 				toh = (struct T_opthdr *)dstopt;
4549 				toh->level = IPPROTO_IP;
4550 				toh->name = IP_RECVIF;
4551 				toh->len = sizeof (struct T_opthdr) +
4552 				    sizeof (uint_t);
4553 				toh->status = 0;
4554 				dstopt += sizeof (struct T_opthdr);
4555 				dstptr = (uint_t *)dstopt;
4556 				*dstptr = pinfo->ip_pkt_ifindex;
4557 				dstopt += sizeof (uint_t);
4558 				udi_size -= toh->len;
4559 			}
4560 
4561 			if (cr != NULL) {
4562 				struct T_opthdr *toh;
4563 
4564 				toh = (struct T_opthdr *)dstopt;
4565 				toh->level = SOL_SOCKET;
4566 				toh->name = SCM_UCRED;
4567 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4568 				toh->status = 0;
4569 				dstopt += sizeof (struct T_opthdr);
4570 				(void) cred2ucred(cr, cpid, dstopt, rcr);
4571 				dstopt += ucredsize;
4572 				udi_size -= toh->len;
4573 			}
4574 
4575 			if (udp_bits.udpb_timestamp) {
4576 				struct	T_opthdr *toh;
4577 
4578 				toh = (struct T_opthdr *)dstopt;
4579 				toh->level = SOL_SOCKET;
4580 				toh->name = SCM_TIMESTAMP;
4581 				toh->len = sizeof (struct T_opthdr) +
4582 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4583 				toh->status = 0;
4584 				dstopt += sizeof (struct T_opthdr);
4585 				/* Align for gethrestime() */
4586 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
4587 				    sizeof (intptr_t));
4588 				gethrestime((timestruc_t *)dstopt);
4589 				dstopt = (char *)toh + toh->len;
4590 				udi_size -= toh->len;
4591 			}
4592 
4593 			/*
4594 			 * CAUTION:
4595 			 * Due to aligment issues
4596 			 * Processing of IP_RECVTTL option
4597 			 * should always be the last. Adding
4598 			 * any option processing after this will
4599 			 * cause alignment panic.
4600 			 */
4601 			if (udp_bits.udpb_recvttl) {
4602 				struct	T_opthdr *toh;
4603 				uint8_t	*dstptr;
4604 
4605 				toh = (struct T_opthdr *)dstopt;
4606 				toh->level = IPPROTO_IP;
4607 				toh->name = IP_RECVTTL;
4608 				toh->len = sizeof (struct T_opthdr) +
4609 				    sizeof (uint8_t);
4610 				toh->status = 0;
4611 				dstopt += sizeof (struct T_opthdr);
4612 				dstptr = (uint8_t *)dstopt;
4613 				*dstptr = ((ipha_t *)rptr)->ipha_ttl;
4614 				dstopt += sizeof (uint8_t);
4615 				udi_size -= toh->len;
4616 			}
4617 
4618 			/* Consumed all of allocated space */
4619 			ASSERT(udi_size == 0);
4620 		}
4621 	} else {
4622 		sin6_t *sin6;
4623 
4624 		/*
4625 		 * Handle both IPv4 and IPv6 packets for IPv6 sockets.
4626 		 *
4627 		 * Normally we only send up the address. If receiving of any
4628 		 * optional receive side information is enabled, we also send
4629 		 * that up as options.
4630 		 */
4631 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
4632 
4633 		if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
4634 		    IPPF_RTHDR|IPPF_IFINDEX)) {
4635 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4636 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4637 				size_t hlen;
4638 
4639 				UDP_STAT(us, udp_in_recvhopopts);
4640 				hlen = copy_hop_opts(&ipp, NULL);
4641 				if (hlen == 0)
4642 					ipp.ipp_fields &= ~IPPF_HOPOPTS;
4643 				udi_size += hlen;
4644 			}
4645 			if (((udp_bits.udpb_ipv6_recvdstopts) ||
4646 			    udp_bits.udpb_old_ipv6_recvdstopts) &&
4647 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4648 				udi_size += sizeof (struct T_opthdr) +
4649 				    ipp.ipp_dstoptslen;
4650 				UDP_STAT(us, udp_in_recvdstopts);
4651 			}
4652 			if ((((udp_bits.udpb_ipv6_recvdstopts) &&
4653 			    udp_bits.udpb_ipv6_recvrthdr &&
4654 			    (ipp.ipp_fields & IPPF_RTHDR)) ||
4655 			    (udp_bits.udpb_ipv6_recvrthdrdstopts)) &&
4656 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4657 				udi_size += sizeof (struct T_opthdr) +
4658 				    ipp.ipp_rtdstoptslen;
4659 				UDP_STAT(us, udp_in_recvrtdstopts);
4660 			}
4661 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4662 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4663 				udi_size += sizeof (struct T_opthdr) +
4664 				    ipp.ipp_rthdrlen;
4665 				UDP_STAT(us, udp_in_recvrthdr);
4666 			}
4667 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4668 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4669 				udi_size += sizeof (struct T_opthdr) +
4670 				    sizeof (struct in6_pktinfo);
4671 				UDP_STAT(us, udp_in_recvpktinfo);
4672 			}
4673 
4674 		}
4675 		if ((udp_bits.udpb_recvucred) &&
4676 		    (cr = DB_CRED(mp)) != NULL) {
4677 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4678 			cpid = DB_CPID(mp);
4679 			UDP_STAT(us, udp_in_recvucred);
4680 		}
4681 
4682 		/*
4683 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4684 		 * buffer. Since gethrestime() expects a pointer aligned
4685 		 * argument, we allocate space necessary for extra
4686 		 * alignment (even though it might not be used).
4687 		 */
4688 		if (udp_bits.udpb_timestamp) {
4689 			udi_size += sizeof (struct T_opthdr) +
4690 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4691 			UDP_STAT(us, udp_in_timestamp);
4692 		}
4693 
4694 		if (udp_bits.udpb_ipv6_recvhoplimit) {
4695 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4696 			UDP_STAT(us, udp_in_recvhoplimit);
4697 		}
4698 
4699 		if (udp_bits.udpb_ipv6_recvtclass) {
4700 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
4701 			UDP_STAT(us, udp_in_recvtclass);
4702 		}
4703 
4704 		mp1 = allocb(udi_size, BPRI_MED);
4705 		if (mp1 == NULL) {
4706 			freemsg(mp);
4707 			if (options_mp != NULL)
4708 				freeb(options_mp);
4709 			BUMP_MIB(&us->us_udp_mib, udpInErrors);
4710 			return;
4711 		}
4712 		mp1->b_cont = mp;
4713 		mp = mp1;
4714 		mp->b_datap->db_type = M_PROTO;
4715 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4716 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4717 		tudi->PRIM_type = T_UNITDATA_IND;
4718 		tudi->SRC_length = sizeof (sin6_t);
4719 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4720 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4721 		    sizeof (sin6_t);
4722 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
4723 		tudi->OPT_length = udi_size;
4724 		sin6 = (sin6_t *)&tudi[1];
4725 		if (ipversion == IPV4_VERSION) {
4726 			in6_addr_t v6dst;
4727 
4728 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src,
4729 			    &sin6->sin6_addr);
4730 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst,
4731 			    &v6dst);
4732 			sin6->sin6_flowinfo = 0;
4733 			sin6->sin6_scope_id = 0;
4734 			sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst,
4735 			    connp->conn_zoneid, us->us_netstack);
4736 		} else {
4737 			sin6->sin6_addr = ip6h->ip6_src;
4738 			/* No sin6_flowinfo per API */
4739 			sin6->sin6_flowinfo = 0;
4740 			/* For link-scope source pass up scope id */
4741 			if ((ipp.ipp_fields & IPPF_IFINDEX) &&
4742 			    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
4743 				sin6->sin6_scope_id = ipp.ipp_ifindex;
4744 			else
4745 				sin6->sin6_scope_id = 0;
4746 			sin6->__sin6_src_id = ip_srcid_find_addr(
4747 			    &ip6h->ip6_dst, connp->conn_zoneid,
4748 			    us->us_netstack);
4749 		}
4750 		sin6->sin6_port = udpha->uha_src_port;
4751 		sin6->sin6_family = udp->udp_family;
4752 
4753 		if (udi_size != 0) {
4754 			uchar_t *dstopt;
4755 
4756 			dstopt = (uchar_t *)&sin6[1];
4757 			if ((udp_bits.udpb_ip_recvpktinfo) &&
4758 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
4759 				struct T_opthdr *toh;
4760 				struct in6_pktinfo *pkti;
4761 
4762 				toh = (struct T_opthdr *)dstopt;
4763 				toh->level = IPPROTO_IPV6;
4764 				toh->name = IPV6_PKTINFO;
4765 				toh->len = sizeof (struct T_opthdr) +
4766 				    sizeof (*pkti);
4767 				toh->status = 0;
4768 				dstopt += sizeof (struct T_opthdr);
4769 				pkti = (struct in6_pktinfo *)dstopt;
4770 				if (ipversion == IPV6_VERSION)
4771 					pkti->ipi6_addr = ip6h->ip6_dst;
4772 				else
4773 					IN6_IPADDR_TO_V4MAPPED(
4774 					    ((ipha_t *)rptr)->ipha_dst,
4775 					    &pkti->ipi6_addr);
4776 				pkti->ipi6_ifindex = ipp.ipp_ifindex;
4777 				dstopt += sizeof (*pkti);
4778 				udi_size -= toh->len;
4779 			}
4780 			if (udp_bits.udpb_ipv6_recvhoplimit) {
4781 				struct T_opthdr *toh;
4782 
4783 				toh = (struct T_opthdr *)dstopt;
4784 				toh->level = IPPROTO_IPV6;
4785 				toh->name = IPV6_HOPLIMIT;
4786 				toh->len = sizeof (struct T_opthdr) +
4787 				    sizeof (uint_t);
4788 				toh->status = 0;
4789 				dstopt += sizeof (struct T_opthdr);
4790 				if (ipversion == IPV6_VERSION)
4791 					*(uint_t *)dstopt = ip6h->ip6_hops;
4792 				else
4793 					*(uint_t *)dstopt =
4794 					    ((ipha_t *)rptr)->ipha_ttl;
4795 				dstopt += sizeof (uint_t);
4796 				udi_size -= toh->len;
4797 			}
4798 			if (udp_bits.udpb_ipv6_recvtclass) {
4799 				struct T_opthdr *toh;
4800 
4801 				toh = (struct T_opthdr *)dstopt;
4802 				toh->level = IPPROTO_IPV6;
4803 				toh->name = IPV6_TCLASS;
4804 				toh->len = sizeof (struct T_opthdr) +
4805 				    sizeof (uint_t);
4806 				toh->status = 0;
4807 				dstopt += sizeof (struct T_opthdr);
4808 				if (ipversion == IPV6_VERSION) {
4809 					*(uint_t *)dstopt =
4810 					    IPV6_FLOW_TCLASS(ip6h->ip6_flow);
4811 				} else {
4812 					ipha_t *ipha = (ipha_t *)rptr;
4813 					*(uint_t *)dstopt =
4814 					    ipha->ipha_type_of_service;
4815 				}
4816 				dstopt += sizeof (uint_t);
4817 				udi_size -= toh->len;
4818 			}
4819 			if ((udp_bits.udpb_ipv6_recvhopopts) &&
4820 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
4821 				size_t hlen;
4822 
4823 				hlen = copy_hop_opts(&ipp, dstopt);
4824 				dstopt += hlen;
4825 				udi_size -= hlen;
4826 			}
4827 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4828 			    (udp_bits.udpb_ipv6_recvrthdr) &&
4829 			    (ipp.ipp_fields & IPPF_RTHDR) &&
4830 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
4831 				struct T_opthdr *toh;
4832 
4833 				toh = (struct T_opthdr *)dstopt;
4834 				toh->level = IPPROTO_IPV6;
4835 				toh->name = IPV6_DSTOPTS;
4836 				toh->len = sizeof (struct T_opthdr) +
4837 				    ipp.ipp_rtdstoptslen;
4838 				toh->status = 0;
4839 				dstopt += sizeof (struct T_opthdr);
4840 				bcopy(ipp.ipp_rtdstopts, dstopt,
4841 				    ipp.ipp_rtdstoptslen);
4842 				dstopt += ipp.ipp_rtdstoptslen;
4843 				udi_size -= toh->len;
4844 			}
4845 			if ((udp_bits.udpb_ipv6_recvrthdr) &&
4846 			    (ipp.ipp_fields & IPPF_RTHDR)) {
4847 				struct T_opthdr *toh;
4848 
4849 				toh = (struct T_opthdr *)dstopt;
4850 				toh->level = IPPROTO_IPV6;
4851 				toh->name = IPV6_RTHDR;
4852 				toh->len = sizeof (struct T_opthdr) +
4853 				    ipp.ipp_rthdrlen;
4854 				toh->status = 0;
4855 				dstopt += sizeof (struct T_opthdr);
4856 				bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
4857 				dstopt += ipp.ipp_rthdrlen;
4858 				udi_size -= toh->len;
4859 			}
4860 			if ((udp_bits.udpb_ipv6_recvdstopts) &&
4861 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
4862 				struct T_opthdr *toh;
4863 
4864 				toh = (struct T_opthdr *)dstopt;
4865 				toh->level = IPPROTO_IPV6;
4866 				toh->name = IPV6_DSTOPTS;
4867 				toh->len = sizeof (struct T_opthdr) +
4868 				    ipp.ipp_dstoptslen;
4869 				toh->status = 0;
4870 				dstopt += sizeof (struct T_opthdr);
4871 				bcopy(ipp.ipp_dstopts, dstopt,
4872 				    ipp.ipp_dstoptslen);
4873 				dstopt += ipp.ipp_dstoptslen;
4874 				udi_size -= toh->len;
4875 			}
4876 
4877 			if (cr != NULL) {
4878 				struct T_opthdr *toh;
4879 
4880 				toh = (struct T_opthdr *)dstopt;
4881 				toh->level = SOL_SOCKET;
4882 				toh->name = SCM_UCRED;
4883 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4884 				toh->status = 0;
4885 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
4886 				dstopt += toh->len;
4887 				udi_size -= toh->len;
4888 			}
4889 			if (udp_bits.udpb_timestamp) {
4890 				struct	T_opthdr *toh;
4891 
4892 				toh = (struct T_opthdr *)dstopt;
4893 				toh->level = SOL_SOCKET;
4894 				toh->name = SCM_TIMESTAMP;
4895 				toh->len = sizeof (struct T_opthdr) +
4896 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4897 				toh->status = 0;
4898 				dstopt += sizeof (struct T_opthdr);
4899 				/* Align for gethrestime() */
4900 				dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
4901 				    sizeof (intptr_t));
4902 				gethrestime((timestruc_t *)dstopt);
4903 				dstopt = (uchar_t *)toh + toh->len;
4904 				udi_size -= toh->len;
4905 			}
4906 
4907 			/* Consumed all of allocated space */
4908 			ASSERT(udi_size == 0);
4909 		}
4910 #undef	sin6
4911 		/* No IP_RECVDSTADDR for IPv6. */
4912 	}
4913 
4914 	BUMP_MIB(&us->us_udp_mib, udpHCInDatagrams);
4915 	if (options_mp != NULL)
4916 		freeb(options_mp);
4917 
4918 	if (udp_bits.udpb_direct_sockfs) {
4919 		/*
4920 		 * There is nothing above us except for the stream head;
4921 		 * use the read-side synchronous stream interface in
4922 		 * order to reduce the time spent in interrupt thread.
4923 		 */
4924 		ASSERT(udp->udp_issocket);
4925 		udp_rcv_enqueue(connp->conn_rq, udp, mp, mp_len);
4926 	} else {
4927 		/*
4928 		 * Use regular STREAMS interface to pass data upstream
4929 		 * if this is not a socket endpoint, or if we have
4930 		 * switched over to the slow mode due to sockmod being
4931 		 * popped or a module being pushed on top of us.
4932 		 */
4933 		putnext(connp->conn_rq, mp);
4934 	}
4935 	return;
4936 
4937 tossit:
4938 	freemsg(mp);
4939 	if (options_mp != NULL)
4940 		freeb(options_mp);
4941 	BUMP_MIB(&us->us_udp_mib, udpInErrors);
4942 }
4943 
4944 /*
4945  * Handle the results of a T_BIND_REQ whether deferred by IP or handled
4946  * immediately.
4947  */
4948 static void
4949 udp_bind_result(conn_t *connp, mblk_t *mp)
4950 {
4951 	struct T_error_ack	*tea;
4952 
4953 	switch (mp->b_datap->db_type) {
4954 	case M_PROTO:
4955 	case M_PCPROTO:
4956 		/* M_PROTO messages contain some type of TPI message. */
4957 		ASSERT((uintptr_t)(mp->b_wptr - mp->b_rptr) <=
4958 		    (uintptr_t)INT_MAX);
4959 		if (mp->b_wptr - mp->b_rptr < sizeof (t_scalar_t)) {
4960 			freemsg(mp);
4961 			return;
4962 		}
4963 		tea = (struct T_error_ack *)mp->b_rptr;
4964 
4965 		switch (tea->PRIM_type) {
4966 		case T_ERROR_ACK:
4967 			switch (tea->ERROR_prim) {
4968 			case O_T_BIND_REQ:
4969 			case T_BIND_REQ:
4970 				udp_bind_error(connp, mp);
4971 				return;
4972 			default:
4973 				break;
4974 			}
4975 			ASSERT(0);
4976 			freemsg(mp);
4977 			return;
4978 
4979 		case T_BIND_ACK:
4980 			udp_bind_ack(connp, mp);
4981 			return;
4982 
4983 		default:
4984 			break;
4985 		}
4986 		freemsg(mp);
4987 		return;
4988 	default:
4989 		/* FIXME: other cases? */
4990 		ASSERT(0);
4991 		freemsg(mp);
4992 		return;
4993 	}
4994 }
4995 
4996 /*
4997  * Process a T_BIND_ACK
4998  */
4999 static void
5000 udp_bind_ack(conn_t *connp, mblk_t *mp)
5001 {
5002 	udp_t	*udp = connp->conn_udp;
5003 	mblk_t	*mp1;
5004 	ire_t	*ire;
5005 	struct T_bind_ack *tba;
5006 	uchar_t *addrp;
5007 	ipa_conn_t	*ac;
5008 	ipa6_conn_t	*ac6;
5009 	udp_fanout_t	*udpf;
5010 	udp_stack_t	*us = udp->udp_us;
5011 
5012 	ASSERT(udp->udp_pending_op != -1);
5013 	rw_enter(&udp->udp_rwlock, RW_WRITER);
5014 	/*
5015 	 * If a broadcast/multicast address was bound set
5016 	 * the source address to 0.
5017 	 * This ensures no datagrams with broadcast address
5018 	 * as source address are emitted (which would violate
5019 	 * RFC1122 - Hosts requirements)
5020 	 *
5021 	 * Note that when connecting the returned IRE is
5022 	 * for the destination address and we only perform
5023 	 * the broadcast check for the source address (it
5024 	 * is OK to connect to a broadcast/multicast address.)
5025 	 */
5026 	mp1 = mp->b_cont;
5027 	if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) {
5028 		ire = (ire_t *)mp1->b_rptr;
5029 
5030 		/*
5031 		 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast
5032 		 * local address.
5033 		 */
5034 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
5035 		    us->us_bind_fanout_size)];
5036 		if (ire->ire_type == IRE_BROADCAST &&
5037 		    udp->udp_state != TS_DATA_XFER) {
5038 			ASSERT(udp->udp_pending_op == T_BIND_REQ ||
5039 			    udp->udp_pending_op == O_T_BIND_REQ);
5040 			/* This was just a local bind to a broadcast addr */
5041 			mutex_enter(&udpf->uf_lock);
5042 			V6_SET_ZERO(udp->udp_v6src);
5043 			mutex_exit(&udpf->uf_lock);
5044 			if (udp->udp_family == AF_INET6)
5045 				(void) udp_build_hdrs(udp);
5046 		} else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
5047 			/*
5048 			 * Local address not yet set - pick it from the
5049 			 * T_bind_ack
5050 			 */
5051 			tba = (struct T_bind_ack *)mp->b_rptr;
5052 			addrp = &mp->b_rptr[tba->ADDR_offset];
5053 			switch (udp->udp_family) {
5054 			case AF_INET:
5055 				if (tba->ADDR_length == sizeof (ipa_conn_t)) {
5056 					ac = (ipa_conn_t *)addrp;
5057 				} else {
5058 					ASSERT(tba->ADDR_length ==
5059 					    sizeof (ipa_conn_x_t));
5060 					ac = &((ipa_conn_x_t *)addrp)->acx_conn;
5061 				}
5062 				mutex_enter(&udpf->uf_lock);
5063 				IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr,
5064 				    &udp->udp_v6src);
5065 				mutex_exit(&udpf->uf_lock);
5066 				break;
5067 			case AF_INET6:
5068 				if (tba->ADDR_length == sizeof (ipa6_conn_t)) {
5069 					ac6 = (ipa6_conn_t *)addrp;
5070 				} else {
5071 					ASSERT(tba->ADDR_length ==
5072 					    sizeof (ipa6_conn_x_t));
5073 					ac6 = &((ipa6_conn_x_t *)
5074 					    addrp)->ac6x_conn;
5075 				}
5076 				mutex_enter(&udpf->uf_lock);
5077 				udp->udp_v6src = ac6->ac6_laddr;
5078 				mutex_exit(&udpf->uf_lock);
5079 				(void) udp_build_hdrs(udp);
5080 				break;
5081 			}
5082 		}
5083 		mp1 = mp1->b_cont;
5084 	}
5085 	udp->udp_pending_op = -1;
5086 	rw_exit(&udp->udp_rwlock);
5087 	/*
5088 	 * Look for one or more appended ACK message added by
5089 	 * udp_connect or udp_disconnect.
5090 	 * If none found just send up the T_BIND_ACK.
5091 	 * udp_connect has appended a T_OK_ACK and a T_CONN_CON.
5092 	 * udp_disconnect has appended a T_OK_ACK.
5093 	 */
5094 	if (mp1 != NULL) {
5095 		if (mp->b_cont == mp1)
5096 			mp->b_cont = NULL;
5097 		else {
5098 			ASSERT(mp->b_cont->b_cont == mp1);
5099 			mp->b_cont->b_cont = NULL;
5100 		}
5101 		freemsg(mp);
5102 		mp = mp1;
5103 		while (mp != NULL) {
5104 			mp1 = mp->b_cont;
5105 			mp->b_cont = NULL;
5106 			putnext(connp->conn_rq, mp);
5107 			mp = mp1;
5108 		}
5109 		return;
5110 	}
5111 	freemsg(mp->b_cont);
5112 	mp->b_cont = NULL;
5113 	putnext(connp->conn_rq, mp);
5114 }
5115 
5116 static void
5117 udp_bind_error(conn_t *connp, mblk_t *mp)
5118 {
5119 	udp_t	*udp = connp->conn_udp;
5120 	struct T_error_ack *tea;
5121 	udp_fanout_t	*udpf;
5122 	udp_stack_t	*us = udp->udp_us;
5123 
5124 	tea = (struct T_error_ack *)mp->b_rptr;
5125 
5126 	/*
5127 	 * If our O_T_BIND_REQ/T_BIND_REQ fails,
5128 	 * clear out the associated port and source
5129 	 * address before passing the message
5130 	 * upstream. If this was caused by a T_CONN_REQ
5131 	 * revert back to bound state.
5132 	 */
5133 
5134 	rw_enter(&udp->udp_rwlock, RW_WRITER);
5135 	ASSERT(udp->udp_pending_op != -1);
5136 	tea->ERROR_prim = udp->udp_pending_op;
5137 	udp->udp_pending_op = -1;
5138 	udpf = &us->us_bind_fanout[
5139 	    UDP_BIND_HASH(udp->udp_port,
5140 	    us->us_bind_fanout_size)];
5141 	mutex_enter(&udpf->uf_lock);
5142 
5143 	switch (tea->ERROR_prim) {
5144 	case T_CONN_REQ:
5145 		ASSERT(udp->udp_state == TS_DATA_XFER);
5146 		/* Connect failed */
5147 		/* Revert back to the bound source */
5148 		udp->udp_v6src = udp->udp_bound_v6src;
5149 		udp->udp_state = TS_IDLE;
5150 		mutex_exit(&udpf->uf_lock);
5151 		if (udp->udp_family == AF_INET6)
5152 			(void) udp_build_hdrs(udp);
5153 		rw_exit(&udp->udp_rwlock);
5154 		break;
5155 
5156 	case T_DISCON_REQ:
5157 	case T_BIND_REQ:
5158 	case O_T_BIND_REQ:
5159 		V6_SET_ZERO(udp->udp_v6src);
5160 		V6_SET_ZERO(udp->udp_bound_v6src);
5161 		udp->udp_state = TS_UNBND;
5162 		udp_bind_hash_remove(udp, B_TRUE);
5163 		udp->udp_port = 0;
5164 		mutex_exit(&udpf->uf_lock);
5165 		if (udp->udp_family == AF_INET6)
5166 			(void) udp_build_hdrs(udp);
5167 		rw_exit(&udp->udp_rwlock);
5168 		break;
5169 
5170 	default:
5171 		mutex_exit(&udpf->uf_lock);
5172 		rw_exit(&udp->udp_rwlock);
5173 		(void) mi_strlog(connp->conn_rq, 1,
5174 		    SL_ERROR|SL_TRACE,
5175 		    "udp_input_other: bad ERROR_prim, "
5176 		    "len %d", tea->ERROR_prim);
5177 	}
5178 	putnext(connp->conn_rq, mp);
5179 }
5180 
5181 /*
5182  * return SNMP stuff in buffer in mpdata. We don't hold any lock and report
5183  * information that can be changing beneath us.
5184  */
5185 mblk_t *
5186 udp_snmp_get(queue_t *q, mblk_t *mpctl)
5187 {
5188 	mblk_t			*mpdata;
5189 	mblk_t			*mp_conn_ctl;
5190 	mblk_t			*mp_attr_ctl;
5191 	mblk_t			*mp6_conn_ctl;
5192 	mblk_t			*mp6_attr_ctl;
5193 	mblk_t			*mp_conn_tail;
5194 	mblk_t			*mp_attr_tail;
5195 	mblk_t			*mp6_conn_tail;
5196 	mblk_t			*mp6_attr_tail;
5197 	struct opthdr		*optp;
5198 	mib2_udpEntry_t		ude;
5199 	mib2_udp6Entry_t	ude6;
5200 	mib2_transportMLPEntry_t mlp;
5201 	int			state;
5202 	zoneid_t		zoneid;
5203 	int			i;
5204 	connf_t			*connfp;
5205 	conn_t			*connp = Q_TO_CONN(q);
5206 	int			v4_conn_idx;
5207 	int			v6_conn_idx;
5208 	boolean_t		needattr;
5209 	udp_t			*udp;
5210 	ip_stack_t		*ipst = connp->conn_netstack->netstack_ip;
5211 	udp_stack_t		*us = connp->conn_netstack->netstack_udp;
5212 	mblk_t			*mp2ctl;
5213 
5214 	/*
5215 	 * make a copy of the original message
5216 	 */
5217 	mp2ctl = copymsg(mpctl);
5218 
5219 	mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL;
5220 	if (mpctl == NULL ||
5221 	    (mpdata = mpctl->b_cont) == NULL ||
5222 	    (mp_conn_ctl = copymsg(mpctl)) == NULL ||
5223 	    (mp_attr_ctl = copymsg(mpctl)) == NULL ||
5224 	    (mp6_conn_ctl = copymsg(mpctl)) == NULL ||
5225 	    (mp6_attr_ctl = copymsg(mpctl)) == NULL) {
5226 		freemsg(mp_conn_ctl);
5227 		freemsg(mp_attr_ctl);
5228 		freemsg(mp6_conn_ctl);
5229 		freemsg(mpctl);
5230 		freemsg(mp2ctl);
5231 		return (0);
5232 	}
5233 
5234 	zoneid = connp->conn_zoneid;
5235 
5236 	/* fixed length structure for IPv4 and IPv6 counters */
5237 	SET_MIB(us->us_udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t));
5238 	SET_MIB(us->us_udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t));
5239 	/* synchronize 64- and 32-bit counters */
5240 	SYNC32_MIB(&us->us_udp_mib, udpInDatagrams, udpHCInDatagrams);
5241 	SYNC32_MIB(&us->us_udp_mib, udpOutDatagrams, udpHCOutDatagrams);
5242 
5243 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
5244 	optp->level = MIB2_UDP;
5245 	optp->name = 0;
5246 	(void) snmp_append_data(mpdata, (char *)&us->us_udp_mib,
5247 	    sizeof (us->us_udp_mib));
5248 	optp->len = msgdsize(mpdata);
5249 	qreply(q, mpctl);
5250 
5251 	mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL;
5252 	v4_conn_idx = v6_conn_idx = 0;
5253 
5254 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
5255 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
5256 		connp = NULL;
5257 
5258 		while ((connp = ipcl_get_next_conn(connfp, connp,
5259 		    IPCL_UDPCONN))) {
5260 			udp = connp->conn_udp;
5261 			if (zoneid != connp->conn_zoneid)
5262 				continue;
5263 
5264 			/*
5265 			 * Note that the port numbers are sent in
5266 			 * host byte order
5267 			 */
5268 
5269 			if (udp->udp_state == TS_UNBND)
5270 				state = MIB2_UDP_unbound;
5271 			else if (udp->udp_state == TS_IDLE)
5272 				state = MIB2_UDP_idle;
5273 			else if (udp->udp_state == TS_DATA_XFER)
5274 				state = MIB2_UDP_connected;
5275 			else
5276 				state = MIB2_UDP_unknown;
5277 
5278 			needattr = B_FALSE;
5279 			bzero(&mlp, sizeof (mlp));
5280 			if (connp->conn_mlp_type != mlptSingle) {
5281 				if (connp->conn_mlp_type == mlptShared ||
5282 				    connp->conn_mlp_type == mlptBoth)
5283 					mlp.tme_flags |= MIB2_TMEF_SHARED;
5284 				if (connp->conn_mlp_type == mlptPrivate ||
5285 				    connp->conn_mlp_type == mlptBoth)
5286 					mlp.tme_flags |= MIB2_TMEF_PRIVATE;
5287 				needattr = B_TRUE;
5288 			}
5289 
5290 			/*
5291 			 * Create an IPv4 table entry for IPv4 entries and also
5292 			 * any IPv6 entries which are bound to in6addr_any
5293 			 * (i.e. anything a IPv4 peer could connect/send to).
5294 			 */
5295 			if (udp->udp_ipversion == IPV4_VERSION ||
5296 			    (udp->udp_state <= TS_IDLE &&
5297 			    IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) {
5298 				ude.udpEntryInfo.ue_state = state;
5299 				/*
5300 				 * If in6addr_any this will set it to
5301 				 * INADDR_ANY
5302 				 */
5303 				ude.udpLocalAddress =
5304 				    V4_PART_OF_V6(udp->udp_v6src);
5305 				ude.udpLocalPort = ntohs(udp->udp_port);
5306 				if (udp->udp_state == TS_DATA_XFER) {
5307 					/*
5308 					 * Can potentially get here for
5309 					 * v6 socket if another process
5310 					 * (say, ping) has just done a
5311 					 * sendto(), changing the state
5312 					 * from the TS_IDLE above to
5313 					 * TS_DATA_XFER by the time we hit
5314 					 * this part of the code.
5315 					 */
5316 					ude.udpEntryInfo.ue_RemoteAddress =
5317 					    V4_PART_OF_V6(udp->udp_v6dst);
5318 					ude.udpEntryInfo.ue_RemotePort =
5319 					    ntohs(udp->udp_dstport);
5320 				} else {
5321 					ude.udpEntryInfo.ue_RemoteAddress = 0;
5322 					ude.udpEntryInfo.ue_RemotePort = 0;
5323 				}
5324 
5325 				/*
5326 				 * We make the assumption that all udp_t
5327 				 * structs will be created within an address
5328 				 * region no larger than 32-bits.
5329 				 */
5330 				ude.udpInstance = (uint32_t)(uintptr_t)udp;
5331 				ude.udpCreationProcess =
5332 				    (udp->udp_open_pid < 0) ?
5333 				    MIB2_UNKNOWN_PROCESS :
5334 				    udp->udp_open_pid;
5335 				ude.udpCreationTime = udp->udp_open_time;
5336 
5337 				(void) snmp_append_data2(mp_conn_ctl->b_cont,
5338 				    &mp_conn_tail, (char *)&ude, sizeof (ude));
5339 				mlp.tme_connidx = v4_conn_idx++;
5340 				if (needattr)
5341 					(void) snmp_append_data2(
5342 					    mp_attr_ctl->b_cont, &mp_attr_tail,
5343 					    (char *)&mlp, sizeof (mlp));
5344 			}
5345 			if (udp->udp_ipversion == IPV6_VERSION) {
5346 				ude6.udp6EntryInfo.ue_state  = state;
5347 				ude6.udp6LocalAddress = udp->udp_v6src;
5348 				ude6.udp6LocalPort = ntohs(udp->udp_port);
5349 				ude6.udp6IfIndex = udp->udp_bound_if;
5350 				if (udp->udp_state == TS_DATA_XFER) {
5351 					ude6.udp6EntryInfo.ue_RemoteAddress =
5352 					    udp->udp_v6dst;
5353 					ude6.udp6EntryInfo.ue_RemotePort =
5354 					    ntohs(udp->udp_dstport);
5355 				} else {
5356 					ude6.udp6EntryInfo.ue_RemoteAddress =
5357 					    sin6_null.sin6_addr;
5358 					ude6.udp6EntryInfo.ue_RemotePort = 0;
5359 				}
5360 				/*
5361 				 * We make the assumption that all udp_t
5362 				 * structs will be created within an address
5363 				 * region no larger than 32-bits.
5364 				 */
5365 				ude6.udp6Instance = (uint32_t)(uintptr_t)udp;
5366 				ude6.udp6CreationProcess =
5367 				    (udp->udp_open_pid < 0) ?
5368 				    MIB2_UNKNOWN_PROCESS :
5369 				    udp->udp_open_pid;
5370 				ude6.udp6CreationTime = udp->udp_open_time;
5371 
5372 				(void) snmp_append_data2(mp6_conn_ctl->b_cont,
5373 				    &mp6_conn_tail, (char *)&ude6,
5374 				    sizeof (ude6));
5375 				mlp.tme_connidx = v6_conn_idx++;
5376 				if (needattr)
5377 					(void) snmp_append_data2(
5378 					    mp6_attr_ctl->b_cont,
5379 					    &mp6_attr_tail, (char *)&mlp,
5380 					    sizeof (mlp));
5381 			}
5382 		}
5383 	}
5384 
5385 	/* IPv4 UDP endpoints */
5386 	optp = (struct opthdr *)&mp_conn_ctl->b_rptr[
5387 	    sizeof (struct T_optmgmt_ack)];
5388 	optp->level = MIB2_UDP;
5389 	optp->name = MIB2_UDP_ENTRY;
5390 	optp->len = msgdsize(mp_conn_ctl->b_cont);
5391 	qreply(q, mp_conn_ctl);
5392 
5393 	/* table of MLP attributes... */
5394 	optp = (struct opthdr *)&mp_attr_ctl->b_rptr[
5395 	    sizeof (struct T_optmgmt_ack)];
5396 	optp->level = MIB2_UDP;
5397 	optp->name = EXPER_XPORT_MLP;
5398 	optp->len = msgdsize(mp_attr_ctl->b_cont);
5399 	if (optp->len == 0)
5400 		freemsg(mp_attr_ctl);
5401 	else
5402 		qreply(q, mp_attr_ctl);
5403 
5404 	/* IPv6 UDP endpoints */
5405 	optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[
5406 	    sizeof (struct T_optmgmt_ack)];
5407 	optp->level = MIB2_UDP6;
5408 	optp->name = MIB2_UDP6_ENTRY;
5409 	optp->len = msgdsize(mp6_conn_ctl->b_cont);
5410 	qreply(q, mp6_conn_ctl);
5411 
5412 	/* table of MLP attributes... */
5413 	optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[
5414 	    sizeof (struct T_optmgmt_ack)];
5415 	optp->level = MIB2_UDP6;
5416 	optp->name = EXPER_XPORT_MLP;
5417 	optp->len = msgdsize(mp6_attr_ctl->b_cont);
5418 	if (optp->len == 0)
5419 		freemsg(mp6_attr_ctl);
5420 	else
5421 		qreply(q, mp6_attr_ctl);
5422 
5423 	return (mp2ctl);
5424 }
5425 
5426 /*
5427  * Return 0 if invalid set request, 1 otherwise, including non-udp requests.
5428  * NOTE: Per MIB-II, UDP has no writable data.
5429  * TODO:  If this ever actually tries to set anything, it needs to be
5430  * to do the appropriate locking.
5431  */
5432 /* ARGSUSED */
5433 int
5434 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
5435     uchar_t *ptr, int len)
5436 {
5437 	switch (level) {
5438 	case MIB2_UDP:
5439 		return (0);
5440 	default:
5441 		return (1);
5442 	}
5443 }
5444 
5445 static void
5446 udp_report_item(mblk_t *mp, udp_t *udp)
5447 {
5448 	char *state;
5449 	char addrbuf1[INET6_ADDRSTRLEN];
5450 	char addrbuf2[INET6_ADDRSTRLEN];
5451 	uint_t print_len, buf_len;
5452 
5453 	buf_len = mp->b_datap->db_lim - mp->b_wptr;
5454 	ASSERT(buf_len >= 0);
5455 	if (buf_len == 0)
5456 		return;
5457 
5458 	if (udp->udp_state == TS_UNBND)
5459 		state = "UNBOUND";
5460 	else if (udp->udp_state == TS_IDLE)
5461 		state = "IDLE";
5462 	else if (udp->udp_state == TS_DATA_XFER)
5463 		state = "CONNECTED";
5464 	else
5465 		state = "UnkState";
5466 	print_len = snprintf((char *)mp->b_wptr, buf_len,
5467 	    MI_COL_PTRFMT_STR "%4d %5u %s %s %5u %s\n",
5468 	    (void *)udp, udp->udp_connp->conn_zoneid, ntohs(udp->udp_port),
5469 	    inet_ntop(AF_INET6, &udp->udp_v6src, addrbuf1, sizeof (addrbuf1)),
5470 	    inet_ntop(AF_INET6, &udp->udp_v6dst, addrbuf2, sizeof (addrbuf2)),
5471 	    ntohs(udp->udp_dstport), state);
5472 	if (print_len < buf_len) {
5473 		mp->b_wptr += print_len;
5474 	} else {
5475 		mp->b_wptr += buf_len;
5476 	}
5477 }
5478 
5479 /* Report for ndd "udp_status" */
5480 /* ARGSUSED */
5481 static int
5482 udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
5483 {
5484 	zoneid_t zoneid;
5485 	connf_t	*connfp;
5486 	conn_t	*connp = Q_TO_CONN(q);
5487 	udp_t	*udp = connp->conn_udp;
5488 	int	i;
5489 	udp_stack_t *us = udp->udp_us;
5490 	ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
5491 
5492 	/*
5493 	 * Because of the ndd constraint, at most we can have 64K buffer
5494 	 * to put in all UDP info.  So to be more efficient, just
5495 	 * allocate a 64K buffer here, assuming we need that large buffer.
5496 	 * This may be a problem as any user can read udp_status.  Therefore
5497 	 * we limit the rate of doing this using us_ndd_get_info_interval.
5498 	 * This should be OK as normal users should not do this too often.
5499 	 */
5500 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
5501 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
5502 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
5503 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
5504 			return (0);
5505 		}
5506 	}
5507 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
5508 		/* The following may work even if we cannot get a large buf. */
5509 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
5510 		return (0);
5511 	}
5512 	(void) mi_mpprintf(mp,
5513 	    "UDP     " MI_COL_HDRPAD_STR
5514 	/*   12345678[89ABCDEF] */
5515 	    " zone lport src addr        dest addr       port  state");
5516 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
5517 
5518 	zoneid = connp->conn_zoneid;
5519 
5520 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
5521 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
5522 		connp = NULL;
5523 
5524 		while ((connp = ipcl_get_next_conn(connfp, connp,
5525 		    IPCL_UDPCONN))) {
5526 			udp = connp->conn_udp;
5527 			if (zoneid != GLOBAL_ZONEID &&
5528 			    zoneid != connp->conn_zoneid)
5529 				continue;
5530 
5531 			udp_report_item(mp->b_cont, udp);
5532 		}
5533 	}
5534 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
5535 	return (0);
5536 }
5537 
5538 /*
5539  * This routine creates a T_UDERROR_IND message and passes it upstream.
5540  * The address and options are copied from the T_UNITDATA_REQ message
5541  * passed in mp.  This message is freed.
5542  */
5543 static void
5544 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen,
5545     t_scalar_t err)
5546 {
5547 	struct T_unitdata_req *tudr;
5548 	mblk_t	*mp1;
5549 	uchar_t	*optaddr;
5550 	t_scalar_t optlen;
5551 
5552 	if (DB_TYPE(mp) == M_DATA) {
5553 		ASSERT(destaddr != NULL && destlen != 0);
5554 		optaddr = NULL;
5555 		optlen = 0;
5556 	} else {
5557 		if ((mp->b_wptr < mp->b_rptr) ||
5558 		    (MBLKL(mp)) < sizeof (struct T_unitdata_req)) {
5559 			goto done;
5560 		}
5561 		tudr = (struct T_unitdata_req *)mp->b_rptr;
5562 		destaddr = mp->b_rptr + tudr->DEST_offset;
5563 		if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr ||
5564 		    destaddr + tudr->DEST_length < mp->b_rptr ||
5565 		    destaddr + tudr->DEST_length > mp->b_wptr) {
5566 			goto done;
5567 		}
5568 		optaddr = mp->b_rptr + tudr->OPT_offset;
5569 		if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr ||
5570 		    optaddr + tudr->OPT_length < mp->b_rptr ||
5571 		    optaddr + tudr->OPT_length > mp->b_wptr) {
5572 			goto done;
5573 		}
5574 		destlen = tudr->DEST_length;
5575 		optlen = tudr->OPT_length;
5576 	}
5577 
5578 	mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen,
5579 	    (char *)optaddr, optlen, err);
5580 	if (mp1 != NULL)
5581 		qreply(q, mp1);
5582 
5583 done:
5584 	freemsg(mp);
5585 }
5586 
5587 /*
5588  * This routine removes a port number association from a stream.  It
5589  * is called by udp_wput to handle T_UNBIND_REQ messages.
5590  */
5591 static void
5592 udp_unbind(queue_t *q, mblk_t *mp)
5593 {
5594 	udp_t *udp = Q_TO_UDP(q);
5595 	udp_fanout_t	*udpf;
5596 	udp_stack_t	*us = udp->udp_us;
5597 
5598 	if (cl_inet_unbind != NULL) {
5599 		/*
5600 		 * Running in cluster mode - register unbind information
5601 		 */
5602 		if (udp->udp_ipversion == IPV4_VERSION) {
5603 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
5604 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
5605 			    (in_port_t)udp->udp_port);
5606 		} else {
5607 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
5608 			    (uint8_t *)&(udp->udp_v6src),
5609 			    (in_port_t)udp->udp_port);
5610 		}
5611 	}
5612 
5613 	rw_enter(&udp->udp_rwlock, RW_WRITER);
5614 	if (udp->udp_state == TS_UNBND || udp->udp_pending_op != -1) {
5615 		rw_exit(&udp->udp_rwlock);
5616 		udp_err_ack(q, mp, TOUTSTATE, 0);
5617 		return;
5618 	}
5619 	udp->udp_pending_op = T_UNBIND_REQ;
5620 	rw_exit(&udp->udp_rwlock);
5621 
5622 	/*
5623 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
5624 	 * and therefore ip_unbind must never return NULL.
5625 	 */
5626 	mp = ip_unbind(q, mp);
5627 	ASSERT(mp != NULL);
5628 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
5629 
5630 	/*
5631 	 * Once we're unbound from IP, the pending operation may be cleared
5632 	 * here.
5633 	 */
5634 	rw_enter(&udp->udp_rwlock, RW_WRITER);
5635 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
5636 	    us->us_bind_fanout_size)];
5637 	mutex_enter(&udpf->uf_lock);
5638 	udp_bind_hash_remove(udp, B_TRUE);
5639 	V6_SET_ZERO(udp->udp_v6src);
5640 	V6_SET_ZERO(udp->udp_bound_v6src);
5641 	udp->udp_port = 0;
5642 	mutex_exit(&udpf->uf_lock);
5643 
5644 	udp->udp_pending_op = -1;
5645 	udp->udp_state = TS_UNBND;
5646 	if (udp->udp_family == AF_INET6)
5647 		(void) udp_build_hdrs(udp);
5648 	rw_exit(&udp->udp_rwlock);
5649 
5650 	qreply(q, mp);
5651 }
5652 
5653 /*
5654  * Don't let port fall into the privileged range.
5655  * Since the extra privileged ports can be arbitrary we also
5656  * ensure that we exclude those from consideration.
5657  * us->us_epriv_ports is not sorted thus we loop over it until
5658  * there are no changes.
5659  */
5660 static in_port_t
5661 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random)
5662 {
5663 	int i;
5664 	in_port_t nextport;
5665 	boolean_t restart = B_FALSE;
5666 	udp_stack_t *us = udp->udp_us;
5667 
5668 	if (random && udp_random_anon_port != 0) {
5669 		(void) random_get_pseudo_bytes((uint8_t *)&port,
5670 		    sizeof (in_port_t));
5671 		/*
5672 		 * Unless changed by a sys admin, the smallest anon port
5673 		 * is 32768 and the largest anon port is 65535.  It is
5674 		 * very likely (50%) for the random port to be smaller
5675 		 * than the smallest anon port.  When that happens,
5676 		 * add port % (anon port range) to the smallest anon
5677 		 * port to get the random port.  It should fall into the
5678 		 * valid anon port range.
5679 		 */
5680 		if (port < us->us_smallest_anon_port) {
5681 			port = us->us_smallest_anon_port +
5682 			    port % (us->us_largest_anon_port -
5683 			    us->us_smallest_anon_port);
5684 		}
5685 	}
5686 
5687 retry:
5688 	if (port < us->us_smallest_anon_port)
5689 		port = us->us_smallest_anon_port;
5690 
5691 	if (port > us->us_largest_anon_port) {
5692 		port = us->us_smallest_anon_port;
5693 		if (restart)
5694 			return (0);
5695 		restart = B_TRUE;
5696 	}
5697 
5698 	if (port < us->us_smallest_nonpriv_port)
5699 		port = us->us_smallest_nonpriv_port;
5700 
5701 	for (i = 0; i < us->us_num_epriv_ports; i++) {
5702 		if (port == us->us_epriv_ports[i]) {
5703 			port++;
5704 			/*
5705 			 * Make sure that the port is in the
5706 			 * valid range.
5707 			 */
5708 			goto retry;
5709 		}
5710 	}
5711 
5712 	if (is_system_labeled() &&
5713 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
5714 	    port, IPPROTO_UDP, B_TRUE)) != 0) {
5715 		port = nextport;
5716 		goto retry;
5717 	}
5718 
5719 	return (port);
5720 }
5721 
5722 static int
5723 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst)
5724 {
5725 	int err;
5726 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
5727 	udp_t *udp = Q_TO_UDP(wq);
5728 	udp_stack_t	*us = udp->udp_us;
5729 
5730 	err = tsol_compute_label(DB_CREDDEF(mp, udp->udp_connp->conn_cred), dst,
5731 	    opt_storage, udp->udp_connp->conn_mac_exempt,
5732 	    us->us_netstack->netstack_ip);
5733 	if (err == 0) {
5734 		err = tsol_update_options(&udp->udp_ip_snd_options,
5735 		    &udp->udp_ip_snd_options_len, &udp->udp_label_len,
5736 		    opt_storage);
5737 	}
5738 	if (err != 0) {
5739 		DTRACE_PROBE4(
5740 		    tx__ip__log__info__updatelabel__udp,
5741 		    char *, "queue(1) failed to update options(2) on mp(3)",
5742 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
5743 	} else {
5744 		IN6_IPADDR_TO_V4MAPPED(dst, &udp->udp_v6lastdst);
5745 	}
5746 	return (err);
5747 }
5748 
5749 static mblk_t *
5750 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port,
5751     uint_t srcid, int *error, boolean_t insert_spi)
5752 {
5753 	udp_t	*udp = connp->conn_udp;
5754 	queue_t	*q = connp->conn_wq;
5755 	mblk_t	*mp1 = mp;
5756 	mblk_t	*mp2;
5757 	ipha_t	*ipha;
5758 	int	ip_hdr_length;
5759 	uint32_t ip_len;
5760 	udpha_t	*udpha;
5761 	boolean_t lock_held = B_FALSE;
5762 	in_port_t	uha_src_port;
5763 	udpattrs_t	attrs;
5764 	uchar_t	ip_snd_opt[IP_MAX_OPT_LENGTH];
5765 	uint32_t	ip_snd_opt_len = 0;
5766 	ip4_pkt_t  pktinfo;
5767 	ip4_pkt_t  *pktinfop = &pktinfo;
5768 	ip_opt_info_t optinfo;
5769 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
5770 	udp_stack_t	*us = udp->udp_us;
5771 	ipsec_stack_t	*ipss = ipst->ips_netstack->netstack_ipsec;
5772 
5773 
5774 	*error = 0;
5775 	pktinfop->ip4_ill_index = 0;
5776 	pktinfop->ip4_addr = INADDR_ANY;
5777 	optinfo.ip_opt_flags = 0;
5778 	optinfo.ip_opt_ill_index = 0;
5779 
5780 	if (v4dst == INADDR_ANY)
5781 		v4dst = htonl(INADDR_LOOPBACK);
5782 
5783 	/*
5784 	 * If options passed in, feed it for verification and handling
5785 	 */
5786 	attrs.udpattr_credset = B_FALSE;
5787 	if (DB_TYPE(mp) != M_DATA) {
5788 		mp1 = mp->b_cont;
5789 		if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) {
5790 			attrs.udpattr_ipp4 = pktinfop;
5791 			attrs.udpattr_mb = mp;
5792 			if (udp_unitdata_opt_process(q, mp, error, &attrs) < 0)
5793 				goto done;
5794 			/*
5795 			 * Note: success in processing options.
5796 			 * mp option buffer represented by
5797 			 * OPT_length/offset now potentially modified
5798 			 * and contain option setting results
5799 			 */
5800 			ASSERT(*error == 0);
5801 		}
5802 	}
5803 
5804 	/* mp1 points to the M_DATA mblk carrying the packet */
5805 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
5806 
5807 	rw_enter(&udp->udp_rwlock, RW_READER);
5808 	lock_held = B_TRUE;
5809 	/*
5810 	 * Check if our saved options are valid; update if not.
5811 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
5812 	 * to different destination may require different labels,
5813 	 * or worse, UDP packets to same IP address may require
5814 	 * different labels due to use of shared all-zones address.
5815 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
5816 	 * and ip_snd_options_len are consistent for the current
5817 	 * destination and are updated atomically.
5818 	 */
5819 	mutex_enter(&connp->conn_lock);
5820 	if (is_system_labeled()) {
5821 		/* Using UDP MLP requires SCM_UCRED from user */
5822 		if (connp->conn_mlp_type != mlptSingle &&
5823 		    !attrs.udpattr_credset) {
5824 			mutex_exit(&connp->conn_lock);
5825 			DTRACE_PROBE4(
5826 			    tx__ip__log__info__output__udp,
5827 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
5828 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
5829 			*error = ECONNREFUSED;
5830 			goto done;
5831 		}
5832 		/*
5833 		 * update label option for this UDP socket if
5834 		 * - the destination has changed, or
5835 		 * - the UDP socket is MLP
5836 		 */
5837 		if ((!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
5838 		    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
5839 		    connp->conn_mlp_type != mlptSingle) &&
5840 		    (*error = udp_update_label(q, mp, v4dst)) != 0) {
5841 			mutex_exit(&connp->conn_lock);
5842 			goto done;
5843 		}
5844 	}
5845 	if (udp->udp_ip_snd_options_len > 0) {
5846 		ip_snd_opt_len = udp->udp_ip_snd_options_len;
5847 		bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
5848 	}
5849 	mutex_exit(&connp->conn_lock);
5850 
5851 	/* Add an IP header */
5852 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len +
5853 	    (insert_spi ? sizeof (uint32_t) : 0);
5854 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
5855 	if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) ||
5856 	    !OK_32PTR(ipha)) {
5857 		mp2 = allocb(ip_hdr_length + us->us_wroff_extra, BPRI_LO);
5858 		if (mp2 == NULL) {
5859 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5860 			    "udp_wput_end: q %p (%S)", q, "allocbfail2");
5861 			*error = ENOMEM;
5862 			goto done;
5863 		}
5864 		mp2->b_wptr = DB_LIM(mp2);
5865 		mp2->b_cont = mp1;
5866 		mp1 = mp2;
5867 		if (DB_TYPE(mp) != M_DATA)
5868 			mp->b_cont = mp1;
5869 		else
5870 			mp = mp1;
5871 
5872 		ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length);
5873 	}
5874 	ip_hdr_length -= (UDPH_SIZE + (insert_spi ? sizeof (uint32_t) : 0));
5875 #ifdef	_BIG_ENDIAN
5876 	/* Set version, header length, and tos */
5877 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5878 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
5879 	    udp->udp_type_of_service);
5880 	/* Set ttl and protocol */
5881 	*(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP;
5882 #else
5883 	/* Set version, header length, and tos */
5884 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
5885 	    ((udp->udp_type_of_service << 8) |
5886 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
5887 	/* Set ttl and protocol */
5888 	*(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl;
5889 #endif
5890 	if (pktinfop->ip4_addr != INADDR_ANY) {
5891 		ipha->ipha_src = pktinfop->ip4_addr;
5892 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
5893 	} else {
5894 		/*
5895 		 * Copy our address into the packet.  If this is zero,
5896 		 * first look at __sin6_src_id for a hint. If we leave the
5897 		 * source as INADDR_ANY then ip will fill in the real source
5898 		 * address.
5899 		 */
5900 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src);
5901 		if (srcid != 0 && ipha->ipha_src == INADDR_ANY) {
5902 			in6_addr_t v6src;
5903 
5904 			ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid,
5905 			    us->us_netstack);
5906 			IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src);
5907 		}
5908 	}
5909 	uha_src_port = udp->udp_port;
5910 	if (ip_hdr_length == IP_SIMPLE_HDR_LENGTH) {
5911 		rw_exit(&udp->udp_rwlock);
5912 		lock_held = B_FALSE;
5913 	}
5914 
5915 	if (pktinfop->ip4_ill_index != 0) {
5916 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
5917 	}
5918 
5919 	ipha->ipha_fragment_offset_and_flags = 0;
5920 	ipha->ipha_ident = 0;
5921 
5922 	mp1->b_rptr = (uchar_t *)ipha;
5923 
5924 	ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <=
5925 	    (uintptr_t)UINT_MAX);
5926 
5927 	/* Determine length of packet */
5928 	ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha);
5929 	if ((mp2 = mp1->b_cont) != NULL) {
5930 		do {
5931 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
5932 			ip_len += (uint32_t)MBLKL(mp2);
5933 		} while ((mp2 = mp2->b_cont) != NULL);
5934 	}
5935 	/*
5936 	 * If the size of the packet is greater than the maximum allowed by
5937 	 * ip, return an error. Passing this down could cause panics because
5938 	 * the size will have wrapped and be inconsistent with the msg size.
5939 	 */
5940 	if (ip_len > IP_MAXPACKET) {
5941 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
5942 		    "udp_wput_end: q %p (%S)", q, "IP length exceeded");
5943 		*error = EMSGSIZE;
5944 		goto done;
5945 	}
5946 	ipha->ipha_length = htons((uint16_t)ip_len);
5947 	ip_len -= ip_hdr_length;
5948 	ip_len = htons((uint16_t)ip_len);
5949 	udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length);
5950 
5951 	/* Insert all-0s SPI now. */
5952 	if (insert_spi)
5953 		*((uint32_t *)(udpha + 1)) = 0;
5954 
5955 	/*
5956 	 * Copy in the destination address
5957 	 */
5958 	ipha->ipha_dst = v4dst;
5959 
5960 	/*
5961 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
5962 	 */
5963 	if (CLASSD(v4dst))
5964 		ipha->ipha_ttl = udp->udp_multicast_ttl;
5965 
5966 	udpha->uha_dst_port = port;
5967 	udpha->uha_src_port = uha_src_port;
5968 
5969 	if (ip_snd_opt_len > 0) {
5970 		uint32_t	cksum;
5971 
5972 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
5973 		lock_held = B_FALSE;
5974 		rw_exit(&udp->udp_rwlock);
5975 		/*
5976 		 * Massage source route putting first source route in ipha_dst.
5977 		 * Ignore the destination in T_unitdata_req.
5978 		 * Create a checksum adjustment for a source route, if any.
5979 		 */
5980 		cksum = ip_massage_options(ipha, us->us_netstack);
5981 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5982 		cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) +
5983 		    (ipha->ipha_dst & 0xFFFF);
5984 		if ((int)cksum < 0)
5985 			cksum--;
5986 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5987 		/*
5988 		 * IP does the checksum if uha_checksum is non-zero,
5989 		 * We make it easy for IP to include our pseudo header
5990 		 * by putting our length in uha_checksum.
5991 		 */
5992 		cksum += ip_len;
5993 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5994 		/* There might be a carry. */
5995 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
5996 #ifdef _LITTLE_ENDIAN
5997 		if (us->us_do_checksum)
5998 			ip_len = (cksum << 16) | ip_len;
5999 #else
6000 		if (us->us_do_checksum)
6001 			ip_len = (ip_len << 16) | cksum;
6002 		else
6003 			ip_len <<= 16;
6004 #endif
6005 	} else {
6006 		/*
6007 		 * IP does the checksum if uha_checksum is non-zero,
6008 		 * We make it easy for IP to include our pseudo header
6009 		 * by putting our length in uha_checksum.
6010 		 */
6011 		if (us->us_do_checksum)
6012 			ip_len |= (ip_len << 16);
6013 #ifndef _LITTLE_ENDIAN
6014 		else
6015 			ip_len <<= 16;
6016 #endif
6017 	}
6018 	ASSERT(!lock_held);
6019 	/* Set UDP length and checksum */
6020 	*((uint32_t *)&udpha->uha_length) = ip_len;
6021 	if (DB_CRED(mp) != NULL)
6022 		mblk_setcred(mp1, DB_CRED(mp));
6023 
6024 	if (DB_TYPE(mp) != M_DATA) {
6025 		ASSERT(mp != mp1);
6026 		freeb(mp);
6027 	}
6028 
6029 	/* mp has been consumed and we'll return success */
6030 	ASSERT(*error == 0);
6031 	mp = NULL;
6032 
6033 	/* We're done.  Pass the packet to ip. */
6034 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
6035 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6036 	    "udp_wput_end: q %p (%S)", q, "end");
6037 
6038 	if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 ||
6039 	    CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) ||
6040 	    connp->conn_dontroute ||
6041 	    connp->conn_nofailover_ill != NULL ||
6042 	    connp->conn_outgoing_ill != NULL || optinfo.ip_opt_flags != 0 ||
6043 	    optinfo.ip_opt_ill_index != 0 ||
6044 	    ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION ||
6045 	    IPP_ENABLED(IPP_LOCAL_OUT, ipst) ||
6046 	    ipst->ips_ip_g_mrouter != NULL) {
6047 		UDP_STAT(us, udp_ip_send);
6048 		ip_output_options(connp, mp1, connp->conn_wq, IP_WPUT,
6049 		    &optinfo);
6050 	} else {
6051 		udp_send_data(udp, connp->conn_wq, mp1, ipha);
6052 	}
6053 
6054 done:
6055 	if (lock_held)
6056 		rw_exit(&udp->udp_rwlock);
6057 	if (*error != 0) {
6058 		ASSERT(mp != NULL);
6059 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6060 	}
6061 	return (mp);
6062 }
6063 
6064 static void
6065 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha)
6066 {
6067 	conn_t	*connp = udp->udp_connp;
6068 	ipaddr_t src, dst;
6069 	ire_t	*ire;
6070 	ipif_t	*ipif = NULL;
6071 	mblk_t	*ire_fp_mp;
6072 	boolean_t retry_caching;
6073 	udp_stack_t *us = udp->udp_us;
6074 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
6075 
6076 	dst = ipha->ipha_dst;
6077 	src = ipha->ipha_src;
6078 	ASSERT(ipha->ipha_ident == 0);
6079 
6080 	if (CLASSD(dst)) {
6081 		int err;
6082 
6083 		ipif = conn_get_held_ipif(connp,
6084 		    &connp->conn_multicast_ipif, &err);
6085 
6086 		if (ipif == NULL || ipif->ipif_isv6 ||
6087 		    (ipif->ipif_ill->ill_phyint->phyint_flags &
6088 		    PHYI_LOOPBACK)) {
6089 			if (ipif != NULL)
6090 				ipif_refrele(ipif);
6091 			UDP_STAT(us, udp_ip_send);
6092 			ip_output(connp, mp, q, IP_WPUT);
6093 			return;
6094 		}
6095 	}
6096 
6097 	retry_caching = B_FALSE;
6098 	mutex_enter(&connp->conn_lock);
6099 	ire = connp->conn_ire_cache;
6100 	ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT));
6101 
6102 	if (ire == NULL || ire->ire_addr != dst ||
6103 	    (ire->ire_marks & IRE_MARK_CONDEMNED)) {
6104 		retry_caching = B_TRUE;
6105 	} else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) {
6106 		ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr;
6107 
6108 		ASSERT(ipif != NULL);
6109 		if (stq_ill != ipif->ipif_ill && (stq_ill->ill_group == NULL ||
6110 		    stq_ill->ill_group != ipif->ipif_ill->ill_group))
6111 			retry_caching = B_TRUE;
6112 	}
6113 
6114 	if (!retry_caching) {
6115 		ASSERT(ire != NULL);
6116 		IRE_REFHOLD(ire);
6117 		mutex_exit(&connp->conn_lock);
6118 	} else {
6119 		boolean_t cached = B_FALSE;
6120 
6121 		connp->conn_ire_cache = NULL;
6122 		mutex_exit(&connp->conn_lock);
6123 
6124 		/* Release the old ire */
6125 		if (ire != NULL) {
6126 			IRE_REFRELE_NOTR(ire);
6127 			ire = NULL;
6128 		}
6129 
6130 		if (CLASSD(dst)) {
6131 			ASSERT(ipif != NULL);
6132 			ire = ire_ctable_lookup(dst, 0, 0, ipif,
6133 			    connp->conn_zoneid, MBLK_GETLABEL(mp),
6134 			    MATCH_IRE_ILL_GROUP, ipst);
6135 		} else {
6136 			ASSERT(ipif == NULL);
6137 			ire = ire_cache_lookup(dst, connp->conn_zoneid,
6138 			    MBLK_GETLABEL(mp), ipst);
6139 		}
6140 
6141 		if (ire == NULL) {
6142 			if (ipif != NULL)
6143 				ipif_refrele(ipif);
6144 			UDP_STAT(us, udp_ire_null);
6145 			ip_output(connp, mp, q, IP_WPUT);
6146 			return;
6147 		}
6148 		IRE_REFHOLD_NOTR(ire);
6149 
6150 		mutex_enter(&connp->conn_lock);
6151 		if (CONN_CACHE_IRE(connp) && connp->conn_ire_cache == NULL &&
6152 		    !(ire->ire_marks & IRE_MARK_CONDEMNED)) {
6153 			irb_t		*irb = ire->ire_bucket;
6154 
6155 			/*
6156 			 * IRE's created for non-connection oriented transports
6157 			 * are normally initialized with IRE_MARK_TEMPORARY set
6158 			 * in the ire_marks. These IRE's are preferentially
6159 			 * reaped when the hash chain length in the cache
6160 			 * bucket exceeds the maximum value specified in
6161 			 * ip[6]_ire_max_bucket_cnt. This can severely affect
6162 			 * UDP performance if IRE cache entries that we need
6163 			 * to reuse are continually removed. To remedy this,
6164 			 * when we cache the IRE in the conn_t, we remove the
6165 			 * IRE_MARK_TEMPORARY bit from the ire_marks if it was
6166 			 * set.
6167 			 */
6168 			if (ire->ire_marks & IRE_MARK_TEMPORARY) {
6169 				rw_enter(&irb->irb_lock, RW_WRITER);
6170 				if (ire->ire_marks & IRE_MARK_TEMPORARY) {
6171 					ire->ire_marks &= ~IRE_MARK_TEMPORARY;
6172 					irb->irb_tmp_ire_cnt--;
6173 				}
6174 				rw_exit(&irb->irb_lock);
6175 			}
6176 			connp->conn_ire_cache = ire;
6177 			cached = B_TRUE;
6178 		}
6179 		mutex_exit(&connp->conn_lock);
6180 
6181 		/*
6182 		 * We can continue to use the ire but since it was not
6183 		 * cached, we should drop the extra reference.
6184 		 */
6185 		if (!cached)
6186 			IRE_REFRELE_NOTR(ire);
6187 	}
6188 	ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION);
6189 	ASSERT(!CLASSD(dst) || ipif != NULL);
6190 
6191 	/*
6192 	 * Check if we can take the fast-path.
6193 	 * Note that "incomplete" ire's (where the link-layer for next hop
6194 	 * is not resolved, or where the fast-path header in nce_fp_mp is not
6195 	 * available yet) are sent down the legacy (slow) path
6196 	 */
6197 	if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) ||
6198 	    (ire->ire_flags & RTF_MULTIRT) || (ire->ire_stq == NULL) ||
6199 	    (ire->ire_max_frag < ntohs(ipha->ipha_length)) ||
6200 	    ((ire->ire_nce == NULL) ||
6201 	    ((ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL)) ||
6202 	    connp->conn_nexthop_set || (MBLKL(ire_fp_mp) > MBLKHEAD(mp))) {
6203 		if (ipif != NULL)
6204 			ipif_refrele(ipif);
6205 		UDP_STAT(us, udp_ip_ire_send);
6206 		IRE_REFRELE(ire);
6207 		ip_output(connp, mp, q, IP_WPUT);
6208 		return;
6209 	}
6210 
6211 	if (src == INADDR_ANY && !connp->conn_unspec_src) {
6212 		if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC))
6213 			ipha->ipha_src = ipif->ipif_src_addr;
6214 		else
6215 			ipha->ipha_src = ire->ire_src_addr;
6216 	}
6217 
6218 	if (ipif != NULL)
6219 		ipif_refrele(ipif);
6220 
6221 	udp_xmit(connp->conn_wq, mp, ire, connp, connp->conn_zoneid);
6222 }
6223 
6224 static void
6225 udp_xmit(queue_t *q, mblk_t *mp, ire_t *ire, conn_t *connp, zoneid_t zoneid)
6226 {
6227 	ipaddr_t src, dst;
6228 	ill_t	*ill;
6229 	mblk_t	*ire_fp_mp;
6230 	uint_t	ire_fp_mp_len;
6231 	uint16_t *up;
6232 	uint32_t cksum, hcksum_txflags;
6233 	queue_t	*dev_q;
6234 	udp_t	*udp = connp->conn_udp;
6235 	ipha_t	*ipha = (ipha_t *)mp->b_rptr;
6236 	udp_stack_t	*us = udp->udp_us;
6237 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
6238 	boolean_t ll_multicast = B_FALSE;
6239 
6240 	dev_q = ire->ire_stq->q_next;
6241 	ASSERT(dev_q != NULL);
6242 
6243 	ill = ire_to_ill(ire);
6244 	ASSERT(ill != NULL);
6245 
6246 	/* is queue flow controlled? */
6247 	if (q->q_first != NULL || connp->conn_draining ||
6248 	    DEV_Q_FLOW_BLOCKED(dev_q)) {
6249 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
6250 		BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
6251 		if (ipst->ips_ip_output_queue)
6252 			(void) putq(connp->conn_wq, mp);
6253 		else
6254 			freemsg(mp);
6255 		ire_refrele(ire);
6256 		return;
6257 	}
6258 
6259 	ire_fp_mp = ire->ire_nce->nce_fp_mp;
6260 	ire_fp_mp_len = MBLKL(ire_fp_mp);
6261 	ASSERT(MBLKHEAD(mp) >= ire_fp_mp_len);
6262 
6263 	dst = ipha->ipha_dst;
6264 	src = ipha->ipha_src;
6265 
6266 
6267 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
6268 
6269 	ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1);
6270 #ifndef _BIG_ENDIAN
6271 	ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8);
6272 #endif
6273 
6274 	if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
6275 		ASSERT(ill->ill_hcksum_capab != NULL);
6276 		hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags;
6277 	} else {
6278 		hcksum_txflags = 0;
6279 	}
6280 
6281 	/* pseudo-header checksum (do it in parts for IP header checksum) */
6282 	cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF);
6283 
6284 	ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION);
6285 	up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH);
6286 	if (*up != 0) {
6287 		IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags,
6288 		    mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH,
6289 		    ntohs(ipha->ipha_length), cksum);
6290 
6291 		/* Software checksum? */
6292 		if (DB_CKSUMFLAGS(mp) == 0) {
6293 			UDP_STAT(us, udp_out_sw_cksum);
6294 			UDP_STAT_UPDATE(us, udp_out_sw_cksum_bytes,
6295 			    ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH);
6296 		}
6297 	}
6298 
6299 	if (!CLASSD(dst)) {
6300 		ipha->ipha_fragment_offset_and_flags |=
6301 		    (uint32_t)htons(ire->ire_frag_flag);
6302 	}
6303 
6304 	/* Calculate IP header checksum if hardware isn't capable */
6305 	if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) {
6306 		IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0],
6307 		    ((uint16_t *)ipha)[4]);
6308 	}
6309 
6310 	if (CLASSD(dst)) {
6311 		boolean_t ilm_exists;
6312 
6313 		ILM_WALKER_HOLD(ill);
6314 		ilm_exists = (ilm_lookup_ill(ill, dst, ALL_ZONES) != NULL);
6315 		ILM_WALKER_RELE(ill);
6316 		if (ilm_exists) {
6317 			ip_multicast_loopback(q, ill, mp,
6318 			    connp->conn_multicast_loop ? 0 :
6319 			    IP_FF_NO_MCAST_LOOP, zoneid);
6320 		}
6321 
6322 		/* If multicast TTL is 0 then we are done */
6323 		if (ipha->ipha_ttl == 0) {
6324 			freemsg(mp);
6325 			ire_refrele(ire);
6326 			return;
6327 		}
6328 		ll_multicast = B_TRUE;
6329 	}
6330 
6331 	ASSERT(DB_TYPE(ire_fp_mp) == M_DATA);
6332 	mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len;
6333 	bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len);
6334 
6335 	UPDATE_OB_PKT_COUNT(ire);
6336 	ire->ire_last_used_time = lbolt;
6337 
6338 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
6339 	UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
6340 	    ntohs(ipha->ipha_length));
6341 
6342 	DTRACE_PROBE4(ip4__physical__out__start,
6343 	    ill_t *, NULL, ill_t *, ill, ipha_t *, ipha, mblk_t *, mp);
6344 	FW_HOOKS(ipst->ips_ip4_physical_out_event,
6345 	    ipst->ips_ipv4firewall_physical_out, NULL, ill, ipha, mp, mp,
6346 	    ll_multicast, ipst);
6347 	DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);
6348 	if (ipst->ips_ipobs_enabled && mp != NULL) {
6349 		zoneid_t szone;
6350 
6351 		szone = ip_get_zoneid_v4(ipha->ipha_src, mp,
6352 		    ipst, ALL_ZONES);
6353 		ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone,
6354 		    ALL_ZONES, ill, IPV4_VERSION, ire_fp_mp_len, ipst);
6355 	}
6356 
6357 	if (mp != NULL) {
6358 		DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
6359 		    void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
6360 		    ipha_t *, ipha, ip6_t *, NULL, int, 0);
6361 
6362 		if (ILL_DIRECT_CAPABLE(ill)) {
6363 			ill_dld_direct_t *idd = &ill->ill_dld_capab->idc_direct;
6364 
6365 			(void) idd->idd_tx_df(idd->idd_tx_dh, mp,
6366 			    (uintptr_t)connp, 0);
6367 		} else {
6368 			putnext(ire->ire_stq, mp);
6369 		}
6370 	}
6371 	IRE_REFRELE(ire);
6372 }
6373 
6374 static boolean_t
6375 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst)
6376 {
6377 	udp_t *udp = Q_TO_UDP(wq);
6378 	int err;
6379 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
6380 	udp_stack_t		*us = udp->udp_us;
6381 
6382 	err = tsol_compute_label_v6(DB_CREDDEF(mp, udp->udp_connp->conn_cred),
6383 	    dst, opt_storage, udp->udp_connp->conn_mac_exempt,
6384 	    us->us_netstack->netstack_ip);
6385 	if (err == 0) {
6386 		err = tsol_update_sticky(&udp->udp_sticky_ipp,
6387 		    &udp->udp_label_len_v6, opt_storage);
6388 	}
6389 	if (err != 0) {
6390 		DTRACE_PROBE4(
6391 		    tx__ip__log__drop__updatelabel__udp6,
6392 		    char *, "queue(1) failed to update options(2) on mp(3)",
6393 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
6394 	} else {
6395 		udp->udp_v6lastdst = *dst;
6396 	}
6397 	return (err);
6398 }
6399 
6400 void
6401 udp_output_connected(void *arg, mblk_t *mp)
6402 {
6403 	conn_t	*connp = (conn_t *)arg;
6404 	udp_t	*udp = connp->conn_udp;
6405 	udp_stack_t	*us = udp->udp_us;
6406 	ipaddr_t	v4dst;
6407 	in_port_t	dstport;
6408 	boolean_t	mapped_addr;
6409 	struct sockaddr_storage ss;
6410 	sin_t		*sin;
6411 	sin6_t		*sin6;
6412 	struct sockaddr	*addr;
6413 	socklen_t	addrlen;
6414 	int		error;
6415 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
6416 
6417 	/* M_DATA for connected socket */
6418 
6419 	ASSERT(udp->udp_issocket);
6420 	UDP_DBGSTAT(us, udp_data_conn);
6421 
6422 	mutex_enter(&connp->conn_lock);
6423 	if (udp->udp_state != TS_DATA_XFER) {
6424 		mutex_exit(&connp->conn_lock);
6425 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6426 		UDP_STAT(us, udp_out_err_notconn);
6427 		freemsg(mp);
6428 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6429 		    "udp_wput_end: connp %p (%S)", connp,
6430 		    "not-connected; address required");
6431 		return;
6432 	}
6433 
6434 	mapped_addr = IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst);
6435 	if (mapped_addr)
6436 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst);
6437 
6438 	/* Initialize addr and addrlen as if they're passed in */
6439 	if (udp->udp_family == AF_INET) {
6440 		sin = (sin_t *)&ss;
6441 		sin->sin_family = AF_INET;
6442 		dstport = sin->sin_port = udp->udp_dstport;
6443 		ASSERT(mapped_addr);
6444 		sin->sin_addr.s_addr = v4dst;
6445 		addr = (struct sockaddr *)sin;
6446 		addrlen = sizeof (*sin);
6447 	} else {
6448 		sin6 = (sin6_t *)&ss;
6449 		sin6->sin6_family = AF_INET6;
6450 		dstport = sin6->sin6_port = udp->udp_dstport;
6451 		sin6->sin6_flowinfo = udp->udp_flowinfo;
6452 		sin6->sin6_addr = udp->udp_v6dst;
6453 		sin6->sin6_scope_id = 0;
6454 		sin6->__sin6_src_id = 0;
6455 		addr = (struct sockaddr *)sin6;
6456 		addrlen = sizeof (*sin6);
6457 	}
6458 	mutex_exit(&connp->conn_lock);
6459 
6460 	if (mapped_addr) {
6461 		/*
6462 		 * Handle both AF_INET and AF_INET6; the latter
6463 		 * for IPV4 mapped destination addresses.  Note
6464 		 * here that both addr and addrlen point to the
6465 		 * corresponding struct depending on the address
6466 		 * family of the socket.
6467 		 */
6468 		mp = udp_output_v4(connp, mp, v4dst, dstport, 0, &error,
6469 		    insert_spi);
6470 	} else {
6471 		mp = udp_output_v6(connp, mp, sin6, &error);
6472 	}
6473 	if (error == 0) {
6474 		ASSERT(mp == NULL);
6475 		return;
6476 	}
6477 
6478 	UDP_STAT(us, udp_out_err_output);
6479 	ASSERT(mp != NULL);
6480 	/* mp is freed by the following routine */
6481 	udp_ud_err(connp->conn_wq, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
6482 	    (t_scalar_t)error);
6483 }
6484 
6485 /*
6486  * This routine handles all messages passed downstream.  It either
6487  * consumes the message or passes it downstream; it never queues a
6488  * a message.
6489  *
6490  * Also entry point for sockfs when udp is in "direct sockfs" mode.  This mode
6491  * is valid when we are directly beneath the stream head, and thus sockfs
6492  * is able to bypass STREAMS and directly call us, passing along the sockaddr
6493  * structure without the cumbersome T_UNITDATA_REQ interface for the case of
6494  * connected endpoints.
6495  */
6496 void
6497 udp_wput(queue_t *q, mblk_t *mp)
6498 {
6499 	sin6_t		*sin6;
6500 	sin_t		*sin;
6501 	ipaddr_t	v4dst;
6502 	uint16_t	port;
6503 	uint_t		srcid;
6504 	conn_t		*connp = Q_TO_CONN(q);
6505 	udp_t		*udp = connp->conn_udp;
6506 	int		error = 0;
6507 	struct sockaddr	*addr;
6508 	socklen_t	addrlen;
6509 	udp_stack_t *us = udp->udp_us;
6510 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
6511 
6512 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START,
6513 	    "udp_wput_start: queue %p mp %p", q, mp);
6514 
6515 	/*
6516 	 * We directly handle several cases here: T_UNITDATA_REQ message
6517 	 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for connected
6518 	 * socket.
6519 	 */
6520 	switch (DB_TYPE(mp)) {
6521 	case M_DATA:
6522 		/*
6523 		 * Quick check for error cases. Checks will be done again
6524 		 * under the lock later on
6525 		 */
6526 		if (!udp->udp_direct_sockfs || udp->udp_state != TS_DATA_XFER) {
6527 			/* Not connected; address is required */
6528 			BUMP_MIB(&us->us_udp_mib, udpOutErrors);
6529 			UDP_STAT(us, udp_out_err_notconn);
6530 			freemsg(mp);
6531 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6532 			    "udp_wput_end: connp %p (%S)", connp,
6533 			    "not-connected; address required");
6534 			return;
6535 		}
6536 		udp_output_connected(connp, mp);
6537 		return;
6538 
6539 	case M_PROTO:
6540 	case M_PCPROTO: {
6541 		struct T_unitdata_req *tudr;
6542 
6543 		ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX);
6544 		tudr = (struct T_unitdata_req *)mp->b_rptr;
6545 
6546 		/* Handle valid T_UNITDATA_REQ here */
6547 		if (MBLKL(mp) >= sizeof (*tudr) &&
6548 		    ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) {
6549 			if (mp->b_cont == NULL) {
6550 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6551 				    "udp_wput_end: q %p (%S)", q, "badaddr");
6552 				error = EPROTO;
6553 				goto ud_error;
6554 			}
6555 
6556 			if (!MBLKIN(mp, 0, tudr->DEST_offset +
6557 			    tudr->DEST_length)) {
6558 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6559 				    "udp_wput_end: q %p (%S)", q, "badaddr");
6560 				error = EADDRNOTAVAIL;
6561 				goto ud_error;
6562 			}
6563 			/*
6564 			 * If a port has not been bound to the stream, fail.
6565 			 * This is not a problem when sockfs is directly
6566 			 * above us, because it will ensure that the socket
6567 			 * is first bound before allowing data to be sent.
6568 			 */
6569 			if (udp->udp_state == TS_UNBND) {
6570 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6571 				    "udp_wput_end: q %p (%S)", q, "outstate");
6572 				error = EPROTO;
6573 				goto ud_error;
6574 			}
6575 			addr = (struct sockaddr *)
6576 			    &mp->b_rptr[tudr->DEST_offset];
6577 			addrlen = tudr->DEST_length;
6578 			if (tudr->OPT_length != 0)
6579 				UDP_STAT(us, udp_out_opt);
6580 			break;
6581 		}
6582 		/* FALLTHRU */
6583 	}
6584 	default:
6585 		udp_wput_other(q, mp);
6586 		return;
6587 	}
6588 	ASSERT(addr != NULL);
6589 
6590 	switch (udp->udp_family) {
6591 	case AF_INET6:
6592 		sin6 = (sin6_t *)addr;
6593 		if (!OK_32PTR((char *)sin6) || (addrlen != sizeof (sin6_t)) ||
6594 		    (sin6->sin6_family != AF_INET6)) {
6595 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6596 			    "udp_wput_end: q %p (%S)", q, "badaddr");
6597 			error = EADDRNOTAVAIL;
6598 			goto ud_error;
6599 		}
6600 
6601 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
6602 			/*
6603 			 * Destination is a non-IPv4-compatible IPv6 address.
6604 			 * Send out an IPv6 format packet.
6605 			 */
6606 			mp = udp_output_v6(connp, mp, sin6, &error);
6607 			if (error != 0)
6608 				goto ud_error;
6609 
6610 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6611 			    "udp_wput_end: q %p (%S)", q, "udp_output_v6");
6612 			return;
6613 		}
6614 		/*
6615 		 * If the local address is not zero or a mapped address
6616 		 * return an error.  It would be possible to send an IPv4
6617 		 * packet but the response would never make it back to the
6618 		 * application since it is bound to a non-mapped address.
6619 		 */
6620 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) &&
6621 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
6622 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6623 			    "udp_wput_end: q %p (%S)", q, "badaddr");
6624 			error = EADDRNOTAVAIL;
6625 			goto ud_error;
6626 		}
6627 		/* Send IPv4 packet without modifying udp_ipversion */
6628 		/* Extract port and ipaddr */
6629 		port = sin6->sin6_port;
6630 		IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst);
6631 		srcid = sin6->__sin6_src_id;
6632 		break;
6633 
6634 	case AF_INET:
6635 		sin = (sin_t *)addr;
6636 		if ((!OK_32PTR((char *)sin) || addrlen != sizeof (sin_t)) ||
6637 		    (sin->sin_family != AF_INET)) {
6638 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6639 			    "udp_wput_end: q %p (%S)", q, "badaddr");
6640 			error = EADDRNOTAVAIL;
6641 			goto ud_error;
6642 		}
6643 		/* Extract port and ipaddr */
6644 		port = sin->sin_port;
6645 		v4dst = sin->sin_addr.s_addr;
6646 		srcid = 0;
6647 		break;
6648 	}
6649 
6650 	mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error, insert_spi);
6651 	if (error != 0) {
6652 ud_error:
6653 		UDP_STAT(us, udp_out_err_output);
6654 		ASSERT(mp != NULL);
6655 		/* mp is freed by the following routine */
6656 		udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
6657 		    (t_scalar_t)error);
6658 	}
6659 }
6660 
6661 /*
6662  * udp_output_v6():
6663  * Assumes that udp_wput did some sanity checking on the destination
6664  * address.
6665  */
6666 static mblk_t *
6667 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error)
6668 {
6669 	ip6_t		*ip6h;
6670 	ip6i_t		*ip6i;	/* mp1->b_rptr even if no ip6i_t */
6671 	mblk_t		*mp1 = mp;
6672 	mblk_t		*mp2;
6673 	int		udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
6674 	size_t		ip_len;
6675 	udpha_t		*udph;
6676 	udp_t		*udp = connp->conn_udp;
6677 	queue_t		*q = connp->conn_wq;
6678 	ip6_pkt_t	ipp_s;	/* For ancillary data options */
6679 	ip6_pkt_t	*ipp = &ipp_s;
6680 	ip6_pkt_t	*tipp;	/* temporary ipp */
6681 	uint32_t	csum = 0;
6682 	uint_t		ignore = 0;
6683 	uint_t		option_exists = 0, is_sticky = 0;
6684 	uint8_t		*cp;
6685 	uint8_t		*nxthdr_ptr;
6686 	in6_addr_t	ip6_dst;
6687 	udpattrs_t	attrs;
6688 	boolean_t	opt_present;
6689 	ip6_hbh_t	*hopoptsptr = NULL;
6690 	uint_t		hopoptslen = 0;
6691 	boolean_t	is_ancillary = B_FALSE;
6692 	udp_stack_t	*us = udp->udp_us;
6693 	size_t		sth_wroff = 0;
6694 
6695 	*error = 0;
6696 
6697 	/*
6698 	 * If the local address is a mapped address return
6699 	 * an error.
6700 	 * It would be possible to send an IPv6 packet but the
6701 	 * response would never make it back to the application
6702 	 * since it is bound to a mapped address.
6703 	 */
6704 	if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) {
6705 		*error = EADDRNOTAVAIL;
6706 		goto done;
6707 	}
6708 
6709 	ipp->ipp_fields = 0;
6710 	ipp->ipp_sticky_ignored = 0;
6711 
6712 	/*
6713 	 * If TPI options passed in, feed it for verification and handling
6714 	 */
6715 	attrs.udpattr_credset = B_FALSE;
6716 	opt_present = B_FALSE;
6717 	if (DB_TYPE(mp) != M_DATA) {
6718 		mp1 = mp->b_cont;
6719 		if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) {
6720 			attrs.udpattr_ipp6 = ipp;
6721 			attrs.udpattr_mb = mp;
6722 			if (udp_unitdata_opt_process(q, mp, error,
6723 			    &attrs) < 0) {
6724 				goto done;
6725 			}
6726 			ASSERT(*error == 0);
6727 			opt_present = B_TRUE;
6728 		}
6729 	}
6730 	rw_enter(&udp->udp_rwlock, RW_READER);
6731 	ignore = ipp->ipp_sticky_ignored;
6732 
6733 	/* mp1 points to the M_DATA mblk carrying the packet */
6734 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
6735 
6736 	if (sin6->sin6_scope_id != 0 &&
6737 	    IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
6738 		/*
6739 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
6740 		 * option nor ancillary data.  It needs to be
6741 		 * explicitly set in options_exists.
6742 		 */
6743 		option_exists |= IPPF_SCOPE_ID;
6744 	}
6745 
6746 	/*
6747 	 * Compute the destination address
6748 	 */
6749 	ip6_dst = sin6->sin6_addr;
6750 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
6751 		ip6_dst = ipv6_loopback;
6752 
6753 	/*
6754 	 * If we're not going to the same destination as last time, then
6755 	 * recompute the label required.  This is done in a separate routine to
6756 	 * avoid blowing up our stack here.
6757 	 *
6758 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
6759 	 * to different destination may require different labels,
6760 	 * or worse, UDP packets to same IP address may require
6761 	 * different labels due to use of shared all-zones address.
6762 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
6763 	 * and sticky ipp_hopoptslen are consistent for the current
6764 	 * destination and are updated atomically.
6765 	 */
6766 	mutex_enter(&connp->conn_lock);
6767 	if (is_system_labeled()) {
6768 		/* Using UDP MLP requires SCM_UCRED from user */
6769 		if (connp->conn_mlp_type != mlptSingle &&
6770 		    !attrs.udpattr_credset) {
6771 			DTRACE_PROBE4(
6772 			    tx__ip__log__info__output__udp6,
6773 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
6774 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
6775 			*error = ECONNREFUSED;
6776 			rw_exit(&udp->udp_rwlock);
6777 			mutex_exit(&connp->conn_lock);
6778 			goto done;
6779 		}
6780 		/*
6781 		 * update label option for this UDP socket if
6782 		 * - the destination has changed, or
6783 		 * - the UDP socket is MLP
6784 		 */
6785 		if ((opt_present ||
6786 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst) ||
6787 		    connp->conn_mlp_type != mlptSingle) &&
6788 		    (*error = udp_update_label_v6(q, mp, &ip6_dst)) != 0) {
6789 			rw_exit(&udp->udp_rwlock);
6790 			mutex_exit(&connp->conn_lock);
6791 			goto done;
6792 		}
6793 	}
6794 
6795 	/*
6796 	 * If there's a security label here, then we ignore any options the
6797 	 * user may try to set.  We keep the peer's label as a hidden sticky
6798 	 * option. We make a private copy of this label before releasing the
6799 	 * lock so that label is kept consistent with the destination addr.
6800 	 */
6801 	if (udp->udp_label_len_v6 > 0) {
6802 		ignore &= ~IPPF_HOPOPTS;
6803 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
6804 	}
6805 
6806 	if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) {
6807 		/* No sticky options nor ancillary data. */
6808 		mutex_exit(&connp->conn_lock);
6809 		goto no_options;
6810 	}
6811 
6812 	/*
6813 	 * Go through the options figuring out where each is going to
6814 	 * come from and build two masks.  The first mask indicates if
6815 	 * the option exists at all.  The second mask indicates if the
6816 	 * option is sticky or ancillary.
6817 	 */
6818 	if (!(ignore & IPPF_HOPOPTS)) {
6819 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
6820 			option_exists |= IPPF_HOPOPTS;
6821 			udp_ip_hdr_len += ipp->ipp_hopoptslen;
6822 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
6823 			option_exists |= IPPF_HOPOPTS;
6824 			is_sticky |= IPPF_HOPOPTS;
6825 			ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0);
6826 			hopoptsptr = kmem_alloc(
6827 			    udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
6828 			if (hopoptsptr == NULL) {
6829 				*error = ENOMEM;
6830 				mutex_exit(&connp->conn_lock);
6831 				goto done;
6832 			}
6833 			hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen;
6834 			bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr,
6835 			    hopoptslen);
6836 			udp_ip_hdr_len += hopoptslen;
6837 		}
6838 	}
6839 	mutex_exit(&connp->conn_lock);
6840 
6841 	if (!(ignore & IPPF_RTHDR)) {
6842 		if (ipp->ipp_fields & IPPF_RTHDR) {
6843 			option_exists |= IPPF_RTHDR;
6844 			udp_ip_hdr_len += ipp->ipp_rthdrlen;
6845 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
6846 			option_exists |= IPPF_RTHDR;
6847 			is_sticky |= IPPF_RTHDR;
6848 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen;
6849 		}
6850 	}
6851 
6852 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
6853 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
6854 			option_exists |= IPPF_RTDSTOPTS;
6855 			udp_ip_hdr_len += ipp->ipp_rtdstoptslen;
6856 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
6857 			option_exists |= IPPF_RTDSTOPTS;
6858 			is_sticky |= IPPF_RTDSTOPTS;
6859 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen;
6860 		}
6861 	}
6862 
6863 	if (!(ignore & IPPF_DSTOPTS)) {
6864 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
6865 			option_exists |= IPPF_DSTOPTS;
6866 			udp_ip_hdr_len += ipp->ipp_dstoptslen;
6867 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
6868 			option_exists |= IPPF_DSTOPTS;
6869 			is_sticky |= IPPF_DSTOPTS;
6870 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen;
6871 		}
6872 	}
6873 
6874 	if (!(ignore & IPPF_IFINDEX)) {
6875 		if (ipp->ipp_fields & IPPF_IFINDEX) {
6876 			option_exists |= IPPF_IFINDEX;
6877 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
6878 			option_exists |= IPPF_IFINDEX;
6879 			is_sticky |= IPPF_IFINDEX;
6880 		}
6881 	}
6882 
6883 	if (!(ignore & IPPF_ADDR)) {
6884 		if (ipp->ipp_fields & IPPF_ADDR) {
6885 			option_exists |= IPPF_ADDR;
6886 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) {
6887 			option_exists |= IPPF_ADDR;
6888 			is_sticky |= IPPF_ADDR;
6889 		}
6890 	}
6891 
6892 	if (!(ignore & IPPF_DONTFRAG)) {
6893 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
6894 			option_exists |= IPPF_DONTFRAG;
6895 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
6896 			option_exists |= IPPF_DONTFRAG;
6897 			is_sticky |= IPPF_DONTFRAG;
6898 		}
6899 	}
6900 
6901 	if (!(ignore & IPPF_USE_MIN_MTU)) {
6902 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
6903 			option_exists |= IPPF_USE_MIN_MTU;
6904 		} else if (udp->udp_sticky_ipp.ipp_fields &
6905 		    IPPF_USE_MIN_MTU) {
6906 			option_exists |= IPPF_USE_MIN_MTU;
6907 			is_sticky |= IPPF_USE_MIN_MTU;
6908 		}
6909 	}
6910 
6911 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
6912 		option_exists |= IPPF_HOPLIMIT;
6913 	/* IPV6_HOPLIMIT can never be sticky */
6914 	ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
6915 
6916 	if (!(ignore & IPPF_UNICAST_HOPS) &&
6917 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
6918 		option_exists |= IPPF_UNICAST_HOPS;
6919 		is_sticky |= IPPF_UNICAST_HOPS;
6920 	}
6921 
6922 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
6923 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
6924 		option_exists |= IPPF_MULTICAST_HOPS;
6925 		is_sticky |= IPPF_MULTICAST_HOPS;
6926 	}
6927 
6928 	if (!(ignore & IPPF_TCLASS)) {
6929 		if (ipp->ipp_fields & IPPF_TCLASS) {
6930 			option_exists |= IPPF_TCLASS;
6931 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
6932 			option_exists |= IPPF_TCLASS;
6933 			is_sticky |= IPPF_TCLASS;
6934 		}
6935 	}
6936 
6937 	if (!(ignore & IPPF_NEXTHOP) &&
6938 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_NEXTHOP)) {
6939 		option_exists |= IPPF_NEXTHOP;
6940 		is_sticky |= IPPF_NEXTHOP;
6941 	}
6942 
6943 no_options:
6944 
6945 	/*
6946 	 * If any options carried in the ip6i_t were specified, we
6947 	 * need to account for the ip6i_t in the data we'll be sending
6948 	 * down.
6949 	 */
6950 	if (option_exists & IPPF_HAS_IP6I)
6951 		udp_ip_hdr_len += sizeof (ip6i_t);
6952 
6953 	/* check/fix buffer config, setup pointers into it */
6954 	ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len];
6955 	if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) ||
6956 	    !OK_32PTR(ip6h)) {
6957 
6958 		/* Try to get everything in a single mblk next time */
6959 		if (udp_ip_hdr_len > udp->udp_max_hdr_len) {
6960 			udp->udp_max_hdr_len = udp_ip_hdr_len;
6961 			sth_wroff = udp->udp_max_hdr_len + us->us_wroff_extra;
6962 		}
6963 
6964 		mp2 = allocb(udp_ip_hdr_len + us->us_wroff_extra, BPRI_LO);
6965 		if (mp2 == NULL) {
6966 			*error = ENOMEM;
6967 			rw_exit(&udp->udp_rwlock);
6968 			goto done;
6969 		}
6970 		mp2->b_wptr = DB_LIM(mp2);
6971 		mp2->b_cont = mp1;
6972 		mp1 = mp2;
6973 		if (DB_TYPE(mp) != M_DATA)
6974 			mp->b_cont = mp1;
6975 		else
6976 			mp = mp1;
6977 
6978 		ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len);
6979 	}
6980 	mp1->b_rptr = (unsigned char *)ip6h;
6981 	ip6i = (ip6i_t *)ip6h;
6982 
6983 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp)
6984 	if (option_exists & IPPF_HAS_IP6I) {
6985 		ip6h = (ip6_t *)&ip6i[1];
6986 		ip6i->ip6i_flags = 0;
6987 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
6988 
6989 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
6990 		if (option_exists & IPPF_SCOPE_ID) {
6991 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6992 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
6993 		} else if (option_exists & IPPF_IFINDEX) {
6994 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
6995 			ASSERT(tipp->ipp_ifindex != 0);
6996 			ip6i->ip6i_flags |= IP6I_IFINDEX;
6997 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
6998 		}
6999 
7000 		if (option_exists & IPPF_ADDR) {
7001 			/*
7002 			 * Enable per-packet source address verification if
7003 			 * IPV6_PKTINFO specified the source address.
7004 			 * ip6_src is set in the transport's _wput function.
7005 			 */
7006 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
7007 		}
7008 
7009 		if (option_exists & IPPF_DONTFRAG) {
7010 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
7011 		}
7012 
7013 		if (option_exists & IPPF_USE_MIN_MTU) {
7014 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
7015 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
7016 		}
7017 
7018 		if (option_exists & IPPF_NEXTHOP) {
7019 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
7020 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
7021 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
7022 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
7023 		}
7024 
7025 		/*
7026 		 * tell IP this is an ip6i_t private header
7027 		 */
7028 		ip6i->ip6i_nxt = IPPROTO_RAW;
7029 	}
7030 
7031 	/* Initialize IPv6 header */
7032 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
7033 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
7034 
7035 	/* Set the hoplimit of the outgoing packet. */
7036 	if (option_exists & IPPF_HOPLIMIT) {
7037 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
7038 		ip6h->ip6_hops = ipp->ipp_hoplimit;
7039 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7040 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
7041 		ip6h->ip6_hops = udp->udp_multicast_ttl;
7042 		if (option_exists & IPPF_MULTICAST_HOPS)
7043 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7044 	} else {
7045 		ip6h->ip6_hops = udp->udp_ttl;
7046 		if (option_exists & IPPF_UNICAST_HOPS)
7047 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7048 	}
7049 
7050 	if (option_exists & IPPF_ADDR) {
7051 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
7052 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
7053 		ip6h->ip6_src = tipp->ipp_addr;
7054 	} else {
7055 		/*
7056 		 * The source address was not set using IPV6_PKTINFO.
7057 		 * First look at the bound source.
7058 		 * If unspecified fallback to __sin6_src_id.
7059 		 */
7060 		ip6h->ip6_src = udp->udp_v6src;
7061 		if (sin6->__sin6_src_id != 0 &&
7062 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
7063 			ip_srcid_find_id(sin6->__sin6_src_id,
7064 			    &ip6h->ip6_src, connp->conn_zoneid,
7065 			    us->us_netstack);
7066 		}
7067 	}
7068 
7069 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
7070 	cp = (uint8_t *)&ip6h[1];
7071 
7072 	/*
7073 	 * Here's where we have to start stringing together
7074 	 * any extension headers in the right order:
7075 	 * Hop-by-hop, destination, routing, and final destination opts.
7076 	 */
7077 	if (option_exists & IPPF_HOPOPTS) {
7078 		/* Hop-by-hop options */
7079 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
7080 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
7081 		if (hopoptslen == 0) {
7082 			hopoptsptr = tipp->ipp_hopopts;
7083 			hopoptslen = tipp->ipp_hopoptslen;
7084 			is_ancillary = B_TRUE;
7085 		}
7086 
7087 		*nxthdr_ptr = IPPROTO_HOPOPTS;
7088 		nxthdr_ptr = &hbh->ip6h_nxt;
7089 
7090 		bcopy(hopoptsptr, cp, hopoptslen);
7091 		cp += hopoptslen;
7092 
7093 		if (hopoptsptr != NULL && !is_ancillary) {
7094 			kmem_free(hopoptsptr, hopoptslen);
7095 			hopoptsptr = NULL;
7096 			hopoptslen = 0;
7097 		}
7098 	}
7099 	/*
7100 	 * En-route destination options
7101 	 * Only do them if there's a routing header as well
7102 	 */
7103 	if (option_exists & IPPF_RTDSTOPTS) {
7104 		ip6_dest_t *dst = (ip6_dest_t *)cp;
7105 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
7106 
7107 		*nxthdr_ptr = IPPROTO_DSTOPTS;
7108 		nxthdr_ptr = &dst->ip6d_nxt;
7109 
7110 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
7111 		cp += tipp->ipp_rtdstoptslen;
7112 	}
7113 	/*
7114 	 * Routing header next
7115 	 */
7116 	if (option_exists & IPPF_RTHDR) {
7117 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
7118 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
7119 
7120 		*nxthdr_ptr = IPPROTO_ROUTING;
7121 		nxthdr_ptr = &rt->ip6r_nxt;
7122 
7123 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
7124 		cp += tipp->ipp_rthdrlen;
7125 	}
7126 	/*
7127 	 * Do ultimate destination options
7128 	 */
7129 	if (option_exists & IPPF_DSTOPTS) {
7130 		ip6_dest_t *dest = (ip6_dest_t *)cp;
7131 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
7132 
7133 		*nxthdr_ptr = IPPROTO_DSTOPTS;
7134 		nxthdr_ptr = &dest->ip6d_nxt;
7135 
7136 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
7137 		cp += tipp->ipp_dstoptslen;
7138 	}
7139 	/*
7140 	 * Now set the last header pointer to the proto passed in
7141 	 */
7142 	ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE));
7143 	*nxthdr_ptr = IPPROTO_UDP;
7144 
7145 	/* Update UDP header */
7146 	udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE);
7147 	udph->uha_dst_port = sin6->sin6_port;
7148 	udph->uha_src_port = udp->udp_port;
7149 
7150 	/*
7151 	 * Copy in the destination address
7152 	 */
7153 	ip6h->ip6_dst = ip6_dst;
7154 
7155 	ip6h->ip6_vcf =
7156 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
7157 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
7158 
7159 	if (option_exists & IPPF_TCLASS) {
7160 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
7161 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
7162 		    tipp->ipp_tclass);
7163 	}
7164 	rw_exit(&udp->udp_rwlock);
7165 
7166 	if (option_exists & IPPF_RTHDR) {
7167 		ip6_rthdr_t	*rth;
7168 
7169 		/*
7170 		 * Perform any processing needed for source routing.
7171 		 * We know that all extension headers will be in the same mblk
7172 		 * as the IPv6 header.
7173 		 */
7174 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
7175 		if (rth != NULL && rth->ip6r_segleft != 0) {
7176 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
7177 				/*
7178 				 * Drop packet - only support Type 0 routing.
7179 				 * Notify the application as well.
7180 				 */
7181 				*error = EPROTO;
7182 				goto done;
7183 			}
7184 
7185 			/*
7186 			 * rth->ip6r_len is twice the number of
7187 			 * addresses in the header. Thus it must be even.
7188 			 */
7189 			if (rth->ip6r_len & 0x1) {
7190 				*error = EPROTO;
7191 				goto done;
7192 			}
7193 			/*
7194 			 * Shuffle the routing header and ip6_dst
7195 			 * addresses, and get the checksum difference
7196 			 * between the first hop (in ip6_dst) and
7197 			 * the destination (in the last routing hdr entry).
7198 			 */
7199 			csum = ip_massage_options_v6(ip6h, rth,
7200 			    us->us_netstack);
7201 			/*
7202 			 * Verify that the first hop isn't a mapped address.
7203 			 * Routers along the path need to do this verification
7204 			 * for subsequent hops.
7205 			 */
7206 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
7207 				*error = EADDRNOTAVAIL;
7208 				goto done;
7209 			}
7210 
7211 			cp += (rth->ip6r_len + 1)*8;
7212 		}
7213 	}
7214 
7215 	/* count up length of UDP packet */
7216 	ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN;
7217 	if ((mp2 = mp1->b_cont) != NULL) {
7218 		do {
7219 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
7220 			ip_len += (uint32_t)MBLKL(mp2);
7221 		} while ((mp2 = mp2->b_cont) != NULL);
7222 	}
7223 
7224 	/*
7225 	 * If the size of the packet is greater than the maximum allowed by
7226 	 * ip, return an error. Passing this down could cause panics because
7227 	 * the size will have wrapped and be inconsistent with the msg size.
7228 	 */
7229 	if (ip_len > IP_MAXPACKET) {
7230 		*error = EMSGSIZE;
7231 		goto done;
7232 	}
7233 
7234 	/* Store the UDP length. Subtract length of extension hdrs */
7235 	udph->uha_length = htons(ip_len + IPV6_HDR_LEN -
7236 	    (int)((uchar_t *)udph - (uchar_t *)ip6h));
7237 
7238 	/*
7239 	 * We make it easy for IP to include our pseudo header
7240 	 * by putting our length in uh_checksum, modified (if
7241 	 * we have a routing header) by the checksum difference
7242 	 * between the ultimate destination and first hop addresses.
7243 	 * Note: UDP over IPv6 must always checksum the packet.
7244 	 */
7245 	csum += udph->uha_length;
7246 	csum = (csum & 0xFFFF) + (csum >> 16);
7247 	udph->uha_checksum = (uint16_t)csum;
7248 
7249 #ifdef _LITTLE_ENDIAN
7250 	ip_len = htons(ip_len);
7251 #endif
7252 	ip6h->ip6_plen = ip_len;
7253 	if (DB_CRED(mp) != NULL)
7254 		mblk_setcred(mp1, DB_CRED(mp));
7255 
7256 	if (DB_TYPE(mp) != M_DATA) {
7257 		ASSERT(mp != mp1);
7258 		freeb(mp);
7259 	}
7260 
7261 	/* mp has been consumed and we'll return success */
7262 	ASSERT(*error == 0);
7263 	mp = NULL;
7264 
7265 	/* We're done. Pass the packet to IP */
7266 	BUMP_MIB(&us->us_udp_mib, udpHCOutDatagrams);
7267 	ip_output_v6(connp, mp1, q, IP_WPUT);
7268 
7269 done:
7270 	if (sth_wroff != 0) {
7271 		(void) mi_set_sth_wroff(RD(q),
7272 		    udp->udp_max_hdr_len + us->us_wroff_extra);
7273 	}
7274 	if (hopoptsptr != NULL && !is_ancillary) {
7275 		kmem_free(hopoptsptr, hopoptslen);
7276 		hopoptsptr = NULL;
7277 	}
7278 	if (*error != 0) {
7279 		ASSERT(mp != NULL);
7280 		BUMP_MIB(&us->us_udp_mib, udpOutErrors);
7281 	}
7282 	return (mp);
7283 }
7284 
7285 
7286 static int
7287 udp_getpeername(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
7288 {
7289 	sin_t *sin = (sin_t *)sa;
7290 	sin6_t *sin6 = (sin6_t *)sa;
7291 
7292 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
7293 
7294 	if (udp->udp_state != TS_DATA_XFER)
7295 		return (ENOTCONN);
7296 
7297 	switch (udp->udp_family) {
7298 	case AF_INET:
7299 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
7300 
7301 		if (*salenp < sizeof (sin_t))
7302 			return (EINVAL);
7303 
7304 		*salenp = sizeof (sin_t);
7305 		*sin = sin_null;
7306 		sin->sin_family = AF_INET;
7307 		sin->sin_port = udp->udp_dstport;
7308 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6dst);
7309 		break;
7310 
7311 	case AF_INET6:
7312 		if (*salenp < sizeof (sin6_t))
7313 			return (EINVAL);
7314 
7315 		*salenp = sizeof (sin6_t);
7316 		*sin6 = sin6_null;
7317 		sin6->sin6_family = AF_INET6;
7318 		sin6->sin6_port = udp->udp_dstport;
7319 		sin6->sin6_addr = udp->udp_v6dst;
7320 		sin6->sin6_flowinfo = udp->udp_flowinfo;
7321 		break;
7322 	}
7323 
7324 	return (0);
7325 }
7326 
7327 static int
7328 udp_getmyname(udp_t *udp, struct sockaddr *sa, uint_t *salenp)
7329 {
7330 	sin_t *sin = (sin_t *)sa;
7331 	sin6_t *sin6 = (sin6_t *)sa;
7332 
7333 	ASSERT(RW_LOCK_HELD(&udp->udp_rwlock));
7334 
7335 	switch (udp->udp_family) {
7336 	case AF_INET:
7337 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
7338 
7339 		if (*salenp < sizeof (sin_t))
7340 			return (EINVAL);
7341 
7342 		*salenp = sizeof (sin_t);
7343 		*sin = sin_null;
7344 		sin->sin_family = AF_INET;
7345 		sin->sin_port = udp->udp_port;
7346 
7347 		/*
7348 		 * If udp_v6src is unspecified, we might be bound to broadcast
7349 		 * / multicast.  Use udp_bound_v6src as local address instead
7350 		 * (that could also still be unspecified).
7351 		 */
7352 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
7353 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
7354 			sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_v6src);
7355 		} else {
7356 			sin->sin_addr.s_addr =
7357 			    V4_PART_OF_V6(udp->udp_bound_v6src);
7358 		}
7359 		break;
7360 
7361 	case AF_INET6:
7362 		if (*salenp < sizeof (sin6_t))
7363 			return (EINVAL);
7364 
7365 		*salenp = sizeof (sin6_t);
7366 		*sin6 = sin6_null;
7367 		sin6->sin6_family = AF_INET6;
7368 		sin6->sin6_port = udp->udp_port;
7369 		sin6->sin6_flowinfo = udp->udp_flowinfo;
7370 
7371 		/*
7372 		 * If udp_v6src is unspecified, we might be bound to broadcast
7373 		 * / multicast.  Use udp_bound_v6src as local address instead
7374 		 * (that could also still be unspecified).
7375 		 */
7376 		if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))
7377 			sin6->sin6_addr = udp->udp_v6src;
7378 		else
7379 			sin6->sin6_addr = udp->udp_bound_v6src;
7380 		break;
7381 	}
7382 
7383 	return (0);
7384 }
7385 
7386 /*
7387  * Handle special out-of-band ioctl requests (see PSARC/2008/265).
7388  */
7389 static void
7390 udp_wput_cmdblk(queue_t *q, mblk_t *mp)
7391 {
7392 	void	*data;
7393 	mblk_t	*datamp = mp->b_cont;
7394 	udp_t	*udp = Q_TO_UDP(q);
7395 	cmdblk_t *cmdp = (cmdblk_t *)mp->b_rptr;
7396 
7397 	if (datamp == NULL || MBLKL(datamp) < cmdp->cb_len) {
7398 		cmdp->cb_error = EPROTO;
7399 		qreply(q, mp);
7400 		return;
7401 	}
7402 	data = datamp->b_rptr;
7403 
7404 	rw_enter(&udp->udp_rwlock, RW_READER);
7405 	switch (cmdp->cb_cmd) {
7406 	case TI_GETPEERNAME:
7407 		cmdp->cb_error = udp_getpeername(udp, data, &cmdp->cb_len);
7408 		break;
7409 	case TI_GETMYNAME:
7410 		cmdp->cb_error = udp_getmyname(udp, data, &cmdp->cb_len);
7411 		break;
7412 	default:
7413 		cmdp->cb_error = EINVAL;
7414 		break;
7415 	}
7416 	rw_exit(&udp->udp_rwlock);
7417 
7418 	qreply(q, mp);
7419 }
7420 
7421 static void
7422 udp_wput_other(queue_t *q, mblk_t *mp)
7423 {
7424 	uchar_t	*rptr = mp->b_rptr;
7425 	struct datab *db;
7426 	struct iocblk *iocp;
7427 	cred_t	*cr;
7428 	conn_t	*connp = Q_TO_CONN(q);
7429 	udp_t	*udp = connp->conn_udp;
7430 	udp_stack_t *us;
7431 
7432 	TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START,
7433 	    "udp_wput_other_start: q %p", q);
7434 
7435 	us = udp->udp_us;
7436 	db = mp->b_datap;
7437 
7438 	cr = DB_CREDDEF(mp, connp->conn_cred);
7439 
7440 	switch (db->db_type) {
7441 	case M_CMD:
7442 		udp_wput_cmdblk(q, mp);
7443 		return;
7444 
7445 	case M_PROTO:
7446 	case M_PCPROTO:
7447 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
7448 			freemsg(mp);
7449 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7450 			    "udp_wput_other_end: q %p (%S)", q, "protoshort");
7451 			return;
7452 		}
7453 		switch (((t_primp_t)rptr)->type) {
7454 		case T_ADDR_REQ:
7455 			udp_addr_req(q, mp);
7456 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7457 			    "udp_wput_other_end: q %p (%S)", q, "addrreq");
7458 			return;
7459 		case O_T_BIND_REQ:
7460 		case T_BIND_REQ:
7461 			udp_bind(q, mp);
7462 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7463 			    "udp_wput_other_end: q %p (%S)", q, "bindreq");
7464 			return;
7465 		case T_CONN_REQ:
7466 			udp_connect(q, mp);
7467 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7468 			    "udp_wput_other_end: q %p (%S)", q, "connreq");
7469 			return;
7470 		case T_CAPABILITY_REQ:
7471 			udp_capability_req(q, mp);
7472 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7473 			    "udp_wput_other_end: q %p (%S)", q, "capabreq");
7474 			return;
7475 		case T_INFO_REQ:
7476 			udp_info_req(q, mp);
7477 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7478 			    "udp_wput_other_end: q %p (%S)", q, "inforeq");
7479 			return;
7480 		case T_UNITDATA_REQ:
7481 			/*
7482 			 * If a T_UNITDATA_REQ gets here, the address must
7483 			 * be bad.  Valid T_UNITDATA_REQs are handled
7484 			 * in udp_wput.
7485 			 */
7486 			udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL);
7487 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7488 			    "udp_wput_other_end: q %p (%S)", q, "unitdatareq");
7489 			return;
7490 		case T_UNBIND_REQ:
7491 			udp_unbind(q, mp);
7492 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7493 			    "udp_wput_other_end: q %p (%S)", q, "unbindreq");
7494 			return;
7495 		case T_SVR4_OPTMGMT_REQ:
7496 			if (!snmpcom_req(q, mp, udp_snmp_set, ip_snmp_get,
7497 			    cr)) {
7498 				(void) svr4_optcom_req(q,
7499 				    mp, cr, &udp_opt_obj, B_TRUE);
7500 			}
7501 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7502 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
7503 			return;
7504 
7505 		case T_OPTMGMT_REQ:
7506 			(void) tpi_optcom_req(q, mp, cr, &udp_opt_obj, B_TRUE);
7507 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7508 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
7509 			return;
7510 
7511 		case T_DISCON_REQ:
7512 			udp_disconnect(q, mp);
7513 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7514 			    "udp_wput_other_end: q %p (%S)", q, "disconreq");
7515 			return;
7516 
7517 		/* The following TPI message is not supported by udp. */
7518 		case O_T_CONN_RES:
7519 		case T_CONN_RES:
7520 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
7521 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7522 			    "udp_wput_other_end: q %p (%S)", q,
7523 			    "connres/disconreq");
7524 			return;
7525 
7526 		/* The following 3 TPI messages are illegal for udp. */
7527 		case T_DATA_REQ:
7528 		case T_EXDATA_REQ:
7529 		case T_ORDREL_REQ:
7530 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
7531 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7532 			    "udp_wput_other_end: q %p (%S)", q,
7533 			    "data/exdata/ordrel");
7534 			return;
7535 		default:
7536 			break;
7537 		}
7538 		break;
7539 	case M_FLUSH:
7540 		if (*rptr & FLUSHW)
7541 			flushq(q, FLUSHDATA);
7542 		break;
7543 	case M_IOCTL:
7544 		iocp = (struct iocblk *)mp->b_rptr;
7545 		switch (iocp->ioc_cmd) {
7546 		case TI_GETPEERNAME:
7547 			if (udp->udp_state != TS_DATA_XFER) {
7548 				/*
7549 				 * If a default destination address has not
7550 				 * been associated with the stream, then we
7551 				 * don't know the peer's name.
7552 				 */
7553 				iocp->ioc_error = ENOTCONN;
7554 				iocp->ioc_count = 0;
7555 				mp->b_datap->db_type = M_IOCACK;
7556 				qreply(q, mp);
7557 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7558 				    "udp_wput_other_end: q %p (%S)", q,
7559 				    "getpeername");
7560 				return;
7561 			}
7562 			/* FALLTHRU */
7563 		case TI_GETMYNAME: {
7564 			/*
7565 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
7566 			 * need to copyin the user's strbuf structure.
7567 			 * Processing will continue in the M_IOCDATA case
7568 			 * below.
7569 			 */
7570 			mi_copyin(q, mp, NULL,
7571 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
7572 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7573 			    "udp_wput_other_end: q %p (%S)", q, "getmyname");
7574 			return;
7575 			}
7576 		case ND_SET:
7577 			/* nd_getset performs the necessary checking */
7578 		case ND_GET:
7579 			if (nd_getset(q, us->us_nd, mp)) {
7580 				qreply(q, mp);
7581 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7582 				    "udp_wput_other_end: q %p (%S)", q, "get");
7583 				return;
7584 			}
7585 			break;
7586 		case _SIOCSOCKFALLBACK:
7587 			/*
7588 			 * Either sockmod is about to be popped and the
7589 			 * socket would now be treated as a plain stream,
7590 			 * or a module is about to be pushed so we could
7591 			 * no longer use read-side synchronous stream.
7592 			 * Drain any queued data and disable direct sockfs
7593 			 * interface from now on.
7594 			 */
7595 			if (!udp->udp_issocket) {
7596 				DB_TYPE(mp) = M_IOCNAK;
7597 				iocp->ioc_error = EINVAL;
7598 			} else {
7599 				udp->udp_issocket = B_FALSE;
7600 				if (udp->udp_direct_sockfs) {
7601 					/*
7602 					 * Disable read-side synchronous
7603 					 * stream interface and drain any
7604 					 * queued data.
7605 					 */
7606 					udp_rcv_drain(RD(q), udp,
7607 					    B_FALSE);
7608 					ASSERT(!udp->udp_direct_sockfs);
7609 					UDP_STAT(us, udp_sock_fallback);
7610 				}
7611 				DB_TYPE(mp) = M_IOCACK;
7612 				iocp->ioc_error = 0;
7613 			}
7614 			iocp->ioc_count = 0;
7615 			iocp->ioc_rval = 0;
7616 			qreply(q, mp);
7617 			return;
7618 		default:
7619 			break;
7620 		}
7621 		break;
7622 	case M_IOCDATA:
7623 		udp_wput_iocdata(q, mp);
7624 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7625 		    "udp_wput_other_end: q %p (%S)", q, "iocdata");
7626 		return;
7627 	default:
7628 		/* Unrecognized messages are passed through without change. */
7629 		break;
7630 	}
7631 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7632 	    "udp_wput_other_end: q %p (%S)", q, "end");
7633 	ip_output(connp, mp, q, IP_WPUT);
7634 }
7635 
7636 /*
7637  * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA
7638  * messages.
7639  */
7640 static void
7641 udp_wput_iocdata(queue_t *q, mblk_t *mp)
7642 {
7643 	mblk_t	*mp1;
7644 	struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
7645 	STRUCT_HANDLE(strbuf, sb);
7646 	udp_t	*udp = Q_TO_UDP(q);
7647 	int	error;
7648 	uint_t	addrlen;
7649 
7650 	/* Make sure it is one of ours. */
7651 	switch (iocp->ioc_cmd) {
7652 	case TI_GETMYNAME:
7653 	case TI_GETPEERNAME:
7654 		break;
7655 	default:
7656 		ip_output(udp->udp_connp, mp, q, IP_WPUT);
7657 		return;
7658 	}
7659 
7660 	switch (mi_copy_state(q, mp, &mp1)) {
7661 	case -1:
7662 		return;
7663 	case MI_COPY_CASE(MI_COPY_IN, 1):
7664 		break;
7665 	case MI_COPY_CASE(MI_COPY_OUT, 1):
7666 		/*
7667 		 * The address has been copied out, so now
7668 		 * copyout the strbuf.
7669 		 */
7670 		mi_copyout(q, mp);
7671 		return;
7672 	case MI_COPY_CASE(MI_COPY_OUT, 2):
7673 		/*
7674 		 * The address and strbuf have been copied out.
7675 		 * We're done, so just acknowledge the original
7676 		 * M_IOCTL.
7677 		 */
7678 		mi_copy_done(q, mp, 0);
7679 		return;
7680 	default:
7681 		/*
7682 		 * Something strange has happened, so acknowledge
7683 		 * the original M_IOCTL with an EPROTO error.
7684 		 */
7685 		mi_copy_done(q, mp, EPROTO);
7686 		return;
7687 	}
7688 
7689 	/*
7690 	 * Now we have the strbuf structure for TI_GETMYNAME
7691 	 * and TI_GETPEERNAME.  Next we copyout the requested
7692 	 * address and then we'll copyout the strbuf.
7693 	 */
7694 	STRUCT_SET_HANDLE(sb, iocp->ioc_flag, (void *)mp1->b_rptr);
7695 	addrlen = udp->udp_family == AF_INET ? sizeof (sin_t) : sizeof (sin6_t);
7696 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
7697 		mi_copy_done(q, mp, EINVAL);
7698 		return;
7699 	}
7700 
7701 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
7702 	if (mp1 == NULL)
7703 		return;
7704 
7705 	rw_enter(&udp->udp_rwlock, RW_READER);
7706 	switch (iocp->ioc_cmd) {
7707 	case TI_GETMYNAME:
7708 		error = udp_getmyname(udp, (void *)mp1->b_rptr, &addrlen);
7709 		break;
7710 	case TI_GETPEERNAME:
7711 		error = udp_getpeername(udp, (void *)mp1->b_rptr, &addrlen);
7712 		break;
7713 	}
7714 	rw_exit(&udp->udp_rwlock);
7715 
7716 	if (error != 0) {
7717 		mi_copy_done(q, mp, error);
7718 	} else {
7719 		mp1->b_wptr += addrlen;
7720 		STRUCT_FSET(sb, len, addrlen);
7721 
7722 		/* Copy out the address */
7723 		mi_copyout(q, mp);
7724 	}
7725 }
7726 
7727 static int
7728 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
7729     udpattrs_t *udpattrs)
7730 {
7731 	struct T_unitdata_req *udreqp;
7732 	int is_absreq_failure;
7733 	cred_t *cr;
7734 	conn_t	*connp = Q_TO_CONN(q);
7735 
7736 	ASSERT(((t_primp_t)mp->b_rptr)->type);
7737 
7738 	cr = DB_CREDDEF(mp, connp->conn_cred);
7739 
7740 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
7741 
7742 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
7743 	    udreqp->OPT_offset, cr, &udp_opt_obj,
7744 	    udpattrs, &is_absreq_failure);
7745 
7746 	if (*errorp != 0) {
7747 		/*
7748 		 * Note: No special action needed in this
7749 		 * module for "is_absreq_failure"
7750 		 */
7751 		return (-1);		/* failure */
7752 	}
7753 	ASSERT(is_absreq_failure == 0);
7754 	return (0);	/* success */
7755 }
7756 
7757 void
7758 udp_ddi_init(void)
7759 {
7760 	udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr,
7761 	    udp_opt_obj.odb_opt_arr_cnt);
7762 
7763 	/*
7764 	 * We want to be informed each time a stack is created or
7765 	 * destroyed in the kernel, so we can maintain the
7766 	 * set of udp_stack_t's.
7767 	 */
7768 	netstack_register(NS_UDP, udp_stack_init, NULL, udp_stack_fini);
7769 }
7770 
7771 void
7772 udp_ddi_destroy(void)
7773 {
7774 	netstack_unregister(NS_UDP);
7775 }
7776 
7777 /*
7778  * Initialize the UDP stack instance.
7779  */
7780 static void *
7781 udp_stack_init(netstackid_t stackid, netstack_t *ns)
7782 {
7783 	udp_stack_t	*us;
7784 	udpparam_t	*pa;
7785 	int		i;
7786 
7787 	us = (udp_stack_t *)kmem_zalloc(sizeof (*us), KM_SLEEP);
7788 	us->us_netstack = ns;
7789 
7790 	us->us_num_epriv_ports = UDP_NUM_EPRIV_PORTS;
7791 	us->us_epriv_ports[0] = 2049;
7792 	us->us_epriv_ports[1] = 4045;
7793 
7794 	/*
7795 	 * The smallest anonymous port in the priviledged port range which UDP
7796 	 * looks for free port.  Use in the option UDP_ANONPRIVBIND.
7797 	 */
7798 	us->us_min_anonpriv_port = 512;
7799 
7800 	us->us_bind_fanout_size = udp_bind_fanout_size;
7801 
7802 	/* Roundup variable that might have been modified in /etc/system */
7803 	if (us->us_bind_fanout_size & (us->us_bind_fanout_size - 1)) {
7804 		/* Not a power of two. Round up to nearest power of two */
7805 		for (i = 0; i < 31; i++) {
7806 			if (us->us_bind_fanout_size < (1 << i))
7807 				break;
7808 		}
7809 		us->us_bind_fanout_size = 1 << i;
7810 	}
7811 	us->us_bind_fanout = kmem_zalloc(us->us_bind_fanout_size *
7812 	    sizeof (udp_fanout_t), KM_SLEEP);
7813 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7814 		mutex_init(&us->us_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT,
7815 		    NULL);
7816 	}
7817 
7818 	pa = (udpparam_t *)kmem_alloc(sizeof (udp_param_arr), KM_SLEEP);
7819 
7820 	us->us_param_arr = pa;
7821 	bcopy(udp_param_arr, us->us_param_arr, sizeof (udp_param_arr));
7822 
7823 	(void) udp_param_register(&us->us_nd,
7824 	    us->us_param_arr, A_CNT(udp_param_arr));
7825 
7826 	us->us_kstat = udp_kstat2_init(stackid, &us->us_statistics);
7827 	us->us_mibkp = udp_kstat_init(stackid);
7828 	return (us);
7829 }
7830 
7831 /*
7832  * Free the UDP stack instance.
7833  */
7834 static void
7835 udp_stack_fini(netstackid_t stackid, void *arg)
7836 {
7837 	udp_stack_t *us = (udp_stack_t *)arg;
7838 	int i;
7839 
7840 	for (i = 0; i < us->us_bind_fanout_size; i++) {
7841 		mutex_destroy(&us->us_bind_fanout[i].uf_lock);
7842 	}
7843 
7844 	kmem_free(us->us_bind_fanout, us->us_bind_fanout_size *
7845 	    sizeof (udp_fanout_t));
7846 
7847 	us->us_bind_fanout = NULL;
7848 
7849 	nd_free(&us->us_nd);
7850 	kmem_free(us->us_param_arr, sizeof (udp_param_arr));
7851 	us->us_param_arr = NULL;
7852 
7853 	udp_kstat_fini(stackid, us->us_mibkp);
7854 	us->us_mibkp = NULL;
7855 
7856 	udp_kstat2_fini(stackid, us->us_kstat);
7857 	us->us_kstat = NULL;
7858 	bzero(&us->us_statistics, sizeof (us->us_statistics));
7859 	kmem_free(us, sizeof (*us));
7860 }
7861 
7862 static void *
7863 udp_kstat2_init(netstackid_t stackid, udp_stat_t *us_statisticsp)
7864 {
7865 	kstat_t *ksp;
7866 
7867 	udp_stat_t template = {
7868 		{ "udp_ip_send",		KSTAT_DATA_UINT64 },
7869 		{ "udp_ip_ire_send",		KSTAT_DATA_UINT64 },
7870 		{ "udp_ire_null",		KSTAT_DATA_UINT64 },
7871 		{ "udp_drain",			KSTAT_DATA_UINT64 },
7872 		{ "udp_sock_fallback",		KSTAT_DATA_UINT64 },
7873 		{ "udp_rrw_busy",		KSTAT_DATA_UINT64 },
7874 		{ "udp_rrw_msgcnt",		KSTAT_DATA_UINT64 },
7875 		{ "udp_out_sw_cksum",		KSTAT_DATA_UINT64 },
7876 		{ "udp_out_sw_cksum_bytes",	KSTAT_DATA_UINT64 },
7877 		{ "udp_out_opt",		KSTAT_DATA_UINT64 },
7878 		{ "udp_out_err_notconn",	KSTAT_DATA_UINT64 },
7879 		{ "udp_out_err_output",		KSTAT_DATA_UINT64 },
7880 		{ "udp_out_err_tudr",		KSTAT_DATA_UINT64 },
7881 		{ "udp_in_pktinfo",		KSTAT_DATA_UINT64 },
7882 		{ "udp_in_recvdstaddr",		KSTAT_DATA_UINT64 },
7883 		{ "udp_in_recvopts",		KSTAT_DATA_UINT64 },
7884 		{ "udp_in_recvif",		KSTAT_DATA_UINT64 },
7885 		{ "udp_in_recvslla",		KSTAT_DATA_UINT64 },
7886 		{ "udp_in_recvucred",		KSTAT_DATA_UINT64 },
7887 		{ "udp_in_recvttl",		KSTAT_DATA_UINT64 },
7888 		{ "udp_in_recvhopopts",		KSTAT_DATA_UINT64 },
7889 		{ "udp_in_recvhoplimit",	KSTAT_DATA_UINT64 },
7890 		{ "udp_in_recvdstopts",		KSTAT_DATA_UINT64 },
7891 		{ "udp_in_recvrtdstopts",	KSTAT_DATA_UINT64 },
7892 		{ "udp_in_recvrthdr",		KSTAT_DATA_UINT64 },
7893 		{ "udp_in_recvpktinfo",		KSTAT_DATA_UINT64 },
7894 		{ "udp_in_recvtclass",		KSTAT_DATA_UINT64 },
7895 		{ "udp_in_timestamp",		KSTAT_DATA_UINT64 },
7896 #ifdef DEBUG
7897 		{ "udp_data_conn",		KSTAT_DATA_UINT64 },
7898 		{ "udp_data_notconn",		KSTAT_DATA_UINT64 },
7899 #endif
7900 	};
7901 
7902 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, "udpstat", "net",
7903 	    KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
7904 	    KSTAT_FLAG_VIRTUAL, stackid);
7905 
7906 	if (ksp == NULL)
7907 		return (NULL);
7908 
7909 	bcopy(&template, us_statisticsp, sizeof (template));
7910 	ksp->ks_data = (void *)us_statisticsp;
7911 	ksp->ks_private = (void *)(uintptr_t)stackid;
7912 
7913 	kstat_install(ksp);
7914 	return (ksp);
7915 }
7916 
7917 static void
7918 udp_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
7919 {
7920 	if (ksp != NULL) {
7921 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7922 		kstat_delete_netstack(ksp, stackid);
7923 	}
7924 }
7925 
7926 static void *
7927 udp_kstat_init(netstackid_t stackid)
7928 {
7929 	kstat_t	*ksp;
7930 
7931 	udp_named_kstat_t template = {
7932 		{ "inDatagrams",	KSTAT_DATA_UINT64, 0 },
7933 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
7934 		{ "outDatagrams",	KSTAT_DATA_UINT64, 0 },
7935 		{ "entrySize",		KSTAT_DATA_INT32, 0 },
7936 		{ "entry6Size",		KSTAT_DATA_INT32, 0 },
7937 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
7938 	};
7939 
7940 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, UDP_MOD_NAME, "mib2",
7941 	    KSTAT_TYPE_NAMED,
7942 	    NUM_OF_FIELDS(udp_named_kstat_t), 0, stackid);
7943 
7944 	if (ksp == NULL || ksp->ks_data == NULL)
7945 		return (NULL);
7946 
7947 	template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t);
7948 	template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t);
7949 
7950 	bcopy(&template, ksp->ks_data, sizeof (template));
7951 	ksp->ks_update = udp_kstat_update;
7952 	ksp->ks_private = (void *)(uintptr_t)stackid;
7953 
7954 	kstat_install(ksp);
7955 	return (ksp);
7956 }
7957 
7958 static void
7959 udp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
7960 {
7961 	if (ksp != NULL) {
7962 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
7963 		kstat_delete_netstack(ksp, stackid);
7964 	}
7965 }
7966 
7967 static int
7968 udp_kstat_update(kstat_t *kp, int rw)
7969 {
7970 	udp_named_kstat_t *udpkp;
7971 	netstackid_t	stackid = (netstackid_t)(uintptr_t)kp->ks_private;
7972 	netstack_t	*ns;
7973 	udp_stack_t	*us;
7974 
7975 	if ((kp == NULL) || (kp->ks_data == NULL))
7976 		return (EIO);
7977 
7978 	if (rw == KSTAT_WRITE)
7979 		return (EACCES);
7980 
7981 	ns = netstack_find_by_stackid(stackid);
7982 	if (ns == NULL)
7983 		return (-1);
7984 	us = ns->netstack_udp;
7985 	if (us == NULL) {
7986 		netstack_rele(ns);
7987 		return (-1);
7988 	}
7989 	udpkp = (udp_named_kstat_t *)kp->ks_data;
7990 
7991 	udpkp->inDatagrams.value.ui64 =	us->us_udp_mib.udpHCInDatagrams;
7992 	udpkp->inErrors.value.ui32 =	us->us_udp_mib.udpInErrors;
7993 	udpkp->outDatagrams.value.ui64 = us->us_udp_mib.udpHCOutDatagrams;
7994 	udpkp->outErrors.value.ui32 =	us->us_udp_mib.udpOutErrors;
7995 	netstack_rele(ns);
7996 	return (0);
7997 }
7998 
7999 /*
8000  * Read-side synchronous stream info entry point, called as a
8001  * result of handling certain STREAMS ioctl operations.
8002  */
8003 static int
8004 udp_rinfop(queue_t *q, infod_t *dp)
8005 {
8006 	mblk_t	*mp;
8007 	uint_t	cmd = dp->d_cmd;
8008 	int	res = 0;
8009 	int	error = 0;
8010 	udp_t	*udp = Q_TO_UDP(q);
8011 	struct stdata *stp = STREAM(q);
8012 
8013 	mutex_enter(&udp->udp_drain_lock);
8014 	/* If shutdown on read has happened, return nothing */
8015 	mutex_enter(&stp->sd_lock);
8016 	if (stp->sd_flag & STREOF) {
8017 		mutex_exit(&stp->sd_lock);
8018 		goto done;
8019 	}
8020 	mutex_exit(&stp->sd_lock);
8021 
8022 	if ((mp = udp->udp_rcv_list_head) == NULL)
8023 		goto done;
8024 
8025 	ASSERT(DB_TYPE(mp) != M_DATA && mp->b_cont != NULL);
8026 
8027 	if (cmd & INFOD_COUNT) {
8028 		/*
8029 		 * Return the number of messages.
8030 		 */
8031 		dp->d_count += udp->udp_rcv_msgcnt;
8032 		res |= INFOD_COUNT;
8033 	}
8034 	if (cmd & INFOD_BYTES) {
8035 		/*
8036 		 * Return size of all data messages.
8037 		 */
8038 		dp->d_bytes += udp->udp_rcv_cnt;
8039 		res |= INFOD_BYTES;
8040 	}
8041 	if (cmd & INFOD_FIRSTBYTES) {
8042 		/*
8043 		 * Return size of first data message.
8044 		 */
8045 		dp->d_bytes = msgdsize(mp);
8046 		res |= INFOD_FIRSTBYTES;
8047 		dp->d_cmd &= ~INFOD_FIRSTBYTES;
8048 	}
8049 	if (cmd & INFOD_COPYOUT) {
8050 		mblk_t *mp1 = mp->b_cont;
8051 		int n;
8052 		/*
8053 		 * Return data contents of first message.
8054 		 */
8055 		ASSERT(DB_TYPE(mp1) == M_DATA);
8056 		while (mp1 != NULL && dp->d_uiop->uio_resid > 0) {
8057 			n = MIN(dp->d_uiop->uio_resid, MBLKL(mp1));
8058 			if (n != 0 && (error = uiomove((char *)mp1->b_rptr, n,
8059 			    UIO_READ, dp->d_uiop)) != 0) {
8060 				goto done;
8061 			}
8062 			mp1 = mp1->b_cont;
8063 		}
8064 		res |= INFOD_COPYOUT;
8065 		dp->d_cmd &= ~INFOD_COPYOUT;
8066 	}
8067 done:
8068 	mutex_exit(&udp->udp_drain_lock);
8069 
8070 	dp->d_res |= res;
8071 
8072 	return (error);
8073 }
8074 
8075 /*
8076  * Read-side synchronous stream entry point.  This is called as a result
8077  * of recv/read operation done at sockfs, and is guaranteed to execute
8078  * outside of the interrupt thread context.  It returns a single datagram
8079  * (b_cont chain of T_UNITDATA_IND plus data) to the upper layer.
8080  */
8081 static int
8082 udp_rrw(queue_t *q, struiod_t *dp)
8083 {
8084 	mblk_t	*mp;
8085 	udp_t	*udp = Q_TO_UDP(q);
8086 	udp_stack_t *us = udp->udp_us;
8087 
8088 	/*
8089 	 * Dequeue datagram from the head of the list and return
8090 	 * it to caller; also ensure that RSLEEP sd_wakeq flag is
8091 	 * set/cleared depending on whether or not there's data
8092 	 * remaining in the list.
8093 	 */
8094 	mutex_enter(&udp->udp_drain_lock);
8095 	if (!udp->udp_direct_sockfs) {
8096 		mutex_exit(&udp->udp_drain_lock);
8097 		UDP_STAT(us, udp_rrw_busy);
8098 		return (EBUSY);
8099 	}
8100 	if ((mp = udp->udp_rcv_list_head) != NULL) {
8101 		uint_t size = msgdsize(mp);
8102 
8103 		/* Last datagram in the list? */
8104 		if ((udp->udp_rcv_list_head = mp->b_next) == NULL)
8105 			udp->udp_rcv_list_tail = NULL;
8106 		mp->b_next = NULL;
8107 
8108 		udp->udp_rcv_cnt -= size;
8109 		udp->udp_rcv_msgcnt--;
8110 		UDP_STAT(us, udp_rrw_msgcnt);
8111 
8112 		/* No longer flow-controlling? */
8113 		if (udp->udp_rcv_cnt < udp->udp_rcv_hiwat &&
8114 		    udp->udp_rcv_msgcnt < udp->udp_rcv_hiwat)
8115 			udp->udp_drain_qfull = B_FALSE;
8116 	}
8117 	if (udp->udp_rcv_list_head == NULL) {
8118 		/*
8119 		 * Either we just dequeued the last datagram or
8120 		 * we get here from sockfs and have nothing to
8121 		 * return; in this case clear RSLEEP.
8122 		 */
8123 		ASSERT(udp->udp_rcv_cnt == 0);
8124 		ASSERT(udp->udp_rcv_msgcnt == 0);
8125 		ASSERT(udp->udp_rcv_list_tail == NULL);
8126 		STR_WAKEUP_CLEAR(STREAM(q));
8127 	} else {
8128 		/*
8129 		 * More data follows; we need udp_rrw() to be
8130 		 * called in future to pick up the rest.
8131 		 */
8132 		STR_WAKEUP_SET(STREAM(q));
8133 	}
8134 	mutex_exit(&udp->udp_drain_lock);
8135 	dp->d_mp = mp;
8136 	return (0);
8137 }
8138 
8139 /*
8140  * Enqueue a completely-built T_UNITDATA_IND message into the receive
8141  * list; this is typically executed within the interrupt thread context
8142  * and so we do things as quickly as possible.
8143  */
8144 static void
8145 udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, uint_t pkt_len)
8146 {
8147 	ASSERT(q == RD(q));
8148 	ASSERT(pkt_len == msgdsize(mp));
8149 	ASSERT(mp->b_next == NULL && mp->b_cont != NULL);
8150 	ASSERT(DB_TYPE(mp) == M_PROTO && DB_TYPE(mp->b_cont) == M_DATA);
8151 	ASSERT(MBLKL(mp) >= sizeof (struct T_unitdata_ind));
8152 
8153 	mutex_enter(&udp->udp_drain_lock);
8154 	/*
8155 	 * Wake up and signal the receiving app; it is okay to do this
8156 	 * before enqueueing the mp because we are holding the drain lock.
8157 	 * One of the advantages of synchronous stream is the ability for
8158 	 * us to find out when the application performs a read on the
8159 	 * socket by way of udp_rrw() entry point being called.  We need
8160 	 * to generate SIGPOLL/SIGIO for each received data in the case
8161 	 * of asynchronous socket just as in the strrput() case.  However,
8162 	 * we only wake the application up when necessary, i.e. during the
8163 	 * first enqueue.  When udp_rrw() is called, we send up a single
8164 	 * datagram upstream and call STR_WAKEUP_SET() again when there
8165 	 * are still data remaining in our receive queue.
8166 	 */
8167 	STR_WAKEUP_SENDSIG(STREAM(q), udp->udp_rcv_list_head);
8168 	if (udp->udp_rcv_list_head == NULL)
8169 		udp->udp_rcv_list_head = mp;
8170 	else
8171 		udp->udp_rcv_list_tail->b_next = mp;
8172 	udp->udp_rcv_list_tail = mp;
8173 	udp->udp_rcv_cnt += pkt_len;
8174 	udp->udp_rcv_msgcnt++;
8175 
8176 	/* Need to flow-control? */
8177 	if (udp->udp_rcv_cnt >= udp->udp_rcv_hiwat ||
8178 	    udp->udp_rcv_msgcnt >= udp->udp_rcv_hiwat)
8179 		udp->udp_drain_qfull = B_TRUE;
8180 
8181 	mutex_exit(&udp->udp_drain_lock);
8182 }
8183 
8184 /*
8185  * Drain the contents of receive list to the module upstream; we do
8186  * this during close or when we fallback to the slow mode due to
8187  * sockmod being popped or a module being pushed on top of us.
8188  */
8189 static void
8190 udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing)
8191 {
8192 	mblk_t *mp;
8193 	udp_stack_t *us = udp->udp_us;
8194 
8195 	ASSERT(q == RD(q));
8196 
8197 	mutex_enter(&udp->udp_drain_lock);
8198 	/*
8199 	 * There is no race with a concurrent udp_input() sending
8200 	 * up packets using putnext() after we have cleared the
8201 	 * udp_direct_sockfs flag but before we have completed
8202 	 * sending up the packets in udp_rcv_list, since we are
8203 	 * either a writer or we have quiesced the conn.
8204 	 */
8205 	udp->udp_direct_sockfs = B_FALSE;
8206 	mutex_exit(&udp->udp_drain_lock);
8207 
8208 	if (udp->udp_rcv_list_head != NULL)
8209 		UDP_STAT(us, udp_drain);
8210 
8211 	/*
8212 	 * Send up everything via putnext(); note here that we
8213 	 * don't need the udp_drain_lock to protect us since
8214 	 * nothing can enter udp_rrw() and that we currently
8215 	 * have exclusive access to this udp.
8216 	 */
8217 	while ((mp = udp->udp_rcv_list_head) != NULL) {
8218 		udp->udp_rcv_list_head = mp->b_next;
8219 		mp->b_next = NULL;
8220 		udp->udp_rcv_cnt -= msgdsize(mp);
8221 		udp->udp_rcv_msgcnt--;
8222 		if (closing) {
8223 			freemsg(mp);
8224 		} else {
8225 			putnext(q, mp);
8226 		}
8227 	}
8228 	ASSERT(udp->udp_rcv_cnt == 0);
8229 	ASSERT(udp->udp_rcv_msgcnt == 0);
8230 	ASSERT(udp->udp_rcv_list_head == NULL);
8231 	udp->udp_rcv_list_tail = NULL;
8232 	udp->udp_drain_qfull = B_FALSE;
8233 }
8234 
8235 static size_t
8236 udp_set_rcv_hiwat(udp_t *udp, size_t size)
8237 {
8238 	udp_stack_t *us = udp->udp_us;
8239 
8240 	/* We add a bit of extra buffering */
8241 	size += size >> 1;
8242 	if (size > us->us_max_buf)
8243 		size = us->us_max_buf;
8244 
8245 	udp->udp_rcv_hiwat = size;
8246 	return (size);
8247 }
8248 
8249 /*
8250  * For the lower queue so that UDP can be a dummy mux.
8251  * Nobody should be sending
8252  * packets up this stream
8253  */
8254 static void
8255 udp_lrput(queue_t *q, mblk_t *mp)
8256 {
8257 	mblk_t *mp1;
8258 
8259 	switch (mp->b_datap->db_type) {
8260 	case M_FLUSH:
8261 		/* Turn around */
8262 		if (*mp->b_rptr & FLUSHW) {
8263 			*mp->b_rptr &= ~FLUSHR;
8264 			qreply(q, mp);
8265 			return;
8266 		}
8267 		break;
8268 	}
8269 	/* Could receive messages that passed through ar_rput */
8270 	for (mp1 = mp; mp1; mp1 = mp1->b_cont)
8271 		mp1->b_prev = mp1->b_next = NULL;
8272 	freemsg(mp);
8273 }
8274 
8275 /*
8276  * For the lower queue so that UDP can be a dummy mux.
8277  * Nobody should be sending packets down this stream.
8278  */
8279 /* ARGSUSED */
8280 void
8281 udp_lwput(queue_t *q, mblk_t *mp)
8282 {
8283 	freemsg(mp);
8284 }
8285